Astrocytic PTGDS inflection defines a neuropathological transition boundary during prodromal Alzheimer’s disease

Astrocytic PTGDS inflection defines a neuropathological transition boundary during prodromal 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 Research Article Astrocytic PTGDS inflection defines a neuropathological transition boundary during prodromal Alzheimer’s disease YoungOuk Kim, WooMyung Heo, Se Jin Park, YoungChul Kim, Ye Eun Cho, and 2 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9499795/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 5 You are reading this latest preprint version Abstract Alzheimer’s disease (AD) is characterized by prolonged prodromal stability preceding abrupt, often irreversible cognitive decline, yet the neuropathological event governing this transition remains undefined. Here, we identify a statistically bounded astrocytic inflection anchored to prostaglandin D2 synthase (PTGDS) dynamics that delineates a transition boundary between reversible compensatory buffering and accelerated neurodegeneration during the mild cognitive impairment (MCI)-to-AD continuum. Integrating pseudo-progression analysis of 1.3 million single nuclei from the SEA-AD atlas (84 donors) with cross-species validation in zebrafish and murine systems and longitudinal ADNI cerebrospinal fluid (CSF) proteomics (n = 735), we demonstrate that astrocytic PTGDS undergoes a biphasic trajectory with a statistically significant inflection point (segmented regression breakpoint: Bin 0.23, 95% CI: 0.13–0.33; Davies’ p = 0.032; quadratic vertex: continuous pseudo-progression score [CPS] 0.46, β₂ = −2.07, p < 2.2 × 10 −16 ). This inflection defines a neuropathological boundary at which metabolic exhaustion precedes and gates lipocalin-2 (LCN2)-mediated inflammatory amplification and CREB/NGFR-linked neurogenic suppression — repositioning inflammation as a downstream consequence of astrocytic metabolic failure. The boundary is conserved across human, zebrafish, and murine systems and aligns clinically with MMSE 26.0, beyond which pharmacological rescue of cognitive deficits is substantially attenuated. In longitudinal ADNI CSF proteomics, the PTGDS/LCN2 ratio independently predicts MCI-to-AD conversion (AUC = 0.743; HR = 3.2, 95% CI: 2.3–4.4), preceding structural atrophy. These findings establish astrocytic PTGDS exhaustion as a cross-scale neuropathological boundary with direct implications for stage-stratified therapeutic intervention in prodromal AD. Alzheimer’s disease astrocyte PTGDS neuropathological transition MCI lipocalin-2 NGFR phase boundary SEA-AD CSF biomarker Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Introduction Mild cognitive impairment (MCI) represents a pivotal yet poorly resolved stage in Alzheimer’s disease (AD) progression. Clinically defined as a prodromal state preceding dementia, annual MCI-to-AD conversion rates span 5–39%, and a substantial proportion of individuals remain stable or revert to normal cognition [ 1 – 3 ]. This heterogeneity implies that MCI does not represent a single biological state but rather a spectrum of substates differing fundamentally in reversibility and pathological commitment. Existing staging frameworks based on amyloid burden, tau phosphorylation, and clinical cognitive scores [ 4 , 5 ] fail to resolve reversible compensation from progressive neurodegenerative commitment at the molecular level. The central unresolved question in prodromal AD is therefore the identification of a transition boundary — the molecular event that separates reversible compensatory states from irreversible neurodegeneration. The limited disease-modifying efficacy of amyloid-targeted therapies [ 6 , 7 ] has redirected attention toward upstream events preceding overt neurodegeneration. Evidence increasingly implicates glial metabolic remodeling and impaired neurogenic support as determinants of neuronal instability that emerge before structural neuronal loss [ 8 – 15 ], raising the possibility that the critical disease transition is metabolic rather than proteopathic in origin. Neuroinflammatory activation alone does not adequately explain this transition, as inflammatory markers are elevated across the MCI spectrum without reliably distinguishing stable from progressive cases. A biologically meaningful transition boundary would therefore require a threshold-like molecular event capable of collapsing accumulated compensatory capacity and initiating downstream amplification cascades. The SEA-AD atlas (Gabitto et al., 2024) [ 16 ], comprising 1.3 million nuclei from 84 donors aligned along a validated continuous pseudo-progression score (CPS), provides an unprecedented framework to interrogate this question. Within this dataset, astrocytic trajectories reveal a biphasic pattern characterized by early compensatory activation followed by abrupt attenuation at intermediate disease stages. Whether this inflection constitutes a true neuropathological transition boundary — defined by non-linear dynamics, statistical threshold behavior, and cross-species conservation — or reflects continuous gradual drift has remained unresolved. Among candidate molecular mediators, lipocalin-2 (LCN2), an NF-κB-regulated inflammatory amplifier elevated in MCI and AD [ 17 – 19 ], links astrocytic dysfunction to synaptic instability and impaired CREB/NGFR-associated neurogenic signaling [ 20 – 25 ]. Upstream of this cascade, prostaglandin D₂ synthase (PTGDS) represents a compelling astrocytic regulator. PTGDS integrates inflammatory tone with neurogenic support [ 26 – 28 ] and functions as a major amyloid-β chaperone [ 29 ], placing it at the convergence of metabolic buffering and neuroprotection. Among astrocyte-enriched transcripts in the SEA-AD atlas, PTGDS demonstrated the most pronounced biphasic inflection and the highest astrocyte-selective expression ratio within CPS 0.3–0.5, distinguishing it from other reactive astrocytic markers including CLU, GFAP, and AQP4. Conflicting human CSF data — PTGDS elevation in some cohorts [ 30 , 31 ] and reduction in others [ 32 ] — are consistent with stage-dependent biphasic dynamics obscured by cross-sectional sampling. Here, we test the hypothesis that astrocytic PTGDS defines a statistically bounded neuropathological transition boundary during the MCI-to-AD continuum. Using an integrated framework comprising (1) segmented regression and quadratic modeling within SEA-AD single-cell pseudo-progression, (2) a reversible zebrafish MCI model enabling temporal dissection of compensatory failure, (3) murine perturbation systems for mechanistic validation, and (4) longitudinal ADNI CSF proteomics, we identify a lipid-metabolic exhaustion event at CPS 0.46 that marks a measurable neuropathological transition boundary, establish PTGDS as its molecular anchor, and define a pre-boundary therapeutic window for stage-stratified intervention. Materials and methods Ethical approval and animal husbandry All mouse experiments were approved by the IACUC of Kangwon National University (Approval No. KW-241104-1) and conducted under controlled environmental conditions (23 ± 2°C, 50 ± 10% humidity, 12 h light/dark cycle). All zebrafish experiments were performed at Zefit Inc., an FDA-registered CRO (Approval No. ZEFIT-IACUC-26010601-0001). Zebrafish were maintained at 28.5°C, 14:10 h light:dark cycle, pH 7.0–7.5. Human dataset analyses utilized de-identified secondary data from public repositories (SEA-AD, ADNI); no additional IRB approval was required. SEA-AD snRNA-seq trajectory analysis We analyzed 1.3 million nuclei from the SEA-AD middle temporal gyrus atlas [16] (quality control: ≥500 genes/nucleus, <20% mitochondrial reads; Scrublet-based doublet removal [33]). CPS values were rounded to one decimal to define nine discrete progression bins (Bins 0.1–0.9), approximately aligned to Braak stages: Bin 0.0–0.3 (~Braak I–II), Bin 0.4–0.6 (~Braak III–IV), Bin 0.7–0.9 (~Braak V–VI). This mapping is heuristic and does not imply direct pathological equivalence. LCN2 showed an extremely low detection rate (0.04%) and was independently validated by CSF proteomics (ADNI; Fig. 5), zebrafish qPCR (Fig. 2F), and murine qPCR (Fig. 4). Quality control metrics and bin-to-Braak stage mapping are detailed in Supplementary Table S1. Trajectory smoothing and cross-correlation analysis Bin-level means were computed from log-normalized expression per cell type and smoothed using a 3-bin centered moving average to mitigate stochastic dropout. Lagged cross-correlation functions (ccf() in R, lag.max = 3) were computed on the inflection window (Bins 0.4–0.8) and full trajectory (Bins 0.1–0.9). Approximate p-values were derived from t-test transformation of the peak correlation coefficient, with effective sample size adjusted for lag (neff = n − |lag|). The inflection window was selected to encompass the complete PTGDS biphasic trajectory from compensatory peak to exhaustion. All analyses are fully reproducible using R v4.3.2 with set.seed(42); analysis code and intermediate outputs are deposited at https://github.com/YoungOukKim/MCI-to-AD. Segmented regression and change-point analysis Segmented regression was performed using the segmented R package [34] with initial breakpoint estimate at 0.5; Davies’ test assessed breakpoint significance. Change-point estimation was conducted independently of ADNI clinical staging (Supplementary Fig. S1). Quadratic regression at single-cell resolution (n = 67,419 astrocytes) used ordinary least squares. Model comparison employed ANOVA and AIC criteria. Peak location derived from the vertex (−β1 / 2β2). LOESS smoothing applied independently to visualize trend consistency. Computational analysis of neuronal vulnerability Intercellular signaling was inferred using CellChat (v1.6.1) [35], focusing on LCN2–SLC22A17 interactions. Neuronal stress responses were quantified using SCENIC (v1.2.4) [36] for ATF3 and EGR1 regulon activity. Modules analyzed included NF-κB Priming (PTGS2, IL6ST, NFKBIA), Purinergic/Ca²⁺ Sensing (P2RY1, P2RY12, GJA1, ITPR2), Metabolic Buffering (HMOX1, SOD2, MT1E, MT2A, CLU, SLC1A2), and Immediate Early Stress (FOS, JUN, EGR1, ATF3). Detailed gene modules and intermediate signaling analyses are described in Supplementary Fig. S2. Reversible zebrafish MCI model Zebrafish (Danio rerio, AB strain) [37,38] were maintained at 28.5°C (14:10 h light:dark). A chronic triple-stressor paradigm (LPS 5 μg/L, D-galactose 0.2 mg/L, 10% lard-based high-fat diet [39,40]) was applied from 2–14 dpf. Survival rate was ≥95% across all groups, with 30–45% cognitive impairment confirmed by the red ball avoidance task. Cognitive assessment was performed at 14 dpf using the red ball avoidance paradigm (Zefit Inc. protocol) [41,42]. Conservation of NF-κB binding motifs and orthologous ptgdsb.1/2 status were verified [43,44]. Brain tissue was collected at 14 and 21 dpf for qPCR (ptgdsb.1/2, ngfr, bdnf; primer sequences in Supplementary Table S2). Pharmacological modulation and BV-2 microglial assay BXP-101 is a standardized multi-component formulation containing honokiol [45], wedelolactone [46], and atractylodin [47] as principal active constituents. Chemical standardization is shown in Supplementary Fig. S3. BXP-101 (0.3–0.6 μg/ml) was administered pre-inflection (6–14 dpf; early) or post-inflection (15–21 dpf; late). BV-2 murine microglial cells (passages 8–15) were cultured in DMEM supplemented with 10% FBS at 37°C, 5% CO₂. Cell viability was assessed by MTT assay after 24 h exposure to BXP-101. For anti-inflammatory assays, cells were pre-treated with BXP-101 for 1 h before LPS stimulation (1 μg/mL, 24 h). Nitric oxide production was quantified using Griess reagent, and NF-κB p65 nuclear translocation was measured by immunofluorescence (anti-p65 antibody, 1:200; secondary Alexa Fluor 488, 1:500). Images were acquired on a Zeiss LSM 880 confocal microscope and quantified using ImageJ (nuclear/cytoplasmic ratio). All experiments were performed in triplicate (Supplementary Fig. S4). Behavioral and molecular analyses Behavioral assessments used EthoVision XT 17 for zebrafish (locomotor and visual avoidance [41,42]) and mice (Y-maze [48], passive avoidance). General locomotor activity was assessed in a separate cohort (24-well plate, total distance moved) and is not included in the current analysis. Antibodies and primers used are listed in Supplementary Table S2. Y-maze and passive avoidance testing (murine) The Y-maze was constructed from black polyvinyl plastic and consisted of three arms (40 cm in length, 4 cm in width, and 12 cm in height), positioned at 120° angles. Distinct visual cues were placed at the end of each arm. Each mouse was placed at one arm end and allowed to freely explore for 8 minutes. Spontaneous alternation (%) = [actual alternations / (total arm entries − 2)] × 100. The passive avoidance test was performed in a light/dark apparatus (grid floor; 0.5 mA, 3 s foot shock). On Day 1 (acquisition trial), the latency to enter the dark compartment was recorded, followed by immediate foot shock delivery upon entry. On Day 2 (retention trial), mice were placed in the light compartment and the latency to enter the dark compartment was recorded with a cutoff time of 300 s. Mammalian validation Male ICR mice (5 weeks; Koatech, Korea) received intracerebroventricular Aβ1–42 injection (20 μM) [49]. Groups: Sham, Aβ1–42, Aβ1–42 + BXP-101 (50/100/200/400 mg/kg), Aβ1–42 + Donepezil (5 mg/kg) (n = 10/group; final n = 9 for Aβ1–42 group after exclusion of one statistical outlier; Supplementary Table S3). Animals were housed at 23 ± 2°C, 50 ± 10% humidity, 12 h light/dark cycle with ad libitum access to standard diet (2018S; Envigo) and water. Network pharmacology and molecular docking Target prediction used SwissTargetPrediction [50], PharmMapper [51], and DisGeNET [52], with supplementary databases OMIM [53] and GeneCards [54]; PPI network construction via STRING [55] in Cytoscape [56]; pathway enrichment via Metascape [57]. Molecular docking with AutoDock Vina [58] (exhaustiveness = 32); key targets: NF-κB p65 (PDB: 1NFI), GSK3B (PDB: 1Q3D), PTGS2 (PDB: 5KIR). ADMET profiles were predicted using pkCSM and SwissADME. ADNI CSF proteomics ADNI CSF proteomics data included Emory TMT-MS (n = 415) and SomaScan 7K (n = 735) platforms (SeqIds: PTGDS X10514-5, LCN2 X2836-68, NEFL X10082-251, Tau X5854-60). Clinical variables (MMSE, MoCA, diagnosis, age, sex, education, APOE ε4) were obtained from ADNIMERGE. CSF Aβ42, total Tau, and pTau181 were measured using Roche Elecsys assays (UPENNBIOMK dataset). Data were ComBat batch-corrected [59] and LOESS-smoothed [60]. PTGDS/LCN2 ratios were adjusted for age, sex, education, and APOE ε4 [61]. ROC analysis used logistic regression with predictors including PTGDS/LCN2 ratio, NEFL, total Tau, age, APOE ε4, Aβ42, pTau181, and ratio × APOE ε4 interaction; AUC computed with 95% DeLong CI. Cox regression (n = 379 MCI participants) modeled time-to-AD conversion stratified by 75th percentile risk score. Sensitivity analysis confirmed PTGDS inflection was consistent across platforms (TMT-MS and SomaScan) at MMSE ~26. Threshold sensitivity was assessed from MMSE 24 to 30 (Supplementary Table S4). Statistical analysis Analyses used R (v4.3.2) and Prism 10. Mixed-effects models implemented using lme4 [62]; survival dynamics evaluated via Kaplan–Meier estimation [63] and Cox proportional hazards regression [64] with Benjamini–Hochberg FDR correction [65]. Data are mean ± SEM; p < 0.05 was considered significant. For murine behavioral assays, Y-maze data were analyzed by one-way ANOVA with Newman–Keuls post hoc, and passive avoidance data by two-way ANOVA with Bonferroni post hoc. Results A conserved astrocytic inflection demarcates the neuropathological transition from compensatory to vulnerable states Analysis of 1.3 million nuclei from the SEA-AD atlas (84 donors; Fig. 1 A) identified a conserved astrocytic inflection demarcating the transition from compensatory to vulnerable states during the MCI-to-AD continuum (Fig. 1 A–E). By projecting nuclei onto CPS, we established a stage-resolved neurodegenerative trajectory framework (Fig. 1 B) and identified a reproducible astrocytic trajectory centered on PTGDS dynamics. The earliest detectable event was neuronal NDUFS1 decline beginning at Bin 0.1, a core mitochondrial complex I subunit consistent with its established role as an early site of vulnerability during MCI and AD progression (Fig. 1 C; Table 1 ) [ 66 , 67 ]. Astrocytic PTGDS expression increased in parallel and exhibited immediate inverse synchronization with neuronal NDUFS1 decline (Table 1 ), consistent with compensatory metabolic buffering — potentially through PTGDS-mediated prostaglandin signaling supporting astrocytic lactate production or anti-inflammatory buffering in response to neuronal energy deficit. Table 1 Cell-type-resolved expression trajectories in SEA-AD astrocytes and neurons across CPS bins Marker (Cell Type) Bin 0.1 (Early) Bin 0.5–0.6 (Mid) Bin 0.9 (Late) Trajectory Characterization NDUFS1 (Neuron) 0.7817 0.6980 0.7061 Overall declining trend from Bin 0.1 PTGDS (Astrocyte) 1.6531 1.9115 1.6310 Biphasic: slope shift at Bin 0.23; peak near CPS 0.46 LCN2 (Astrocyte) 0.0000 0.0001 0.0001 Sparse; induction follows PTGDS decline NGFR (Neuron) 0.00262 0.00058 0.00073 Early decline with sustained low plateau APOE (Microglia) 1.1820 1.1004 1.2186 U-shaped; mid-stage nadir Note: Expression values represent 3-bin moving averages of log-normalized expression from the SEA-AD middle temporal gyrus atlas (1.3 million nuclei, 84 donors). Segmented regression on PTGDS identified a significant inflection at Bin 0.23 (95% CI: 0.13–0.33; Davies’ p = 0.032). LCN2 detection rate in astrocytes was 0.04% (5/67,419); all LCN2-related interpretations are supported by orthogonal datasets (ADNI CSF proteomics, zebrafish qPCR, murine qPCR). Detailed bin-resolved data are provided in Supplementary Table S5. Quantitative modeling confirms a statistically bounded biphasic transition boundary PTGDS expression demonstrated a robust biphasic trajectory supporting a discrete neuropathological transition boundary (Fig. 1 D; Table 2 ). At the single-cell level (n = 67,419 astrocytes), quadratic modeling significantly improved fit relative to a linear model (β₂ = −2.07, p < 2.2 × 10 − 16 ), confirming strong parabolic curvature with vertex at CPS 0.46. Linear regression stratified at CPS 0.46 revealed a modest but significant pre-peak rise (β = +0.244, p = 0.000345) followed by a markedly steeper post-peak decline (β = −1.213, p < 2.2 × 10 − 16 ), indicating asymmetric dynamics of gradual compensation and accelerated collapse. Segmented regression on unsmoothed bin-level means (n = 9 bins) identified a significant early inflection at Bin 0.23 (95% CI: 0.13–0.33; Davies’ p = 0.032), marking transition from compensatory upregulation to gradual deceleration prior to peak. Change-point estimation confirmed this inflection independently of clinical staging (Supplementary Fig. S1 ). Table 2 Statistical validation of biphasic PTGDS dynamics at single-cell resolution (n = 67,419 astrocytes) Component Estimate p-value Interpretation Quadratic term β2 = − 2.07 < 2.2×10⁻¹⁶ Strong biphasic curvature Vertex CPS 0.46 — Peak inflection point Pre-peak slope + 0.244 0.000345 Compensatory rise Post-peak slope −1.213 < 2.2×10⁻¹⁶ Progressive decline Segmented inflection Bin 0.23 (95% CI: 0.13–0.33) p = 0.032 Early deceleration onset Note: Quadratic and linear regression models fitted using CPS as a continuous variable at single-cell resolution (n = 67,419 astrocytes). Vertex position derived from the quadratic model (−β1 / 2β2). Pre-peak: CPS ≤ 0.46, n = 17,874; Post-peak: CPS > 0.46, n = 49,545. Segmented regression performed on unsmoothed bin-level means (n = 9 bins) using Davies’ test. PTGDS inflection reflects astrocytic phenotypic reprogramming without evidence of cell loss Astrocyte abundance remained stable across Bins 0.4–0.8 (n = 41,141; 4,691–14,108 per bin), excluding population loss as the primary explanation for PTGDS decline. Pro-apoptotic markers showed negligible correlation with CPS (CASP3 ρ = −0.027; BAX ρ = −0.010; Supplementary Table S6), and astrocyte identity markers SLC1A2 (ρ = −0.041) and AQP4 (ρ = −0.078) were preserved. Across the CPS 0.5 boundary, astrocytes exhibited increased NF-κB-associated programs and gliosis markers (Supplementary Fig. S2), accompanied by reduction of ferroptosis-protective modules (GPX4, FTH1, SLC7A11; Supplementary Table S7). LCN2 expression increased following PTGDS decline (Table 1 ); given sparse snRNA-seq detection (0.04%), all LCN2-related interpretations rely on orthogonal validation (Fig. 2 ; Supplementary Fig. S5). Together, these findings indicate coordinated astrocytic state modulation accompanying PTGDS inflection, independent of astrocyte population loss. A temporally ordered astrocyte-neuron-microglia cascade emerges across the PTGDS inflection window Within the PTGDS inflection window (Bins 0.4–0.8), lagged cross-correlation analysis revealed a temporally ordered cascade (Table 3 ; Supplementary Table S7). Purinergic/Ca²⁺ signaling preceded PTGDS decline by one bin (Lag − 1, r = − 0.886), consistent with early compensatory sensing. NF-κB exhibited largely concurrent dynamics with PTGDS (Lag 0, r = − 0.678). PTGDS decline was inversely associated with PPARG module expression (Lag 0, r = − 0.775), indicating coordinated metabolic attenuation. LCN2 induction followed PTGDS decline (Lag + 2, r = − 0.527), coincided with strong inverse synchronization with NDUFS1 (Lag 0, r = − 0.857), and preceded TREM2 induction by one bin (Lag + 1, r = 0.743). Microglial C3 showed strong inverse correlation with neuronal NGFR (Lag 0, r = − 0.940, p = 0.0176), positioning astrocytes upstream in the cascade with microglia as secondary amplifiers. Table 3 Lagged cross-correlation analysis within the PTGDS inflection window (Bins 0.4–0.8) Interaction Pair Cell Types Lag (Bins) r-value Interpretation Purinergic → PTGDS Astro ↔ Astro −1 −0.886 Sensing precedes compensation NF-κB → PTGDS Astro ↔ Astro 0 −0.678 Concurrent inflammatory association PTGDS → PPARG Astro ↔ Astro 0 −0.775 Coordinated metabolic attenuation PPARG → LCN2 Astro ↔ Astro −1 −0.646 Concurrent rather than sequential modulation PTGDS → LCN2 Astro ↔ Astro + 2 −0.527 PTGDS decline precedes LCN2 induction LCN2 → TREM2 Astro ↔ Micro + 1 + 0.743 Astrocyte signal precedes microglial activation C3 → NGFR Micro ↔ Neuron 0 −0.940* Complement-mediated neurotoxicity LCN2 → NDUFS1 Astro ↔ Neuron 0 −0.857 Concurrent metabolic crisis coupling Note: Lagged cross-correlations computed on 3-bin moving average trajectories within the inflection window (Bins 0.4–0.8). Negative lag indicates source precedes target. Due to limited effective sample size after lagging (neff = 3–5), p-values are not reported except *p = 0.0176 for C3 → NGFR. LCN2-involving pairs should be interpreted alongside orthogonal proteomic and in vivo validation data. Full trajectory analysis in Supplementary Table S7. Post-inflection destabilization reveals subtype-selective neuronal vulnerability Comparison of excitatory neurons and SST⁺ inhibitory interneurons across CPS bins (Supplementary Fig. S6; Supplementary Table S8) demonstrated that while the PTGDS-LCN2-NGFR cascade was preserved across both neuronal classes, quantitative differences emerged. SST⁺ interneurons showed proportional reduction from 11.24% to 5.49% among classified neuronal subtypes, with higher NDUFS1 expression variance (0.19–0.26 vs 0.12–0.16 in excitatory neurons), indicating increased metabolic heterogeneity under post-inflection stress. Excitatory neurons exhibited steep early NGFR decline (0.00476 to 0.00048, Bin 0.1→0.2), consistent with trophic signaling collapse, whereas SST⁺ neurons demonstrated more pronounced BCL2 reduction (6.8% decline, Bins 0.1–0.4) and elevated BAX/BCL2 ratio. These findings indicate a division of vulnerability roles: excitatory neurons primarily reflect trophic signaling collapse, while SST⁺ interneurons exhibit reduced apoptotic buffering capacity under PTGDS-associated metabolic stress. Cross-species conservation of biphasic PTGDS dynamics in a reversible zebrafish MCI model In a reversible zebrafish MCI model (Fig. 2 A), selective cognitive impairment was confirmed by a 22.7% reduction in red ball avoidance (70.5% to 54.5%; p = 0.020; Fig. 2 B). Whole-mount double immunofluorescence (BLBP/TRITC + Nestin/GFP) revealed reactive glial dysfunction: BLBP fluorescence intensity was elevated 51% in MCI larvae (20.33 vs 13.43 a.u.; p < 0.01; Fig. 2 C–D), suppressed 47% following BXP-101 treatment (10.74 a.u.; p < 0.001 vs MCI). Nestin signal intensity declined across MCI and treatment groups (MCI: 9.27 a.u.; BXP-101 0.4 µg/ml: 8.50 a.u.; Fig. 2 E), consistent with reduced neural progenitor-associated activity under sustained inflammatory load. Longitudinal qPCR of ptgdsb.1 and ptgdsb.2 at 14 and 21 dpf revealed progressive compensatory induction, with MCI larvae exceeding control expression by 21 dpf (ptgdsb.1: 1.063 ± 0.08; ptgdsb.2: 1.066 ± 0.14; Fig. 2 F) [ 43 , 44 , 68 ], mirroring the SEA-AD compensatory PTGDS upregulation phase (CPS 0.1–0.46). An inverse relationship between ptgdsb expression and ngfr levels was observed across conditions and timepoints [ 68 ], consistent with LCN2-mediated suppression of neurogenic signaling downstream of PTGDS exhaustion. Zebrafish ptgdsb.1/2 show 85.6% amino acid identity and high structural conservation with human PTGDS (RMSD < 1.3 Å), with conserved NF-κB promoter motifs (Supplementary Fig. S5). Timing-dependent pharmacological rescue validates the therapeutic window defined by the transition boundary Early intervention during the pre-peak phase (6–14 dpf) significantly rescued cognitive deficits (BXP-101 0.6 µg/ml: 67.6% vs MCI 54.5%; p = 0.016; Fig. 2 B), whereas late intervention (15–21 dpf) yielded substantially attenuated rescue — consistent with the asymmetric trajectory dynamics at CPS 0.46. BXP-101 reduced upstream inflammatory demand on PTGDS, evidenced by dose-dependent suppression of TNF-α (MCI: 39.7 pg/ml; BXP0.6: 21.8 pg/ml, − 45% vs MCI) and IL-6 (MCI: 91.8 pg/ml; BXP0.6: 23.4 pg/ml, − 75% vs MCI) at the protein level (ELISA; Fig. 2 G). This anti-inflammatory effect was more pronounced at the protein than the mRNA level, suggesting post-transcriptional regulatory mechanisms consistent with the known modes of action of honokiol and wedelolactone as NF-κB modulators. BXP-101 treatment additionally suppressed reactive gliosis (BLBP: BXP0.4 10.74 vs MCI 20.33 a.u., − 47%; p < 0.001; Fig. 2 C–D), and preserved ptgdsb.1/2 induction capacity at 21 dpf (ptgdsb.1: 1.080 vs MCI: 1.063; Fig. 2 F), consistent with a demand-conservation model in which NF-κB suppression reduces the metabolic burden on astrocytic PTGDS buffering. BDNF expression was elevated (1.71-fold vs control) and synaptic integrity partially restored (PSD-95/dlg4: 1.27-fold vs MCI; Fig. 2 F) [ 69 , 70 ]. Network pharmacology analysis identified 12 hub targets enriched in neurogenesis and NF-κB pathways, with molecular docking confirming high-affinity engagement of GSK3B (binding affinity − 8.3 kcal/mol; Fig. 3 A–D). Behavioral assessments used the red ball avoidance paradigm (n = 30/group) [ 41 , 42 ] and Y-maze [ 48 ]. Murine validation (Fig. 4 A–F; Supplementary Fig. S4) confirmed concordant dose-dependent effects on LCN2, NGFR, and inflammatory markers. Collectively, these data demonstrate timing-dependent preservation of PTGDS-centered trajectory dynamics following NF-κB modulation, providing functional validation of the CPS 0.46 phase boundary. CSF PTGDS/LCN2 dynamics define a clinically translatable transition boundary in ADNI CSF PTGDS is a constitutively secreted lipocalin and directly reflects astrocytic PTGDS synthetic activity [ 32 ]. Analysis of 735 ADNI participants (Supplementary Table S9) revealed biphasic CSF PTGDS dynamics: elevation in EMCI (+ 23%, p = 0.008), a peak in LMCI (+ 41%, p < 0.001), and subsequent decline in AD (− 16% vs LMCI, p = 0.02; Fig. 5 A). Cross-platform validation confirmed this biphasic pattern across both TMT-MS (n = 415) and SomaScan 7K (n = 735) platforms (Supplementary Fig. S7). Segmented regression identified a PTGDS–LCN2 crossover at MMSE 26.0 (95% CI: 24.8–27.2; [ 71 ]), followed by NEFL inflection at MMSE 24.7 (95% CI: 23.1–26.3; p < 0.001; Fig. 5 B–C), defining a narrow 1.3-point MMSE interval that demarcates the clinically accessible transition window. Pseudo-time analysis independently recapitulated biphasic PTGDS dynamics with abrupt LCN2 induction at Bin 0.5–0.6 (Supplementary Fig. S2), consistent with a mid-stage astrocytic molecular inflection and corroborating the SEA-AD-derived phase boundary. The PTGDS/LCN2 ratio correlated with MMSE (r = 0.34, p < 0.001) and NEFL (β = −0.37, p < 10 − 28 ). ROC analysis demonstrated AUC = 0.743 (95% CI: 0.698–0.788) for MCI vs AD discrimination (Fig. 5 E), and the index independently predicted conversion risk (HR 3.2, 95% CI: 2.3–4.4; Fig. 5 F; Supplementary Table S4). MoCA-based segmented regression showed a concordant inflection at MoCA 21.3 (Fig. 5 D), confirming cross-instrument robustness of the clinical boundary. Threshold sensitivity analysis across MMSE inclusion criteria (24–30) demonstrated consistent AUC performance (range: 0.56–0.72), confirming robustness across clinical severity strata (Supplementary Table S4). Collectively, these multi-scale observations converge on a conserved astrocytic PTGDS inflection associated with progression from compensatory buffering toward accelerated neurodegeneration, with direct alignment to clinically measurable cognitive decline. Discussion A directionally ordered astrocyte-neuron cascade structured around a neuropathological transition boundary By integrating single-cell pseudo-progression with cross-species validation, we characterize a temporally ordered astrocyte–neuron cascade that is structured around a statistically bounded metabolic inflection during the MCI-to-AD transition. As used herein, ‘transition boundary’ denotes a statistically defined inflection interval characterized by asymmetric trajectory dynamics and cross-scale conservation; it does not imply a strict thermodynamic discontinuity. Trajectory modeling revealed marked asymmetry: compensatory dynamics accumulated gradually, whereas post-inflection destabilization accelerated disproportionately, consistent with progressive compensation followed by accelerated vulnerability once metabolic buffering capacity is exceeded. This framework is supported by: (i) pseudo-progression–based directional relationships (Table 1 , Table 3 ), (ii) pathway-level dissection of regulatory modules (Supplementary Fig. S2, Fig. S4), and (iii) functional validation across zebrafish and murine models (Fig. 2 , Fig. 4 ). Three convergent observations support a directionally consistent PTGDS–LCN2–NGFR coupling framework. The limited statistical power of lagged CCF analysis (neff = 3–5) precludes definitive causal inference from correlation coefficients alone; however, the directional consistency of cascade ordering — combined with pharmacological rescue and cross-species validation — supports the proposed temporal hierarchy as a biologically coherent framework. First, purinergic sensing precedes PTGDS decline (Lag − 1, r = − 0.886), consistent with early compensatory mechanisms. Second, PTGDS decline is followed by LCN2 induction (Lag + 2, r = − 0.527), with LCN2 showing strong inverse coupling with NDUFS1 (Lag 0, r = − 0.857) and preceding TREM2 activation (Lag + 1, r = 0.743). Third, microglial C3 inversely correlates with neuronal NGFR (Lag 0, r = − 0.940, p = 0.0176), positioning astrocytes upstream in the observed cascade, with microglia acting as secondary amplifiers. PPARG module changes were inversely rather than sequentially related to PTGDS and LCN2 trajectories, suggesting coordinated metabolic attenuation rather than a strict upstream regulatory role. While lagged correlations showed directional consistency in the core cascade, statistical significance was limited in several pairs due to small effective sample sizes after lagging (neff = 3–5). Early neuronal mitochondrial vulnerability, reflected by NDUFS1 decline (Table 1 ), coincides with compensatory astrocytic PTGDS upregulation. Following the inflection at CPS 0.46, PTGDS decline is associated with coordinated LCN2 elevation and suppression of neuronal NGFR (Supplementary Table S7). Microglial activation follows this astrocytic shift (Table 3 ), positioning astrocytes upstream in the observed cascade, with microglia acting as secondary amplifiers [ 72 ]. Subtype analysis further reveals hierarchical neuronal vulnerability (Supplementary Fig. S6): SST⁺ interneurons show proportional reduction (11.24% to 5.49%; Supplementary Table S8) and increased metabolic variance, consistent with early excitation–inhibition imbalance [ 16 ] preceding circuit destabilization. Lipid-metabolic exhaustion as a phase-triggering event in neuropathological progression Neuronal NDUFS1 decline reflects early mitochondrial stress [ 66 , 67 ]. Within astrocyte–neuron metabolic coupling [ 73 ], PTGDS functions as a homeostatic effector [ 69 ]: its gradual compensatory induction represents the cell’s attempt to maintain lipid-metabolic homeostasis under increasing energetic demand. This is not a passive biomarker response but an active buffering process whose exhaustion constitutes the phase-triggering event. By analogy to tipping point dynamics in complex systems [ 74 ], the PTGDS inflection represents a bifurcation: prior to CPS 0.46, the system retains resilience and can return to baseline; beyond it, positive feedback through LCN2-mediated inflammatory amplification drives the system toward an attractor state of accelerated vulnerability from which recovery becomes progressively less feasible. This architecture — gradual compensatory loading followed by threshold collapse — is structurally analogous to metabolic phase transitions described in oncology and cardiac remodeling [ 75 ], suggesting a conserved principle of cellular stress tolerance across disease contexts. Importantly, this inflection does not reflect astrocyte loss. Astrocyte abundance remains stable across bins 0.4–0.8 (n = 41,141), and correlations with apoptotic signatures are negligible (Supplementary Table S6). Instead, reactive markers (NFKBIA, GFAP) increase significantly across the CPS 0.46 boundary (Table 2 ), indicating phenotypic reprogramming rather than population collapse. Module-level analysis (Supplementary Fig. S2) identifies layered regulatory architecture. Purinergic and calcium-associated genes (P2RY1, GJA1) precede PTGDS induction (Table 3 ), while buffering modules (HMOX1, CLU) shift from acute to sustained compensation. NF-κB activation was concurrent with PTGDS dynamics (Lag 0, r = − 0.678; Table 3 ), suggesting that inflammatory priming and metabolic attenuation co-occur rather than following a strict sequential order. Instead, PTGDS-associated metabolic tone — potentially mediated through 15d-PGJ₂–PPAR-γ signaling [ 76 , 77 ] — may transiently restrain downstream effector induction. Transition beyond CPS 0.46 coincides with LCN2 amplification and coordinated ferroptosis-related vulnerability [ 78 , 79 ]. Quadratic modeling confirms significant curvature (β² = −2.07, p < 2.2e − 16) with a vertex at CPS 0.46 (Table 2 ), defining a statistically bounded inflection interval. Critically, the asymmetry between pre-peak slope (+ 0.244) and post-peak slope (− 1.213) quantifies the irreversibility gradient: compensatory capacity accumulates slowly and collapses rapidly, consistent with a tipping point architecture rather than symmetric oscillation. Together, these data support a structured phase boundary model in which the therapeutic window is defined not by symptom severity but by position relative to the metabolic inflection point. Intervention before CPS 0.46 — corresponding clinically to MMSE ~ 26 — targets the compensatory phase, when metabolic buffering capacity is maximal and inflammatory amplification has not yet been established. Intervention after this boundary faces a qualitatively different and less tractable target landscape. LCN2 as a boundary-associated amplifier and clinical translation LCN2 transcripts are detected in only 0.04% of astrocytes under baseline conditions, yet surge robustly at the phase boundary — a pattern consistent with a threshold-gated switch rather than a graded response. This binary-like induction profile, analogous to ‘all-or-nothing’ inflammatory priming, reinforces the existence of a discrete phase transition rather than continuous disease progression. Despite sparse LCN2 transcript detection in snRNA-seq (0.04%), consistent late-stage induction was supported across orthogonal platforms including CSF proteomics (Fig. 5 ), zebrafish qPCR (Fig. 2 F), and murine models (Fig. 4 C). Integration of longitudinal ADNI CSF proteomics (n = 735) establishes quantitative clinical alignment: the PTGDS/LCN2 crossover coincides with MMSE 26.0 and precedes NEFL inflection at 24.7 (Fig. 5 B–C), mirroring the CPS 0.46 metabolic boundary identified in SEA-AD. This narrow 1.3-point MMSE interval predicts conversion risk (HR 3.2; Fig. 5 F), demonstrating cross-scale concordance between a cellular phase boundary and clinically measurable decline. Pharmacological suppression of LCN2 restores NGFR expression and cognitive performance only prior to PTGDS exhaustion (Fig. 2 H; Fig. 4 ), reinforcing the concept that therapeutic responsiveness is constrained by this metabolically defined boundary. While pseudo-time inference cannot establish definitive causality [ 70 , 74 ], convergence across 1.3 million nuclei, orthogonal perturbation models, and longitudinal clinical proteomics supports the biological and clinical significance of this temporally bounded transition. Threshold sensitivity analysis across MMSE inclusion criteria (24–30) demonstrated consistent AUC improvement with broader inclusion, confirming robustness of the phase boundary index across clinical severity strata (Supplementary Table S4). Molecular redundancy, metabolic-trophic coupling, and oligodendrocyte vulnerability LCN2 induction occurs within a broader remodeling of lipid-associated pathways. APOE and CLU display distinct but complementary trajectories (Table 1 ), indicating partial redundancy within lipocalin-associated buffering systems. As PTGDS declines, LCN2 induction aligns with iron-linked inflammatory signaling, suggesting a qualitative shift in astrocytic metabolic state. Lagged cross-correlation analysis (Table 3 ) demonstrates that PTGDS decline is followed by PPARG module attenuation (Lag 0, r = − 0.775), consistent with weakening anti-inflammatory restraint. LCN2 induction coincides with suppression of neuronal NGFR and NDUFS1 (Lag 0, r = − 0.857; Table 3 ), indicating tight metabolic–trophic coupling between astrocytes and neurons. This strong temporal synchrony (zero-lag, high negative correlation) suggests that astrocyte-derived LCN2 acts as a secreted mediator that propagates metabolic stress to neurons. LCN2 is known to induce iron dysregulation and ferroptosis sensitivity [ 78 , 79 ], which is consistent with mitochondrial complex I vulnerability reflected by NDUFS1 decline and impairs neurotrophic signaling pathways (NGFR downregulation). Consequently, PTGDS exhaustion and subsequent LCN2 emergence in astrocytes constrain neuronal bioenergetic capacity and trophic support, reflecting a coordinated astrocyte-to-neuron metabolic–trophic failure during disease progression. Together, these data support a temporally ordered and directionally consistent coupling model in which astrocytic metabolic reprogramming precedes and predicts neuronal vulnerability. Beyond the astrocyte–neuron axis, preliminary analysis of SEA-AD oligodendrocyte trajectories revealed parallel metabolic vulnerability. Oligodendrocytic MCT4 (SLC16A3) declined by 42.2% (rho = − 0.905, p = 0.002), accompanied by significant reduction of myelin-associated genes MOG (− 14.4%, rho = − 0.810, p = 0.015) and MAG (− 10.1%, rho = − 0.762, p = 0.028; data not shown). Notably, MAPT expression increased by 22.2% in oligodendrocytes (rho = + 0.905, p = 0.002), inversely correlated with MCT4 (rho = − 0.833, p = 0.010), suggesting that energy metabolic compromise extends beyond neurons to myelinating glia. Since oligodendrocytes supply lactate to axons