Drug delivery compositions and methods using thereof

Abstract

The present disclosure is related to drug delivery compositions containing extracellular vesicles having a therapeutic by utilizing visceral sensory neurons, a method using thereof for delivering the drugs to central nervous system and an organoid on chip model comprising a three-compartment axis-on-a chip for testing delivery of the drug containing EV over BBB, wherein the compositions comprising bacterial extracellular vesicles (EV) and therapeutics, wherein the drug or the therapeutic is encapsulated within the EV or attached or conjugated to EV.

Description

[0001] Docket No. 043846.00009 (PCT)

[0002] DRUG DELIVERY COMPOSITIONSAND METHODS USING THEREOF

[0003] CROSS-REFERENCE TO RELATED APPLICATION

[0004] This application claims the benefit of priority, under 35 U.S.C. § 119, from U.S.

[0005] Provisional Patent Application No. 63 / 760,269, filed February 19, 2025, the entire contents of which are incorporated herein by reference.

[0006] GOVERNMENT LICENSE RIGHTS STATEMENT

[0007] The present disclosure was made with the support of the SRnD Project number A0132-20250001 and Seoul National University's Regional Innovation-Oriented University Support System (RISE) Leading Global Industry-Academic Cooperation under Research project Seoul City RISE Project managed by Seoul RISE Center under Ministry of Education, Project Number 2025-RISE-01-016-01. The government of City of Seoul and Ministry of Education may have certain rights in the invention.

[0008] SEQUENCE LISTING

[0009] This application incorporates by reference in its entirety the Sequence Listing XML file entitled “Sequence Listing 043846.00009. xml (8 KB)”, which was created on February 19, 2026, and filed electronically herewith.

[0010] FIELD

[0011] The present disclosure is related to drug delivery compositions containing extracellular vesicles having drugs by utilizing visceral sensory neurons and a method using thereof for delivering the drugs to central nervous system.

[0012] BACKGROUND

[0013] The blood-brain barrier (BBB) is a complex barrier that controls and limits the delivery of therapeutics to the central nervous system. The BBB in healthy brain is a diffusion barrier essential for protecting normal brain function by impeding most compounds from transiting from the blood to the brain. Under certain pathological conditions of diseases such as stroke, diabetes, seizures, multiple sclerosis, Parkinson's disease and Alzheimer disease, the BBB is disrupted. It's indispensable for treatment of diseases, but it's also a formidable obstacle in treating disorders such as neurological diseases.

[0014] Delivering drugs through the BBB is challenging because the BBB is highly selective and almost impermeable. Recent efforts to develop therapeutics for neurological disorders have focused on improving drug delivery across the blood-brain barrier (BBB). The global drug development for brain diseases has to grow rapidly in the next 20 years as theDocket No. 043846.00009 (PCT)

[0015] populations of seniors and patients with CNS disorders are increasing. However, drug development for brain diseases has the poorest success rates compared to other therapeutic areas. The time for developing CNS drugs is normally much longer than for non-CNS drugs. Clinical trials of CNS drugs become challenging because of the complexity of the brain, side effects and the impermeable blood-brain barrier (BBB).

[0016] Despite extensive research, there remains need for effective method to efficiently transport drugs into the brain. In addition to the complexity of brain diseases, the lack of efficient technologies to deliver drugs across the BBB hinders CNS drug development. Both small molecules and macromolecules are investigated as effective therapeutic agents to treat various brain diseases. However, only small molecules that are lipid soluble and also have a molecular weight <400 Da were found to cross the BBB, but most macromolecules cannot penetrate the brain endothelium. This physiological hurdle of the BBB stops 95% of molecules for drug development.

[0017] Recent advancements in neuropharmacology have led to the development of various therapeutic agents targeting neurological disorders. However, one of the greatest challenges in establishing effective treatment strategies still lies in the need for these agents to cross the blood-brain barrier (BBB) to exert their effects on the central nervous system (CNS). The BBB strictly regulates the passage of substances from the peripheral circulation into the brain, allowing only highly selective molecules to permeate.

[0018] SUMMARY

[0019] The present disclosure provides compositions comprising bacterial extracellular vesicles (EV) and therapeutics.

[0020] The present disclosure also provides methods for delivering therapeutics to brain using a composition comprising extracellular vesicles (EV) containing therapeutics through visceral sensory neurons (VSNs).

[0021] The present disclosure further provides a organoid on chip model comprising a three-compartment axis-on-a chip for testing delivery of the drug containing EV over BBB.

[0022] In an embodiment of the present disclosure the EV is bacterial EV.

[0023] In an embodiment of the present disclosure the three compartments on the organoid on chip comprises gut, nerve, and brain.

[0024] In an embodiment of the present disclosure the drug or the therapeutic is encapsulated within the EV or attached or conjugated to EV.

[0025] BRIEF DESCRIPTION OF DRAWINGS FIG. 1 presents Comparison of the Relationship Between Vagotomy and LPS Deposition in the Hippocampus: (a) Comparison of hippocampal LPS deposition using anDocket No. 043846.00009 (PCT)

[0026] LPS-specific antibody (LPS 2D7 / 1, red); (b) Quantification of LPS deposition levels in WT, AS, and AV mice; (c) High-magnification image of hippocampal LPS signals, revealing morphology resembling glial cells; (d) Immunohistochemical staining using a different LPS antibody (LPS C6, purple) to validate the findings observed in (a); and (e) Quantification of hippocampal LPS signals, similar to (b), for comparative analysis. Statistical analysis was performed using Brown-Forsythe ANOVA followed by Tamhane’s T2 post-hoc test. Statistical significance was denoted as p < 0.05. AS: Alzheimer's disease (AD) model mouse sham-operated, AV: Alzheimer's disease model mouse vagotomized, LPS: Lipopolysaccharide, WT: Wild type mouse sham-operated.

[0027] FIG. 1 presents Analysis of the Brain Entry Pathway of Lipopolysaccharides (LPS): (a) Immunohistochemical staining of the medulla in AS mice. LPS is labeled in red, CHAT in green, and DAPI in blue. LPS signals are predominantly concentrated in mcNTS. Scale bar: 150 pm; and (b) Quantification of LPS signals across different regions of the medulla.

[0028] Statistical analysis was performed using the Friedman test followed by the Conover post-hoc test. Statistical significance was denoted as p < 0.05.AP: Area postrema, AS: Alzheimer’s disease model mouse sham-operated, DMVN: Dorsal motor nucleus of vagus nerve, IcNTS: lateral-caudal nucleus tractus solitarius, LPS: Lipopolysaccharide, mcNTS: medial-caudal nucleus tractus solitarius.

