Amyloid-β-regulated gene circuits for programmable Alzheimer’s disease therapy

18 19 Alzheimer’s disease (AD) is a neurodegenerative disease characterized in part by the 20 accumulation of the protein amyloid-β (Aβ). Monoclonal antibodies (mAbs) that target Aβ 21 for clearance from the brain have received FDA approval; however, these therapies are 22 accompanied by serious side effects , and their cognitive benefit for patients remains of 23 tremendous debate. Here, we present a potential engineered cell therapy for AD in which 24 we enlist cells of the central nervous system as programmable agents for sculpting the 25 neurodegenerative niche toward one that mitigates glial reactivity and neuronal loss. We 26 constructed a suite of A β-sensitive synthetic Notch (synNotch) receptors from clinically 27 tested anti -Aβ mAbs and show that cells expressing these receptors can recognize 28 synthetic A β42 and A β40 with differential sensitivity . We express these receptors in 29 astrocytes, cells native to the brain that are known to become dysfunctional in AD. These 30 synNotch astrocytes, which upregulate selected transgenes upon exposure to synthetic 31 and human brain -derived amyloid, were engineered to express potential therapeutic 32 transgenes in response to Aβ, including brain-derived neurotrophic factor and antagonists 33 of the cytokines tumor necrosis factor and interleukin-1. SynNotch astrocytes that express 34 such antagonists in response to A β partially attenuate a cytokine -induced reactive 35 astrocyte phenotype and promote barrier properties in brain microvascular endothelial 36 cells. Additionally, engineered Aβ-synNotch cells potently upregulate transgene 37 expression in response to Aβ deposited in the 5xFAD mouse brain , demonstrating the 38 capacity to recognize Aβ in situ. Overall, our work supports A β-synNotch receptors as 39 promising tools to generate a cell-based therapy for AD with targeted functionalities to 40 positively influence the AD niche. 41 42 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 17, 2025. ; https://doi.org/10.1101/2025.03.12.642808doi: bioRxiv preprint 2

43 44 Alzheimer’s Disease (AD) is a progressive, neurodegenerative disorder. The 45 pathophysiology underlying AD is complex, though it is most often characterized by 46 accumulation of amyloid-β (Aβ) and tau aggregates.1–3 These proteins have been the 47 subject of intensive research in AD therapeutic development. Much of this work has 48 focused on anti-Aβ immunotherapy, in the form of monoclonal antibodies, that clear A β. 49 Interest in this approach is based on a disease model in which Aβ aggregation is posited 50 as the disease trigger and other disease facets, including tau propagation, chronic 51 neuroinflammation, brain atrophy, and ultimately diminished cognition all occur 52 downstream.3 Anti-Aβ antibodies aim to mitigate such symptoms by clearing the putative 53 trigger, Aβ. Dozens of monoclonal antibodies (mAbs) targeting various species of Aβ have 54 been developed and investigated in clinical trials.4–9 Though many of them have failed to 55 reach their trial benchmarks, three, aducanumab (Aduhelm, 2021), lecanemab (Leqembi, 56 2023), and donanemab (Kisunla, 2024), have received approval from the FDA. These 57 drugs are the first to be approved for AD in ~20 years. However, the effect s of these 58 treatments on cognitive decline is small , and some have questioned whether the 59 associated therapeutic benefits are clinically meangingful.10 This class of antibodies is 60 also associated with side effects – most notably amyloid -related imaging abnormalities 61 (ARIA), which is due to edema and/or hemorrhage in the brain .11 In the aducanumab 62 (EMERGE and ENGAGE), lecanemab (Clarity AD), and donanemab (TRAILBLAZER -63 ALZ2) clinical trials, 12.6 -36% of participants receiving drug (compared to 0.8 -3% of 64 placebo group) experienced ARIA-Edema, while ARIA-Hemorrhage was present in 17.3-65 35% of participants receiving drug (compared to 7.2 -10% of placebo group). 12–15 While 66 most episodes of ARIA resolved, a minority of ARIA cases have resulted in poor outcomes 67 for patients, including a small number of fatalities.11 There is currently no therapy for AD 68 that stops or reverses disease progression; those that slow cognitive decline do so at 69 substantial risk to patient s. Nevertheless, the A β plaque is a critically important niche 70 within AD brain tissue where many neuropathological abnormalities converge. 71 The next generation of experimental therapies for AD will likely need to address a 72 broader range of factors beyond Aβ accumulation, and chronic inflammation has emerged 73 as a potential major driver of disease. Microglia and astrocytes, both glial cells, become 74 activated in response to A β and other features of the AD niche .16,17 Microglia take on a 75 pro-inflammatory phenotype and produce inflammatory cytokines, including interleukin-1 76 (IL-1α), which induces synapse loss, and tumor necrosis factor (TNF), which can directly 77 cause neuronal death.18,19 Significantly, these inflammatory signals from microglia induce 78 pro-inflammatory reactive astrocytes.20 Reactive astrocytes are highly present in AD, are 79 neurotoxic, and contribute to cerebrovascular dysfunction .21–23 New treatment 80 approaches, that do not rely solely on the clearance of protein aggregates, and are 81 instead capable of mitigating neuroinflammation, are needed. 82 Engineered cell therapies, in which cells can be programmed to survey their 83 environment and exert therapeutic functions, have shown great success in other 84 diseases. The most notable is chimeric antigen receptor T (CAR -T) cells in oncology. 85 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 17, 2025. ; https://doi.org/10.1101/2025.03.12.642808doi: bioRxiv preprint 3 CAR-T cell therapy involves ex vivo transduction of a patient’s T cells to program them 86 with an antibody -based receptor that can specifically recognize a tumor -associated 87 antigen, thus coupling recognition of a tumor antigen to activation of the native T cell 88 effector mechanisms.24 Six CD19 or BCMA CAR-T cell products are FDA -approved to 89 treat blood cancers, including lymphomas, leukemias, and multiple myeloma.25,26 Building 90 on the success of CAR -T therapy, CAR macrophages (CAR-M) have been investigated 91 pre-clinically for AD; in these designs, the CAR is constructed from an anti -Aβ antibody 92 fragment and results in macrophage activation through Fc receptor signaling, leading to 93 phagocytosis of A β.27 While this approach capitalize s on native macrophage cellular 94 functions, the intended outcome still depen ds on Aβ clearance, which has not shown 95 evidence of substantially modifying disease progression. To date, the success of 96 engineered cell therapies has not been translated to neurodegenerative disease. 97 Here we present a potential cell-based therapy platform for AD, where cells can 98 intelligently survey their microenvironment for disease markers (i.e., the Aβ plaque niche). 99 As with CAR technology, this is accomplished through a synthetic receptor capable of 100 recognizing Aβ. However, unlike Aβ CAR-M cells, we design cells to respond to Aβ in a 101 manner that is independent of A β clearance and instead activates a selected gene 102 expression program. We utilize the synthetic Notch (synNotch) platform to program cells 103 with custom output responses to A β recognition. SynNotch is based on the native 104 juxtacrine Notch signaling pathway, in which recognition of the ligand Delta results in a 105 series of sequential cleavages, mediated by an ADAM metalloproteinase and g-106 secretase, in the transmembrane core of the Notch protein. Proteolytic cleavage frees the 107 Notch intracellular domain from the membrane. 28 The intracellular domain travels to the 108 nucleus, where it results in the transcription of downstream genes (Fig. 1A). Thus, in 109 Notch signaling, there is a direct link between an extracellular cue and a transcriptional 110 event, mediated by receptor proteolysis subsequent to ligand recognition. Importantly, the 111 transmembrane core where this proteolysis occurs is conserved in the synNotch platform, 112 while the extracellular and intracellular domains are exchanged for synthetic components: 113 the extracellular domain for the single chain variable fragment (scFv) of an anti -Aβ 114 antibody, and the intracellular domain for a synthetic transcription factor. Recognition of 115 Aβ drives proteolytic cleavage of the transmembrane core, releasing the synthetic 116 transcription factor to translocate to the nucleus to drive expression of a programmed 117 transgene (Fig. 1A). Cells that express these receptors sense Aβ as a marker of disease 118 pathology and, via synNotch, respond with the expression of a programmed transgene. 119 This Aβ-synNotch platform regulates expression of anti-inflammatory or neuroprotective 120 transgenes, demonstrating the capacity to engineer central nervous system (CNS) cells 121 to overcome deleterious factors within the AD brain. 122 123

