Methods and compositions for treating alzheimer's disease

Abstract

Provided are methods and compositions for treating Alzheimer's disease (AD). Aspects of the present disclosure include methods of treating Alzheimer's disease (AD), the methods comprising administering to a subject having Alzheimer's disease a therapeutically effective amount of irinotecan and letrozole. Aspects of the present disclosure further include pharmaceutical compositions comprising irinotecan, letrozole, and a pharmaceutically acceptable carrier. Use of such compositions for treating Alzheimer's disease (AD) are also provided.

Description

[0001] METHODS AND COMPOSITIONS FOR T REATING ALZHEIMER’S DISEASE

[0002] CROSS-REFERENCE TO RELATED APPLICATIONS

[0003] This application claims the benefit of U.S. Provisional Patent Application No. 63 / 603,372, filed November 28, 2023, which application is incorporated herein by reference in its entirety.

[0004] STATEMENT OF GOVERNMENT SUPPORT

[0005] This invention was made with Government support under contracts AG060393 and AG057683 awarded by the National Institutes of Health. The Government has certain rights in the invention.

[0006] INTRODUCTION

[0007] Alzheimer’s disease (AD) is a neurodegenerative disorder with severe impacts on individuals, families, and society1, and yet without a cure. AD patients suffer progressive memory loss, behavior changes, and cognitive and motor deficits, leading to diminished quality of life2. Over 50 million people currently live with AD or other related dementias worldwide, a number projected to triple by 205034. With global costs exceeding $1 trillion annually, AD is one of the most costly health conditions worldwide5. Urgent action is needed to develop effective and accessible treatments.

[0008] Despite rigorous preclinical and clinical research efforts, AD drug development faces significant challenges, with a 98% failure rate in recent decades6. Current treatments are mostly limited to symptom-managing treatments7. Furthermore, recently approved immunotherapies have only modest effects on disease progression8. The lack of effective treatments stems from the pathological heterogeneity in AD. While AD’s most prominent disease hallmark is proteopathy, characterized by extracellular amyloid-p (Ap) plaques and intracellular tau neurofibrillary tangles (NFTs), their interplay and exact mechanisms leading to disease remain unclear9. Genetic heterogeneity further complicates the disease, including risk mutations in the amyloid precursor protein (APP) and presenilin genes, and APOE4, a risk isoform of apolipoprotein E (APOE)10. Emerging evidence highlights the critical roles of different brain cell types, with neuroinflammation and inadequate neuronal support from malfunctioning glial cells contributing to AD progression11. Considering the multifactorial nature of AD, traditional therapeutic approaches focusing on single disease hallmarks or bulk-tissue-level pathologies are often insufficient, leading to variable treatment outcomes.

[0009] Given the unmet need for disease-modifying treatments, drug repurposing has gained interest due to its faster development, lower costs, and improved safety12. In addition, technical advancements in mining large-scale databases offer new opportunities for discovering promising candidates. These developments, combined with various drug screening approaches such as in vitro'3and in vivo assays14, gene signature matching15, network modeling16, machine learning17, and data mining18, have identified numerous repurposed candidates for AD over the past decade. However, the extensive array of potential candidates complicates the establishment of priorities for clinical translation19.

[0010] SUMMARY

[0011] Provided are methods and compositions for treating Alzheimer's disease (AD). Aspects of the present disclosure include methods of treating Alzheimer's disease (AD), the methods comprising administering to a subject having Alzheimer's disease a therapeutically effective amount of irinotecan and letrozole. Aspects of the present disclosure further include pharmaceutical compositions comprising irinotecan, letrozole, and a pharmaceutically acceptable carrier. Use of such compositions for treating Alzheimer's disease (AD) are also provided.

[0012] BRIEF DESCRIPTION OF THE FIGURES

[0013] FIG. 1 : Single nucleus transcriptomic profiling reveals both shared and cell-type- specific gene expression signatures in human AD samples, a. Data sources (publicly- available) and cohort summary, including sample filtering criteria and case-control standardization strategies, b. UMAP plot of merged samples from three independent studies. Study-associated variations are apparent without integration, c-e. UMAP plots of harmonized dataset using Seurat CCA integration algorithm, with study variation eliminated, and labeled by study (c), cell type identity (d), and case-control identification (e). f. Scattered bar plot of cell type abundance percentage for inhibitory neurons, comparing AD and controls (unpaired two-sided t- test). g. Projection plots for cell type to cell type communication measured by expression of receptor and ligand pairs and represented as arrows connecting two cell types. Red connections indicate increased communication and blue indicate decreased communications comparing AD vs controls. Arrows points to the directions from sender to receiver of the communication, h. Bar plot of differentially expressed gene (DEG) counts of AD versus controls in the six major brain cell types and separated into up and down regulated subgroups. Counts of DEGs unique to one cell type were superimposed on the total counts, i. DEG heatmap showing overlaps of top DEGs (highest absolute log fold changes) in each cell type, DEGs with opposite regulatory patterns in AD across cell types are bolded, j. Heatmap showing AD enriched KEGG pathways across cell types. Few functional pathways are shared by all six cell types. Oligodendrocytes cluster with excitatory and inhibitory neurons, suggesting more similarity among these cell types, while microglia cluster with astrocytes and OPCs.

[0014] FIG. 2: Computational drug repurposing pipeline predicts drug candidates reversing cell-type- specific transcriptomic signature profiles of AD. a, Schematic showing computational drug repurposing workflow, b, Input DEG counts (filtered list by mapping AD vs control DEGs to existing gene probes in the CMap database) per cell type for the computational drug repurposing pipeline, c, Disease-drug network depicting the connections between the six cell types and drug candidates that significantly reverse disease profiles within the respective cell types. Drug names in red denote drug candidates validated in humans using electronic medical records (EMR), d. Heatmap showing drug candidates that significantly reverse AD profiles in more than one cell type. Black frames label drugs validated in humans using the EMR, and red asteroids label drugs selected for validation in vivo using a mouse model of AD. e, AD prevalence assessment in drug-exposed individuals, only drugs with significant reduced AD risks are shown, f, Schematic illustration demonstrating the rationale of prioritizing letrozole and irinotecan as potential combination therapy for further validation in a mouse model of AD. g, Heatmaps of each cell-type-specific AD transcriptomic signature profiles, rank ordered genes from the most upregulated to the most downregulated and color coded by log fold changes, in comparisons with the gene probe ranks by letrozole or irinotecan treatments, colored by corresponding fold change ranks in the CMap.

[0015] FIG. 3: Combination treatment with letrozole and irinotecan rescues AD-like memory impairments in aged 5xFAD / PS19 mice with both Ab and tau pathologies, a. The schematic of treatment cohort design and experimental timeline (Created with BioRender.com). . Four treatment groups (n=20 per group, both sexes), including treatment with vehicle, letrozole (1 mg / kg), irinotecan (1 Omg / kg) or combination of both drugs every other day via i.p. injection, b. Escape latency during hidden platform training days 1-6 did not differ statistically between groups. One-way repeated-measures ANOVA test was applied to compare vehicle group with treatment groups. c,d. Memory probes of percent time spent in target quadrant versus average of other quadrants demonstrated a significant preference of the target quadrant solely by mice with combination treatment at 24 hour (c) and 72 hour (d) after removing the hidden platform. e,f. Memory probes measuring number of platform-location crossings in the target quadrant versus average of the other quadrants. Significantly more crossings in the target quadrant where the platform used to be only observed in the combination treatment group at 24 hour (e) and 72 hour (d) after the platform was removed, c-f. Ordinary one-way ANOVA test was performed between target and other quadrants for each group, and p- value adjusted with Bonferroni multiplecomparisons testing.

[0016] FIG. 4: AD pathologies are significantly reduced in 9-month-old 5xFAD / PS19 mice after drug treatments, with the strongest rescue in the combination treatment group, a. Representative images of the ventral hippocampus from 9-month-old 5xFAD / PS19 mice with Sudan Black staining to enhance hippocampal visualization (scale bar, 2mm). b. Quantification of hippocampal volume in 9-month-old 5xFAD / PS19 mice across treatment groups, c. Representative images of hippocampus from 9-month-old 5xFAD / PS19 mice with immunostaining of phosphorylated tau (p-tau) using AT8 monoclonal antibody (scale bar, 500mm). d. Quantification of AT8-positive p-tau percent coverage area in 9-month-old 5xFAD / PS19 mice across treatment groups, e. Representative images of Thio-S staining in the hippocampus from 9-month-old 5xFAD / PS19 mice (scale bar, 500mm). f. Quantification of Thio- S-positive percent coverage area and plaque counts in the hippocampus of 9-month-old 5xFAD / PS19 mice across treatment groups, g. Representative images of microglia immunostaining with anti-lba1 in the hippocampus of 9-month-old 5xFAD / PS19 mice (scale bar, 500mm). h. Quantification of the percent Iba1 coverage area in the hippocampus of 9-month-old 5xFAD / PS19 mice across treatment groups, i. Representative images of astrocyte immunostaining with anti-GFAP in the hippocampus of 9-month-old 5xFAD / PS19 mice (scale bar, 500mm). j. Quantification of percent GFAP coverage area in hippocampus of 9-month-old 5xFAD / PS19 mice . k. Representative images of CA1 neurons with neuronal marker NeuN immunostaining (scale bar, 100mm). I. Quantification of the thickness of the CA1 neuronal cell layer of 9-month-old 5xFAD / PS19 mice across treatment groups, b, d, f, h, j, I. Ordinary one-way ANOVA were applied between all treatment groups and vehicle control group, p-values were adjusted by Dunnett’s multiple comparison test.

[0017] FIG. 5: Single-nucleus RNA-sequencing (snRNA-seq) analysis in 9-month-old 5xFAD / PS19 mice across treatment groups, a. UMAP plot of all 31 distinct cell clusters in hippocampus of mice from combination-treatment and vehicle-treatment groups, b. Projection plots for cell type to cell type communication measured by expression of receptor and ligand pairs and represented as arrows connecting two cell types. Red connections indicate increased communication and blue indicate decreased communications comparing L+l-treated versus vehicle-treated groups. Arrows points to the directions from sender to receiver of the communication, c. UMAP plots labeled by hippocampal specific cell types and split by treatments to highlight difference in distribution density for neuronal subtypes between treatments, d. Scattered bar plots of cell type abundance (percentage) for CA1 and CA3 pyramidal neurons (unpaired two- sided t-test), comparing L+l-treated and vehicle-treated mice (n=8 per group, including both sexes), e. Differentially expressed gene (DEG) counts of L+l-treated versus vehicle-treated mice in six major brain cell types. Counts of DEGs per cell type do not correlate with cell counts, f. Top combination treatment-enriched Gene Ontology (GO) terms across six cell types, g. Heatmap illustration of enriched KEGG pathways across six cell types. Some treatment-enriched pathways overlap with AD signature pathways from integrated human snRNA-seq analysis (see Fig. 1 ) and were labeled in red.

[0018] FIG. 6: Combination-treatment with letrozole and irinotecan reverses cell-type- specific transcriptomic signatures of AD. a. Comparison of cell-type-specific transcriptom ic signature profiles of AD in humans with gene expression changes in combination treatment versus vehicle treatment groups of mice. Only treatment- reversed genes from AD profiles are included (abs (LFC) >0.01 in L+l-treated versus vehicle-treated mice), and those with FDR- adjusted p value < 0.05 are labeled in red. Heatmap colors indicate directions and magnitude of gene expression changes, with downregulations in blue and upregulations in red. b. Gene set enrichment analysis on treatment-reversed genes, which reached statistical significance, in excitatory and inhibitory neurons revealed reversal of multiple AD relevant functional pathways, c. Gene set enrichment analysis on treatment-reversed genes, which reached statistical significance, in astrocytes, microglia, or oligodendrocyte and OPCs demonstrated reversal of glial cell specific AD gene signatures and functional pathways.

[0019] DETAILED DESCRIPTION

[0020] Before the methods and compositions of the present disclosure are described in greater detail, it is to be understood that the methods and compositions are not limited to particular embodiments described, as such may, of course, vary. It is also to be understood that the terminology used herein is for the purpose of describing particular embodiments only, and is not intended to be limiting, since the scope of the methods and compositions will be limited only by the appended claims.

[0021] Where a range of values is provided, it is understood that each intervening value, to the tenth of the unit of the lower limit unless the context clearly dictates otherwise, between the upper and lower limit of that range and any other stated or intervening value in that stated range, is encompassed within the methods and compositions. The upper and lower limits of these smaller ranges may independently be included in the smaller ranges and are also encompassed within the methods and compositions, subject to any specifically excluded limit in the stated range. Where the stated range includes one or both of the limits, ranges excluding either or both of those included limits are also included in the methods and compositions.

[0022] Certain ranges are presented herein with numerical values being preceded by the term “about.” The term “about” is used herein to provide literal support for the exact number that it precedes, as well as a number that is near to or approximately the number that the term precedes. In determining whether a number is near to or approximately a specifically recited number, the near or approximating unrecited number may be a number which, in the context in which it is presented, provides the substantial equivalent of the specifically recited number.

[0023] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the methods and compositions belong. Although any methods and compositions similar or equivalent to those described herein can also be used in the practice or testing of the methods and compositions, representative illustrative methods and compositions are now described.

[0024] All publications and patents cited in this specification are herein incorporated by reference as if each individual publication or patent were specifically and individually indicated to be incorporated by reference and are incorporated herein by reference to disclose and describe the materials and / or methods in connection with which the publications are cited. The citation of any publication is for its disclosure prior to the filing date and should not be construed as an admission that the present methods and compositions are not entitled to antedate such publication, as the date of publication provided may be different from the actual publication date which may need to be independently confirmed.

[0025] It is noted that, as used herein and in the appended claims, the singular forms “a”, “an”, and “the” include plural referents unless the context clearly dictates otherwise. It is further noted that the claims may be drafted to exclude any optional element. As such, this statement is intended to serve as antecedent basis for use of such exclusive terminology as “solely,” “only” and the like in connection with the recitation of claim elements, or use of a “negative” limitation.