[0029] FIG. 2 presents Comparison of LPS Deposition in the Medulla Based on Vagotomy Status: (a) Bilateral comparison of LPS signals in the medulla of AS mice; (b) Quantification of LPS deposition in the NTS of the medulla, comparing the left and right sides; (c) Bilateral comparison of LPS signals in the medulla of AV mice; and (d) Quantification of LPS deposition in the NTS of the medulla, comparing the vagotomized side and the non-vagotomized side. Statistical analysis was performed using the Student’s t-test. Statistical significance was denoted as p < 0.01, and ns indicates non-significant results. AS: Alzheimer's disease model mouse sham-operated, AV: Alzheimer's disease model mouse vagotomized, AP: Area postrema, DMVN: Dorsal motor nucleus of vagus nerve, LPS: Lipopolysaccharide, mcNTS: Medial caudal nucleus tractus solitarius, NTS: Nucleus tractus solitarius.

[0030] FIG. 3 presents Accumulation of LPS in Nodose ganglion: (a) LPS accumulation in PHOX2B-positive Nodose ganglion of WT mice. PHOX2B is labeled as green, TUJ1 in yellow and LPS in red; (b) LPS accumulation in PHOX2B-positive Nodose ganglion of AD mice. PHOX2B is labeled as green, TUJ1 in yellow and LPS in red; and (c) Quantification of LPS positive PHOX2B positive visceral sensory neuron proportion in WT mice and AD mice.Docket No. 043846.00009 (PCT)

[0031] Statistical analysis was performed using the Student’s t-test. Statistical significance was denoted as p < 0.01, and ns indicates non-significant results.

[0032] FIG. 4 presents Propagation of bacterial extracellular vesicles occurs via visceral sensory neuron: (a) Overview of the gut-nerve-brain axis on a chip, showing the structural connection between the HCO, VSGO and HBO. DSG2 (red), TLIJ1 (green), DAPI (blue); and (b) Propagation of LPS along the gut-nerve-brain axis. LPS (2D7 / 1, red) was detected in the VSGO and along neuronal projections, demonstrating the movement of bacterial extracellular vesicles via visceral sensory neurons. TLIJ1 (green) marks neuronal structures, and DAPI (blue) labels cell nuclei. High-magnification images reveal co-localization of LPS and TLIJ1, supporting the role of visceral sensory neurons in mediating gut-to-brain transport of bacterial components. HBO: Human brain organoids, HCO: Human colon organoids, VSGO: Visceral sensory ganglion organoids

[0033] FIG. 5 presents TNF-a stimulation can activate visceral sensory neurons in gut-nerve brain axis: (a) Schematic representation of live calcium imaging in the gut-nerve-brain axis on a chip. TNF-a was applied to the colon chamber, and calcium imaging was performed in the VSGO over 10 minutes to assess neuronal activation; (b) Baseline calcium fluorescence before TNF-a treatment, showing minimal neuronal activity; (c) Calcium fluorescence increase after TNF-a treatment, indicating neuronal activation in response to inflammatory signals transmitted through the gut-nerve axis; and (d) Quantification of calcium activity following TNF-a stimulation, showing a gradual and sustained increase in fluorescence intensity overtime. (Mean ± SEM). HBO: Human brain organoid, HCO: Human colon organoid, VSGO: Visceral sensory ganglion organoids

[0034] FIG. 6 presents TNF-a enhances the transport of bacterial extracellular vesicles via the gut-nerve-brain axis: (a) Schematic representation of experimental design. Alexa-488 labeled E.co / / -derived OMVs the present disclosure re treated into the colon chamber in two conditions: EV only treatment (blue arrow) and EV with TNF- a treatment (red arrow).

[0035] Fluorescence signals the present disclosure re measured in the brain chamber every day. Immunostaining was held 3 days after treatment; (b) Comparing Alexa-488 fluorescent intensity between EV-only and EV + TNF-a treated groups. For statistical analysis, repetitive t-test was performed; (c) Fluorescent imaging of E.coli EVs in neurite of VSGO under EV-only condition. TLIJ1 (red), EV (green). Image was captured in brain chamber; (d) Fluorescent imaging of E.coli EVs in neurite of VSGO under EV+TNF-a treated condition. TLIJ1 (red), EV (green). Image was captured in brain chamber; and (e) Quantification of propagated EV puncta count in VSGO neurite.

[0036] FIG. 8 presents LPS-responsive microglia precede DAM transformation of microglia in AD: (a) Immunofluorescence staining of LPS, Ap and microglia (MG) in Alizheimer’s disease (AD) model mice. Note the colocalization of LPS, A and microglia. Scale bars,Docket No. 043846.00009 (PCT)

[0037] 20pm; (b) Schematic diagram of public MG single nuclei RNA sequencing dataset; (c) LIMAP clustering of conventional MG subtypes, labeled by previous study; (d) Schematic diagram of calculating LPS score by stat-weighted univariate linear model (ULM), utilizing DEGs provided in previous study; (e) LIMAP based on LPS score calculated by ULM; (f) Violin plot demonstrating LPS score among conventional clusters; (g) Upset plot of intersecting DEGs between DAM and LPS-responsiveness. Number of DEGs specific to the cluster was plotted; (h) UMAP clustering by semi-supervised clustering using intersecting DEGs between LPS and DAM; (i) Sankey diagram comparing conventional MG type and newly defined clusters by LPS and DAM signatures; (j) Matrix plot showing scaled expression levels of canonical gene markers among LPS and DAM clusters; (k) Matrix plot demonstrates scaled gene set variational analysis (GSVA) scores of Reactome pathway 2024 among groups. FDR was calculated by Welch’s t-test with Benjamini-Hochberg correction; (I) Fractions of MG in each DAM / LPS category plotted against the continuous pseudo-progression score. Curves show generalized additive model (GAM) fits and shaded ribbons indicate 95% confidence intervals; (m) Fractions of MG in each DAM / LPS category stratified by overall AD neuropathological change (ADNC) score. (Point indicates mean, error bar indicates SEM; p value was calculated by Spearman’s Rank Correlation Coefficient).

[0038] FIG. 9 presents Quality control matrix for pseudobulk analysis in human postmortem analysis between DAM and others (a-c) and Quality control matrix for pseudobulk analysis between LPS and DAM clusters (d-f), wherein (a) Additional immunostaining data of LPS accumulated in hippocampus. Scale bars, 100pm; (b) Additional immunostaining data of LPS accumulation in dorsal medulla of AD mouse, supported by ChAT staining of dorsal motor vagal nucleus (DM N). Scale bars, 150pm.