124 125 Receptor construction and plasmid cloning 126 Each Aβ receptor was constructed by reverse translating the amino acid sequences of 127 the variable heavy (vH) and variable light (vL) chains of their respective monoclonal 128 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 17, 2025. ; https://doi.org/10.1101/2025.03.12.642808doi: bioRxiv preprint 4 antibodies (donanemab, bapineuzumab, and gantenerumab). The scFv nucleic acid 129 sequences were designed by combining cognate vH and vL chains via a flexible linker 130 ((G4S)3). These scFvs were assembled with the Notch transmembrane core and 131 tetracycline transactivator (tTA) synthetic transcription factor. In a series of experiments, 132 a previously described, GFP -responsive receptor based on the GFP -specific LaG16 133 nanobody was used as a control.29,30 Lentiviral plasmids encoding for each receptor were 134 cloned using NEBuilder HiFi DNA Assembly mix. “Payload” plasmids encoding the 135 response transgene downstream of the tTA response element (TRE) were cloned using 136 the same methods. Plasmids were transformed into DH5 α E. coli competent cells 137 (SMObio) and plated on LB agar with ampicillin and incubated overnight at 37˚C. Colonies 138 were picked and cultured in LB with ampicillin overnight. Plasmids were purified via 139 miniprep (Qiagen). All plasmids were sequence-verified before use. 140 141 Cell culture 142 L929 mouse fibroblasts. Fibroblasts (ATCC) were cultured in DMEM + GlutaMax (Gibco) 143 supplemented with 10% heat-inactivated FBS (Gibco) at 37˚C with 5% CO 2. For routine 144 passaging, cells were dissociated using TrypLE (Gibco) by incubation at 37˚C for 5 145 minutes. For synNotch activation experiments, cells were detached using Accutase 146 (Gibco) by incubation at 37˚C for 5 minutes. Suspensions were quenched in medium and 147 centrifuged at 300xg for 5 minutes and resuspended for plating. 148 149 Mouse mesenchymal stem cells. Bone marrow -derived mMSCs from C57BL/6 mice 150 (Cyagen) were cultured in MEM -α + GlutaMax (Gibco) supplemented with 15% FBS 151 (Gibco) at 37˚C with 5% CO2. For passaging, cells were washed with DPBS (Gibco) then 152 detached with TrypLE by incubation at 37˚C for 5 minutes. For transplanting to slice 153 cultures, cells were detached using Accutase (Gibco) by incubation at 37˚C for 5 minutes. 154 Suspensions were quenched in medium and centrifuged at 300x g for 5 minutes and 155 resuspended for seeding. 156 157 CC3 human induced pluripotent stem cells (hiPSCs) . CC3 hiPSCs31 were maintained in 158 Essential 8 (E8) medium 32 on Matrigel (Corning) -coated wells. Routine cell passaging 159 was carried out using by incubation with ReLeSR (Stem Cell Technologies) for 1 minute 160 at RT, followed by ReLeSR removal and incubation at 37˚C for 6 minutes. 161 162 Astrocyte differentiation from CC3 hiPSCs. Differentiations were adapted from a 163 previously described protocol.22 On Day -1 of differentiation, CC3 hiPSCs were detached 164 using Accutase and plated on Matrigel in E8 with 10 µM Y-27632 (ROCKi) (Tocris) at a 165 density of 2x105 cells/cm2 (Fig. 3A). The following day, the medium was replaced with E6 166 containing 10 µM SB431542 (STEMCELL Technologies) and 1 µM dorsomorphin (Tocris). 167 Medium was changed every day until Day 6 of differentiation. On Day 6, clumps of cells 168 were carved out from the confluent cell layer using a P200 pipette tip and replated on a 169 fresh Matrigel well containing E6 with 10 ng/mL CNTF (Peprotech) and 10 ng/mL EGF 170 (Peprotech). The plates were shaken at 37˚C for several hours to allow cells to attach. 171 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 17, 2025. ; https://doi.org/10.1101/2025.03.12.642808doi: bioRxiv preprint 5 The medium was replaced every 3 days and cells were passaged when nearing 172 confluence, with the first passage occurring around Day 30 of differentiation. To reach 173 astrocyte maturity, cells were maintained in differentiation medium for 7 to 10 passages 174 until about 65 days of culture. Astrocytic phenotype was confirmed by GFAP expression 175 by immunocytochemistry. 176 177 Brain microvascular endothelial cell (BMEC) differentiation from CC3 hiPSCs. 178 Differentiations were performed as previously described. 33 On Day -1 of differentiation, 179 CC3 hiPSCs were detached using Accutase and plated on Matrigel in E8 with 10 µM 180 ROCKi at a density of 15,600 cells/cm2. The following day, the medium was replaced with 181 E6 to induce differentiation. The medium was replaced every day until Day 4 of 182 differentiation. On Day 4, the medium was switched to endothelial cell medium composed 183 of human endothelial serum free medium (HESFM; Gibco) with GlutaMAX (Gibco), B27 184 supplement (Gibco), 10 µM retinoic acid (RA; Millipore -Sigma), and 20 ng/mL bFGF 185 (Peprotech). On Day 6 of differentiation, cells were seeded on transwell membranes 186 coated with a mixture of 400 µg/mL collagen IV (Sigma Aldrich) and 100 µg/mL fibronectin 187 (Sigma Aldrich). After 24 hours, cells were switched to endothelial cell medium without 188 bFGF and RA. 189 190 Lentivirus production and transduction 191 Lentivirus was produced by transfecting Lx293T cells (Clontech) plated in a 6-well plate 192 with 2.0 µg of transfer plasmid (i.e., receptor or payload plasmids), 1.5 µg of pCMV -193 dR8.91 gag/pol packaging plasmid,34 and 0.6 µg of pMD2.G envelope plasmid (Addgene 194 #12259) with Lipofectamine 3000 (Thermo Scientific). One day following transfection, the 195 medium was replaced with complete medium composed of DMEM + GlutaMax 196 supplemented with 10% heat -inactivated FBS (Gibco). On days 2 and 3 following 197 transfection, viral medium was collected and filtered with a 0.45 µm PVDF filter 198 (CELLTREAT). 199 200 Transduction of L929 fibroblasts and mMSCs. Viral medium was concentrated in a 100 201 kDa MWCO filter (Millipore) via centrifugation and resuspended in fresh cell culture 202 medium (DMEM + 10% heat-inactivated FBS; MEM-α + 15% FBS) for cell transduction. 203 Cells were transduced in a 6-well plate format using the virus collected from one well and 204 media was supplemented with 4 µg/mL polybrene (Sigma Aldrich) to facilitate viral 205 transduction. 206 207 Transduction of hiPSC-derived astrocytes. Viral medium was incubated overnight at 4˚C 208 with Lenti -X concentrator (Takara). Lentivirus was pelleted via centrifugation and 209 resuspended in astrocyte medium (E6 + CNTF + EGF) before being added to astrocytes. 210 Cells were transduced in a 6-well plate format using the half the virus collected from one 211 well and media was supplemented with 4 µg/mL polybrene (Sigma Aldrich) to facilitate 212 viral transduction. 213 214 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 17, 2025. ; https://doi.org/10.1101/2025.03.12.642808doi: bioRxiv preprint 6 Flow cytometry and fluorescence activated cell sorting (FACS) 215 Prior to analytical flow cytometry or FACS, cells were dissociated with Accutase. Cells 216 were centrifuged at 300x g for 5 minutes and resuspended in flow buffer (1% BSA in 217 DPBS) and blocked on ice for 15 minutes. Cells were stained with an anti -c-myc tag 218 antibody conjugated to Alexa647 (Cell Signaling Technologies) diluted 1:50 in flow buffer 219 for 1 hr on ice. Cells were washed 2 times by centrifugation and resuspended in flow 220 buffer for analysis on a Cellstream flow cytometer to assess cells double-positive for myc-221 tag (incorporated on the N-terminus of the synNotch receptor) and BFP (encoded for 222 constitutive expression in the payload transgene vector). Flow cytometry analysis was 223 done using FlowJo. To obtain a pure population of synNotch-positive cells, samples were 224 run on a 4 -laser FACSAria III cell sorter , and double -positive cells were collected. 225 Collected cells were spun down at 300x g for 5 minutes and resuspended in cell culture 226 medium for plating. 227 228 Synthetic Aβ preparation 229 Biotinylated and unmodified synthetic A β40 and A β42 preparations (Anaspec) were 230 resuspended according to manufacturer’s instructions and aliquoted for long-term storage 231 at -80˚C. Protein was originally resuspended at 1 mg/mL for dynamic light scattering 232 (DLS) analysis then diluted to 50 µg/mL for cell culture assays. 233 234 SynNotch L929 activation 235 For immobilized , biotinylated Aβ, tissue culture plates were treated with 10 µg/mL of 236 streptavidin (Thermo Fisher) in DPBS and incubated for a minimum of 1 hr at 37˚C. The 237 streptavidin was then removed, and biotinylated Aβ was added to the coated wells at 50 238 µg/mL and incubated for an additional hour at 37˚C. The A β solution was removed and 239 rinsed with DPB S, leaving 0.6 µg A β per well by BCA . SynNotch L929 cells were then 240 plated at 20,000 cells/well. For adsorbed unmodified Aβ, the cell culture plate was coated 241 directly with 50 µg/mL Aβ and incubated for 1 hr at 37˚C. The wells were rinsed and cells 242 were added as above. For medium supplemented A β, the cells were plated first at the 243 same density and an equivalent amount of A β was then supplemented to the medium. 244 For anti-c-myc bead activation assays, the cells were plated and anti -c-myc beads were 245 added at a final concentration of 0.1 mg/mL (Thermo Fisher). SynNotch activation was 246 visualized by mCherry by fluorescence microscopy. After 72 hr, the production of the firefly 247 luciferase transgene was measured using a BrightGlo luminescence assay (Promega) on 248 a Tecan Infinite M1000 Pro plate reader. 249 250 Immunostaining of astrocytes 251 Immunolabeling of hiPSC-derived astrocytes was performed at day 60 of differentiation 252 to confirm fate. W ells were fixed with ice -cold 4% paraformaldehyde in DPBS (Thomas 253 Scientific) for 15 minutes at room temperature. Cells were then washed 3 times with 254 DPBS and blocked with 5% FBS, 0.3% Triton -X in DPBS at 4˚C overnight. Cells were 255 incubated with anti -GFAP (Aves Labs, 1:300) and anti -CD44 (Cell Signaling 256 Technologies, 1:500) primary antibodies at 4˚C overnight. After primary antibody 257 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 17, 2025. ; https://doi.org/10.1101/2025.03.12.642808doi: bioRxiv preprint 7 incubation, cells were rinsed 3 times with DPBS and then incubated with secondary 258 antibodies for 1 hr at room temperature (goat anti -chicken Alexa488; goat anti -mouse 259 Alexa647, Invitrogen, 1:1000). Cells were then incubated with nuclear stain DAPI 260 (Thermo Scientific) and washed before imaging. 261 262 Human brain-derived Aβ seed preparation 263 Brain tissue from de -identified human cases with cerebral amyloid angiopathy (CAA) 264 pathology but minimal parenchymal A β and tau (i.e., pure CAA -Aβ) from the Vanderbilt 265 Brain and Biospecimen Bank (VBBB) were used to isolate Aβ seeds. Ethical oversight of 266 the VBBB is provided by the Vanderbilt Institutional Review Board; recruitment and uses 267 of the tissue conforms to the ethical principles of the Belmont Report. Microvessels were 268 isolated from identified brains and A β microparticles ranging from 500 nm to 7 µm 269 diameter were collected via collagenase digestion. 35 Aβ was then separated from the 270 vascular elements using a 3D printed custom-designed micro-sieve. Western blots were 271 performed on samples to confirm that they were not contaminated by tau. 272 273 SynNotch astrocyte activation on CAA-Aβ 274 SynNotch astrocytes were plated on Matrigel coated wells in E6 with 10 ng/mL CNTF and 275 10 ng/mL EGF at a density of 50,000 cells/cm 2. 