[0026] It is appreciated that certain features of the methods and compositions, which are, for clarity, described in the context of separate embodiments, may also be provided in combination in a single embodiment. Conversely, various features of the methods and compositions, which are, for brevity, described in the context of a single embodiment, may also be provided separately or in any suitable sub-combination. All combinations of the embodiments are specifically embraced by the present disclosure and are disclosed herein just as if each and every combination was individually and explicitly disclosed, to the extent that such combinations embrace operable processes and / or compositions. In addition, all sub-combinations listed in the embodiments describing such variables are also specifically embraced by the present methods and compositions and are disclosed herein just as if each and every such sub-combination was individually and explicitly disclosed herein.

[0027] As will be apparent to those of skill in the art upon reading this disclosure, each of the individual embodiments described and illustrated herein has discrete components and features which may be readily separated from or combined with the features of any of the other several embodiments without departing from the scope or spirit of the present methods. Any recited method can be carried out in the order of events recited or in any other order that is logically possible.

[0028] METHODS OF TREATING ALZHEIMER'S DISEASE

[0029] Aspects of the present disclosure include methods of treating Alzheimer’s disease (AD). In certain embodiments, the methods comprise administering to a subject having Alzheimer's disease a therapeutically effective amount of irinotecan and letrozole. The methods are based at least in part on a drug screening strategy driven entirely by human data, utilizing large-scale omic datasets from post-mortem brains22 24, a drug perturbation library generated in human cell lines25, and clinical records encompassing millions of individuals, thereby maximizing the chance of clinical translation. Proposed from the results of the screening strategy was the repurposing a combination of two existing drugs, letrozole20and irinotecan21, to reverse cell-type-specific gene expression alterations across multiple cell types implicated in AD. The predicted combination therapy was validated through dosing experiments in an AD mouse model2627, demonstrating that the combination therapy targeting both neurons and glial cells significantly ameliorated memory deficits and AD-related pathologies compared to vehicle treatment, and outperformed single-drug treatments targeting either neurons (letrozole) or glial cells (irinotecan) alone. Details regarding embodiments of the methods of the present disclosure will now be described.

[0030] Irinotecan (CAS No. 97682-44-5) is a derivative of camptothecin. The target of drugs of the camptothecin family is DNA topoisomerase I, a nuclear enzyme involved in the relaxation of the DNA double helix required for replication and transcription activities. The safety of irinotecan has been established and it is approved by the United States Food and Drug Administration (FDA) for the treatment of colon cancer, small cell lung cancer and metastatic pancreatic adenocarcinoma.

[0031] Letrozole (CAS No. 1 12809-51 -5) is a non-steroidal type II aromatase inhibitor. It is a third generation aromatase inhibitor, meaning it does not significantly affect cortisol, aldosterone, and thyroxine. The safety of letrozole has been established and it is approved by the FDA for the treatment of local or metastatic breast cancer that is hormone receptor positive or has an unknown receptor status in postmenopausal women.

[0032] The irinotecan and letrozole may be administered via a route of administration independently selected from oral, parenteral (e.g., by intravenous, intra-arterial, intra-cerebral subcutaneous, intramuscular, or epidural injection), topical, or intranasal administration. In certain embodiments, the irinotecan is administered parenterally, e.g., intravenously. According to some embodiments, the letrozole is administered orally. In some instances, the irinotecan is administered intravenously and the letrozole is administered orally.

[0033] The irinotecan and letrozole are administered to a subject (e.g., a human subject) having Alzheimer's disease (AD). That is, prior to the administration of irinotecan and letrozole, the subject has been identified (diagnosed) as having AD. Physicians use several methods and tools to help determine if a subject with memory or cognitive problems has AD. To diagnose AD, doctors may: ask the subject experiencing symptoms, as well as a family member or friend, questions about overall health, use of prescription and over-the-counter medicines, diet, past medical problems, ability to carry out daily activities, and changes in behavior and personality; conduct tests of memory, problem solving, attention, counting, and language; order blood, urine, and other standard medical tests that can help identify other possible causes of the problem; administer a psychiatric evaluation to determine if depression or another mental health condition is causing or contributing to a person's symptoms; collect cerebrospinal fluid (CSF) via a spinal tap and measure the levels of proteins associated with Alzheimer's and related dementias; and / or perform brain scans, such as computed tomography (CT), magnetic resonance imaging (MRI), or positron emission tomography (PET), to support an Alzheimer’s diagnosis or rule out other possible causes for symptoms.

[0034] AD tends to develop slowly and gradually worsens over several years. Eventually, AD affects most areas of the brain. Memory, thinking, judgment, language, problem-solving, personality and movement can all be affected by the disease. There are five stages associated with AD. They include: preclinical Alzheimer's disease; mild cognitive impairment (MCI) due to Alzheimer's disease; mild dementia due to Alzheimer's disease; moderate dementia due to Alzheimer's disease; and severe dementia due to Alzheimer's disease.

[0035] In certain embodiments, the subject to whom the irinotecan and letrozole are administered has been identified as having preclinical Alzheimer's disease. According to some embodiments, the subject to whom the irinotecan and letrozole are administered has been identified as having mild cognitive impairment (MCI) due to Alzheimer's disease. In some instances, the subject to whom the irinotecan and letrozole are administered has been identified as having mild dementia due to Alzheimer's disease. In certain embodiments, the subject to whom the irinotecan and letrozole are administered has been identified as having moderate dementia due to Alzheimer's disease. According to some embodiments, the subject to whom the irinotecan and letrozole are administered has been identified as having severe dementia due to Alzheimer's disease.

[0036] The irinotecan and letrozole are administered in a therapeutically effective amount. By “therapeutically effective amount” is meant a dosage sufficient to produce a desired result, e.g., an amount sufficient to effect beneficial or desired therapeutic (including preventative) results, such as a reduction in a symptom of Alzheimer's disease, non-limiting examples of which include memory loss, confusion, language problems, planning and problem-solving difficulties, personality and behavior changes, movement difficulties, and sleep disturbances. An effective amount can be administered in one or more administrations.

[0037] In some instances, the irinotecan and letrozole are administered concurrently. By “concurrently” is meant the irinotecan and letrozole are administered in the same composition to the subject or in separate compositions within six hours of each other to the subject. In other embodiments, the irinotecan and letrozole are administered sequentially, i.e., in separate compositions and spaced by six hours or more.

[0038] According to certain embodiments, the irinotecan is administered to the subject prior to administration of the letrozole, concurrently with administration of the letrozole, or both. In some embodiments, the letrozole is administered to the subject prior to administration of the irinotecan, concurrently with administration of the irinotecan, or both.

[0039] In some instances, the irinotecan and letrozole are administered according to a dosing regimen approved for individual use. In some embodiments, the administration of the letrozole permits the irinotecan to be administered according to a dosing regimen that involves one or more lower and / or less frequent doses, and / or a reduced number of cycles as compared with that utilized when the irinotecan is administered without administration of the letrozole. In some embodiments, the administration of the irinotecan permits the letrozole to be administered according to a dosing regimen that involves one or more lower and / or less frequent doses, and / or a reduced number of cycles as compared with that utilized when the letrozole is administered without administration of the irinotecan. As noted above, in some instances, one or more doses of the letrozole and irinotecan are administered at the same time; in some such embodiments, such agents may be administered present in the same pharmaceutical composition. In some embodiments, however, the letrozole and irinotecan are administered to the subject in different compositions and / or at different times. For example, the letrozole may be administered prior to administration of the irinotecan (e.g., in a particular cycle). Alternatively, the irinotecan may be administered prior to administration of the letrozole (e.g., in a particular cycle). The letrozole may be administered a period of time that starts at least 1 hour, 3 hours, 6 hours, 12 hours, 24 hours, 48 hours, 72 hours, or up to 5 days or more after the administration of the irinotecan. The irinotecan may be administered a period of time that starts at least 1 hour, 3 hours, 6 hours, 12 hours, 24 hours, 48 hours, 72 hours, or up to 5 days or more after the administration of the letrozole.

[0040] In one example, the letrozole is administered to the subject for a desirable period of time prior to administration of the irinotecan. In some embodiments, such a regimen “primes” the subject for the activity of the irinotecan. In another example, the irinotecan is administered to the subject for a desirable period of time prior to administration of the letrozole. In some embodiments, such a regimen “primes” the subject for the activity of the letrozole.

[0041] In some embodiments, administration of the irinotecan is specifically timed relative to administration of the letrozole. For example, in some embodiments, irinotecan is administered so that a particular effect is observed (or expected to be observed, for example based on population studies showing a correlation between a given dosing regimen and the particular effect of interest).

[0042] In certain embodiments, desired relative dosing regimens for agents administered in combination may be assessed or determined empirically, for example using ex vivo, in vivo and / or in vitro models; in some embodiments, such assessment or empirical determination is made in vivo, in a patient population (e.g., so that a correlation is established), or alternatively in a particular subject of interest.

[0043] In some embodiments, the irinotecan and letrozole are administered according to an intermittent dosing regimen including at least two cycles. Where two or more agents are administered in combination, and each by such an intermittent, cycling, regimen, individual doses of different agents may be interdigitated with one another. In certain embodiments, one or more doses of the second agent is administered a period of time after a dose of the first agent. In some embodiments, each dose of the second agent is administered a period of time after a dose of the first agent. In certain embodiments, each dose of the first agent is followed after a period of time by a dose of the second agent. In some embodiments, two or more doses of the first agent are administered between at least one pair of doses of the second agent; in certain aspects, two or more doses of the second agent are administered between al least one pair of doses of the first agent. In some instances, different doses of the same agent are separated by a common interval of time; in some embodiments, the interval of time between different doses of the same agent varies. In certain embodiments, different doses of the different agents are separated from one another by a common interval of time; in some embodiments, different doses of the different agents are separated from one another by different intervals of time.

[0044] One exemplary protocol for interdigitating two intermittent, cycled dosing regimens (e.g, for potentiating the effect of the letrozole, the irinotecan, or both), may include: (a) a first dosing period during which a therapeutically effective amount a first agent is administered to a subject; (b) a first resting period; (c) a second dosing period during which a therapeutically effective amount of a second agent and, optionally, a third agent, is administered to the subject; and (d) a second resting period.

[0045] In some embodiments, the first resting period and second resting period may correspond to an identical number of hours or days. Alternatively, in some embodiments, the first resting period and second resting period are different, with either the first resting period being longer than the second one or, vice versa. In some embodiments, each of the resting periods corresponds to 120 hours, 96 hours, 72 hours, 48 hours, 24 hours, 12 hours, 6 hours, 30 hours, 1 hour, or less. In some embodiments, if the second resting period is longer than the first resting period, it can be defined as a number of days or weeks rather than hours (for instance 1 day, 3 days, 5 days, 1 week, 2, weeks, 4 weeks or more).

[0046] If the first resting period’s length is determined by existence or development of a particular biological or therapeutic event, then the second resting period’s length may be determined on the basis of different factors, separately or in combination. Exemplary such factors may include the stage of AD for which the agents are administered; identity and / or properties (e.g., pharmacokinetic properties) of the first agent, and / or one or more features of the subject’s response to therapy with the first agent. In some embodiments, length of one or both resting periods may be adjusted in light of pharmacokinetic properties (e.g., as assessed via plasma concentration levels) of one or the other (or both) of the administered agents. For example, a relevant resting period might be deemed to be completed when plasma concentration of the relevant agent is below about 1 pg / ml, 0.1 pg / ml, 0.01 pg / ml or 0.001 pg / ml, optionally upon evaluation or other consideration of one or more features of the subject’s response.

[0047] In certain embodiments, the number of cycles for which a particular agent is administered may be determined empirically. Also, in some embodiments, the precise regimen followed (e.g, number of doses, spacing of doses (e.g., relative to each other or to another event such as administration of another therapy), amount of doses, etc.) may be different for one or more cycles as compared with one or more other cycles.

[0048] As described above, the subject methods are for treating Alzheimer's disease (AD). By treatment is meant at least an amelioration of one or more symptoms associated with the AD of the subject, where amelioration is used in a broad sense to refer to at least a reduction in the magnitude of a parameter, e.g. symptom, associated with the AD being treated. As such, treatment also includes situations where the AD, or at least one or more symptoms associated therewith, are completely inhibited, e.g., prevented from happening, or stopped, e.g., terminated, such that the subject no longer suffers from the AD, or at least the symptoms that characterize the AD.

[0049] COMPOSITIONS

[0050] Aspects of the present disclosure include compositions. In some embodiments, provided are pharmaceutical compositions comprising irinotecan, letrozole and a pharmaceutically acceptable carrier.

[0051] The irinotecan and letrozole can be incorporated into a variety of formulations for therapeutic administration. More particularly, the irinotecan and letrozole can be formulated into pharmaceutical compositions by combination with appropriate, pharmaceutically acceptable excipients or diluents, and may be formulated into preparations in solid, semi-solid, liquid or gaseous forms, such as tablets, capsules, powders, granules, ointments, solutions, injections, inhalants and aerosols.

[0052] Formulations of the irinotecan and letrozole for administration to the subject (e.g., suitable for human administration) are generally sterile and may further be free of detectable pyrogens or other contaminants contraindicated for administration to a patient according to a selected route of administration.

[0053] In pharmaceutical dosage forms, the irinotecan and letrozole can be administered in the form of their pharmaceutically acceptable salts, or they may also be used alone or in appropriate association, as well as in combination, with other pharmaceutically active compounds. The following methods and carriers / excipients are merely examples and are in no way limiting.

[0054] For oral preparations, the irinotecan and letrozole can be used alone or in combination with appropriate additives to make tablets, powders, granules or capsules, for example, with conventional additives, such as lactose, mannitol, corn starch or potato starch; with binders, such as crystalline cellulose, cellulose derivatives, acacia, corn starch or gelatins; with disintegrators, such as corn starch, potato starch or sodium carboxymethylcellulose; with lubricants, such as talc or magnesium stearate; and if desired, with diluents, buffering agents, moistening agents, preservatives and flavoring agents.