[0039] FIG. 10 presents vagal sensory neurons are the main route of LPS transmission: (a) Immunofluorescence staining of LPS in the dorsal medulla of AD mouse. Area postrema (AP), medial caudal nucelus tractus solitarius (mcNTS), lateral caudal nucleus tractus solitarius (IcNTS), dorsal motor vagal nucleus (DMVN) were anatomically distinguished, supported by ChAT staining of DMVN. Scale bars, 150pm; (b) Quantification of lipopolysaccharaide (LPS) positive area in A (n=4, Friedman test with conover post-hoc test; P=0.0173 between mcNTS and IcNTS; P=0.0173 between mcNTS and AP; P=0.0024 between mcNTS and DMVN); (c) Immunofluorescence staining of LPS in nodose ganglia of WT and AD mouse. Scale bars, 50pm; (d) Quantification of LPS positive neurons among Phox2B positive neurons (n=5 for each group; Student t-test P=0.0345); (e) Overview of the experimental design. ADLPmand ADLPAPTmice were either vagotomized (AV) or sham-operated (WT for ADLPwrand AS for ADLPAPT); (f) Immunofluorescence staining of LPS in the dorsal medulla of AV mouse. Scale bars, 100pm; (g) Quantification of LPS positive area in mcNTS of vagotomized mice (n=4 for left side; n=7 for right side; Mann Whitney’s U-Docket No. 043846.00009 (PCT)

[0040] test; P=0.0064); (h) Immunofluorescence staining of LPS accumulated in hippocampus. Scale bars, 100pm; (i) Quantification of d (n=4, 11, 9 for WT, AS and AD, respectively;

[0041] Brown-Forsythe ANOVA test with Tamhane’s T2 post-hoc test; P=0.0041 between WT and AS; P=0.0006 between AS and AV).

[0042] FIG. 11 presents additional immunofluorescence staining data of AD model mouse brain: (a) Additional immunostaining data of LPS accumulation in dorsal medulla of AD mouse, supported by ChAT staining of dorsal motor vagal nucleus (DMVN). Scale bars, 150pm; (b) Additional immunostaining data of LPS accumulated in hippocampus. Scale bars, 100pm.

[0043] FIG. 12 presents reduced lipopolysaccharide (LPS) accumulation upon vagotomy: (a) schematic description of operation. Posterior vagal trunk was exposed (Sham-operation) and cut (posterior subdiaphragmatic vagotomy); (b) Comparison of food consumption upon vagotomy (CCK-8 test). (n=6 for each group, Student t-test); (c) Fluorogold retrograde tracer accumulation between right and left dorsal motor vagal nuclei. Scale bars, 100pm; (d) Quantification of right-to-left ratio of Fluorogold positive neuronal counts (n=5 for each group, Welch’s t-test); (e) Immunofluorescence staining of LPS in the dorsal medulla of sham-operated mouse. Scale bars, 100pm; (f) Quantification of LPS (+) area in mcNTS of sham-operated mice (n=6 for left side, n=4 for right side, Mann-Whitney U test); (g) Immunofluorescence staining of LPS lipid A in hippocampus between WT and AD. Scale bars, 20pm; (h) Quantification of lipid A positive area among conditions (n=4 for WT and n=5 for AD, Mann-Whitney U test); (i) Immunofluorescence staining of LPS in hippocampus with C6 antibody between AS and AV. Scale bars, 100pm; (j) Quantification of LPS C6 antibody positive area among conditions (n=6 for each group, Mann-Whitney U test).

[0044] FIG. 13 presents Vagotomy ameliorates amyloidopathy and neuroinflammation in an AD mouse model: (a) Immunofluorescence staining of Ap accumulation in AD mouse cortex. Scale bars, 100pm; (b) Quantification of f (n=6 for AS and n=5 for AV; Student’s t-test; P=0.0365); (c-h) Ap ELISA quantification of cortex (n=3 for each group’ Kruskal-wallis test with conover post-hoc test); (c) RIPA insoluble fraction of Api-42 (P=0.0003 between WT and AS; P=0.0104 between WT and AV; P=0.0104 between AS and AV); (d) RIPA insoluble fraction of Api-40 (P=0.0003 between WT and AS; P=0.0104 between WT and AV; P=0.0104 between AS and AV); (e) Api-42 / Api-40 ratio among RIPA insoluble fraction; (f) RIPA soluble fraction of Api-42; (g) RIPA soluble fraction of Api-42; (h) RIPA soluble fraction of Api-40; (I and j) Immunofluorescence staining of astrocytes and microglia in hippocampus. Scale bars, 100pm (AS for j; AV for k); (k and I) Quantification of j and k. (n=5 for AS, n=4 for AV; P=0.0422 for Iba1 and P=0.3342 for GFAP, Welch’s t-test). ns, not significant.Docket No. 043846.00009 (PCT)

[0045] FIG. 14 presents Additional immunofluorescence staining data of AD model mouse brain: (a) Immunofluorescence staining of phosphorylated tau (AT8) in hippocampus. Scale bars, 100pm; (b) Quantification of AT8 positive area upon vagotomy (n=6 for AS and n=5 for AV, p=0.6623, Mann-Whitney ll-test); (c) Immunofluorescence staining of phosphorylated tau (AT180) in hippocampus. Scale bars, 100pm; (d) Quantification of AT180 positive area upon vagotomy (n=6 for AS and n=5 for AV, p=0.5499, Student t-test).

[0046] FIG. 15 presents Single-nuclei RNA sequencing of Nodose ganglion reveals characteristics of AD Nodose ganglia: (a) Schematic overview of nodose ganglia single nucleus RNA-sequencing; (b) LIMAP clustering of cell types in the samples; (c) Dot plot for representative markers of each cell types; (d) Volcano plot comparing AD and WT nodose differentially expressed genes (DEGs). Significantly upregulated motor proteins are annotated; (e) GO terms of upregulated genes in AD nodose; dashed line denotes adjusted P-value of 0.05; (f) GO terms of downregulated genes in AD nodose; dashed line denotes adjusted P-value of 0.05; (g) Gene-set enrichement analysis (GSEA) for upregulated GO terms in AD nodose; (h) Volcano plot of predicted differentially expressed transcription factors between AD and WT analyzed with CollecTRI gene-regulatory network analysis; (i) Quantification of mouse colon TNF-a level with ELISA. (n=9 for each group; Welch’s t-test; P=0.0078).

[0047] FIG. 16 presents snRNA-seq analysis baseline information: (a) Quality control violin plots for integrated AD and WT Nodose ganglion (NG); (b) Quality control metrics of snRNA-seq libraries; (c) LIMAP of canonical markers for each cell type; (d) LIMAP of nodose ganglia sample distribution.

