Cross-tissue immune profiling of APOE ε4 reveals early dysregulation in Alzheimer’s disease

Cross-tissue immune profiling of APOE ε4 reveals early dysregulation in Alzheimer’s disease | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Cross-tissue immune profiling of APOE ε4 reveals early dysregulation in Alzheimer’s disease Caitlin Finney, Artur Shvetcov, Shannon Thomson, Mark Graham, and 16 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7089423/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted You are reading this latest preprint version Abstract APOE ε4 is the strongest genetic risk factor for late-onset Alzheimer’s disease (AD), but its contribution to disease pathogenesis remains incompletely understood. Here, we integrate proteomic profiling of plasma, cerebrospinal fluid (CSF), and brain tissue from over 10,000 individuals to define the immune phenotype associated with APOE ε4. We identify a conserved, allele dose-dependent pro-inflammatory immune signature across peripheral and central tissues independent of AD diagnosis. This signature is enriched in adaptive immune cells and white matter-resident glial and vascular cells. It also emerges in patient-derived cortical organoids prior to amyloid-β and tau pathology, supporting a causal, genotype-driven mechanism. Cross-tissue comparisons reveal shared innate and antiviral responses alongside tissue-specific immune signaling. Notably, a 12-week medical ketogenic diet partially reversed the APOE ε4 immune signature. These findings position immune dysregulation as an early and tractable driver of AD risk in APOE ε4 carriers with direct implications for targeted prevention strategies. Biological sciences/Neuroscience/Diseases of the nervous system/Alzheimer's disease Biological sciences/Neuroscience/Neuroimmunology Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Main The functional consequences of the largest genetic risk factor for late onset Alzheimer’s disease (AD), the ε4 variant of the apolipoprotein E ( APOE ) gene, and how they drive AD pathogenesis remain poorly understood. Recent accumulating evidence suggests that APOE ε4 exerts broad modulatory effects on both innate and adaptive immunity across peripheral tissues and the central nervous system. APOE ε4 is associated with a shift toward a heightened pro-inflammatory state 1,2 . This is further evidenced by peripherally-derived immune cells showing exaggerated responses to innate immune stimulation 3,4 , altered antigen presentation 5 , disrupted T-cell homeostasis 6 , and signs of accelerated immune aging 7 . In the brain, APOE ε4 carriage is linked to pre-activated microglial phenotypes 8 , impaired lipid and autophagic homeostasis 9,10 , and amplified responses to amyloid-β pathology 10,11 . Together, these findings highlight a genotype-dependent dysregulation of immune homeostasis in APOE ε4 carriers. However, several key questions remain, including whether immune changes represent innate genetic effects or secondary responses to disease, how these signatures evolve temporally with progression to AD, which aspects are systemic, central, or peripheral in origin, and whether they are modifiable. Resolving these questions is essential to understanding APOE ε4-driven mechanisms of neurodegeneration and developing precision biomarkers and early intervention strategies for AD. To address how APOE ε4 influences immune function across tissues and time, we systematically profiled the plasma, CSF, and brain proteomes of APOE ε4 carriers and non-carriers with and without AD (Fig. 1a). We identified a conserved, allele dose-dependent pro-inflammatory immune signature spanning both the periphery and central nervous system (CNS), independent of disease. This signature was recapitulated in patient-derived cortical organoids and emerged prior to the development of amyloid-β and tau pathology, indicating that APOE ε4-driven immune changes are early, intrinsic, and may initiate downstream neurodegenerative processes. Cross-tissue comparisons revealed that many APOE ε4-associated pathways were systemic, reflecting broad innate and adaptive immune activation. Others were tissue-specific and localized to the CNS or periphery. Importantly, short-term treatment with an anti-inflammatory medical ketogenic diet in APOE ε4 carriers with AD partially reprogrammed this immune phenotype, reducing pro-inflammatory signaling and promoting regulatory and tissue-supportive immune functions. Together, these findings demonstrate that APOE ε4 confers a systemic but locally modulated pro-inflammatory immune phenotype that emerges early, evolves with disease progression, and may promote AD pathology. This provides new insight into APOE ε4-driven mechanisms of neurodegeneration and highlights the potential of this immune state to be therapeutically modulated. Results APOE ε4 carriers have a distinct pro-inflammatory immune phenotype in the plasma and cerebrospinal fluid We profiled 6,340 plasma proteins of n = 2,929 AD and n = 6,099 non-impaired controls from the Global Neurodegeneration Proteomics Consortium (GNPC) using the SomaScan 7k assay (Supplementary Table 1). Nine APOE ε4-associated plasma proteins were identified (mutual information > 0.1), which clustered by carrier status and allele dose (Supplementary Table 2; Extended Data Fig. 1a-c). Machine learning using classification and regression trees (CART) showed these proteins robustly classified APOE ε4 carriers versus non-carriers (AUC > 0.91 across subgroups; Extended Data Table 1; Fig. 1b), independent of AD status, sex, or race. No plasma proteins differentiated AD APOE ε4 carriers from non-impaired carriers (mutual information 0.93). We then analyzed 6,340 CSF proteins from n = 526 AD and n = 573 non-impaired controls (Supplementary Table 1), identifying 51 APOE ε4-associated proteins (mutual information > 0.1) that similarly clustered by carrier status and allele dose (Supplementary Table 3; Extended Data Fig. 1d-f). CART models accurately classified APOE ε4 carriers (AUC = 0.99), again independent of AD status or sex, and distinguished between heterozygous and homozygous individuals (AUC > 0.94; Extended Data Table 2; Fig. 1c). As in plasma, no CSF proteins distinguished AD from non-impaired APOE ε4 carriers (mutual information < 0.01; AUC = 0.75), further supporting an AD-independent immune signature. Six of the nine plasma proteins overlapped with CSF APOE ε4 proteins (Fig. 1d). Pathway analyses revealed significant enrichment for viral processes (FDR < 0.05) in both plasma and CSF (PANTHER; Fig. 1e; Supplementary Table 4). KEGG and Reactome analyses confirmed enrichment for immune pathways including interferon, interleukin, Th17, TGF-β signaling, Epstein-Barr virus and hepatitis (Fig. 1e-f; Supplementary Table 4). CSF-specific pathways included toll-like receptor signaling and NK/B cell pathways, while plasma-specific enrichments were dominated by interferon signaling (Fig. 1e-f; Supplementary Table 4). Cell-type enrichment using Human Protein Atlas single-cell RNA-sequencing data 12,13 , showed plasma APOE ε4 proteins were enriched in innate immune cells (neutrophils, basophils) and T cells, particularly central memory CD8 + and CD4 + TFH subsets (Fig. 1h-i). CSF proteins were enriched in NK cells and adaptive T-cell subsets, including CD4 + Th1 and CD8 + terminal effector memory cells (Fig. 1i). Both plasma and CSF APOE ε4 proteins were preferentially expressed in white matter, with further enrichment in microglia, oligodendrocytes, and vascular-associated cells in CSF (Fig. 1j-k). Together, these findings define a robust, AD-independent, allele dose-sensitive immune signature in APOE ε4 carriers that spans the plasma and CNS and is enriched in adaptive immune and white matter-resident cell types. The APOE ε4 phenotype extends into the brain, with the superior temporal gyrus especially affected To assess whether the APOE ε4-associated immune phenotype extended into the brain, we analysed TMT-based proteomic data from the Accelerating Medicine Partnerships in AD (AMP-AD) Diverse Cohorts study in the dorsolateral prefrontal cortex (dlPFC; n = 530 AD, n = 190 non-impaired controls) and superior temporal gyrus (STG; n = 161 AD, n = 49 non-impaired controls; Supplementary Table 5). We identified 85 APOE ε4-associated proteins in dlPFC and 56 in STG of non-impaired controls (mutual information > 0.1; Supplementary Table 6), with only two proteins overlapping across regions, GORAB and PHOSPHO1, suggesting region-specific effects of APOE ε4. Coverage differences between SomaScan and TMT precluded direct plasma/CSF comparisons. Random forest modelling showed that APOE ε4 brain proteins reliably classified APOE ε4 carriers versus non-carriers with AD (dlPFC AUC = 0.94; STG AUC = 0.91), independent of sex and race for dlPFC (AUC > 0.88; Extended Data Table 3; Fig. 2a). Insufficient sample size precluded subgroup analysis in STG. Consistent with plasma and CSF findings, viral processes were the most significantly enriched APOE ε4 associated biological processes (PANTHER; Fig. 2b; Supplementary Table 7). KEGG analyses revealed significant enrichment for pro-inflammatory, infection-related, and adaptive immune pathways. There were also region-specific pathways including Fc receptor and platelet activation in dlPFC, and TLR and RLR signaling in STG (Fig. 2c; Supplementary Table 7). Reactome pathways overlapped across regions for adaptive immune, B cell receptor, and HIV-related pathways. dlPFC-specific pathways reflected immune cell activation and cytokine regulation whereas STG pathways were more focused on pathogen sensing and antiviral defense (Fig. 2d; Supplementary Table 7). Across both regions, APOE ε4 brain proteins were preferentially enriched in microglia and endothelial cells (Fig. 2e). These findings identify distinct, region-specific APOE ε4 brain protein signatures, independent of sex and race, that mirror the viral, pro-inflammatory, and adaptive immune pathway enrichment seen in plasma and CSF. The regional differences suggest nuanced immune and pathogen-sensing functions in APOE ε4 brains, with microglia and cerebrovascular endothelial enrichment implicating white matter and vascular immune processes in the APOE ε4 brain phenotype. Stem cell-derived cortical organoids show the APOE ε4 phenotype before Alzheimer’s pathology onset Across two clinical cohorts, we demonstrated that APOE ε4 carriers exhibit a pro-inflammatory immune phenotype both peripherally and centrally. We previously showed that this phenotype is not correlated with hallmark AD pathologies including amyloid-β, tau, gliosis, or angiopathy 2 . However, since these clinical samples are derived from symptomatic individuals or asymptomatic individuals who may harbor subclinical AD pathology, it remains unclear whether this APOE ε4 immune phenotype precedes or follows the development of AD pathology, a critical question for guiding diagnosis and treatment in APOE ε4 carriers. To address this, we generated cortical organoids 15,16 from induced pluripotent stem cell (iPSC) lines of two donors from the University of Kansas Alzheimer’s Disease Research Center cohort: a heterozygous APOE ε4 carrier who developed AD and a non-carrier control. To better model APOE ε4 immune interactions, iPSC-derived microglia from the same donors were integrated into cortical organoids (Methods; Extended Data Fig. 2). At four weeks maturation, neither APOE ε4 nor non-carrier organoids showed evidence of p-tau217 or amyloid-β accumulation (Fig. 3a,b) and no differences in secreted amyloid-β42/40 or p-tau217 (Fig. 3d,f,g,i). By eight weeks, however, APOE ε4 organoids exhibited marked accumulation of p-tau217 (t (2.5) = 4.48, p = 0.0203, unpaired two-tailed t-test; Fig. 3a,c) and amyloid-β (t (2.5) = 10.02, p = 0.0010, unpaired two-tailed t-test; Fig. 3a,c). This was accompanied by significantly reduced levels of secreted amyloid-β42/40 compared to non-carriers (t (8) = 2.40, p = 0.043; unpaired two-tailed t-test; Fig. 3e), consistent with amyloid-β retention in APOE ε4 carrier organoids. A significant time × APOE genotype interaction in amyloid-β42/40 secretion was also observed (F (1,8) = 9.12, p = 0.017; two-way ANOVA; Fig. 3f). Despite intracellular p-tau217 accumulation in APOE ε4 organoids at eight weeks, no significant differences were observed in secreted p-tau217 across groups or over time (Fig. 3g-i). At four weeks of maturation, APOE ε4 cortical organoids lacked hallmark AD pathology, providing an opportunity to study early pre-pathological processes. To investigate how the APOE ε4 proteome evolves with disease onset, we performed label-free mass spectrometry on 10,828 proteins from APOE ε4 and non-carrier organoids at four and eight weeks. At four weeks, 642 proteins were differentially expressed between APOE ε4 and non-carriers, increasing to 2,655 by eight weeks (adjusted p < 0.05; Supplementary Table 8; Fig. 4a-b). KEGG enrichment confirmed absence of AD pathway proteins at four weeks, with significant enrichment emerging by eight weeks (FDR = 2.1 × 10 -6 Fig. 5c). Consistent with our plasma, CSF, and brain findings, viral processes were the most enriched biological process at both time points (PANTHER; Fig. 4d; Supplementary Table 9). Pathways evident at four weeks were largely exacerbated at eight weeks, as reflected by increased protein representation (Fig. 4e). New eight-week-specific enrichments included pathways related to RNA metabolism, oxidative phosphorylation, lipid metabolism, glycolysis, and stress responses, potentially reflecting changes secondary to accumulating pathology (Fig. 4d; Supplementary Table 9). KEGG immune pathway analysis showed substantial overlap between early and late time points, with progressive intensification over time (Fig. 4f-g; Supplementary Table 9). Early pathways reflected innate immune sensing and antigen presentation, suggesting early APOE ε4 microglial priming. By eight weeks, pathways were reflective of chronic innate activation, phagocytosis, and adaptive immune responses, consistent with a transition to a disease-associated microglial (DAM) phenotype in response to pathology (Fig. 4f; Supplementary Table 9). Reactome enrichment further supported this finding. Early time points were dominated by NF-κB, type I interferon, and MHC class I antigen presentation while late time points showed increased pathways for phagocytosis, cytokine signaling, and potential astrocyte involvement via TGF-β signaling (Fig. 4h; Supplementary Table 9). Pathways common to both time points showed progressive activation over time, as evidenced by increased numbers of contributing proteins (Fig. 4i). Collectively, these data demonstrate that APOE ε4 cortical organoids display an early pro-inflammatory, viral-related immune phenotype prior to the onset of AD pathology, which progressively intensifies with maturation and disease development. The observed transition from innate immune priming to sustained activation and phagocytic responses is consistent with the emergence of DAM. These results highlight cortical organoids as a valuable model for dissecting temporal immune mechanisms underlying APOE ε4-driven AD pathogenesis. Cross-tissue proteomic analysis reveals systemic as well as central- and peripheral-specific dysregulated pathways in APOE ε4 carriers Thus far, we independently characterized the proteomes of plasma, CSF, two brain regions, and cortical organoids from APOE ε4 carriers and non-carriers. While broad enrichment of pro-inflammatory immune pathways was observed across compartments, it remained unclear which APOE ε4-associated molecular changes were systemic versus compartment-specific. To address this, we performed a cross-tissue comparison of enriched KEGG and Reactome immune pathways to delineate molecular processes that are systemic, CNS-specific, or peripheral in nature. Systemic APOE ε4 pathways were broadly representative of conserved innate and adaptive immune responses to diverse viral and bacterial infections and pro-inflammatory stimuli, as well as generalized cytokine signaling and antigen presentation (Fig. 5a). Enrichment was also observed for pattern recognition receptor signaling, interferon responses, and MHC class I antigen presentation, core elements of viral sensing, antiviral immunity, and systemic immune activation (Fig. 5a). In contrast, CNS-specific pathways were associated with more complex immune functions, including adaptive immune signaling through T and B cells, intracellular pathogen responses, chronic inflammatory responses via TNF, NF-κB, and JAK-STAT pathways, and immune modulation (Fig. 5b). Enrichment of pathways involved in immune-microbe interactions at tissue barriers and immune cell recruitment was also observed, potentially reflecting infiltrating adaptive immune cells or blood-brain barrier (BBB)-associated immune processes (Fig. 5b). Notably, five central APOE ε4-enriched pathways were absent in cortical organoids at both time points (Fig. 5b), suggesting that these pathways, particularly those requiring mature T/B cell interactions, antigen presentation, and immune cell crosstalk, cannot be fully recapitulated in organoids due to the absence of adaptive, peripherally derived immune cells. Only four peripheral-only pathways were identified in plasma, primarily involving the fine-tuning and regulation of interferon-driven transcriptional responses (Fig. 5c). These results demonstrate that the APOE ε4 pro-inflammatory immune phenotype includes both systemic and CNS-specific components, with core antiviral and innate immune pathways shared across tissues and more complex adaptive immune and BBB-associated processes enriched in the CNS. The absence of certain adaptive pathways in organoids further underscores the importance of immune cell interactions in shaping the APOE ε4 CNS immune environment. Together, these findings highlight a multi-tissue, pro-inflammatory immune phenotype that may contribute to APOE ε4-driven AD risk. Short-term treatment with an anti-inflammatory ketogenic diet can modulate the pro-inflammatory phenotype of APOE ε4 carriers APOE ε4 carriers have a pro-inflammatory phenotype that spans both the periphery and CNS. To test whether this phenotype is modifiable, we analyzed plasma samples from a 12-week trial of a medical ketogenic diet in 15 individuals with mild AD ( n = 9 APOE ε4 carriers, n = 6 non-carriers; Supplementary Table 10). Medical ketogenic diets, focused on increasingly quality fat intake and reducing carbohydrate consumption to promote production of ketone bodies are widely regarded as a gold-standard systemic anti-inflammatory diet 17 . They have demonstrated treatment benefits for epilepsy, general neurological conditions including AD, and cancers 18-21 . The intervention increased serum β-hydroxybutyrate in both groups, confirming ketone body induction (Supplementary Table 10). Plasma proteomic profiling (7,595 proteins via SomaScan 7k) across baseline and study endpoint revealed no significant changes in non-carriers (Fig. 6a). In contrast, APOE ε4 carriers had 264 differentially expressed proteins (161 downregulated, 103 upregulated; adjusted p < 0.05; Fig. 6b; Supplementary Table 11). PANTHER analysis identified viral processes as the top enriched biological process across both up- and downregulated proteins (Fig. 6c; Supplementary Table 12). A KEGG immune pathway analysis revealed that the medical ketogenic diet led to immune remodelling in APOE ε4 carriers rather than broad suppression or activation. Pathways enriched for both up- and downregulated proteins included those involved in innate and adaptive pathways (Fig. 6d). Pathways enriched in upregulated proteins were associated with adaptive immunity, phagocytosis, and chemotaxis, suggesting enhanced immune surveillance. Downregulated proteins, however, mapped to chronic inflammation and metabolic-immune signaling, indicating reduced chronic pro-inflammatory signaling (Fig. 6d; Supplementary Table 12). These findings were mirrored by a Reactome pathway analysis. This showed a bidirectional regulation of TLR and cytokine signaling, with selective upregulation of interferon responses and Fc-mediated phagocytosis, and downregulation of NF-κB, TGF-β, TCR, and IL-1 signaling pathways (Fig. 6e; Supplementary Table 12). Cell type enrichment analysis showed that both up- and downregulated proteins were mapped to T-regs, NK cells, memory CD8 + T cells, and non-Vδ2 γδ T cells, suggesting functional reprogramming within these populations rather than uniform activation or suppression (Fig. 6f,g). Notably, both protein sets were also enriched in white matter, highlighting that the CNS may have similar immune shifts in response to the ketogenic diet. Taken together, these findings indicate that a medical ketogenic diet leads to reprogramming rather than uniform activation or suppression of immune cell populations. Protein expression patterns and pathway analyses suggest a shift in T-regs and NK cells from pro-inflammatory or cytotoxic states toward more regulatory, tissue supportive phenotypes. Similarly, memory CD8 + T cells likely have reduced signatures of chronic activation alongside enhanced effector memory features, consistent with restored adaptive immune competence. Non-Vδ2 γδ T cells also displayed bidirectional modulation, suggesting altered roles at the innate-adaptive interface and reduced IL-17-associated inflammatory signaling. Discussion Our study provides the most comprehensive cross-tissue characterization to date of the pro-inflammatory immune phenotype associated with APOE ε4, integrating large-scale clinical proteomic data with experimental validation in iPSC-derived cortical organoids and a dietary intervention trial. We demonstrate that APOE ε4 carriers exhibit a conserved, allele dose-dependent pro-inflammatory immune signature across plasma, CSF, and brain tissue, independent of AD status. Importantly, this immune phenotype was also recapitulated in cortical organoids, emerging prior to the development of amyloid-β or tau pathology. This indicates that key features of the APOE ε4 immune signature arise early in disease pathogenesis and reflect an intrinsic, genotype-driven mechanism of immune dysregulation. Cross-tissue analyses further revealed a core set of systemic immune pathways involving broad innate and adaptive responses to infection, alongside distinct tissue-specific signatures, including inflammatory signaling and BBB-related interactions within the CNS. Notably, we show that this immune phenotype is not fixed. Short-term treatment with a medical ketogenic diet in APOE ε4 carriers with AD partially reversed pro-inflammatory signatures, promoting more regulatory and tissue-supportive immune states. Together, these findings illuminate how peripheral and central tissue environments modulate APOE ε4-driven immune responses, provide new mechanistic insight into AD pathogenesis, and support the development of precision biomarkers and immunomodulatory interventions targeting early-stage disease processes. Our results are consistent with and extend prior work. Multiple studies have reported heightened peripheral immune responses in APOE ε4 carriers, including increased cytokine production following TLR2/4/5 activation 3 , elevated TNF-α and IL-6 levels after a lipopolysaccharide (LPS) challenge 4 , increased cytokine levels 6,22-24 , and enhanced susceptibility to systemic inflammatory stressors such as sepsis 4 . Similarly, transcriptomic and epigenomic profiling of PBMCs has demonstrated broad dysregulation of innate immune pathways, altered NF-κB activation, increased chromatin accessibility in CD14 + monocytes, and clonally expanded CD8 + effector memory T cells 25 . Our plasma APOE ε4 proteomic signature mirrors these findings, with robust enrichment for innate immune and antiviral responses, including TLR-NF-κB and interferon signaling cascades. Moreover, we observed allele dose-dependent effects, consistent with prior reports of exacerbated immune dysfunction in homozygous APOE ε4 carriers 3,25 . Notably, our findings also align with the emerging hypothesis that APOE ε4 carriers exhibit increased susceptibility to viral reactivation and higher viral titers and that this plays a role in the development of AD 26 . The strong enrichment of viral response pathways we find across plasma, CSF, and brain in our study supports this hypothesis. Prior studies measuring C-reactive protein (CRP) levels in APOE ε4 carriers have yielded mixed results, with some reporting decreased baseline CRP 22,27-29 , while others report increased associated brain atrophy 30 . Our findings suggest that APOE ε4-driven chronic immune dysregulation is not adequately captured by individual markers of acute inflammation such as CRP. A limitation of our study is the lack of longitudinal data. Although we show that the APOE ε4 immune phenotype