via MCT1, their concurrent metabolic failure implies that neurons may face energy deprivation at both the soma (via astrocytic ANLS) and along the axon (via oligodendrocytic myelin), constituting a dual supply-line disruption that may accelerate neuronal vulnerability. However, whether astrocytic reactive transformation directly contributes to oligodendrocyte pathology — for example, through A1-associated toxic lipid release — or whether these represent parallel but independent responses to a shared upstream stressor remains to be determined. Conclusions Our findings identify astrocytic PTGDS inflection as a conserved neuropathological boundary that separates reversible prodromal compensation from irreversible neurodegenerative progression in Alzheimer’s disease. The convergence of this boundary across 1.3 million single nuclei, orthogonal perturbation models, and longitudinal clinical proteomics — with cross-scale alignment at MMSE 26.0 — establishes a quantitative framework for stage-stratified intervention. The therapeutic window defined by this boundary is determined not by symptom severity alone but by molecular position relative to the PTGDS inflection point. These findings reframe prodromal AD therapeutics from symptom-driven timing to mechanism-anchored staging, with direct implications for clinical trial design and patient stratification. Limitations Several limitations warrant consideration. First, SEA-AD provides pseudo-time trajectories rather than true longitudinal sampling, limiting causal inference. Although lagged cross-correlation and timing-dependent rescue experiments support directional interpretation, definitive causality requires targeted perturbation in human-relevant systems. Second, LCN2 transcript detection in droplet-based snRNA-seq was sparse (0.04%; 5 astrocytes among 67,419), a known limitation for low-abundance or secreted transcripts. Conclusions regarding LCN2 trajectory rely on orthogonal validation platforms, including CSF proteomics and cross-species qPCR, which consistently supported pathway engagement. Third, the SEA-AD dataset is restricted to the middle temporal gyrus (MTG); thus, our findings may not fully capture the regional heterogeneity of AD progression. Future studies integrating spatial transcriptomics or multi-region atlases will be essential to validate these metabolic phase boundaries across differentially affected brain areas. Finally, pharmacologic stabilization using BXP-101 demonstrated timing-dependent efficacy consistent with the phase boundary model, with pre-inflection intervention yielding substantially greater rescue than post-inflection treatment. However, BXP-101 broadly modulates NF-κB signaling, and the observed anti-inflammatory effects — TNF-α suppression (− 45%), IL-6 suppression (− 75% at protein level) — should be interpreted as downstream consequences of upstream metabolic stabilization rather than primary mechanism. NF-κB activation in this context is better understood as amplification noise enabled by PTGDS exhaustion, not the causal driver. More specific PTGDS-directed approaches — such as PGD₂ analogs or conditional PTGDS restoration — will be required to establish direct metabolic causality and to determine whether re-crossing the phase boundary is achievable. Declarations Ethics approval and consent to participate: All animal experiments were conducted under approved IACUC protocols (Kangwon National University: KW-241104-1; Zefit Inc.: ZEFIT-IACUC-26010601-0001). Human data analysis utilized de-identified secondary data from public repositories; no additional IRB approval was required. Availability of data and materials: SEA-AD data are available at the Allen Brain Cell Atlas (https://portal.brain-map.org). ADNI CSF proteomics data are available through https://adni.loni.usc.edu upon completion of standard data use agreements. Analysis code is available at https://github.com/YoungOukKim/MCI-to-AD. Competing interests: W.M. Heo is Chief Executive Officer of BioXP, Inc., and BXP-101 is a product under development by BioXP, Inc. All other authors declare no competing interests. Funding: This research was supported by internal research funds from BioXP, Inc. Funders had no role in study design, data collection, analysis, or publication decisions. Authors’ contributions: Y.O.K., W.M.H., and S.J.P. contributed equally as co-first authors. Y.O.K. and W.M.H. conceived the study, designed all experiments, and supervised all analyses. S.J.P. led murine validation. Zebrafish experiments were conducted by Zefit Inc. Y.C.K., Y.E.C., Y.W.L., and J.Y.K. contributed to experimental work and manuscript preparation. All authors read and approved the final manuscript. Acknowledgements: The authors thank the SEA-AD consortium and ADNI investigators for publicly available datasets. The authors thank SeungHwan Kang and the Zefit team for expert experimental support. References Lopez OL (2013) Mild cognitive impairment. Continuum 19:411–424. https://doi.org/10.1212/01.CON.0000429175.29601.97 Livingston G, Huntley J, Sommerlad A et al (2020) Dementia prevention, intervention, and care: 2020 report of the Lancet Commission. Lancet 396:413–446. https://doi.org/10.1016/S0140-6736(20)30367-6 Petersen RC, Aisen PS, Beckett LA et al (2010) Alzheimer's Disease Neuroimaging Initiative (ADNI): clinical characterization. Neurology 74:201–209. https://doi.org/10.1212/WNL.0b013e3181cb3e25 Long JM, Holtzman DM (2019) Alzheimer disease: an update on pathobiology and treatment strategies. Cell 179:312–339. https://doi.org/10.1016/j.cell.2019.09.001 Selkoe DJ, Hardy J (2016) The amyloid hypothesis of Alzheimer's disease at 25 years. EMBO Mol Med 8:595–608. https://doi.org/10.15252/emmm.201606210 Sims JR, Zimmer JA, Evans CD et al (2023) Donanemab in early symptomatic Alzheimer disease. JAMA 330:512–527. https://doi.org/10.1001/jama.2023.13239 van Dyck CH, Swanson CJ, Aisen P et al (2023) Lecanemab in early Alzheimer's disease. N Engl J Med 388:9–21. https://doi.org/10.1056/NEJMoa2212948 Bradburn S, Murgatroyd C, Ray N (2019) Neuroinflammation in mild cognitive impairment and Alzheimer's disease: a meta-analysis. Ageing Res Rev 50:1–8. https://doi.org/10.1016/j.arr.2019.01.002 Calsolaro V, Edison P (2016) Neuroinflammation in Alzheimer's disease: current evidence and future directions. Alzheimers Dement 12:719–732. https://doi.org/10.1016/j.jalz.2016.02.010 Colonna M, Butovsky O (2017) Microglia function in the central nervous system during health and neurodegeneration. Annu Rev Immunol 35:441–468. https://doi.org/10.1146/annurev-immunol-051116-052358 Moreno-Jiménez EP, Flor-García M, Terreros-Roncal J et al (2019) Adult hippocampal neurogenesis is abundant in neurologically healthy subjects and drops sharply in patients with Alzheimer's disease. Nat Med 25:554–560. https://doi.org/10.1038/s41591-019-0375-9 Parkitny L, Maletic-Savatic M (2021) Glial PAMPering and DAMPening of adult hippocampal neurogenesis. Brain Sci 11:1299. https://doi.org/10.3390/brainsci11101299 Jack CR Jr, Bennett DA, Blennow K et al (2018) NIA-AA Research Framework: toward a biological definition of Alzheimer's disease. Alzheimers Dement 14:535–562. https://doi.org/10.1016/j.jalz.2018.02.018 Scheltens P, De Strooper B, Kivipelto M et al (2021) Alzheimer's disease. Lancet 397:1577–1590. https://doi.org/10.1016/S0140-6736(20)32205-4 Blennow K, Zetterberg H (2018) Biomarkers for Alzheimer's disease: current status and prospects for the future. J Intern Med 284:643–663. https://doi.org/10.1111/joim.12816 Gabitto MI, Travaglini KJ, Rachleff VM et al (2024) Integrated multimodal cell atlas of Alzheimer's disease. Nat Neurosci 27:2366–2383. https://doi.org/10.1038/s41593-024-01774-5 Li X, Wang X, Guo L et al (2023) Association between lipocalin-2 and mild cognitive impairment or dementia: a systematic review and meta-analysis. Ageing Res Rev 89:101984. https://doi.org/10.1016/j.arr.2023.101984 Naudé PJW, Nyakas C, Eiden LE et al (2012) Lipocalin 2: novel component of proinflammatory signaling in Alzheimer's disease. FASEB J 26:2811–2823. https://doi.org/10.1096/fj.11-202457 Jha MK, Lee S, Park DH et al (2015) Diverse functional roles of lipocalin-2 in the central nervous system. Neurosci Biobehav Rev 49:135–156. https://doi.org/10.1016/j.neubiorev.2014.12.006 Ferreira AC, Pinto V, Mesquita SD et al (2013) Lipocalin-2 is involved in emotional behaviors and cognitive function. Front Cell Neurosci 7:122. https://doi.org/10.3389/fncel.2013.00122 Dekens DW, Naudé PJW, Keijser JN et al (2018) Lipocalin 2 contributes to brain iron dysregulation but does not affect cognition, plaque load, and glial activation in the J20 Alzheimer mouse model. J Neuroinflammation 15:330. https://doi.org/10.1186/s12974-018-1372-5 Siddiqui T, Cosacak MI, Popova S et al (2023) Nerve growth factor receptor (Ngfr) induces neurogenic plasticity by suppressing reactive astroglial Lcn2/Slc22a17 signaling in Alzheimer's disease. npj Regen Med 8:33. https://doi.org/10.1038/s41536-023-00311-5 Devireddy LR, Gazin C, Zhu X, Green MR (2005) A cell-surface receptor for lipocalin 24p3 selectively mediates apoptosis and iron uptake. Cell 123:1293–1305. https://doi.org/10.1016/j.cell.2005.10.027 Hayden MS, Ghosh S (2012) NF-κB, the first quarter-century: remarkable progress and outstanding questions. Genes Dev 26:203–234. https://doi.org/10.1101/gad.183434.111 Liu T, Zhang L, Joo D, Sun S-C (2017) NF-κB signaling in inflammation. Signal Transduct Target Ther 2:17023. https://doi.org/10.1038/sigtrans.2017.23 Mohri I, Taniike M, Taniguchi H et al (2006) Prostaglandin D2-mediated microglia/astrocyte interaction enhances astrogliosis and demyelination in twitcher. J Neurosci 26:4383–4393. https://doi.org/10.1523/JNEUROSCI.4531-05.2006 Mohri I, Kadoyama K, Kanekiyo T et al (2007) Hematopoietic prostaglandin D synthase and DP1 receptor are selectively upregulated in microglia and astrocytes within senile plaques from human patients and in a mouse model of Alzheimer disease. J Neuropathol Exp Neurol 66:469–480. https://doi.org/10.1097/01.jnen.0000240472.43038.27 Choi D-J, An J, Jou I et al (2019) A Parkinson's disease gene, DJ-1, regulates anti-inflammatory roles of astrocytes through prostaglandin D2 synthase expression. Neurobiol Dis 127:482–491. https://doi.org/10.1016/j.nbd.2019.04.003 Kanekiyo T, Ban T, Aritake K et al (2007) Lipocalin-type prostaglandin D synthase/beta-trace is a major amyloid beta-chaperone in human cerebrospinal fluid. Proc Natl Acad Sci USA 104:6412–6417. https://doi.org/10.1073/pnas.0701585104 Higginbotham L, Ping L, Dammer EB et al (2020) Integrated proteomics reveals brain-based cerebrospinal fluid biomarkers in asymptomatic and symptomatic Alzheimer's disease. Sci Adv 6:eaaz9360. https://doi.org/10.1126/sciadv.aaz9360 Johnson ECB, Dammer EB, Duong DM et al (2020) Large-scale proteomic analysis of Alzheimer's disease brain and cerebrospinal fluid reveals early changes in energy metabolism associated with microglia and astrocyte activation. Nat Med 26:769–780. https://doi.org/10.1038/s41591-020-0815-6 Urade Y (2021) Biochemical and structural characteristics, gene regulation, physiological, pathological and clinical features of lipocalin-type prostaglandin D2 synthase as a multifunctional lipocalin. Front Physiol 12:718002. https://doi.org/10.3389/fphys.2021.718002 Wolock SL, Lopez R, Klein AM (2019) Scrublet: computational identification of cell doublets in single-cell transcriptomic data. Cell Syst 8:281–291. https://doi.org/10.1016/j.cels.2018.11.005 Lu K-P, Chang S-T (2023) An advanced segmentation approach to piecewise regression models. Mathematics 11:4959. https://doi.org/10.3390/math11244959 Jin S, Guerrero-Juarez CF, Zhang L et al (2021) Inference and analysis of cell-cell communication using CellChat. Nat Commun 12:1088. https://doi.org/10.1038/s41467-021-21246-9 Aibar S, Bravo González-Blas C, Moerman T et al (2017) SCENIC: single-cell regulatory network inference and clustering. Nat Methods 14:1083–1086. https://doi.org/10.1038/nmeth.4463 Howe K, Clark MD, Torroja CF et al (2013) The zebrafish reference genome sequence and its relationship to the human genome. Nature 496:498–503. https://doi.org/10.1038/nature12111 Fontana BD, Mezzomo NJ, Kalueff AV, Rosemberg DB (2018) The developing utility of zebrafish models of neurological and neuropsychiatric disorders: a critical review. Exp Neurol 299:157–171. https://doi.org/10.1016/j.expneurol.2017.10.004 Azman KF, Zakaria R (2019) D-galactose-induced accelerated aging model: an overview. Biogerontology 20:763–782. https://doi.org/10.1007/s10522-019-09837-y Adams MM, Kafaligonul H (2018) Zebrafish — a model organism for studying the neurobiological mechanisms underlying cognitive brain aging and use of potential interventions. Front Cell Dev Biol 6:135. https://doi.org/10.3389/fcell.2018.00135 Colwill RM, Creton R (2011) Imaging escape and avoidance behavior in zebrafish larvae. Rev Neurosci 22:63–73. https://doi.org/10.1515/RNS.2011.008 Pelkowski SD, Kapoor M, Richendrfer HA et al (2011) A novel high-throughput imaging system for automated analyses of avoidance behavior in zebrafish larvae. Behav Brain Res 223:135–144. https://doi.org/10.1016/j.bbr.2011.04.033 Fujimori K, Inui T, Uodome N et al (2006) Zebrafish and chicken lipocalin-type prostaglandin D synthase homologues: conservation of mammalian gene structure and binding ability for lipophilic molecules. Gene 375:14–25. https://doi.org/10.1016/j.gene.2006.01.037 Mihaljevic I, Popovic M, Zaja R, Smital T (2016) Phylogenetic, syntenic, and tissue expression analysis of slc22 genes in zebrafish (Danio rerio). BMC Genomics 17:626. https://doi.org/10.1186/s12864-016-2981-y Lee Y-J, Lee YM, Lee C-K et al (2011) Therapeutic applications of compounds in the Magnolia family. Pharmacol Ther 130:157–176. https://doi.org/10.1016/j.pharmthera.2011.01.010 Tao X, Zhao M, Gao K et al (2025) Wedelolactone ameliorates ischemic stroke by inhibiting oxidative damage and ferroptosis via HIF-1α/SLC7A11/GPX4 signaling. Drug Des Devel Ther 19:6849–6868. https://doi.org/10.2147/DDDT.S528831 Yang L, Ji C, Li Y et al (2020) Natural potent NAAA inhibitor atractylodin counteracts LPS-induced microglial activation. Front Pharmacol 11:577319. https://doi.org/10.3389/fphar.2020.577319 Kraeuter A-K, Guest PC, Sarnyai Z (2019) The Y-maze for assessment of spatial working and reference memory in mice. Methods Mol Biol 1916:105–111. https://doi.org/10.1007/978-1-4939-8994-2_10 Mucke L, Selkoe DJ (2012) Neurotoxicity of amyloid β-protein: synaptic and network dysfunction. Cold Spring Harb Perspect Med 2:a006338. https://doi.org/10.1101/cshperspect.a006338 Daina A, Michielin O, Zoete V (2019) SwissTargetPrediction: updated data and new features for efficient prediction of protein targets of small molecules. Nucleic Acids Res 47:W357–W364. https://doi.org/10.1093/nar/gkz382 Wang X, Shen Y, Wang S et al (2017) PharmMapper 2017 update: a web server for potential drug target identification with a comprehensive target pharmacophore database. Nucleic Acids Res 45:W356–W360. https://doi.org/10.1093/nar/gkx374 Piñero J, Ramírez-Anguita JM, Saüch-Pitarch J et al (2020) The DisGeNET knowledge platform for disease genomics: 2019 update. Nucleic Acids Res 48:D845–D855. https://doi.org/10.1093/nar/gkz1021 Amberger JS, Bocchini CA, Schiettecatte F et al (2015) OMIM.org: Online Mendelian Inheritance in Man, an online catalog of human genes and genetic disorders. Nucleic Acids Res 43:D789–D798. https://doi.org/10.1093/nar/gku1205 Stelzer G, Rosen N, Plaschkes I et al (2016) The GeneCards suite: from gene data mining to disease genome sequence analyses. Curr Protoc Bioinformatics 55:1.30.1–1.30.33. https://doi.org/10.1002/cpbi.5 Szklarczyk D, Kirsch R, Koutrouli M et al (2023) The STRING database in 2023: protein-protein association networks and functional enrichment analyses for any sequenced genome of interest. Nucleic Acids Res 51:D638–D646. https://doi.org/10.1093/nar/gkac1000 Shannon P, Markiel A, Ozier O et al (2003) Cytoscape: a software environment for integrated models of biomolecular interaction networks. Genome Res 13:2498–2504. https://doi.org/10.1101/gr.1239303 Zhou Y, Zhou B, Pache L et al (2019) Metascape provides a biologist-oriented resource for the analysis of systems-level datasets. Nat Commun 10:1523. https://doi.org/10.1038/s41467-019-09234-6 Eberhardt J, Santos-Martins D, Tillack AF, Forli S (2021) AutoDock Vina 1.2.0: new docking methods, expanded force field, and Python bindings. J Chem Inf Model 61:3891–3898. https://doi.org/10.1021/acs.jcim.1c00203 Johnson WE, Li C, Rabinovic A (2007) Adjusting batch effects in microarray expression data using empirical Bayes methods. Biostatistics 8:118–127. https://doi.org/10.1093/biostatistics/kxj037 Cleveland WS (1981) LOWESS: a program for smoothing scatter plots by robust locally weighted regression. Am Stat 35:54. https://doi.org/10.2307/2683591 Liu C-C, Kanekiyo T, Xu H, Bu G (2013) Apolipoprotein E and Alzheimer disease: risk, mechanisms and therapy. Nat Rev Neurol 9:106–118. https://doi.org/10.1038/nrneurol.2012.263 Bates D, Mächler M, Bolker B, Walker S (2015) Fitting linear mixed-effects models using lme4. J Stat Softw 67:1–48. https://doi.org/10.18637/jss.v067.i01 Kaplan EL, Meier P (1958) Nonparametric estimation from incomplete observations. J Am Stat Assoc 53:457–481 Cox DR (1972) Regression models and life-tables. J R Stat Soc B 34:187–202 Benjamini Y, Hochberg Y (1995) Controlling the false discovery rate: a practical and powerful approach to multiple testing. J R Stat Soc B 57:289–300 Adav SS, Park JE, Sze SK (2019) Quantitative profiling brain proteomes revealed mitochondrial dysfunction in Alzheimer's disease. Mol Brain 12:8. https://doi.org/10.1186/s13041-019-0430-y Wang X, Zhang Z (2020) Mitochondria dysfunction in the pathogenesis of Alzheimer's disease: recent advances. Mol Neurodegener 15:30. https://doi.org/10.1186/s13024-020-00376-6 Underwood CK, Coulson EJ (2008) The p75 neurotrophin receptor. Int J Biochem Cell Biol 40:1664–1668. https://doi.org/10.1016/j.biocel.2007.06.010 Pitta S, Augustine BB, Kasala ER, Sulakhiya K, Ravindranath V, Lahkar M (2013) Honokiol reverses depressive-like behavior and decrease in brain BDNF levels induced by chronic corticosterone injections in mice. Pharmacognosy J 5:211–215. https://doi.org/10.1016/j.phcgj.2013.08.004 Bustos FJ, Ampuero E, Jury N, Aguilar R, Falahi F, Toledo J, Ahumada J, Lata J, Cubillos P, Henríquez B, Guerra MV, Stehberg J, Neve RL, Inestrosa NC, Wyneken U, Fuenzalida M, Härtel S, Sena-Esteves M, Varela-Nallar L, Rots MG, Montecino M, van Zundert B (2017) Epigenetic editing of the Dlg4/PSD95 gene improves cognition in aged and Alzheimer's disease mice. Brain 140:3252–3268. https://doi.org/10.1093/brain/awx272 Folstein MF, Folstein SE, McHugh PR (1975) 'Mini-mental state'. A practical method for grading the cognitive state of patients for the clinician. J Psychiatr Res 12:189–198 Prater KE, Green KJ, Mamde S et al (2023) Human microglia show unique transcriptional changes in Alzheimer's disease. Nat Aging 3:894–907. https://doi.org/10.1038/s43587-023-00424-y Magistretti PJ, Allaman I (2018) Lactate in the brain: from metabolic end-product to signalling molecule. Nat Rev Neurosci 19:235–249. https://doi.org/10.1038/nrn.2018.19 Scheffer M, Carpenter SR, Lenton TM et al (2012) Anticipating critical transitions. Science 338:344–348. https://doi.org/10.1126/science.1225244 Heineke J, Molkentin JD (2006) Regulation of cardiac hypertrophy by intracellular signalling pathways. Nat Rev Mol Cell Biol 7:589–600. https://doi.org/10.1038/nrm1983 Yagami T, Yamamoto Y, Koma H (2018) Physiological and pathological roles of 15-deoxy-delta12,14-prostaglandin J2 in the central nervous system and neurological diseases. Mol Neurobiol 55:2227–2248. https://doi.org/10.1007/s12035-017-0435-4 Scher JU, Pillinger MH (2005) 15d-PGJ2: the anti-inflammatory prostaglandin? Clin Immunol 114:100–109. https://doi.org/10.1016/j.clim.2004.09.008 Zhou Z, Zhang Y, Liu S et al (2025) Ferroptosis in Alzheimer's disease: molecular mechanisms and advances in therapeutic strategies. Front Neurosci 19:1673315. https://doi.org/10.3389/fnins.2025.1673315 Ashraf A, Jeandriens J, Parkes HG, So P-W (2020) Iron dyshomeostasis, lipid peroxidation