[0048] FIG. 17 presents TNF-a enhances propagation of LPS containing OMV via VSN in a 3-compartment microfluidic chip: (a) Schematic depiction of gut-nerve-brain axis on a 3-compartment chamber microfluidic chip. Brain organoid (BO), visceral sensory ganglion organoid (VSGO), and colon organoid (CO) were differentiated from induced pluripotent stem cells (iPSC), then seeded on the chip; (b) Immunofluorescence staining of entire 3-compartment microfluidic chip. Scale bar, 1000 pm; (c and d) Calcium imaging of VSGO upon TNF-a treatment in the colon chamber; (e) Time-dependent change in fluorescence signal of VSGO; (f) Quantification of maximum signal intensity of J (n=5; paired t-test;

[0049] P=0.0229); (g) Schematic depiction of outer membrane vesicle (OMV) treatment experiment; (h) Schematic flow of fluorescence measurement in the media of brain chamber. OMV was treated every 24 hours for 3 days, and the media was changed into Opti after 12 hours. The Opti media was harvested for measurement; (i, j) Comparison of OMV-tagged Alexa-488 fluorescence between OMV-only treated group and OMV+ TNF-a treated group in day 1 media and day 3 media, respectively (n= 22 for each group; Student t-test; P=0.0182 for day 3); (k) Immunofluorescence staining of brain-side visceral sensory neurite for visualization ofDocket No. 043846.00009 (PCT)

[0050] Alexa-488-tagged OMV. Scale bars, 5pm; (I) Quantification of OMV puncta in k (n= 5 for OMV only group, n=4 for OMV + TNF-a group; Mann whitney’s U-test;P=0.0159).

[0051] FIG. 18 presents general experimental process of using microfluidic chip: (a) Additional immunostaining data of gut-nerve-brain axis-on-a-chip. Scale bar, 1000 pm; (b) Overall design of axis-on-a-chip; (c) Microfluidic chip is assembled with media chip, channel chip and cover glass; (d) Representative photo of assembled microfluidic chip; (e) Schematic drawing of assembled axis-on-a-chip; (f) Schematic drawing for the gut-nerve-brain axis-on-a-chip experimental timeline.; (g) Schematic indicating top view and cross-section view for setting entire process of gut-nerve-brain axis-on-a-chip.

[0052] FIG. 19 presents modeling 3 compartment axis-on-a-chip: (a) Brightfield microscopic image and FITC-dextran signal in 3-compartment microfluidic chip. 40kDa FITC-dextran was treated in the colon chamber; (b) Quantification of FITC-dextran signal in each compartment of the microfluidic chip; (c) Schematic overview of electrophysiologic activity measurement upon treatment of bile acids; (d) Depiction of probe measuring electrical activity (field potential) of brain organoid. Scale bar, 50pm; (e) Image of actual electrical measurement; (f) Quantitative comparison of brain organoid field potential band power between ‘media only’ and ‘bile acid treatment’ (n=9 for each group, Student t-test); (g) Quantitative comparison of brain organoid field potential coherence recorded in each channel (n=9 for each group, Student t-test); (h) Schematic overview of electrophysiologic activity measurement upon treatment of D-glucose; (i) Average band power of brain organoid electrical activity; (j) Quantitative comparison of brain organoid field potential band power between before and after D-glucose treatment (n=7, paired t-test); (k) Coherence of electrical activity recorded in each channel; (I) Quantitative comparison of brain organoid field potential coherence recorded in each channel (n=7, paired t-test); (m) Schematic overview of pseudorabies virus (PRV) transmission experiment. PRV-infected colon organoid was seeded and connected with VSGO to validate anatomical connection of the whole gut-nerve-brain axis-on-a-chip; (n) Representative immunofluorescence image of PRV transmission through VSGO to BO. Scale bar, 1000pm; (o) Immunofluorescence staining of VSGO connected with PRV-CO. Scale bar, 50pm; (p) Immunofluorescence staining of BO connected with PRV-CO via VSGO. Scale bar, 50pm.

[0053] FIG. 20 presents Spatial transcriptomics of nodose ganglia reveal gut VSN as the primary transmitter of LPS: (a) Schematic overview of spatial transcriptomic analysis; (b) Transcriptome-derived cell type assignments (colored dots) are overlaid on the corresponding immunofluorescence-stained tissue section Immunofluorescence staining image of nodose ganglia on each slide used for the analysis; (c) LIMAP clustering of cell types in the samples; (d) Distribution and relative fraction of WT and AD cells on the clustered LIMAP; (e) Dot plot for representative markers of each cell types; (f) IntegratedDocket No. 043846.00009 (PCT)

[0054] subclusters of nodose ganglia from ST, snRNA-seq and previous scRNA-seq study of Kupari et al; (g) Subcluster matching with the study of Kupari et al; (h) Dot plot for representative markers of each subclusters, i, Innervating organ-based classification of subclusters according to previous study of Zhao et al; (j) Visualization of organ-based subclustering on the LIMAP; (k) Quantified LPS signal intensity of each cells plotted on immunofluorescence image; (I) Distribution of LPS signal intensity per cell. Upper 20% was named ‘LPS high’ cells, and lower 20% was named ‘LPS low’ cells; (m) Proportion of organ-innervating VSN subclusters among LPS high cells; (n) Ratio of LPS high cells to LPS low cells in each cluster. Clusters with <2% of total cells were excluded.

[0055] FIG. 21 presents source coherence heatmap of D-glucose challenge electrophysiologic experiments: (a) Schematic overview of electrophysiologic analysis; (b) Source coherence of electrical activity recorded in each channels before D-glucose treatment; (c) Source coherence of electrical activity recorded in each channels after D-glucose treatment.

[0056] FIG. 22 presents source coherence heatmap of bile acid challenge electrophysiologic experiments: (a) Schematic overview of electrophysiologic analysis; (b) Source coherence of electrical activity recorded in each channels without bile acids treatment; (c) Source coherence of electrical activity recorded in each channels after bile acids treatment.

[0057] FIG. 23 presents spatial transcriptomics of mouse nodose ganglion: (a) Quality control violin plots for integrated AD and WT Nodose ganglia; (b) Summary of quality control metrics for each mouse nodose ganglion sample; (c) UMAP of cell type specific marker genes; (d) UMAP of mouse nodose ganglia showing 18 clusters (NG1-NG18) identified from snRNA-seq (sn_) and spatial transcriptomics (st_). NG clusters without prefixes indicate the original public scRNA-seq dataset; € Matrixplot of representative subcluster marker genes across nodose ganglia; (f-g) Violin plots showing App expression in gut VSNs between LPS low and LPS high groups. Statistical significance was determined using the Mann-Whitney’s U test (** p<0.01, **** p<0.0001). Red dashed lines indicate mean App expression values. (2.61, 2.91, 2.86 and 3.05, respectively).

[0058] FIG. 24 presents marker data for AD Nodose.

[0059] FIG 25 presents marker data for WT Nodose.