is present in cognitively unimpaired individuals, future work should investigate whether this signature evolves with disease progression in APOE ε4 carriers. In the CNS, our findings offer new mechanistic insight into how APOE ε4 may prime resident immune cells, particularly microglia, toward a disease-associated phenotype. Prior studies have demonstrated that APOE ε4 microglia exhibit impaired phagocytosis 11 , disrupted autophagy with lipid droplet accumulation 9,10 , and exaggerated inflammatory responses to amyloid-β 10,11 . Our CSF and brain proteomic data confirm upregulation of inflammatory pathways (NF-κB, TNF signaling), consistent with a chronically activated or DAM-like state and reveal enrichment of BBB-related immune pathways. These findings align with evidence of APOE ε4-associated BBB dysfunction 31,32 and increased infiltration of peripheral immune cells such as IL-17 + neutrophils 7 . Notably, we observed stronger pro-inflammatory enrichment in the STG compared to the dlPFC. This likely reflects the greater vascularization and CSF-blood interface in the STG, which may facilitate immune interactions. A major strength of our study is the use of patient-derived cortical organoids to model APOE ε4-driven immune changes in human neural tissue before AD pathology emerges. While prior studies of APOE ε4 immunopathology have largely focused on peripheral immune cells 3-7,22-25,30,33,34 , postmortem brain tissue 8,35 , or single iPSC-derived cell types 9-11 , our organoid model enables longitudinal study of APOE ε4 effects within a more complex network. We experimentally confirmed early activation of inflammatory pathways, including NF-κB and DAM-like signatures, prior to amyloid-β or tau pathology. This provides direct evidence that APOE ε4 immune dysregulation is intrinsic and genotype-driven, rather than a secondary response to existing AD pathology. Moreover, the fact that APOE ε4 immune activation precedes amyloid-β and tau accumulation supports the view that these hallmark pathologies may be downstream consequences rather than primary disease drivers in APOE ε4 carriers. These insights challenge current therapeutic paradigms focused solely on amyloid-β and tau, targets that, despite effective clearance, have shown limited clinical benefit and in some cases worsening disease 36-39 . Instead, our findings suggest that tempering the persistent APOE ε4-driven immune response through immunomodulation or anti-inflammatory approaches will be critical for altering disease risk and progression in this high-risk population. Importantly, this aligns with emerging evidence that vaccinations reduce AD risk 40,41 , including the shingles vaccines Zostavax 42 and Shingrix 43 , potentially by reducing reactivation of varicella zoster virus and/or inducing a virus-specific immunomodulatory effect 42 . Targeting immune dysregulation should therefore be prioritized in future precision medicine strategies for AD. Our finding that a medical ketogenic diet partially reversed the APOE ε4 immune phenotype has important implications for therapeutic development. Twelve weeks of treatment led to the downregulation of chronic inflammatory signaling pathways, including NF-κB, IL-1, and TGF-β and upregulation of regulatory and phagocytic functions across T-regs, NK cells, and memory CD8 + T cells. These changes suggest a shift away from maladaptive, chronic inflammation toward a more balanced and functional immune state. This supports the idea that immune modulation may be a viable disease-modifying strategy, particularly if implemented early, before irreversible pathology has accumulated. Future studies would benefit from larger-scale clinical trials of a medical ketogenic diet as well as other immunomodulatory therapies before the onset of AD in APOE ε4 carriers. One limitation of our study is that cross-platform differences limited direct comparisons of APOE ε4 tissue proteomes. Plasma and CSF were profiled using SomaScan aptamer-based technology, while TMT and label-free mass spectrometry were used for brain and organoid analyses, respectively. As a result, we focused on pathway-level rather than protein-level comparisons across samples. However, this limitation is also a strength. The consistency of pathway-level findings across multiple platforms provides orthogonal validation of the robustness and generalizability of our results. Moreover, our study demonstrates that aptamer-based approaches, despite known limitations in proteoform sensitivity 44 , can reliably capture APOE ε4-related pathway alterations. Future work using harmonized proteomic platforms will further refine these insights. In conclusion, our findings establish that APOE ε4 drives a conserved pro-inflammatory immune phenotype that emerges early, prior to the development of hallmark AD pathology. These results advance the mechanistic understanding of APOE ε4-mediated AD risk and challenge the prevailing amyloid-β- and tau-centric therapeutic approaches, especially for APOE ε4 carriers. By demonstrating that immune dysregulation is an intrinsic, upstream feature of APOE ε4, and showing that it is modifiable, our study underscores the need to prioritize immunomodulatory and anti-inflammatory strategies in precision medicine approaches for this high-risk group. Moving forward, longitudinal studies and harmonized proteomic platforms will be critical to further delineate the temporal evolution of the APOE ε4 immune signature and to guide the development of targeted interventions aimed at altering disease trajectories in APOE ε4 carriers. Methods Participants Global Neurodegeneration Proteomics Consortium (GNPC) cohort. The GNPC cohort is the largest collection of proteomic data for individuals with neurodegenerative diseases and non-impaired controls from >20 clinical sites from across the USA, UK, and Europe. In the current study, we included non-impaired controls (n = 6,672) and individuals with AD (n = 3,455). Across both groups, n = 6,107 were non- APOE ε4 carriers and n = 4,767 had at least one APOE ε4 allele (Supplementary Table 1). AD was clinically diagnosed within each study site as previously described including Clinical Dementia Rating (CDR) score of > 1, Mini-Mental State Exam (MMSE) < 24, and/or Montreal Cognitive Assessment (MOCA) score of < 23 45 . Individual plasma and CSF samples, demographic, and clinical variables were collected at a single timepoint (Supplementary Table 1). Participants from each study site in the GNPC cohort provided written informed consent and studies were approved by the relevant institution’s ethics committee 45 . Accelerating Medicines in AD Partnership (AMP-AD) Diverse Cohorts study. The AMP-AD Diverse Cohorts study is a cross-consortium project that created harmonized high-throughput multi-omic data for post-mortem brain tissue samples. AD cases were pathologically defined as described elsewhere 46 . Samples from the dorsolateral prefrontal cortex (dlPFC) and superior temporal gyrus (STG) were collected across four institutions: Mayo Clinic, Rush University, Mount Sinai University Hospital, and Emory University. In the present study, we included dlPFC data from n = 190 non-impaired controls and n = 530 AD cases. For the STG, n = 49 non-impaired controls and n = 161 AD cases were included. Supplementary Table 6 lists the demographic characteristics for the included donors. Participants from each study site in the AMP-AD Diverse Cohorts study provided written informed consent and studies were approved by the relevant institution’s ethics committee. University of Kansas Alzheimer’s Disease Research Center donors for cortical organoids. Two donors from the University of Kansas Alzheimer’s Disease Research Center provided post-mortem skin samples. The non-carrier donor was a 72-year-old female with an APOE ε3/ε2 genotype and APOE ε4 carrier donor was an 84-year-old female with an APOE ε3/ε4 genotype. Therapeutic Diets in Alzheimer’s Disease (TDAD) clinical trial. The TDAD clinical trial is a single-blind, randomized, controlled study of the effects of a 12-week medical ketogenic dietary intervention in patients with mild AD. Participants were recruited through the Clinical Translational Science Unit at the University of Kansas Medical Center Clinical Research Center. Fifteen participants ( n = 9 APOE ε4 carriers and n = 6 non-carriers) were given a ketogenic diet intervention. All participants met the criteria for probable AD based on the National Institutes of Aging-Alzheimer’s Association (NIA-AA) criteria 47 and had a baseline CDR of 0.5 to 1 (see Supplementary Table 10 for demographics). Participants with moderate to severe AD, including those residing in a nursing home or dementia special care unit, were excluded from the trial based on our previous experience showing that these individuals have a low retention rate to dietary interventions 21 . Participants with serious medical conditions, those participating in another clinical trial, women of child-bearing capacity who seek to become pregnant, and non-English speakers were also excluded. Table 1 lists the inclusion and exclusion criteria for the TDAD clinical trial. Methods Table 1. Therapeutic Diets in Alzheimer’s Disease (TDAD) clinical trial inclusion and exclusion criteria. Inclusion Criteria 1. Diagnosis of AD by NIA-AA criteria 47 2. CDR global score of 0.5 or 1 3. Agreed cooperation from an appropriate study partner 4. Speaks English as a primary language 5. Age 50 to 90 6. No medication changes within the past 30 days Exclusion Criteria 1. Resides in a nursing home or dementia special care unit, or cannot control diet 2. A potentially confounding serious medical risk including insulin-requiring diabetes, cancer requiring chemotherapy or radiation within the past five years, or recent cardiac event (e.g. heart attack, angioplasty, etc.) 3. Participating in another clinical trial or using an investigational drug or therapy within 30 days of the screening visit 4. A history of renal stones 5. Women with child-bearing capacity who seek to become pregnant 6. Non-English speakers Ketogenic diets focus on increasing quality fat intake and reducing carbohydrate consumption to promote the production of ketone bodies, mimicking the effects of prolonged fasting and vigorous exercise 17 . Macronutrients were maintained at approximately 70% fat, 20% protein, and < 10% carbohydrates. A 1:1 emulsified medium chain triglyceride (MCT) oil was also provided to participants in the ketogenic diet group to increase fat intake, enhance ketosis, and increase palatability in line with other medical ketogenic diets 21 . To reduce gastrointestinal side effects from the MCT, the supplement dose increased on a weekly basis across a four-week period, starting from an intake level corresponding to 10% of total fat energy and gradually increasing to 40%. The Mifflin-St. Jeor equation was used to calculate target energy goals 48 . All participants received weekly dietary counselling from the study dietitian. Participants also received dietitian-created ketogenic diet educational material and a manual. Manuals included sample menus, recipes, and strategies for maintaining their respective diet during travel or holidays. Participants’ APOE genotype, medical history, medication use, comorbidities, smoking, and alcohol intake were recorded at their initial screening visit. Blood samples were taken at two time points, a pre-intervention baseline and post-intervention (12 weeks), for plasma proteomics and to measure depth of ketosis using a routine serum β-hydroxybutyrate assay performed commercially by Quest Diagnostics (Lenexa, KS). Participant demographic characteristics are listed in Supplementary Table 10. The TDAD clinical trial was approved by the University of Kansas Center IRB (STUDY00143457). Prior to participating in the study, all participants provided informed consent. The trial was registered on ClinicalTrials.gov under registration number NCT03860792. Cortical Organoids Induced pluripotent stem cell derivation. Induced pluripotent stem cells (iPSCs) were generated from fibroblast samples from both donors using a CytoTune-iPSC 2.0 Sendai Reprogramming kit (ThermoFisher Scientific, A16517) according to the manufacturer’s instructions. iPSCs were plated down into 6-well plates pre-treated with 0.08mg/ml Matrigel (Corning, 356234) in Dulbecco’s Phosphate-Buffered Saline (DPBS). StemFlex medium (Gibco, A3349401) with 100U/ml penicillin-streptomycin added (Gibco, 15140122) was added to each well and changed every two days. iPSCs were passaged once a week using ReLeSR (StemCell Technologies, 15140122) as per the manufacturer’s protocol. Following each passage, medium was supplemented with 0.01mM Y-27632 (StemCell Technologies, 72302) for 24 hours. iPSC-derived microglia differentiation. iPSCs were differentiated into hematopoietic progenitor cells (HPCs) using a STEMdiff Hematopoietic kit (StemCell Technologies, 5310) as per the manufacturer’s instructions. First, iPSCs were passaged into HPC Medium A in 24-well plates pre-treated with 0.04mg/ml Matrigel. After three days, they were cultured with HPC Medium B for nine more days. Mature HPCs were then collected and passaged into STEMDiff Microglia Differentiation medium (StemCell Technologies, 100-0019) as per the manufacturer’s instructions. Fresh media was added every other day for 24 days and cells were passaged once on day 12. After differentiation, microglia were matured in BrainPhys medium with N2-A and SM1 supplements (StemCell Technologies, 05793) for four days prior to integration into embryoid bodies. Cortical organoid development. iPSCs were first passaged three times to stabilize the line. Embryoid bodies were generated in ultra-low attachment dishes (Corning, 3261) using StemFlex medium supplemented with 100U/ml penicillin-streptomycin and 0.01mM y-27632 ROCK inhibitor. Media was changed every two days. Embryoid body medium was replaced into N2/B27 medium (Methods Table 2) and transferred into a 24-well plate. Mature microglia were added to each well at a concentration of 0.8 x 10 5 cells per well. Fresh N2/B27 media was added every two days and organoids were passaged once every two weeks. Cortical organoids were aged for either four or eight weeks. Methods Table 2. List of N2/B27 media components. Reagent Concentration (in 1000ml) Source DMEM/F12 478ml Gibco, 11320033 Neurobasal medium 480ml Gibco, 21103049 N2 supplement 5ml Gibco, 17502048 Human insulin 2.5ug/ml Sigma-Aldrich, I9278-5ML MEM NEAA 5ml Gibco, 11140050 Beta-mercaptoethanol 25mM Gibco, 21985023 B27 (without vitamin A) 10ml Gibco, 12587010 GlutaMAX supplement 10ml Gibco, 35050061 Penicillin-streptomycin 100u/ml Gibco, 15140122 Dosomorphin 1uM StemCell Technologies, 72102 SB431542 10uM StemCell Technologies, 72234 CHIR99021 3uM StemCell Technologies, 72054 Thiazovivin 2uM StemCell Technologies, 72254 Immunofluorescent staining. At each timepoint, cortical organoids were removed from the media and washed in DPBS three times for 10 minutes. They were then fixed in 4% paraformaldehyde (Sigma-Aldrich, 158127) for 20 minutes, washed in DPBS once for 10 minutes, and stored in cyroprotectant (30% glycerol (Sigma-Aldrich, G5516), 30% ethylene glycol (Sigma-Aldrich, 324558), 40% PBS) at -20 o C. Organoids were washed in PBS three times for 10 minutes and incubated with blocking solution (5% donkey serum (Sigma-Aldrich, D9663) or 5% bovine serum albumin (BSA), 0.1% Triton X-100 (Sigma-Aldrich, #X100), and PBS) overnight at 4 o C. Organoids were then incubated with primary antibodies (Methods Table 3) made up in blocking solution for two days at 4 o C. They were then washed three times with PBS for 10 minutes followed by an overnight incubation with secondary antibodies (Methods Table 3) made up in blocking solution at 4 o C. This process was repeated for each primary antibody. The MAP2, GFAP, and IBA1 stained organoids (Extended Data Fig. 2) were co-stained for DAPI using NucBlue Live ReadyProbes (Invitrogen, R37605) as per the manufacturer’s instructions and incubated for three days at 4 o C. The amyloid-β and p-tau217 organoids (Fig. 3) were co-stained for DAPI using ProLong Gold AntiFade with DAPI (Invitrogen, P36931). Methods Table 3. List of primary and secondary antibodies. Antibody Concentration Source Rabbit anti-MAP2 1:250 Invitrogen, PA5-17646 Goat anti-IBA1 1:500 Abcam, ab5076 Rabbit anti- GFAP 1:250 Invitrogen, PA5-16291 Rabbit anti- pTau217 1:200 Invitrogen, 44-744 Mouse anti-APP6E10 5 μg/ml BioLegend, SIG-39320 Donkey anti-goat Cy3 1:100 Jackson Immunoresearch, 705-165-147 Donkey anti-rabbit AlexaFluor 488 1:100 Invitrogen, A21206 Donkey anti-rabbit AlexaFluor 647 1:100 Invitrogen, A31573 Donkey anti-rabbit Cy3 1:100 Jackson Immunoresearch, 711-165-152 Donkey anti-mouse AlexaFluor 555 1:100 Invitrogen, A31570 Prior to imaging, stained cortical organoids were washed three times in PBS for 10 minutes and incubated with CytoVista 3D Cell Culture clearing reagent (Invitrogen, V11315) overnight at 4 o C. To visualize and quantify amyloid-β and p-tau217 staining, whole cortical organoids were imaged using a Nikon TI2-E inverted microscope attached to a Yokagawa CSU-W1 spinning disk confocal with SoRa super-resolution. Images were acquired with a Hamamatsu Fusion BT camera using Nikon NIS-Elements Advanced Research software. To visualize and quantify MAP2, IBA1, and GFAP staining, whole cortical organoids were imaged using a Marianas 3i Spinning Disk Confocal microscope with Super-Resolution by Optical Re-Assignment (SoRa). Representative images were taken using Fiji ImageJ v1.54p 49 . ELISA for secreted Aβ 42 /Aβ 40 and p-tau217. Media was collected from cortical organoids at four and eight weeks of age and proteins were extracted using acetone precipitation. Briefly, 1ml of ice-cold acetone (cooled to -20 o C) was added to 0.5ml of collected media in a tube. Tubes were vortexed, incubated for 60 minutes at -20 o C, and then centrifuged for 10 minutes at 15,000 x g. The supernatant was removed and pellets were washed in 70% ethanol before being resuspended in 8M urea. Secreted human amyloid-β 40 and amyloid-β 42 were measured with an ELISA as per the manufacturer’s instructions (ThermoFisher, KHB3481 and KHB3544). Secreted human p-tau217 was also measured with an ELISA as per the manufacturer’s instructions (Cell Signaling Technology, 59672). All samples were diluted 1:5 and concentrations were normalized to protein content measured via Pierce BCA protein assay (ThermoFisher, 23225). Graphs were made in GraphPad Prism v10.5.0. Proteomics GNPC cohort SomaScan assay. Proteomics of plasma and CSF samples from the GNPC cohort was done using the SomaScan v4.1 assay (SomaLogic). The SomaScan assay detects approximately 7,000 proteins using aptamer-based technology called slow off-rate modified aptamers (SOMAmers). These contain chemically modified nucelotides that bind with high specificity and affinity to target proteins 50,51 . Raw data is provided by SomaLogic following standardization, normalization, and calibration, including adaptive normalization by maximum likelihood (ANML). Protein measurements are provided in relative fluorescent units (RFU). Prior to their inclusion in the GNPC dataset, aptamers are mapped to Uniprot. Details on the creation and harmonization of the GNPC dataset are described elsewhere 45 . Data from the GNPC dataset was log2 transformed and standardized prior to analyses in the current study and was done separately on training and testing datasets. AMP-AD Diverse Cohorts study TMT quantitation. Proteomics on post-mortem dlPFC and STG homogenates was done using TMT mass spectrometry as previously described 46 . Using a conservative approach, we only included proteins with < 30% missing values across samples and imputed these values using the median. Batch effects were accounted for by fitting a linear regression model, as in the original study 46 . Cortical organoids label free mass spectrometry. At 4 and 8 weeks of age, cortical organoids were snap frozen in liquid nitrogen and stored at -20 o C. To prepare samples for mass spectrometry, 0.2% n-dodecyl-β-D-maltoside (DDM) in 50 mM triethylammonium bicarbonate (TEAB) with 5 mM Tris(2-carboxyethyl)phosphine (TCEP) was added to each organoid. A pestle suitable for a 0.5 mL tube attached to a drill was used to homogenize the organoids for 10 s. The tubes were incubated at 85 °C for 10 min, then cooled in incubated with 10 mM iodoacetamide for 30 min at 22 °C. The samples were frozen in dry ice and then lyophilized. To each sample was added 10 µL of 0.1% DDM in 50 mM TEAB with 2.5 mM TCEP, 0.5 ug of LysC (WAKO Fujifilm) and 1 ug of trypsin (Sigma, trypzean). The samples were incubated at 42 °C for 2 h, then a further 0.5 ug of trypsin was added and incubated at 33 °C for 4 h. Each sample was acidified with 0.2 µL of formic acid and diluted with 30 µL of 0.1% trifluoroacetic acid and then desalted using the STAGEtip method 52 . Each sample was analyzed by LC-MS/MS using the Vanquish Neo system and Astral Orbitrap mass spectrometer (ThermoFisher Scientific). Samples were loaded onto a Pepmap Neo C18 5 µm particle 5 mm long by 300 µm inside diameter trap column (ThermoFisher Scientific) at up to 10 µL/min and at a maximum pressure of 800 bar. They were then eluted through a 15 cm long 150 µm inside diameter Pepmap C18 EASY-spray column with 2 μm 100 Å particles (ThermoFisher Scientific). The column was heated to 40 °C using the EASY-spray ion source operating a 1.9 kV. The S lens radio frequency level was 40 and capillary temperature was 280 °C. The liquid chromatography used buffer A (solution of 0.1% formic acid) and buffer B (0.1% formic acid and 99.9% acetonitrile). After loading the sample in 3.6% buffer A, the gradient at 1 µL/min was from 7.2% to 25.2% buffer B in 19.7 min, to 31.5% buffer B in 3.7 min, to 49.5% buffer B in 0.4 min and to 90% buffer B at 3 µL/min in 0.5 min and held for 0.7 min. MS acquisition was for 26 min. The MS scans in the orbitrap analyzer were at a resolution of 240,000 with automatic gain control set to 5,000,000 and a maximum ion time of 3 ms for m/z 380 to 980. The data-independent acquisition MS/MS scans with a window of 2 m/z in the Astral analyser had automatic gain control at 50,000 for a maximum ion time of 3 ms. The loop was controlled to 0.6 seconds. The MS/MS scan range was 150-2000 m/z. The normalized collision energy was 25. The raw LC-MS/MS data was processed with DIA-NN v1.9.2 53 . The Homo Sapiens reference proteome downloaded Feb 24 2025 with 20,644 genes, using canonical sequences only, was used to create an in silico library for peptide-spectrum matching. N-terminal methionine excision was allowed. Carbamidomethyl (C) was a fixed modification. Peptide length was 7-30. Initial mass accuracy was 10 ppm and MS1 accuracy was 4 pm. Digestion was set to trypsin/P with a maximum of 1 missed cleavage. Precursor false discovery rate was 1%. Match between runs was disabled. Heuristic protein inference was enabled, as were all other default algorithm settings for DIA-NN v1.9.2. TDAD clinical trial SomaScan assay. Proteomics of plasma samples before and after the dietary intervention were done using the SomaScan v4.1 assay (SomaLogic) that detects approximately 7,000 proteins. Raw data was provided by SomaLogic following standardization, normalization, and calibration, including adaptive normalization by maximum likelihood (ANML), and mapped to UniProt. Protein measurements are provided in relative fluorescent units (RFU). Data was log2 transformed and standardized prior to analyses in the current study. Statistical analyses Feature selection. In the plasma, CSF, dlPFC, and STG, APOE4 proteins were identified using mutual information 54 as previously reported 1,2 . In plasma and CSF, we identified APOE ε4-specific proteins using all APOE ε4 carriers and non-carriers to validate our previous findings where APOE ε4 proteins were identified only in non-impaired controls 2 . In the dlPFC and STG, we identified APOE ε4 proteins in non-impaired controls. Proteins with a mutual information value >0.1 were selected for machine learning analyses. We confirmed our feature selection method using principal component analysis (PCA). For plasma and CSF, we also performed a second feature selection using mutual information on APOE ε4 carriers with AD relative to non-impaired control APOE ε4 carriers. Mutual information was calculated in R (v4.4.1) using the package 'FSelectorRcpp' 55 and PCA plots were made using ‘ggplot2’ 56 . Machine learning. We used classification and regression trees (CART) and random forest to test the predictive performance of our identified APOE ε4 proteins for differentiating between APOE ε4 carriers and non-carriers. The dataset was split into a 70% training and validation set and a 30% withheld (unseen) testing set. Model training and evaluation were done