0.6 µg of CAA -Aβ seeds was added 276 directly to plated astrocytes. To prevent contamination by the patient-derived protein, the 277 astrocytes were treated with 1X antibiotic -antimycotic (Gibco). SynNotch activation was 278 visualized by mCherry by fluorescence microscopy after 72 hr. 279 280 SynNotch astrocyte activation on synthetic Aβ 281 Stem cell-derived astrocytes are typically cultured on a basement membrane such as 282 Matrigel, which may interfere with our use of surface immobilized, biotinylated A β42 for 283 synNotch activation studies. We and others have previously shown that culture surfaces 284 decorated with glycosaminoglycan -binding peptide (GBP) and cyclic arginine -285 asparagine-aspartic acid (cRGD) can serve as a substitute for Matrigel to promote cell 286 adhesion and support stem cell -derived CNS cell phenotypes. 30,36,37 Thus, biotinylated 287 GBP (Genscript Express) and cRGD (Carbosynth) adhesion peptides were reconstituted 288 in UltraPure distilled water (Invitrogen) and mixed with biotinylated A β42 to final 289 concentrations of 5 µM GBP, 2.15 µM cRGD, and 0.4 µM of Aβ. Non-tissue culture treated 290 well plates were first coated with 10 µg/mL of streptavidin for a minimum of 1 hr to 291 overnight at 37˚C. The streptavidin was aspirated and replaced with the peptide solution 292 and incubated for 1 -2 hr at 37˚C. SynNotch astrocytes were dissociated with Accutase 293 and plated in the coated wells after the peptide solution was removed. 294 295 Astrocyte monoculture inflammation experiments 296 The Aβ peptide surface was prepared as above. SynNotch astrocytes were dissociated 297 with Accutase and plated in the coated wells after the peptide solution was removed. 298 Astrocytes were plated a density of ~35,000 cells/cm 2. 48 hr after cell plating, the cells 299 were supplemented with inflammatory cytokines IL -1α (STEMCELL Technologies) and 300 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 17, 2025. ; https://doi.org/10.1101/2025.03.12.642808doi: bioRxiv preprint 8 TNF-α (STEMCELL Technologies). Cytokines were added at 5 ng/mL IL -1α, 10 ng/mL 301 TNF. After an additional 48 hr, medium was collected for analysis by ELISA or secreted 302 embryonic alkaline phosphatase ( SEAP) assay and mRNA was isolated. The presence 303 of the SEAP transgene was detected using a chemiluminescence assay (Takara Bio) 304 measured on a Tecan Infinite M1000 Pro plate reader. 305 306 ELISA 307 The Human TNF R1/TNFRSF1A DuoSet ELISA kit and Human Il -1RA/IL-1F3 DuoSet 308 ELISA kits were used with the DuoSet Ancillary Reagents kit to detect anti -inflammatory 309 transgenes (R&D). Cell culture medium from Bap-Notch_sTNFR1-IL-1Ra astrocytes was 310 diluted 1:50 in diluent; medium from Bap -Notch_SEAP astrocytes was diluted 1:5 in 311 diluent. For detection of BDNF payload, the Total BDNF Quantikine ELISA kit was used 312 (R&D). Cell culture medium was added directly with no dilution. Absorbance was 313 measured at 450 nm on a Tecan Infinite M1000 Pro plate reader and a correction read at 314 570 nm was subtracted. The average absorbance of the 0 pg/mL standard was subtracted 315 from all samples. Sample concentrations were calculated using a four-parameter logistic 316 curve fit of the standards using Graphpad Prism 10. 317 318 BMEC:Astrocyte co-culture 319 BMECs were plated on transwell filter culture inserts (Corning) after coating the 320 membranes with a mixture of 400 µg/mL collagen IV (Sigma Aldrich) and 100 µg/mL 321 fibronectin (Sigma Aldrich) in endothelial cell medium (HESFM with GlutaMAX and B27) 322 supplemented with bFGF and RA at a density of 100,000 cells/cm2. Separately, synNotch 323 astrocytes were plated on immobilized GBP, cRGD, and Aβ42 as in the monoculture 324 experiments. SynNotch astrocytes were also plated in endothelial cell medium at a 325 density of ~50,000 cells/cm 2. After synNotch astrocytes attached, the BMEC transwells 326 were transferred to synNotch astrocytes and were maintained in endothelial medium 327 without bFGF or RA. After 48 hours of co-culture, inflammatory cytokines IL-1α (5 ng/mL) 328 and TNF (10 ng/mL) were added both above and below the transwell. The 329 transendothelial electrical resistance (TEER) was measured using an STX2 chopsticks 330 electrode set and an EVOM2 volt -ohmmeter (World Precision Instruments) daily. The 331 following formula was utilized to calculate the reported TEER (TE): 332 333 𝑇! = (𝑇" − 𝑇#) 𝑥 𝐴𝑟𝑒𝑎 334 335 The measured TEER from a BMEC monolayer (T M) was subtracted by the measured 336 TEER from a blank transwell membrane with no cells (T B). This quantity was then 337 multiplied by the surface area of the transwell membrane (0.33 cm 2) to determine TE. All 338 TEER measurements in this study are reported as Ω × cm2. TEER was plotted over time 339 using GraphPad Prism; the area under the curve (AUC) was calculated in Prism for each 340 well and averaged across groups. 341 342 Gene expression analysis 343 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 17, 2025. ; https://doi.org/10.1101/2025.03.12.642808doi: bioRxiv preprint 9 Cells were lysed, and mRNA was isolated using PureLink Mini Kits (Invitrogen) following 344 the manufacturer’s instructions. mRNA was reverse transcribed to cDNA using the 345 SuperScript IV VILO Master Mix (Invitrogen). Quantitative PCR was performed with 346 PowerTrack SYBR Green Master Mix (Applied Biosystems) on a Bio-Rad CFX96 with the 347 mass of cDNA across all samples. Primer pairs used for gene expression detection are 348 listed in Supplementary Table 1. Relative gene expression was calculated using the delta-349 delta Ct method using GAPDH as a reference gene. In all inflammation experiments, 350 samples from the Aβ-free/cytokine-free condition was used as the reference group. 351 352 Organotypic slice culture preparation and cell transplantation 353 All animal protocols were approved by the Institutional Animal Care and Use Committee 354 (IACUC) at Vanderbilt University. Three adult (11 -month old) hemizygous 5xFAD mice 355 and 3 adult (11-month old) wild-type C57/BL6 mice were sacrificed. 5xFAD and wild-type 356 mice were littermates. Brains were removed and mounted on a vibrating blade microtome 357 (Leica) in slicing media composed of Hibernate A (BrainBits) with B27, GlutaMax, and 358 Gentamycin (Gibco). Coronal slices 300 µm thick were cultured on a membrane insert 359 (Millipore Sigma) in Neurobasal A (Gibco) with B27, GlutaMax, and Gentamycin. After 360 slices were prepared, synNotch mMSCs were dissociated with Accutase and 361 resuspended in MEM-α with 15% FBS. 200,000 synNotch cells were plated in a drop of 362 10 µL on top of each slice. Save for a few slices maintained as no cell controls, half the 363 slices from each mouse received Bap-Notch mMSCs; half received control LaG16-Notch 364 mMSCs. Serial slices were alternately seeded with Bap -Notch mMSCs or LaG16-Notch 365 mMSCs to remove the confounding effects of variable A β accumulation on different 366 sections of the brain. After 72 hr, medium was collected from the culture and the presence 367 of Aβ-driven SEAP transgene was determined using a chemiluminescence assay read on 368 a Tecan Infinite Pro plate reader. Samples were run in technical duplicate. 369 370 Organotypic slice culture cryosection preparation and staining 371 Slices were fixed in 4% paraformaldehyde at 4˚C overnight and cryopreserved in 30% 372 sucrose. Samples were embedded in optimal cutting temperature (OCT) compound and 373 tissue was sectioned into 10 µm sections. After thaw, sections were surrounded with a 374 hydrophobic barrier, washed with DPBS, and blocked with 1% BSA, 0.25% Triton-X, and 375 10% normal goat serum for 1 hr at RT. Sections were incubated with anti-Aβ (6E10, 1:400) 376 and anti -mCherry (Cell Signaling Technologies, 1:200) primary antibodies at 4˚C 377 overnight. After primary antibody incubation, cells were rinsed 3 times with DPBS and 378 incubated with secondary antibodies for 1 hr at room temperature (goat anti -mouse 379 Alexa488 (1:1000); goat anti -rabbit Alexa555 (1:500) (Invitrogen). After secondary 380 antibody incubation, tissues were incubated with TrueVIEW Autofluorescence Quenching 381 reagent (VectorLabs) for 5 min at RT. Finally, sections were washed and sealed with a 382 coverslip. Sections were cured for 2 hours at RT prior to imaging. 383 384 Fluorescence microscopy 385 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 17, 2025. ; https://doi.org/10.1101/2025.03.12.642808doi: bioRxiv preprint 10 All images were taken on a Leica Dmi8 epifluorescent microscope at 10 or 20X 386 magnification. 387 388 Statistical analysis 389 All statistical analyses were performed with GraphPad Prism 10. Plotted values represent 390 mean ± standard deviation. Unpaired t tests with α=0.05 were performed when comparing 391 two groups. For TEER analysis, a three-way ANOVA was performed (α=0.05) to include 392 time as a factor. For all other multi-factorial comparisons, one- or two-way ANOVAs with 393 Tukey’s post-hoc analysis were performed with α=0.05. Outliers were identified by robust 394 regression and outlier removal (ROUT) and were excluded. Statistical analysis of qPCR 395 gene expression was performed on the log2 transform of fold change. 396 397