[0055] The irinotecan and letrozole can be formulated for parenteral (e.g., intravenous, intraarterial, intraosseous, intramuscular, intracerebral, intracerebroventricular, intrathecal, subcutaneous, etc.) administration. In certain embodiments, the irinotecan and letrozole are formulated for injection by dissolving, suspending or emulsifying the irinotecan and letrozole in an aqueous or non-aqueous solvent, such as vegetable or other similar oils, synthetic aliphatic acid glycerides, esters of higher aliphatic acids or propylene glycol; and if desired, with conventional additives such as solubilizers, isotonic agents, suspending agents, emulsifying agents, stabilizers and preservatives.

[0056] Pharmaceutical compositions that include the irinotecan and letrozole may be prepared by mixing the irinotecan and letrozole having the desired degree of purity with optional physiologically acceptable carriers, excipients, stabilizers, surfactants, buffers and / or tonicity agents. Acceptable carriers, excipients and / or stabilizers are nontoxic to recipients at the dosages and concentrations employed, and include buffers such as phosphate, citrate, and other organic acids; antioxidants including ascorbic acid, glutathione, cysteine, methionine and citric acid; preservatives (such as ethanol, benzyl alcohol, phenol, m-cresol, p-chlor-m-cresol, methyl or propyl parabens, benzalkonium chloride, or combinations thereof); amino acids such as arginine, glycine, ornithine, lysine, histidine, glutamic acid, aspartic acid, isoleucine, leucine, alanine, phenylalanine, tyrosine, tryptophan, methionine, serine, proline and combinations thereof; monosaccharides, disaccharides and other carbohydrates; low molecular weight (less than about 10 residues) polypeptides; proteins, such as gelatin or serum albumin; chelating agents such as EDTA; sugars such as trehalose, sucrose, lactose, glucose, mannose, maltose, galactose, fructose, sorbose, raffinose, glucosamine, N-methylglucosamine, galactosamine, and neuraminic acid; and / or non-ionic surfactants such as Tween, Brij Pluronics, Triton-X, or polyethylene glycol (PEG).

[0057] The pharmaceutical composition may be in a liquid form, a lyophilized form or a liquid form reconstituted from a lyophilized form, wherein the lyophilized preparation is to be reconstituted with a sterile solution prior to administration. The standard procedure for reconstituting a lyophilized composition is to add back a volume of pure water (typically equivalent to the volume removed during lyophilization); however solutions comprising antibacterial agents may be used for the production of pharmaceutical compositions for parenteral administration.

[0058] An aqueous formulation of the irinotecan and letrozole may be prepared in a pH-buffered solution, e.g., at pH ranging from about 4.0 to about 7.0, or from about 5.0 to about 6.0, or alternatively about 5.5. Examples of buffers that are suitable for a pH within this range include phosphate-, histidine-, citrate-, succinate-, acetate-buffers and other organic acid buffers. The buffer concentration can be from about 1 mM to about 100 mM, or from about 5 mM to about 50 mM, depending, e.g., on the buffer and the desired tonicity of the formulation.

[0059] A tonicity agent may be included to modulate the tonicity of the formulation. Example tonicity agents include sodium chloride, potassium chloride, glycerin and any component from the group of amino acids, sugars as well as combinations thereof. In some embodiments, the aqueous formulation is isotonic, although hypertonic or hypotonic solutions may be suitable. The term "isotonic" denotes a solution having the same tonicity as some other solution with which it is compared, such as physiological salt solution or serum. Tonicity agents may be used in an amount of about 5 mM to about 350 mM, e.g., in an amount of 100 mM to 350 mM. A surfactant may also be added to the formulation to reduce aggregation and / or minimize the formation of particulates in the formulation and / or reduce adsorption. Example surfactants include polyoxyethylensorbitan fatty acid esters (Tween), polyoxyethylene alkyl ethers (Brij), alkylphenylpolyoxyethylene ethers (Triton-X), polyoxyethylene-polyoxypropylene copolymer (Poloxamer, Pluronic), and sodium dodecyl sulfate (SDS). Examples of suitable polyoxyethylenesorbitan-fatty acid esters are polysorbate 20, (sold under the trademark Tween 20™) and polysorbate 80 (sold under the trademark Tween 80™). Examples of suitable polyethylene-polypropylene copolymers are those sold under the names Pluronic® F68 or Poloxamer 188™. Examples of suitable Polyoxyethylene alkyl ethers are those sold under the trademark Brij™. Example concentrations of surfactant may range from about 0.001% to about 1% w / v.

[0060] A lyoprotectant may also be added in order to protect the irinotecan and letrozole against destabilizing conditions during a lyophilization process. For example, known lyoprotectants include sugars (including glucose and sucrose); polyols (including mannitol, sorbitol and glycerol); and amino acids (including alanine, glycine and glutamic acid). Lyoprotectants can be included in an amount of about 10 mM to 500 nM.

[0061] In some embodiments, the pharmaceutical composition includes the irinotecan and letrozole, and one or more of the above-identified components (e.g., a surfactant, a buffer, a stabilizer, a tonicity agent) and is essentially free of one or more preservatives, such as ethanol, benzyl alcohol, phenol, m-cresol, p-chlor-m-cresol, methyl or propyl parabens, benzalkonium chloride, and combinations thereof. In other embodiments, a preservative is included in the formulation, e.g., at concentrations ranging from about 0.001 to about 2% (w / v).

[0062] KITS

[0063] Aspects of the present disclosure further include kits. The kits find use, e.g., in practicing the methods of the present disclosure. In some embodiments, a kit of the present disclosure comprises irinotecan and letrozole suitable for administration to a subject, e.g., a human subject having AD. The irinotecan and letrozole may be provided as separate pharmaceutical compositions. Alternatively, the irinotecan and letrozole may be provided in a single (i.e., the same) pharmaceutical composition. Whether provided as separate pharmaceutical compositions or in the same pharmaceutical composition(s), the compositions may be according to any of those described hereinabove in the Compositions section of the present disclosure.

[0064] The kits of the present disclosure may include a quantity of the compositions, present in unit dosages, e.g., ampoules, or a multi-dosage format. As such, in certain embodiments, the kits may include one or more (e.g., two or more) unit dosages (e.g., ampoules) of a composition that includes irinotecan, letrozole, or both. The term “unit dosage”, as used herein, refers to physically discrete units suitable as unitary dosages for human and animal subjects, each unit containing a predetermined quantity of the composition calculated in an amount sufficient to produce the desired effect. The amount of the unit dosage depends on various factors, such as the particular agent employed, the effect to be achieved, and the pharmacodynamics associated with the irinotecan and letrozole, in the subject. In yet other embodiments, the kits may include a single multi dosage amount of the composition(s).

[0065] Components of the kits may be present in separate containers, or multiple components may be present in a single container.

[0066] In some embodiments, a kit of the present disclosure may comprise instructions for administering the irinotecan and letrozole to a subject having Alzheimer’s Disease (AD). The instructions (e.g., instructions for use (IFU)) included in the kits may be recorded on a suitable recording medium. For example, the instructions may be printed on a substrate, such as paper or plastic, etc. As such, the instructions may be present in the kits as a package insert, in the labeling of the container of the kit or components thereof (i.e., associated with the packaging or sub-packaging) etc. In other embodiments, the instructions are present as an electronic storage data file present on a suitable computer readable storage medium, e.g., portable flash drive, DVD, CD-ROM, diskette, etc. In yet other embodiments, the actual instructions are not present in the kit, but means for obtaining the instructions from a remote source, e.g. via the internet, are provided. An example of this embodiment is a kit that includes a web address where the instructions can be viewed and / or from which the instructions can be downloaded. As with the instructions, the means for obtaining the instructions is recorded on a suitable substrate.

[0067] For purposes of completeness, non-limiting aspects and embodiments of the present disclosure are further disclosed in the following numbered clauses.

[0068] 1 . A method of treating Alzheimer's disease (AD), the method comprising: administering to a subject having Alzheimer's disease a therapeutically effective amount of irinotecan and letrozole.

[0069] 2. The method of claim 1 , wherein the irinotecan and letrozole are administered sequentially.

[0070] 3. The method of claim 2, wherein the irinotecan and letrozole are administered to the subject in separate compositions.

[0071] 4. The method of claim 1 , wherein the irinotecan and letrozole are administered to the subject in a single composition.

[0072] 5. The method of any one of claims 1 -4, wherein the irinotecan and letrozole are administered orally or parenterally to the subject.

[0073] 6. A pharmaceutical composition comprising: irinotecan; letrozole; and a pharmaceutically acceptable carrier.

[0074] 7. The pharmaceutical composition of claim 6, wherein the composition is formulated for oral administration to a subject in need thereof.

[0075] 8. The pharmaceutical composition of claim 6, wherein the composition is formulated for parenteral administration to a subject in need thereof.

[0076] 9. A kit comprising: irinotecan; letrozole; and instructions for administering the irinotecan and letrozole to a subject having Alzheimer’s Disease (AD).

[0077] 10. The kit of claim 9, wherein the irinotecan and letrozole are provided as separate pharmaceutical compositions.

[0078] 11 . The kit of claim 9, wherein the irinotecan and letrozole are provided in a single pharmaceutical composition.

[0079] 12. The kit of claim 10 or 1 1 , comprising two or more unit dosages of the pharmaceutical composition(s).

[0080] The following examples are offered by way of illustration and not by way of limitation.

[0081] EXPERIMENTAL

[0082] Example 1 - Characterization of combination therapy prediction for Alzheimer’s Disease from Single Cell Data

[0083] Study summary

[0084] Alzheimer's disease (AD) is a multifactorial condition1characterized by heterogeneous molecular alternations across various brain cell types23, which poses significant challenges in the development of effective treatments45. Recent advancements in mining multi-domain, large-scale databases include powerful and innovative approaches to identifying promising drug candidates6. By integrating diverse datasets from single-cell human transcriptomics, drug perturbations, and clinical records, predicted were combination therapeutic candidates aimed at rectifying gene expression changes in multiple cell types implicated in this disease. By testing the predicted candidates in an AD mouse model, demonstrated herein is that a combination of drugs targeting both neurons and glial cells significantly rescued memory deficit and hippocampal atrophy compared to vehicle, as well as to single-drug treatments focusing on one cell type. The successful validation of the human data-driven, cell-specific drug discovery approach not only demonstrates its potential to uncover more potent therapies for complex diseases but also propels the field of precision medicine forward, offering significant promise not only for AD but also other conditions traditionally challenging to treat.

[0085] Methods

[0086] Cell-type-specific transcriptomic signatures of AD patients were established in a metaanalysis of publicly available single-nucleus-RNA-seq (snRNA) datasets. Three independent datasets7 9, generated from the prefrontal cortices of AD patients (n=41 ) and controls (n=40), were computationally integrated by Canonical Correlation Analysis (CCA) in Seurat. Case and control classification across studies were standardized using histological measurements of tau tangle severity and amyloid beta (Ap) burdens1. Differential gene expression analysis of AD versus controls was performed and stratified by cell types using FindMarker function in Seurat10(logFC>0.1 & p-adj<0.05). Cell-type-specific transcriptomic profiles of AD patients, each consisting of about 200 differentially expressed genes (DEGs), were compared to drug expression profiles from the connectivity map (CMap)1database to identify drugs that reverse the gene expression signatures of AD. The beneficial effects of the top drug hits were examined by analyzing AD prevalence (% AD diagnosis) in drug-exposed individuals and propensity- matched controls, via UC-wide Electronic Medical Records (EMR). To validate combination drug efficacy, a sex-balanced double transgenic cohort crossing 5xFAD12and PS1913mice were evenly divided into four groups and treated with vehicle, the two identified drugs alone, and both drugs, respectively, for three months starting at four months of age. Morris water maze was performed at seven months of age to assess spatial learning and memory14. The mice were dosed every other day via intraperitoneal injections until they were sacrificed.

[0087] Results

[0088] Our integrated dataset, consisting of a total of 47,119 genes in 261 ,309 cells, has no technical variation associated with individual studies or case and control, while cell-type-specific biological variation is preserved.

[0089] Differential gene expression analysis of AD versus controls was performed and stratified by cell types. The cell-type-specific transcriptomic signatures were overlaid with the drug expression profiles provided by the cMAP, encompassing 1 ,300 either FDA-approved or previously investigated drugs, and identified 25 drug candidates that could reverse AD transcriptomic signatures in a cell-type-specific manner. Further analysis of UC-wide EMR, covering 1 ,441 ,778 individuals from six University of California health centers15, reduced this list to two drugs. One reverses AD signatures in glial cells and the other in neurons. Both drugs displayed a remarkable association with a significantly reduced risk of AD. Irinotecan, the drug candidate that targets glial cells, has a relative risk score of 0.195 compared to propensity- matched controls. Irinotecan is a DNA topoisomerase I inhibitor currently used as a treatment for colorectal cancer16and no prior study has demonstrated its potential neuroprotective effect on AD. It was hypothesized that this drug might exert an anti-inflammatory effect on glial cells. The other compound, letrozole, which targets neurons and has a relative risk score of 0.466, is an aromatase inhibitor used in the treatment of breast cancer. Although no obvious mechanistic links were previously established between the drug and neurodegeneration, an epidemiological study17indicated a reduced risk of dementia in patients taking this drug compared to other aromatase inhibitors.

[0090] To functionally validate cognitive and pathological improvement by the predicted drug combination, dosing experiments were performed in an AD mouse model carrying overexpression of mutant human tau and Ap (MAPT P301 S13x 5xFAD12). A sex-balanced double transgenic cohort was evenly divided into four groups and treated with either vehicle, individual predicted drugs alone, or both drugs, for three months. Morris Water Maze14experiments were performed following the treatments to assess spatial learning and memory. The behavioral test revealed that only the combination-treated group showed a statistically significant preference for the target quadrant at both 24 and 72 hours after the learning trials, indicating rescue of both short-term and long-term memory deficits. The combination-treated mice demonstrated better performance of location recall also by more frequent crossing of the target quadrant where the platform used to be and shorter latency of first platform crossing.