using a 5-fold cross-validation procedure repeated 10 times. Machine learning was done in R (v4.4.1) using the package ‘caret’. Cortical organoid proteomics. Label-free mass spectrometry proteomic data was analyzed in R (v4.4.1) using ‘limma’ package. The raw data consisted of protein group intensities across biological samples, with sample names encoding both experimental condition ( APOE ε4 carrier or non-carrier) and time point (four weeks or eight weeks). The protein intensity matrix was filtered to retain proteins with valid quantification in at least 30% of samples. Missing values were imputed per protein by replacing missing entries with the minimum observed intensity for that protein, assuming missingness was due to low abundance below the detection limit. Protein intensities were log2-transformed to stabilize variance. Differential protein abundance between APOE ε4 and non-carrier cortical organoid samples was assessed at both four week and eight-week time points using a linear modeling framework with empirical Bayes moderation, implemented in limma. Specific contrasts ( APOE ε4 vs non-carrier at each time point) were tested. Resulting p -values were corrected for multiple testing using the Benjamini-Hochberg method to control the false discovery rate (FDR). Proteins with adjusted p-values (FDR) below 0.05 were considered significantly differentially abundant. Volcano plots were constructed using -log 10 -transformed adjusted p-values on the y-axis. Plots show significance thresholds that were applied at FDR-adjusted p < 0.05. All plots and downstream analyses were performed in R using the ‘ggplot2’ package. Cortical organoid AD pathology. ELISA data for secreted amyloid-β42/40 and p-tau217 was analyzed at each time point using a two-tailed unpaired t-test. To analyze interactions between genotype and time, a two-way ANOVA was used. All statistical analyses were performed in GraphPad Prism (v10.5.0). TDAD clinical trial proteomics. Log2 transformed SomaScan assay data was analyzed in R (v4.4.1) using ‘limma’ package. Differential protein abundance between baseline and post-study in APOE ε4 carriers and non-carriers was assessed using a linear modeling framework with empirical Bayes moderation, implemented in limma. Resulting p -values were corrected for multiple testing using the Benjamini-Hochberg method to control the false discovery rate (FDR). Proteins with adjusted p-values (FDR) below 0.05 were considered significantly differentially abundant. Volcano plots were constructed using -log 10 -transformed adjusted p-values on the y-axis. Plots show significance thresholds that were applied at FDR-adjusted p < 0.05. All plots and downstream analyses were performed in R using the ‘ggplot2’ package. Enrichment analysis. Enriched biological functions and pathways across APOE4 proteins were assessed using NetworkAnalyst (v3.0) 57-59 . Protein-protein interactions were identified using a first order network in the International Molecular Exchange Consortium (IMEx) interactome database 60 and InnateDB 61 . Network enrichments for biological processes and pathways were done using the PANTHER classification system 62 and Kyoto Encyclopedia of Genes and Genomes (KEGG) database 63 , respectively. Statistical significance of the enriched networks was determined by a false discovery rate (FDR) of > 0.05. Proteins enriched in the KEGG AD pathway were visualized using KEGG mapper 64 . Cell type-specific enrichment analyses for immune cells, brain regions, and white matter cells were done using single-cell RNA sequencing data from the Human Protein Atlas (v23, Ensembl v109) 12-14 . Here, the corresponding protein-coding transcripts per million for each APOE ε4 protein was identified. Expression for each cell type was normalized using min-max scaling. Heatmaps were generated to visually represent these enrichments using R (v4.4.1). Declarations Data availability The harmonized GNPC data used to generate these findings was provided to Consortium Members in June 2024 and will be made available for public request by the AD Data Initiative by July 1, 2025. Members of the global research community will be able to access the metadata and place a data use request via the AD Discovery Portal (https://discover.alzheimersdata.org/). Access is contingent on adherence to the GNPC Data Use Agreement and the Publication Policies. The AMP-AD Diverse Cohorts study data is available through the AD Knowledge Portal (https://adknowledgeportal.synapse.org/). Researchers who wish to access this controlled dataset are required to submit a Data Use Agreement. More information can be found here: https://adknowledgeportal.synapse.org/Data%20Access. Code Availability All code used in this study are publicly available at https://github.com/Art83/AD_apoe. Acknowledgements The authors are grateful to the cohort contributors, patients, donors, and families who helped make this research possible. This work was supported by the Australian Government’s Medical Research Future Fund MRF2040081 (C.A.F, A-N.C, A.S.); Dementia Australia Research Foundation Bondi2BlueMtns Project Grant (C.A.F., A-N.C., H.M.W., R.H.S.); Neil & Norma Hill Foundation (C.A.F.); Annemarie & Arturo Gandioli-Fumagali Foundation (C.A.F.); Perpetual Foundation – John Williams Endowment (C.A.F.); Hillcrest Foundation (C.A.F.); John & Anne Leece Family (C.A.F.); Paul & Valeria Ainsworth Precision Medicine Research Fellowship (C.A.F.); University of Kansas Alzheimer’s Disease Developmental Projects Grant (C.Sl., C.A.F., A.S.); the NIH P30AG072973 (H.M.W., J.M.B., C.Sl., R.H.S.), R01AG064227 (C.Sl.), R21TR003589 (H.M.W.), R01AG07816 (H.M.W.), U19AG068054 (H.M.W.), R01AG060733 (J.E.K., M.K.T., J.D.M., D.K.S., J.M.B., R.H.S.); Alzheimer’s Association 23AARG-1023 (H.M.W.); Sydney Horizon Fellowship (A-N.C.); and National Health and Medical Research Council 2025529 (J.H.R.). The results published here are in whole or in part based on data obtained from the AD Knowledge Portal (https://adknowledgeportal.org/). Data generation was supported by the following NIH grants: U01AG046139, U01AG046170, U01AG061357, U01AG061356, U01AG061359, and R01AG067025. We thank the participants of participants of the Religious Order Study, Memory and Aging Project, the Minority Aging Research Study, Rush Alzheimer’s Disease Research Center, Mount Sinai/JJ Peters VA Medical Center NIH Brain and Tissue Repository, National Institute of Mental Health Human Brain Collection Core (NIMH HBCC), Mayo Clinic Brain Bank, Sun Health Research Institute Brain and Body Donation Program, Goizueta Alzheimer’s Disease Research Center, New York Brain Bank at Columbia University, New York Genome Center and the Biggs Institute Brain Bank for their generous donations. Data and analysis contributing investigators include Nilüfer Ertekin-Taner, Minerva Carrasquillo, Mariet Allen (Mayo Clinic, Jacksonville, FL), David Bennett, Lisa Barnes (Rush University), Philip De Jager, Vilas Menon (Columbia University), Bin Zhang, Vahram Haroutanian (Icahn School of Medicine at Mount Sinai), Allan Levey, Nick Seyfried (Emory University), Rima Kaddurah-Daouk (Duke University), Steve Finkbeiner (University of California-San Francisco/Gladstone Institutes), Daifeng Wang (University of Wisconsin-Madison), Stefano Marenco (NIMH HBCC), Anna Greenwood, Abby Vander Linden, Laura Heath, William Poehlman (Sage Bionetworks). Confocal microscopy was performed at the University of Kansas Medical Center supported by NIH S10 OD 032207 and the University of Sydney’s Australian Centre for Microscopy & Microanalysis. The funders of this work played no role in the design of the study, the running of experiments and analyses, the interpretation of the results, and the writing of the manuscript. Author contributions A.S. and C.A.F. conceptualized the study and led the study design. S.T., M.E.G., B.H., C.Sm., C.T., A-N.C., H.M.W. and C.A.F. performed the cortical organoid experiments. J.H.R. provided key immunological insights. J.M.B. and R.H.S. provided clinical neurological insights. V.K. and F.B.I. provided assistance and insight for the GNPC dataset. M.E.G. and C.Sl. provided proteomic insights. J.E.K., M.K.T., J.D.M., D.K.S., J.M.B., and R.H.S. performed the TDAD pilot clinical trial. A.S., S.T., and C.A.F. produced the figures. C.A.F. wrote the paper and supervised the study. All authors revised the paper for intellectual content and read and approved the final version. Competing interests The authors declare no competing interests. References Shvetcov, A. et al. Proteome profiling of cerebrospinal fluid using machine learning shows a unique protein signature associated with APOE genotype. Aging Cell (2024). Shvetcov, A. et al. APOE ε4 carriers share immune-related proteomic changes across neurodegenerative diseases. Nature Medicine (2025). Gale, S. C. et al. APOε4 is associated with enhanced in vivo innate immune responses in humans. Journal of Allergy and Clinical Immunology 134 , 127-134 (2015). Shao, Y. et al. Presence of the apolipoprotein E-ε4 allele is associated with an increased risk of sepsis progression. Scientific Reports 10 (2020). Bonacina, F. et al. Myeloid apolipoprotein E controls dendritic cell antigen presentation and T cell activation. Nature Communications 9 (2018). Zhang, Y.-J. et al. APOE ε4-associated downregulation of the IL-7/IL-7R pathway in effector memory T cells: Implications for Alzheimer's disease. Alzheimer's & Dementia 20 , 6441-6455 (2024). Rosenzweig, N. et al. Sex-dependent APOE4 neutrophil-microglia interactions drive cognitive impairment in Alzheimer's disease. Nature Medicine 30 , 2990-3003 (2024). Friedberg, J. S. et al. Associations between brain inflammatory profiles and human neuropathology are altered based on apolipoprotein E ε4 genotype. Scientific Reports 10 (2020). Haney, M. S. et al. APOE4/4 is linked to damaging lipid droplets in Alzheimer's disease microglia. Nature 628 , 154-161 (2024). Hellen, M. et al. Inflammation-induced lysosomal dysfunction in human iPSC-derived microglia is exacerbated by APOE 4/4 genotype. Journal of Neuroinflammation 22 (2025). Konttinen, H. et al. PSEN1ΔE9, APPswe, and APOE4 confer disparate phenotypes in human iPSC-derived microglia. Stem Cell Reports 13 , 669-683 (2019). Karlsson, M. et al. A single-cell type transcriptomics map of human tissues. Science Advances 7 (2021). Uhlen, M. et al. A genome-wide transcriptomic analysis of protein-coding genes in human blood cells. Science 366 , 6472 (2019). Sjostedt, E. et al. An atlas of protein-coding genes in the human, pig, and mouse brain. Science 367 (2020). Cho, A.-N. et al. Microfluidic device with brain extracellular matrix promotes structural and functional maturation of human brain organoids. Nature Communications 12 (2021). Truong, D.-J. J. et al. Non-invasive and high-throughput investigation of exon-specific isoform expression. Nature Cell Biology 23 , 652-663 (2021). Harman, A. L., Gasior, M., Vining, E. P. & Rogawski, M. A. The neuropharmacology of the ketogenic diet. Pediatric Neurology 36 , 281-292 (2007). Kelement, R. J., Brehm, N. & Sweeney, R. A. Ketogenic diets in medical oncology: A systematic review with focus on clinical outcomes. Medical Oncology 37 (2020). Cervenka, M. C. et al. International recommendations for the management of adults treated with ketogenic diet therapies. Neurology Clinical Practice 11 , 385-397 (2020). Weber, D. D., Aminazdeh-Gohari, S. & Kofler, B. Ketogenic diet in cancer therapy. Aging 10 , 164-165 (2018). Taylor, M. K., Sullivan, D. K., Mahnken, J. D., Burns, J. M. & Swerdlow, R. H. Feasibility and efficacy data from a ketogenic diet intervention in Alzheimer's disease. Alzheimer's & Dementia 4 , 28-36 (2017). Soares, H. D. et al. Plasma biomarkers associated with the apolipoprotein E genotype and Alzheimer disease. Archives of Neurology 69 , 1310-1317 (2012). Fan, Y.-Y. et al. APOE ε4 allele elevates the expressions of inflammatory factors and promotes Alzheimer's diseae progression: A comparative study based on Han and She populations in the Wenzhou area. Brain Research Bulletin 132 , 39-43 (2017). Olgiati, P. et al. APOE epislon-4 allele and cytokine production in Alzheimer's disease. International Journal of Geriatric Psychiatry 25 , 338-344 (2009). Ramakrishnan, A. et al. Epigenetic dysregulation in Alzheimer's disease peripheral immunity. Neuron 112 , 1235-1248 (2024). Wainberg, M. et al. The viral hypothesis: How herpesvirus may contribute to Alzheimer's disease. Molecular Psychiatry 26 , 5476-5480 (2021). Eiriksdottir, G. et al. Apolipoprotein E genotype and statins affect CRP levels through independent and different mechanisms. Atherosclerosis 186 , 222-224 (2006). Garcia, A. R. et al. APOE4 is associated with elevated blood lipids and lower levels of innate immune biomarkers in a tropical Amerindian subsistence population. Elife 10 (2021). Vasunilashorn, S. et al. Apolipoprotein E, biomarkers, and mortality in Taiwanese older adults. Atherosclerosis 219 , 349-354 (2011). Tao, Q. et al. Association of chronic low-grade inflammation with risk of Alzheimer disease in ApoE4 carriers. JAMA Network Open 1 , e183597 (2018). Sweeney, M. D., Montagne, A., Sagare, A. P., Nation, D. A. & Zlokovic, B. V. Broken blood-brain barrier in Alzheimer's disease and other neurodegenerative disorders. Nature Reviews Neurology 15 , 133-150 (2019). Montagne, A. et al. APOE4 leads to blood-brain barrier dysfunction predictive cognitive decline. Nature 581 , 71-76 (2020). Bachstetter, A. D. et al. A blunted T H 17 cytokine signature in women with mild cognitive impairment: Insights from inflammatory profiling of a community-based cohort of older adults. Brain Communications 5 (2023). Begum, A. N. et al. Women with the Alzheimer's risk marker ApoE4 lose Aβ-specific CD4 + T cells 10-20 years before men. Translational Psychiatry 4 (2014). Minett, T. et al. Microglial immunophenotype in dementia with Alzheimer's pathology. Journal of Neuroinflammation 13 (2016). Musiek, E. S. & Bennett, D. A. Aducanumab and the "post-amyloid" era of Alzheimer research. Neuron 109 , 3045-3047 (2021). Knopman, D. S., Jones, D. T. & Greicius, M. D. Failure to demonstrate efficacy of aducanumab: An analysis of the EMERGE and ENGAGE trials as reported by Biogen, December 2019. Alzheimer's & Dementia 17 , 696-701 (2020). Burke, J. F., Kerber, K. A., Langa, K. M., Albin, R. L. & Kotagal, V. Lecanemab: Looking before we leap. Neurology 101 , 661-665 (2023). Imbimbo, B. P., Balducci, C., Ippati, S. & Watling, M. Initial failures of anti-tau antibodies in Alzheimer's disease are reminiscent of the amyloid-β story. Neural Regeneration Research 18 , 117-118 (2022). Harris, K. et al. The impact of routine vaccinations on Alzheimer's disease risk in persons 65 years and older: A claims-based cohort study using propensity score matching. Journal of Alzheimer's Disease 95 , 703-718 (2023). Wu, X. et al. Adult vaccination as a protective factor for dementia: A meta-analysis and systematic review of population-based observational studies. Frontiers in Immunology 13 , 872542 (2022). Eyting, M. et al. A natural experiment on the effect of herpes zoster vaccination on dementia. Nature (2025). Taquet, M., Dercon, Q., Todd, J. A. & Harrison, P. J. The recombinant shingles vaccine is associated with lower risk of dementia. Nature Medicine 30 , 2777-2781 (2024). Rutledge, J. et al. Comprehensive proteomics of CSF, plasma, and urine identify DDC and other biomarkers of early Parkinson's disease. Acta Neuropathologica 147 , 52 (2024). Imam, F. B. et al. The Global Neurodegeneration Proteomics Consortium: Biomarker and drug target discovery across ~40,000 biosample analyses for AD, PD, ALS, FTD, and aging. Nature Medicine (2025). Seifar, F. et al. Large-scale deep proteomic analysis in Alzheimer's disease brain regions across race and ethnicity. Alzheimer's & Dementia 20 , 8878-8897 (2024). McKhann, G. M. et al. The diagnosis of dementia due to Alzheimer's disease: Recommendations from the National Institute on Aging-Alzheimer's Association workgroups on diagnostic guidelines for Alzheimer's disease. Alzheimer's & Dementia 7 , 263-269 (2011). Mifflin, M. D. et al. A new predictive equation for resting energy expenditure in healthy individuals. American Journal of Clinical Nutrition 51 , 241-247 (1990). Shindelin, J. et al. Fiji: An open-source platform for biological-image analysis. Nature Methods 9 , 676-682 (2012). Gold, L. et al. Aptamer-based multiplexed proteomic technology for biomarker discovery. PLoS One 5 , e15004 (2010). Williams, S. A. et al. Plasma protein patterns as comprehensive indicators of health. Nature Medicine 25 , 1851-1857 (2019). Rappsilber, J., Ishihama, Y. & Mann, M. Stop and go extraction tips for matrix-assisted laser desorption / ionization, nanoelectrospray, and LC/MS sample pretreatment in proteomics. Annals of Chemistry 75 , 663-670 (2003). Demichev, V., Messner, C. B., Vernardis, S. I., Lilley, K. S. & Ralser, M. DIA-NN: Neural networks and interference correction enable deep proteome coverage in high throughput. Nature Methods 17 , 41-44 (2020). Ross, B. C. Mutual information between discrete and continuous data sets. PLoS One 9 , e87357 (2014). Fayyad, U. M. & Irani, K. B. in 13th International Joint Conference on Uncertainty in Artificial Intelligence 1022-1029 (Chambery, France, 1993). Wickham, H. ggplot2: Elegant Graphics for Data Analysis . (Springer-Verlag, 2016). Xia, J., Benner, M. J. & Hancock, R. E. W. NetworkAnalyst - Integrative approaches for protein-protein interaction network analysis and visual exploration. Nucleic Acids Research 42 , W167-W174 (2014). Xia, J., Gill, E. & Hancock, R. E. W. NetworkAnalyst for statistical, visual and network-based appraoches for meta-analysis of expression data. Nature Protocols 10 , 823-844 (2015). Zhou, G. et al. NetworkAnalyst 3.0: A visual analytics platform for comprehensive gene expression profiling and meta-analysis. Nucleic Acids Research 47 , W234-W241 (2019). Orchard, S. et al. Protein interaction data curation - The International Molecular Exchange Consortium (IMEx). Nature Methods 9 , 345-350 (2013). Breuer, K. et al. InnateDB: Systems biology of innate immunity and beyond - Recent updates and continuing curation. Nucleic Acids Research 41 , 1228-1233 (2013). Mi, H., Muruganujan, A., Casagrande, J. T. & Thomas, P. D. Large-scale gene function analysis with PANTHER classification system. Nature Protocols 8 , 1551-1566 (2019). Kanehisa, M. & Goto, S. KEGG: Kyoto Encyclopedia of Genes and Genomes. Nucleic Acids Research 38 , 27-30 (2000). Kanehisa, M. & Sato, Y. K., M. KEGG mapping tools for uncovering hidden featuers in biological data. Protein Science 31 , 47-53 (2021). Additional Declarations There is NO Competing Interest. Supplementary Files Shvetcovetal.2025SupplementaryTables.xlsx Supplementary Tables Extendeddata.docx Cite Share Download PDF Status: Under Review Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-7089423","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":485432585,"identity":"6442c28e-388f-4483-9301-05d5640a0845","order_by":0,"name":"Caitlin Finney","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAABDUlEQVRIiWNgGAWjYFACxgaGBwYMDOzNYJ4NAxsRWhobEoBaeA6DeWnEaAFakwAkeQ6AOYcJq+dvP9z+IKHgDgMPO++xBx/bzif2iZ19wPBxTy0D/4wErFokziSCHPaMgYeZL91wZtvtxDbpdAPGGc+OM0jcwK7FgAGs5TCDPTOPmTRv2+3cNuk0BmaeA8cYGHBp4X8I0cID0vK37RxCizwuLRKJSFoY2w7AtNQwGODQInHjYeMMoBYekBbJnnPJ9SAtB2ccOMBjeOYB9hDrT3/w4cOfw3I8/GfMJH6U2RnLz05jfPDhQJ2c3HHstsAAD5hkhMbjAWAE8eBVjwB/4Kw6InWMglEwCkbBCAAAjw1cR850+m0AAAAASUVORK5CYII=","orcid":"https://orcid.org/0000-0002-9357-8316","institution":"Westmead Institute for Medical Research","correspondingAuthor":true,"prefix":"","firstName":"Caitlin","middleName":"","lastName":"Finney","suffix":""},{"id":485432586,"identity":"7a28c17a-18b1-4271-a3c9-617ccbf3c53a","order_by":1,"name":"Artur Shvetcov","email":"","orcid":"","institution":"Westmead Institute for Medical Research","correspondingAuthor":false,"prefix":"","firstName":"Artur","middleName":"","lastName":"Shvetcov","suffix":""},{"id":485432587,"identity":"1e75d2e4-d9e7-40a1-8164-3cd761ead11c","order_by":2,"name":"Shannon Thomson","email":"","orcid":"","institution":"Westmead Institute for Medical Research","correspondingAuthor":false,"prefix":"","firstName":"Shannon","middleName":"","lastName":"Thomson","suffix":""},{"id":485432588,"identity":"1910b95f-c298-4300-bd05-0ecae040db77","order_by":3,"name":"Mark Graham","email":"","orcid":"https://orcid.org/0000-0002-7290-1217","institution":"Children's Medical Research Institute","correspondingAuthor":false,"prefix":"","firstName":"Mark","middleName":"","lastName":"Graham","suffix":""},{"id":485432589,"identity":"cc558c15-7160-481c-92ac-d2d0dc1d2c8b","order_by":4,"name":"Brittany Hauger","email":"","orcid":"","institution":"University of Kansas Alzheimer's Disease Research Center","correspondingAuthor":false,"prefix":"","firstName":"Brittany","middleName":"","lastName":"Hauger","suffix":""},{"id":485432590,"identity":"b01abfed-a3ec-4d6a-9993-277411350c55","order_by":5,"name":"Jessica Keller","email":"","orcid":"","institution":"University of Kansas Alzheimer's Disease Research Center","correspondingAuthor":false,"prefix":"","firstName":"Jessica","middleName":"","lastName":"Keller","suffix":""},{"id":485432591,"identity":"c883daed-652d-4739-b1ee-21b6b9911808","order_by":6,"name":"Christine Smoyer","email":"","orcid":"","institution":"University of Kansas Medical Center","correspondingAuthor":false,"prefix":"","firstName":"Christine","middleName":"","lastName":"Smoyer","suffix":""},{"id":485432592,"identity":"c66c792c-b6e9-45e2-97a5-be55247b4b7a","order_by":7,"name":"Sarah Tague","email":"","orcid":"","institution":"University of Kansas Medical Center","correspondingAuthor":false,"prefix":"","firstName":"Sarah","middleName":"","lastName":"Tague","suffix":""},{"id":485432593,"identity":"355cc1eb-faf4-41c5-9bf8-d9990a113b0f","order_by":8,"name":"Ann-Na Cho","email":"","orcid":"https://orcid.org/0000-0003-0047-8867","institution":"University of Sydney","correspondingAuthor":false,"prefix":"","firstName":"Ann-Na","middleName":"","lastName":"Cho","suffix":""},{"id":485432594,"identity":"a143161e-de45-400e-9049-4804371c03f5","order_by":9,"name":"Farhad Imam","email":"","orcid":"https://orcid.org/0000-0003-2854-2568","institution":"Gates Ventures","correspondingAuthor":false,"prefix":"","firstName":"Farhad","middleName":"","lastName":"Imam","suffix":""},{"id":485432595,"identity":"d1e18f26-b688-4bb9-803a-b79de6ef1dcb","order_by":10,"name":"Varsha Krish","email":"","orcid":"","institution":"Gates Ventures","correspondingAuthor":false,"prefix":"","firstName":"Varsha","middleName":"","lastName":"Krish","suffix":""},{"id":485432596,"identity":"a0b90f1d-598b-4f9f-9083-390ed20929b7","order_by":11,"name":"Global Neurodegeneration Proteomics Consortium Global Neurodegeneration Proteomics Consortium","email":"","orcid":"","institution":"Gates Ventures","correspondingAuthor":false,"prefix":"","firstName":"Global","middleName":"Neurodegeneration Proteomics Consortium Global Neurodegeneration Proteomics","lastName":"Consortium","suffix":""},{"id":485432597,"identity":"d6f11bbd-e617-4a8d-84d6-fc49d71bed59","order_by":12,"name":"Matthew Taylor","email":"","orcid":"","institution":"University of Kansas Alzheimer's Disease Research Center","correspondingAuthor":false,"prefix":"","firstName":"Matthew","middleName":"","lastName":"Taylor","suffix":""},{"id":485432598,"identity":"f6afe27e-557d-4c86-ad9e-66d170639f4b","order_by":13,"name":"Jonathan Mahnken","email":"","orcid":"","institution":"University of Kansas Alzheimer's Disease Research Center","correspondingAuthor":false,"prefix":"","firstName":"Jonathan","middleName":"","lastName":"Mahnken","suffix":""},{"id":485432599,"identity":"679c1d4a-46da-43dc-b82d-88ae6dbdab82","order_by":14,"name":"Debra Sullivan","email":"","orcid":"","institution":"University of Kansas Alzheimer's Disease Research Center","correspondingAuthor":false,"prefix":"","firstName":"Debra","middleName":"","lastName":"Sullivan","suffix":""},{"id":485432600,"identity":"bf4ad592-b3ef-4018-b623-2c1ee268c9af","order_by":15,"name":"Joanne Reed","email":"","orcid":"","institution":"Westmead Institute for Medical Research","correspondingAuthor":false,"prefix":"","firstName":"Joanne","middleName":"","lastName":"Reed","suffix":""},{"id":485432601,"identity":"617a9485-efeb-4036-954d-17e202127611","order_by":16,"name":"Jeffrey Burns","email":"","orcid":"https://orcid.org/0000-0001-7609-8954","institution":"Univesrity of Kansas Medical Center","correspondingAuthor":false,"prefix":"","firstName":"Jeffrey","middleName":"","lastName":"Burns","suffix":""},{"id":485432602,"identity":"8d97f9d5-c832-4c4e-8e3d-099ea3cd1bcf","order_by":17,"name":"Chad Slawson","email":"","orcid":"","institution":"University of Kansas Alzheimer's Disease Research