398 399 SynNotch receptors constructed from anti -Aβ monoclonal antibodies enable A β-driven 400 transgene expression 401 402 Receptors were designed from a panel of anti -Aβ monoclonal antibodies previously 403 tested in clinical trial s for AD: gantenerumab (binds conformational epitope found in 404 fibrillar Aβ);5 donanemab (recognizes N -terminal truncated pyroglutamate form of A β 405 [Aβp3-42] found in plaques)38; bapineuzumab (recognizes amino acids 1-5; binds soluble 406 and insoluble Aβ40/42)39,40 (Fig. 1B). These receptors were denoted as Gant-Notch, Don-407 Notch, and Bap -Notch, respectively. For initial characterization, the receptors were 408 expressed in mouse L929 fibroblasts by lentiviral transduction, and cells were sorted for 409 matched levels of receptor expression by fluorescence-activated cell sorting ( FACS) 410 (Supplementary Fig. 1). Notably, prior to cell sorting, only a small fraction of fibroblasts 411 successfully expressed Don -Notch, though receptor expression persisted in the sorted 412 population (Supplementary Fig. 1D). Cells were programmed to express mCherry 413 fluorescent protein and firefly luciferase in response to receptor activation by A β. To 414 characterize the ability of the synNotch receptors to recognize A β, we immobilized C-415 terminally biotinylated, synthetic Aβ42 and A β40 to streptavidin -coated well -plates then 416 seeded synNotch cells. Gant-Notch, Don -Notch, and Bap -Notch recognize d synthetic 417 Aβ42 immobilized to the plate surface, as indicated by Aβ-dependent luciferase transgene 418 expression. Fold inductions of luminescence compared to synNotch cells plated on a 419 control, Aβ-free surface range from 28x (Don-Notch) to 70x (Bap-Notch) (Fig. 1C). Gant-420 Notch, Don -Notch, and Bap -Notch also recognize d Aβ40, the more prevalent but less 421 pathogenic variant, 41,42 immobilized by biotin -streptavidin interaction, indicated by 422 luciferase transgene expression (Fig. 1D). This is expected based on the antibody 423 epitopes, though overall, receptors exhibited a lower fold activation on Aβ40 than on Aβ42. 424 Notably, synNotch activation, as measured by luciferase activity, was not detected in cells 425 expressing a control, GFP -responsive LaG16-Notch receptor, confirming that activation 426 is dependent on selectively programming synNotch with Aβ-detection motifs. 427 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 17, 2025. ; https://doi.org/10.1101/2025.03.12.642808doi: bioRxiv preprint 11 We also constructed synNotch receptors with recognition domains derived from the 428 other two anti -Aβ monoclonal antibodies approved by the FDA for AD: aducanumab 429 (recognizes amino acids 3 -7; binds soluble oligomers and insoluble fibrils of A β42) (Ad-430 Notch);43,44 and lecanemab (binds a conformational epitope found in Aβ protofibrils) (Lec-431 Notch)45,46. Ad -Notch and Lec -Notch d id not recognize synthetic A β in our studies 432 (Supplementary Fig. 2A), perhaps due to the conformation of the aggregates in our 433 preparations, which tend to be dominated by high molecular weight species with 434 hydrodynamic radii of ~2 -100 nm (Supplementa ry Fig. 3A-B), making them preferred 435 substrates for recognition via our original panel of receptors. To validate that both Lec- 436 and Ad-Notch receptors are functional in the fibroblasts, we treated the cell cultures with 437 1 µm beads decorated with antibodies specific to the c -myc epitope tag. Such beads 438 serve as surrogate s for Aβ ligand, as the N -terminal domain of the synNotch receptor 439 contains the c-myc tag and thus engages the beads in a manner similar to immobilized 440 ligand (i.e., Aβ). Results reveal that the Ad -Notch and Lec -Notch receptors can be 441 activated (Supplementary Fig. 2B-C), suggesting that the lack of response to A β is not 442 due to a defect in the receptor design. Due to their underperformance in converting A β 443 presence to a productive, engineered cell response, subsequent experiments did not 444 include Ad-Notch or Lec-Notch. 445 446 Aβ-synNotch receptors differentially recognize Aβ42 and Aβ40 447 448 We next tested whether the A β synNotch receptors recognize unmodified A β (i.e., not 449 biotinylated). We first coated the culture surface with synthetic A β42 via passive 450 adsorption. Gant-Notch, Don-Notch, and Bap-Notch recognized synthetic Aβ42 adsorbed 451 to the plate, as demonstrated by the fold induction in luciferase transgene expression 452 compared to cells plated on a control surface (Fig. 2A). Remarkably, for all three receptor 453 variants, passive adsorption of unmodified Aβ42 results in higher luciferase fold induction 454 than on biotinylated A β42, ranging from 53x (Don-Notch) to 155x (Bap-Notch). We were 455 also interested in whether the synNotch receptors recognize A β supplemented in the 456 medium. Of note, synNotch activation requires a mechanical force of 4-12 pN to activate; 457 this cannot be provided by soluble , monomeric ligands, and thus synNotch requires an 458 immobilized ligand.29,30,47 We hypothesize that the oligomeric nature of A β (specifically, 459 multiple instances of the same epitope on an Aβ oligomer) may allow for recognition of 460 ligand captured and presented by a neighboring synNotch receptor in trans. In support of 461 this concept, t here is substantial aggregation seen in the synthetic A β preparations 462 (Supplementary Fig. 3). Bap-Notch resulted in significant ly higher luminescence fold 463 induction than all other receptors when Aβ42 was supplemented to the medium (Fig. 2B). 464 Unlike Aβ42, Aβ40 must be anchored to the surface to result in synNotch activation (Fig. 465 2C); medium supplemented A β40 resulted in less than 2 -fold induction in synNotch 466 activation (Fig. 2D). This discrepancy between Aβ42- and Aβ40-driven synNotch activation 467 is likely due to the decreased propensity of Aβ40 to form stable aggregates compared to 468 Aβ42, resulting in fewer epitopes available for multiple synNotch receptors to 469 simultaneously engage Aβ40 to generate the mechanical stress required for activation.48 470 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 17, 2025. ; https://doi.org/10.1101/2025.03.12.642808doi: bioRxiv preprint 12 471 Aβ-synNotch governs therapeutic transgene expression in human astrocytes 472 473 We next sought to express the Aβ synNotch system in astrocytes, brain cells that have a 474 major role in maintaining the blood -brain barrier (BBB) , regulating synapses, and 475 coordinating the immune response in the brain. 49 In AD, these roles become 476 dysregulated.50,51 In particular, astrocytes deeply interact with and shape the 477 amyloidogenic microenvironment by phagocytosing A β and releasing pro -inflammatory 478 factors.16,21 To capitalize on the native homeostatic function of astrocytes, we utilized 479 synNotch to enhance astrocytic Aβ recognition and implement programmable therapeutic 480 responses. 481 Astrocytes were differentiated from CC3 human iPSCs via dual -SMAD inhibition 482 (SB431542 and dorsomorphin) followed by culture in the presence of EGF and CNTF 483 (Fig. 3A-B).22 To encourage expression of the receptor in astrocytes, the constitutive 484 promoter driving the synNotch receptor was transitioned from EF1A to an astrocyte-485 specific truncated glial fibrillary acidic protein (GFAP) promoter (gfaABC1D).52 Astrocytes 486 were engineered to express the Gant -Notch or Bap-Notch receptor s via lentiviral 487 transduction. Don-Notch was excluded from the panel tested in astrocytes due to the 488 inefficient expression of the receptor in unsorted fibroblasts (Supplementary Fig. 1B). We 489 were interested in whether the synNotch receptors could recognize patient-derived Aβ, in 490 addition to synthetic A β. Aβ isolated from brain bank donors with cerebral amyloid 491 angiopathy (CAA) was added to the culture medium of astrocytes. Gant-Notch and Bap-492 Notch astrocytes recognized the patient-derived Aβ, as indicated by the expression of an 493 Aβ-driven mCherry transgene (Fig. 3C). These results indicate that human brain cells 494 engineered with Aβ-synNotch receptors recognize human brain-derived Aβ and respond 495 with programmable transgene output. 496 We were next interested in regulating potential therapeutic transgene expression 497 in synNotch astrocytes. Brain -derived neurotrophic factor (BDNF) is a neuronal growth 498 factor that has shown great promise in encouraging neuronal growth and synapse 499 formation after brain injury in non -human primates (NHPs).53–55 It has been investigated 500 as a therapy for neurodegeneration when delivered in AAV via direct injection to the 501 entorhinal cortex;29,54 however, this strategy results in expression of BDNF in all cell types 502 and regions of the brain. Widespread delivery of BDNF to the brain has been associated 503 with severe side effects (weight loss, sensory disturbances, and inappropriate cellular 504 migratory patterns).56 It is therefore a promising candidate as a payload transgene in the 505 synNotch system , as ectopic BDNF expression may be regulated by local A β 506 accumulation. To explore the feasibility of using synNotch to govern ectopic BDNF 507 expression, Bap -Notch astrocytes were plated on synthetic A β42. Recognition of A β 508 resulted in the expression of BDNF and mCherry transgenes from a bicistronic cassette. 509 Bap-Notch astrocytes expressed BDNF only when cultured with Aβ (Fig. 3D). Astrocytes 510 engineered to express only mCherry as a synNotch payload implement a synNotch -511 governed response to Aβ, as indicated by mCherry expression (Fig. 3E); however, such 512 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 17, 2025. ; https://doi.org/10.1101/2025.03.12.642808doi: bioRxiv preprint 13 cells did not upregulate BDNF production, confirming that enhanced BDNF expression is 513 potentiated by synNotch and is not attributable to a native astrocyte response. 514 515 Bap-Notch attenuates astrocyte inflammation in response to Aβ 516 517 Chronic neuroinflammation, mediated by astrocytes and microglia, plays a major role in 518 AD progression. Many AD risk genes (e.g., APOE, TREM2, CD33) identified in genome-519 wide association studies (GWAS) skew immune cell function toward a pro-inflammatory 520 phenotype.57–59 Additionally, elevated levels of inflammatory cytokines are found in the 521 cerebral spinal fluid (CSF) of AD patients,60 suggesting the AD brain may be conducive to 522 a reactive astrocyte phenotype. Indeed, pro-inflammatory, reactive astrocytes have been 523 identified in post -mortem brain tissue in AD. 21 Further, in vitro models suggest that 524 microglial responses to amyloid produced by APPSwe neurons exacerbate astrocyte 525 reactivity.61 We were therefore interested in whether we could use synNotch to re-route 526 this inflammatory, reactive astrocyte phenotype via an engineered response to Aβ. To this 527 end, we programmed Bap -Notch astrocytes to mitigate pro -inflammatory cues via 528 expression of soluble TNF receptor (sTNFR1) and IL -1 receptor antagonist (IL -1Ra) 529 payloads in response to Aβ. A control astrocyte line was generated that expressed SEAP 530 downstream of synNotch. Bap-Notch astrocytes were plated on synthetic A β42 for 48 hr 531 prior to supplementing the medium with IL-1α (5 ng/mL) and TNF (10 ng/mL) to induce a 532 reactive astrocyte phenotype (Fig. 4A).21,61–63 After 48 hr exposure to cytokines , 533 synNotch-SEAP astrocytes produced no detectable levels of sTNFR1 and IL1-Ra, despite 534 responding potently via the synNotch signaling channel, as represented by significant Aβ-535 driven SEAP production (Supplementary Fig. 4A-C). In contradistinction, astrocytes 536 engineered to antagonize TNF and IL -1α dramatically upregulated sTNFR1 and IL -1Ra 537 payloads in an A β-dependent manner (Fig. 4B-C), while SEAP activity was negligible 538 (Supplementary Fig. 4C). Correspondingly, qRT-PCR revealed that Bap-Notch mediated 539 significant attenuation of CSF2 and SERPINA3 expression when Aβ was present to drive 540 expression of the anti -inflammatory transgenes (Fig. 4D). Although not statistically 541 significant, IL6 expression was also reduced by ~18% upon Aβ-stimulated induction of 542 sTNFR1 and IL-1Ra production. Bap-Notch SEAP-astrocytes, however, did not mitigate 543 reactive gene expression in response to Aβ (Fig. 4E), again reflecting the ability to 544 deliberately modulate cell responses based on transgene selection . These results 545 demonstrate the utility of the synNotch platform in moderating the inflammatory 546 environment that leads to glial reactivity and eventual neuron loss. 547 Brain microvascular endothelial cells ( BMECs), along with astrocytes and 548 pericytes, compose the blood-brain barrier (BBB), the breakdown of which contributes to 549 cognitive decline in AD.64 In AD, this disruption is mediated in part by reactive astrocyte 550 production of SERPINA3.22 Building off our results in the astrocyte monoculture studies, 551 we postulated that synNotch-driven attenuation of a reactive astrocyte phenotype would 552 positively impact the function of the BBB. Anti -inflammatory Bap-Notch astrocytes were 553 co-cultured with BMECs, where the BMECs are cultured in a Transwell filter (Fig. 4F). 