[0091] Microscopic pathology analysis also revealed that the drug-treated groups had significantly reduced brain atrophy compared to controls, with the most significant rescue observed in the combination-treated mice. Immunolabeling of pathological tau using AT8 antibody18demonstrated significantly reduced positive AT8 percent area in the combination- treated group.

[0092] Conclusions

[0093] Therapeutic candidates for AD were identified based on its cell-type-specific transcriptomic signatures, several of which are further supported by analysis of EMR data showing lower AD prevalence after exposure to the drugs. The present validation in an AD mouse model demonstrated that the combination treatment significantly rescued spatial memory, brain atrophy, and tau burden and outperformed the individual drug treatments, suggesting correcting expression profiles in both neuronal and glial cells is critical for rescuing AD-specific deficits.

[0094] Our unique approach, emphasizing cell-type precision, predicts multi-target drug combinations entirely using human-derived data. This approach leverages extensive omic data obtained from post-mortem human brains and clinical records representing millions of diverse patients across multiple health centers. By prioritizing human-based cellular-precision drug screening, treatments that significantly enhance both efficacy and patient outcomes can be discovered. Materials and Methods

[0095] Human single nuclei RNA-sequencing data curation

[0096] To ensure a diverse representation of AD patients, publicly available human sn-RNA sequencing datasets were curated from three independent sources. Both Mathys et al. and Zhou et al. studies were obtained from the Accelerating Medicines Partnership Alzheimer’s Disease Project (AMP-AD) Knowledge Portal (adknowledgeportal.org) under the Religious Order Study and Memory and Aging Project (ROSMAP). The Mathys et al dataset is accessible through doi.org / 10.7303 / syn2580853. The Zhou et al dataset is available under study snRNAseqAD_TREM2 and are also accessible through doi.org / 10.7303 / syn21125841 . Only individuals without TREM2 mutations were included in the integrated dataset. The third dataset by Lau et al. was obtained from Gene Expression Omnibus (GEO) under the accession number GSE157827.

[0097] Case-control standardization across datasets

[0098] To standardize AD identification across studies, samples were re-classified into AD or control groups based on tau tangles severity with BRAAK1clinical staging scores and Ap burden using Consortium to Establish a Registry for Alzheimer’s Disease (CERAD)2scores as proxy. BRAAK clinical staging ranges from I to VI to represent low to high level of tau deposition. The CERAD scoring ranges from 1 to 4 to indicate high to low Ap burden severity. Based on the available metadata from the original studies, AD case was defined as individuals with severe tau deposition (BRAAK > IV) and high Ap load (CERAD < 2), and non-AD controls as individuals with low tau (Braak < III) and low Ap load (CERAD > 3). Individuals with scores that do not fulfill either criterion were excluded. The final integrated dataset consisted of 41 AD cases and 40 controls.

[0099] Sn-RNA-seq dataset integration, normalization, and batch correction

[0100] Before merging the datasets, cells deemed as poor quality were removed with the following criteria: total feature count less than 500, total features less than 250, or mitochondria gene ratio higher than 10%. Sparse features, expressed in fewer than 100 cells, were also removed.

[0101] To obtain conserved disease markers across datasets and correct for batch effects, canonical correlation analysis (CCA)3was performed using the Seurat package4to harmonize the merged dataset. The default settings were applied to first log-normalize the merged dataset with Seurat’s NormalizeData function. Then, for each dataset independently, the top 2000 variable features were identified using the FindVariableFeatures function with "vst" as the selection method. Integration anchors were identified with the FindlntegrationAnchors function based on the top variable features. Finally, a harmonized dataset consisting of 168,521 cells and 29,120 features was generated using the IntegrateData function in Seurat. The data matrix was linearly transformed using Seurat’s ScaleData function. Dimensional reduction through Uniform Manifold Approximation and Projection (UMAP) was performed with the RunUMAP function, which considers the top 30 dimensions selected from the corresponding principal component analysis (PCA) obtained by running the unPCA function in Seurat.

[0102] Clustering was determined based on the first 30 PCs using the FindNeighbors function in Seurat, which embeds cells in a K-nearest neighbor graph based on Euclidean distance in PCA space and refines the edge weights between any two cells based on the shared overlap in their local neighborhoods. Clustering was implemented using a resolution of 0.5 in the FindClusters function, which applies modularity optimization techniques such as the Louvain algorithm, resulting in a set of 21 distinct clusters.

[0103] Cell type annotation

[0104] Cell type identities were determined by applying Seurat’s AddModuleScore function to lists of human brain marker genes ( ~ 8 per cell type) collected from PanglaoDB and referenced by Mckenzi et al5. Cell type assignment included astrocyte, microglia, oligodendrocytes, oligodendrocyte precursor cells, endothelial cells, pericytes, excitatory and inhibitory neurons. Each cell was assigned the corresponding cell type identity that generated the highest scores among scores for all cell types. If the highest and second highest scores of a cell were within 20% of the highest score, then the cell were deemed hybrids and excluded from further analysis. The validity of the assigned cell type identities was assessed by examining the homogeneity, distribution, and separation of cell types by clustering in UMAP plots.

[0105] Cell-type-specific differential expression analysis

[0106] For each cell type, differential gene expression analysis comparing AD samples to controls was performed using the FindMarkers function in the Seurat package4. The test. use parameter was set to MAST, which uses a two-part generalized linear model that models gene expression rate using linear regression and expression level using Gaussian distribution, as recommended by Mou et al.6in a study comparing 9 DE methods for single-cell RNA sequencing analysis. Significant differentially expressed genes(DEGs) were determined as those with adjusted p value < 0.05 based on Bonferroni correction using all genes in the dataset and a Iog2 fold change (LFC) greater than 0.1 .

[0107] Pathway analysis

[0108] Pathway enrichment analysis was conducted using gProfiler7, a web tool that performs functional over-representation analysis by mapping a given gene list to known functional information sources and detects statistically significantly enriched terms. As input gene lists, each cell-type-specific DEG list was split into upregulated and downregulated subsets and queried them independently with adjusted p-value cutoff of 0.05 to obtain significant pathway enrichment in both directions. Multiple testing correction was performed using g:SCS algorithm7, which specifically correct for p-values obtained from GO and pathway enrichment analysis. This method is designed to correct for multiple tests that may be potentially dependent of each other due to term associations. In addition to considering Gene Ontology cellular components, biological processes, and molecular functions, also included were the provided pathways from the Human Protein Atlas, Human Phenotype Ontology, KEGG, Reactome, and Wiki pathways in the subsequent network analysis.

[0109] Network visualization of functional enrichment analysis

[0110] Network analysis was conducted on pathway results derived from the cell-type-specific DEG lists, following a well-established protocol, to better visualize and interpret the enriched functional pathways. In summary, pathways results were imported into Cytoscape8, an open- source software, using the visualization application EnrichmentMap. Subsequently, redundant and related pathways were collapsed into clusters, which were then annotated concisely with the associated functional biological themes using the Cytoscape application AutoAnnotate.

[0111] Computational drug screening pipeline

[0112] Previously published drug repositioning pipeline9was applied to each cell-type-specific disease gene expression profile to identify potential therapeutic candidates that reverse transcriptomic effects in brain cell types pertinent to AD. At a high level, the algorithm takes two inputs. The first is an ordered list of DEGs specific to cell types, arranged based on degrees of log fold change, from the most upregulated to most down regulated genes in the disease. The second is drug gene expression profiles sourced from Connectivity Map (CMap)10. CMap, generated by the Broad Institute, contains 6100 gene expression profiles generated of 1300 compounds treated in five different human cancer cell lines. A reversal score is calculated for each drug and disease pair by applying a Kolmogorov-Smirnov test (K-S test) that compares the gene expression ranks in the disease relative to drug-induced expression change profiles. A strong negative CMap score (< -0.25) suggest that the drug has an opposing mechanism of action to the disease, with potential therapeutic effect on the disease. Permutation analysis was carried out to assess significance, drug hits with FDR < 0.05 were considered for further examination and in silico human data validation.

[0113] Drug selections

[0114] While the drug screening approach generated around 100 drug candidates targeting six major brain cell types, prioritized were approximately 25 drugs that significantly reverse more than one brain cell type. This aims to maximize the coverage of multiple cell types by the combination of two drugs.

[0115] Human validation using Electronic Medical Records (EMR) Data

[0116] The beneficial effects of the top drug hits were examined by analyzing AD prevalence (% AD diagnosis) in drug-exposed individuals and compared to propensity-matched controls, via UC-wide Electronic Medical Records (EMR). At the time of surveying, the UC-wide EMR aggregated clinical data covering 1 ,441 ,778 individuals across six University of California (UC) campuses, encompassing more than 10 million patient records including diagnosis and medication prescriptions. For each screened drug candidate, patients prescribed or taken the drug will be identified using the medication order table and matched controls will be identified using a propensity score matching approach based on age at diagnosis, age at death (if applicable), race, sex, original indications, AD comorbidity diagnosis, UC center location. AD was measured relative risk score in the drug-exposed group and compared to the matched control groups by calculating the ratios of AD to total patients in each group. Bootstrapped x-squared tests across 10 permutations of the iterative matching of the control group was applied to establish significance.

[0117] References for Materials and Methods

[0118] 1. Braak, H., Alafuzoff, I., Arzberger, T., Kretzschmar, H. & Del Tredici, K. Staging of Alzheimer disease-associated neurofibrillary pathology using paraffin sections and immunocytochemistry. Acta Neuropathol 112, 389-404 (2006).

[0119] 2. Rossetti, H. C., Munro Cullum, C., Hynan, L. S. & Lacritz, L. The CERAD Neuropsychological Battery Total Score and the Progression of Alzheimer’s Disease. Alzheimer Dis Assoc Disord 24, 138-142 (2010).

[0120] 3. Integrative analysis in Seurat v5. satijalab.org / seurat / articles / seurat5Jntegration.

[0121] 4. Tools for Single Cell Genomics, satijalab.org / seurat / .

[0122] 5. McKenzie, A. T. etal. Brain Cell Type Specific Gene Expression and Co-expression Network Architectures. Scientific Reports 8, 1-19 (2018).

[0123] 6. Mou, T., Deng, W., Gu, F., Pawitan, Y. & Vu, T. N. Reproducibility of Methods to Detect Differentially Expressed Genes from Single-Cell RNA Sequencing. Front. Genet. 10, (2020).

[0124] 7. Raudvere, U. et al. g:Profiler: a web server for functional enrichment analysis and conversions of gene lists (2019 update). Nucleic Acids Research 47, W191-W198 (2019).

[0125] 8. Otasek, D., Morris, J. H., Bougas, J., Pico, A. R. & Demchak, B. Cytoscape Automation: empowering workflow-based network analysis. Genome Biol 20, 185 (2019).

[0126] 9. Sirota, M. et al. Discovery and preclinical validation of drug indications using compendia of public gene expression data. Sci Trans! Med 3, 96ra77 (2011 ).

[0127] 10. Lamb, J. et al. The Connectivity Map: Using Gene-Expression Signatures to Connect Small Molecules, Genes, and Disease. Science 313, 1929-1935 (2006).

[0128] References for Example 1

[0129] 1. Bloom, G. S. Amyloid-p and tau: the trigger and bullet in Alzheimer disease pathogenesis. JAMA Neurol 71 , 505-508 (2014). 2. Hansen, D. V., Hanson, J. E. & Sheng, M. Microglia in Alzheimer’s disease. J Cell Biol 217, 459-472 (2018).

[0130] 3. Gonzalez-Reyes, R. E., Nava-Mesa, M. O., Vargas-Sanchez, K., Ariza-Salamanca, D. & Mora-Munoz, L. Involvement of Astrocytes in Alzheimer’s Disease from a Neuroinflammatory and Oxidative Stress Perspective. Front Mol Neurosci 10, (2017).

[0131] 4. Frozza, R. L., Lourenco, M. V. & De Felice, F. G. Challenges for Alzheimer’s Disease Therapy: Insights from Novel Mechanisms Beyond Memory Defects. Front Neurosci 12, (2018).

[0132] 5. Cummings, J. L, Morstorf, T. & Zhong, K. Alzheimer's disease drug-development pipeline: few candidates, frequent failures. Alzheimers Res TherG, 37 (2014).

[0133] 6. Guo, S. etal. Editorial: Computational and systematic analysis of multi-omics data for drug discovery and development. Frontiers in Medicine 10, (2023).

[0134] 7. Mathys, H. et al. Single-cell transcriptom ic analysis of Alzheimer’s disease. Nature 570, 332-337 (2019).

[0135] 8. Zhou, Y. et al. Human and mouse single-nucleus transcriptomics reveal TREM2-dependent and TREM2-independent cellular responses in Alzheimer’s disease. Nat Med2G, 131-142 (2020).

[0136] 9. Lau, S.-F., Cao, H., Fu, A. K. Y. & Ip, N. Y. Single-nucleus transcriptome analysis reveals dysregulation of angiogenic endothelial cells and neuroprotective glia in Alzheimer’s disease. PM4S 117, 25800-25809 (2020).

[0137] 10. Tools for Single Cell Genomics, satijalab.org / seurat / .

[0138] 11 . Lamb, J. et al. The Connectivity Map: Using Gene-Expression Signatures to Connect Small Molecules, Genes, and Disease. Science 313, 1929-1935 (2006).

[0139] 12. 5xFAD (B6SJL) | ALZFORUM. www.alzforum.org / research-models / 5xfad-b6sjl.

[0140] 13. Tau P301 S (Line PS19) | ALZFORUM. www.alzforum.org / research-models / tau-p301s-line- ps19.