Center","correspondingAuthor":false,"prefix":"","firstName":"Chad","middleName":"","lastName":"Slawson","suffix":""},{"id":485432603,"identity":"69a88d9f-aa98-4c22-921e-62e67fa59b94","order_by":18,"name":"Russell Swerdlow","email":"","orcid":"https://orcid.org/0000-0003-2948-7230","institution":"University of Kansas","correspondingAuthor":false,"prefix":"","firstName":"Russell","middleName":"","lastName":"Swerdlow","suffix":""},{"id":485432604,"identity":"d7dcac08-a08a-457a-9a7c-742a2145159b","order_by":19,"name":"Heather Wilkins","email":"","orcid":"","institution":"University of Kansas Alzheimer’s Disease Research Center, University of Kansas Medical Center, Kansas City","correspondingAuthor":false,"prefix":"","firstName":"Heather","middleName":"","lastName":"Wilkins","suffix":""}],"badges":[],"createdAt":"2025-07-10 06:10:14","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7089423/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7089423/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":86773290,"identity":"bf8d1145-0866-4746-8196-5d6bcc32ca39","added_by":"auto","created_at":"2025-07-15 12:09:05","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":1250632,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eStudy design and characterization of the plasma and cerebrospinal fluid proteome of \u003c/strong\u003e\u003cem\u003e\u003cstrong\u003eAPOE\u003c/strong\u003e\u003c/em\u003e\u003cstrong\u003e ε4 carriers and non-carriers with and without Alzheimer’s disease. \u003c/strong\u003e(\u003cstrong\u003ea\u003c/strong\u003e)\u003cstrong\u003e \u003c/strong\u003eDesign of the current study using plasma and cerebrospinal (CSF) proteomic data from the Global Neurodegeneration Proteomics Consortium cohort, brain proteomic data from the dorsolateral prefrontal cortex (dlPFC) and superior temporal gyrus (STG) of donors from the Accelerating Medicines Partnership in Alzheimer’s Disease Diverse Cohorts study, cortical organoids generated from donors from the University of Kansas Alzheimer’s Disease Research Center cohort, and plasma proteomic data from the Therapeutic Diets in AD pilot clinical trial of 12 week treatment with an anti-inflammatory medical ketogenic diet. (\u003cstrong\u003eb\u003c/strong\u003e) Performance metrics of machine learning models using \u003cem\u003eAPOE\u003c/em\u003e ε4 plasma proteins as features. The graph shows mean area under the curve (AUC) \u003cu\u003e+\u003c/u\u003e SD for both testing and training datasets across five folds repeated 10 times. Models were trained and validated on a 70% training dataset and tested using a 30% withheld testing dataset. Race 1 represents Black or African American participants and Race 2 represents American Indian or Alaskan Native participants. (\u003cstrong\u003ec\u003c/strong\u003e) Performance metrics of machine learning models using the 51 \u003cem\u003eAPOE\u003c/em\u003e ε4 CSF proteins as features. The graph shows mean area under the curve (AUC) \u003cu\u003e+\u003c/u\u003e SD for both testing and training datasets across five folds repeated 10 times. Models were trained and validated on a 70% training dataset and tested using a 30% withheld testing dataset. (\u003cstrong\u003ed\u003c/strong\u003e) Venn diagram showing the overlap between identified CSF and plasma \u003cem\u003eAPOE\u003c/em\u003e ε4 proteins. (\u003cstrong\u003ee\u003c/strong\u003e) Significantly (FDR \u0026lt; 0.05) enriched PANTHER biological processes \u003cem\u003eAPOE\u003c/em\u003e ε4 plasma and CSF proteins, with viral processes as the top enriched pathway across both. (\u003cstrong\u003ef\u003c/strong\u003e) Significantly (FDR \u0026lt; 0.05) enriched KEGG immune pathways for \u003cem\u003eAPOE\u003c/em\u003e ε4 plasma and CSF proteins. Enriched pathways included those involved in pro-inflammatory, cytokine, infection, and both adaptive and innate immune function. (\u003cstrong\u003eg\u003c/strong\u003e) Significantly (FDR \u0026lt; 0.05) enriched Reactome immune pathways for \u003cem\u003eAPOE\u003c/em\u003e ε4 plasma and CSF proteins. Enriched pathways mirror those found with KEGG, with enrichment for pro-inflammatory, cytokine, and broad adaptive and innate immune functions. (\u003cstrong\u003eh\u003c/strong\u003e) Enrichment of \u003cem\u003eAPOE\u003c/em\u003e ε4 plasma and CSF proteins across distinct immune cell subsets. Plasma proteins were especially enriched across neutrophils, basophils, and T cells. CSF proteins were more enriched in T cells and NK cells. (\u003cstrong\u003ei\u003c/strong\u003e) Given that we found high enrichment for T cells, broadly, we did a further enrichment for distinct T-cell subsets. \u003cem\u003eAPOE\u003c/em\u003e ε4 plasma proteins were the most enriched for central memory CD8\u003csup\u003e+\u003c/sup\u003e T cells and CD4\u003csup\u003e+\u003c/sup\u003e memory TFH cells. CSF \u003cem\u003eAPOE\u003c/em\u003e ε4 proteins were also enriched in CD4\u003csup\u003e+\u003c/sup\u003e memory TFH cells as well as CD4\u003csup\u003e+\u003c/sup\u003e memory Th1, CD8\u003csup\u003e+\u003c/sup\u003e terminal effector memory, and non-Vδ2 γδ T cells. (\u003cstrong\u003ej\u003c/strong\u003e) Enrichment of \u003cem\u003eAPOE\u003c/em\u003e ε4 plasma and CSF proteins across commonly affected brain regions in AD, with the highest enrichment for white matter. (\u003cstrong\u003ek\u003c/strong\u003e) Further enrichment of \u003cem\u003eAPOE\u003c/em\u003e ε4 CSF proteins across white matter cell subtypes. Microglia and oligodendrocytes were the most enriched immune cells and endothelial and vascular cells were the most enriched vascular cells. Enrichments are based on single cell RNA-sequencing data from the Human Protein Atlas \u003csup\u003e12-14\u003c/sup\u003e and plots show mix-max scaling of protein-coding transcripts per million for each identified \u003cem\u003eAPOE\u003c/em\u003e ε4 plasma protein. Abbreviations: AD: Alzheimer’s disease; COPC: committed oligodendrocyte precursor cell; CSF: cerebrospinal fluid; DC: dendritic cell; dlPFC: dorsolateral prefrontal cortex; Het: heterozygous; Hom: homozygous; NI: non-impaired control; OPC: oligodendrocyte precursor cell; STG: superior temporal gyrus; TFH: T follicular helper T cells; VASMC: vascular-associated smooth muscle cells.\u003c/p\u003e","description":"","filename":"image1.png","url":"https://assets-eu.researchsquare.com/files/rs-7089423/v1/fd3a39098d8f1f99b36442f8.png"},{"id":86773295,"identity":"bb8a74ac-7ec5-492c-ab29-1c8e98dc0d70","added_by":"auto","created_at":"2025-07-15 12:09:05","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":1058370,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eCharacterization of the dorsolateral prefrontal cortex and superior temporal gyrus brain proteomes of \u003c/strong\u003e\u003cem\u003e\u003cstrong\u003eAPOE\u003c/strong\u003e\u003c/em\u003e\u003cstrong\u003e ε4 carriers and non-carriers with and without Alzheimer’s disease. \u003c/strong\u003e(\u003cstrong\u003ea\u003c/strong\u003e) Performance metrics of machine learning models using the \u003cem\u003eAPOE\u003c/em\u003e ε4 dorsolateral prefrontal cortex (dlPFC) or superior temporal gyrus (STG) proteins as features. The graph shows mean area under the curve (AUC) \u003cu\u003e+\u003c/u\u003e SD for both testing and training datasets across five folds repeated 10 times. Models were trained and validated on a 70% training dataset and tested using a 30% withheld testing dataset. Race 1 refers to Black or African American donors. (\u003cstrong\u003eb\u003c/strong\u003e) Significantly (FDR \u0026lt; 0.05) enriched PANTHER biological processes for the \u003cem\u003eAPOE\u003c/em\u003e ε4 dlPFC and STG proteins. The top enriched biological process across both regions was viral processes (FDR = 3.95E-17 and 2.18E-18, respectively). The scale bar represents -1og10(FDR). (\u003cstrong\u003ec\u003c/strong\u003e) Significantly (FDR \u0026lt; 0.05) enriched KEGG immune pathways for the \u003cem\u003eAPOE\u003c/em\u003e ε4 dlPFC and STG proteins. Enriched pathways included those involved in pro-inflammatory, cytokine, infection, and both adaptive and innate immune function. The scale bar represents -1og10(FDR). (\u003cstrong\u003ed\u003c/strong\u003e) Significantly (FDR \u0026lt; 0.05) enriched Reactome immune pathways for the \u003cem\u003eAPOE\u003c/em\u003e ε4 dlPFC and STG proteins. Enriched pathways mirror those found with KEGG, with enrichment for pro-inflammatory, cytokine, and broad adaptive and innate immune functions. Enrichment for these pathways was more significant in the STG than the dlPFC. The scale bar represents -1og10(FDR). (\u003cstrong\u003ee\u003c/strong\u003e,\u003cstrong\u003ef\u003c/strong\u003e) Given our previous findings in the plasma and CSF showing \u003cem\u003eAPOE\u003c/em\u003e ε4 proteins were enriched in white matter, we performed an enrichment analysis for white matter cell subtypes. (\u003cstrong\u003ee\u003c/strong\u003e) dlPFC \u003cem\u003eAPOE\u003c/em\u003e ε4 proteins were enriched for microglia as well as endothelial cells. (\u003cstrong\u003ef\u003c/strong\u003e) STG \u003cem\u003eAPOE\u003c/em\u003e ε4 proteins were similarly enriched for microglia and endothelial cells. Enrichments are based on single cell RNA-sequencing data from the Human Protein Atlas \u003csup\u003e14\u003c/sup\u003e and plots show mix-max scaling of protein-coding transcripts per million for each identified \u003cem\u003eAPOE\u003c/em\u003e ε4 dlPFC or STG protein. Abbreviations: AD: Alzheimer’s disease; COPC: committed oligodendrocyte precursor cell; CSF: cerebrospinal fluid; dlPFC: dorsolateral prefrontal cortex; NI: non-impaired control; OPC: oligodendrocyte precursor cell; STG: superior temporal gyrus; VASMC: vascular-associated smooth muscle cells.\u003c/p\u003e","description":"","filename":"image2.png","url":"https://assets-eu.researchsquare.com/files/rs-7089423/v1/067dbade86d16163542c6476.png"},{"id":86773289,"identity":"ef7dfa1c-aed7-4e66-8c01-d8a6956e7443","added_by":"auto","created_at":"2025-07-15 12:09:05","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":547447,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eAlzheimer’s pathology development in induced pluripotent stem cell-derived cortical organoids of an \u003c/strong\u003e\u003cem\u003e\u003cstrong\u003eAPOE\u003c/strong\u003e\u003c/em\u003e\u003cstrong\u003e ε4 carrier and non-carrier. \u003c/strong\u003e(\u003cstrong\u003ea\u003c/strong\u003e) Representative immunofluorescent confocal microscopy images of the presence of AD pathological markers p-tau217 and amyloid-β in cortical organoids from an \u003cem\u003eAPOE\u003c/em\u003e ε4 carrier and non-carrier at four and eight weeks of maturation. (\u003cstrong\u003eb\u003c/strong\u003e) Quantification of the mean fluorescent intensity of p-tau217 and amyloid-β, normalized to DAPI, in cortical organoids from an \u003cem\u003eAPOE\u003c/em\u003e ε4 carrier and non-carrier at four weeks of maturation (\u003cem\u003en\u003c/em\u003e = 3 technical replicates/group). (\u003cstrong\u003ec\u003c/strong\u003e) Quantification of the mean fluorescent intensity of p-tau217 and amyloid-β, normalized to DAPI, in cortical organoids from an \u003cem\u003eAPOE\u003c/em\u003eε4 carrier and non-carrier at four weeks of maturation (*\u003cem\u003ep \u003c/em\u003e= 0.0203, unpaired two-tailed t-test; **\u003cem\u003ep\u003c/em\u003e = 0.0010, unpaired two-tailed t-test; \u003cem\u003en\u003c/em\u003e= 3 technical replicates/group). (\u003cstrong\u003ed\u003c/strong\u003e,\u003cstrong\u003ee\u003c/strong\u003e,\u003cstrong\u003ef\u003c/strong\u003e) ELISA analysis of secreted amyloid-β42/40 into the culture media\u003cem\u003e \u003c/em\u003efrom \u003cem\u003eAPOE\u003c/em\u003e ε4 carrier and non-carrier cortical organoids at (\u003cstrong\u003ed\u003c/strong\u003e) 4 weeks and (\u003cstrong\u003ee\u003c/strong\u003e) 8 weeks of maturation. There was no difference at 4 weeks between the \u003cem\u003eAPOE\u003c/em\u003eε4 carrier and non-carrier whereas the non-carrier cortical organoids showed significantly (*\u003cem\u003ep \u003c/em\u003e= 0.0432, unpaired t-test, n = 5 technical replicates/group) higher levels of secreted amyloid-β42/40 at 8 weeks. (\u003cstrong\u003ef\u003c/strong\u003e) The non-carrier cortical organoids showed significantly different levels of secreted amyloid-β42/40 over time relative to the\u003cem\u003e APOE\u003c/em\u003e ε4 cortical organoids (** \u003cem\u003ep\u003c/em\u003e= 0.0166, two-way ANOVA, n = 5 technical replicates/group). (\u003cstrong\u003eg\u003c/strong\u003e,\u003cstrong\u003eh\u003c/strong\u003e,\u003cstrong\u003ei\u003c/strong\u003e) ELISA analysis of secreted p-tau217 into the culture media from\u003cem\u003e APOE\u003c/em\u003e ε4 carrier and non-carrier cortical organoids at (\u003cstrong\u003eg\u003c/strong\u003e) 4 weeks and (\u003cstrong\u003eh\u003c/strong\u003e) 8 weeks of maturation (\u003cem\u003en\u003c/em\u003e = 5 technical replicates/group). There were no significant differences between groups. (\u003cstrong\u003ei\u003c/strong\u003e) Time course of \u003cem\u003eAPOE\u003c/em\u003eε4 carrier and non-carrier cortical organoid-secreted p-tau217 showing no significant changes over time (\u003cem\u003en\u003c/em\u003e = 5 technical replicates/group).\u003c/p\u003e","description":"","filename":"image3.png","url":"https://assets-eu.researchsquare.com/files/rs-7089423/v1/d8e02361ab5aa5ad5ecf92d5.png"},{"id":86773296,"identity":"abaf82f8-8132-4530-95c1-758a1f19cf9c","added_by":"auto","created_at":"2025-07-15 12:09:05","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":1114341,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eCharacterization of the \u003c/strong\u003e\u003cem\u003e\u003cstrong\u003eAPOE\u003c/strong\u003e\u003c/em\u003e\u003cstrong\u003e ε4 proteome over time in stem cell-derived cortical organoids from an \u003c/strong\u003e\u003cem\u003e\u003cstrong\u003eAPOE\u003c/strong\u003e\u003c/em\u003e\u003cstrong\u003e ε4 carrier and non-carrier. \u003c/strong\u003e(\u003cstrong\u003ea\u003c/strong\u003e) Volcano plot representing 642 significantly differentially expressed proteins (in red) in \u003cem\u003eAPOE\u003c/em\u003e ε4 cortical organoids at four weeks of maturation before the onset of AD pathologies; \u003cem\u003en\u003c/em\u003e = 4 technical replicates/group (see Supplementary Table 8 for a full list of proteins, adjusted \u003cem\u003ep\u003c/em\u003e-values, and fold changes). (\u003cstrong\u003eb\u003c/strong\u003e) Volcano plot representing 2,655 significantly differentially expressed proteins (in red) in \u003cem\u003eAPOE\u003c/em\u003e ε4 cortical organoids at eight weeks of maturation after the onset of AD pathologies; \u003cem\u003en\u003c/em\u003e = 4 technical replicates/group (see Supplementary Table 8 for a full list of proteins, adjusted \u003cem\u003ep\u003c/em\u003e-values, and fold changes). (\u003cstrong\u003ec\u003c/strong\u003e) KEGG pathway enrichment analysis of differentially expressed proteins in \u003cem\u003eAPOE\u003c/em\u003e ε4 cortical organoids at eight weeks of maturation showing significant (FDR = 0.00000211) enrichment in the AD pathway. Pink colored proteins represent those identified as differentially expressed proteins at eight weeks. (\u003cstrong\u003ed\u003c/strong\u003e) Significantly (FDR \u0026lt; 0.05) enriched PANTHER biological processes for \u003cem\u003eAPOE\u003c/em\u003eε4 cortical organoids at four and eight weeks of maturation. The top enriched process across both timepoints was viral processes (FDR = 1.29E-12 and 3.99E-36, respectively). The scale bar represents -1og10(FDR). (\u003cstrong\u003ee\u003c/strong\u003e) Graph showing the change in the number of identified (“hit”) proteins in overlapping \u003cem\u003eAPOE\u003c/em\u003eε4 PANTHER biological processes across cortical organoids at four and eight weeks of maturation. The y-axis represents the “hit” proteins expressed as a percentage of expected proteins (“hit” proteins in pathway / total number of proteins in pathway). (\u003cstrong\u003ef\u003c/strong\u003e) Significantly (FDR \u0026lt; 0.05) enriched KEGG immune pathways for \u003cem\u003eAPOE\u003c/em\u003e ε4 cortical organoids at four and eight weeks of maturation. Enriched pathways were reflective of processes involved in pro-inflammatory, cytokine, infection, and immune function and time point-specific pathways indicated a changing role of microglia from surveillance and immune sensing to activated and phagocytic. (\u003cstrong\u003eg\u003c/strong\u003e) Graph showing the change in the number of identified (“hit”) proteins in overlapping \u003cem\u003eAPOE\u003c/em\u003eε4 KEGG immune pathways across cortical organoids at four and eight weeks of maturation. The y-axis represents the “hit” proteins expressed as a percentage of expected proteins (“hit” proteins in pathway / total number of proteins in pathway). (\u003cstrong\u003eh\u003c/strong\u003e) Significantly (FDR \u0026lt; 0.05) enriched Reactome immune pathways for \u003cem\u003eAPOE\u003c/em\u003e ε4 cortical organoids at four and eight weeks of maturation. Enriched pathways mirrored those seen in the KEGG analysis and further indicated a change in microglia over time from innate immune sensing and inflammatory priming to phagocytic, disease-associated microglia (DAM). (\u003cstrong\u003ei\u003c/strong\u003e) Graph showing the change in the number of identified (“hit”) proteins in overlapping \u003cem\u003eAPOE\u003c/em\u003e ε4 Reactome immune pathways across cortical organoids at four and eight weeks of maturation. The y-axis represents the “hit” proteins expressed as a percentage of expected proteins (“hit” proteins in pathway / total number of proteins in pathway).\u003c/p\u003e","description":"","filename":"image4.png","url":"https://assets-eu.researchsquare.com/files/rs-7089423/v1/03662d6b20fea4587c613f10.png"},{"id":86773286,"identity":"553bdfa3-7280-4ac7-a861-f8dd70a476c4","added_by":"auto","created_at":"2025-07-15 12:09:04","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":481716,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eCross-tissue comparison of immune pathways in \u003c/strong\u003e\u003cem\u003e\u003cstrong\u003eAPOE\u003c/strong\u003e\u003c/em\u003e\u003cstrong\u003e ε4 carriers. \u003c/strong\u003e(\u003cstrong\u003ea\u003c/strong\u003e) Systemic \u003cem\u003eAPOE\u003c/em\u003eε4 immune KEGG and Reactome pathways identified across the plasma, cerebrospinal fluid (CSF), dorsolateral prefrontal cortex (dlPFC), superior temporal gyrus (STG), and cortical organoids (CO). (\u003cstrong\u003eb\u003c/strong\u003e) Central nervous system-specific \u003cem\u003eAPOE\u003c/em\u003eε4 immune KEGG and Reactome pathways identified across the CSF, dlPFC, STG, and CO. (\u003cstrong\u003ec\u003c/strong\u003e) Peripheral system-specific \u003cem\u003eAPOE\u003c/em\u003e ε4 immune KEGG and Reactome pathways identified only in the plasma. Scale bars represent -1og10(FDR). Abbreviations: AD: Alzheimer’s disease; CO: cortical organoids; CSF: cerebrospinal fluid; dlPFC: dorsolateral prefrontal cortex; STG: superior temporal gyrus.\u003c/p\u003e","description":"","filename":"image5.png","url":"https://assets-eu.researchsquare.com/files/rs-7089423/v1/f9f8b32295043daa832f3ae5.png"},{"id":86773292,"identity":"9a63327f-0f4b-49a8-bf3e-c34166a61315","added_by":"auto","created_at":"2025-07-15 12:09:05","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":1406483,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eCharacterization of the plasma proteome of Alzheimer’s patients after 12 weeks of an anti-inflammatory medical ketogenic diet from the Therapeutic Diets in Alzheimer’s Disease clinical trial. \u003c/strong\u003e(\u003cstrong\u003ea\u003c/strong\u003e) Volcano plot representing no significantly differentially expressed proteins in non-carriers at the end of the 12-week study; \u003cem\u003en\u003c/em\u003e = 9 \u003cem\u003eAPOE\u003c/em\u003e ε4 carriers and \u003cem\u003en\u003c/em\u003e = 6 non-carriers. (\u003cstrong\u003eb\u003c/strong\u003e) Volcano plot representing 264 significantly (adjusted \u003cem\u003ep\u003c/em\u003e-value \u0026lt; 0.05) differentially expressed proteins in \u003cem\u003eAPOE\u003c/em\u003e ε4 carriers at the end of the 12-week study; \u003cem\u003en\u003c/em\u003e = 9 \u003cem\u003eAPOE\u003c/em\u003e ε4 carriers and \u003cem\u003en\u003c/em\u003e = 6 non-carriers (see Supplementary Table 11 for a full list of proteins, adjusted \u003cem\u003ep\u003c/em\u003e-values, and fold changes). (\u003cstrong\u003ec\u003c/strong\u003e) PANTHER biological processes analysis of pathways enriched for up- or down-regulated differentially expressed proteins in \u003cem\u003eAPOE\u003c/em\u003e ε4 carriers. (\u003cstrong\u003ed\u003c/strong\u003e) KEGG immune pathway analysis of pathways enriched for up- or down-regulated differentially expressed proteins in \u003cem\u003eAPOE\u003c/em\u003e ε4 carriers. (\u003cstrong\u003ee\u003c/strong\u003e) Reactome immune pathway analysis of pathways enriched for up- or down-regulated differentially expressed proteins in \u003cem\u003eAPOE\u003c/em\u003e ε4 carriers. (\u003cstrong\u003ef\u003c/strong\u003e) Enrichment of significantly up- or down-regulated \u003cem\u003eAPOE\u003c/em\u003e ε4 proteins across distinct immune cell subsets. Downregulated proteins were primarily enriched in T-regs and NK cells. Upregulated proteins were primarily enriched across memory CD8\u003csup\u003e+\u003c/sup\u003e T cells, naïve CD8\u003csup\u003e+\u003c/sup\u003e T cells, T-regs, and NK cells. There was also more enrichment across the spectrum of innate immune cells relative to \u003cem\u003eAPOE\u003c/em\u003e ε4 downregulated proteins. (\u003cstrong\u003eg\u003c/strong\u003e) Given that we found high enrichment for T cells, broadly, we did a further enrichment for distinct T-cell subsets. Downregulated \u003cem\u003eAPOE\u003c/em\u003e ε4 proteins were primarily enriched in effector memory and central memory CD8\u003csup\u003e+\u003c/sup\u003e T cells and non-Vδ2 γδ T cells. Upregulated \u003cem\u003eAPOE\u003c/em\u003e ε4 proteins were similarly enriched for effector memory CD8\u003csup\u003e+\u003c/sup\u003e T cells and non-Vδ2 γδ T cells as well as CD4\u003csup\u003e+\u003c/sup\u003e T cells including memory TFH and terminal effector memory. (\u003cstrong\u003eh\u003c/strong\u003e) Enrichment of significantly up- and down-regulated \u003cem\u003eAPOE\u003c/em\u003e ε4 proteins across commonly affected brain regions in AD, with the highest enrichment for white matter. Enrichments are based on single cell RNA-sequencing data from the Human Protein Atlas \u003csup\u003e12-14\u003c/sup\u003e and plots show mix-max scaling of protein-coding transcripts per million for each identified \u003cem\u003eAPOE\u003c/em\u003e ε4 plasma protein. Abbreviations: AD: Alzheimer’s disease; DC: dendritic cell; TFH: T follicular helper T cells.\u003c/p\u003e","description":"","filename":"image6.png","url":"https://assets-eu.researchsquare.com/files/rs-7089423/v1/c2a068ca930c9958267c9493.png"},{"id":87517863,"identity":"738ade7e-2b17-498b-b6fb-a24c58fa4f28","added_by":"auto","created_at":"2025-07-24 16:54:06","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":7850397,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7089423/v1/1f332ada-4498-40bc-bf5a-56108934eceb.pdf"},{"id":86773285,"identity":"70f042eb-70d8-49e2-a17f-4bcff7c9c611","added_by":"auto","created_at":"2025-07-15 12:09:03","extension":"xlsx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":196922,"visible":true,"origin":"","legend":"Supplementary Tables","description":"","filename":"Shvetcovetal.2025SupplementaryTables.