and perturbed expression of cystine/glutamate antiporter in Alzheimer's disease: evidence of ferroptosis. Redox Biol 32:101494. https://doi.org/10.1016/j.redox.2020.101494 Additional Declarations Competing interest reported. Y.O.K., W.M.H., Y.C.K. are employees of BioXP, Inc. J.Y.K. is an employee of Zefit Inc. The authors declare that the research was conducted in the absence of any other financial or non-financial relationships that could be construed as a potential conflict of interest. Supplementary Files SupplementaryDatafile.docx Cite Share Download PDF Status: Under Review Version 1 posted Reviewers agreed at journal 10 May, 2026 Reviewers invited by journal 26 Apr, 2026 Editor assigned by journal 25 Apr, 2026 Submission checks completed at journal 24 Apr, 2026 First submitted to journal 22 Apr, 2026 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-9499795","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":628396985,"identity":"ba4c4f3c-206f-4032-b547-722e785fba0f","order_by":0,"name":"YoungOuk Kim","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAwUlEQVRIiWNgGAWjYFAC5gaGhAobIIOx8QCRWhgbGB6cSQMziNfC+LDlMJhJnBb59sbGD4kN5+3Wth8G2lJjE01Qi8GZg80SiTtuJ287kwjUciwtt4GgFonEBonEM7eTzQ4AtTA2HCasRX5GYvOPxLZzyWbnHxKpheFGYptEYtsBO7MbxNoC9EubRcKZ5ASzG0BbEojxi3x78+GbPyrs7M3Opz988KHGhgiHQUEiWGUCscpBwJ4UxaNgFIyCUTDCAACevE3rXwHGWgAAAABJRU5ErkJggg==","orcid":"","institution":"BioXP Research Institute","correspondingAuthor":true,"prefix":"","firstName":"YoungOuk","middleName":"","lastName":"Kim","suffix":""},{"id":628396986,"identity":"a8df589e-4a68-4dca-b024-36bcb43db6cb","order_by":1,"name":"WooMyung Heo","email":"","orcid":"","institution":"BioXP Research Institute","correspondingAuthor":false,"prefix":"","firstName":"WooMyung","middleName":"","lastName":"Heo","suffix":""},{"id":628396987,"identity":"b5fabf76-deb4-47d4-938d-c144a95ddb1f","order_by":2,"name":"Se Jin Park","email":"","orcid":"","institution":"Kangwon National University","correspondingAuthor":false,"prefix":"","firstName":"Se","middleName":"Jin","lastName":"Park","suffix":""},{"id":628396988,"identity":"611ab0ea-4fa6-4fbf-9ae4-87a6d8e68bc1","order_by":3,"name":"YoungChul Kim","email":"","orcid":"","institution":"BioXP Research Institute","correspondingAuthor":false,"prefix":"","firstName":"YoungChul","middleName":"","lastName":"Kim","suffix":""},{"id":628396989,"identity":"56e6ef8a-6475-4d35-a66b-92ec4e27c71f","order_by":4,"name":"Ye Eun Cho","email":"","orcid":"","institution":"Kangwon National University","correspondingAuthor":false,"prefix":"","firstName":"Ye","middleName":"Eun","lastName":"Cho","suffix":""},{"id":628396990,"identity":"00e3cb88-8016-4f45-8662-f5656ba02a4d","order_by":5,"name":"Ye-Won Lee","email":"","orcid":"","institution":"Kangwon National University","correspondingAuthor":false,"prefix":"","firstName":"Ye-Won","middleName":"","lastName":"Lee","suffix":""},{"id":628396991,"identity":"c4ad4e44-d21e-4e84-8dc0-7d229311e728","order_by":6,"name":"JungYeon Kim","email":"","orcid":"","institution":"Zefit Inc","correspondingAuthor":false,"prefix":"","firstName":"JungYeon","middleName":"","lastName":"Kim","suffix":""}],"badges":[],"createdAt":"2026-04-22 19:09:27","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-9499795/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9499795/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":107694313,"identity":"4d0cbc5d-28e9-4afe-aa12-ae0fc099ae7b","added_by":"auto","created_at":"2026-04-24 06:41:32","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":296195,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSingle-nucleus trajectories in SEA-AD reveal an astrocytic PTGDS phase boundary.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cem\u003e\u003cstrong\u003eA.\u003c/strong\u003e\u003c/em\u003e\u003cem\u003e Study schema integrating SEA-AD snRNA-seq (1.3M nuclei, n=84 donors), reversible zebrafish MCI model, and longitudinal ADNI CSF proteomics (n=735). \u003c/em\u003e\u003cem\u003e\u003cstrong\u003eB.\u003c/strong\u003e\u003c/em\u003e\u003cem\u003e Mechanistic framework: PTGDS-LCN2-NGFR axis from homeostatic buffering to neurotoxic signaling. \u003c/em\u003e\u003cem\u003e\u003cstrong\u003eC.\u003c/strong\u003e\u003c/em\u003e\u003cem\u003e Neuronal NDUFS1 decline from Bin 0.1; concurrent compensatory PTGDS upregulation (Table 1). \u003c/em\u003e\u003cem\u003e\u003cstrong\u003eD.\u003c/strong\u003e\u003c/em\u003e\u003cem\u003eBiphasic PTGDS trajectory: significant slope change at Bin 0.23 (Davies’ p=0.032), peaking at CPS 0.46 (raw mean=2.02). \u003c/em\u003e\u003cem\u003e\u003cstrong\u003eE.\u003c/strong\u003e\u003c/em\u003e\u003cem\u003e Quadratic model curvature (β2=−2.07, p\u0026lt;2.2e−16); vertex at CPS 0.46. Pre-peak slope +0.244, post-peak slope −1.213. Data are mean±SEM. *p\u0026lt;0.05.\u003c/em\u003e\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-9499795/v1/48630b18edfee56062086c00.png"},{"id":107694317,"identity":"fb04ed12-0235-4543-83ce-6aafc883f25f","added_by":"auto","created_at":"2026-04-24 06:41:33","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":164438,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eCross-species conservation of the PTGDS phase boundary and timing-dependent pharmacological rescue in zebrafish MCI.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cem\u003e\u003cstrong\u003eA.\u003c/strong\u003e\u003c/em\u003e\u003cem\u003e Experimental timeline: triple-stressor MCI induction (from 6 dpf); early BXP-101 intervention (6–14 dpf) vs late (15–21 dpf). \u003c/em\u003e\u003cem\u003e\u003cstrong\u003eB.\u003c/strong\u003e\u003c/em\u003e\u003cem\u003eRed ball avoidance at 14 dpf: MCI 22.7% reduction (70.5% to 54.5%; p=0.020 vs Control). BXP-101 0.6 μg/ml rescued to 67.6% (p=0.016 vs MCI). Donepezil positive control: 70.7%. \u003c/em\u003e\u003cem\u003e\u003cstrong\u003eC–D.\u003c/strong\u003e\u003c/em\u003e\u003cem\u003e Whole-mount double immunofluorescence (BLBP/TRITC + Nestin/GFP) at 14 dpf: reactive gliosis (BLBP) and neural progenitor state (Nestin) quantified by mean fluorescence intensity (ImageJ, fixed ROI; n=10 larvae/group). \u003c/em\u003e\u003cem\u003e\u003cstrong\u003eE.\u003c/strong\u003e\u003c/em\u003e\u003cem\u003e Representative fluorescence images (Control / MCI / MCI + BXP-101 0.4 μg/ml). \u003c/em\u003e\u003cem\u003e\u003cstrong\u003eF.\u003c/strong\u003e\u003c/em\u003e\u003cem\u003e Longitudinal ptgdsb.1/2 qPCR at 14 and 21 dpf. MCI exceeds control by 21 dpf (ptgdsb.1: 1.063±0.08; ptgdsb.2: 1.066±0.14), consistent with ongoing compensatory induction (SEA-AD CPS 0.1–0.46 phase). BXP-101 0.4 μg/ml preserved induction (ptgdsb.1: 1.080 vs MCI: 1.063). BDNF: 1.71-fold vs control; PSD-95/dlg4: 1.27-fold vs MCI. \u003c/em\u003e\u003cem\u003e\u003cstrong\u003eG.\u003c/strong\u003e\u003c/em\u003e\u003cem\u003eELISA (14 dpf): TNF-α −45% (MCI: 39.7 pg/ml; BXP0.6: 21.8 pg/ml); IL-6 −75% (MCI: 91.8 pg/ml; BXP0.6: 23.4 pg/ml). Anti-inflammatory effect more pronounced at protein than mRNA level, indicating post-transcriptional NF-κB modulation. \u003c/em\u003e\u003cem\u003e\u003cstrong\u003eH.\u003c/strong\u003e\u003c/em\u003e\u003cem\u003eMolecular recovery schematic: NF-κB suppression → preserved ptgdsb.1/2 capacity → TNF-α↓ / IL-6↓ → NGFR re-expression. Data are mean±SEM. n=10 larvae/group (WIF); n=3 replicates/group (14 dpf qPCR); n=4 replicates/group (21 dpf qPCR). *p\u0026lt;0.05, **p\u0026lt;0.01, ***p\u0026lt;0.001 vs Control; #p\u0026lt;0.05, ###p\u0026lt;0.001 vs MCI.\u003c/em\u003e\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-9499795/v1/d9d890ad15633324195bbdaf.png"},{"id":107694315,"identity":"6d4c5ed1-73df-4ea7-9378-ecc3ff9d158d","added_by":"auto","created_at":"2026-04-24 06:41:33","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":293223,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eNetwork pharmacology and molecular docking identify NF-κB/GSK3B as mechanistic hubs of BXP-101 action.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cem\u003e\u003cstrong\u003eA.\u003c/strong\u003e\u003c/em\u003e\u003cem\u003e Venn diagram of target overlap among atractylodin, honokiol, and wedelolactone. \u003c/em\u003e\u003cem\u003e\u003cstrong\u003eB.\u003c/strong\u003e\u003c/em\u003e\u003cem\u003e PPI network of top 12 hub targets. \u003c/em\u003e\u003cem\u003e\u003cstrong\u003eC.\u003c/strong\u003e\u003c/em\u003e\u003cem\u003eTop enriched pathways and GO terms (−log₁₀P). \u003c/em\u003e\u003cem\u003e\u003cstrong\u003eD.\u003c/strong\u003e\u003c/em\u003e\u003cem\u003e Representative docking poses: wedelolactone–GSK3B, honokiol–PTGS2. *p\u0026lt;0.05; Benjamini–Hochberg correction applied.\u003c/em\u003e\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-9499795/v1/37d167dc7a165b763ca87e07.png"},{"id":107694320,"identity":"54b365d1-ac93-4218-9b2c-a1acaed565d1","added_by":"auto","created_at":"2026-04-24 06:41:33","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":161620,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eBXP-101 rescues cognitive and molecular phenotypes in Aβ1–42 mice.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cem\u003e\u003cstrong\u003eA–B.\u003c/strong\u003e\u003c/em\u003e\u003cem\u003e Dose-dependent rescue of Y-maze spontaneous alternation and passive avoidance latency. \u003c/em\u003e\u003cem\u003e\u003cstrong\u003eC–D.\u003c/strong\u003e\u003c/em\u003e\u003cem\u003e Suppression of astrocytic LCN2 and pro-inflammatory cytokines (TNF-α, IL-6, IL-1β). \u003c/em\u003e\u003cem\u003e\u003cstrong\u003eE.\u003c/strong\u003e\u003c/em\u003e\u003cem\u003e Preservation of neuronal NGFR and BDNF. \u003c/em\u003e\u003cem\u003e\u003cstrong\u003eF.\u003c/strong\u003e\u003c/em\u003e\u003cem\u003e Normalization of Keap1; upregulation of APOE and ABCA1. Data are mean±SEM (n=8–10/group). One-way ANOVA + Tukey post-hoc. *p\u0026lt;0.05 vs Sham; #p\u0026lt;0.05 vs Aβ1–42.\u003c/em\u003e\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-9499795/v1/3a8b6eeee7a1c59d5d0afee3.png"},{"id":107694321,"identity":"da61a26d-8b68-4edf-8ed7-f570ce666644","added_by":"auto","created_at":"2026-04-24 06:41:33","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":235657,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eClinical translation of the PTGDS/LCN2 phase boundary in ADNI CSF (n=735).\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cem\u003e\u003cstrong\u003eA.\u003c/strong\u003e\u003c/em\u003e\u003cem\u003e Biphasic CSF PTGDS trajectory across ADNI diagnostic groups: EMCI +23% (p=0.008), LMCI +41% (p\u0026lt;0.001), AD −16% vs LMCI (p=0.02). \u003c/em\u003e\u003cem\u003e\u003cstrong\u003eB–C.\u003c/strong\u003e\u003c/em\u003e\u003cem\u003eSegmented regression: PTGDS–LCN2 crossover at MMSE 26.0 (95% CI 24.8–27.2); NEFL inflection at MMSE 24.7 (95% CI 23.1–26.3; p\u0026lt;0.001), defining a 1.3-point MMSE transition interval. \u003c/em\u003e\u003cem\u003e\u003cstrong\u003eD.\u003c/strong\u003e\u003c/em\u003e\u003cem\u003e MoCA-based segmented regression showing concordant inflection at MoCA 21.3, confirming cross-instrument robustness of the clinical boundary. \u003c/em\u003e\u003cem\u003e\u003cstrong\u003eE.\u003c/strong\u003e\u003c/em\u003e\u003cem\u003e ROC analysis: AUC 0.743 (95% CI 0.698–0.788) for MCI vs AD discrimination (logistic regression adjusted for PTGDS/LCN2 ratio, NEFL, total Tau, age, APOE ε4, Aβ42, pTau181). \u003c/em\u003e\u003cem\u003e\u003cstrong\u003eF.\u003c/strong\u003e\u003c/em\u003e\u003cem\u003eKaplan–Meier survival analysis: PTGDS/LCN2 index predicts MCI-to-AD conversion risk (HR 3.2, 95% CI 2.3–4.4; n=379 MCI participants, median follow-up 3.8 years). *p\u0026lt;0.05. All analyses adjusted for age, sex, education, APOE ε4.\u003c/em\u003e\u003c/p\u003e","description":"","filename":"5.png","url":"https://assets-eu.researchsquare.com/files/rs-9499795/v1/c8569e07e15032586a483c20.png"},{"id":107708046,"identity":"b23548d9-8e74-4013-aac5-bf2ef797d5dc","added_by":"auto","created_at":"2026-04-24 09:21:45","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1398781,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9499795/v1/340218fd-efb4-4d68-bb6f-201224f46042.pdf"},{"id":107694312,"identity":"bd361ff6-9a6c-46bb-98df-63142b91ba8b","added_by":"auto","created_at":"2026-04-24 06:41:32","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":1186901,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryDatafile.docx","url":"https://assets-eu.researchsquare.com/files/rs-9499795/v1/14c82ed77a88fa3eeb949621.docx"}],"financialInterests":"Competing interest reported. Y.O.K., W.M.H., Y.C.K. are employees of BioXP, Inc. J.Y.K. is an employee of Zefit Inc. The authors declare that the research was conducted in the absence of any other financial or non-financial relationships that could be construed as a potential conflict of interest.","formattedTitle":"Astrocytic PTGDS inflection defines a neuropathological transition boundary during prodromal Alzheimer’s disease","fulltext":[{"header":"Introduction","content":"\u003cp\u003eMild cognitive impairment (MCI) represents a pivotal yet poorly resolved stage in Alzheimer\u0026rsquo;s disease (AD) progression. Clinically defined as a prodromal state preceding dementia, annual MCI-to-AD conversion rates span 5\u0026ndash;39%, and a substantial proportion of individuals remain stable or revert to normal cognition [\u003cspan additionalcitationids=\"CR2\" citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]. This heterogeneity implies that MCI does not represent a single biological state but rather a spectrum of substates differing fundamentally in reversibility and pathological commitment. Existing staging frameworks based on amyloid burden, tau phosphorylation, and clinical cognitive scores [\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e, \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e] fail to resolve reversible compensation from progressive neurodegenerative commitment at the molecular level. The central unresolved question in prodromal AD is therefore the identification of a \u003cem\u003etransition boundary\u003c/em\u003e \u0026mdash; the molecular event that separates reversible compensatory states from irreversible neurodegeneration.\u003c/p\u003e \u003cp\u003eThe limited disease-modifying efficacy of amyloid-targeted therapies [\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e, \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e] has redirected attention toward upstream events preceding overt neurodegeneration. Evidence increasingly implicates glial metabolic remodeling and impaired neurogenic support as determinants of neuronal instability that emerge before structural neuronal loss [\u003cspan additionalcitationids=\"CR9 CR10 CR11 CR12 CR13 CR14\" citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e], raising the possibility that the critical disease transition is metabolic rather than proteopathic in origin. Neuroinflammatory activation alone does not adequately explain this transition, as inflammatory markers are elevated across the MCI spectrum without reliably distinguishing stable from progressive cases. A biologically meaningful transition boundary would therefore require a threshold-like molecular event capable of collapsing accumulated compensatory capacity and initiating downstream amplification cascades.\u003c/p\u003e \u003cp\u003eThe SEA-AD atlas (Gabitto et al., 2024) [\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e], comprising 1.3\u0026nbsp;million nuclei from 84 donors aligned along a validated continuous pseudo-progression score (CPS), provides an unprecedented framework to interrogate this question. Within this dataset, astrocytic trajectories reveal a biphasic pattern characterized by early compensatory activation followed by abrupt attenuation at intermediate disease stages. Whether this inflection constitutes a true neuropathological transition boundary \u0026mdash; defined by non-linear dynamics, statistical threshold behavior, and cross-species conservation \u0026mdash; or reflects continuous gradual drift has remained unresolved.\u003c/p\u003e \u003cp\u003eAmong candidate molecular mediators, lipocalin-2 (LCN2), an NF-κB-regulated inflammatory amplifier elevated in MCI and AD [\u003cspan additionalcitationids=\"CR18\" citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e], links astrocytic dysfunction to synaptic instability and impaired CREB/NGFR-associated neurogenic signaling [\u003cspan additionalcitationids=\"CR21 CR22 CR23 CR24\" citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e]. Upstream of this cascade, prostaglandin D₂ synthase (PTGDS) represents a compelling astrocytic regulator. PTGDS integrates inflammatory tone with neurogenic support [\u003cspan additionalcitationids=\"CR27\" citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e] and functions as a major amyloid-β chaperone [\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e], placing it at the convergence of metabolic buffering and neuroprotection. Among astrocyte-enriched transcripts in the SEA-AD atlas, PTGDS demonstrated the most pronounced biphasic inflection and the highest astrocyte-selective expression ratio within CPS 0.3\u0026ndash;0.5, distinguishing it from other reactive astrocytic markers including CLU, GFAP, and AQP4. Conflicting human CSF data \u0026mdash; PTGDS elevation in some cohorts [\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e, \u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e] and reduction in others [\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e] \u0026mdash; are consistent with stage-dependent biphasic dynamics obscured by cross-sectional sampling.\u003c/p\u003e \u003cp\u003eHere, we test the hypothesis that astrocytic PTGDS defines a statistically bounded neuropathological transition boundary during the MCI-to-AD continuum. Using an integrated framework comprising (1) segmented regression and quadratic modeling within SEA-AD single-cell pseudo-progression, (2) a reversible zebrafish MCI model enabling temporal dissection of compensatory failure, (3) murine perturbation systems for mechanistic validation, and (4) longitudinal ADNI CSF proteomics, we identify a lipid-metabolic exhaustion event at CPS 0.46 that marks a measurable neuropathological transition boundary, establish PTGDS as its molecular anchor, and define a pre-boundary therapeutic window for stage-stratified intervention.\u003c/p\u003e"},{"header":"Materials and methods","content":"\u003cp\u003eEthical approval and animal husbandry\u003c/p\u003e\n\u003cp\u003eAll mouse experiments were approved by the IACUC of Kangwon National University (Approval No. KW-241104-1) and conducted under controlled environmental conditions (23 \u0026plusmn; 2\u0026deg;C, 50 \u0026plusmn; 10% humidity, 12 h light/dark cycle). All zebrafish experiments were performed at Zefit Inc., an FDA-registered CRO (Approval No. ZEFIT-IACUC-26010601-0001). Zebrafish were maintained at 28.5\u0026deg;C, 14:10 h light:dark cycle, pH 7.0\u0026ndash;7.5. Human dataset analyses utilized de-identified secondary data from public repositories (SEA-AD, ADNI); no additional IRB approval was required.\u003c/p\u003e\n\u003cp\u003eSEA-AD snRNA-seq trajectory analysis\u003c/p\u003e\n\u003cp\u003eWe analyzed 1.3 million nuclei from the SEA-AD middle temporal gyrus atlas [16] (quality control: \u0026ge;500 genes/nucleus, \u0026lt;20% mitochondrial reads; Scrublet-based doublet removal [33]). CPS values were rounded to one decimal to define nine discrete progression bins (Bins 0.1\u0026ndash;0.9), approximately aligned to Braak stages: Bin 0.0\u0026ndash;0.3 (~Braak I\u0026ndash;II), Bin 0.4\u0026ndash;0.6 (~Braak III\u0026ndash;IV), Bin 0.7\u0026ndash;0.9 (~Braak V\u0026ndash;VI). This mapping is heuristic and does not imply direct pathological equivalence. LCN2 showed an extremely low detection rate (0.04%) and was independently validated by CSF proteomics (ADNI; Fig. 5), zebrafish qPCR (Fig. 2F), and murine qPCR (Fig. 4). Quality control metrics and bin-to-Braak stage mapping are detailed in Supplementary Table S1.