[0141] 14. Morris water maze: procedures for assessing spatial and related forms of learning and memory | Nature Protocols, www.nature.com / articles / nprot.2006.116.

[0142] 15. UCSF Clinical Data. UCSF Data Resources data.ucsf.edu / research / ucsf-data.

[0143] 16. Reyhanoglu, G. & Smith, T. Irinotecan, in StatPearls (StatPearls Publishing, 2023).

[0144] 17. Branigan, G. L., Soto, M., Neumayer, L., Rodgers, K. & Brinton, R. D. Association Between Hormone-Modulating Breast Cancer Therapies and Incidence of Neurodegenerative Outcomes for Women With Breast Cancer. JAMA Network Open 3, e201541 (2020).

[0145] 18. Goedert, M., Jakes, R. & Vanmechelen, E. Monoclonal antibody AT8 recognises tau protein phosphorylated at both serine 202 and threonine 205. Neurosci Letf 89, 167-169 (1995). Example 2 - Computational integration of large-scale single-nucleus transcriptomic datasets from multi-studies identifies cell-type-specific signatures of AD

[0146] A comprehensive single nucleus RNA-seguencing (snRNA-seg) dataset was organized by combining published data from three independent studies22 24(Fig. 1 a), covering 37 AD patients (15 females and 22 males) and 29 controls (13 females and 22 males). Samples from individuals who did not meet both CERAD28and Braak29criteria for AD or control were excluded from the dataset. The Uniform Manifold Approximation and Projection (UMAP) visualization reveals clustering predominantly by the datasets of origin rather than biological variations, such as cell types or disease status (Fig. 1 b). After batch harmonization with the integration algorithm, technical artifacts between datasets were effectively removed (Fig. 1 c), and distinct cell populations clustered by cell types (Fig. 1 d) while no discrete separation was observed based on disease status (Fig. 1 e). The integrated dataset consisted of expressions of 29,120 features in 137,065 cells.

[0147] To systematically characterize cell-type-specific AD pathophysiological features, comprehensive analyses were conducted focusing on six disease-relevant cell types: excitatory neurons (ex.neu), inhibitory neurons (in.neu) microglia (mic), astrocytes (ast), oligodendrocytes (oli), and oligodendrocyte precursor cells (OPC). In AD patients, the proportion of inhibitory neurons significantly decreased compared to controls (Fig. 1 f), as previously reported in human and mouse models of AD30 32. Cell-cell-communication analysis revealed heterogeneous signaling patterns among neuronal subpopulations in AD compared to controls. Signaling to inhibitory neurons increased from all cell types except microglia, while signaling to excitatory neurons decreased from all cell types except oligodendrocytes (Fig. 1 g).

[0148] Cell-type-specific differential gene expression analysis between AD and control groups revealed significant variations in differentially expressed genes (DEGs) across cell types (Fig. 1 h). Many DEGs were unique to specific cell types, while others were shared but displayed opposite regulatory patterns in AD (Fig. 1 i). For example, the APOE gene, a major genetic risk factor for AD33, was significantly upregulated in microglia but downregulated in astrocytes and OPCs in AD samples. Additionally, gene set enrichment analysis showed that AD-enriched Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways and gene ontology (GO) terms exhibited extensive cell-type heterogeneity, with most pathways and terms unique to individual cell types (Fig. 1 j). While pathways like chemical carcinogenesis - reactive oxygen species and prion disease were enriched in AD across all six cell types, unique pathways delineated them into two clusters. The neuronal-centric cluster included excitatory and inhibitory neurons, with oligodendrocytes sharing estrogen signaling with excitatory neurons and cAMP signaling with inhibitory neurons. The glial-centric cluster was comprised of astrocytes, microglia, and OPCs, with astrocytes and OPCs sharing gap junction pathway enrichment, and microglia and OPCs enriched for oxidative phosphorylation. These findings suggest AD pathogenesis involves heterogeneous transcriptomic-driven molecular alterations manifested in discordant behaviors of multiple cell types. Effective treatment likely needs to correct malfunctions in multiple, if not all, cell types.

[0149] Example 3 - Computational drug repurposing pipeline predicts cell-type-specific therapeutic candidates

[0150] After establishing cell-type-specific AD profiles from the integrated human dataset, screened for were network-correcting drug candidates targeting AD-specific transcriptom ic changes across multiple cell types. Each cell-type-specific AD transcriptomic profile was queried against the Connectivity Map (CMap) drug expression database25using a computational pipeline that matches gene expression profiles of diseases and existing drugs (Fig. 2a). Most AD signatures overlapped with drug profile features, which were then used as inputs in the patternmatching algorithm (Fig. 2b). With a false discovery rate (FDR) < 0.05, 35 hits were predicted for excitatory neurons, 12 hits for inhibitory neurons, 8 hits for microglia, 33 hits for astrocytes, 4 hits for oligodendrocytes, and 8 hits for OPCs. Several drug hits overlapped across cell types, forming linkages in a network visualization of drug interactions with cell types (Fig. 2c).

[0151] Notably, 25 repurposed drugs significantly reversed cell-type-specific AD profiles in multiple cell types (Fig. 2d), indicating a multi-targeted potential with these drugs. These multi- cell-type drug candidates span various therapeutic classes, including cardiac glycosides (digitoxigenin, digoxigenin, helveticoside), chemotherapeutic agents (methotrexate, irinotecan, etoposide, mitoxantrone), aromatase inhibitors (letrozole), histone deacetylase inhibitors (trichostatinA, vorinostat, scriptaid), immunosuppressants / mTOR inhibitors (sirolimus), anti- inflammatory / anti-tumor agents (15-delta prostaglandin J2), nonsteroidal anti-inflammatory drugs (fenbufen), antifungal agents (ciclopirox), antibiotic (monensin, ionomycin), antiepileptics (valproic acid), antipsychotics (haloperidol), and several experimental compounds with undefined pharmacological classifications (LY-294002, CP-320650-01 , CP-690334-01 , syrosingopine, hycanthone, 5155877)34.

[0152] Example 4 - Validation of candidate drug effects in humans using real-world evidence derived from electronic medical records

[0153] Sought next was to validate the effects of the repurposed drug candidates in humans using real-world data. An advantage of existing pharmaceutical agents is their potential for population-based analyses using real-world patient data to investigate associations between drug use and AD outcomes. The University of California (UC)-wide Electronic Medical Records (EMR) database was explored. This database includes clinical records of more than 10 million individuals across six university health centers in California. 25,257 individuals diagnosed with AD were identified within the UC-wide EMR database, out of 1.4 million individuals aged 65 or older, creating a substantial clinical dataset for analysis. Focusing on the 25 multi-cell-type candidate drugs, usage records were found for 10 of these drugs, as only a subset of the drugs from CMap is FDA-approved or prescribed (Fig 2e). Of these ten drugs, five (letrozole, irinotecan, methotrexate, ciclopirox, and sirolimus) were associated with a significantly reduced risk of AD compared to matched controls, suggesting potential protective effects of them against AD. Although three drugs (etoposide, trifluoperazine, and vorinostat) showed reduced risk, statistical significance was not achieved due to insufficient patient representations. Lastly, two drugs (valproic acid and haloperidol), both used for neurological conditions, showed higher relative risk scores.

[0154] Example 5 - Prioritization of letrozole and irinotecan as combination therapy for AD based on human-derived evidence

[0155] It was hypothesized that combination therapy targeting both neuronal and glial transcriptomic profiles might more effectively alleviate AD pathologies. Among five drug candidates showing significant AD risk reduction in UC-wide EMR, letrozole was prioritized for its predicted reversal effects in excitatory and inhibitory neurons, and irinotecan for its effects on the glial-centric cluster, including astrocytes, microglia, and OPCs (Fig. 2f). Therefore, a combination of letrozole and irinotecan potentially targets five cell types in AD.

[0156] When visualizing AD and drug profiles side by side, genes upregulated in AD neurons shifted downward, while downregulated genes shifted upward in letrozole-treated profiles. Irinotecan-treated profiles showed similar shifts across glial cells (Fig. 2g). Letrozole, an aromatase inhibitor primarily prescribed for breast cancer treatment and occasionally for male infertility2035, had a relative risk ratio of 0.466 for AD in UC-wide EMR. It demonstrated reversal scores of -0.37 and -0.61 for excitatory and inhibitory neurons, respectively. Despite a female- skewed cohort (33 females to 1 male), the risk reduction ratios were comparable for both sexes when analyzed individually. Irinotecan, a DNA topoisomerase I inhibitor used for colorectal cancer treatment21, showed a relative risk ratio of 0.195 for AD. It demonstrated AD reversal capabilities across all three cell types of the glial-centric clusters, with reversal scores of -0.29, - 0.31 , and -0.45 for astrocytes, microglia, and OPCs, respectively.

[0157] Example 6 - The combination treatment with letrozole and irinotecan rescues both short-term and long-term spatial memory in a mouse model of AD

[0158] To experimentally test the efficacy of the combination therapy with letrozole and irinotecan, dosing experiments were conducted in an AD mouse model expressing mutant human APP / PS1 and tau (5xFAD27x PS1926), which recapitulates many AD-related phenotypes including amyloid plaque formation, tau tangles, and gliosis, with an early and aggressive pathology onset3637(Fig. 3a). A sex-balanced double transgenic cohort was evenly divided into four groups (n=20, including both sexes) and each treated with either vehicle, letrozole, irinotecan, or the combination of both drugs every other day for three months. Spatial learning and memory performance were evaluated using the Morris Water Maze (MWM) test38. While no significant difference was observed in hidden platform training trials across 6 days between each treatment and vehicle group (Fig. 3b), the memory test in probe trials revealed that only the combination-treatment group exhibited a statistically significant preference for the target quadrant at both 24 and 72 hours post-training (Fig. 3c, d), suggesting a rescue of both short-term and long-term memory deficits by the combination treatment. Furthermore, only the combinationtreatment group demonstrated significantly better location recall by making more frequent crossings to the platform location at both time points after platform removal (Fig. 3e,f). Mice in the combination-treatment group also significantly outperformed the vehicle mice in latency to the first platform crossing during the 72-hour probe trial. Average swim speeds were similar across all groups, indicating that the observed behavioral differences were not confounded by visual or motor impairments.

[0159] Additionally, sex differences were observed in the dosing experiment, with significantly improved learning performance only observed in combination-treated males compared to vehicle- treated males, and not in females. Memory rescue by the single drug treatments was also evident in males at 24 and 72 hours of the probe trials but not in females. Taken together, AD-related behavior assessments validated the efficacy of the combination therapy with letrozole and irinotecan, especially in males, in an AD mouse model with both amyloid and tau pathologies, and single-drug treatments alone had much less efficiency.

[0160] Example 7 - The combination treatment with letrozole and irinotecan rescues AD pathologies in a mouse model of AD

[0161] To determine whether the single or combination drug treatments can also rescue AD- related pathologies, the cohort was morphologically assessed at 9 months of age (4 months posttreatment). According to previous literature, this AD mouse model develops extensive neurodegeneration, Ap deposits, hyperphosphorylated tau (p-tau), gliosis, and severe loss of CA1 neurons in the hippocampus at this age3637. Neurodegeneration in the hippocampal region was evaluated first, as atrophy in this region is a hallmark of AD progression. Analyses of hippocampal volume showed rescue of atrophy in all treatment groups, with the most significant improvement in the combination-treatment group compared to the control group (Fig. 4a, b). Thioflavin S (Thio-S) staining for p-amyloid pathology revealed significant reductions in all treatment groups compared to vehicle-treated controls, measured by the Thio-S-positive percent area and plaque counts normalized by hippocampal size (Fig. 4c, d). P-tau pathology was assessed by immunofluorescent staining using the p-tau-specific AT8 antibody. While all treatment groups showed a trend of reduced p-tau pathology compared to vehicle-treated controls, only the combination-treatment group had a statistically significant reduction in the p- tau coverage area of the hippocampus (Fig. 4e,f).

[0162] Given the cell-type-precision therapeutic design, investigated next were the effects of each treatment on gliosis and neuronal loss through immunostaining of specific cell types. As major contributors to neuroinflammation, microgliosis and astrogliosis were evaluated by quantifying the coverage area of cell type-specific markers Iba1 and GFAP, respectively. Mice in irinotecan and combination-treatment groups showed a significant reduction in the Iba1 -positive percent area in the hippocampus, indicating alleviation of microgliosis (Fig. 4g, h). Astrogliosis reduction was moderate, with significant reduction observed only in the irinotecan-treated group (Fig. 4i,j). Neuronal loss was assessed by measuring neuronal layer thickness via NeuN staining in the CA1 region of the hippocampus. Significant rescue of neuronal loss was evident in the CA1 region (Fig. 4k, I). The most significant rescues were observed in letrozole and combinationtreatment groups. Interestingly, irinotecan-treated mice also showed moderate rescue in neuronal loss, likely attributable to the alleviation of gliosis that may help prevent neuronal loss.

[0163] To evaluate the relationship between cognitive performance and underlying neurodegenerative changes, correlation analysis was conducted between behavioral test metrics and pathological measurements. This analysis aimed to better elucidate the efficacy of therapeutic interventions. Notably, the combination-treatment group exhibited consistently stronger correlations compared to the vehicle-treated group, particularly between the percent time spent in the target quadrant during memory probes and the thickness of NeuN-positive layers. These findings suggest that improvements in cognitive performance in the combinationtreatment group were more closely associated with reductions in pathologies, reinforcing the potential therapeutic synergy observed.

[0164] Example 8 - The combination treatment with letrozole and irinotecan promotes neuroprotective functional pathways in a cell-type-specific manner

[0165] To investigate the cell-type-specific transcriptomic changes in response to the combination treatment, snRNA-seq was performed on dissected hippocampi obtained from mice on combination treatment of letrozole and irinotecan (L+l) or vehicle treatment (n=8 per group, including both sexes). After standard processing and quality control, a filtered dataset containing 25,642 gene features expressed across 237,853 nuclei was obtained for further analysis. Through graph-based clustering and visualization using UMAP, 31 distinct cell clusters were identified, including clusters assigned to the six major cell types (Fig. 5a). Additionally, it was found that the L+l treatment reduced the excessive communication between inhibitory neurons and other cell types (Fig. 5b), evident in AD (see Fig. 1g).