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-7089423/v1/3fef80f260f46fe4bb1bb782.xlsx"},{"id":86773284,"identity":"ebeaf3a4-fcfc-4507-8576-d90e7fb1f579","added_by":"auto","created_at":"2025-07-15 12:09:03","extension":"docx","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":587752,"visible":true,"origin":"","legend":"","description":"","filename":"Extendeddata.docx","url":"https://assets-eu.researchsquare.com/files/rs-7089423/v1/4b4abe5a1942c4c23de189db.docx"}],"financialInterests":"There is \u003cb\u003eNO\u003c/b\u003e Competing Interest.","formattedTitle":"Cross-tissue immune profiling of APOE ε4 reveals early dysregulation in Alzheimer’s disease","fulltext":[{"header":"Main","content":"\u003cp\u003eThe functional consequences of the largest genetic risk factor for late onset Alzheimer\u0026rsquo;s disease (AD), the \u0026epsilon;4 variant of the apolipoprotein E (\u003cem\u003eAPOE\u003c/em\u003e) gene, and how they drive AD pathogenesis remain poorly understood. Recent accumulating evidence suggests that \u003cem\u003eAPOE\u0026nbsp;\u003c/em\u003e\u0026epsilon;4 exerts broad modulatory effects on both innate and adaptive immunity across peripheral tissues and the central nervous system. \u003cem\u003eAPOE\u0026nbsp;\u003c/em\u003e\u0026epsilon;4 is associated with a shift toward a heightened pro-inflammatory state \u003csup\u003e1,2\u003c/sup\u003e. This is further evidenced by peripherally-derived immune cells showing exaggerated responses to innate immune stimulation \u003csup\u003e3,4\u003c/sup\u003e, altered antigen presentation \u003csup\u003e5\u003c/sup\u003e, disrupted T-cell homeostasis \u003csup\u003e6\u003c/sup\u003e, and signs of accelerated immune aging \u003csup\u003e7\u003c/sup\u003e. In the brain, \u003cem\u003eAPOE\u0026nbsp;\u003c/em\u003e\u0026epsilon;4 carriage is linked to pre-activated microglial phenotypes \u003csup\u003e8\u003c/sup\u003e, impaired lipid and autophagic homeostasis \u003csup\u003e9,10\u003c/sup\u003e, and amplified responses to amyloid-\u0026beta; pathology \u003csup\u003e10,11\u003c/sup\u003e. Together, these findings highlight a genotype-dependent dysregulation of immune homeostasis in \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4 carriers. However, several key questions remain, including whether immune changes represent innate genetic effects or secondary responses to disease, how these signatures evolve temporally with progression to AD, which aspects are systemic, central, or peripheral in origin, and whether they are modifiable. Resolving these questions is essential to understanding \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4-driven mechanisms of neurodegeneration and developing precision biomarkers and early intervention strategies for AD.\u003c/p\u003e\n\u003cp\u003eTo address how \u003cem\u003eAPOE\u0026nbsp;\u003c/em\u003e\u0026epsilon;4 influences immune function across tissues and time, we systematically profiled the plasma, CSF, and brain proteomes of APOE \u0026epsilon;4 carriers and non-carriers with and without AD (Fig. 1a). We identified a conserved, allele dose-dependent pro-inflammatory immune signature spanning both the periphery and central nervous system (CNS), independent of disease. This signature was recapitulated in patient-derived cortical organoids and emerged prior to the development of amyloid-\u0026beta; and tau pathology, indicating that \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4-driven immune changes are early, intrinsic, and may initiate downstream neurodegenerative processes. Cross-tissue comparisons revealed that many \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4-associated pathways were systemic, reflecting broad innate and adaptive immune activation. Others were tissue-specific and localized to the CNS or periphery. Importantly, short-term treatment with an anti-inflammatory medical ketogenic diet in \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4 carriers with AD partially reprogrammed this immune phenotype, reducing pro-inflammatory signaling and promoting regulatory and tissue-supportive immune functions. Together, these findings demonstrate that \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4 confers a systemic but locally modulated pro-inflammatory immune phenotype that emerges early, evolves with disease progression, and may promote AD pathology. This provides new insight into \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4-driven mechanisms of neurodegeneration and highlights the potential of this immune state to be therapeutically modulated.\u0026nbsp;\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003e\u003cstrong\u003e\u003cem\u003eAPOE\u003c/em\u003e\u003c/strong\u003e\u003cstrong\u003e\u0026nbsp;\u0026epsilon;4 carriers have a distinct pro-inflammatory immune phenotype in the plasma and cerebrospinal fluid\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe profiled 6,340 plasma proteins of \u003cem\u003en\u003c/em\u003e = 2,929 AD and \u003cem\u003en\u003c/em\u003e = 6,099 non-impaired controls from the Global Neurodegeneration Proteomics Consortium (GNPC) using the SomaScan 7k assay (Supplementary Table 1). Nine \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4-associated plasma proteins were identified (mutual information \u0026gt; 0.1), which clustered by carrier status and allele dose (Supplementary Table 2; Extended Data Fig. 1a-c). Machine learning using classification and regression trees (CART) showed these proteins robustly classified \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4 carriers versus non-carriers (AUC \u0026gt; 0.91 across subgroups; Extended Data Table 1; Fig. 1b), independent of AD status, sex, or race. No plasma proteins differentiated AD \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4 carriers from non-impaired carriers (mutual information \u0026lt; 0.01; AUC = 0.70), confirming an AD-independent signature. Multi-class CART further distinguished heterozygous from homozygous carriers (AUC \u0026gt; 0.93).\u003c/p\u003e\n\u003cp\u003eWe then analyzed 6,340 CSF proteins from \u003cem\u003en\u003c/em\u003e = 526 AD and \u003cem\u003en\u003c/em\u003e = 573 non-impaired controls (Supplementary Table 1), identifying 51 \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4-associated proteins (mutual information \u0026gt; 0.1) that similarly clustered by carrier status and allele dose (Supplementary Table 3; Extended Data Fig. 1d-f). CART models accurately classified \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4 carriers (AUC = 0.99), again independent of AD status or sex, and distinguished between heterozygous and homozygous individuals (AUC \u0026gt; 0.94; Extended Data Table 2; Fig. 1c). As in plasma, no CSF proteins distinguished AD from non-impaired \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4 carriers (mutual information \u0026lt; 0.01; AUC = 0.75), further supporting an AD-independent immune signature.\u003c/p\u003e\n\u003cp\u003eSix of the nine plasma proteins overlapped with CSF\u003cem\u003e\u0026nbsp;APOE\u003c/em\u003e \u0026epsilon;4 proteins (Fig. 1d). Pathway analyses revealed significant enrichment for viral processes (FDR \u0026lt; 0.05) in both plasma and CSF (PANTHER; Fig. 1e; Supplementary Table 4). KEGG and Reactome analyses confirmed enrichment for immune pathways including interferon, interleukin, Th17, TGF-\u0026beta; signaling, Epstein-Barr virus and hepatitis (Fig. 1e-f; Supplementary Table 4). CSF-specific pathways included toll-like receptor signaling and NK/B cell pathways, while plasma-specific enrichments were dominated by interferon signaling (Fig. 1e-f; Supplementary Table 4). Cell-type enrichment using Human Protein Atlas single-cell RNA-sequencing data \u003csup\u003e12,13\u003c/sup\u003e, showed plasma \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4 proteins were enriched in innate immune cells (neutrophils, basophils) and T cells, particularly central memory CD8\u003csup\u003e+\u003c/sup\u003e and CD4\u003csup\u003e+\u003c/sup\u003e TFH subsets (Fig. 1h-i). CSF proteins were enriched in NK cells and adaptive T-cell subsets, including CD4\u003csup\u003e+\u003c/sup\u003e Th1 and CD8\u003csup\u003e+\u003c/sup\u003e terminal effector memory cells (Fig. 1i). Both plasma and CSF \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4 proteins were preferentially expressed in white matter, with further enrichment in microglia, oligodendrocytes, and vascular-associated cells in CSF (Fig. 1j-k).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eTogether, these findings define a robust, AD-independent, allele dose-sensitive immune signature in \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4 carriers that spans the plasma and CNS and is enriched in adaptive immune and white matter-resident cell types.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eThe \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4 phenotype extends into the brain, with the superior temporal gyrus especially affected \u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo assess whether the \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4-associated immune phenotype extended into the brain, we analysed TMT-based proteomic data from the Accelerating Medicine Partnerships in AD (AMP-AD) Diverse Cohorts study in the dorsolateral prefrontal cortex (dlPFC; \u003cem\u003en\u003c/em\u003e = 530 AD, \u003cem\u003en\u003c/em\u003e = 190 non-impaired controls) and superior temporal gyrus (STG; \u003cem\u003en\u003c/em\u003e = 161 AD, \u003cem\u003en\u003c/em\u003e = 49 non-impaired controls; Supplementary Table 5). We identified 85 \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4-associated proteins in dlPFC and 56 in STG of non-impaired controls (mutual information \u0026gt; 0.1; Supplementary Table 6), with only two proteins overlapping across regions, GORAB and PHOSPHO1, suggesting region-specific effects of \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4. Coverage differences between SomaScan and TMT precluded direct plasma/CSF comparisons. Random forest modelling showed that \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4 brain proteins reliably classified \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4 carriers versus non-carriers with AD (dlPFC AUC = 0.94; STG AUC = 0.91), independent of sex and race for dlPFC (AUC \u0026gt; 0.88; Extended Data Table 3; Fig. 2a). Insufficient sample size precluded subgroup analysis in STG.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eConsistent with plasma and CSF findings, viral processes were the most significantly enriched \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4 associated biological processes (PANTHER; Fig. 2b; Supplementary Table 7). KEGG analyses revealed significant enrichment for pro-inflammatory, infection-related, and adaptive immune pathways. There were also region-specific pathways including Fc receptor and platelet activation in dlPFC, and TLR and RLR signaling in STG (Fig. 2c; Supplementary Table 7). Reactome pathways overlapped across regions for adaptive immune, B cell receptor, and HIV-related pathways. dlPFC-specific pathways reflected immune cell activation and cytokine regulation whereas STG pathways were more focused on pathogen sensing and antiviral defense (Fig. 2d; Supplementary Table 7). Across both regions, \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4 brain proteins were preferentially enriched in microglia and endothelial cells (Fig. 2e).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThese findings identify distinct, region-specific \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4 brain protein signatures, independent of sex and race, that mirror the viral, pro-inflammatory, and adaptive immune pathway enrichment seen in plasma and CSF. The regional differences suggest nuanced immune and pathogen-sensing functions in \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4 brains, with microglia and cerebrovascular endothelial enrichment implicating white matter and vascular immune processes in the \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4 brain phenotype.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eStem cell-derived cortical organoids show the \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4 phenotype before Alzheimer\u0026rsquo;s pathology onset\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAcross two clinical cohorts, we demonstrated that \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4 carriers exhibit a pro-inflammatory immune phenotype both peripherally and centrally. We previously showed that this phenotype is not correlated with hallmark AD pathologies including amyloid-\u0026beta;, tau, gliosis, or angiopathy \u003csup\u003e2\u003c/sup\u003e. However, since these clinical samples are derived from symptomatic individuals or asymptomatic individuals who may harbor subclinical AD pathology, it remains unclear whether this \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4 immune phenotype precedes or follows the development of AD pathology, a critical question for guiding diagnosis and treatment in \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4 carriers.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eTo address this, we generated cortical organoids \u003csup\u003e15,16\u003c/sup\u003e from induced pluripotent stem cell (iPSC) lines of two donors from the University of Kansas Alzheimer\u0026rsquo;s Disease Research Center cohort: a heterozygous \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4 carrier who developed AD and a non-carrier control. To better model \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4 immune interactions, iPSC-derived microglia from the same donors were integrated into cortical organoids (Methods; Extended Data Fig. 2). At four weeks maturation, neither \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4 nor non-carrier organoids showed evidence of p-tau217 or amyloid-\u0026beta; accumulation (Fig. 3a,b) and no differences in secreted amyloid-\u0026beta;42/40 or p-tau217 (Fig. 3d,f,g,i). By eight weeks, however, \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4 organoids exhibited marked accumulation of p-tau217 (t\u003csub\u003e(2.5)\u003c/sub\u003e = 4.48, \u003cem\u003ep\u003c/em\u003e = 0.0203, unpaired two-tailed t-test; Fig. 3a,c) and amyloid-\u0026beta; (t\u003csub\u003e(2.5)\u003c/sub\u003e = 10.02, \u003cem\u003ep\u003c/em\u003e = 0.0010, unpaired two-tailed t-test; Fig. 3a,c). This was accompanied by significantly reduced levels of secreted amyloid-\u0026beta;42/40 compared to non-carriers (t\u003csub\u003e(8)\u003c/sub\u003e = 2.40, p = 0.043; unpaired two-tailed t-test; Fig. 3e), consistent with amyloid-\u0026beta; retention in \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4 carrier organoids. A significant time \u0026times; \u003cem\u003eAPOE\u003c/em\u003e genotype interaction in amyloid-\u0026beta;42/40 secretion was also observed (F\u003csub\u003e(1,8)\u003c/sub\u003e = 9.12, p = 0.017; two-way ANOVA; Fig. 3f). Despite intracellular p-tau217 accumulation in \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4 organoids at eight weeks, no significant differences were observed in secreted p-tau217 across groups or over time (Fig. 3g-i).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eAt four weeks of maturation, \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4 cortical organoids lacked hallmark AD pathology, providing an opportunity to study early pre-pathological processes. To investigate how the \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4 proteome evolves with disease onset, we performed label-free mass spectrometry on 10,828 proteins from \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4 and non-carrier organoids at four and eight weeks. At four weeks, 642 proteins were differentially expressed between \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4 and non-carriers, increasing to 2,655 by eight weeks (adjusted p \u0026lt; 0.05; Supplementary Table 8; Fig. 4a-b). KEGG enrichment confirmed absence of AD pathway proteins at four weeks, with significant enrichment emerging by eight weeks (FDR = 2.1 \u0026times; 10\u003csup\u003e-6\u003c/sup\u003e Fig. 5c). Consistent with our plasma, CSF, and brain findings, viral processes were the most enriched biological process at both time points (PANTHER; Fig. 4d; Supplementary Table 9). Pathways evident at four weeks were largely exacerbated at eight weeks, as reflected by increased protein representation (Fig. 4e). New eight-week-specific enrichments included pathways related to RNA metabolism, oxidative phosphorylation, lipid metabolism, glycolysis, and stress responses, potentially reflecting changes secondary to accumulating pathology (Fig. 4d; Supplementary Table 9).\u003c/p\u003e\n\u003cp\u003eKEGG immune pathway analysis showed substantial overlap between early and late time points, with progressive intensification over time (Fig. 4f-g; Supplementary Table 9). Early pathways reflected innate immune sensing and antigen presentation, suggesting early APOE \u0026epsilon;4 microglial priming. By eight weeks, pathways were reflective of chronic innate activation, phagocytosis, and adaptive immune responses, consistent with a transition to a disease-associated microglial (DAM) phenotype in response to pathology (Fig. 4f; Supplementary Table 9). Reactome enrichment further supported this finding. Early time points were dominated by NF-\u0026kappa;B, type I interferon, and MHC class I antigen presentation while late time points showed increased pathways for phagocytosis, cytokine signaling, and potential astrocyte involvement via TGF-\u0026beta; signaling (Fig. 4h; Supplementary Table 9). Pathways common to both time points showed progressive activation over time, as evidenced by increased numbers of contributing proteins (Fig. 4i).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eCollectively, these data demonstrate that \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4 cortical organoids display an early pro-inflammatory, viral-related immune phenotype prior to the onset of AD pathology, which progressively intensifies with maturation and disease development. The observed transition from innate immune priming to sustained activation and phagocytic responses is consistent with the emergence of DAM. These results highlight cortical organoids as a valuable model for dissecting temporal immune mechanisms underlying \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4-driven AD pathogenesis.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCross-tissue proteomic analysis reveals systemic as well as central- and peripheral-specific dysregulated pathways in \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4 carriers\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThus far, we independently characterized the proteomes of plasma, CSF, two brain regions, and cortical organoids from \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4 carriers and non-carriers. While broad enrichment of pro-inflammatory immune pathways was observed across compartments, it remained unclear which \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4-associated molecular changes were systemic versus compartment-specific. To address this, we performed a cross-tissue comparison of enriched KEGG and Reactome immune pathways to delineate molecular processes that are systemic, CNS-specific, or peripheral in nature.\u003c/p\u003e\n\u003cp\u003eSystemic \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4 pathways were broadly representative of conserved innate and adaptive immune responses to diverse viral and bacterial infections and pro-inflammatory stimuli, as well as generalized cytokine signaling and antigen presentation (Fig. 5a). Enrichment was also observed for pattern recognition receptor signaling, interferon responses, and MHC class I antigen presentation, core elements of viral sensing, antiviral immunity, and systemic immune activation (Fig. 5a). In contrast, CNS-specific pathways were associated with more complex immune functions, including adaptive immune signaling through T and B cells, intracellular pathogen responses, chronic inflammatory responses via TNF, NF-\u0026kappa;B, and JAK-STAT pathways, and immune modulation (Fig. 5b). Enrichment of pathways involved in immune-microbe interactions at tissue barriers and immune cell recruitment was also observed, potentially reflecting infiltrating adaptive immune cells or blood-brain barrier (BBB)-associated immune processes (Fig. 5b). Notably, five central \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4-enriched pathways were absent in cortical organoids at both time points (Fig. 5b), suggesting that these pathways, particularly those requiring mature T/B cell interactions, antigen presentation, and immune cell crosstalk, cannot be fully recapitulated in organoids due to the absence of adaptive, peripherally derived immune cells. Only four peripheral-only pathways were identified in plasma, primarily involving the fine-tuning and regulation of interferon-driven transcriptional responses (Fig. 5c).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThese results demonstrate that the \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4 pro-inflammatory immune phenotype includes both systemic and CNS-specific components, with core antiviral and innate immune pathways shared across tissues and more complex adaptive immune and BBB-associated processes enriched in the CNS. The absence of certain adaptive pathways in organoids further underscores the importance of immune cell interactions in shaping the \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4 CNS immune environment. Together, these findings highlight a multi-tissue, pro-inflammatory immune phenotype that may contribute to \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4-driven AD risk.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eShort-term treatment with an anti-inflammatory ketogenic diet can modulate the pro-inflammatory phenotype of \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4 carriers\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;\u0026nbsp;\u003c/strong\u003e\u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4 carriers have a pro-inflammatory phenotype that spans both the periphery and CNS. To test whether this phenotype is modifiable, we analyzed plasma samples from a 12-week trial of a medical ketogenic diet in 15 individuals with mild AD (\u003cem\u003en\u003c/em\u003e = 9 \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4 carriers, n = 6 non-carriers; Supplementary Table 10). Medical ketogenic diets, focused on increasingly quality fat intake and reducing carbohydrate consumption to promote production of ketone bodies are widely regarded as a gold-standard systemic anti-inflammatory diet \u003csup\u003e17\u003c/sup\u003e. They have demonstrated treatment benefits for epilepsy, general neurological conditions including AD, and cancers \u003csup\u003e18-21\u003c/sup\u003e. The intervention increased serum \u0026beta;-hydroxybutyrate in both groups, confirming ketone body induction (Supplementary Table 10).\u003c/p\u003e\n\u003cp\u003ePlasma proteomic profiling (7,595 proteins via SomaScan 7k) across baseline and study endpoint revealed no significant changes in non-carriers (Fig. 6a). In contrast, \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4 carriers had 264 differentially expressed proteins (161 downregulated, 103 upregulated; adjusted \u003cem\u003ep\u003c/em\u003e \u0026lt; 0.05; Fig. 6b; Supplementary Table 11). PANTHER analysis identified viral processes as the top enriched biological process across both up- and downregulated proteins (Fig. 6c; Supplementary Table 12). A KEGG immune pathway analysis revealed that the medical ketogenic diet led to immune remodelling in \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4 carriers rather than broad suppression or activation. Pathways enriched for both up- and downregulated proteins included those involved in innate and adaptive pathways (Fig. 6d). Pathways enriched in upregulated proteins were associated with adaptive immunity, phagocytosis, and chemotaxis, suggesting enhanced immune surveillance. Downregulated proteins, however, mapped to chronic inflammation and metabolic-immune signaling, indicating reduced chronic pro-inflammatory signaling (Fig. 6d; Supplementary Table 12). These findings were mirrored by a Reactome pathway analysis. This showed a bidirectional regulation of TLR and cytokine signaling, with selective upregulation of interferon responses and Fc-mediated phagocytosis, and downregulation of NF-\u0026kappa;B, TGF-\u0026beta;, TCR, and IL-1 signaling pathways (Fig. 6e; Supplementary Table 12). Cell type enrichment analysis showed that both up- and downregulated proteins were mapped to T-regs, NK cells, memory CD8\u003csup\u003e+\u003c/sup\u003e T cells, and non-V\u0026delta;2 \u0026gamma;\u0026delta; T cells, suggesting functional reprogramming within these populations rather than uniform activation or suppression (Fig. 6f,g). Notably, both protein sets were also enriched in white matter, highlighting that the CNS may have similar immune shifts in response to the ketogenic diet.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eTaken together, these findings indicate that a medical ketogenic diet leads to reprogramming rather than uniform activation or suppression of immune cell populations. Protein expression patterns and pathway analyses suggest a shift in T-regs and NK cells from pro-inflammatory or cytotoxic states toward more regulatory, tissue supportive phenotypes. Similarly, memory CD8\u003csup\u003e+\u003c/sup\u003e T cells likely have reduced signatures of chronic activation alongside enhanced effector memory features, consistent with restored adaptive immune competence. Non-V\u0026delta;2 \u0026gamma;\u0026delta; T cells also displayed bidirectional modulation, suggesting altered roles at the innate-adaptive interface and reduced IL-17-associated inflammatory signaling.