554 The Transwell format allows for the measurement of the transendothelial electrical 555 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 17, 2025. ; https://doi.org/10.1101/2025.03.12.642808doi: bioRxiv preprint 14 resistance (TEER), an indirect, quantitative readout of the integrity of the tight junctions 556 formed between BMECs. The co-culture was treated with inflammatory cytokines IL -1α 557 (5 ng/mL) and TNF (10 ng/mL) to mimic inflammation that disrupts the endothelial cell 558 tight junctions , as in our prior report. 22 While BMECs co -cultured with Bap -Notch 559 astrocytes engineered to express SEAP displayed a sharp overall decline in TEER values 560 in both the presence and absence of Aβ, synNotch -driven sTNFR1 and IL -1Ra 561 significantly attenuated this loss of barrier function when Aβ was included in cultures (Fig. 562 4G, Supplementary Fig. 5A). Analysis of these results revealed time and the interaction 563 of time and transgene to be significant factors (p = 0.003 and p = 0.0045, respectively). 564 Subsequent analysis of each timepoint separately revealed significant effects of 565 transgene and Αβ on BMEC function (Supplementary Table 2). Given the influence of 566 time, we binned an area under the curve (AUC) analysis into early (Days 0 -2) and late 567 (Days 3-5) phases, which indicated a significant protective effect of Αβ-induced synNotch-568 driven sTNFR1 and IL-1Ra that emerged in the Day 3-5 phase (Supplementary Fig. 5C 569 Fig. 4H). Taken together, these studies highlight that programmable transgene 570 expression enables synNotch astrocytes to resist neuroinflammation that characterizes 571 AD and suggest that such cells have the capacity to positively influence functional 572 properties of other CNS cells that are dysregulated in the AD niche. 573 574 The artificial Aβ-responsive platform detects Αβ in situ and is compatible with CNS -575 targeting gene delivery vehicles 576 577 We next sought to demonstrate the feasibility and functional consequence of delivering 578 Aβ-synNotch cells in a model of transplanted cell therapy. We expressed Bap -Notch in 579 mouse mesenchymal stromal cells ( mMSCs) via lentiviral transduction and sorted for a 580 receptor-positive population via FACS. mMSCs were chosen for this application for 581 several reasons: (1) they have been investigated as cell therapy for AD and can be 582 transplanted to mice directly; 65,66 (2) they are easily engineered to express synNotch 583 receptors by lentiviral transduction and can be sorted to obtain a population of synNotch-584 expressing cells; 29,67 and (3) they can migrate through tissue, allowing them to reach 585 targets beyond the injection site .68 As a model of a transplanted cell therapy and to 586 determine whether the Aβ synNotch receptors can recognize Aβ in situ, we delivered Bap-587 Notch mMSCs to brain organotypic slice cultures (OSCs) prepared from aged (11-month) 588 hemizygous 5xFAD mice and wild-type (WT) controls (Fig. 5A). The 5xFAD mouse model 589 (named for the 5 Familial AD mutations; 3 in APP, 2 in PSEN1) has been widely used to 590 investigate AD progression and the preclinical efficacy of AD therapeutics. By 6 to 9 591 months of age, the mice show significant amyloid plaque burden in addition to impaired 592 memory and cognitive function. 69,70 For each 5xFAD and WT mouse, half the total 593 number of slices were seeded with Bap-Notch cells; the other half were seeded with GFP-594 responsive LaG16-Notch cells as a negative control. mMSCs were engineered to express 595 SEAP and mCherry upon Αβ engagement. We observed a significant increase in SEAP 596 expression only in Bap-Notch cells seeded on slices from hemizygous 5xFAD mice (Fig. 597 5B). Levels of SEAP from Bap -Notch cells on WT OSCs were comparable to those 598 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 17, 2025. ; https://doi.org/10.1101/2025.03.12.642808doi: bioRxiv preprint 15 measured from negative control LaG16-Notch cells on 5xFAD or WT OSCs, or from OSCs 599 that did not receive mMSC transplantation. To validate the SEAP findings, we 600 cryosectioned OSCs and immunolabeled for A β and mCherry. The transplanted cells 601 were visualized with constitutive BFP. Bap-Notch mMSCs plated on the A β-containing 602 hemizygous 5xFAD slices demonstrated strong mCherry positivity co-localized with A β 603 deposits (Fig. 5C). As expected, wild -type tissue lacks Αβ, and transplanted mMSCs 604 (BFP+) are not mCherry+ (Fig. 5D). These results indicate that Bap-Notch recognizes Aβ 605 in native brain tissue and drives cells to implement potent responses. 606 While transplanted, engineered cells represent one potential avenue for deploying 607 synNotch cells in AD therapy, transducing endogenous astrocytes to express the 608 synNotch circuits offers an alternative translational approach that may be more 609 compatible with clinical utility as an off -the-shelf gene therapy . Towards this goal, we 610 constructed an AAV plasmid capable of delivering the A β-responsive receptor and 611 payload transgenes simultaneously. The packaging capacity of AAV is capped at 4.7 kb 612 of DNA. 71 Expression of the complete synNotch circuit would therefore require dual 613 transduction, which may drastically limit the number of cells able to drive therapeutic 614 output in response to Aβ. Synthetic Intramembrane Proteolysis Receptors (SNIPRs) form 615 a class of synthetic receptors that includes the original synNotch receptor but that also 616 include receptor variants composed of truncated Notch transmembrane and 617 juxtamembrane domains. 72 The smaller size of such SNIPR architecture s allows for 618 expression of both the receptor and the response transgenes from one AAV vector. Thus, 619 we constructed a Bap -SNIPR receptor. To validate Bap -SNIPR responsiveness to Aβ, 620 Bap-SNIPR and Bap -Notch L929 fibroblast cells were plated on immobilized A β42 and 621 Aβ40. Bap-Notch and Bap-SNIPR resulted in similar fold changes in luciferase expression 622 compared to control , Αβ-free conditions (Supplementary Fig. 6A-B). Bap -SNIPR and 623 Bap-Notch also display comparable recognition of unmodified Aβ42 and Aβ40 when these 624 species are supplemented in the medium or adsorbed to the plate surface 625 (Supplementary Fig. 6C-E). These results demonstrate the feasibility of installing the 626 artificial Aβ-responsive signaling channel in a format compatible with AAV, which are 627 therapeutically relevant and widely investigated for use as gene delivery vehicles in the 628 CNS. 629 630

631 632 Here, we present a new cell therapy approach for AD. In the last several decades, 633 many monoclonal antibodies targeting A β have been developed as passive 634 immunotherapy for treating AD.43,73 We repurpos ed these antibodies as synNotch 635 recognition domains and generated a panel of A β-sensitive synNotch receptors that 636 maintain the A β recognition capability of the antibodies. This represents a major 637 innovation over existing A β-targeted therapies for AD, which rely on A β clearance to 638 provide therapeutic benefit. In contrast, we demonstrate the capacity to program cells that 639 recognize Aβ as a marker of local disease and consequently execute prescribed functions 640 independent of Aβ clearance. The ability to enlist synNotch cells to regulate any 641 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 17, 2025. ; https://doi.org/10.1101/2025.03.12.642808doi: bioRxiv preprint 16 transcriptional output distinguishes this cell engineering strategy not only from passive 642 immunotherapy approaches, but also from CAR -macrophage approaches, which aim to 643 enhance Αβ phagocytosis upon CAR stimulation. The synNotch platform has been used 644 alongside CAR technology to program additional layers of logic into the T cell design;74,75 645 this approach is now being deployed in human clinical trials in the context of glioblastoma 646 (NCT06186401), demonstrating clinical relevance of the synNotch platform being applied 647 in the brain. However, reports demonstrating the use of synNotch to treat inputs specific 648 to neurodegenerative disease have not been published. Additionally, ours is the first use 649 of synNotch to program astrocytes, thus capitalizing on their native CNS surveillance 650 functions. 