[0060] FIG. 26 presents Potential mediators of LPS-OMV uptake in AD. (a) Volcano plot of differentially expressed genes (DEGs) in gut VSNs (WT and AD groups or LPS high and LPS low groups). Significant DEGs as adjusted p-value < 0.05 and Iog2 fold changes > ±0.3; (b) Upset plot of gut VSN DEGs. Number of DEGs specific to the cluster was plotted; (c) Gene ontology (GO) of overlapping genes in Biological Process / Cellular Component and Reactome pathways; dashed red line denote adjusted p-value of 0.05; (d) Overlapping genes upregulated in both AD and LPS high cells that are relevant to ‘GQ:0009986 CellDocket No. 043846.00009 (PCT)

[0061] surface’; (e) Scatter plot of overlapping genes showing adjusted p-value across AD and LPS-high contrasts. Selected genes are denoted in red dots; (f) Matrix plot of normalized expression of candidates across nodose ganglion clusters; color denotes mean expression within group; (g) Representative image of AD nodose ganglion showing concurrence of LPS signal (intensity scale at right) and App expression; (h) Summary of AlphaFold 3 template-augmented co-folding with the E. coli LPS R1 outer core: interchain predicted TM-score (ipTM) versus minimum of the mean predicted aligned error (PAE) for each candidate; shaded area denotes acceptance thresholds; (i) Inter-chain PAE matrix for the APP E2-R1 outer-core model; (j) Orientation of the predicted APP E2-R1 outer-core complex; the ligand sits on the outward face of E2 toward the E1 side and away from the membrane side (arrows).

[0062] FIG. 27 presents sequence for AlphaFold 3 co-folding analysis.

[0063] FIG. 28 presents additional PAE plots for other candidates.

[0064] FIG. 29 presents a graphical summary of changes in Alzheimer’s disease based on transportation of pathological compositions

[0065] DETAILED DESCRIPTION

[0066] The present disclosure provides compositions comprising bacterial extracellular vesicles (EV) and therapeutics.

[0067] The present disclosure also provides methods for delivering therapeutics to brain using a composition comprising extracellular vesicles (EV) containing therapeutics through visceral sensory neurons (VSNs).

[0068] The present disclosure further provides a organoid on chip model comprising a three-compartment axis-on-a chip for testing delivery of the drug containing EV over BBB.

[0069] In an embodiment of the present disclosure the EV is bacterial EV.

[0070] In an embodiment of the present disclosure the three compartments on the organoid on chip comprises gut, nerve, and brain.

[0071] In an embodiment of the present disclosure the drug or the therapeutic is encapsulated within the EV or attached or conjugated to EV.

[0072] The EV of the present disclosure containing therapeutics also can be in incorporated into other drug delivery carriers such as nanocarriers, nanoparticles, or other efficient drug delivery system to the brain.

[0073] The present disclosure also provides methods for delivering therapeutics to brain using a composition comprising extracellular vesicles (EV) containing therapeutics through visceral sensory neurons (VSNs).

[0074] The method of the present disclosure comprises:Docket No. 043846.00009 (PCT)

[0075] preparing extracellular vesicles comprising a therapeutic;

[0076] administering the extracellular vesicles comprising the therapeutic to a subject in need thereof.

[0077] In another embodiment of the present disclosure, the composition comprising extracellular vesicles (EV) containing therapeutics is injected to the subject in need thereof near the visceral sensory neurons.

[0078] Despite extensive research, there remains no effective method to efficiently transport drugs into the brain. This patent proposes a novel approach to bypass the BBB by utilizing visceral sensory neurons (VSNs) as a drug delivery route to the central nervous system (CNS). Conducting vagotomy and immunostaining of the brainstem in an Alzheimer’s disease mouse model, the present disclosure demonstrated that bacterial extracellular vesicles (EVs) reaches the brain more efficiently via VSNs rather than through the bloodstream.

[0079] The present disclosure provides that bacterial EVs accumulated more prominently in the nucleus tractus solitarius (NTS) than in the area postrema (AP), a brainstem region lacking the BBB, indicating that VSN-mediated transport is not only effective but significantly more efficient than passive diffusion through the bloodstream.

[0080] In addition, the in vitro model of the present disclosure provides, on the gut-nerve-brain axis-on-a-chip, that bacterial EVs can be transported to the brain through activitydependent transport mechanisms of VSNs.

[0081] Therefore, the present disclosure provides that 1) VSNs can serve as a CNS drug delivery route; 2) Therapeutic agents can be loaded into bacterial EVs or other nanocarriers, such as nanoparticles, for efficient drug delivery to the brain; and 3) Visceral sensory ganglion organoids (VSGOs) can be used as a screening platform to evaluate which drug carriers are most effectively transported to the brain via VSNs.

[0082] The present disclosure provides a composition and methods for CNS drug delivery, bypassing the BBB to overcome the limitations of conventional drug delivery methods.

[0083] Recent advancements in neuropharmacology have led to the development of various therapeutic agents targeting neurological disorders. However, one of the greatest challenges in establishing effective treatment strategies lies in the need for these agents to cross the blood-brain barrier (BBB) to exert their effects on the central nervous system (CNS). The BBB strictly regulates the passage of substances from the peripheral circulation into the brain, allowing only highly selective molecules to permeate.

[0084] Although some studies suggest that the BBB is impermeable to most peripheral molecules, other studies indicate that certain short-chain fatty acids (SCFAs) and, in someDocket No. 043846.00009 (PCT)

[0085] cases, peripheral immune cells such as T cells and neutrophils could traverse the BBB under specific conditions. However, cerebrospinal fluid (CSF) analyses in humans reveal that white blood cell (WBC) counts are exceedingly low, typically fewer than five cells per milliliter, with most detected cases arising from inadvertent blood contamination during sampling, suggesting that the true count is close to zero.

[0086] Many clinical trials targeting neurological diseases have failed due to the inability of therapeutic agents to efficiently penetrate the BBB. To circumvent this issue, researchers have explored alternative drug delivery strategies, one of which involves the use of extracellular vesicles (EVs) as carriers. While in vitro models have shown some success in EV-mediated drug delivery, in vivo studies indicate that only a limited amount of EVs successfully reach the brain.

[0087] Another alternative to bypass the BBB is the utilization of the vagus nerve. The vagus nerve provides a direct anatomical connection between the peripheral organs and the brain, independent of the circulatory system. Notably, it serves as a critical conduit linking the gut and the brain. The vagus nerve transmits signals bidirectionally, with visceral sensory neurons (VSNs) relaying signals from the gut to the brain and visceral motor neurons (VMNs) conveying signals from the brain to the gut. It is suggested that VSNs facilitate the transmission of pathological proteins implicated in neurodegenerative disorders such as Alzheimer's disease and Parkinson's disease.

[0088] Utilizing the result, the present disclosure provides that VSNs can transport therapeutic agents, providing a platform for CNS drug delivery that bypasses the BBB.

[0089] The present disclosure utilizes extracellular vesicles (EVs) such as bacterial extracellular vesicles (EVs) as carriers. The in vivo mouse model demonstrated that endogenous bacterial EVs can reach the brain through VSNs rather than the BBB. To establish a in vitro system, the present disclosure employed a protocol for differentiating visceral sensory ganglion organoids (VSGOs) from induced pluripotent stem cells (iPSCs). By integrating VSGOs with colonic and brain organoids in a three-compartment axis-on-a-chip system, the present disclosure further provides the mechanism of VSN-mediated transport and establish a robust platform for CNS-targeted drug delivery.