\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eOur study provides the most comprehensive cross-tissue characterization to date of the pro-inflammatory immune phenotype associated with \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4, integrating large-scale clinical proteomic data with experimental validation in iPSC-derived cortical organoids and a dietary intervention trial. We demonstrate that \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4 carriers exhibit a conserved, allele dose-dependent pro-inflammatory immune signature across plasma, CSF, and brain tissue, independent of AD status. Importantly, this immune phenotype was also recapitulated in cortical organoids, emerging prior to the development of amyloid-\u0026beta; or tau pathology. This indicates that key features of the \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4 immune signature arise early in disease pathogenesis and reflect an intrinsic, genotype-driven mechanism of immune dysregulation. Cross-tissue analyses further revealed a core set of systemic immune pathways involving broad innate and adaptive responses to infection, alongside distinct tissue-specific signatures, including inflammatory signaling and BBB-related interactions within the CNS. Notably, we show that this immune phenotype is not fixed. Short-term treatment with a medical ketogenic diet in \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4 carriers with AD partially reversed pro-inflammatory signatures, promoting more regulatory and tissue-supportive immune states. Together, these findings illuminate how peripheral and central tissue environments modulate \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4-driven immune responses, provide new mechanistic insight into AD pathogenesis, and support the development of precision biomarkers and immunomodulatory interventions targeting early-stage disease processes.\u003c/p\u003e\n\u003cp\u003eOur results are consistent with and extend prior work. Multiple studies have reported heightened peripheral immune responses in \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4 carriers, including increased cytokine production following TLR2/4/5 activation \u003csup\u003e3\u003c/sup\u003e, elevated TNF-\u0026alpha; and IL-6 levels after a lipopolysaccharide (LPS) challenge \u003csup\u003e4\u003c/sup\u003e, increased cytokine levels \u003csup\u003e6,22-24\u003c/sup\u003e, and enhanced susceptibility to systemic inflammatory stressors such as sepsis \u003csup\u003e4\u003c/sup\u003e. Similarly, transcriptomic and epigenomic profiling of PBMCs has demonstrated broad dysregulation of innate immune pathways, altered NF-\u0026kappa;B activation, increased chromatin accessibility in CD14\u003csup\u003e+\u003c/sup\u003e monocytes, and clonally expanded CD8\u003csup\u003e+\u003c/sup\u003e effector memory T cells \u003csup\u003e25\u003c/sup\u003e. Our plasma \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4 proteomic signature mirrors these findings, with robust enrichment for innate immune and antiviral responses, including TLR-NF-\u0026kappa;B and interferon signaling cascades. Moreover, we observed allele dose-dependent effects, consistent with prior reports of exacerbated immune dysfunction in homozygous \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4 carriers \u003csup\u003e3,25\u003c/sup\u003e. Notably, our findings also align with the emerging hypothesis that \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4 carriers exhibit increased susceptibility to viral reactivation and higher viral titers and that this plays a role in the development of AD \u003csup\u003e26\u003c/sup\u003e. The strong enrichment of viral response pathways we find across plasma, CSF, and brain in our study supports this hypothesis. Prior studies measuring C-reactive protein (CRP) levels in \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4 carriers have yielded mixed results, with some reporting decreased baseline CRP \u003csup\u003e22,27-29\u003c/sup\u003e, while others report increased associated brain atrophy \u003csup\u003e30\u003c/sup\u003e. Our findings suggest that \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4-driven chronic immune dysregulation is not adequately captured by individual markers of acute inflammation such as CRP. A limitation of our study is the lack of longitudinal data. Although we show that the\u003cem\u003e\u0026nbsp;APOE\u003c/em\u003e \u0026epsilon;4 immune phenotype is present in cognitively unimpaired individuals, future work should investigate whether this signature evolves with disease progression in \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4 carriers.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eIn the CNS, our findings offer new mechanistic insight into how \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4 may prime resident immune cells, particularly microglia, toward a disease-associated phenotype. Prior studies have demonstrated that \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4 microglia exhibit impaired phagocytosis \u003csup\u003e11\u003c/sup\u003e, disrupted autophagy with lipid droplet accumulation \u003csup\u003e9,10\u003c/sup\u003e, and exaggerated inflammatory responses to amyloid-\u0026beta; \u003csup\u003e10,11\u003c/sup\u003e. Our CSF and brain proteomic data confirm upregulation of inflammatory pathways (NF-\u0026kappa;B, TNF signaling), consistent with a chronically activated or DAM-like state and reveal enrichment of BBB-related immune pathways. These findings align with evidence of \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4-associated BBB dysfunction \u003csup\u003e31,32\u003c/sup\u003e and increased infiltration of peripheral immune cells such as IL-17\u003csup\u003e+\u003c/sup\u003e neutrophils \u003csup\u003e7\u003c/sup\u003e. Notably, we observed stronger pro-inflammatory enrichment in the STG compared to the dlPFC. This likely reflects the greater vascularization and CSF-blood interface in the STG, which may facilitate immune interactions.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eA major strength of our study is the use of patient-derived cortical organoids to model \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4-driven immune changes in human neural tissue before AD pathology emerges. While prior studies of \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4 immunopathology have largely focused on peripheral immune cells \u003csup\u003e3-7,22-25,30,33,34\u003c/sup\u003e, postmortem brain tissue \u003csup\u003e8,35\u003c/sup\u003e, or single iPSC-derived cell types \u003csup\u003e9-11\u003c/sup\u003e, our organoid model enables longitudinal study of \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4 effects within a more complex network. We experimentally confirmed early activation of inflammatory pathways, including NF-\u0026kappa;B and DAM-like signatures, prior to amyloid-\u0026beta; or tau pathology. This provides direct evidence that \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4 immune dysregulation is intrinsic and genotype-driven, rather than a secondary response to existing AD pathology. Moreover, the fact that\u003cem\u003e\u0026nbsp;APOE\u003c/em\u003e \u0026epsilon;4 immune activation precedes amyloid-\u0026beta; and tau accumulation supports the view that these hallmark pathologies may be downstream consequences rather than primary disease drivers in \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4 carriers. These insights challenge current therapeutic paradigms focused solely on amyloid-\u0026beta; and tau, targets that, despite effective clearance, have shown limited clinical benefit and in some cases worsening disease \u003csup\u003e36-39\u003c/sup\u003e. Instead, our findings suggest that tempering the persistent \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4-driven immune response through immunomodulation or anti-inflammatory approaches will be critical for altering disease risk and progression in this high-risk population. Importantly, this aligns with emerging evidence that vaccinations reduce AD risk \u003csup\u003e40,41\u003c/sup\u003e, including the shingles vaccines Zostavax \u003csup\u003e42\u003c/sup\u003e and Shingrix \u003csup\u003e43\u003c/sup\u003e, potentially by reducing reactivation of varicella zoster virus and/or inducing a virus-specific immunomodulatory effect \u003csup\u003e42\u003c/sup\u003e. Targeting immune dysregulation should therefore be prioritized in future precision medicine strategies for AD.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eOur finding that a medical ketogenic diet partially reversed the \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4 immune phenotype has important implications for therapeutic development. Twelve weeks of treatment led to the downregulation of chronic inflammatory signaling pathways, including NF-\u0026kappa;B, IL-1, and TGF-\u0026beta; and upregulation of regulatory and phagocytic functions across T-regs, NK cells, and memory CD8\u003csup\u003e+\u003c/sup\u003e T cells. These changes suggest a shift away from maladaptive, chronic inflammation toward a more balanced and functional immune state. This supports the idea that immune modulation may be a viable disease-modifying strategy, particularly if implemented early, before irreversible pathology has accumulated. Future studies would benefit from larger-scale clinical trials of a medical ketogenic diet as well as other immunomodulatory therapies before the onset of AD in \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4 carriers.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eOne limitation of our study is that cross-platform differences limited direct comparisons of APOE \u0026epsilon;4 tissue proteomes. Plasma and CSF were profiled using SomaScan aptamer-based technology, while TMT and label-free mass spectrometry were used for brain and organoid analyses, respectively. As a result, we focused on pathway-level rather than protein-level comparisons across samples. However, this limitation is also a strength. The consistency of pathway-level findings across multiple platforms provides orthogonal validation of the robustness and generalizability of our results. Moreover, our study demonstrates that aptamer-based approaches, despite known limitations in proteoform sensitivity \u003csup\u003e44\u003c/sup\u003e, can reliably capture \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4-related pathway alterations. Future work using harmonized proteomic platforms will further refine these insights.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eIn conclusion, our findings establish that \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4 drives a conserved pro-inflammatory immune phenotype that emerges early, prior to the development of hallmark AD pathology. These results advance the mechanistic understanding of \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4-mediated AD risk and challenge the prevailing amyloid-\u0026beta;- and tau-centric therapeutic approaches, especially for \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4 carriers. By demonstrating that immune dysregulation is an intrinsic, upstream feature of \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4, and showing that it is modifiable, our study underscores the need to prioritize immunomodulatory and anti-inflammatory strategies in precision medicine approaches for this high-risk group. Moving forward, longitudinal studies and harmonized proteomic platforms will be critical to further delineate the temporal evolution of the \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4 immune signature and to guide the development of targeted interventions aimed at altering disease trajectories in \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4 carriers.\u003c/p\u003e"},{"header":"Methods","content":"\u003cp\u003e\u003cstrong\u003eParticipants\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eGlobal Neurodegeneration Proteomics Consortium (GNPC) cohort.\u0026nbsp;\u003c/strong\u003eThe GNPC cohort is the largest collection of proteomic data for individuals with neurodegenerative diseases and non-impaired controls from \u0026gt;20 clinical sites from across the USA, UK, and Europe. In the current study, we included non-impaired controls (n = 6,672) and individuals with AD (n = 3,455). Across both groups, n = 6,107 were non-\u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4 carriers and n = 4,767 had at least one \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4 allele (Supplementary Table 1). AD was clinically diagnosed within each study site as previously described including Clinical Dementia Rating (CDR) score of \u003cu\u003e\u0026gt;\u003c/u\u003e 1, Mini-Mental State Exam (MMSE) \u003cu\u003e\u0026lt;\u003c/u\u003e 24, and/or Montreal Cognitive Assessment (MOCA) score of \u003cu\u003e\u0026lt;\u003c/u\u003e 23 \u003csup\u003e45\u003c/sup\u003e. Individual plasma and CSF samples, demographic, and clinical variables were collected at a single timepoint (Supplementary Table 1). Participants from each study site in the GNPC cohort provided written informed consent and studies were approved by the relevant institution\u0026rsquo;s ethics committee \u003csup\u003e45\u003c/sup\u003e.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAccelerating Medicines in AD Partnership (AMP-AD) Diverse Cohorts study.\u0026nbsp;\u003c/strong\u003eThe AMP-AD Diverse Cohorts study is a cross-consortium project that created harmonized high-throughput multi-omic data for post-mortem brain tissue samples. AD cases were pathologically defined as described elsewhere \u003csup\u003e46\u003c/sup\u003e. Samples from the dorsolateral prefrontal cortex (dlPFC) and superior temporal gyrus (STG) were collected across four institutions: Mayo Clinic, Rush University, Mount Sinai University Hospital, and Emory University. In the present study, we included dlPFC data from \u003cem\u003en\u003c/em\u003e = 190 non-impaired controls and \u003cem\u003en\u003c/em\u003e = 530 AD cases. For the STG, \u003cem\u003en\u003c/em\u003e = 49 non-impaired controls and \u003cem\u003en\u003c/em\u003e = 161 AD cases were included. Supplementary Table 6 lists the demographic characteristics for the included donors. Participants from each study site in the AMP-AD Diverse Cohorts study provided written informed consent and studies were approved by the relevant institution\u0026rsquo;s ethics committee.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eUniversity of Kansas Alzheimer\u0026rsquo;s Disease Research Center donors for cortical organoids.\u0026nbsp;\u003c/strong\u003eTwo donors from the University of Kansas Alzheimer\u0026rsquo;s Disease Research Center provided post-mortem skin samples. The non-carrier donor was a 72-year-old female with an\u0026nbsp;\u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;3/\u0026epsilon;2 genotype and\u0026nbsp;\u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4\u0026nbsp;carrier donor was an 84-year-old female with an\u0026nbsp;\u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;3/\u0026epsilon;4 genotype. \u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTherapeutic Diets in Alzheimer\u0026rsquo;s Disease (TDAD) clinical trial.\u0026nbsp;\u003c/strong\u003eThe TDAD clinical trial is a single-blind, randomized, controlled study of the effects of a 12-week medical ketogenic dietary intervention in patients with mild AD. Participants were recruited through the Clinical Translational Science Unit at the University of Kansas Medical Center Clinical Research Center. Fifteen participants (\u003cem\u003en\u003c/em\u003e = 9 \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4 carriers and \u003cem\u003en\u003c/em\u003e = 6 non-carriers) were given a ketogenic diet intervention. All participants met the criteria for probable AD based on the National Institutes of Aging-Alzheimer\u0026rsquo;s Association (NIA-AA) criteria \u003csup\u003e47\u003c/sup\u003e and had a baseline CDR of 0.5 to 1 (see Supplementary Table 10 for demographics). Participants with moderate to severe AD, including those residing in a nursing home or dementia special care unit, were excluded from the trial based on our previous experience showing that these individuals have a low retention rate to dietary interventions \u003csup\u003e21\u003c/sup\u003e. Participants with serious medical conditions, those participating in another clinical trial, women of child-bearing capacity who seek to become pregnant, and non-English speakers were also excluded. Table 1 lists the inclusion and exclusion criteria for the TDAD clinical trial.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMethods Table 1.\u003c/strong\u003e Therapeutic Diets in Alzheimer\u0026rsquo;s Disease (TDAD) clinical trial inclusion and exclusion criteria.\u0026nbsp;\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"6\" valign=\"top\" style=\"width: 151px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003eInclusion Criteria\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 450px;\"\u003e\n \u003cp\u003e1. Diagnosis of AD by NIA-AA criteria \u003csup\u003e47\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 450px;\"\u003e\n \u003cp\u003e2. CDR global score of 0.5 or 1\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 450px;\"\u003e\n \u003cp\u003e3. Agreed cooperation from an appropriate study partner\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 450px;\"\u003e\n \u003cp\u003e4. Speaks English as a primary language\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 450px;\"\u003e\n \u003cp\u003e5. Age 50 to 90\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 450px;\"\u003e\n \u003cp\u003e6. No medication changes within the past 30 days\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"5\" valign=\"top\" style=\"width: 151px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003eExclusion Criteria\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 450px;\"\u003e\n \u003cp\u003e1. Resides in a nursing home or dementia special care unit, or cannot control diet\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 450px;\"\u003e\n \u003cp\u003e2. A potentially confounding serious medical risk including insulin-requiring diabetes, cancer requiring chemotherapy or radiation within the past five years, or recent cardiac event (e.g. heart attack, angioplasty, etc.)\u003c/p\u003e\n \u003cp\u003e3. Participating in another clinical trial or using an investigational drug or therapy within 30 days of the screening visit\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 450px;\"\u003e\n \u003cp\u003e4. A history of renal stones\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 450px;\"\u003e\n \u003cp\u003e5. Women with child-bearing capacity who seek to become pregnant\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 450px;\"\u003e\n \u003cp\u003e6. Non-English speakers\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eKetogenic diets focus on increasing quality fat intake and reducing carbohydrate consumption to promote the production of ketone bodies, mimicking the effects of prolonged fasting and vigorous exercise \u003csup\u003e17\u003c/sup\u003e. Macronutrients were maintained at approximately 70% fat, 20% protein, and \u003cu\u003e\u0026lt;\u003c/u\u003e 10% carbohydrates. A 1:1 emulsified medium chain triglyceride (MCT) oil was also provided to participants in the ketogenic diet group to increase fat intake, enhance ketosis, and increase palatability in line with other medical ketogenic diets \u003csup\u003e21\u003c/sup\u003e. To reduce gastrointestinal side effects from the MCT, the supplement dose increased on a weekly basis across a four-week period, starting from an intake level corresponding to 10% of total fat energy and gradually increasing to 40%. The Mifflin-St. Jeor equation was used to calculate target energy goals \u003csup\u003e48\u003c/sup\u003e. All participants received weekly dietary counselling from the study dietitian. Participants also received dietitian-created ketogenic diet educational material and a manual. Manuals included sample menus, recipes, and strategies for maintaining their respective diet during travel or holidays.\u003c/p\u003e\n\u003cp\u003eParticipants\u0026rsquo; APOE genotype, medical history, medication use, comorbidities, smoking, and alcohol intake were recorded at their initial screening visit. Blood samples were taken at two time points, a pre-intervention baseline and post-intervention (12 weeks), for plasma proteomics and to measure depth of ketosis using a routine serum \u0026beta;-hydroxybutyrate assay performed commercially by Quest Diagnostics (Lenexa, KS). Participant demographic characteristics are listed in Supplementary Table 10. \u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe TDAD clinical trial was approved by the University of Kansas Center IRB (STUDY00143457). Prior to participating in the study, all participants provided informed consent. The trial was registered on ClinicalTrials.gov under registration number NCT03860792.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCortical Organoids\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eInduced pluripotent stem cell derivation.\u0026nbsp;\u003c/strong\u003eInduced pluripotent stem cells (iPSCs) were generated from fibroblast samples from both donors using a CytoTune-iPSC 2.0 Sendai Reprogramming kit (ThermoFisher Scientific, A16517) according to the manufacturer\u0026rsquo;s instructions. iPSCs were plated down into 6-well plates pre-treated with 0.08mg/ml Matrigel (Corning, 356234) in Dulbecco\u0026rsquo;s Phosphate-Buffered Saline (DPBS). StemFlex medium (Gibco, A3349401) with 100U/ml penicillin-streptomycin added (Gibco, 15140122) was added to each well and changed every two days. iPSCs were passaged once a week using ReLeSR (StemCell Technologies, 15140122) as per the manufacturer\u0026rsquo;s protocol. Following each passage, medium was supplemented with 0.01mM Y-27632 (StemCell Technologies, 72302) for 24 hours.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eiPSC-derived microglia differentiation.\u0026nbsp;\u003c/strong\u003eiPSCs were differentiated into hematopoietic progenitor cells (HPCs) using a STEMdiff Hematopoietic kit (StemCell Technologies, 5310) as per the manufacturer\u0026rsquo;s instructions. First, iPSCs were passaged into HPC Medium A in 24-well plates pre-treated with 0.04mg/ml Matrigel. After three days, they were cultured with HPC Medium B for nine more days. Mature HPCs were then collected and passaged into STEMDiff Microglia Differentiation medium (StemCell Technologies, 100-0019) as per the manufacturer\u0026rsquo;s instructions. Fresh media was added every other day for 24 days and cells were passaged once on day 12. After differentiation, microglia were matured in BrainPhys medium with N2-A and SM1 supplements (StemCell Technologies, 05793) for four days prior to integration into embryoid bodies.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCortical organoid development.\u0026nbsp;\u003c/strong\u003eiPSCs were first passaged three times to stabilize the line. Embryoid bodies were generated in ultra-low attachment dishes (Corning, 3261) using StemFlex medium supplemented with 100U/ml penicillin-streptomycin and 0.01mM y-27632 ROCK inhibitor. Media was changed every two days. Embryoid body medium was replaced into N2/B27 medium (Methods Table 2) and transferred into a 24-well plate. Mature microglia were added to each well at a concentration of 0.8 x 10\u003csup\u003e5\u003c/sup\u003e cells per well. Fresh N2/B27 media was added every two days and organoids were passaged once every two weeks. Cortical organoids were aged for either four or eight weeks.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMethods Table 2.\u003c/strong\u003e List of N2/B27 media components.