651 The modularity of synNotch lends immense utility to this system for developing 652 targeted therapies for AD. Here, w e have used the platform to express both 653 neuroprotective (e.g., BDNF) and anti -inflammatory (e.g., sTNFR1, IL -1Ra) transgenes 654 from the same A β-synNotch receptor. SynNotch-astrocytes exquisitely regulated BDNF 655 production in an Aβ-dependent manner. Additionally, Aβ-dependent sTNFR1 and IL-1Ra 656 expression by Bap-Notch astrocytes significantly attenuated astrocyte reactivity based on 657 the expression of representative genes CSF2 and SERPINA3. Further, astrocytes play a 658 major role in maintaining the BBB; in AD, breakdown of the BBB contributes to disease 659 progression.64 Notably, previous work by our collaborators has revealed a major role for 660 SERPINA3 in mediating inflammation -driven breakdown of the BBB through the TNF -661 STAT3 signaling axis .22 Indeed, w hen we co -cultured anti -inflammatory Bap -Notch 662 astrocytes with BMEC cells, synNotch-regulated, Aβ-driven sTNFR1 and IL-1Ra provided 663 protection against an inflammation-mediated disruption in the BMEC layer, demonstrating 664 the potential for these anti-inflammatory transgenes to combat BBB disruption in AD. The 665 use of the A β synNotch system is not restricted to the therapeutic transgenes selected 666 here. As more about the disease becomes understood, the platform can be adapted to 667 regulate relevant therapeutic candidate transgenes . Because synNotch transcriptional 668 responses are potentiated by direct contact with immobilized ligand , the platform allows 669 for the possibility of using synNotch to express therapeutic compounds that would be 670 detrimental if delivered systemically to the brain and instead gates their production on 671 disease-dependent features that accumulate as AD progresses . Future work will be 672 needed to demonstrate a long -term benefit of A β-regulated payload production on 673 neuronal health, BBB integrity, and cognition. 674 Translation of the Aβ-synNotch system into mouse models of AD will be imperative 675 to expand its utility in the context of an AD cell therapy. We demonstrated the ability of 676 Aβ-synNotch receptors to recognize Aβ in situ by delivering Bap-Notch mMSCs to OSCs 677 prepared from aged 5xFAD mice. This result lends confidence in the ability of this system 678 to be translated to an in vivo setting. The delivery of Bap-Notch mMSCs to OSCs mimics 679 the deployment of ex vivo engineered CAR-T cell therapy, the most successful and most 680 prevalent engineered cell therapy to date. In all the FDA -approved CAR-T products, T 681 cells are engineered ex vivo by lentivirus or g-retrovirus.76 Such a strategy allows for 682 exogenous brain cells to be used as the therapeutic cell , thereby enabling the massive 683 expansion of engineered, therapeutic cells outside the body and eliminating risk of 684 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 17, 2025. ; https://doi.org/10.1101/2025.03.12.642808doi: bioRxiv preprint 17 transducing off -target cell populations .77 Here, we demonstrate the ability to deploy 685 engineered, hiPSC-derived astrocytes with Bap-Notch, indicating potential for generating 686 designer CNS-resident cells for transplantation from autologous or allogeneic sources.78 687 Importantly, the concept of ex vivo engineering is particularly relevant for MSCs, which 688 we also tested here, and which have been investigated in Phase I/II clinical trials for AD.79 689 Of note, one investigational AD therapy, Lomecel-B, is composed of allogenic MSCs.80,81 690 The ability to use allogenic cells with a different genetic background is of particular interest 691 to this application. Presenilin 1 (commonly mutated in familial AD) is a component of the 692 g-secretase complex, which functions in Aβ processing, but is also crucial to the function 693 of the synNotch receptor platform. While the majority of AD cases do not involve FAD 694 presenilin mutations, and PSEN1 mutations are not clearly loss-of-function, the ability to 695 use allogenic cells would mitigate potential influence of genetic background on the 696 function of the receptor platform. Altogether, these considerations foreshadow 697 widespread applications of an ex vivo engineering strategy that is compatible with use of 698 patient-specific, stem cell -derived CNS cells and allogeneic cell sources that may 699 circumvent concerns related to patient disease-associated genotypes. 700 The ability to express the Aβ-synNotch system in brain resident astrocytes offers 701 an attractive alternative approach to cell transplantation . This method of delivery would 702 allow for the synNotch cell to take advantage of the native CNS surveillance functions of 703 astrocytes. Additionally, advancements in adeno -associated virus (AAV) capsid 704 engineering have resulted in capsids that can efficiently cross the BBB . Further, AAV 705 capsids with engineered tropism allow for preferential targeting of cell types for 706 transduction.82 Specifically, the PHP.B capsid transduces brain cells at lower viral doses 707 than the previous benchmark for CNS delivery , AAV9, and results in more efficient 708 astrocyte transduction.83 Delivery of the synNotch circuit via PHP.B AAV may allow for 709 delivery via systemic injection and could open the door to an off-the-shelf therapeutic that 710 does not require ex vivo cell engineering . Due to inherent AAV packaging capacity 711

(<4.7 kb), the construction and validation of the smaller Bap-SNIPR receptor, 712 possessing the same Aβ recognition motif and similar intramembrane proteolysis 713 regulation as the Bap -Notch receptor , represents a major step towards porting the 714 programmable Aβ-responsive receptor system to AAV. Future work is required to deliver 715 AAV to aged 5xFAD mice and confirm Aβ-dependent transgene expression in vivo. 716 Bap-Notch was selected for our work generating Aβ-dependent BDNF expression, 717 attenuating astrocyte inflammation, and detecting A β in the 5xFAD mouse brain. Bap -718 Notch consistently resulted in a wide dynamic signaling range against A β in a variety of 719 contexts and is expressed efficiently in multiple cell types, including fibroblasts, MSCs, 720 and astrocytes. Notably, bapineuzumab, the antibody from which the Bap -Notch 721 recognition domain is derived, can detect a range of A β species, including both soluble 722 and insoluble Aβ species.39,40 This supports the case for using Bap -Notch to generate a 723 synNotch-based cell therapy for AD. However, the other receptors constructed here, 724 Gant-Notch and Don-Notch, also recognize both synthetic Aβ42 and Aβ40. It is interesting 725 that there are differences in their ability to detect A β preparations, which is likely due to 726 the different species of A β the antibodies are designed to detect. Future work 727 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 17, 2025. ; https://doi.org/10.1101/2025.03.12.642808doi: bioRxiv preprint 18 characterizing the ability of the synNotch receptors constructed from the antibodies to 728 recognize specific species may result in a cell capable of finely discriminating between 729 Aβ species, as well as between Aβ and non-Aβ components of the periplaque niche. This 730 may be of particular interest in the development of an A β-aggregation sensor system. 731 Tau propagation sensors have been developed, 84 but to our knowledge, there is no 732 equivalent Aβ system. 733 In conclusion , this work integrates a cell control module into a precision 734 regenerative medicine strategy to treat AD. We have generated cells that can detect AD 735 pathology and, in response, regulate therapeutic transgene expression. This platform 736 functions both in brain -resident cells (here, astrocytes) and a clinical AD cell therapy 737 candidate (MSCs), and thus represents a significant advance in the potential therapies 738 for AD. The modularity of the system allows for flexibility in the choice of targeted 739 transgene therapeutics. In all the AD drugs approved in the last 25 years, including 740 Leqembi (lecanemab) and Kisunla (donanemab), the therapeutic benefit is reliant on 741 clearance of A β protein aggregates. Such anti -Aβ monoclonal antibodies have 742 demonstrated extremely limited cognitive benefit and are associated with potentially 743 serious side effects, despite their ability to engage A β. Our approach builds upon the 744 massive research effort to discover such anti-Aβ antibodies by repurposing them as the 745 recognition domains of synthetic receptors. Crucially, we selected transgenes to address 746 aspects of disease (BDNF, inflammatory cytokine antagonists) that intersect with facets 747 of AD pathology that are not resolved by simply clearing A β from the AD brain (synapse 748 and neuron loss; neuroinflammation; astrocyte -mediated BBB disruption). Further, the 749 platform affords for autoregulation of therapeutic factors on the basis of local amyloid 750 disease burden, thereby potentially avoiding complications from wide -spread, 751 unregulated production of biologic drugs throughout the brain. Thus, this work represents 752 a powerful advance over traditional therapies or investigative cell -based/gene therapies 753 applied to AD, which lack a means for feedback -controlled cell functions based on 754 pathology within the AD niche. 755 756

757 The authors acknowledge funding support from NIH R21 AG086883 (JMB, ESL, MSS) , 758 NSF CAREER Award CBET-2237639 (JMB) , the Vanderbilt Memory and Alzheimer’s 759 Center ( VMAC) P20 AG068082 , the NSF GRFP (MRS), the Integrated Training in 760 Engineering and Diabetes T32 DK101003 (DC), the SyBBURE Searle Undergraduate 761 Research and Vanderbilt University School of Engineering Summer Research Programs 762 (DJC), and the Vanderbilt Training Program in Environment Toxicity T32 ES007028 (AL). 763 FACS was performed at the Vanderbilt Flow Cytometry Shared Resource, which is 764 supported by the Vanderbilt Ingram Cancer Center (P30 CA068485) and the Vanderbilt 765 Digestive Disease Research Center (DK058404). We thank Zach Lamantia for technical 766 assistance. 767 768 Disclosures 769 JMB, MRS, and ESL have disclosed intellectual property pertaining to this manuscript. 770 771 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 17, 2025. ; https://doi.org/10.1101/2025.03.12.642808doi: bioRxiv preprint 19 Figure 1. Engineered synNotch receptors enable programmable cellular 772 recognition of immobilized, synthetic Aβ. (A) Schematic of Notch and synNotch. Notch 773 recognizes its ligand Delta presented in trans by a neighboring cell, resulting in cleavage 774 of the transmembrane core and releasing the intracellular domain to the nucleus for target 775 gene regulation. Aβ-synNotch receptors were constructed by replacing the extracellular 776 domain with anti-Aβ monoclonal antibody single chain variable fragments (scFvs) and the 777 intracellular domain (ICD) with the tet-transactivator synthetic transcription factor (TF). 