\u003c/p\u003e\n\u003cp\u003eTrajectory smoothing and cross-correlation analysis\u003c/p\u003e\n\u003cp\u003eBin-level means were computed from log-normalized expression per cell type and smoothed using a 3-bin centered moving average to mitigate stochastic dropout. Lagged cross-correlation functions (ccf() in R, lag.max = 3) were computed on the inflection window (Bins 0.4\u0026ndash;0.8) and full trajectory (Bins 0.1\u0026ndash;0.9). Approximate p-values were derived from t-test transformation of the peak correlation coefficient, with effective sample size adjusted for lag (neff = n \u0026minus; |lag|). The inflection window was selected to encompass the complete PTGDS biphasic trajectory from compensatory peak to exhaustion. All analyses are fully reproducible using R v4.3.2 with set.seed(42); analysis code and intermediate outputs are deposited at https://github.com/YoungOukKim/MCI-to-AD.\u003c/p\u003e\n\u003cp\u003eSegmented regression and change-point analysis\u003c/p\u003e\n\u003cp\u003eSegmented regression was performed using the segmented R package [34] with initial breakpoint estimate at 0.5; Davies\u0026rsquo; test assessed breakpoint significance. Change-point estimation was conducted independently of ADNI clinical staging (Supplementary Fig. S1). Quadratic regression at single-cell resolution (n = 67,419 astrocytes) used ordinary least squares. Model comparison employed ANOVA and AIC criteria. Peak location derived from the vertex (\u0026minus;\u0026beta;1 / 2\u0026beta;2). LOESS smoothing applied independently to visualize trend consistency.\u003c/p\u003e\n\u003cp\u003eComputational analysis of neuronal vulnerability\u003c/p\u003e\n\u003cp\u003eIntercellular signaling was inferred using CellChat (v1.6.1) [35], focusing on LCN2\u0026ndash;SLC22A17 interactions. Neuronal stress responses were quantified using SCENIC (v1.2.4) [36] for ATF3 and EGR1 regulon activity. Modules analyzed included NF-\u0026kappa;B Priming (PTGS2, IL6ST, NFKBIA), Purinergic/Ca\u0026sup2;⁺ Sensing (P2RY1, P2RY12, GJA1, ITPR2), Metabolic Buffering (HMOX1, SOD2, MT1E, MT2A, CLU, SLC1A2), and Immediate Early Stress (FOS, JUN, EGR1, ATF3). Detailed gene modules and intermediate signaling analyses are described in Supplementary Fig. S2.\u003c/p\u003e\n\u003cp\u003eReversible zebrafish MCI model\u003c/p\u003e\n\u003cp\u003eZebrafish (Danio rerio, AB strain) [37,38] were maintained at 28.5\u0026deg;C (14:10 h light:dark). A chronic triple-stressor paradigm (LPS 5 \u0026mu;g/L, D-galactose 0.2 mg/L, 10% lard-based high-fat diet [39,40]) was applied from 2\u0026ndash;14 dpf. Survival rate was \u0026ge;95% across all groups, with 30\u0026ndash;45% cognitive impairment confirmed by the red ball avoidance task. Cognitive assessment was performed at 14 dpf using the red ball avoidance paradigm (Zefit Inc. protocol) [41,42]. Conservation of NF-\u0026kappa;B binding motifs and orthologous ptgdsb.1/2 status were verified [43,44]. Brain tissue was collected at 14 and 21 dpf for qPCR (ptgdsb.1/2, ngfr, bdnf; primer sequences in Supplementary Table S2).\u003c/p\u003e\n\u003cp\u003ePharmacological modulation and BV-2 microglial assay\u003c/p\u003e\n\u003cp\u003eBXP-101 is a standardized multi-component formulation containing honokiol [45], wedelolactone [46], and atractylodin [47] as principal active constituents. Chemical standardization is shown in Supplementary Fig. S3. BXP-101 (0.3\u0026ndash;0.6 \u0026mu;g/ml) was administered pre-inflection (6\u0026ndash;14 dpf; early) or post-inflection (15\u0026ndash;21 dpf; late). BV-2 murine microglial cells (passages 8\u0026ndash;15) were cultured in DMEM supplemented with 10% FBS at 37\u0026deg;C, 5% CO₂. Cell viability was assessed by MTT assay after 24 h exposure to BXP-101. For anti-inflammatory assays, cells were pre-treated with BXP-101 for 1 h before LPS stimulation (1 \u0026mu;g/mL, 24 h). Nitric oxide production was quantified using Griess reagent, and NF-\u0026kappa;B p65 nuclear translocation was measured by immunofluorescence (anti-p65 antibody, 1:200; secondary Alexa Fluor 488, 1:500). Images were acquired on a Zeiss LSM 880 confocal microscope and quantified using ImageJ (nuclear/cytoplasmic ratio). All experiments were performed in triplicate (Supplementary Fig. S4).\u003c/p\u003e\n\u003cp\u003eBehavioral and molecular analyses\u003c/p\u003e\n\u003cp\u003eBehavioral assessments used EthoVision XT 17 for zebrafish (locomotor and visual avoidance [41,42]) and mice (Y-maze [48], passive avoidance). General locomotor activity was assessed in a separate cohort (24-well plate, total distance moved) and is not included in the current analysis. Antibodies and primers used are listed in Supplementary Table S2.\u003c/p\u003e\n\u003cp\u003eY-maze and passive avoidance testing (murine)\u003c/p\u003e\n\u003cp\u003eThe Y-maze was constructed from black polyvinyl plastic and consisted of three arms (40 cm in length, 4 cm in width, and 12 cm in height), positioned at 120\u0026deg; angles. Distinct visual cues were placed at the end of each arm. Each mouse was placed at one arm end and allowed to freely explore for 8 minutes. Spontaneous alternation (%) = [actual alternations / (total arm entries \u0026minus; 2)] \u0026times; 100.\u003c/p\u003e\n\u003cp\u003eThe passive avoidance test was performed in a light/dark apparatus (grid floor; 0.5 mA, 3 s foot shock). On Day 1 (acquisition trial), the latency to enter the dark compartment was recorded, followed by immediate foot shock delivery upon entry. On Day 2 (retention trial), mice were placed in the light compartment and the latency to enter the dark compartment was recorded with a cutoff time of 300 s.\u003c/p\u003e\n\u003cp\u003eMammalian validation\u003c/p\u003e\n\u003cp\u003eMale ICR mice (5 weeks; Koatech, Korea) received intracerebroventricular A\u0026beta;1\u0026ndash;42 injection (20 \u0026mu;M) [49]. Groups: Sham, A\u0026beta;1\u0026ndash;42, A\u0026beta;1\u0026ndash;42 + BXP-101 (50/100/200/400 mg/kg), A\u0026beta;1\u0026ndash;42 + Donepezil (5 mg/kg) (n = 10/group; final n = 9 for A\u0026beta;1\u0026ndash;42 group after exclusion of one statistical outlier; Supplementary Table S3). Animals were housed at 23 \u0026plusmn; 2\u0026deg;C, 50 \u0026plusmn; 10% humidity, 12 h light/dark cycle with \u003cem\u003ead libitum\u003c/em\u003e access to standard diet (2018S; Envigo) and water.\u003c/p\u003e\n\u003cp\u003eNetwork pharmacology and molecular docking\u003c/p\u003e\n\u003cp\u003eTarget prediction used SwissTargetPrediction [50], PharmMapper [51], and DisGeNET [52], with supplementary databases OMIM [53] and GeneCards [54]; PPI network construction via STRING [55] in Cytoscape [56]; pathway enrichment via Metascape [57]. Molecular docking with AutoDock Vina [58] (exhaustiveness = 32); key targets: NF-\u0026kappa;B p65 (PDB: 1NFI), GSK3B (PDB: 1Q3D), PTGS2 (PDB: 5KIR). ADMET profiles were predicted using pkCSM and SwissADME.\u003c/p\u003e\n\u003cp\u003eADNI CSF proteomics\u003c/p\u003e\n\u003cp\u003eADNI CSF proteomics data included Emory TMT-MS (n = 415) and SomaScan 7K (n = 735) platforms (SeqIds: PTGDS X10514-5, LCN2 X2836-68, NEFL X10082-251, Tau X5854-60). Clinical variables (MMSE, MoCA, diagnosis, age, sex, education, APOE \u0026epsilon;4) were obtained from ADNIMERGE. CSF A\u0026beta;42, total Tau, and pTau181 were measured using Roche Elecsys assays (UPENNBIOMK dataset). Data were ComBat batch-corrected [59] and LOESS-smoothed [60]. PTGDS/LCN2 ratios were adjusted for age, sex, education, and APOE \u0026epsilon;4 [61]. ROC analysis used logistic regression with predictors including PTGDS/LCN2 ratio, NEFL, total Tau, age, APOE \u0026epsilon;4, A\u0026beta;42, pTau181, and ratio \u0026times; APOE \u0026epsilon;4 interaction; AUC computed with 95% DeLong CI. Cox regression (n = 379 MCI participants) modeled time-to-AD conversion stratified by 75th percentile risk score. Sensitivity analysis confirmed PTGDS inflection was consistent across platforms (TMT-MS and SomaScan) at MMSE ~26. Threshold sensitivity was assessed from MMSE 24 to 30 (Supplementary Table S4).\u003c/p\u003e\n\u003cp\u003eStatistical analysis\u003c/p\u003e\n\u003cp\u003eAnalyses used R (v4.3.2) and Prism 10. Mixed-effects models implemented using lme4 [62]; survival dynamics evaluated via Kaplan\u0026ndash;Meier estimation [63] and Cox proportional hazards regression [64] with Benjamini\u0026ndash;Hochberg FDR correction [65]. Data are mean \u0026plusmn; SEM; p \u0026lt; 0.05 was considered significant. For murine behavioral assays, Y-maze data were analyzed by one-way ANOVA with Newman\u0026ndash;Keuls post hoc, and passive avoidance data by two-way ANOVA with Bonferroni post hoc.\u003c/p\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec16\" class=\"Section2\"\u003e \u003ch2\u003eA conserved astrocytic inflection demarcates the neuropathological transition from compensatory to vulnerable states\u003c/h2\u003e \u003cp\u003eAnalysis of 1.3\u0026nbsp;million nuclei from the SEA-AD atlas (84 donors; Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e1\u003c/span\u003eA) identified a conserved astrocytic inflection demarcating the transition from compensatory to vulnerable states during the MCI-to-AD continuum (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e1\u003c/span\u003eA\u0026ndash;E). By projecting nuclei onto CPS, we established a stage-resolved neurodegenerative trajectory framework (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e1\u003c/span\u003eB) and identified a reproducible astrocytic trajectory centered on PTGDS dynamics. The earliest detectable event was neuronal NDUFS1 decline beginning at Bin 0.1, a core mitochondrial complex I subunit consistent with its established role as an early site of vulnerability during MCI and AD progression (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e1\u003c/span\u003eC; Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e) [\u003cspan citationid=\"CR66\" class=\"CitationRef\"\u003e66\u003c/span\u003e, \u003cspan citationid=\"CR67\" class=\"CitationRef\"\u003e67\u003c/span\u003e]. Astrocytic PTGDS expression increased in parallel and exhibited immediate inverse synchronization with neuronal NDUFS1 decline (Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e), consistent with compensatory metabolic buffering \u0026mdash; potentially through PTGDS-mediated prostaglandin signaling supporting astrocytic lactate production or anti-inflammatory buffering in response to neuronal energy deficit.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eCell-type-resolved expression trajectories in SEA-AD astrocytes and neurons across CPS bins\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"5\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMarker (Cell Type)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eBin 0.1 (Early)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eBin 0.5\u0026ndash;0.6 (Mid)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eBin 0.9 (Late)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eTrajectory Characterization\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNDUFS1 (Neuron)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.7817\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.6980\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.7061\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eOverall declining trend from Bin 0.1\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePTGDS (Astrocyte)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e1.6531\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e1.9115\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e1.6310\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eBiphasic: slope shift at Bin 0.23; peak near CPS 0.46\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLCN2 (Astrocyte)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.0000\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.0001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.0001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eSparse; induction follows PTGDS decline\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNGFR (Neuron)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.00262\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.00058\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.00073\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eEarly decline with sustained low plateau\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAPOE (Microglia)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e1.1820\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e1.1004\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e1.2186\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eU-shaped; mid-stage nadir\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"5\"\u003e \u003cem\u003eNote: Expression values represent 3-bin moving averages of log-normalized expression from the SEA-AD middle temporal gyrus atlas (1.3\u0026nbsp;million nuclei, 84 donors). Segmented regression on PTGDS identified a significant inflection at Bin 0.23 (95% CI: 0.13\u0026ndash;0.33; Davies\u0026rsquo; p\u0026thinsp;=\u0026thinsp;0.032). LCN2 detection rate in astrocytes was 0.04% (5/67,419); all LCN2-related interpretations are supported by orthogonal datasets (ADNI CSF proteomics, zebrafish qPCR, murine qPCR). Detailed bin-resolved data are provided in Supplementary Table S5.\u003c/em\u003e \u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec17\" class=\"Section2\"\u003e \u003ch2\u003eQuantitative modeling confirms a statistically bounded biphasic transition boundary\u003c/h2\u003e \u003cp\u003ePTGDS expression demonstrated a robust biphasic trajectory supporting a discrete neuropathological transition boundary (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e1\u003c/span\u003eD; Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e). At the single-cell level (n\u0026thinsp;=\u0026thinsp;67,419 astrocytes), quadratic modeling significantly improved fit relative to a linear model (β₂ = \u0026minus;2.07, p\u0026thinsp;\u0026lt;\u0026thinsp;2.2 \u0026times; 10\u003csup\u003e\u0026minus;\u0026thinsp;16\u003c/sup\u003e), confirming strong parabolic curvature with vertex at CPS 0.46. Linear regression stratified at CPS 0.46 revealed a modest but significant pre-peak rise (β = +0.244, p\u0026thinsp;=\u0026thinsp;0.000345) followed by a markedly steeper post-peak decline (β = \u0026minus;1.213, p\u0026thinsp;\u0026lt;\u0026thinsp;2.2 \u0026times; 10\u003csup\u003e\u0026minus;\u0026thinsp;16\u003c/sup\u003e), indicating asymmetric dynamics of gradual compensation and accelerated collapse. Segmented regression on unsmoothed bin-level means (n\u0026thinsp;=\u0026thinsp;9 bins) identified a significant early inflection at Bin 0.23 (95% CI: 0.13\u0026ndash;0.33; Davies\u0026rsquo; p\u0026thinsp;=\u0026thinsp;0.032), marking transition from compensatory upregulation to gradual deceleration prior to peak. Change-point estimation confirmed this inflection independently of clinical staging (Supplementary Fig. \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e).\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eStatistical validation of biphasic PTGDS dynamics at single-cell resolution (n\u0026thinsp;=\u0026thinsp;67,419 astrocytes)\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"4\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eComponent\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eEstimate\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003ep-value\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eInterpretation\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eQuadratic term\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eβ2\u0026thinsp;=\u0026thinsp;\u0026minus;\u0026thinsp;2.07\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;2.2\u0026times;10⁻\u0026sup1;⁶\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eStrong biphasic curvature\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eVertex CPS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.46\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u0026mdash;\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003ePeak inflection point\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePre-peak slope\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e+\u0026thinsp;0.244\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.000345\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eCompensatory rise\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePost-peak slope\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u0026minus;1.213\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;2.2\u0026times;10⁻\u0026sup1;⁶\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eProgressive decline\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSegmented inflection\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eBin 0.23 (95% CI: 0.13\u0026ndash;0.33)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003ep\u0026thinsp;=\u0026thinsp;0.032\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eEarly deceleration onset\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"4\"\u003e\u003cem\u003eNote: Quadratic and linear regression models fitted using CPS as a continuous variable at single-cell resolution (n\u0026thinsp;=\u0026thinsp;67,419 astrocytes). Vertex position derived from the quadratic model (\u0026minus;β1 / 2β2). Pre-peak: CPS\u0026thinsp;\u0026le;\u0026thinsp;0.46, n\u0026thinsp;=\u0026thinsp;17,874; Post-peak: CPS\u0026thinsp;\u0026gt;\u0026thinsp;0.46, n\u0026thinsp;=\u0026thinsp;49,545. Segmented regression performed on unsmoothed bin-level means (n\u0026thinsp;=\u0026thinsp;9 bins) using Davies\u0026rsquo; test.\u003c/em\u003e\u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec18\" class=\"Section2\"\u003e \u003ch2\u003ePTGDS inflection reflects astrocytic phenotypic reprogramming without evidence of cell loss\u003c/h2\u003e \u003cp\u003eAstrocyte abundance remained stable across Bins 0.4\u0026ndash;0.8 (n\u0026thinsp;=\u0026thinsp;41,141; 4,691\u0026ndash;14,108 per bin), excluding population loss as the primary explanation for PTGDS decline. Pro-apoptotic markers showed negligible correlation with CPS (CASP3 ρ = \u0026minus;0.027; BAX ρ = \u0026minus;0.010; Supplementary Table S6), and astrocyte identity markers SLC1A2 (ρ = \u0026minus;0.041) and AQP4 (ρ = \u0026minus;0.078) were preserved. Across the CPS 0.5 boundary, astrocytes exhibited increased NF-κB-associated programs and gliosis markers (Supplementary Fig. S2), accompanied by reduction of ferroptosis-protective modules (GPX4, FTH1, SLC7A11; Supplementary Table S7). LCN2 expression increased following PTGDS decline (Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e); given sparse snRNA-seq detection (0.04%), all LCN2-related interpretations rely on orthogonal validation (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e2\u003c/span\u003e; Supplementary Fig. S5). Together, these findings indicate coordinated astrocytic state modulation accompanying PTGDS inflection, independent of astrocyte population loss.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec19\" class=\"Section2\"\u003e \u003ch2\u003eA temporally ordered astrocyte-neuron-microglia cascade emerges across the PTGDS inflection window\u003c/h2\u003e \u003cp\u003eWithin the PTGDS inflection window (Bins 0.4\u0026ndash;0.8), lagged cross-correlation analysis revealed a temporally ordered cascade (Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e; Supplementary Table S7). Purinergic/Ca\u0026sup2;⁺ signaling preceded PTGDS decline by one bin (Lag\u0026thinsp;\u0026minus;\u0026thinsp;1, r\u0026thinsp;=\u0026thinsp;\u0026minus;\u0026thinsp;0.886), consistent with early compensatory sensing. NF-κB exhibited largely concurrent dynamics with PTGDS (Lag 0, r\u0026thinsp;=\u0026thinsp;\u0026minus;\u0026thinsp;0.678). PTGDS decline was inversely associated with PPARG module expression (Lag 0, r\u0026thinsp;=\u0026thinsp;\u0026minus;\u0026thinsp;0.775), indicating coordinated metabolic attenuation. LCN2 induction followed PTGDS decline (Lag\u0026thinsp;+\u0026thinsp;2, r\u0026thinsp;=\u0026thinsp;\u0026minus;\u0026thinsp;0.527), coincided with strong inverse synchronization with NDUFS1 (Lag 0, r\u0026thinsp;=\u0026thinsp;\u0026minus;\u0026thinsp;0.857), and preceded TREM2 induction by one bin (Lag\u0026thinsp;+\u0026thinsp;1, r\u0026thinsp;=\u0026thinsp;0.743). Microglial C3 showed strong inverse correlation with neuronal NGFR (Lag 0, r\u0026thinsp;=\u0026thinsp;\u0026minus;\u0026thinsp;0.940, p\u0026thinsp;=\u0026thinsp;0.0176), positioning astrocytes upstream in the cascade with microglia as secondary amplifiers.