[0166] To account for the granularity and selective vulnerabilities among neuronal subtypes, the excitatory and inhibitory neurons were subdivided into CA1 pyramidal cells, CA3 pyramidal cells, dentate granule cells (DGC), subiculum neurons, and interneurons based on expressions of hippocampal subregion marker genes. Notably, in the UMAP plot split by treatment groups, there is a discernible higher density of cells observed in the L+l-treated group compared to the vehicle- treated group within the pyramidal neuron clusters of the CA1 and CA3 regions (Fig. 5c). This was quantified via cell type abundance analysis, revealing that the proportions of CA1 and CA3 pyramidal neurons in L+l-treated mice were significantly higher than in vehicle-treated mice (Fig. 5d), consistent with the findings from pathological analysis.

[0167] Differential gene expression analysis comparing L+l and vehicle treatments revealed varying counts and compositions of DEGs (abs(logFC) > 0.1 , padj < 0.05) across cell types, with the fewest DEGs observed in oligodendrocytes, suggesting they were least affected (Fig. 5e). Gene set enrichment analysis of cell-type-specific DEGs revealed enrichments in diseaserelevant functional pathways (Fig. 5f). Notably, the treatment enriched pathways related to nervous system development and synaptic activities in neurons. For example, in excitatory neurons, the dendrite morphogenesis pathway was enriched, while the axonogenesis pathway was enriched in inhibitory neurons. In glial cells, oligodendrocyte differentiation pathway was enriched in OPCs, and actin filament-based movement and synapse assembly pathways in oligodendrocytes. Essential microglial functions like pathways of response to oxidative stress and histamine, regulation of neuron death and synaptic plasticity, and neuron projection development were enriched, as were neuron development and regulation of synaptic activity pathways in astrocytes. KEGG pathway analysis revealed that L+l-treatment perturbed pathways associated with AD, such as long-term potentiation, circadian entrainment, cAMP signaling, and calcium signaling (Fig. 5g), suggesting that the combination treatment promotes neuroprotective functional pathways in a cell-type-specific manner.

[0168] Example 9 - The combination treatment with letrozole and irinotecan effectively reverses multiple cell-tvoe-specific transcriptomic signatures of AD

[0169] Lastly, drug treatment-reversed transcriptomic signatures of AD were explored to elucidate potential molecular and cellular mechanisms underlying the observed benefits of the L+l treatment. Human AD signatures were mapped to homologous mouse genes. With an absolute log fold change cutoff of 0.01 , the L+l treatment reversed the expression patterns of AD signatures in multiple cell types: 56 (27%) in excitatory neurons, 36 (29%) in inhibitory neurons, 90 (27%) in astrocytes, 52 (20%) in microglia, 37 (15%) in oligodendrocytes, and 64 (34%) in OPCs (Fig. 6a). With a p-adj-value < 0.05 cutoff, 49 and 27 AD signatures were significantly reversed by L+l treatment in excitatory and inhibitory neurons, respectively, with 13 shared between them (Fig. 6a). In glial cells, fewer reversed gene expressions reached statistical significance: 13 in astrocytes, 6 in microglia, 1 1 in oligodendrocytes, and 19 in OPCs. However, they unveiled intriguing patterns. For instance, APOE, identified as a DEG in the human integrated analysis, was upregulated in microglia and downregulated in astrocytes and OPCs. The expression patterns of Apoe in the tested AD mouse model were reversed by L+l treatment across all three cell types (Fig. 6a), although statistical significance was only achieved in astrocytes and OPCs, possibly due to the presence of a small but diverse microglia population in the sequenced cohort. Gene set enrichment analysis was then conducted on significantly reversed genes to identify potential mechanistic targets. Combining reversed genes in excitatory and inhibitory neurons revealed associations with the estrogen signaling pathway among others (Fig. 6b). Reversed genes involved in this pathway also regulate protein kinase and phosphatase activity, along with tau kinase activity specifically, which could be directly relevant to AD. Furthermore, certain reversed genes are associated with synaptic activities and neuron projections, potentially contributing to the rescue of neurodegeneration noted in the pathological analysis. In astrocytes, reversed genes are associated with the regulation of long-term synaptic potentiation, chemical synaptic transmission, and cholesterol metabolism (Fig. 6c). In microglia, reversed genes are associated with pathways regulating synapses and cell growth (Fig. 6c). Additionally, reversed genes from oligodendrocytes and OPCs were combined, revealing associations with pathways related to cell growth regulation, response to reactive oxygen species, and neuron projection regulation (Fig. 6c). These findings provide a transcriptomic foundation, supporting the network correction concept that L+l combination treatment rescues AD-related behavioral and pathological deficits by rectifying complex dysregulated gene networks across multiple diseaserelevant cell types.

[0170] Methods for Examples 2-9

[0171] Human single nuclei RNA-sequencing (snRNA-seq) data curation.

[0172] To ensure a diverse representation of AD patients, publicly available human snRNA-seq datasets were curated from three independent sources. Both Mathys et al. and Zhou et al. studies were obtained from the Accelerating Medicines Partnership Alzheimer’s Disease Project (AMP- AD) Knowledge Portal (adknowledgeportal.org) under the Religious Order Study and Memory and Aging Project (ROSMAP). The Mathys et al. dataset is accessible through doi.org / 10.7303 / syn2580853. The Zhou et al. dataset is available under the study snRNAseqAD_TREM2 and is also accessible through doi.org / 10.7303 / syn21125841. Only individuals without TREM2 mutations were included in the integrated dataset. The third dataset by Lau et al. was obtained from Gene Expression Omnibus (GEO) under the accession number GSE157827.

[0173] Case-control standardization across datasets.

[0174] To standardize AD identification across studies, samples were re-classified into AD or control groups based on tau tangles severity with Braak clinical staging29scores and Ap burden using Consortium to Establish a Registry for Alzheimer’s Disease (CERAD) scores28as a proxy. Braak staging ranges from I to VI to represent low to high levels of tau deposition. The CERAD scoring ranges from 1 to 4 to indicate high to low Ap burden severity. Based on the available metadata from the original studies, AD cases were defined as individuals with severe tau deposition (Braak > IV) and high Ap load (CERAD < 2), and non-AD controls as individuals with low tau (Braak < III) and low Ap load (CERAD > 3). Individuals with scores that did not fulfill both criteria were excluded. The final integrated dataset consisted of 37 AD cases and 29 controls.

[0175] Human snRNA-seq dataset integration, normalization, and batch correction.

[0176] Before merging the datasets, cells deemed as poor quality were removed with the following criteria: total feature count less than 500, total features less than 250, or mitochondrial gene ratio higher than 10%. Sparse features, expressed in fewer than 10 cells, were also removed.

[0177] Simply merging datasets based on principle component analysis (PCA) reveals apparent study-associated variations. To perform batch harmonization while maximizing the conservation of disease-relevant biological variance, canonical correlation analysis (CCA)48was performed using the Seurat package v4.0.4 to harmonize the merged dataset. The default settings were applied to first log-normalize the merged dataset with Seurat’s NormalizeData function. Then, for each dataset independently, the top 2000 variable features were identified using the FindVariableFeatures function with "vst" as the selection method. Integration anchors were identified with the FindlntegrationAnchors function based on the top variable features. Finally, a harmonized dataset consisting of 137,065 cells and 29,120 features was generated using the IntegrateData function in Seurat.

[0178] The data matrix was linearly transformed using Seurat’s ScaleData function. Dimensionality reduction through Uniform Manifold Approximation and Projection (UMAP) was performed with the RunUMAP function, which considers the top 30 dimensions selected from the corresponding principal component analysis (PCA) obtained by running the RunPCA function in Seurat.

[0179] Clustering was determined based on the first 30 PCs using the FindNeighbors function in Seurat, which embeds cells in a K-nearest neighbor graph based on Euclidean distance in PCA space and refines the edge weights between any two cells based on the shared overlap in their local neighborhoods. Clustering was implemented using a resolution of 0.5 in the FindClusters function, which applies modularity optimization techniques such as the Louvain algorithm, resulting in a set of 21 distinct clusters.

[0180] Human cell type annotation.

[0181] Cell type identities were determined by applying Seurat’s AddModuleScore function to lists of known human brain marker genes ( ~ 8 per cell type) collected from PanglaoDB and referenced by Jiang et al. Cell type assignment included astrocytes, microglia, oligodendrocytes, oligodendrocyte precursor cells, endothelial cells, pericytes, and excitatory and inhibitory neurons. Each cell was assigned the corresponding cell type identity that generated the highest scores among scores for all cell types. If the highest and second highest scores of a cell were within 20% of the highest score, then the cells were deemed hybrids and excluded from further analysis. The validity of the assigned cell type identities was assessed by examining the homogeneity, distribution, and separation of cell types by clusterings in UMAP plots.

[0182] Cell-cell communication analysis

[0183] Cell-cell communication (CCC) was calculated using the CellChat method49. Briefly, CellChat utilizes ligand, receptor, and cofactor expression from transcriptomic data to calculate a CCC probability. First, based on a Ce / ZChat-curated database of ligand-receptor interactions, differentially expressed signaling genes are used to calculate the ensemble average expression of signaling genes. Communication probability is modeled using the law of mass action, and statistically significant communications are identified using a permutation test. Evaluated next was the interaction across cell types (excitatory neurons, inhibitory neurons, oligodendrocytes, oligodendrocyte precursor cells, and astrocytes). For the mouse vehicle-treatment comparison, and the human control-AD comparison, the netVisual_difflnteraction was used to calculate the net differential interaction strength between each of the two groups.

[0184] Cell-type-specific differential expression analysis.

[0185] For each cell type, differential gene expression analysis was performed comparing AD samples to controls using the FindMarkers function in the Seurat package. The test. use parameter was set to MAST, which uses a two-part generalized linear model that models gene expression rate using linear regression and expression level using Gaussian distribution, as recommended by Mou et al50, in a study comparing 9 DE methods for single-cell RNA sequencing analysis. Differentially expressed genes (DEGs) were determined as those with adjusted p-value < 0.05 based on Bonferroni correction using all genes in the dataset and a Iog2 fold change (LFC) greater than 0.1 .

[0186] Pathway analysis

[0187] Pathway enrichment analysis was conducted using g:Profiler, a web tool that performs functional over-representation analysis by mapping a given gene list to known functional information sources and detects statistically significant enriched terms. As input gene lists, each cell-type-specific DEG list was split into upregulated and downregulated subsets and queried independently with an adjusted p-value cutoff of 0.05 to obtain significant pathway enrichment in both directions. Multiple testing correction was performed using the g:SCS algorithm, which is specifically correct for p-values obtained from GO and pathway enrichment analysis. This method is designed to correct for multiple tests that may be potentially dependent on each other due to term associations. In addition to considering Gene Ontology cellular components, biological processes, and molecular functions as supplementary data, the analysis focused on the Kyoto Encyclopedia of Genes and Genomes (KEGG) functional pathways. Computational drug repurposing analysis

[0188] The computational drug repurposing algorithm, which was developed by Sirota et al. and Dudley et al., and taken from Chen et al., was applied to each disease gene signature profile using the publicly available Connectivity Map (CMap) database, consisting of treatment profiles of more than 1300 FDA-approved drugs or previously investigated compounds. The employed drug repurposing algorithm takes two inputs: 1 ) an ordered list of up and down-regulated genes from individuals with disease as compared to controls and 2) the data from CMap, consisting of rank FC of each gene after drug treatment relative to vehicle controls on the same plate. The pipeline assesses the disease-drug relationship using CMap scores derived from a Kolmogorov- Smirnov test (K-S test), comparing gene expression ranks in disease and by a drug. The pipeline was adapted for single-cell analysis, where each cell-type-specific DEG signature set was overlaid with 6100 drug profiles on the 1300 drugs provided by the CMap. A drug with a strong negative CMap score indicates an opposing mechanistic relationship with the disease, suggesting therapeutic potential by reversing the regulation direction of disease signatures. The absolute values of CMap scores reflect the degrees to which the drug “flips” the signature of the disease. To address variations in input counts across cell types, significantly reversed drug profiles were identified for each cell type separately using a permutation-based approach. The false discovery rate (FDR; Benjamini-Hochberg) was calculated to adjust P-values. P-values for individual drug hits were determined by comparing reversal scores to a distribution of random scores that were generated by the permutation strategy. Negative reversal scores were deemed significant if they met the criterion of FDR < 0.05. For drugs tested multiple times (e.g., in different cell lines), the profile with the most substantial reversal (lowest negative score) was used.

[0189] Validation in real-world human electronic medical records (EMR)

[0190] While the drug screening approach generated a total of 86 drug candidates targeting six major brain cell types, prioritized were 25 drugs that significantly reverse more than one brain cell type. This allows maximizing coverage of multiple disease-relevant cell types by combining just two drugs. The beneficial effects of the top drug hits were examined by analyzing AD prevalence (% AD diagnosis) in drug-exposed individuals and compared to propensity-matched controls, via UC-wide EMR. AD diagnosis was defined with ICD10 codes G30.0, G30.1 , G30.8, and G30.9. At the time of surveying, the UC-wide EMR aggregated clinical data covering 1 ,441 ,778 individuals across six University of California (UC) campuses, encompassing more than 10 million patient records including diagnosis and medication prescriptions. For each screened drug candidate, patients prescribed or taken the drug will be identified using the medication order table and string-matching for drug names. Only individuals above the age of 65 were considered.