\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 208px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eReagent\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 189px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eConcentration (in 1000ml)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 204px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eSource\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 208px;\"\u003e\n \u003cp\u003eDMEM/F12\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 189px;\"\u003e\n \u003cp\u003e478ml\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 204px;\"\u003e\n \u003cp\u003eGibco, 11320033\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 208px;\"\u003e\n \u003cp\u003eNeurobasal medium\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 189px;\"\u003e\n \u003cp\u003e480ml\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 204px;\"\u003e\n \u003cp\u003eGibco, 21103049\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 208px;\"\u003e\n \u003cp\u003eN2 supplement\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 189px;\"\u003e\n \u003cp\u003e5ml\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 204px;\"\u003e\n \u003cp\u003eGibco, 17502048\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 208px;\"\u003e\n \u003cp\u003eHuman insulin\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 189px;\"\u003e\n \u003cp\u003e2.5ug/ml\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 204px;\"\u003e\n \u003cp\u003eSigma-Aldrich, I9278-5ML\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 208px;\"\u003e\n \u003cp\u003eMEM NEAA\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 189px;\"\u003e\n \u003cp\u003e5ml\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 204px;\"\u003e\n \u003cp\u003eGibco, 11140050\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 208px;\"\u003e\n \u003cp\u003eBeta-mercaptoethanol\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 189px;\"\u003e\n \u003cp\u003e25mM\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 204px;\"\u003e\n \u003cp\u003eGibco, 21985023\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 208px;\"\u003e\n \u003cp\u003eB27 (without vitamin A)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 189px;\"\u003e\n \u003cp\u003e10ml\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 204px;\"\u003e\n \u003cp\u003eGibco, 12587010\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 208px;\"\u003e\n \u003cp\u003eGlutaMAX supplement\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 189px;\"\u003e\n \u003cp\u003e10ml\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 204px;\"\u003e\n \u003cp\u003eGibco, 35050061\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 208px;\"\u003e\n \u003cp\u003ePenicillin-streptomycin\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 189px;\"\u003e\n \u003cp\u003e100u/ml\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 204px;\"\u003e\n \u003cp\u003eGibco, 15140122\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 208px;\"\u003e\n \u003cp\u003eDosomorphin\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 189px;\"\u003e\n \u003cp\u003e1uM\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 204px;\"\u003e\n \u003cp\u003eStemCell Technologies, 72102\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 208px;\"\u003e\n \u003cp\u003eSB431542\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 189px;\"\u003e\n \u003cp\u003e10uM\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 204px;\"\u003e\n \u003cp\u003eStemCell Technologies, 72234\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 208px;\"\u003e\n \u003cp\u003eCHIR99021\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 189px;\"\u003e\n \u003cp\u003e3uM\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 204px;\"\u003e\n \u003cp\u003eStemCell Technologies, 72054\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 208px;\"\u003e\n \u003cp\u003eThiazovivin\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 189px;\"\u003e\n \u003cp\u003e2uM\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 204px;\"\u003e\n \u003cp\u003eStemCell Technologies, 72254\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eImmunofluorescent staining.\u0026nbsp;\u003c/strong\u003eAt each timepoint, cortical organoids were removed from the media and washed in DPBS three times for 10 minutes. They were then fixed in 4% paraformaldehyde (Sigma-Aldrich, 158127) for 20 minutes, washed in DPBS once for 10 minutes, and stored in cyroprotectant (30% glycerol (Sigma-Aldrich, G5516), 30% ethylene glycol (Sigma-Aldrich, 324558), 40% PBS) at -20\u003csup\u003eo\u003c/sup\u003eC. Organoids were washed in PBS three times for 10 minutes and incubated with blocking solution (5% donkey serum (Sigma-Aldrich, D9663) or 5% bovine serum albumin (BSA), 0.1% Triton X-100 (Sigma-Aldrich, #X100), and PBS) overnight at 4\u003csup\u003eo\u003c/sup\u003eC. Organoids were then incubated with primary antibodies (Methods Table 3) made up in blocking solution for two days at 4\u003csup\u003eo\u003c/sup\u003eC. They were then washed three times with PBS for 10 minutes followed by an overnight incubation with secondary antibodies (Methods Table 3) made up in blocking solution at 4\u003csup\u003eo\u003c/sup\u003eC. This process was repeated for each primary antibody. The MAP2, GFAP, and IBA1 stained organoids (Extended Data Fig. 2) were co-stained for DAPI using NucBlue Live ReadyProbes (Invitrogen, R37605) as per the manufacturer\u0026rsquo;s instructions and incubated for three days at 4\u003csup\u003eo\u003c/sup\u003eC. The amyloid-\u0026beta; and p-tau217 organoids (Fig. 3) were co-stained for DAPI using ProLong Gold AntiFade with DAPI (Invitrogen, P36931).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMethods Table 3.\u003c/strong\u003e List of primary and secondary antibodies.\u0026nbsp;\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 236px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eAntibody\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 113px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eConcentration\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 251px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eSource\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 236px;\"\u003e\n \u003cp\u003eRabbit anti-MAP2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 113px;\"\u003e\n \u003cp\u003e1:250\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 251px;\"\u003e\n \u003cp\u003eInvitrogen, PA5-17646\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 236px;\"\u003e\n \u003cp\u003eGoat anti-IBA1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 113px;\"\u003e\n \u003cp\u003e1:500\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 251px;\"\u003e\n \u003cp\u003eAbcam, ab5076\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 236px;\"\u003e\n \u003cp\u003eRabbit anti- GFAP\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 113px;\"\u003e\n \u003cp\u003e1:250\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 251px;\"\u003e\n \u003cp\u003eInvitrogen, PA5-16291\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 236px;\"\u003e\n \u003cp\u003eRabbit anti- pTau217\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 113px;\"\u003e\n \u003cp\u003e1:200\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 251px;\"\u003e\n \u003cp\u003eInvitrogen, 44-744\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 236px;\"\u003e\n \u003cp\u003eMouse anti-APP6E10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 113px;\"\u003e\n \u003cp\u003e5 \u0026mu;g/ml\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 251px;\"\u003e\n \u003cp\u003eBioLegend, SIG-39320\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 236px;\"\u003e\n \u003cp\u003eDonkey anti-goat Cy3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 113px;\"\u003e\n \u003cp\u003e1:100\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 251px;\"\u003e\n \u003cp\u003eJackson Immunoresearch, 705-165-147\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 236px;\"\u003e\n \u003cp\u003eDonkey anti-rabbit AlexaFluor 488\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 113px;\"\u003e\n \u003cp\u003e1:100\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 251px;\"\u003e\n \u003cp\u003eInvitrogen, A21206\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 236px;\"\u003e\n \u003cp\u003eDonkey anti-rabbit AlexaFluor 647\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 113px;\"\u003e\n \u003cp\u003e1:100\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 251px;\"\u003e\n \u003cp\u003eInvitrogen, A31573\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 236px;\"\u003e\n \u003cp\u003eDonkey anti-rabbit Cy3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 113px;\"\u003e\n \u003cp\u003e1:100\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 251px;\"\u003e\n \u003cp\u003eJackson Immunoresearch, 711-165-152\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 236px;\"\u003e\n \u003cp\u003eDonkey anti-mouse AlexaFluor 555\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 113px;\"\u003e\n \u003cp\u003e1:100\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 251px;\"\u003e\n \u003cp\u003eInvitrogen, A31570\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003ePrior to imaging, stained cortical organoids were washed three times in PBS for 10 minutes and incubated with CytoVista 3D Cell Culture clearing reagent (Invitrogen, V11315) overnight at 4\u003csup\u003eo\u003c/sup\u003eC. To visualize and quantify amyloid-\u0026beta; and p-tau217 staining, whole cortical organoids were imaged using a Nikon TI2-E inverted microscope attached to a Yokagawa CSU-W1 spinning disk confocal with SoRa super-resolution. Images were acquired with a Hamamatsu Fusion BT camera using Nikon NIS-Elements Advanced Research software. To visualize and quantify MAP2, IBA1, and GFAP staining, whole cortical organoids were imaged using a Marianas 3i Spinning Disk Confocal microscope with Super-Resolution by Optical Re-Assignment (SoRa). Representative images were taken using Fiji ImageJ v1.54p \u003csup\u003e49\u003c/sup\u003e.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eELISA for secreted A\u0026beta;\u003csub\u003e42\u003c/sub\u003e/A\u0026beta;\u003csub\u003e40\u0026nbsp;\u003c/sub\u003eand p-tau217.\u0026nbsp;\u003c/strong\u003eMedia was collected from cortical organoids at four and eight weeks of age and proteins were extracted using acetone precipitation. Briefly, 1ml of ice-cold acetone (cooled to -20\u003csup\u003eo\u003c/sup\u003eC) was added to 0.5ml of collected media in a tube. Tubes were vortexed, incubated for 60 minutes at -20\u003csup\u003eo\u003c/sup\u003eC, and then centrifuged for 10 minutes at 15,000 x g. The supernatant was removed and pellets were washed in 70% ethanol before being resuspended in 8M urea. Secreted human amyloid-\u0026beta;\u003csub\u003e40\u003c/sub\u003e and amyloid-\u0026beta;\u003csub\u003e42\u003c/sub\u003e were measured with an ELISA as per the manufacturer\u0026rsquo;s instructions (ThermoFisher, KHB3481 and KHB3544). Secreted human p-tau217 was also measured with an ELISA as per the manufacturer\u0026rsquo;s instructions (Cell Signaling Technology, 59672). All samples were diluted 1:5 and concentrations were normalized to protein content measured via Pierce BCA protein assay (ThermoFisher, 23225). Graphs were made in GraphPad Prism v10.5.0.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eProteomics\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eGNPC cohort SomaScan assay.\u0026nbsp;\u003c/strong\u003eProteomics of plasma and CSF samples from the GNPC cohort was done using the SomaScan v4.1 assay (SomaLogic). The SomaScan assay detects approximately 7,000 proteins using aptamer-based technology called slow off-rate modified aptamers (SOMAmers). These contain chemically modified nucelotides that bind with high specificity and affinity to target proteins \u003csup\u003e50,51\u003c/sup\u003e. Raw data is provided by SomaLogic following standardization, normalization, and calibration, including adaptive normalization by maximum likelihood (ANML). Protein measurements are provided in relative fluorescent units (RFU). Prior to their inclusion in the GNPC dataset, aptamers are mapped to Uniprot. Details on the creation and harmonization of the GNPC dataset are described elsewhere \u003csup\u003e45\u003c/sup\u003e. Data from the GNPC dataset was log2 transformed and standardized prior to analyses in the current study and was done separately on training and testing datasets. \u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAMP-AD Diverse Cohorts study TMT quantitation.\u0026nbsp;\u003c/strong\u003eProteomics on post-mortem dlPFC and STG homogenates was done using TMT mass spectrometry as previously described \u003csup\u003e46\u003c/sup\u003e. Using a conservative approach, we only included proteins with \u003cu\u003e\u0026lt;\u003c/u\u003e30% missing values across samples and imputed these values using the median. Batch effects were accounted for by fitting a linear regression model, as in the original study \u003csup\u003e46\u003c/sup\u003e.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCortical organoids label free mass spectrometry.\u0026nbsp;\u003c/strong\u003eAt 4 and 8 weeks of age, cortical organoids were snap frozen in liquid nitrogen and stored at -20 \u003csup\u003eo\u003c/sup\u003eC. To prepare samples for mass spectrometry, 0.2% n-dodecyl-\u0026beta;-D-maltoside (DDM) in 50 mM triethylammonium bicarbonate (TEAB) with 5 mM Tris(2-carboxyethyl)phosphine (TCEP) was added to each organoid. A pestle suitable for a 0.5 mL tube attached to a drill was used to homogenize the organoids for 10 s. The tubes were incubated at 85 \u0026deg;C for 10 min, then cooled in incubated with 10 mM iodoacetamide for 30 min at 22 \u0026deg;C. The samples were frozen in dry ice and then lyophilized. To each sample was added 10 \u0026micro;L of 0.1% DDM in 50 mM TEAB with 2.5 mM TCEP, 0.5 ug of LysC (WAKO Fujifilm) and 1 ug of trypsin (Sigma, trypzean). The samples were incubated at 42 \u0026deg;C for 2 h, then a further 0.5 ug of trypsin was added and incubated at 33 \u0026deg;C for 4 h. Each sample was acidified with 0.2 \u0026micro;L of formic acid and diluted with 30 \u0026micro;L of 0.1% trifluoroacetic acid and then desalted using the STAGEtip method \u003csup\u003e52\u003c/sup\u003e.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eEach sample was analyzed by LC-MS/MS using the Vanquish Neo system and Astral Orbitrap mass spectrometer (ThermoFisher Scientific). Samples were loaded onto a Pepmap Neo C18 5 \u0026micro;m particle 5 mm long by 300 \u0026micro;m inside diameter trap column (ThermoFisher Scientific) at up to 10 \u0026micro;L/min and at a maximum pressure of 800 bar. They were then eluted through a 15 cm long 150 \u0026micro;m inside diameter Pepmap C18 EASY-spray column with 2 \u0026mu;m 100 \u0026Aring; particles (ThermoFisher Scientific). The column was heated to 40 \u0026deg;C using the EASY-spray ion source operating a 1.9 kV. The S lens radio frequency level was 40 and capillary temperature was 280 \u0026deg;C.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe liquid chromatography used buffer A (solution of 0.1% formic acid) and buffer B (0.1% formic acid and 99.9% acetonitrile). After loading the sample in 3.6% buffer A, the gradient at 1 \u0026micro;L/min was from 7.2% to 25.2% buffer B in 19.7 min, to 31.5% buffer B in 3.7 min, to 49.5% buffer B in 0.4 min and to 90% buffer B at 3 \u0026micro;L/min in 0.5 min and held for 0.7 min. MS acquisition was for 26 min. The MS scans in the orbitrap analyzer were at a resolution of 240,000 with automatic gain control set to 5,000,000 and a maximum ion time of 3 ms for m/z 380 to 980. The data-independent acquisition MS/MS scans with a window of 2 m/z in the Astral analyser had automatic gain control at 50,000 for a maximum ion time of 3 ms. The loop was controlled to 0.6 seconds. The MS/MS scan range was 150-2000 m/z. The normalized collision energy was 25.\u003c/p\u003e\n\u003cp\u003eThe raw LC-MS/MS data was processed with DIA-NN v1.9.2 \u003csup\u003e53\u003c/sup\u003e. The \u003cem\u003eHomo Sapiens\u003c/em\u003e reference proteome downloaded Feb 24 2025 with 20,644 genes, using canonical sequences only, was used to create an \u003cem\u003ein silico\u003c/em\u003e library for peptide-spectrum matching. N-terminal methionine excision was allowed. Carbamidomethyl (C) was a fixed modification. Peptide length was 7-30. Initial mass accuracy was 10 ppm and MS1 accuracy was 4 pm. Digestion was set to trypsin/P with a maximum of 1 missed cleavage. Precursor false discovery rate was 1%. Match between runs was disabled. Heuristic protein inference was enabled, as were all other default algorithm settings for DIA-NN v1.9.2.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTDAD clinical trial SomaScan assay.\u0026nbsp;\u003c/strong\u003eProteomics of plasma samples before and after the dietary intervention were done using the SomaScan v4.1 assay (SomaLogic) that detects approximately 7,000 proteins. Raw data was provided by SomaLogic following standardization, normalization, and calibration, including adaptive normalization by maximum likelihood (ANML), and mapped to UniProt. Protein measurements are provided in relative fluorescent units (RFU). Data was log2 transformed and standardized prior to analyses in the current study.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eStatistical analyses\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFeature selection.\u0026nbsp;\u003c/strong\u003eIn the plasma, CSF, dlPFC, and STG, APOE4 proteins were identified using mutual information \u003csup\u003e54\u003c/sup\u003e as previously reported \u003csup\u003e1,2\u003c/sup\u003e. In plasma and CSF, we identified \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4-specific proteins using all \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4 carriers and non-carriers to validate our previous findings where \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4 proteins were identified only in non-impaired controls \u003csup\u003e2\u003c/sup\u003e. In the dlPFC and STG, we identified \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4 proteins in non-impaired controls. Proteins with a mutual information value \u0026gt;0.1 were selected for machine learning analyses. We confirmed our feature selection method using principal component analysis (PCA). For plasma and CSF, we also performed a second feature selection using mutual information on \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4 carriers with AD relative to non-impaired control \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4 carriers. Mutual information was calculated in R (v4.4.1) using the package \u0026apos;FSelectorRcpp\u0026apos; \u003csup\u003e55\u003c/sup\u003e and PCA plots were made using \u0026lsquo;ggplot2\u0026rsquo; \u003csup\u003e56\u003c/sup\u003e.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMachine learning.\u0026nbsp;\u003c/strong\u003eWe used classification and regression trees (CART) and random forest to test the predictive performance of our identified \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4 proteins for differentiating between \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4 carriers and non-carriers. The dataset was split into a 70% training and validation set and a 30% withheld (unseen) testing set. Model training and evaluation were done using a 5-fold cross-validation procedure repeated 10 times. Machine learning was done in R (v4.4.1) using the package \u0026lsquo;caret\u0026rsquo;.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCortical organoid proteomics.\u0026nbsp;\u003c/strong\u003eLabel-free mass spectrometry proteomic data was analyzed in R (v4.4.1) using \u0026lsquo;limma\u0026rsquo; package. The raw data consisted of protein group intensities across biological samples, with sample names encoding both experimental condition (\u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4 carrier or non-carrier) and time point (four weeks or eight weeks). The protein intensity matrix was filtered to retain proteins with valid quantification in at least 30% of samples. Missing values were imputed per protein by replacing missing entries with the minimum observed intensity for that protein, assuming missingness was due to low abundance below the detection limit. Protein intensities were log2-transformed to stabilize variance. Differential protein abundance between \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4 and non-carrier cortical organoid samples was assessed at both four week and eight-week time points using a linear modeling framework with empirical Bayes moderation, implemented in limma. Specific contrasts (\u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;4 vs non-carrier at each time point) were tested. Resulting \u003cem\u003ep\u003c/em\u003e-values were corrected for multiple testing using the Benjamini-Hochberg method to control the false discovery rate (FDR). Proteins with adjusted p-values (FDR) below 0.05 were considered significantly differentially abundant. Volcano plots were constructed using -log\u003csub\u003e10\u003c/sub\u003e-transformed adjusted p-values on the y-axis. Plots show significance thresholds that were applied at FDR-adjusted p \u0026lt; 0.05. All plots and downstream analyses were performed in R using the \u0026lsquo;ggplot2\u0026rsquo; package.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCortical organoid AD pathology.\u0026nbsp;\u003c/strong\u003eELISA data for secreted amyloid-\u0026beta;42/40 and p-tau217 was analyzed at each time point using a two-tailed unpaired t-test. To analyze interactions between genotype and time, a two-way ANOVA was used. All statistical analyses were performed in GraphPad Prism (v10.5.0). \u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTDAD clinical trial proteomics.\u0026nbsp;\u003c/strong\u003eLog2 transformed SomaScan assay data was analyzed in R (v4.4.1) using \u0026lsquo;limma\u0026rsquo; package. Differential protein abundance between baseline and post-study in\u003cem\u003e\u0026nbsp;APOE\u003c/em\u003e \u0026epsilon;4 carriers and non-carriers was assessed using a linear modeling framework with empirical Bayes moderation, implemented in limma. Resulting \u003cem\u003ep\u003c/em\u003e-values were corrected for multiple testing using the Benjamini-Hochberg method to control the false discovery rate (FDR). Proteins with adjusted p-values (FDR) below 0.05 were considered significantly differentially abundant. Volcano plots were constructed using -log\u003csub\u003e10\u003c/sub\u003e-transformed adjusted p-values on the y-axis. Plots show significance thresholds that were applied at FDR-adjusted p \u0026lt; 0.05. All plots and downstream analyses were performed in R using the \u0026lsquo;ggplot2\u0026rsquo; package.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEnrichment analysis.\u0026nbsp;\u003c/strong\u003eEnriched biological functions and pathways across APOE4 proteins were assessed using NetworkAnalyst (v3.0) \u003csup\u003e57-59\u003c/sup\u003e. Protein-protein interactions were identified using a first order network in the International Molecular Exchange Consortium (IMEx) interactome database \u003csup\u003e60\u003c/sup\u003e and InnateDB \u003csup\u003e61\u003c/sup\u003e. Network enrichments for biological processes and pathways were done using the PANTHER classification system \u003csup\u003e62\u003c/sup\u003e and Kyoto Encyclopedia of Genes and Genomes (KEGG) database \u003csup\u003e63\u003c/sup\u003e, respectively. Statistical significance of the enriched networks was determined by a false discovery rate (FDR) of \u0026gt; 0.05. Proteins enriched in the KEGG AD pathway were visualized using KEGG mapper \u003csup\u003e64\u003c/sup\u003e. Cell type-specific enrichment analyses for immune cells, brain regions, and white matter cells were done using single-cell RNA sequencing data from the Human Protein Atlas (v23, Ensembl v109) \u003csup\u003e12-14\u003c/sup\u003e. Here, the corresponding protein-coding transcripts per million for each \u003cem\u003eAPOE\u0026nbsp;\u003c/em\u003e\u0026epsilon;4 protein was identified. Expression for each cell type was normalized using min-max scaling. Heatmaps were generated to visually represent these enrichments using R (v4.4.1).\u0026nbsp;\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eData availability\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe harmonized GNPC data used to generate these findings was provided to Consortium Members in June 2024 and will be made available for public request by the AD Data Initiative by July 1, 2025. Members of the global research community will be able to access the metadata and place a data use request via the AD Discovery Portal (https://discover.alzheimersdata.org/). Access is contingent on adherence to the GNPC Data Use Agreement and the Publication Policies. \u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe AMP-AD Diverse Cohorts study data is available through the AD Knowledge Portal (https://adknowledgeportal.synapse.org/). Researchers who wish to access this controlled dataset are required to submit a Data Use Agreement. More information can be found here: https://adknowledgeportal.synapse.org/Data%20Access.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCode Availability\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAll code used in this study are publicly available at https://github.com/Art83/AD_apoe.