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab3\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eLagged cross-correlation analysis within the PTGDS inflection window (Bins 0.4\u0026ndash;0.8)\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"5\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eInteraction Pair\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eCell Types\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eLag (Bins)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003er-value\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eInterpretation\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePurinergic \u0026rarr; PTGDS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eAstro \u0026harr; Astro\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u0026minus;1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e\u0026minus;0.886\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eSensing precedes compensation\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNF-κB \u0026rarr; PTGDS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eAstro \u0026harr; Astro\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e\u0026minus;0.678\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eConcurrent inflammatory association\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePTGDS \u0026rarr; PPARG\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eAstro \u0026harr; Astro\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e\u0026minus;0.775\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eCoordinated metabolic attenuation\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePPARG \u0026rarr; LCN2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eAstro \u0026harr; Astro\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u0026minus;1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e\u0026minus;0.646\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eConcurrent rather than sequential modulation\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePTGDS \u0026rarr; LCN2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eAstro \u0026harr; Astro\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e+\u0026thinsp;2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e\u0026minus;0.527\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003ePTGDS decline precedes LCN2 induction\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLCN2 \u0026rarr; TREM2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eAstro \u0026harr; Micro\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e+\u0026thinsp;1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e+\u0026thinsp;0.743\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eAstrocyte signal precedes microglial activation\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eC3 \u0026rarr; NGFR\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eMicro \u0026harr; Neuron\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e\u0026minus;0.940*\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eComplement-mediated neurotoxicity\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLCN2 \u0026rarr; NDUFS1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eAstro \u0026harr; Neuron\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e\u0026minus;0.857\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eConcurrent metabolic crisis coupling\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"5\"\u003e \u003cem\u003eNote: Lagged cross-correlations computed on 3-bin moving average trajectories within the inflection window (Bins 0.4\u0026ndash;0.8). Negative lag indicates source precedes target. Due to limited effective sample size after lagging (neff\u0026thinsp;=\u0026thinsp;3\u0026ndash;5), p-values are not reported except *p\u0026thinsp;=\u0026thinsp;0.0176 for C3 \u0026rarr; NGFR. LCN2-involving pairs should be interpreted alongside orthogonal proteomic and in vivo validation data. Full trajectory analysis in Supplementary Table S7.\u003c/em\u003e \u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec20\" class=\"Section2\"\u003e \u003ch2\u003ePost-inflection destabilization reveals subtype-selective neuronal vulnerability\u003c/h2\u003e \u003cp\u003eComparison of excitatory neurons and SST⁺ inhibitory interneurons across CPS bins (Supplementary Fig. S6; Supplementary Table S8) demonstrated that while the PTGDS-LCN2-NGFR cascade was preserved across both neuronal classes, quantitative differences emerged. SST⁺ interneurons showed proportional reduction from 11.24% to 5.49% among classified neuronal subtypes, with higher NDUFS1 expression variance (0.19\u0026ndash;0.26 vs 0.12\u0026ndash;0.16 in excitatory neurons), indicating increased metabolic heterogeneity under post-inflection stress. Excitatory neurons exhibited steep early NGFR decline (0.00476 to 0.00048, Bin 0.1\u0026rarr;0.2), consistent with trophic signaling collapse, whereas SST⁺ neurons demonstrated more pronounced BCL2 reduction (6.8% decline, Bins 0.1\u0026ndash;0.4) and elevated BAX/BCL2 ratio. These findings indicate a division of vulnerability roles: excitatory neurons primarily reflect trophic signaling collapse, while SST⁺ interneurons exhibit reduced apoptotic buffering capacity under PTGDS-associated metabolic stress.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec21\" class=\"Section2\"\u003e \u003ch2\u003eCross-species conservation of biphasic PTGDS dynamics in a reversible zebrafish MCI model\u003c/h2\u003e \u003cp\u003eIn a reversible zebrafish MCI model (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e2\u003c/span\u003eA), selective cognitive impairment was confirmed by a 22.7% reduction in red ball avoidance (70.5% to 54.5%; p\u0026thinsp;=\u0026thinsp;0.020; Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e2\u003c/span\u003eB). Whole-mount double immunofluorescence (BLBP/TRITC\u0026thinsp;+\u0026thinsp;Nestin/GFP) revealed reactive glial dysfunction: BLBP fluorescence intensity was elevated 51% in MCI larvae (20.33 vs 13.43 a.u.; p\u0026thinsp;\u0026lt;\u0026thinsp;0.01; Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e2\u003c/span\u003eC\u0026ndash;D), suppressed 47% following BXP-101 treatment (10.74 a.u.; p\u0026thinsp;\u0026lt;\u0026thinsp;0.001 vs MCI). Nestin signal intensity declined across MCI and treatment groups (MCI: 9.27 a.u.; BXP-101 0.4 \u0026micro;g/ml: 8.50 a.u.; Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e2\u003c/span\u003eE), consistent with reduced neural progenitor-associated activity under sustained inflammatory load. Longitudinal qPCR of ptgdsb.1 and ptgdsb.2 at 14 and 21 dpf revealed progressive compensatory induction, with MCI larvae exceeding control expression by 21 dpf (ptgdsb.1: 1.063\u0026thinsp;\u0026plusmn;\u0026thinsp;0.08; ptgdsb.2: 1.066\u0026thinsp;\u0026plusmn;\u0026thinsp;0.14; Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e2\u003c/span\u003eF) [\u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e43\u003c/span\u003e, \u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e44\u003c/span\u003e, \u003cspan citationid=\"CR68\" class=\"CitationRef\"\u003e68\u003c/span\u003e], mirroring the SEA-AD compensatory PTGDS upregulation phase (CPS 0.1\u0026ndash;0.46). An inverse relationship between ptgdsb expression and ngfr levels was observed across conditions and timepoints [\u003cspan citationid=\"CR68\" class=\"CitationRef\"\u003e68\u003c/span\u003e], consistent with LCN2-mediated suppression of neurogenic signaling downstream of PTGDS exhaustion. Zebrafish ptgdsb.1/2 show 85.6% amino acid identity and high structural conservation with human PTGDS (RMSD\u0026thinsp;\u0026lt;\u0026thinsp;1.3 \u0026Aring;), with conserved NF-κB promoter motifs (Supplementary Fig. S5).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec22\" class=\"Section2\"\u003e \u003ch2\u003eTiming-dependent pharmacological rescue validates the therapeutic window defined by the transition boundary\u003c/h2\u003e \u003cp\u003eEarly intervention during the pre-peak phase (6\u0026ndash;14 dpf) significantly rescued cognitive deficits (BXP-101 0.6 \u0026micro;g/ml: 67.6% vs MCI 54.5%; p\u0026thinsp;=\u0026thinsp;0.016; Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e2\u003c/span\u003eB), whereas late intervention (15\u0026ndash;21 dpf) yielded substantially attenuated rescue \u0026mdash; consistent with the asymmetric trajectory dynamics at CPS 0.46. BXP-101 reduced upstream inflammatory demand on PTGDS, evidenced by dose-dependent suppression of TNF-α (MCI: 39.7 pg/ml; BXP0.6: 21.8 pg/ml, \u0026minus;\u0026thinsp;45% vs MCI) and IL-6 (MCI: 91.8 pg/ml; BXP0.6: 23.4 pg/ml, \u0026minus;\u0026thinsp;75% vs MCI) at the protein level (ELISA; Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e2\u003c/span\u003eG). This anti-inflammatory effect was more pronounced at the protein than the mRNA level, suggesting post-transcriptional regulatory mechanisms consistent with the known modes of action of honokiol and wedelolactone as NF-κB modulators. BXP-101 treatment additionally suppressed reactive gliosis (BLBP: BXP0.4 10.74 vs MCI 20.33 a.u., \u0026minus;\u0026thinsp;47%; p\u0026thinsp;\u0026lt;\u0026thinsp;0.001; Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e2\u003c/span\u003eC\u0026ndash;D), and preserved ptgdsb.1/2 induction capacity at 21 dpf (ptgdsb.1: 1.080 vs MCI: 1.063; Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e2\u003c/span\u003eF), consistent with a demand-conservation model in which NF-κB suppression reduces the metabolic burden on astrocytic PTGDS buffering. BDNF expression was elevated (1.71-fold vs control) and synaptic integrity partially restored (PSD-95/dlg4: 1.27-fold vs MCI; Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e2\u003c/span\u003eF) [\u003cspan citationid=\"CR69\" class=\"CitationRef\"\u003e69\u003c/span\u003e, \u003cspan citationid=\"CR70\" class=\"CitationRef\"\u003e70\u003c/span\u003e]. Network pharmacology analysis identified 12 hub targets enriched in neurogenesis and NF-κB pathways, with molecular docking confirming high-affinity engagement of GSK3B (binding affinity\u0026thinsp;\u0026minus;\u0026thinsp;8.3 kcal/mol; Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e3\u003c/span\u003eA\u0026ndash;D). Behavioral assessments used the red ball avoidance paradigm (n\u0026thinsp;=\u0026thinsp;30/group) [\u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e41\u003c/span\u003e, \u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e42\u003c/span\u003e] and Y-maze [\u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e48\u003c/span\u003e]. Murine validation (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e4\u003c/span\u003eA\u0026ndash;F; Supplementary Fig. S4) confirmed concordant dose-dependent effects on LCN2, NGFR, and inflammatory markers. Collectively, these data demonstrate timing-dependent preservation of PTGDS-centered trajectory dynamics following NF-κB modulation, providing functional validation of the CPS 0.46 phase boundary.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cdiv id=\"Sec23\" class=\"Section3\"\u003e \u003ch2\u003eCSF PTGDS/LCN2 dynamics define a clinically translatable transition boundary in ADNI\u003c/h2\u003e \u003cp\u003eCSF PTGDS is a constitutively secreted lipocalin and directly reflects astrocytic PTGDS synthetic activity [\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e]. Analysis of 735 ADNI participants (Supplementary Table S9) revealed biphasic CSF PTGDS dynamics: elevation in EMCI (+\u0026thinsp;23%, p\u0026thinsp;=\u0026thinsp;0.008), a peak in LMCI (+\u0026thinsp;41%, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001), and subsequent decline in AD (\u0026minus;\u0026thinsp;16% vs LMCI, p\u0026thinsp;=\u0026thinsp;0.02; Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e5\u003c/span\u003eA). Cross-platform validation confirmed this biphasic pattern across both TMT-MS (n\u0026thinsp;=\u0026thinsp;415) and SomaScan 7K (n\u0026thinsp;=\u0026thinsp;735) platforms (Supplementary Fig. S7). Segmented regression identified a PTGDS\u0026ndash;LCN2 crossover at MMSE 26.0 (95% CI: 24.8\u0026ndash;27.2; [\u003cspan citationid=\"CR71\" class=\"CitationRef\"\u003e71\u003c/span\u003e]), followed by NEFL inflection at MMSE 24.7 (95% CI: 23.1\u0026ndash;26.3; p\u0026thinsp;\u0026lt;\u0026thinsp;0.001; Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e5\u003c/span\u003eB\u0026ndash;C), defining a narrow 1.3-point MMSE interval that demarcates the clinically accessible transition window. Pseudo-time analysis independently recapitulated biphasic PTGDS dynamics with abrupt LCN2 induction at Bin 0.5\u0026ndash;0.6 (Supplementary Fig. S2), consistent with a mid-stage astrocytic molecular inflection and corroborating the SEA-AD-derived phase boundary. The PTGDS/LCN2 ratio correlated with MMSE (r\u0026thinsp;=\u0026thinsp;0.34, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001) and NEFL (β = \u0026minus;0.37, p\u0026thinsp;\u0026lt;\u0026thinsp;10\u003csup\u003e\u0026minus;\u0026thinsp;28\u003c/sup\u003e). ROC analysis demonstrated AUC\u0026thinsp;=\u0026thinsp;0.743 (95% CI: 0.698\u0026ndash;0.788) for MCI vs AD discrimination (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e5\u003c/span\u003eE), and the index independently predicted conversion risk (HR 3.2, 95% CI: 2.3\u0026ndash;4.4; Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e5\u003c/span\u003eF; Supplementary Table S4). MoCA-based segmented regression showed a concordant inflection at MoCA 21.3 (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e5\u003c/span\u003eD), confirming cross-instrument robustness of the clinical boundary. Threshold sensitivity analysis across MMSE inclusion criteria (24\u0026ndash;30) demonstrated consistent AUC performance (range: 0.56\u0026ndash;0.72), confirming robustness across clinical severity strata (Supplementary Table S4). Collectively, these multi-scale observations converge on a conserved astrocytic PTGDS inflection associated with progression from compensatory buffering toward accelerated neurodegeneration, with direct alignment to clinically measurable cognitive decline.\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e"},{"header":"Discussion","content":"\u003cdiv id=\"Sec25\" class=\"Section2\"\u003e \u003ch2\u003eA directionally ordered astrocyte-neuron cascade structured around a neuropathological transition boundary\u003c/h2\u003e \u003cp\u003eBy integrating single-cell pseudo-progression with cross-species validation, we characterize a temporally ordered astrocyte\u0026ndash;neuron cascade that is structured around a statistically bounded metabolic inflection during the MCI-to-AD transition. As used herein, \u0026lsquo;transition boundary\u0026rsquo; denotes a statistically defined inflection interval characterized by asymmetric trajectory dynamics and cross-scale conservation; it does not imply a strict thermodynamic discontinuity. Trajectory modeling revealed marked asymmetry: compensatory dynamics accumulated gradually, whereas post-inflection destabilization accelerated disproportionately, consistent with progressive compensation followed by accelerated vulnerability once metabolic buffering capacity is exceeded. This framework is supported by: (i) pseudo-progression\u0026ndash;based directional relationships (Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e, Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e), (ii) pathway-level dissection of regulatory modules (Supplementary Fig. S2, Fig. S4), and (iii) functional validation across zebrafish and murine models (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e2\u003c/span\u003e, Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e4\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThree convergent observations support a directionally consistent PTGDS\u0026ndash;LCN2\u0026ndash;NGFR coupling framework. The limited statistical power of lagged CCF analysis (neff\u0026thinsp;=\u0026thinsp;3\u0026ndash;5) precludes definitive causal inference from correlation coefficients alone; however, the directional consistency of cascade ordering \u0026mdash; combined with pharmacological rescue and cross-species validation \u0026mdash; supports the proposed temporal hierarchy as a biologically coherent framework. First, purinergic sensing precedes PTGDS decline (Lag\u0026thinsp;\u0026minus;\u0026thinsp;1, r\u0026thinsp;=\u0026thinsp;\u0026minus;\u0026thinsp;0.886), consistent with early compensatory mechanisms. Second, PTGDS decline is followed by LCN2 induction (Lag\u0026thinsp;+\u0026thinsp;2, r\u0026thinsp;=\u0026thinsp;\u0026minus;\u0026thinsp;0.527), with LCN2 showing strong inverse coupling with NDUFS1 (Lag 0, r\u0026thinsp;=\u0026thinsp;\u0026minus;\u0026thinsp;0.857) and preceding TREM2 activation (Lag\u0026thinsp;+\u0026thinsp;1, r\u0026thinsp;=\u0026thinsp;0.743). Third, microglial C3 inversely correlates with neuronal NGFR (Lag 0, r\u0026thinsp;=\u0026thinsp;\u0026minus;\u0026thinsp;0.940, p\u0026thinsp;=\u0026thinsp;0.0176), positioning astrocytes upstream in the observed cascade, with microglia acting as secondary amplifiers.\u003c/p\u003e \u003cp\u003ePPARG module changes were inversely rather than sequentially related to PTGDS and LCN2 trajectories, suggesting coordinated metabolic attenuation rather than a strict upstream regulatory role. While lagged correlations showed directional consistency in the core cascade, statistical significance was limited in several pairs due to small effective sample sizes after lagging (neff\u0026thinsp;=\u0026thinsp;3\u0026ndash;5). Early neuronal mitochondrial vulnerability, reflected by NDUFS1 decline (Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e), coincides with compensatory astrocytic PTGDS upregulation. Following the inflection at CPS 0.46, PTGDS decline is associated with coordinated LCN2 elevation and suppression of neuronal NGFR (Supplementary Table S7). Microglial activation follows this astrocytic shift (Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e), positioning astrocytes upstream in the observed cascade, with microglia acting as secondary amplifiers [\u003cspan citationid=\"CR72\" class=\"CitationRef\"\u003e72\u003c/span\u003e]. Subtype analysis further reveals hierarchical neuronal vulnerability (Supplementary Fig. S6): SST⁺ interneurons show proportional reduction (11.24% to 5.49%; Supplementary Table S8) and increased metabolic variance, consistent with early excitation\u0026ndash;inhibition imbalance [\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e] preceding circuit destabilization.\u003c/p\u003e \u003cdiv id=\"Sec26\" class=\"Section3\"\u003e \u003ch2\u003eLipid-metabolic exhaustion as a phase-triggering event in neuropathological progression\u003c/h2\u003e \u003cp\u003eNeuronal NDUFS1 decline reflects early mitochondrial stress [\u003cspan citationid=\"CR66\" class=\"CitationRef\"\u003e66\u003c/span\u003e, \u003cspan citationid=\"CR67\" class=\"CitationRef\"\u003e67\u003c/span\u003e]. Within astrocyte\u0026ndash;neuron metabolic coupling [\u003cspan citationid=\"CR73\" class=\"CitationRef\"\u003e73\u003c/span\u003e], PTGDS functions as a homeostatic effector [\u003cspan citationid=\"CR69\" class=\"CitationRef\"\u003e69\u003c/span\u003e]: its gradual compensatory induction represents the cell\u0026rsquo;s attempt to maintain lipid-metabolic homeostasis under increasing energetic demand. This is not a passive biomarker response but an active buffering process whose exhaustion constitutes the phase-triggering event. By analogy to tipping point dynamics in complex systems [\u003cspan citationid=\"CR74\" class=\"CitationRef\"\u003e74\u003c/span\u003e], the PTGDS inflection represents a bifurcation: prior to CPS 0.46, the system retains resilience and can return to baseline; beyond it, positive feedback through LCN2-mediated inflammatory amplification drives the system toward an attractor state of accelerated vulnerability from which recovery becomes progressively less feasible. This architecture \u0026mdash; gradual compensatory loading followed by threshold collapse \u0026mdash; is structurally analogous to metabolic phase transitions described in oncology and cardiac remodeling [\u003cspan citationid=\"CR75\" class=\"CitationRef\"\u003e75\u003c/span\u003e], suggesting a conserved principle of cellular stress tolerance across disease contexts.