[0191] The matched controls were identified using a propensity score matching approach based on age, age at death (if applicable), race, ethnicity, sex, original indications (for example, breast cancer was used for Letrozole and colorectal cancer was used for Irinotecan), AD comorbidity (such as hypertension and edema), and UC center locations (including UC San Francisco, UC Davis, UC Los Angeles, UC Irvine, UC San Diego, and UC Riverside). The relative risk score for AD diagnosis (not including patients with only a diagnosis of Mild Cognitive Impairment) in the drug-exposed group was measured and compared to the matched control groups by calculating the ratios of AD to total patients in each group. Bootstrapped x-squared tests across 10 permutations of the iterative matching of the control group were applied to establish significance. Drugs with a significantly reduced risk of AD were included in Fig 2e.

[0192] Drug selection rationale for validation in AD mouse model

[0193] Our integrated human single-cell transcriptomic analysis revealed two distinct pathological clusters among the six cell types analyzed. Additionally, it was observed that AD inhibitory neurons exhibited the most pronounced disparity between AD cases and controls, underscoring its significance in AD pathology. These discoveries informed the prioritization of drug candidates targeting these specific clusters and accommodating inhibitory neurons for treating AD. Among a selection of five promising drug candidates that demonstrated a significant reduction in AD risk in humans, letrozole and irinotecan were prioritized, since letrozole targets the neuronal pathological cluster while irinotecan targets the glial pathological cluster. It was hypothesized that the combination of letrozole and irinotecan holds synergistic therapeutic effects, with each pathological cell type cluster addressed by one EMR-validated drug candidate. Consequently, AD outcomes in patients receiving both drugs under their original disease indications were examined in the UC-wide EMR. Unfortunately, fewer than 100 patients were ever prescribed both medications, rendering statistical analysis inconclusive. To explore the synergistic effects of this combination in the context of AD, direct experimentations in animal models were turned to.

[0194] Mouse cohort generation

[0195] The 5xFAD (6SJL) Tg mice (The Jackson Laboratory, ME, USA), overexpress both mutant human amyloid beta (A4) precursor protein 695 (APP) with the Swedish (KM670 / 671 NL), Florida (1716V), and London (V717I) Familial Alzheimer’s Disease (FAD) mutation and human PS1 harboring two FAD mutations, M146L and L286V. These mice were crossed with the widely used tauopathy model, PS19 transgenic mice expressing P301 S mutation and including four microtubule-binding domains and one N-terminal insert. A large cohort of the cross, including both sexes, was genotyped, and the littermates’ carrying transgenes from both lines were subsequently used for the dosing experiment.

[0196] Drug treatments

[0197] A solution of either letrozole (1 mg / kg) or irinotecan (10 mg / kg) alone, or of the two combined was prepared weekly in 0.9% sterile saline containing 10% Tween-20 to obtain the final concentrations. Treatment dosages were established based on previously tolerated doses given during cancer studies in mice. All treatments were sonicated for 30 minutes before injections to ensure proper mixing of the drugs into the solution. The treatments were given by i.p. injection every other day to the 5xFAD x PS19 double transgenic cohort, evenly split into four groups and sex-balanced, starting 12 weeks before and continuing throughout the behavioral assessment. Mice were aged 4-5 months old before the start of the treatment and the treatments lasted about 4 months until sacrificed. Body weight was measured weekly during drug treatment; injection volume was calculated based on body weight. Control mice were injected with a matched volume of vehicle, 10% Tween in 0.9% sterile saline at pH 8.5. Injections were well tolerated and had no adverse effects on health.

[0198] Behavioral test

[0199] Mice were housed with littermate controls. Each mouse was assigned a random number, so researchers were blinded to genotype and treatment information. Male and female mice were tested in separate rooms with the same settings and test duration to avoid olfactory cues becoming a distraction with testing a mixed sex cohort on the same equipment

[0200] The Morris Water Maze pool (diameter, 122 cm) contained opaque water (21 + 13°C) with a platform 10 cm in diameter. Mice were first trained for 6 days to locate a hidden platform submerged 1 .5cm below the surface of the water using distal cues surrounding the pool. These 6 days consisted of two training sessions, 2 hours apart, each consisting of two trials with a maximum latency of 60s, with a 15-minute inter-trial interval. Entry points were changed semirandomly between trials, but distal visual cues on the walls of the behavioral testing room remained constant throughout the test. To test for memory retention, at 24 and 72 hours after the last hidden platform training, a 60-s probe trial (platform removed) was done. The entry point for the probe trial was opposite to the target. Finally, as a control for visual acuity and motor ability, mice were tested with a cued platform where the submerged platform is indicated by a black and white striped mast 15cm high extending out the surface of the water. The platform location remained constant in the hidden platform sessions but was changed for each visible platform session. Three sessions of two trials, with a maximum latency of 60s, was performed over 2 days. Performance was monitored with an EthoVision video-tracking system (Noldus Information Technology). For the probe trials, analyzed was (1 ) target quadrant preference- the percent time spent in the target quadrant versus average time spent in the three other quadrants, (2) platform crossing- the number of crossings over the position of the target platform versus the average number of crossings over the equivalent positions in the three other quadrants as well as (3) escape latency- a memory probe measuring how fast mice arrive at the platform location after placement in the maze. An ANOVA was used to analyze the effect of treatment, genotype, and probe timing on percent on-target crosses. Histopathological analyses

[0201] One hemibrain per mouse was drop-fixed in 4% paraformaldehyde prepared in 1xPBS, washed for 24 hours in 1xPBS, then cryoprotected in 30% sucrose for 48 hours at 4°C. The fixed hemispheres were cut into 30 pm thick coronal slices using a freeze-sliding microtome (Leica). These sections were then stored in a cryoprotectant solution of 30% ethylene glycol, 30% glycerol, and 40% 1 xPBS at -20 °C.

[0202] Hippocampal brain sections (ten sections per mouse spaced approximately 300 pm apart, 30-pm thick) were mounted onto microscope slides from Fisher Scientific. A 0.1% Sudan Black solution was prepared by dissolving Sudan Black powder (Sigma) in 70% ethanol (KOPTEC) and mixing it with a magnetic stirrer. After centrifuging the solution at 1 ,100g for 10 minutes, the supernatant was filtered using a 0.2-pm filter syringe (Thermo Scientific). The brain sections were stained with the 0.1 % Sudan Black solution for 10 minutes, then washed in 70% ethanol followed by Milli-Q water. The sections were then coverslipped with Prolong Gold mounting medium (Invitrogen). All sections were imaged; eight consecutive brain sections with consistent anatomic locations from AP -1 .2 to -3.6 were quantified.

[0203] For Thio-S staining, several brain sections spaced 300 pm apart were mounted onto slides, following a protocol adapted from a previous study. The sections were washed with 1 x PBS-T and then incubated in a solution of 0.06% Thio-S in PBS for 8 minutes. After this incubation, the sections were washed for 1 minute in 80% ethanol and then for 5 minutes in PBS- T. The sections were counterstained with DAPI for 8 minutes, washed again with PBS-T, and coverslipped. Three sections per mouse were quantified and averaged for percent Thio-S positive area and number of plaque counts (normalized by hippocampal area size per section).

[0204] For immunofluorescent staining, several brain sections spaced approximately 300 pm apart, were washed three times with 1 x PBS-T (PBS + 0.1 % Tween-20) (Millipore Sigma) and then incubated for 5 minutes in boiling antigen retrieval buffer (Tris buffer, pH 7.6; TEKNOVA). After this, the sections were rinsed in PBS-T before being placed in a blocking solution composed of 5% normal donkey serum (Jackson Labs) and 0.2% Triton-X (Millipore Sigma) in 1 x PBS for 1 hour at room temperature. Next, the sections were washed again in PBS-T and incubated in Mouse-on-Mouse (MOM) Blocking Buffer (one drop MOM IgG in 4 ml PBS-T) (Vector Labs) for another hour at room temperature. Following the MOM block, the sections were incubated overnight at 4°C in primary antibodies diluted to their optimal concentrations. The antibodies and their dilutions included: anti-AT8 (ms, 1 :300, Invitrogen), anti-GFAP (ms, 1 SOO, Millipore Sigma), anti-lba1 (rbt, 1 SOO, Wako), anti-NeuN (GP, 1 SOO, Millipore Sigma). Three sections per mouse were quantified and averaged for a positive percent area.

[0205] After the primary antibody incubation, the sections were washed in PBS-T and then incubated for 1 hour at room temperature in secondary antibodies. These included donkey antimouse 488 (1 :1 ,000, Abeam), donkey anti-rabbit 594 (1 :1 ,000, Abeam), donkey anti-guinea pig 594 (1 :1 ,000, Jackson Immuno), and donkey anti-guinea pig 647 (1 :1 ,000, Jackson Immuno). Subsequently, the sections were washed in PBS-T and incubated in DAPI (1 :30,000 dilution in PBS-T) (Thermo Fisher) for 8 minutes at room temperature. After a final wash with PBS-T, the sections were mounted onto microscope slides (Fisher Scientific), coverslipped with ProLong Gold mounting medium (Vector Laboratories), and sealed with clear nail polish.

[0206] Images for quantifications were captured using a scanning microscope (Keyence) at magnifications of x10 or x20, depending on the stain. To minimize batch-to-batch variation, all samples for each stain were processed simultaneously and imaged at the same fluorescent intensity. For quantifying the percent coverage area, an optimal threshold was established for each stain in ImageJ, and all samples were quantified using this threshold. To prevent bias, researchers were blinded to the sample identities. Representative images were captured using an Aperio VERSA slide scanning microscope (Leica) at x20 magnification. snRNA-seq library preparation and sequencing

[0207] The other hemibrains were dissected by brain subregions, rapidly frozen on dry ice, and kept at -80 °C. Hippocampal samples were used for single-nuclei preparation. One frozen mouse hippocampus was placed into a pre-chilled 2 mL Dounce with 1 mL of cold 1 X Homogenization Buffer (1 X HB) (250 mM Sucrose, 25 mM KCL, 5 mM MgCh, 20 mM Tricine-KOH pH7.8, 1 mM DTT, 0.5 mM Sermidine, 0.15 mM Sermine, 0.3% NP40, 0.2 units / pL RNase inhibitor, -0.07 tabs / ml complete Protease inhibitor). Dounce with “A” loose pestle (-10 strokes) and then with “B” tight pestle (-10 strokes). The homogenate was filtered using a 70 pM Flowmi strainer (Bel- Art) and transferred to a pre-chilled 2 mL LoBind tube (Fischer Scientific). Nuclei were pelleted by spinning for 5 min at 4°C at 350 RCF. The supernatant was removed and the nuclei were resuspended in 400 pL 1X HB. Next, 400 pL of 50% lodixanol solution was added to the nuclei and then slowly layered with 600 pL of 30% lodixanol solution under the 25% mixture, then layered with 600 pL of 40% lodixanol solution under the 30% mixture. The nuclei were then spun for 20 min at 4°C at 3,000RCF in a pre-chilled swinging bucket centrifuge. 200 pL of the nuclei band at the 30%-40% interface was collected and transferred to a fresh tube. Then, 800 pL of 2.5% BSA in PBS plus 0.2 units / pL of RNase inhibitor was added to the nuclei and then were spun for 10 min at 500 RCF at 4C. The nuclei were resuspended with 2% BSA in PBS plus 0.2 units / pL RNase inhibitor to reach at least 500 nuclei / pL. The nuclei were then filtered with a 40 pM Flowmi cell strainer. The nuclei were counted and then -13,000 nuclei per sample were loaded onto 10x Genomics Next GEM chip M. The snRNA-seq libraries were prepared using the Chromium Next GEM Single Cell 3' HT kit v3.1 (10x Genomics) according to the manufacturer’s instructions. Libraries were sequenced on an Illumina NovaSeq 6000 sequencer at the UCSF CAT sequencing core. Sequence alignment, filtering, and counting

[0208] The demultiplexed fastq files were processed following the procedure previously described by Zalocusky et al. In summary, the fastq files were aligned to the mouse reference genome, mm10-1.2.0, which includes introns, using the cellranger count function (version 4.0.0) with default parameters, as detailed in the Cell Ranger documentation. Subsequently, a single UMI count file per animal / sample was generated by the cellranger count function. Individual UMI count files were then combined into a single count matrix using the merge function in the Seurat package v4.0.4. Metadata, including age, sex, and treatment information, were added to each cell.

[0209] Pre-processing and quality control

[0210] The count matrix was further processed with Seurat by first calculating the percentage of mitochondria genes mapped per cell. The distribution of feature count, total mapped gene count, and percentage of mitochondria genes were visualized across biological samples as violin plots, and no obvious outlier was identified. The count matrix was filtered to only include cells with higher than 250 gene features, at least 500 gene counts, and mitochondria gene percentages lower than 10%. Potential misaligned or ambiguous gene features expressed in fewer than 10 cells were also removed. These quality assurance steps resulted in a final Seurat object containing 25,642 gene features expressed by 237,853 nuclei.

[0211] Normalization, dimensional reduction, and clustering

[0212] Count normalization and dimensionality reduction were conducted following standard procedure in the Seurat package. In brief, normalization and variance stabilization were performed with an updated version of sctransform, v2, and principal component analysis (PCA) with RunPCA (npcs = 30). Dimensional reduction through Uniform Manifold Approximation and Projection (UMAP) was performed with the RunUMAP function and considering the top 15 dimensions selected from the corresponding PCA.

[0213] The clustering was based on the first 15 principal components (PCs) using the FindNeighbors function in Seurat. This function embeds cells in a K-nearest neighbor graph, considering the Euclidean distance in PCA space and refining the edge weights between any two cells based on the shared overlap in their local neighborhoods. The clustering was obtained using the FindClusters function, which employs modularity optimization techniques such as the Louvain algorithm, with a resolution parameter of, resulting in a set of 35 distinct clusters.