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgements\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors are grateful to the cohort contributors, patients, donors, and families who helped make this research possible. This work was supported by the Australian Government\u0026rsquo;s Medical Research Future Fund MRF2040081 (C.A.F, A-N.C, A.S.); Dementia Australia Research Foundation Bondi2BlueMtns Project Grant (C.A.F., A-N.C., H.M.W., R.H.S.); Neil \u0026amp; Norma Hill Foundation (C.A.F.); Annemarie \u0026amp; Arturo Gandioli-Fumagali Foundation (C.A.F.); Perpetual Foundation \u0026ndash; John Williams Endowment (C.A.F.); Hillcrest Foundation (C.A.F.); John \u0026amp; Anne Leece Family (C.A.F.); Paul \u0026amp; Valeria Ainsworth Precision Medicine Research Fellowship (C.A.F.); University of Kansas Alzheimer\u0026rsquo;s Disease Developmental Projects Grant (C.Sl., C.A.F., A.S.); the NIH P30AG072973 (H.M.W., J.M.B., C.Sl., R.H.S.), R01AG064227 (C.Sl.), R21TR003589 (H.M.W.), R01AG07816 (H.M.W.), U19AG068054 (H.M.W.), R01AG060733 (J.E.K., M.K.T., J.D.M., D.K.S., J.M.B., R.H.S.); Alzheimer\u0026rsquo;s Association 23AARG-1023 (H.M.W.); Sydney Horizon Fellowship (A-N.C.); and National Health and Medical Research Council 2025529 (J.H.R.). The results published here are in whole or in part based on data obtained from the AD Knowledge Portal (https://adknowledgeportal.org/). Data generation was supported by the following NIH grants: U01AG046139, U01AG046170, U01AG061357, U01AG061356, U01AG061359, and R01AG067025. We thank the participants of participants of the Religious Order Study, Memory and Aging Project, the Minority Aging Research Study, Rush Alzheimer\u0026rsquo;s Disease Research Center, Mount Sinai/JJ Peters VA Medical Center NIH Brain and Tissue Repository, National Institute of Mental Health Human Brain Collection Core (NIMH HBCC), Mayo Clinic Brain Bank, Sun Health Research Institute Brain and Body Donation Program, Goizueta Alzheimer\u0026rsquo;s Disease Research Center, New York Brain Bank at Columbia University, New York Genome Center and the Biggs Institute Brain Bank for their generous donations. Data and analysis contributing investigators include Nil\u0026uuml;fer Ertekin-Taner, Minerva Carrasquillo, Mariet Allen (Mayo Clinic, Jacksonville, FL), David Bennett, Lisa Barnes (Rush University), Philip De Jager, Vilas Menon (Columbia University), Bin Zhang, Vahram Haroutanian (Icahn School of Medicine at Mount Sinai), Allan Levey, Nick Seyfried (Emory University), Rima Kaddurah-Daouk (Duke University), Steve Finkbeiner (University of California-San Francisco/Gladstone Institutes), Daifeng Wang (University of Wisconsin-Madison), Stefano Marenco (NIMH HBCC), Anna Greenwood, Abby Vander Linden, Laura Heath, William Poehlman (Sage Bionetworks). Confocal microscopy was performed at the University of Kansas Medical Center supported by NIH S10 OD 032207 and the University of Sydney\u0026rsquo;s Australian Centre for Microscopy \u0026amp; Microanalysis. The funders of this work played no role in the design of the study, the running of experiments and analyses, the interpretation of the results, and the writing of the manuscript.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor contributions\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eA.S. and C.A.F. conceptualized the study and led the study design. S.T., M.E.G., B.H., C.Sm., C.T., A-N.C., H.M.W. and C.A.F. performed the cortical organoid experiments. J.H.R. provided key immunological insights. J.M.B. and R.H.S. provided clinical neurological insights. V.K. and F.B.I. provided assistance and insight for the GNPC dataset. M.E.G. and C.Sl. provided proteomic insights. J.E.K., M.K.T., J.D.M., D.K.S., J.M.B., and R.H.S. performed the TDAD pilot clinical trial. A.S., S.T., and C.A.F. produced the figures. C.A.F. wrote the paper and supervised the study. All authors revised the paper for intellectual content and read and approved the final version.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting interests\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare no competing interests.\u0026nbsp;\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eShvetcov, A.\u003cem\u003e et al.\u003c/em\u003e Proteome profiling of cerebrospinal fluid using machine learning shows a unique protein signature associated with APOE genotype. \u003cem\u003eAging Cell\u003c/em\u003e (2024).\u003c/li\u003e\n\u003cli\u003eShvetcov, A.\u003cem\u003e et al.\u003c/em\u003e APOE \u0026epsilon;4 carriers share immune-related proteomic changes across neurodegenerative diseases. \u003cem\u003eNature Medicine\u003c/em\u003e (2025).\u003c/li\u003e\n\u003cli\u003eGale, S. C.\u003cem\u003e et al.\u003c/em\u003e APO\u0026epsilon;4 is associated with enhanced in vivo innate immune responses in humans. \u003cem\u003eJournal of Allergy and Clinical Immunology\u003c/em\u003e \u003cstrong\u003e134\u003c/strong\u003e, 127-134 (2015).\u003c/li\u003e\n\u003cli\u003eShao, Y.\u003cem\u003e et al.\u003c/em\u003e Presence of the apolipoprotein E-\u0026epsilon;4 allele is associated with an increased risk of sepsis progression. \u003cem\u003eScientific Reports\u003c/em\u003e \u003cstrong\u003e10\u003c/strong\u003e (2020).\u003c/li\u003e\n\u003cli\u003eBonacina, F.\u003cem\u003e et al.\u003c/em\u003e Myeloid apolipoprotein E controls dendritic cell antigen presentation and T cell activation. \u003cem\u003eNature Communications\u003c/em\u003e \u003cstrong\u003e9\u003c/strong\u003e (2018).\u003c/li\u003e\n\u003cli\u003eZhang, Y.-J.\u003cem\u003e et al.\u003c/em\u003e APOE \u0026epsilon;4-associated downregulation of the IL-7/IL-7R pathway in effector memory T cells: Implications for Alzheimer\u0026apos;s disease. \u003cem\u003eAlzheimer\u0026apos;s \u0026amp; Dementia\u003c/em\u003e \u003cstrong\u003e20\u003c/strong\u003e, 6441-6455 (2024).\u003c/li\u003e\n\u003cli\u003eRosenzweig, N.\u003cem\u003e et al.\u003c/em\u003e Sex-dependent APOE4 neutrophil-microglia interactions drive cognitive impairment in Alzheimer\u0026apos;s disease. \u003cem\u003eNature Medicine\u003c/em\u003e \u003cstrong\u003e30\u003c/strong\u003e, 2990-3003 (2024).\u003c/li\u003e\n\u003cli\u003eFriedberg, J. S.\u003cem\u003e et al.\u003c/em\u003e Associations between brain inflammatory profiles and human neuropathology are altered based on apolipoprotein E \u0026epsilon;4 genotype. \u003cem\u003eScientific Reports\u003c/em\u003e \u003cstrong\u003e10\u003c/strong\u003e (2020).\u003c/li\u003e\n\u003cli\u003eHaney, M. S.\u003cem\u003e et al.\u003c/em\u003e APOE4/4 is linked to damaging lipid droplets in Alzheimer\u0026apos;s disease microglia. \u003cem\u003eNature\u003c/em\u003e \u003cstrong\u003e628\u003c/strong\u003e, 154-161 (2024).\u003c/li\u003e\n\u003cli\u003eHellen, M.\u003cem\u003e et al.\u003c/em\u003e Inflammation-induced lysosomal dysfunction in human iPSC-derived microglia is exacerbated by APOE 4/4 genotype. \u003cem\u003eJournal of Neuroinflammation\u003c/em\u003e \u003cstrong\u003e22\u003c/strong\u003e (2025).\u003c/li\u003e\n\u003cli\u003eKonttinen, H.\u003cem\u003e et al.\u003c/em\u003e PSEN1\u0026Delta;E9, APPswe, and APOE4 confer disparate phenotypes in human iPSC-derived microglia. \u003cem\u003eStem Cell Reports\u003c/em\u003e \u003cstrong\u003e13\u003c/strong\u003e, 669-683 (2019).\u003c/li\u003e\n\u003cli\u003eKarlsson, M.\u003cem\u003e et al.\u003c/em\u003e A single-cell type transcriptomics map of human tissues. \u003cem\u003eScience Advances\u003c/em\u003e \u003cstrong\u003e7\u003c/strong\u003e (2021).\u003c/li\u003e\n\u003cli\u003eUhlen, M.\u003cem\u003e et al.\u003c/em\u003e A genome-wide transcriptomic analysis of protein-coding genes in human blood cells. \u003cem\u003eScience\u003c/em\u003e \u003cstrong\u003e366\u003c/strong\u003e, 6472 (2019).\u003c/li\u003e\n\u003cli\u003eSjostedt, E.\u003cem\u003e et al.\u003c/em\u003e An atlas of protein-coding genes in the human, pig, and mouse brain. \u003cem\u003eScience\u003c/em\u003e \u003cstrong\u003e367\u003c/strong\u003e (2020).\u003c/li\u003e\n\u003cli\u003eCho, A.-N.\u003cem\u003e et al.\u003c/em\u003e Microfluidic device with brain extracellular matrix promotes structural and functional maturation of human brain organoids. \u003cem\u003eNature Communications\u003c/em\u003e \u003cstrong\u003e12\u003c/strong\u003e (2021).\u003c/li\u003e\n\u003cli\u003eTruong, D.-J. J.\u003cem\u003e et al.\u003c/em\u003e Non-invasive and high-throughput investigation of exon-specific isoform expression. \u003cem\u003eNature Cell Biology\u003c/em\u003e \u003cstrong\u003e23\u003c/strong\u003e, 652-663 (2021).\u003c/li\u003e\n\u003cli\u003eHarman, A. L., Gasior, M., Vining, E. P. \u0026amp; Rogawski, M. A. The neuropharmacology of the ketogenic diet. \u003cem\u003ePediatric Neurology\u003c/em\u003e \u003cstrong\u003e36\u003c/strong\u003e, 281-292 (2007).\u003c/li\u003e\n\u003cli\u003eKelement, R. J., Brehm, N. \u0026amp; Sweeney, R. A. Ketogenic diets in medical oncology: A systematic review with focus on clinical outcomes. \u003cem\u003eMedical Oncology\u003c/em\u003e \u003cstrong\u003e37\u003c/strong\u003e (2020).\u003c/li\u003e\n\u003cli\u003eCervenka, M. C.\u003cem\u003e et al.\u003c/em\u003e International recommendations for the management of adults treated with ketogenic diet therapies. \u003cem\u003eNeurology Clinical Practice\u003c/em\u003e \u003cstrong\u003e11\u003c/strong\u003e, 385-397 (2020).\u003c/li\u003e\n\u003cli\u003eWeber, D. D., Aminazdeh-Gohari, S. \u0026amp; Kofler, B. Ketogenic diet in cancer therapy. \u003cem\u003eAging\u003c/em\u003e \u003cstrong\u003e10\u003c/strong\u003e, 164-165 (2018).\u003c/li\u003e\n\u003cli\u003eTaylor, M. K., Sullivan, D. K., Mahnken, J. D., Burns, J. M. \u0026amp; Swerdlow, R. H. Feasibility and efficacy data from a ketogenic diet intervention in Alzheimer\u0026apos;s disease. \u003cem\u003eAlzheimer\u0026apos;s \u0026amp; Dementia\u003c/em\u003e \u003cstrong\u003e4\u003c/strong\u003e, 28-36 (2017).\u003c/li\u003e\n\u003cli\u003eSoares, H. D.\u003cem\u003e et al.\u003c/em\u003e Plasma biomarkers associated with the apolipoprotein E genotype and Alzheimer disease. \u003cem\u003eArchives of Neurology\u003c/em\u003e \u003cstrong\u003e69\u003c/strong\u003e, 1310-1317 (2012).\u003c/li\u003e\n\u003cli\u003eFan, Y.-Y.\u003cem\u003e et al.\u003c/em\u003e APOE \u0026epsilon;4 allele elevates the expressions of inflammatory factors and promotes Alzheimer\u0026apos;s diseae progression: A comparative study based on Han and She populations in the Wenzhou area. \u003cem\u003eBrain Research Bulletin\u003c/em\u003e \u003cstrong\u003e132\u003c/strong\u003e, 39-43 (2017).\u003c/li\u003e\n\u003cli\u003eOlgiati, P.\u003cem\u003e et al.\u003c/em\u003e APOE epislon-4 allele and cytokine production in Alzheimer\u0026apos;s disease. \u003cem\u003eInternational Journal of Geriatric Psychiatry\u003c/em\u003e \u003cstrong\u003e25\u003c/strong\u003e, 338-344 (2009).\u003c/li\u003e\n\u003cli\u003eRamakrishnan, A.\u003cem\u003e et al.\u003c/em\u003e Epigenetic dysregulation in Alzheimer\u0026apos;s disease peripheral immunity. \u003cem\u003eNeuron\u003c/em\u003e \u003cstrong\u003e112\u003c/strong\u003e, 1235-1248 (2024).\u003c/li\u003e\n\u003cli\u003eWainberg, M.\u003cem\u003e et al.\u003c/em\u003e The viral hypothesis: How herpesvirus may contribute to Alzheimer\u0026apos;s disease. \u003cem\u003eMolecular Psychiatry\u003c/em\u003e \u003cstrong\u003e26\u003c/strong\u003e, 5476-5480 (2021).\u003c/li\u003e\n\u003cli\u003eEiriksdottir, G.\u003cem\u003e et al.\u003c/em\u003e Apolipoprotein E genotype and statins affect CRP levels through independent and different mechanisms. \u003cem\u003eAtherosclerosis\u003c/em\u003e \u003cstrong\u003e186\u003c/strong\u003e, 222-224 (2006).\u003c/li\u003e\n\u003cli\u003eGarcia, A. R.\u003cem\u003e et al.\u003c/em\u003e APOE4 is associated with elevated blood lipids and lower levels of innate immune biomarkers in a tropical Amerindian subsistence population. \u003cem\u003eElife\u003c/em\u003e \u003cstrong\u003e10\u003c/strong\u003e (2021).\u003c/li\u003e\n\u003cli\u003eVasunilashorn, S.\u003cem\u003e et al.\u003c/em\u003e Apolipoprotein E, biomarkers, and mortality in Taiwanese older adults. \u003cem\u003eAtherosclerosis\u003c/em\u003e \u003cstrong\u003e219\u003c/strong\u003e, 349-354 (2011).\u003c/li\u003e\n\u003cli\u003eTao, Q.\u003cem\u003e et al.\u003c/em\u003e Association of chronic low-grade inflammation with risk of Alzheimer disease in ApoE4 carriers. \u003cem\u003eJAMA Network Open\u003c/em\u003e \u003cstrong\u003e1\u003c/strong\u003e, e183597 (2018).\u003c/li\u003e\n\u003cli\u003eSweeney, M. D., Montagne, A., Sagare, A. P., Nation, D. A. \u0026amp; Zlokovic, B. V. Broken blood-brain barrier in Alzheimer\u0026apos;s disease and other neurodegenerative disorders. \u003cem\u003eNature Reviews Neurology\u003c/em\u003e \u003cstrong\u003e15\u003c/strong\u003e, 133-150 (2019).\u003c/li\u003e\n\u003cli\u003eMontagne, A.\u003cem\u003e et al.\u003c/em\u003e APOE4 leads to blood-brain barrier dysfunction predictive cognitive decline. \u003cem\u003eNature\u003c/em\u003e \u003cstrong\u003e581\u003c/strong\u003e, 71-76 (2020).\u003c/li\u003e\n\u003cli\u003eBachstetter, A. D.\u003cem\u003e et al.\u003c/em\u003e A blunted T\u003csub\u003eH\u003c/sub\u003e17 cytokine signature in women with mild cognitive impairment: Insights from inflammatory profiling of a community-based cohort of older adults. \u003cem\u003eBrain Communications\u003c/em\u003e \u003cstrong\u003e5\u003c/strong\u003e (2023).\u003c/li\u003e\n\u003cli\u003eBegum, A. N.\u003cem\u003e et al.\u003c/em\u003e Women with the Alzheimer\u0026apos;s risk marker ApoE4 lose A\u0026beta;-specific CD4\u003csup\u003e+\u003c/sup\u003e T cells 10-20 years before men. \u003cem\u003eTranslational Psychiatry\u003c/em\u003e \u003cstrong\u003e4\u003c/strong\u003e (2014).\u003c/li\u003e\n\u003cli\u003eMinett, T.\u003cem\u003e et al.\u003c/em\u003e Microglial immunophenotype in dementia with Alzheimer\u0026apos;s pathology. \u003cem\u003eJournal of Neuroinflammation\u003c/em\u003e \u003cstrong\u003e13\u003c/strong\u003e (2016).\u003c/li\u003e\n\u003cli\u003eMusiek, E. S. \u0026amp; Bennett, D. A. Aducanumab and the \u0026quot;post-amyloid\u0026quot; era of Alzheimer research. \u003cem\u003eNeuron\u003c/em\u003e \u003cstrong\u003e109\u003c/strong\u003e, 3045-3047 (2021).\u003c/li\u003e\n\u003cli\u003eKnopman, D. S., Jones, D. T. \u0026amp; Greicius, M. D. Failure to demonstrate efficacy of aducanumab: An analysis of the EMERGE and ENGAGE trials as reported by Biogen, December 2019. \u003cem\u003eAlzheimer\u0026apos;s \u0026amp; Dementia\u003c/em\u003e \u003cstrong\u003e17\u003c/strong\u003e, 696-701 (2020).\u003c/li\u003e\n\u003cli\u003eBurke, J. F., Kerber, K. A., Langa, K. M., Albin, R. L. \u0026amp; Kotagal, V. Lecanemab: Looking before we leap. \u003cem\u003eNeurology\u003c/em\u003e \u003cstrong\u003e101\u003c/strong\u003e, 661-665 (2023).\u003c/li\u003e\n\u003cli\u003eImbimbo, B. P., Balducci, C., Ippati, S. \u0026amp; Watling, M. Initial failures of anti-tau antibodies in Alzheimer\u0026apos;s disease are reminiscent of the amyloid-\u0026beta; story. \u003cem\u003eNeural Regeneration Research\u003c/em\u003e \u003cstrong\u003e18\u003c/strong\u003e, 117-118 (2022).\u003c/li\u003e\n\u003cli\u003eHarris, K.\u003cem\u003e et al.\u003c/em\u003e The impact of routine vaccinations on Alzheimer\u0026apos;s disease risk in persons 65 years and older: A claims-based cohort study using propensity score matching. \u003cem\u003eJournal of Alzheimer\u0026apos;s Disease\u003c/em\u003e \u003cstrong\u003e95\u003c/strong\u003e, 703-718 (2023).\u003c/li\u003e\n\u003cli\u003eWu, X.\u003cem\u003e et al.\u003c/em\u003e Adult vaccination as a protective factor for dementia: A meta-analysis and systematic review of population-based observational studies. \u003cem\u003eFrontiers in Immunology\u003c/em\u003e \u003cstrong\u003e13\u003c/strong\u003e, 872542 (2022).\u003c/li\u003e\n\u003cli\u003eEyting, M.\u003cem\u003e et al.\u003c/em\u003e A natural experiment on the effect of herpes zoster vaccination on dementia. \u003cem\u003eNature\u003c/em\u003e (2025).\u003c/li\u003e\n\u003cli\u003eTaquet, M., Dercon, Q., Todd, J. A. \u0026amp; Harrison, P. J. The recombinant shingles vaccine is associated with lower risk of dementia. \u003cem\u003eNature Medicine\u003c/em\u003e \u003cstrong\u003e30\u003c/strong\u003e, 2777-2781 (2024).\u003c/li\u003e\n\u003cli\u003eRutledge, J.\u003cem\u003e et al.\u003c/em\u003e Comprehensive proteomics of CSF, plasma, and urine identify DDC and other biomarkers of early Parkinson\u0026apos;s disease. \u003cem\u003eActa Neuropathologica\u003c/em\u003e \u003cstrong\u003e147\u003c/strong\u003e, 52 (2024).\u003c/li\u003e\n\u003cli\u003eImam, F. B.\u003cem\u003e et al.\u003c/em\u003e The Global Neurodegeneration Proteomics Consortium: Biomarker and drug target discovery across ~40,000 biosample analyses for AD, PD, ALS, FTD, and aging. \u003cem\u003eNature Medicine\u003c/em\u003e (2025).\u003c/li\u003e\n\u003cli\u003eSeifar, F.\u003cem\u003e et al.\u003c/em\u003e Large-scale deep proteomic analysis in Alzheimer\u0026apos;s disease brain regions across race and ethnicity. \u003cem\u003eAlzheimer\u0026apos;s \u0026amp; Dementia\u003c/em\u003e \u003cstrong\u003e20\u003c/strong\u003e, 8878-8897 (2024).\u003c/li\u003e\n\u003cli\u003eMcKhann, G. M.\u003cem\u003e et al.\u003c/em\u003e The diagnosis of dementia due to Alzheimer\u0026apos;s disease: Recommendations from the National Institute on Aging-Alzheimer\u0026apos;s Association workgroups on diagnostic guidelines for Alzheimer\u0026apos;s disease. \u003cem\u003eAlzheimer\u0026apos;s \u0026amp; Dementia\u003c/em\u003e \u003cstrong\u003e7\u003c/strong\u003e, 263-269 (2011).\u003c/li\u003e\n\u003cli\u003eMifflin, M. D.\u003cem\u003e et al.\u003c/em\u003e A new predictive equation for resting energy expenditure in healthy individuals. \u003cem\u003eAmerican Journal of Clinical Nutrition\u003c/em\u003e \u003cstrong\u003e51\u003c/strong\u003e, 241-247 (1990).\u003c/li\u003e\n\u003cli\u003eShindelin, J.\u003cem\u003e et al.\u003c/em\u003e Fiji: An open-source platform for biological-image analysis. \u003cem\u003eNature Methods\u003c/em\u003e \u003cstrong\u003e9\u003c/strong\u003e, 676-682 (2012).\u003c/li\u003e\n\u003cli\u003eGold, L.\u003cem\u003e et al.\u003c/em\u003e Aptamer-based multiplexed proteomic technology for biomarker discovery. \u003cem\u003ePLoS One\u003c/em\u003e \u003cstrong\u003e5\u003c/strong\u003e, e15004 (2010).\u003c/li\u003e\n\u003cli\u003eWilliams, S. A.\u003cem\u003e et al.\u003c/em\u003e Plasma protein patterns as comprehensive indicators of health. \u003cem\u003eNature Medicine\u003c/em\u003e \u003cstrong\u003e25\u003c/strong\u003e, 1851-1857 (2019).\u003c/li\u003e\n\u003cli\u003eRappsilber, J., Ishihama, Y. \u0026amp; Mann, M. Stop and go extraction tips for matrix-assisted laser desorption / ionization, nanoelectrospray, and LC/MS sample pretreatment in proteomics. \u003cem\u003eAnnals of Chemistry\u003c/em\u003e \u003cstrong\u003e75\u003c/strong\u003e, 663-670 (2003).\u003c/li\u003e\n\u003cli\u003eDemichev, V., Messner, C. B., Vernardis, S. I., Lilley, K. S. \u0026amp; Ralser, M. DIA-NN: Neural networks and interference correction enable deep proteome coverage in high throughput. \u003cem\u003eNature Methods\u003c/em\u003e \u003cstrong\u003e17\u003c/strong\u003e, 41-44 (2020).\u003c/li\u003e\n\u003cli\u003eRoss, B. C. Mutual information between discrete and continuous data sets. \u003cem\u003ePLoS One\u003c/em\u003e \u003cstrong\u003e9\u003c/strong\u003e, e87357 (2014).\u003c/li\u003e\n\u003cli\u003eFayyad, U. M. \u0026amp; Irani, K. B. in \u003cem\u003e13th International Joint Conference on Uncertainty in Artificial Intelligence\u003c/em\u003e 1022-1029 (Chambery, France, 1993).\u003c/li\u003e\n\u003cli\u003eWickham, H. \u003cem\u003eggplot2: Elegant Graphics for Data Analysis\u003c/em\u003e. (Springer-Verlag, 2016).\u003c/li\u003e\n\u003cli\u003eXia, J., Benner, M. J. \u0026amp; Hancock, R. E. W. NetworkAnalyst - Integrative approaches for protein-protein interaction network analysis and visual exploration. \u003cem\u003eNucleic Acids Research\u003c/em\u003e \u003cstrong\u003e42\u003c/strong\u003e, W167-W174 (2014).\u003c/li\u003e\n\u003cli\u003eXia, J., Gill, E. \u0026amp; Hancock, R. E. W. NetworkAnalyst for statistical, visual and network-based appraoches for meta-analysis of expression data. \u003cem\u003eNature Protocols\u003c/em\u003e \u003cstrong\u003e10\u003c/strong\u003e, 823-844 (2015).\u003c/li\u003e\n\u003cli\u003eZhou, G.\u003cem\u003e et al.\u003c/em\u003e NetworkAnalyst 3.0: A visual analytics platform for comprehensive gene expression profiling and meta-analysis. \u003cem\u003eNucleic Acids Research\u003c/em\u003e \u003cstrong\u003e47\u003c/strong\u003e, W234-W241 (2019).\u003c/li\u003e\n\u003cli\u003eOrchard, S.\u003cem\u003e et al.\u003c/em\u003e Protein interaction data curation - The International Molecular Exchange Consortium (IMEx). \u003cem\u003eNature Methods\u003c/em\u003e \u003cstrong\u003e9\u003c/strong\u003e, 345-350 (2013).\u003c/li\u003e\n\u003cli\u003eBreuer, K.\u003cem\u003e et al.\u003c/em\u003e InnateDB: Systems biology of innate immunity and beyond - Recent updates and continuing curation. \u003cem\u003eNucleic Acids Research\u003c/em\u003e \u003cstrong\u003e41\u003c/strong\u003e, 1228-1233 (2013).\u003c/li\u003e\n\u003cli\u003eMi, H., Muruganujan, A., Casagrande, J. T. \u0026amp; Thomas, P. D. Large-scale gene function analysis with PANTHER classification system. \u003cem\u003eNature Protocols\u003c/em\u003e \u003cstrong\u003e8\u003c/strong\u003e, 1551-1566 (2019).\u003c/li\u003e\n\u003cli\u003eKanehisa, M. \u0026amp; Goto, S. KEGG: Kyoto Encyclopedia of Genes and Genomes. \u003cem\u003eNucleic Acids Research\u003c/em\u003e \u003cstrong\u003e38\u003c/strong\u003e, 27-30 (2000).\u003c/li\u003e\n\u003cli\u003eKanehisa, M. \u0026amp; Sato, Y. K., M. KEGG mapping tools for uncovering hidden featuers in biological data. \u003cem\u003eProtein Science\u003c/em\u003e \u003cstrong\u003e31\u003c/strong\u003e, 47-53 (2021).\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"nature-portfolio","isNatureJournal":true,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"","title":"Nature Portfolio","twitterHandle":"","acdcEnabled":false,"dfaEnabled":false,"editorialSystem":"ejp","reportingPortfolio":"","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-7089423/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7089423/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"APOE ε4 is the strongest genetic risk factor for late-onset Alzheimer’s disease (AD), but its contribution to disease pathogenesis remains incompletely understood. Here, we integrate proteomic profiling of plasma, cerebrospinal fluid (CSF), and brain tissue from over 10,000 individuals to define the immune phenotype associated with APOE ε4. We identify a conserved, allele dose-dependent pro-inflammatory immune signature across peripheral and central tissues independent of AD diagnosis. This signature is enriched in adaptive immune cells and white matter-resident glial and vascular cells. It also emerges in patient-derived cortical organoids prior to amyloid-β and tau pathology, supporting a causal, genotype-driven mechanism. Cross-tissue comparisons reveal shared innate and antiviral responses alongside tissue-specific immune signaling. Notably, a 12-week medical ketogenic diet partially reversed the APOE ε4 immune signature. These findings position immune dysregulation as an early and tractable driver of AD risk in APOE ε4 carriers with direct implications for targeted prevention strategies.","manuscriptTitle":"Cross-tissue immune profiling of APOE ε4 reveals early dysregulation in Alzheimer’s disease","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-07-15 12:08:38","doi":"10.21203/rs.3.rs-7089423/v1","editorialEvents":[],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"nature-communications","isNatureJournal":true,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"NCOMMS","sideBox":"Learn more about [Nature Communications](http://www.nature.com/ncomms/)","snPcode":"","submissionUrl":"https://mts-ncomms.nature.com/","title":"Nature Communications","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"ejp","reportingPortfolio":"Nature Communications","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"a6f6e826-58d8-4dfb-a1e7-177c43701a71","owner":[],"postedDate":"July 15th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[{"id":51531089,"name":"Biological sciences/Neuroscience/Diseases of the nervous system/Alzheimer's disease"},{"id":51531090,"name":"Biological sciences/Neuroscience/Neuroimmunology"}],"tags":[],"updatedAt":"2025-08-18T15:56:07+00:00","versionOfRecord":[],"versionCreatedAt":"2025-07-15 12:08:38","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7089423","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7089423","identity":"rs-7089423","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}