\u003c/p\u003e \u003cp\u003eImportantly, this inflection does not reflect astrocyte loss. Astrocyte abundance remains stable across bins 0.4\u0026ndash;0.8 (n\u0026thinsp;=\u0026thinsp;41,141), and correlations with apoptotic signatures are negligible (Supplementary Table S6). Instead, reactive markers (NFKBIA, GFAP) increase significantly across the CPS 0.46 boundary (Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e), indicating phenotypic reprogramming rather than population collapse.\u003c/p\u003e \u003cp\u003eModule-level analysis (Supplementary Fig. S2) identifies layered regulatory architecture. Purinergic and calcium-associated genes (P2RY1, GJA1) precede PTGDS induction (Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e), while buffering modules (HMOX1, CLU) shift from acute to sustained compensation. NF-κB activation was concurrent with PTGDS dynamics (Lag 0, r\u0026thinsp;=\u0026thinsp;\u0026minus;\u0026thinsp;0.678; Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e), suggesting that inflammatory priming and metabolic attenuation co-occur rather than following a strict sequential order. Instead, PTGDS-associated metabolic tone \u0026mdash; potentially mediated through 15d-PGJ₂\u0026ndash;PPAR-γ signaling [\u003cspan citationid=\"CR76\" class=\"CitationRef\"\u003e76\u003c/span\u003e, \u003cspan citationid=\"CR77\" class=\"CitationRef\"\u003e77\u003c/span\u003e] \u0026mdash; may transiently restrain downstream effector induction. Transition beyond CPS 0.46 coincides with LCN2 amplification and coordinated ferroptosis-related vulnerability [\u003cspan citationid=\"CR78\" class=\"CitationRef\"\u003e78\u003c/span\u003e, \u003cspan citationid=\"CR79\" class=\"CitationRef\"\u003e79\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eQuadratic modeling confirms significant curvature (β\u0026sup2; = \u0026minus;2.07, p\u0026thinsp;\u0026lt;\u0026thinsp;2.2e\u0026thinsp;\u0026minus;\u0026thinsp;16) with a vertex at CPS 0.46 (Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e), defining a statistically bounded inflection interval. Critically, the asymmetry between pre-peak slope (+\u0026thinsp;0.244) and post-peak slope (\u0026minus;\u0026thinsp;1.213) quantifies the irreversibility gradient: compensatory capacity accumulates slowly and collapses rapidly, consistent with a tipping point architecture rather than symmetric oscillation. Together, these data support a structured phase boundary model in which the therapeutic window is defined not by symptom severity but by position relative to the metabolic inflection point. Intervention before CPS 0.46 \u0026mdash; corresponding clinically to MMSE\u0026thinsp;~\u0026thinsp;26 \u0026mdash; targets the compensatory phase, when metabolic buffering capacity is maximal and inflammatory amplification has not yet been established. Intervention after this boundary faces a qualitatively different and less tractable target landscape.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec27\" class=\"Section3\"\u003e \u003ch2\u003eLCN2 as a boundary-associated amplifier and clinical translation\u003c/h2\u003e \u003cp\u003eLCN2 transcripts are detected in only 0.04% of astrocytes under baseline conditions, yet surge robustly at the phase boundary \u0026mdash; a pattern consistent with a threshold-gated switch rather than a graded response. This binary-like induction profile, analogous to \u0026lsquo;all-or-nothing\u0026rsquo; inflammatory priming, reinforces the existence of a discrete phase transition rather than continuous disease progression.\u003c/p\u003e \u003cp\u003eDespite sparse LCN2 transcript detection in snRNA-seq (0.04%), consistent late-stage induction was supported across orthogonal platforms including CSF proteomics (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e5\u003c/span\u003e), zebrafish qPCR (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e2\u003c/span\u003eF), and murine models (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e4\u003c/span\u003eC). Integration of longitudinal ADNI CSF proteomics (n\u0026thinsp;=\u0026thinsp;735) establishes quantitative clinical alignment: the PTGDS/LCN2 crossover coincides with MMSE 26.0 and precedes NEFL inflection at 24.7 (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e5\u003c/span\u003eB\u0026ndash;C), mirroring the CPS 0.46 metabolic boundary identified in SEA-AD. This narrow 1.3-point MMSE interval predicts conversion risk (HR 3.2; Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e5\u003c/span\u003eF), demonstrating cross-scale concordance between a cellular phase boundary and clinically measurable decline. Pharmacological suppression of LCN2 restores NGFR expression and cognitive performance only prior to PTGDS exhaustion (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e2\u003c/span\u003eH; Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e4\u003c/span\u003e), reinforcing the concept that therapeutic responsiveness is constrained by this metabolically defined boundary. While pseudo-time inference cannot establish definitive causality [\u003cspan citationid=\"CR70\" class=\"CitationRef\"\u003e70\u003c/span\u003e, \u003cspan citationid=\"CR74\" class=\"CitationRef\"\u003e74\u003c/span\u003e], convergence across 1.3\u0026nbsp;million nuclei, orthogonal perturbation models, and longitudinal clinical proteomics supports the biological and clinical significance of this temporally bounded transition. Threshold sensitivity analysis across MMSE inclusion criteria (24\u0026ndash;30) demonstrated consistent AUC improvement with broader inclusion, confirming robustness of the phase boundary index across clinical severity strata (Supplementary Table S4).\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec28\" class=\"Section2\"\u003e \u003ch2\u003eMolecular redundancy, metabolic-trophic coupling, and oligodendrocyte vulnerability\u003c/h2\u003e \u003cp\u003eLCN2 induction occurs within a broader remodeling of lipid-associated pathways. APOE and CLU display distinct but complementary trajectories (Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e), indicating partial redundancy within lipocalin-associated buffering systems. As PTGDS declines, LCN2 induction aligns with iron-linked inflammatory signaling, suggesting a qualitative shift in astrocytic metabolic state.\u003c/p\u003e \u003cp\u003eLagged cross-correlation analysis (Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e) demonstrates that PTGDS decline is followed by PPARG module attenuation (Lag 0, r\u0026thinsp;=\u0026thinsp;\u0026minus;\u0026thinsp;0.775), consistent with weakening anti-inflammatory restraint. LCN2 induction coincides with suppression of neuronal NGFR and NDUFS1 (Lag 0, r\u0026thinsp;=\u0026thinsp;\u0026minus;\u0026thinsp;0.857; Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e), indicating tight metabolic\u0026ndash;trophic coupling between astrocytes and neurons. This strong temporal synchrony (zero-lag, high negative correlation) suggests that astrocyte-derived LCN2 acts as a secreted mediator that propagates metabolic stress to neurons. LCN2 is known to induce iron dysregulation and ferroptosis sensitivity [\u003cspan citationid=\"CR78\" class=\"CitationRef\"\u003e78\u003c/span\u003e, \u003cspan citationid=\"CR79\" class=\"CitationRef\"\u003e79\u003c/span\u003e], which is consistent with mitochondrial complex I vulnerability reflected by NDUFS1 decline and impairs neurotrophic signaling pathways (NGFR downregulation). Consequently, PTGDS exhaustion and subsequent LCN2 emergence in astrocytes constrain neuronal bioenergetic capacity and trophic support, reflecting a coordinated astrocyte-to-neuron metabolic\u0026ndash;trophic failure during disease progression. Together, these data support a temporally ordered and directionally consistent coupling model in which astrocytic metabolic reprogramming precedes and predicts neuronal vulnerability.\u003c/p\u003e \u003cp\u003eBeyond the astrocyte\u0026ndash;neuron axis, preliminary analysis of SEA-AD oligodendrocyte trajectories revealed parallel metabolic vulnerability. Oligodendrocytic MCT4 (SLC16A3) declined by 42.2% (rho\u0026thinsp;=\u0026thinsp;\u0026minus;\u0026thinsp;0.905, p\u0026thinsp;=\u0026thinsp;0.002), accompanied by significant reduction of myelin-associated genes MOG (\u0026minus;\u0026thinsp;14.4%, rho\u0026thinsp;=\u0026thinsp;\u0026minus;\u0026thinsp;0.810, p\u0026thinsp;=\u0026thinsp;0.015) and MAG (\u0026minus;\u0026thinsp;10.1%, rho\u0026thinsp;=\u0026thinsp;\u0026minus;\u0026thinsp;0.762, p\u0026thinsp;=\u0026thinsp;0.028; data not shown). Notably, MAPT expression increased by 22.2% in oligodendrocytes (rho\u0026thinsp;=\u0026thinsp;+\u0026thinsp;0.905, p\u0026thinsp;=\u0026thinsp;0.002), inversely correlated with MCT4 (rho\u0026thinsp;=\u0026thinsp;\u0026minus;\u0026thinsp;0.833, p\u0026thinsp;=\u0026thinsp;0.010), suggesting that energy metabolic compromise extends beyond neurons to myelinating glia. Since oligodendrocytes supply lactate to axons via MCT1, their concurrent metabolic failure implies that neurons may face energy deprivation at both the soma (via astrocytic ANLS) and along the axon (via oligodendrocytic myelin), constituting a dual supply-line disruption that may accelerate neuronal vulnerability. However, whether astrocytic reactive transformation directly contributes to oligodendrocyte pathology \u0026mdash; for example, through A1-associated toxic lipid release \u0026mdash; or whether these represent parallel but independent responses to a shared upstream stressor remains to be determined.\u003c/p\u003e \u003c/div\u003e"},{"header":"Conclusions","content":"\u003cp\u003eOur findings identify astrocytic PTGDS inflection as a conserved neuropathological boundary that separates reversible prodromal compensation from irreversible neurodegenerative progression in Alzheimer\u0026rsquo;s disease. The convergence of this boundary across 1.3\u0026nbsp;million single nuclei, orthogonal perturbation models, and longitudinal clinical proteomics \u0026mdash; with cross-scale alignment at MMSE 26.0 \u0026mdash; establishes a quantitative framework for stage-stratified intervention. The therapeutic window defined by this boundary is determined not by symptom severity alone but by molecular position relative to the PTGDS inflection point. These findings reframe prodromal AD therapeutics from symptom-driven timing to mechanism-anchored staging, with direct implications for clinical trial design and patient stratification.\u003c/p\u003e\n\u003ch3\u003eLimitations\u003c/h3\u003e\n\u003cp\u003eSeveral limitations warrant consideration. First, SEA-AD provides pseudo-time trajectories rather than true longitudinal sampling, limiting causal inference. Although lagged cross-correlation and timing-dependent rescue experiments support directional interpretation, definitive causality requires targeted perturbation in human-relevant systems.\u003c/p\u003e \u003cp\u003eSecond, LCN2 transcript detection in droplet-based snRNA-seq was sparse (0.04%; 5 astrocytes among 67,419), a known limitation for low-abundance or secreted transcripts. Conclusions regarding LCN2 trajectory rely on orthogonal validation platforms, including CSF proteomics and cross-species qPCR, which consistently supported pathway engagement.\u003c/p\u003e \u003cp\u003eThird, the SEA-AD dataset is restricted to the middle temporal gyrus (MTG); thus, our findings may not fully capture the regional heterogeneity of AD progression. Future studies integrating spatial transcriptomics or multi-region atlases will be essential to validate these metabolic phase boundaries across differentially affected brain areas.\u003c/p\u003e \u003cp\u003eFinally, pharmacologic stabilization using BXP-101 demonstrated timing-dependent efficacy consistent with the phase boundary model, with pre-inflection intervention yielding substantially greater rescue than post-inflection treatment. However, BXP-101 broadly modulates NF-κB signaling, and the observed anti-inflammatory effects \u0026mdash; TNF-α suppression (\u0026minus;\u0026thinsp;45%), IL-6 suppression (\u0026minus;\u0026thinsp;75% at protein level) \u0026mdash; should be interpreted as downstream consequences of upstream metabolic stabilization rather than primary mechanism. NF-κB activation in this context is better understood as amplification noise enabled by PTGDS exhaustion, not the causal driver. More specific PTGDS-directed approaches \u0026mdash; such as PGD₂ analogs or conditional PTGDS restoration \u0026mdash; will be required to establish direct metabolic causality and to determine whether re-crossing the phase boundary is achievable.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eEthics approval and consent to participate:\u0026nbsp;\u003c/strong\u003eAll animal experiments were conducted under approved IACUC protocols (Kangwon National University: KW-241104-1; Zefit Inc.: ZEFIT-IACUC-26010601-0001). Human data analysis utilized de-identified secondary data from public repositories; no additional IRB approval was required.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAvailability of data and materials:\u0026nbsp;\u003c/strong\u003eSEA-AD data are available at the Allen Brain Cell Atlas (https://portal.brain-map.org). ADNI CSF proteomics data are available through https://adni.loni.usc.edu upon completion of standard data use agreements. Analysis code is available at https://github.com/YoungOukKim/MCI-to-AD.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting interests:\u0026nbsp;\u003c/strong\u003eW.M. Heo is Chief Executive Officer of BioXP, Inc., and BXP-101 is a product under development by BioXP, Inc. All other authors declare no competing interests.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding:\u0026nbsp;\u003c/strong\u003eThis research was supported by internal research funds from BioXP, Inc. Funders had no role in study design, data collection, analysis, or publication decisions.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthors\u0026rsquo; contributions:\u0026nbsp;\u003c/strong\u003eY.O.K., W.M.H., and S.J.P. contributed equally as co-first authors. Y.O.K. and W.M.H. conceived the study, designed all experiments, and supervised all analyses. S.J.P. led murine validation. Zebrafish experiments were conducted by Zefit Inc. Y.C.K., Y.E.C., Y.W.L., and J.Y.K. contributed to experimental work and manuscript preparation. All authors read and approved the final manuscript.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgements:\u0026nbsp;\u003c/strong\u003eThe authors thank the SEA-AD consortium and ADNI investigators for publicly available datasets. The authors thank SeungHwan Kang and the Zefit team for expert experimental support.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n \u003cli\u003e\u003cem\u003eLopez OL (2013) Mild cognitive impairment. Continuum 19:411\u0026ndash;424. https://doi.org/10.1212/01.CON.0000429175.29601.97\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eLivingston G, Huntley J, Sommerlad A et al (2020) Dementia prevention, intervention, and care: 2020 report of the Lancet Commission. Lancet 396:413\u0026ndash;446. https://doi.org/10.1016/S0140-6736(20)30367-6\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cem\u003ePetersen RC, Aisen PS, Beckett LA et al (2010) Alzheimer\u0026apos;s Disease Neuroimaging Initiative (ADNI): clinical characterization. Neurology 74:201\u0026ndash;209. https://doi.org/10.1212/WNL.0b013e3181cb3e25\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eLong JM, Holtzman DM (2019) Alzheimer disease: an update on pathobiology and treatment strategies. Cell 179:312\u0026ndash;339. https://doi.org/10.1016/j.cell.2019.09.001\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eSelkoe DJ, Hardy J (2016) The amyloid hypothesis of Alzheimer\u0026apos;s disease at 25 years. EMBO Mol Med 8:595\u0026ndash;608. https://doi.org/10.15252/emmm.201606210\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eSims JR, Zimmer JA, Evans CD et al (2023) Donanemab in early symptomatic Alzheimer disease. JAMA 330:512\u0026ndash;527. https://doi.org/10.1001/jama.2023.13239\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cem\u003evan Dyck CH, Swanson CJ, Aisen P et al (2023) Lecanemab in early Alzheimer\u0026apos;s disease. N Engl J Med 388:9\u0026ndash;21. https://doi.org/10.1056/NEJMoa2212948\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eBradburn S, Murgatroyd C, Ray N (2019) Neuroinflammation in mild cognitive impairment and Alzheimer\u0026apos;s disease: a meta-analysis. Ageing Res Rev 50:1\u0026ndash;8. https://doi.org/10.1016/j.arr.2019.01.002\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eCalsolaro V, Edison P (2016) Neuroinflammation in Alzheimer\u0026apos;s disease: current evidence and future directions. Alzheimers Dement 12:719\u0026ndash;732. https://doi.org/10.1016/j.jalz.2016.02.010\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eColonna M, Butovsky O (2017) Microglia function in the central nervous system during health and neurodegeneration. Annu Rev Immunol 35:441\u0026ndash;468. https://doi.org/10.1146/annurev-immunol-051116-052358\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eMoreno-Jim\u0026eacute;nez EP, Flor-Garc\u0026iacute;a M, Terreros-Roncal J et al (2019) Adult hippocampal neurogenesis is abundant in neurologically healthy subjects and drops sharply in patients with Alzheimer\u0026apos;s disease. Nat Med 25:554\u0026ndash;560. https://doi.org/10.1038/s41591-019-0375-9\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eParkitny L, Maletic-Savatic M (2021) Glial PAMPering and DAMPening of adult hippocampal neurogenesis. Brain Sci 11:1299. https://doi.org/10.3390/brainsci11101299\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eJack CR Jr, Bennett DA, Blennow K et al (2018) NIA-AA Research Framework: toward a biological definition of Alzheimer\u0026apos;s disease. Alzheimers Dement 14:535\u0026ndash;562. https://doi.org/10.1016/j.jalz.2018.02.018\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eScheltens P, De Strooper B, Kivipelto M et al (2021) Alzheimer\u0026apos;s disease. Lancet 397:1577\u0026ndash;1590. https://doi.org/10.1016/S0140-6736(20)32205-4\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eBlennow K, Zetterberg H (2018) Biomarkers for Alzheimer\u0026apos;s disease: current status and prospects for the future. J Intern Med 284:643\u0026ndash;663. https://doi.org/10.1111/joim.12816\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eGabitto MI, Travaglini KJ, Rachleff VM et al (2024) Integrated multimodal cell atlas of Alzheimer\u0026apos;s disease. Nat Neurosci 27:2366\u0026ndash;2383. https://doi.org/10.1038/s41593-024-01774-5\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eLi X, Wang X, Guo L et al (2023) Association between lipocalin-2 and mild cognitive impairment or dementia: a systematic review and meta-analysis. Ageing Res Rev 89:101984. https://doi.org/10.1016/j.arr.2023.101984\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eNaud\u0026eacute; PJW, Nyakas C, Eiden LE et al (2012) Lipocalin 2: novel component of proinflammatory signaling in Alzheimer\u0026apos;s disease. FASEB J 26:2811\u0026ndash;2823. https://doi.org/10.1096/fj.11-202457\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eJha MK, Lee S, Park DH et al (2015) Diverse functional roles of lipocalin-2 in the central nervous system. Neurosci Biobehav Rev 49:135\u0026ndash;156. https://doi.org/10.1016/j.neubiorev.2014.12.006\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eFerreira AC, Pinto V, Mesquita SD et al (2013) Lipocalin-2 is involved in emotional behaviors and cognitive function. Front Cell Neurosci 7:122. https://doi.org/10.3389/fncel.2013.00122\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eDekens DW, Naud\u0026eacute; PJW, Keijser JN et al (2018) Lipocalin 2 contributes to brain iron dysregulation but does not affect cognition, plaque load, and glial activation in the J20 Alzheimer mouse model. J Neuroinflammation 