[0214] Mouse cell type annotation, differential gene expression, and pathway enrichment analysis

[0215] Major cell types, including astrocytes, microglia, oligodendrocytes, oligodendrocyte precursor cells, and excitatory and inhibitory neurons, were classified using mouse brain cell markers in PanglaoDB, a publicly available marker gene database. Further subdivisions of hippocampus cell types, such as CA1 and CA3 pyramidal cells, were queried against hippocampal cell-type-specific marker genes as published in hipposeq (hipposeq.janelia.org). Cell type identities per cluster were determined by applying Seurat’s AddModuleScore function to sets of mouse brain marker genes. A module score for each cell type considered was calculated per cell. Each cell was assigned the corresponding cell type identity that generated the highest scores among scores for all cell types. If the highest and second highest scores of a cell were within 20% of the highest score, then the cells were deemed hybrids and excluded from further analysis. The validity of the assigned cell type identities was assessed by examining the homogeneity, distribution, and separation of cell types by clustering in UMAP plots. Minority cell types in a cluster, defined as a cell type that accounts for less than 5% of the total counts for that cluster, were considered potential hybrid cells and excluded from further analysis.

[0216] For the cell-type-specific differential gene expression analysis, the same procedure and tools were used in the human-integrated analysis. For pathway enrichment analysis, in addition to using g:Profiler, curated was the EnrichR-KG web tool, which facilitates analysis across multiple databases and provides visual representations linking significantly enriched genes with the associated pathways and GO terms. The reversal gene-pathway network was generated and downloaded using EnrichR-KG with significantly differentially reversed genes as inputs. For each cell-type-stratified analysis, the top five KEGG pathways and GO terms were displayed in a network format.

[0217] References for Introduction and Examples 2-9

[0218] 1 . Alzheimer’s Association 2024 Alzheimer’s Disease Facts and Figures.

[0219] 2. What Is Alzheimer’s Disease? National Institute on Aging www.nia.nih.gov / health / alzheimers-and-dementia / what-alzheimers-disease (2021 ).

[0220] 3. GBD 2019 Dementia Forecasting Collaborators. Estimation of the global prevalence of dementia in 2019 and forecasted prevalence in 2050: an analysis for the Global Burden of Disease Study 2019. Lancet Public Health 7, e105-e125 (2022).

[0221] 4. Alzheimer’s Disease Fact Sheet. National Institute on Aging http: / / www.nia.nih.gov / health / alzheimers-disease-fact-sheet.

[0222] 5. Dementia, www.who.int / news-room / fact-sheets / detail / dementia.

[0223] 6. Kim, C. K. et al. Alzheimer’s Disease: Key Insights from Two Decades of Clinical Trial Failures. J Alzheimers Dis 87, 83-100 (2022).

[0224] 7. How Is Alzheimer’s Disease Treated? National Institute on Aging http: / / www.nia.nih.gov / health / how-alzheimers-disease-treated.

[0225] 8. Valiukas, Z. etal. Immunotherapies for Alzheimer’s Disease — A Review. Vaccines (Basel) 10, 1527 (2022). 9. Bloom, G. S. Amyloid-p and tau: the trigger and bullet in Alzheimer disease pathogenesis. JAMA Neurol 7 , 505-508 (2014).

[0226] 10. Guerreiro, R. J., Gustafson, D. R. & Hardy, J. The genetic architecture of Alzheimer’s disease: beyond APP, PSENs and APOE. Neurobiol Aging 33, 437-456 (2012).

[0227] 11 . Toledano-Diaz, A., Alvarez, M. I. & Toledano, A. The relationships between neuroglial alterations and neuronal changes in Alzheimer’s disease, and the related controversies I: Gliopathogenesis and glioprotection. J Cent Nerv Syst Dis 14, 1 1795735221 128703 (2022).

[0228] 12. Appleby, B. S., Nacopoulos, D., Milano, N., Zhong, K. & Cummings, J. L. A Review: Treatment of Alzheimer’s Disease Discovered in Repurposed Agents. DEM 35, 1-22 (2013).

[0229] 13. Cheung, C., Goh, Y. T., Zhang, J., Wu, C. & Guccione, E. Modeling cerebrovascular pathophysiology in amyloid-p metabolism using neural-crest-derived smooth muscle cells. Cell Pep 9, 391-401 (2014).

[0230] 14. Choi, Y. et al. Clozapine Improves Memory Impairment and Reduces Ap Level in the Tg- APPswe / PS1 dE9 Mouse Model of Alzheimer's Disease. Mol Neurobiol 54, 450-460 (2017).

[0231] 15. Taubes, A. et al. Experimental and real-world evidence supporting the computational repurposing of bumetanide for APOE4-related Alzheimer’s disease. Nat Aging 1 , 932-947 (2021 ).

[0232] 16. Gerring, Z. F., Gamazon, E. R., White, A. & Derks, E. M. Integrative Network-Based Analysis Reveals Gene Networks and Novel Drug Repositioning Candidates for Alzheimer Disease. Neurology Genetics 7, e622 (2021 ).

[0233] 17. Rodriguez, S. et al. Machine learning identifies candidates for drug repurposing in Alzheimer’s disease. Nat Common 12, 1033 (2021 ).

[0234] 18. Endophenotype-based in silico network medicine discovery combined with insurance record data mining identifies sildenafil as a candidate drug for Alzheimer’s disease | Nature Aging. www.nature.com / articles / s43587-021 -00138-z.

[0235] 19. Grabowska, M. E., Huang, A., Wen, Z., Li, B. & Wei, W.-Q. Drug repurposing for Alzheimer’s disease from 2012-2022 — a 10-year literature review. Front Pharmacol 14, 1257700 (2023).

[0236] 20. Pajai, S., Potdar, J., Gopal, U. & Banait, T. A Review on the Use of Letrozole in Female and Male Infertility. Cureus 14, e31291 .

[0237] 21 . Fujita, K„ Kubota, Y„ Ishida, H. & Sasaki, Y. Irinotecan, a key chemotherapeutic drug for metastatic colorectal cancer. World J Gastroenterol 21 , 12234-12248 (2015).

[0238] 22. Mathys, H. et al. Single-cell transcriptomic analysis of Alzheimer’s disease. Nature 570, 332-337 (2019). 23. Lau, S.-F., Cao, H., Fu, A. K. Y. & Ip, N. Y. Single-nucleus transcriptome analysis reveals dysregulation of angiogenic endothelial cells and neuroprotective glia in Alzheimer’s disease. PNAS 117, 25800-25809 (2020).

[0239] 24. Zhou, Y. et al. Human and mouse single-nucleus transcriptomics reveal TREM2- dependent and TREM2-independent cellular responses in Alzheimer’s disease. Nat Med 26, 131-142 (2020).

[0240] 25. Lamb, J. et al. The Connectivity Map: Using Gene-Expression Signatures to Connect Small Molecules, Genes, and Disease. Science 313, 1929-1935 (2006).

[0241] 26. Tau P301 S (Line PS19) | ALZFORUM. www.alzforum.org / research-models / tau-p301s- Iine-ps19.

[0242] 27. 5xFAD (B6SJL) | ALZFORUM. www.alzforum.org / research-models / 5xfad-b6sjl.

[0243] 28. Mirra, S. S. et al. The Consortium to Establish a Registry for Alzheimer’s Disease (CERAD). Part II. Standardization of the neuropathologic assessment of Alzheimer’s disease. Neurology 41 , 479-486 (1991).

[0244] 29. Braak, H., Alafuzoff, I., Arzberger, T., Kretzschmar, H. & Del Tredici, K. Staging of Alzheimer disease-associated neurofibrillary pathology using paraffin sections and immunocytochemistry. Acta Neuropathol 112, 389-404 (2006).

[0245] 30. H, M. et al. Single-cell atlas reveals correlates of high cognitive function, dementia, and resilience to Alzheimer’s disease pathology. Ce / / 186, (2023).

[0246] 31 . Leung, L. et al. Apolipoprotein E4 Causes Age- and Sex-Dependent Impairments of Hilar GABAergic Interneurons and Learning and Memory Deficits in Mice. PLoS ONE7, (2012).

[0247] 32. Najm, R., Jones, E. A. & Huang, Y. Apolipoprotein E4, inhibitory network dysfunction, and Alzheimer’s disease. Mol Neurodegeneration 14, 1-13 (2019).

[0248] 33. Blumenfeld, J., Yip, O., Kim, M. J. & Huang, Y. Cell type-specific roles of APOE4 in Alzheimer disease. Nat Rev Neurosci 25, 91-110 (2024).

[0249] 34. Knox, C. et al. DrugBank 6.0: the DrugBank Knowledgebase for 2024. Nucleic Acids Res 52, D1265-D1275 (2024).

[0250] 35. Lamb, H. M. & Adkins, J. C. Letrozole. A review of its use in postmenopausal women with advanced breast cancer. Drugs 56, 1125-1140 (1998).

[0251] 36. Saul, A., Sprenger, F., Bayer, T. A. & Wirths, O. Accelerated tau pathology with synaptic and neuronal loss in a novel triple transgenic mouse model of Alzheimer’s disease. Neurobiology of Aging 34, 2564-2573 (2013).

[0252] 37. Stancu, l.-C. et al. Tauopathy contributes to synaptic and cognitive deficits in a murine model for Alzheimer’s disease. The FASEB Journal 28, 2620-2631 (2014). 38. Morris water maze: procedures for assessing spatial and related forms of learning and memory | Nature Protocols, www.nature.com / articles / nprot.2006.116.

[0253] 39. Peyton, M. E. Single-cell Approach to Repurposing of Drugs for Alzheimer’s Disease. (2023).

[0254] 40. Parolo, S., Mariotti, F., Bora, P., Carboni, L. & Domenici, E. Single-cell-led drug repurposing for Alzheimer’s disease. Sci Rep 13, 222 (2023).

[0255] 41 . Dave, N., Gudelsky, G. A. & Desai, P. B. The pharmacokinetics of letrozole in brain and brain tumor in rats with orthotopically implanted C6 glioma, assessed using intracerebral microdialysis. Cancer Chemother Pharmacol 72, 349-357 (2013).

[0256] 42. Goldwirt, L., Beccaria, K., Carpentier, A., Farinotti, R. & Fernandez, C. Irinotecan and temozolomide brain distribution: a focus on ABCB1 . Cancer Chemother Pharmacol74, 185-193 (2014).

[0257] 43. Branigan, G. L., Soto, M., Neumayer, L., Rodgers, K. & Brinton, R. D. Association Between Hormone-Modulating Breast Cancer Therapies and Incidence of Neurodegenerative Outcomes for Women With Breast Cancer. JAMA Network Open 3, e201541 (2020).

[0258] 44. Stenger, M. Use of Chemotherapy and Risk of Alzheimer’s Disease in Colorectal Cancer Survivors. ascopost.com / news / march-2021 / use-of-chemotherapy-and-risk-of-alzheimer-s- disease-in-colorectal-cancer-survivors / .

[0259] 45. Guo, L., Zhong, M. B., Zhang, L, Zhang, B. & Cai, D. Sex Differences in Alzheimer’s Disease: Insights From the Multiomics Landscape. Biol Psychiatry 91 , 61-71 (2022).

[0260] 46. Sun, Y. et al. The behavioural and neuropathologic sexual dimorphism and absence of MIP-3a in tau P301 S mouse model of Alzheimer’s disease. Journal of Neuroinflammation 17 , 72 (2020).

[0261] 47. Sil, A. etal. Sex Differences in Behavior and Molecular Pathology in the 5XFAD Model. J Alzheimers Dis 85, 755-778 (2022).

[0262] 48. Integrative analysis in Seurat v5 • satijalab.org / seurat / articles / seurat5_integration.

[0263] 49. Jin, S. et al. Inference and analysis of cell-cell communication using CellChat. Nat Common 12, 1088 (2021).

[0264] 50. Mou, T., Deng, W., Gu, F., Pawitan, Y. & Vu, T. N. Reproducibility of Methods to Detect Differentially Expressed Genes from Single-Cell RNA Sequencing. Front. Genet. 10, (2020).

[0265] Accordingly, the preceding merely illustrates the principles of the present disclosure. It will be appreciated that those skilled in the art will be able to devise various arrangements which, although not explicitly described or shown herein, embody the principles of the invention and are included within its spirit and scope. Furthermore, all examples and conditional language recited herein are principally intended to aid the reader in understanding the principles of the invention and the concepts contributed by the inventors to furthering the art, and are to be construed as being without limitation to such specifically recited examples and conditions. Moreover, all statements herein reciting principles, aspects, and embodiments of the invention as well as specific examples thereof, are intended to encompass both structural and functional equivalents thereof. Additionally, it is intended that such equivalents include both currently known equivalents and equivalents developed in the future, i.e., any elements developed that perform the same function, regardless of structure. The scope of the present invention, therefore, is not intended to be limited to the exemplary embodiments shown and described herein.

Claims

WHAT IS CLAIMED IS:1 . A method of treating Alzheimer's disease (AD), the method comprising: administering to a subject having Alzheimer's disease a therapeutically effective amount of irinotecan and letrozole.

2. The method of claim 1 , wherein the irinotecan and letrozole are administered sequentially.

3. The method of claim 2, wherein the irinotecan and letrozole are administered to the subject in separate compositions.

4. The method of claim 1 , wherein the irinotecan and letrozole are administered to the subject in a single composition.

5. The method of any one of claims 1 -4, wherein the irinotecan and letrozole are administered orally or parenterally to the subject.

6. A pharmaceutical composition comprising: irinotecan; letrozole; and a pharmaceutically acceptable carrier.

7. The pharmaceutical composition of claim 6, wherein the composition is formulated for oral administration to a subject in need thereof.

8. The pharmaceutical composition of claim 6, wherein the composition is formulated for parenteral administration to a subject in need thereof.

9. A kit comprising: irinotecan; letrozole; and instructions for administering the irinotecan and letrozole to a subject having Alzheimer’s Disease (AD).

10. The kit of claim 9, wherein the irinotecan and letrozole are provided as separate pharmaceutical compositions.11 . The kit of claim 9, wherein the irinotecan and letrozole are provided in a single pharmaceutical composition.

12. The kit of claim 10 or 11 , comprising two or more unit dosages of the pharmaceutical composition(s).

Images