778 ScFvs were generated from anti-Aβ antibodies gantenerumab (Gant-Notch), donanemab 779 (Don-Notch), and bapineuzumab. Here, activation of synNotch receptors by Aβ will result 780 in expression of a luciferase transgene. (B) Depiction of gantenerumab, donanemab, and 781 bapineuzumab epitopes on the Aβ peptide. (C) Aβ-driven luciferase transgene expression 782 of mouse L929 fibroblasts expressing Bap-Notch, Don-Notch, and Gant-Notch plated on 783 biotinylated A β42 immobilized to plate surface s via streptavidin. Fold induction of 784 luminescence compared to cells plated on a control , Aβ-free surface. LaG16-Notch is a 785 negative control (GFP-responsive) receptor. n = 13 from 3 independent experiments ; 786 mean±SD; **p<0.01, ****p<0.0001 from one -way ANOVA with Tukey’s multiple 787 comparisons. All groups are significantly different from LaG16-Notch control (statistics not 788 shown). (D) Luciferase transgene expression of synNotch cells plated on immobilized 789 biotinylated A β40. n = 11-13 from 3 independent experiments ; mean±SD; **p<0.01, 790 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 17, 2025. ; https://doi.org/10.1101/2025.03.12.642808doi: bioRxiv preprint 20 ***p<0.001 from one -way ANOVA with Tukey’s multiple comparisons. All groups are 791 significantly different from LaG16-Notch control unless shown. 792 793 794 795 Figure 2. Aβ-synNotch receptors display differential levels of recognition of 796 unmodified Aβ peptides. Fold change in Aβ-driven luciferase transgene expression of 797 synNotch cells plated on (A) Aβ42 adsorbed to the plate , (B) Aβ42 supplemented to the 798 medium, (C) Aβ40 adsorbed to the plate, and (D) Aβ40 supplemented to the medium. n = 799 10-13; mean±SD; *p<0.05, **p<0.01, ****p<0.0001 from one -way ANOVA with Tukey’s 800 multiple comparisons. All groups are significantly different from LaG16 -Notch control 801 unless shown. 802 803 804 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 17, 2025. ; https://doi.org/10.1101/2025.03.12.642808doi: bioRxiv preprint 21 805 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 17, 2025. ; https://doi.org/10.1101/2025.03.12.642808doi: bioRxiv preprint 22 Figure 3. SynNotch astrocytes regulate programmable Aβ-dependent transgene 806 expression. (A) Schematic of astrocyte derivation from hiPSCs. (B) Expression of 807 astrocytic markers CD44 and GFAP. DAPI shows cell nuclei. Scale bar = 200 µm. (C) 808 Representative A β-dependent mCherry expression in Gant -Notch and Bap -Notch 809 astrocytes cultured with patient-derived Aβ seeds. Scale bar = 200 µm. (D) Aβ-dependent 810 BDNF expression in Bap-Notch astrocytes programmed with either a BDNF -mCherry 811 transgene or an mCherry only transgene and plated on immobilized synthetic Aβ42. n = 3; 812 mean±SD; ****p<0.0001 from two-way ANOVA with Tukey’s multiple comparisons. (E) 813 Representative Aβ-dependent mCherry expression in Bap-Notch astrocytes programmed 814 with either a BDNF -mCherry transgene or a n mCherry only transgene and plated on 815 immobilized synthetic Aβ42. Scale bar = 200 µm. 816 817 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 17, 2025. ; https://doi.org/10.1101/2025.03.12.642808doi: bioRxiv preprint 23 818 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 17, 2025. ; https://doi.org/10.1101/2025.03.12.642808doi: bioRxiv preprint 24 Figure 4. Bap-Notch attenuates astrocyte reactivity and supports BMEC barrier 819 function. (A) Schematic of experimental configuration. Bap-Notch astrocytes are plated 820 on immobilized synthetic A β42. Inflammatory cytokines TNF and IL -1α are added to the 821 medium. Recognition of A β results in the astrocytes expressing inflammatory cytokine 822 antagonists sTNFR1 and IL -1Ra. (B) sTNFR1 transgene expression of Bap -Notch 823 astrocytes plated on immobilized Aβ42 and treated with TNF and IL-1α. n = 5; mean±SD; 824 ****p<0.0001 from two-way ANOVA with Tukey’s multiple comparisons. (C) IL-1Ra 825 transgene expression of Bap -Notch astrocytes plated on immobilized A β42 and treated 826 with TNF and IL -1α. n =5; mean±SD; ns: not significant from unpaired t test. n.d. = not 827 detected. (D) Reactive gene expression in Bap-Notch astrocytes that express an sTNFR1 828 and IL-1Ra transgene in response to Aβ. Bap-Notch astrocytes were plated on either an 829 Aβ or control surface, and half were treated with TNF and IL-1α. n = 5; mean±SD; *p<0.05, 830 ****p<0.0001 from two-way ANOVA with Tukey’s multiple comparisons . (E) Reactive 831 gene expression in Bap -Notch astrocytes that express a SEAP reporter in response to 832 Aβ. n = 5; mean±SD; ns: not significant from two-way ANOVA with Tukey’s multiple 833 comparisons. (F) Schematic of transwell co -culture of Bap -Notch astrocytes and brain 834 microvascular endothelial cells (BMECs). (G) Transendothelial electrical resistance 835 (TEER) across BMECs co-cultured with Bap-Notch astrocytes and treated with TNF and 836 IL-1α on Day 0. Bap-Notch astrocytes were programmed with either sTNFR1 and IL-1Ra 837 transgenes or a SEAP transgene and were plated on either an Aβ or a control surface. n 838 = 3; mean±SD. (H) Area under the curve (AUC) for TEER measurements from Day 3 -5. 839 n = 3; mean±SD; *p<0.05 from two-way ANOVA with Tukey’s multiple comparisons. 840 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 17, 2025. ; https://doi.org/10.1101/2025.03.12.642808doi: bioRxiv preprint 25 841 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 17, 2025. ; https://doi.org/10.1101/2025.03.12.642808doi: bioRxiv preprint 26 Figure 5. Recognition of amyloid in situ in organotypic slice cultures. (A) Schematic 842 of coronal organotypic slice culture preparation. (B) Aβ-dependent SEAP expression of 843 Bap-Notch or LaG16 -Notch mMSCs plated on hemizygous 5xFAD or wild -type 844 organotypic slice cultures. n = 3-11 biological replicates; mean±SD; ****p<0.0001 from 845 two-way ANOVA with multiple comparisons. (C-D) Fluorescence microscopy of Aβ (6E10 846 immunolabeled), constitutive BFP, and synNotch-driven mCherry (immunolabeled) in 847 cryosections of hemizygous 5xFAD (C) or wild -type (D) OSCs treated with Bap -Notch 848 mMSCs. Scale bar = 200 µm. 849 850 851 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 17, 2025. ; https://doi.org/10.1101/2025.03.12.642808doi: bioRxiv preprint 27

852 853 1. Knopman, D. S. et al. Alzheimer disease. Nat Rev Dis Primers 7, 33 (2021). 854 2. Hanseeuw, B. J. et al. Association of amyloid and tau with cognition in preclinical 855 Alzheimer disease: A longitudinal study. JAMA Neurol 76, 915–924 (2019). 856 3. Hampel, H. et al. The amyloid-β pathway in Alzheimer’s disease. Mol Psychiatry 857 26, 5481–5503 (2021). 858 4. Schneider, L. A resurrection of aducanumab for Alzheimer’s disease. Lancet Neurol 859 19, 111–112 (2020). 860 5. Bohrmann, B. et al. Gantenerumab: A novel human anti-Aβ antibody demonstrates 861 sustained cerebral amyloid -β binding and elicits cell -mediated removal of human 862 amyloid-β. J Alzheimers Dis 28, 49–69 (2012). 863 6. Adolfsson, O. et al. An effector-reduced anti-β-amyloid (Aβ) antibody with unique 864 Aβ binding properties promotes neuroprotection and glial engulfment of A β. J 865 Neurosci 32, 9677–9689 (2012). 866 7. Ultsch, M. et al. Structure of crenezumab complex with Aβ shows loss of β-hairpin. 867 Sci Rep 6, 39374 (2016). 868 8. Mintun, M. A. et al. Donanemab in early Alzheimer’s disease. N Engl J Med 384, 869 1691–1704 (2021). 870 9. Salloway, S. et al. Two phase 3 trials of bapineuzumab in mild -to-moderate 871 Alzheimer’s Disease. N Engl J Med 370, 322–333 (2014). 872 10. Tarawneh, R. & Pankratz, V. S. The search for clarity regarding “clinically 873 meaningful outcomes” in Alzheimer disease clinical trials: CLARITY -AD and 874 beyond. Alzheimers Res Ther 16, 37 (2024). 875 11. Solopova, E. et al. Fatal iatrogenic cerebral β-amyloid-related arteritis in a woman 876 treated with lecanemab for Alzheimer’s disease. Nat Commun 14, 8220 (2023). 877 12. Budd Haeberlein, S. et al. Two randomized phase 3 studies of aducanumab in early 878 Alzheimer’s disease. J Prev Alzheimers Dis 9, 197–210 (2022). 879 13. van Dyck, C. H. et al. Lecanemab in early Alzheimer’s disease. N Engl J Med 388, 880 9–21 (2023). 881 14. Sims, J. R. et al. Donanemab in early symptomatic Alzheimer disease: The 882 TRAILBLAZER-ALZ 2 randomized clinical trial. JAMA 330, 512–527 (2023). 883 15. Salloway, S. et al. Amyloid-related imaging abnormalities in 2 phase 3 studies 884 evaluating aducanumab in patients with early Alzheimer disease. JAMA Neurol 79, 885 13–21 (2022). 886 16. Stanca, S., Rossetti, M. & Bongioanni, P. Astrocytes as neuroimmunocytes in 887 Alzheimer’s disease: A biochemical tool in the neuron –glia crosstalk along the 888 pathogenetic pathways. Int J Mol Sci 24, 13880 (2023). 889 17. Cai, Y., Liu, J., Wang, B., Sun, M. & Yang, H. Microglia in the neuroinflammatory 890 pathogenesis of Alzheimer’s disease and related therapeutic targets. Front 891 Immunol 13, 856376 (2022). 892 18. Mishra, A., Kim, H. J., Shin, A. H. & Thayer, S. A. Synapse loss induced by 893 interleukin-1β requires pre - and post -synaptic mechanisms. J Neuroimmune 894 Pharmacol 7, 571 (2012). 895 19. Olmos, G. & Lladó, J. Tumor necrosis factor alpha: A link between 896 neuroinflammation and excitotoxicity. Mediators Inflamm 2014, 861231 (2014). 897 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 17, 2025. ; https://doi.org/10.1101/2025.03.12.642808doi: bioRxiv preprint 28 20. Liu, L. R., Liu, J. C., Bao, J. S., Bai, Q. Q. & Wang, G. Q. Interaction of microglia 898 and astrocytes in the neurovascular unit. Front Immunol 11, 1024 (2020). 899 21. Liddelow, S. A. et al. Neurotoxic reactive astrocytes are induced by activated 900 microglia. Nature 541, 481–487 (2017). 901 22. Kim, H. et al. Reactive astrocytes transduce inflammation in a blood -brain barrier 902 model through a TNF -STAT3 signaling axis and secretion of alpha 1 -903 antichymotrypsin. Nat Commun 13, 6581 (2022). 904 23. Liu, C. Y., Yang, Y., Ju, W. N., Wang, X. & Zhang, H. L. Emerging roles of astrocytes 905 in neuro-vascular unit and the tripartite synapse with emphasis on reactive gliosis 906 in the context of alzheimer’s disease. Front Cell Neurosci 12, 391246 (2018). 907 24. Alnefaie, A. et al. Chimeric antigen receptor T -cells: An overview of concepts, 908 applications, limitations, and proposed solutions. Front Bioeng Biotechnol 10, 976 909 (2022). 910 25. Sterner, R. C. & Sterner, R. M. CAR-T cell therapy: current limitations and potential 911 strategies. Blood Cancer J 11, 69 (2021). 912 26. Sheykhhasan, M. et al. CAR T therapies in multiple myeloma: unleashing the future. 913 Cancer Gene Ther 31, 667–686 (2024). 914 27. Kim, A. B. et al. Chimeric antigen receptor macrophages target and resorb amyloid 915 plaques. JCI Insight 9, e175015 (2024). 916 28. Kopan, R. & Ilagan, M. X. G. The canonical Notch signaling pathway: Unfolding the 917 activation mechanism. Cell 137, 216–233 (2009). 918 29. Lee, J. C. et al. Instructional materials that control cellular activity through synthetic 919 Notch receptors. Biomaterials 297, 122099 (2023). 920 30. Brien, H. J. et al. Templated pluripotent stem cell differentiation via substratum -921 guided artificial signaling. ACS Biomater Sci Eng 10, 6465–6482 (2024). 922 31. Kumar, K. K. et al. Cellular manganese content is developmentally regulated in 923 human dopaminergic neurons. Sci Rep 4, 6801 (2014). 924 32. Hollmann, E. K. et al. Accelerated differentiation of human induced pluripotent stem 925 cells to blood-brain barrier endothelial cells. Fluids Barriers CNS 14, 9 (2017). 926 33. Neal, E. H. et al. A Simplified, Fully Defined Differentiation Scheme for Producing 927 Blood-Brain Barrier Endothelial Cells from Human iPSCs. Stem Cell Reports 12, 928 1380–1388 (2019). 929 34. Zufferey, R., Nagy, D., Mandel, R. J., Naldini, L. & Trono, D. Multiply attenuated 930 lentiviral vector achieves efficient gene delivery in vivo. Nat Biotechnol 15, 871–875 931 (1997). 932 35. Xu, F. et al. Cerebral vascular amyloid seeds drive amyloid β-protein fibril assembly 933 with a distinct anti-parallel structure. Nat Commun 7, 13527 (2016). 934 36. Klim, J. R., Li, L., Wrighton, P. J., Piekarczyk, M. S. & Kiessling, L. L. A defined 935 glycosaminoglycan-binding substratum for human pluripotent stem cells. Nat 936