[0090] Lipopolysaccharide reaches the brain via vagal afferent neuron in Alzheimer’s disease

[0091] Microbiome-gut-brain axis is an emerging modulator of Alzheimer’s disease.

[0092] However, blood-brain barrier limits the effects of gut microbiome on the brain. The present disclosure identifies the vagus nerve as a bypass tract directly connecting gut environment to brain. The experimental data shows that lipopolysaccharide, a microbiota-derived immunogen, enters brain via vagal afferent, also known as visceral sensory neurons. HumanDocket No. 043846.00009 (PCT)

[0093] postmortem brain transcriptomic data supports an early microglial lipopolysaccharide response that precedes disease-associated microglial transformation in Alzheimer’s disease. Mechanistically, without bound to any theory, gut inflammation activates vagal afferent, boosting propagation of lipopolysaccharides from gut to brain. Blocking lipopolysaccharide propagation by vagotomy ameliorated amyloidopathy and neuroinflammation in the brain. Moreover, amyloid precursor protein may mediate uptake of lipopolysaccharides in visceral sensory neurons. The present disclosure provides that the route of lipopolysaccharides transmission from gut to brain, delineating a therapeutically modifiable gut-vagus-brain axis linking gut dysbiosis and early Alzheimer’s disease pathogenesis and delivery of other compositions from gut to brain through the gut-vag us- brain axis.

[0094] Microbiome-Gut-Brain Axis

[0095] The present disclosure provides that the microbiome-gut-brain axis to be utilized as a significant modulator of Alzheimer’s disease (AD) pathology including delivery of desired compositions from gut to brain. One of the key mediators of microbiome-gut-brain axis is Lipopolysaccharide (LPS), an immunogen that constitutes the major component of the outer membrane of Gram-negative microbiota. Previous experimental and clinical evidence indicated that brain LPS could exacerbate neuroinflammation, amyloidopathy, and tauopathy, the hallmarks of Alzheimer’s disease (AD). Furthermore, clinical studies have detected LPS in the brains of AD patients, where it may nucleate amyloid plaques.

[0096] Even though how LPS gains access to the ostensibly aseptic brain may remain uncertain. The gut and brain are linked via two major communicating routes: Blood circulation and the vagus nerve. Two components of vagus nerve, the afferent visceral sensory neurons (VSNs) and efferent visceral motor neurons (VMNs), enable bidirectional communication between gut and brain. The present disclosure provides VSN as a delivery route for LPS to be moved from gut to brain.

[0097] The experimental study in the present disclosure utilized ADLP (AD-like phenotype) mouse model, which overexpresses A and tau in the central nervous system and shows AD-associated pathological alterations in the gut. In addition, the present disclosure analyzed human postmortem brain microglia single nuclei RNA sequencing data to show whether LPS-associated microglial signatures are enriched in early Alzheimer’s disease. In parallel, the experiments validated, in a human iPSC-derived gut-nerve-brain axis-on-a-chip that connects colon organoids (CO) with a visceral sensory ganglion organoid (VSGO) and brain organoid (BO), on the extent from gut-nerve axis-on-a-chip model. By combining in vitro experimental model with spatially resolved transcriptomic analyses of nodose ganglion (NG) in AD model mice, the data shows that gut-derived LPS reaches the brain preferentiallyDocket No. 043846.00009 (PCT)

[0098] via vagal afferents, identifying AD-associated factors that accelerate this propagation and suggesting possible mechanism of internalization within neurons in AD-relevant contexts.

[0099] LPS-responsive microglia are enriched in postmortem early AD brains

[0100] The experimental study in the present disclosure first assessed the role of LPS in AD pathogenesis in the brain. LPS colocalized with amyloid plaques and microglia in AD model mice, consistent with previous study (FIG. 8a). The data confirms that exposure to LPS could leave a detectable imprint on the microglial transcriptome in the brains of patients with AD. The data was analyzed using an open-source post-mortem human brain microglia single-nucleus RNA-sequencing dataset and re-labelled clusters for clarity based on the authors’ annotation (FIG. 8b, c).

[0101] To define an LPS-response molecular signatures in microglia, differentially expressed genes (DEGs) reported for induced microglia treated with LPS versus saline were used. After filtering significant genes, the LPS score was computed by stat-weighted univariate linear model (ULM) to assess the degree of LPS responsiveness (FIG. 8d, e). LPS scores were comparable between reactive microglia and disease-associated microglia (DAM) (FIG. 8f). The result indicates that LPS-associated transcriptional traits are not captured by conventional reactive-microglia or DAM signatures alone.

[0102] To disentangle overlapping components of the LPS and DAM signatures, discriminative markers across the intersection space were selected for downstream classification (FIG. 8g, FIG. 9a-c). We then stratified microglia into four groups defined by DAM and LPS signature status using semi-supervised clustering with the top 20 ranked genes from each intersection (FIG. 8h). At the population level, DAM (-) LPS (-) microglia comprised 50.2% of cells, whereas 26.0% were DAM (-) LPS (+), 6.9% were DAM (+) LPS (-) and 16.8% were DAM (+) LPS (+) (FIG. 8i). These DAM / LPS-defined states were distributed across conventional microglial clusters rather than being confined to a single subtype (FIG. 8i).

[0103] For understanding molecular signatures of each cluster, canonical gene expressions among the clusters (FIG. 8j) were compared, and calculated the scores of gene set variation analysis (GSVA) based on Reactome pathway database (FIG. 8k and FIG. 9d-f). DAM (-) LPS (-) microglia expressed canonical homeostatic microglial markers including CX3CR1 and P2RY12. This cluster significantly showed anti-inflammatory phenotype including TGF-beta, Wnt and BMP signaling pathways (FIG. 8j, k). DAM (-) LPS (+) showed higher expression of pro-inflammatory genes including TMEM163'8and genes associated with TLR4 downstream signaling pathways including TICAM1, CD14 and MYD88. Enriched pathways were related to interleukin signaling and TLR downstream signaling pathways (FIG. 8j). DAM (+) LPS (-) microglia mainly showed upregulation of classical DAM markersDocket No. 043846.00009 (PCT)

[0104] including APOE, AXL and TREM2. These genes were linked to elevation of cholesterol metabolic processes related to APOE and DAP12 signaling pathway related to TREM2 (FIG.

[0105] 8j). DAM (+) LPS (+) microglia not only shared molecular signatures with DAM (+) LPS (-) and DAM (-) LPS (+), but also expressed specific gene markers including ACSL1, NAMPT and DYPD, which are known as the canonical markers of lipid droplet associated microglia showing neurotoxicity in AD (FIG. 8j).