15:330. https://doi.org/10.1186/s12974-018-1372-5\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eSiddiqui T, Cosacak MI, Popova S et al (2023) Nerve growth factor receptor (Ngfr) induces neurogenic plasticity by suppressing reactive astroglial Lcn2/Slc22a17 signaling in Alzheimer\u0026apos;s disease. npj Regen Med 8:33. https://doi.org/10.1038/s41536-023-00311-5\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eDevireddy LR, Gazin C, Zhu X, Green MR (2005) A cell-surface receptor for lipocalin 24p3 selectively mediates apoptosis and iron uptake. Cell 123:1293\u0026ndash;1305. https://doi.org/10.1016/j.cell.2005.10.027\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eHayden MS, Ghosh S (2012) NF-\u0026kappa;B, the first quarter-century: remarkable progress and outstanding questions. Genes Dev 26:203\u0026ndash;234. https://doi.org/10.1101/gad.183434.111\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eLiu T, Zhang L, Joo D, Sun S-C (2017) NF-\u0026kappa;B signaling in inflammation. Signal Transduct Target Ther 2:17023. https://doi.org/10.1038/sigtrans.2017.23\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eMohri I, Taniike M, Taniguchi H et al (2006) Prostaglandin D2-mediated microglia/astrocyte interaction enhances astrogliosis and demyelination in twitcher. J Neurosci 26:4383\u0026ndash;4393. https://doi.org/10.1523/JNEUROSCI.4531-05.2006\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eMohri I, Kadoyama K, Kanekiyo T et al (2007) Hematopoietic prostaglandin D synthase and DP1 receptor are selectively upregulated in microglia and astrocytes within senile plaques from human patients and in a mouse model of Alzheimer disease. J Neuropathol Exp Neurol 66:469\u0026ndash;480. https://doi.org/10.1097/01.jnen.0000240472.43038.27\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eChoi D-J, An J, Jou I et al (2019) A Parkinson\u0026apos;s disease gene, DJ-1, regulates anti-inflammatory roles of astrocytes through prostaglandin D2 synthase expression. Neurobiol Dis 127:482\u0026ndash;491. https://doi.org/10.1016/j.nbd.2019.04.003\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eKanekiyo T, Ban T, Aritake K et al (2007) Lipocalin-type prostaglandin D synthase/beta-trace is a major amyloid beta-chaperone in human cerebrospinal fluid. Proc Natl Acad Sci USA 104:6412\u0026ndash;6417. https://doi.org/10.1073/pnas.0701585104\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eHigginbotham L, Ping L, Dammer EB et al (2020) Integrated proteomics reveals brain-based cerebrospinal fluid biomarkers in asymptomatic and symptomatic Alzheimer\u0026apos;s disease. Sci Adv 6:eaaz9360. https://doi.org/10.1126/sciadv.aaz9360\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eJohnson ECB, Dammer EB, Duong DM et al (2020) Large-scale proteomic analysis of Alzheimer\u0026apos;s disease brain and cerebrospinal fluid reveals early changes in energy metabolism associated with microglia and astrocyte activation. Nat Med 26:769\u0026ndash;780. https://doi.org/10.1038/s41591-020-0815-6\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eUrade Y (2021) Biochemical and structural characteristics, gene regulation, physiological, pathological and clinical features of lipocalin-type prostaglandin D2 synthase as a multifunctional lipocalin. Front Physiol 12:718002. https://doi.org/10.3389/fphys.2021.718002\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eWolock SL, Lopez R, Klein AM (2019) Scrublet: computational identification of cell doublets in single-cell transcriptomic data. Cell Syst 8:281\u0026ndash;291. https://doi.org/10.1016/j.cels.2018.11.005\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eLu K-P, Chang S-T (2023) An advanced segmentation approach to piecewise regression models. Mathematics 11:4959. https://doi.org/10.3390/math11244959\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eJin S, Guerrero-Juarez CF, Zhang L et al (2021) Inference and analysis of cell-cell communication using CellChat. Nat Commun 12:1088. https://doi.org/10.1038/s41467-021-21246-9\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eAibar S, Bravo Gonz\u0026aacute;lez-Blas C, Moerman T et al (2017) SCENIC: single-cell regulatory network inference and clustering. Nat Methods 14:1083\u0026ndash;1086. https://doi.org/10.1038/nmeth.4463\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eHowe K, Clark MD, Torroja CF et al (2013) The zebrafish reference genome sequence and its relationship to the human genome. Nature 496:498\u0026ndash;503. https://doi.org/10.1038/nature12111\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eFontana BD, Mezzomo NJ, Kalueff AV, Rosemberg DB (2018) The developing utility of zebrafish models of neurological and neuropsychiatric disorders: a critical review. Exp Neurol 299:157\u0026ndash;171. https://doi.org/10.1016/j.expneurol.2017.10.004\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eAzman KF, Zakaria R (2019) D-galactose-induced accelerated aging model: an overview. Biogerontology 20:763\u0026ndash;782. https://doi.org/10.1007/s10522-019-09837-y\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eAdams MM, Kafaligonul H (2018) Zebrafish \u0026mdash; a model organism for studying the neurobiological mechanisms underlying cognitive brain aging and use of potential interventions. Front Cell Dev Biol 6:135. https://doi.org/10.3389/fcell.2018.00135\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eColwill RM, Creton R (2011) Imaging escape and avoidance behavior in zebrafish larvae. Rev Neurosci 22:63\u0026ndash;73. https://doi.org/10.1515/RNS.2011.008\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cem\u003ePelkowski SD, Kapoor M, Richendrfer HA et al (2011) A novel high-throughput imaging system for automated analyses of avoidance behavior in zebrafish larvae. Behav Brain Res 223:135\u0026ndash;144. https://doi.org/10.1016/j.bbr.2011.04.033\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eFujimori K, Inui T, Uodome N et al (2006) Zebrafish and chicken lipocalin-type prostaglandin D synthase homologues: conservation of mammalian gene structure and binding ability for lipophilic molecules. Gene 375:14\u0026ndash;25. https://doi.org/10.1016/j.gene.2006.01.037\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eMihaljevic I, Popovic M, Zaja R, Smital T (2016) Phylogenetic, syntenic, and tissue expression analysis of slc22 genes in zebrafish (Danio rerio). BMC Genomics 17:626. https://doi.org/10.1186/s12864-016-2981-y\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eLee Y-J, Lee YM, Lee C-K et al (2011) Therapeutic applications of compounds in the Magnolia family. Pharmacol Ther 130:157\u0026ndash;176. https://doi.org/10.1016/j.pharmthera.2011.01.010\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eTao X, Zhao M, Gao K et al (2025) Wedelolactone ameliorates ischemic stroke by inhibiting oxidative damage and ferroptosis via HIF-1\u0026alpha;/SLC7A11/GPX4 signaling. Drug Des Devel Ther 19:6849\u0026ndash;6868. https://doi.org/10.2147/DDDT.S528831\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eYang L, Ji C, Li Y et al (2020) Natural potent NAAA inhibitor atractylodin counteracts LPS-induced microglial activation. Front Pharmacol 11:577319. https://doi.org/10.3389/fphar.2020.577319\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eKraeuter A-K, Guest PC, Sarnyai Z (2019) The Y-maze for assessment of spatial working and reference memory in mice. Methods Mol Biol 1916:105\u0026ndash;111. https://doi.org/10.1007/978-1-4939-8994-2_10\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eMucke L, Selkoe DJ (2012) Neurotoxicity of amyloid \u0026beta;-protein: synaptic and network dysfunction. Cold Spring Harb Perspect Med 2:a006338. https://doi.org/10.1101/cshperspect.a006338\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eDaina A, Michielin O, Zoete V (2019) SwissTargetPrediction: updated data and new features for efficient prediction of protein targets of small molecules. Nucleic Acids Res 47:W357\u0026ndash;W364. https://doi.org/10.1093/nar/gkz382\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eWang X, Shen Y, Wang S et al (2017) PharmMapper 2017 update: a web server for potential drug target identification with a comprehensive target pharmacophore database. Nucleic Acids Res 45:W356\u0026ndash;W360. https://doi.org/10.1093/nar/gkx374\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cem\u003ePi\u0026ntilde;ero J, Ram\u0026iacute;rez-Anguita JM, Sa\u0026uuml;ch-Pitarch J et al (2020) The DisGeNET knowledge platform for disease genomics: 2019 update. Nucleic Acids Res 48:D845\u0026ndash;D855. https://doi.org/10.1093/nar/gkz1021\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eAmberger JS, Bocchini CA, Schiettecatte F et al (2015) OMIM.org: Online Mendelian Inheritance in Man, an online catalog of human genes and genetic disorders. Nucleic Acids Res 43:D789\u0026ndash;D798. https://doi.org/10.1093/nar/gku1205\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eStelzer G, Rosen N, Plaschkes I et al (2016) The GeneCards suite: from gene data mining to disease genome sequence analyses. Curr Protoc Bioinformatics 55:1.30.1\u0026ndash;1.30.33. https://doi.org/10.1002/cpbi.5\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eSzklarczyk D, Kirsch R, Koutrouli M et al (2023) The STRING database in 2023: protein-protein association networks and functional enrichment analyses for any sequenced genome of interest. Nucleic Acids Res 51:D638\u0026ndash;D646. https://doi.org/10.1093/nar/gkac1000\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eShannon P, Markiel A, Ozier O et al (2003) Cytoscape: a software environment for integrated models of biomolecular interaction networks. Genome Res 13:2498\u0026ndash;2504. https://doi.org/10.1101/gr.1239303\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eZhou Y, Zhou B, Pache L et al (2019) Metascape provides a biologist-oriented resource for the analysis of systems-level datasets. Nat Commun 10:1523. https://doi.org/10.1038/s41467-019-09234-6\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eEberhardt J, Santos-Martins D, Tillack AF, Forli S (2021) AutoDock Vina 1.2.0: new docking methods, expanded force field, and Python bindings. J Chem Inf Model 61:3891\u0026ndash;3898. https://doi.org/10.1021/acs.jcim.1c00203\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eJohnson WE, Li C, Rabinovic A (2007) Adjusting batch effects in microarray expression data using empirical Bayes methods. Biostatistics 8:118\u0026ndash;127. https://doi.org/10.1093/biostatistics/kxj037\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eCleveland WS (1981) LOWESS: a program for smoothing scatter plots by robust locally weighted regression. Am Stat 35:54. https://doi.org/10.2307/2683591\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eLiu C-C, Kanekiyo T, Xu H, Bu G (2013) Apolipoprotein E and Alzheimer disease: risk, mechanisms and therapy. Nat Rev Neurol 9:106\u0026ndash;118. https://doi.org/10.1038/nrneurol.2012.263\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eBates D, M\u0026auml;chler M, Bolker B, Walker S (2015) Fitting linear mixed-effects models using lme4. J Stat Softw 67:1\u0026ndash;48. https://doi.org/10.18637/jss.v067.i01\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eKaplan EL, Meier P (1958) Nonparametric estimation from incomplete observations. J Am Stat Assoc 53:457\u0026ndash;481\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eCox DR (1972) Regression models and life-tables. J R Stat Soc B 34:187\u0026ndash;202\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eBenjamini Y, Hochberg Y (1995) Controlling the false discovery rate: a practical and powerful approach to multiple testing. J R Stat Soc B 57:289\u0026ndash;300\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eAdav SS, Park JE, Sze SK (2019) Quantitative profiling brain proteomes revealed mitochondrial dysfunction in Alzheimer\u0026apos;s disease. Mol Brain 12:8. https://doi.org/10.1186/s13041-019-0430-y\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eWang X, Zhang Z (2020) Mitochondria dysfunction in the pathogenesis of Alzheimer\u0026apos;s disease: recent advances. Mol Neurodegener 15:30. https://doi.org/10.1186/s13024-020-00376-6\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eUnderwood CK, Coulson EJ (2008) The p75 neurotrophin receptor. Int J Biochem Cell Biol 40:1664\u0026ndash;1668. https://doi.org/10.1016/j.biocel.2007.06.010\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cem\u003ePitta S, Augustine BB, Kasala ER, Sulakhiya K, Ravindranath V, Lahkar M (2013) Honokiol reverses depressive-like behavior and decrease in brain BDNF levels induced by chronic corticosterone injections in mice. Pharmacognosy J 5:211\u0026ndash;215. https://doi.org/10.1016/j.phcgj.2013.08.004\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eBustos FJ, Ampuero E, Jury N, Aguilar R, Falahi F, Toledo J, Ahumada J, Lata J, Cubillos P, Henr\u0026iacute;quez B, Guerra MV, Stehberg J, Neve RL, Inestrosa NC, Wyneken U, Fuenzalida M, H\u0026auml;rtel S, Sena-Esteves M, Varela-Nallar L, Rots MG, Montecino M, van Zundert B (2017) Epigenetic editing of the Dlg4/PSD95 gene improves cognition in aged and Alzheimer\u0026apos;s disease mice. Brain 140:3252\u0026ndash;3268. https://doi.org/10.1093/brain/awx272\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eFolstein MF, Folstein SE, McHugh PR (1975) \u0026apos;Mini-mental state\u0026apos;. A practical method for grading the cognitive state of patients for the clinician. J Psychiatr Res 12:189\u0026ndash;198\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cem\u003ePrater KE, Green KJ, Mamde S et al (2023) Human microglia show unique transcriptional changes in Alzheimer\u0026apos;s disease. Nat Aging 3:894\u0026ndash;907. https://doi.org/10.1038/s43587-023-00424-y\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eMagistretti PJ, Allaman I (2018) Lactate in the brain: from metabolic end-product to signalling molecule. Nat Rev Neurosci 19:235\u0026ndash;249. https://doi.org/10.1038/nrn.2018.19\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eScheffer M, Carpenter SR, Lenton TM et al (2012) Anticipating critical transitions. Science 338:344\u0026ndash;348. https://doi.org/10.1126/science.1225244\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eHeineke J, Molkentin JD (2006) Regulation of cardiac hypertrophy by intracellular signalling pathways. Nat Rev Mol Cell Biol 7:589\u0026ndash;600. https://doi.org/10.1038/nrm1983\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eYagami T, Yamamoto Y, Koma H (2018) Physiological and pathological roles of 15-deoxy-delta12,14-prostaglandin J2 in the central nervous system and neurological diseases. Mol Neurobiol 55:2227\u0026ndash;2248. https://doi.org/10.1007/s12035-017-0435-4\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eScher JU, Pillinger MH (2005) 15d-PGJ2: the anti-inflammatory prostaglandin? Clin Immunol 114:100\u0026ndash;109. https://doi.org/10.1016/j.clim.2004.09.008\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eZhou Z, Zhang Y, Liu S et al (2025) Ferroptosis in Alzheimer\u0026apos;s disease: molecular mechanisms and advances in therapeutic strategies. Front Neurosci 19:1673315. https://doi.org/10.3389/fnins.2025.1673315\u003c/em\u003e\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eAshraf A, Jeandriens J, Parkes HG, So P-W (2020) Iron dyshomeostasis, lipid peroxidation and perturbed expression of cystine/glutamate antiporter in Alzheimer\u0026apos;s disease: evidence of ferroptosis. Redox Biol 32:101494. https://doi.org/10.1016/j.redox.2020.101494\u003c/em\u003e\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":"acta-neuropathologica-communications","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"anec","sideBox":"Learn more about [Acta Neuropathologica Communications](https://actaneurocomms.biomedcentral.com/)","snPcode":"40478","submissionUrl":"https://submission.springernature.com/new-submission/40478/3","title":"Acta Neuropathologica Communications","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"BMC/SO AJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"Alzheimer’s disease, astrocyte, PTGDS, neuropathological transition, MCI, lipocalin-2, NGFR, phase boundary, SEA-AD, CSF biomarker","lastPublishedDoi":"10.21203/rs.3.rs-9499795/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9499795/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eAlzheimer’s disease (AD) is characterized by prolonged prodromal stability preceding abrupt, often irreversible cognitive decline, yet the neuropathological event governing this transition remains undefined. Here, we identify a statistically bounded astrocytic inflection anchored to prostaglandin D2 synthase (PTGDS) dynamics that delineates a transition boundary between reversible compensatory buffering and accelerated neurodegeneration during the mild cognitive impairment (MCI)-to-AD continuum. Integrating pseudo-progression analysis of 1.3 million single nuclei from the SEA-AD atlas (84 donors) with cross-species validation in zebrafish and murine systems and longitudinal ADNI cerebrospinal fluid (CSF) proteomics (n = 735), we demonstrate that astrocytic PTGDS undergoes a biphasic trajectory with a statistically significant inflection point (segmented regression breakpoint: Bin 0.23, 95% CI: 0.13–0.33; Davies’ p = 0.032; quadratic vertex: continuous pseudo-progression score [CPS] 0.46, β₂ = −2.07, p \u0026lt; 2.2 × 10\u003csup\u003e−16\u003c/sup\u003e). This inflection defines a neuropathological boundary at which metabolic exhaustion precedes and gates lipocalin-2 (LCN2)-mediated inflammatory amplification and CREB/NGFR-linked neurogenic suppression — repositioning inflammation as a downstream consequence of astrocytic metabolic failure. The boundary is conserved across human, zebrafish, and murine systems and aligns clinically with MMSE 26.0, beyond which pharmacological rescue of cognitive deficits is substantially attenuated. In longitudinal ADNI CSF proteomics, the PTGDS/LCN2 ratio independently predicts MCI-to-AD conversion (AUC = 0.743; HR = 3.2, 95% CI: 2.3–4.4), preceding structural atrophy. These findings establish astrocytic PTGDS exhaustion as a cross-scale neuropathological boundary with direct implications for stage-stratified therapeutic intervention in prodromal AD.\u003c/p\u003e","manuscriptTitle":"Astrocytic PTGDS inflection defines a neuropathological transition boundary during prodromal Alzheimer’s disease","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-04-24 06:41:27","doi":"10.21203/rs.3.rs-9499795/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"reviewerAgreed","content":"291853855128366596628892560977148026075","date":"2026-05-10T15:27:18+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2026-04-26T14:17:23+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2026-04-25T15:29:59+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2026-04-24T12:53:13+00:00","index":"","fulltext":""},{"type":"submitted","content":"Acta Neuropathologica Communications","date":"2026-04-22T19:02:42+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"acta-neuropathologica-communications","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"anec","sideBox":"Learn more about [Acta Neuropathologica Communications](https://actaneurocomms.biomedcentral.com/)","snPcode":"40478","submissionUrl":"https://submission.springernature.com/new-submission/40478/3","title":"Acta Neuropathologica Communications","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"BMC/SO AJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"4d3c05af-6ee5-4f63-964b-ae89598299ca","owner":[],"postedDate":"April 24th, 2026","published":true,"recentEditorialEvents":[{"type":"reviewerAgreed","content":"291853855128366596628892560977148026075","date":"2026-05-10T15:27:18+00:00","index":23,"fulltext":""}],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[],"tags":[],"updatedAt":"2026-04-26T14:23:20+00:00","versionOfRecord":[],"versionCreatedAt":"2026-04-24 06:41:27","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9499795","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9499795","identity":"rs-9499795","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}