7, 989–994 (2010). 937 37. Wrighton, P. J. et al. Signals from the surface modulate differentiation of human 938 pluripotent stem cells through glycosaminoglycans and integrins. Proc Natl Acad 939 Sci U S A 111, 18126–18131 (2014). 940 38. DeMattos, R. B. et al. A plaque-specific antibody clears existing β-amyloid plaques 941 in Alzheimer’s disease mice. Neuron 76, 908–920 (2012). 942 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 17, 2025. ; https://doi.org/10.1101/2025.03.12.642808doi: bioRxiv preprint 29 39. Bard, F. et al. Peripherally administered antibodies against amyloid β-peptide enter 943 the central nervous system and reduce pathology in a mouse model of Alzheimer 944 disease. Nat Med 6, 916–919 (2000). 945 40. Miles, L. A., Crespi, G. A. N., Doughty, L. & Parker, M. W. Bapineuzumab captures 946 the N -terminus of the Alzheimer’s disease amyloid -beta peptide in a helical 947 conformation. Sci Rep 3, 1302–6 (2013). 948 41. Escott-Price, V. & Schmidt, K. M. Probability of Alzheimer’s disease based on 949 common and rare genetic variants. Alzheimers Res Ther 13, 140 (2021). 950 42. Sperling, R. A. et al. The impact of amyloid-beta and tau on prospective cognitive 951 decline in older individuals. Ann Neurol 85, 181–193 (2019). 952 43. Sevigny, J. et al. The antibody aducanumab reduces A β plaques in Alzheimer’s 953 disease. Nature 537, 50–56 (2016). 954 44. Arndt, J. W. et al. Structural and kinetic basis for the selectivity of aducanumab for 955 aggregated forms of amyloid-β. Sci Rep 8, 6412 (2018). 956 45. Englund, H. et al. Sensitive ELISA detection of amyloid -β protofibrils in biological 957 samples. J Neurochem 103, 334–345 (2007). 958 46. Lord, A. et al. An amyloid-β protofibril-selective antibody prevents amyloid formation 959 in a mouse model of Alzheimer’s disease. Neurobiol Dis 36, 425–434 (2009). 960 47. Morsut, L. et al. Engineering customized cell sensing and response behaviors using 961 synthetic Notch receptors. Cell 164, 780–791 (2016). 962 48. Marina, G. B. et al. Amyloid β-protein (Aβ) assembly: Aβ40 and Aβ42 oligomerize 963 through distinct pathways. Proc Natl Acad Sci U S A 100, 330–335 (2003). 964 49. Garland, E. F., Hartnell, I. J. & Boche, D. Microglia and Astrocyte Function and 965 Communication: What Do We Know in Humans? Front Neurosci 16, 824888 966 (2022). 967 50. Siracusa, R., Fusco, R. & Cuzzocrea, S. Astrocytes: Role and functions in brain 968 pathologies. Front Pharmacol 10, 479091 (2019). 969 51. Lee, H. G., Wheeler, M. A. & Quintana , F. J. Function and therapeutic value of 970 astrocytes in neurological diseases. Nat Rev Drug Discov 21, 339–358 (2022). 971 52. Lee, Y., Messing, A., Su, M. & Brenner, M. GFAP promoter elements required for 972 region-specific and astrocyte-specific expression. Glia 56, 481–493 (2008). 973 53. Nagahara, A. H. et al. Neuroprotective effects of brain-derived neurotrophic factor 974 in rodent and primate models of Alzheimer’s disease. Nat Med 15, 331–337 (2009). 975 54. Nagahara, A. H. et al. MR-guided delivery of AAV2-BDNF into the entorhinal cortex 976 of non-human primates. Gene Ther 25, 104–114 (2018). 977 55. Nagahara, A. H. et al. Early BDNF treatment ameliorates cell loss in the entorhinal 978 cortex of APP transgenic mice. J Neurosci 33, 15596–15602 (2013). 979 56. Nagahara, A. H. & Tuszynski, M. H. Potential therapeutic uses of BDNF in 980 neurological and psychiatric disorders. Nat Rev Drug Discov 10, 209–219 (2011). 981 57. Corder, E. H. et al. Gene dose of apolipoprotein E type 4 allele and the risk of 982 Alzheimer’s disease in late onset families. Science (1979) 261, 921–923 (1993). 983 58. Hollingworth, P. et al. Common variants at ABCA7, MS4A6A/MS4A4E, EPHA1, 984 CD33 and CD2AP are associated with Alzheimer’s disease. Nat Genet 43, 429–985 435 (2011). 986 59. Jonsson, T. et al. Variant of TREM2 associated with the risk of Alzheimer’s disease. 987 N Engl J Med 368, 107–116 (2013). 988 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 17, 2025. ; https://doi.org/10.1101/2025.03.12.642808doi: bioRxiv preprint 30 60. Swardfager, W. et al. A meta-analysis of cytokines in Alzheimer’s disease. Biol 989 Psychiatry 68, 930–941 (2010). 990 61. Guttikonda, S. R. et al. Fully defined human pluripotent stem cell-derived microglia 991 and tri-culture system model C3 production in Alzheimer’s disease. Nat Neurosci 992 24, 343–354 (2021). 993 62. Leng, K. et al. CRISPRi screens in human iPSC -derived astrocytes elucidate 994 regulators of distinct inflammatory reactive states. Nat Neurosci 25, 1528 –1542 995 (2022). 996 63. Guttenplan, K. A. et al. Neurotoxic reactive astrocytes induce cell death via 997 saturated lipids. Nature 599, 102–107 (2021). 998 64. Sweeney, M. D., Sagare, A. P. & Zlokovic, B. V. Blood–brain barrier breakdown in 999 Alzheimer disease and other neurodegenerative disorders. Nat Rev Neurol 14, 1000 133–150 (2018). 1001 65. Regmi, S. et al. Mesenchymal stromal cells for the treatment of Alzheimer’s 1002 disease: Strategies and limitations. Front Mol Neurosci 15, 1011225 (2022). 1003 66. Lee, N. K. et al. Exploring the potential of mesenchymal stem cell-based therapy in 1004 mouse models of vascular cognitive impairment. Int J Mol Sci 21, 5524 (2020). 1005 67. Walton, B. L. et al. A programmable arthritis-specific receptor for guided articular 1006 cartilage regenerative medicine. bioRxiv (2024) doi:10.1101/2024.01.31.578281. 1007 68. Uccelli, A., Moretta, L. & Pistoia, V. Mesenchymal stem cells in health and disease. 1008 Nat Rev Immunol 8, 726–736 (2008). 1009 69. Richard, B. C. et al. Gene dosage dependent aggravation of the neurological 1010 phenotype in the 5XFAD mouse model of Alzheimer’s disease. J Alzheimers Dis 1011 45, 1223–1236 (2015). 1012 70. Oakley, H. et al. Intraneuronal β-Amyloid aggregates, neurodegeneration, and 1013 neuron loss in transgenic mice with five familial Alzheimer’s disease mutations: 1014 Potential factors in amyloid plaque formation. J Neurosci 26, 10129–10140 (2006). 1015 71. Marrone, L., Marchi, P. M. & Azzouz, M. Circumventing the packaging limit of AAV-1016 mediated gene replacement therapy for neurological disorders. Expert Opin Biol 1017 Ther 22, 1163–1176 (2022). 1018 72. Zhu, I. et al. Modular design of synthetic receptors for programmed gene regulation 1019 in cell therapies. Cell 185, 1431-1443.e16 (2022). 1020 73. Tolar, M., Abushakra, S., Hey, J. A., Porsteinsson, A. & Sabbagh, M. Aducanumab, 1021 gantenerumab, BAN2401, and ALZ-801 - The first wave of amyloid-targeting drugs 1022 for Alzheimer’s disease with potential for near term approval. Alzheimers Res Ther 1023 12, 95 (2020). 1024 74. Roybal, K. T. et al. Precision tumor recognition by T cells with combinatorial antigen-1025 sensing circuits. Cell 164, 770–779 (2016). 1026 75. Choe, J. H. et al. SynNotch-CAR T cells overcome challenges of specificity, 1027 heterogeneity, and persistence in treating glioblastoma. Sci Transl Med 13, 1028 eabe7378 (2021). 1029 76. Irving, M., Lanitis, E., Migliorini, D., Ivics, Z. & Guedan, S. Choosing the right tool 1030 for genetic engineering: Clinical lessons from chimeric antigen receptor -T cells. 1031 Hum Gene Ther 32, 1044–1058 (2021). 1032 77. Short, L., Holt, R. A., Cullis, P. R. & Evgin, L. Direct in vivo CAR T cell engineering. 1033 Trends Pharmacol Sci 45, 406–418 (2024). 1034 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 17, 2025. ; https://doi.org/10.1101/2025.03.12.642808doi: bioRxiv preprint 31 78. Sallman, D. A. et al. Ameli-01: A phase I trial of UCART123v1.2, an anti -CD123 1035 allogeneic CAR-T cell product, in adult patients with relapsed or refractory (R/R) 1036 CD123+ acute myeloid leukemia (AML). Blood 140, 2371–2373 (2022). 1037 79. Thompson, M. et al. Cell therapy with intravascular administration of mesenchymal 1038 stromal cells continues to appear safe: An updated systematic review and meta -1039 analysis. EClinicalMedicine 19, 100249 (2020). 1040 80. Brody, M. et al. Results and insights from a phase I clinical trial of Lomecel -B for 1041 Alzheimer’s disease. Alzheimers Dement 19, 261–273 (2023). 1042 81. Oliva Jr, A. A. et al. Safety and efficacy of Lomecel -B in patients with mild 1043 Alzheimer’s disease: Results of a double -blinded, randomized, placebo-controlled 1044 phase 1 clinical trial. Alzheimers Dement 17, e057581 (2021). 1045 82. Lopez-Gordo, E., Chamberlain, K., Riyad, J. M., Kohlbrenner, E. & Weber, T. 1046 Natural adeno-associated virus serotypes and engineered adeno-associated virus 1047 capsid variants: tropism differences and mechanistic insights. Viruses 16, 442 1048 (2024). 1049 83. Deverman, B. E. et al. Cre-dependent selection yields AAV variants for widespread 1050 gene transfer to the adult brain. Nat Biotechnol 34, 204–209 (2016). 1051 84. Kaufman, S. K. et al. Tau prion strains dictate patterns of cell pathology, progression 1052 rate, and regional vulnerability in vivo. Neuron 92, 796–812 (2016). 1053 1054 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 17, 2025. ; https://doi.org/10.1101/2025.03.12.642808doi: bioRxiv preprint