[0106] Ap cannot solely induce such microglial transformation and LPS could be a major trigger of toxic microglial transformation. It was confirmed that DAM (+) LPS (+) microglia highly expressed terms associated with pro-inflammatory effector state with lipid droplet turnover (FIG. 8k).

[0107] This dataset includes a quantitative local neuropathological-burden metric, termed the continuous pseudo-progression score (CPS)10. We next tested whether the abundance of the four states changed along CPS and AD neuropathological change (ADNC) score (FIG.

[0108] 8 l-m). The fraction of DAM (-) LPS (-) microglia remained largely stable along disease progression, whereas DAM (-) LPS (+) microglial fraction decreased with increasing disease burden. In contrast, DAM-positive states increased with progression. Within DAM, LPS-responsive microglia (DAM (+) LPS (+)) were consistently more than twice as abundant as LPS-nonresponsive DAM microglia (DAM (+) LPS (-)) across AD progression (FIG. 8l-m).

[0109] Taken together, the results indicate that LPS-responsive microglial signatures are enriched in early AD and may arise prior to canonical DAM-associated microglia. Moreover, the converged DAM (+) LPS (+) state is associated with a lipid droplet-linked pro-inflammatory effector program, confirming the role of LPS priming microglia besides Ap to manifest a distinct phenotype.

[0110] VSN-dominant LPS delivery to the brain

[0111] To determine the route of LPS delivery, immunostaining of LPS in dorsal medulla of ADLP mice was performed. This brain area contains NTS, DMVN, and area postrema (AP). NTS, especially the medial caudal part of NTS (mcNTS) is known to contain afferent nerve endings from the gut. DMVN is the soma of VMN, and AP is a proxy of hematogenous spread due to lack of BBB. Unexpectedly, mcNTS showed significantly high LPS signal intensity compared to lateral caudal part of the NTS (IcNTS), AP, and DMVN. (FIG. 10 a-b, and FIG. 11a). Further assessment of LPS in nodose ganglion (NG), the sensory ganglion of the vagus nerve. NG of AD mouse, compared to that of WT mouse, exhibited higher portion of neurons containing intracellular LPS signal (FIG. 10c-d).

[0112] It was also evaluated whether posterior subdiaphragmatic vagotomy reduced LPS transmission (FIG. 12a). AD mice (ADLPAPT) were either vagotomized (AV) or sham operated (AS), and ADLPwrmouse was sham operated (WT) at 3 months of age. Pathology wasDocket No. 043846.00009 (PCT)

[0113] assessed 4 months later (FIG. 10e). CCK-8 test proved reduced functional signaling after vagotomy (FIG.12b). Fluorogold tracer injection revealed reduced signal in right DMVN, indicating successful posterior vagal trunk resection (FIG. 12c-d).

[0114] Comparison of left and right mcNTS revealed no difference in LPS signal intensity in sham-operated mice (FIG. 12e, f), while the intensity was lower inside right mcNTS of vagotomized group, confirming the transmission of LPS via VSN (FIG. 10f, g).

[0115] Further immunostaining of LPS in hippocampus revealed extensive accumulation of LPS in the hippocampus of AD mouse, which was drastically reduced by vagotomy (FIG. 10h-i and FIG. 11b). Additional immunostaining using other antibodies validated the results (FIG. 12g-j). Together, the results suggest that VSN is the major route of LPS transmission to the brain.

[0116] Vagotomy reduces brain LPS and Ap burden

[0117] Considering significant impact of LPS in microglia among human AD brain, we aimed to investigate impact of blocking LPS transmission by vagotomy in AD mice. Cortical accumulation of amyloid plaques decreased by vagotomy (FIG. 13a, b). Quantitative measurement of Api-42 and 1-40 species via enzyme-linked immunosorbent assay (ELISA) revealed that both Api-42 and 1-40 species in RIPA insoluble fraction were reduced in the vagotomized group, without significant difference in the ratio of two species (Api-42 / 1-40) (FIG. 13c-e). Interestingly, Api-42 and 1-40 in RIPA soluble fraction showed no difference, indicating LPS-induced aggregation of A (FIG. 13f-h). On the other hand, vagotomy did not significantly reduce tau hyperphosphorylation (FIG. 14).

[0118] Since LPS is a well-known trigger of inflammation, it was also evaluated gliosis in hippocampus. Vagotomy significantly reduced microgliosis, while the extent of astrocytosis showed no difference (FIG. 13i-l). Lowered microgliosis might be the cause of decreased insoluble Ap in vagotomized group, regarding the role of microglia in aggregation of amyloid plaques. Altogether, our results suggest that reduced transmission of LPS by vagotomy mitigates microglia-mediated aggregation of amyloid plaques.

[0119] Upregulated axonal transport in AD VSN

[0120] Experiment was conducted to evaluate if single-nucleus RNA sequencing (snRNA-seq) of NG collected from six WT mice and five AD mice at 7 months of age to analyze transcriptional alterations in relation to LPS transmission (FIG. 15a and FIG. 15a-b). It was identified 11 cell types including NG by unsupervised Leiden clustering. (FIG. 15b-c and FIG.

[0121] 15c).

[0122] Next compared transcriptomes between AD and WT NG (FIG. 14d) . As a result, motor proteins, kinesins and dynein were upregulated in AD NG (FIG. 15d). Over-Docket No. 043846.00009 (PCT)

[0123] representation analysis of gene ontology (GO) terms and Reactome pathway gene sets (FIG. 15e-f) were analyzed. Upregulated terms in AD NG included pathways related to axonal transport bacterial toxin responses, Toll-like receptor (TLR) signaling, and tumor necrosis factor-a (TNF-a) signaling (FIG. 15e). Gene set enrichment analysis (GSEA) validated increased axonal transport and bacterial toxin uptake in AD NG (FIG. 15g). We applied CollecTRI-dervied regulon analysis to infer upstream regulators. Inflammatory cytokine-responsive transcription factors such as Rela and Jun were upregulated in AD NG (FIG. 15h).

[0124] TNF-a is a key cytokine linked to AD. Crohn’s disease, a condition driven by TNF-a, has been reported to increase AD risk by 7.67-fold, highlighting the clinical relevance of this cytokine. Consistent with this association, colonic TNF-a levels were elevated in AD mice (FIG. 15i). TNF-a has been reported to activate VSNs, while chronic sensory neuron activation increases motor protein expression. These findings support a model in which sustained gut inflammation activates VSNs, leading to elevated motor-protein levels that may facilitate LPS propagation in AD.

[0125] TNF-a accelerates LPS transport in VSN

[0126] The present disclosure utilized in vitro analysis of human iPSC-derived CO, VSGO, and BO in microfluidic chip system to validate cytokine-mediated increase of LPS transmission. CO, VSGO and BO were independently differentiated from a single human iPSC line, then seeded into a three-compartment microfluidic chip to assemble a gut-nerve-brain axis-on-a-chip (FIG. 17a-b and FIG. 18). The disclosure confirmed that diffusion between compartments was negligible using FITC-dextran (40kDa) (FIG. 19a, b). It was also assessed connectivity between compartments. Field potential recording showed increase in normalized band power and coherence in BO upon D-glucose or bile-acid treatment in the CO compartment, confirming functional connectivity between the organoids (FIG. 19c-l, FIGs. 21 and 22). Trans-synaptic propagation of EGFP-tagged pseudorabies virus (PRV) from infected CO to BO confirmed anatomical connectivity between organoids (FIG. 19m-p).

[0127] Calcium imaging revealed increased VSN activity after treating TNF-a in the CO chamber (FIG. 17c-f). Subsequently, we assessed difference in propagation of LPS via VSGO upon TNF-a treatment. Since VSNs transport LPS efficiently when packaged in outer membrane vesicles (OMVs), we delivered Alexa Fluor 488-tagged E. coli-derived OMVs to the colon compartment with or without TNF-a (FIG. 17g-h). OMV fluorescence signal intensity was significantly elevated 3 days after co-treatment with TNF-a (FIG. 17p-q). In addition, immunostaining at day3 revealed increased OMV puncta within VSGO neuritesDocket No. 043846.00009 (PCT)

[0128] upon TNF-a treatment (FIG. 17r-s). Together, these results confirm that gut-originated TNF-a accelerates propagation of LPS from gut to brain via VSN.

[0129] Gut VSNs dominate LPS uptake

[0130] Spatial transcriptomics with TUJ1, LPS, and DAPI immunostaining was applied to nodose ganglia from 7-month-old AD and WT mice (n = 7 per group) to identify VSN subtypes mediating LPS transport. 2-micrometer spatial bins were merged into single-cell resolution using bin2cell, enabling quantification of LPS fluorescence at the cellular level (FIG. 20a). After cell binning and quality control, a total of 29,636 cells were retained (AD, 17,584; WT, 12,052). Cell identities were first inferred through marker-guided supervised classification, after which remaining populations were subsequently resolved using the semisupervised algorithm ResolVI (FIG. 20b-e, FIG. 29a-c, FIGs. 24 and 25).

[0131] The dataset was integrated with previously published VSN atlas using CellHint for subtype annotation (FIG. 20f-h). Using established organ-specific marker genes from a previous study, we calculated projection scores for gut, heart, lung, and pancreas, which stratified the 17 NG subtypes by their respective organ associations (FIG. 20i-). We assessed which VSN subtypes preferentially transport LPS. LPS fluorescence intensity was partitioned into high (top 20%) and low (bottom 20%) groups (FIG. 20k, I). Most LPS-high cells (90.6%) corresponded to gut-projecting VSNs (FIG. 20m). We computed the ratio of LPS-high to LPS-low cells; gut-projecting NG17 and NG13 had the highest values, 1.70 and 1.62, respectively (FIG. 20n). Notably, NG17, showing the strongest LPS enrichment, corresponds to a gut mucosa-innervating VSN subtype characterized by high Gpr65 expression (FIG. 20h). Together, these results indicate that gut- projecting VSNs, particularly mucosa-innervating neurons, constitute the principal LPS-positive population within the nodose ganglion.

[0132] Identifying mediators of LPS endocytosis

[0133] Differential expression analysis was performed within gut-projecting VSNs to investigate the mechanisms underlying LPS uptake into these neurons. Compared with WT, AD gut VSNs exhibited a distinct transcriptional profile. Likewise, contrasting LPS-high with LPS-low gut VSNs revealed genes enriched in the LPS-high population (adjusted P < 0.05; log2fold change > 0.3) (FIG. 26a). It was identified 100 genes shared by AD-associated DEGs and genes enriched in LPS-high gut VSNs (FIG. 26b). Over-representation analysis using GO terms and Reactome pathway revealed trafficking- related processes, including vesicle formation and fusion, axo-dendritic transport, and TLR signaling (FIG. 26c).

[0134] From the 100-gene overlap, we selected eight genes mapped to the GO Cellular Component term “cell surface” (G0:0009986) as candidates for surface molecules mediatingDocket No. 043846.00009 (PCT)

[0135] LPS uptake (FIG. 26d). Among these, APP and ANXA2 have been reported to bind heparin, a sulfated glycosaminoglycan that shares key features with the LPS O-antigen. We evaluated the significance and expression differences for the overlapping genes across conditions (AD to WT; LPS-high to LPS-low). App showed significant upregulation in both comparisons (FIG. 26e, and FIG. 26f-g). Moreover, App expression was particularly enriched in NG17 subtype, which exhibited the highest LPS-high / LPS-low ratio (FIG. 26f). Consistent with these findings, spatial mapping showed an overlapping of LPS signal intensity and App expression within the nodose ganglion (FIG. 26g).

[0136] The present disclosure modeled protein-LPS complexes with AlphaFold 3 to evaluate potential interactions. As the present disclosure postulated that OMV-bound LPS enters VSN, we chose the extracellular R1 core of E. coli LPS, due to high variability of O-antigen. The present disclosure modeled the extracellular domains of the candidates, APP, CLSTN1, TIMP2, NGFR and ANXA2. HCN1 was excluded as it lacked extracellular domain. Experimentally determined protein structures orAlphaFold2 models of the extracellular domain sequences were used. (For APP, experimentally determined structure of respective E1 and E2 domains; for CLSTN1, AlphaFold 2 model; for NGFR, AlphaFold2 model predicted with PDB 3BUK chain C template) (FIG. 27).

[0137] Protein-LPS complex structure predictions were evaluated by AlphaFold3 interchain predicted TM-score (ipTM) and the minimum of the mean predicted aligned error (pAE) between chains. APP E2 domain showed the highest confidence AF3 interface predictions (ipTM = 0.91 ; min mean pAE = 1.57 A; FIG. 26h, I and FIG. 27). LPS R1 outer core was predicted to interact with the part of the E2 domain proximal to the E1 domain (FIG. 26j, FIG.

[0138] 27). Together, the transcriptional data and complex structure prediction data suggest APP as a possible mediator of VSN endocytosis in AD.

[0139] The experimental results show that VSN delivers LPS from the gut to the brain in AD. Though prior work has demonstrated that OMV-bound LPS can travel along the vagus nerve7, our work further specified the functional component of the vagus nerve as the main route of transmission. Vagotomy reduced propagation of LPS, attenuating microgliosis and Ap aggregation without significant difference in tauopathy. Although our model exhibits tauopathy by 7 months of age, high-molecular-weight tau and neuronal death were prominent only after 10 months. Given that amyloidopathy typically precedes tauopathy in AD, our experimental condition likely reflects the relatively early pathological stage of the disease. Considering that serum-CSF LPS-level disparity has been reported at the mild cognitive impairment stage, use of relatively early-stage mouse model for delineating the entry route of LPS in AD would be clinically relevant.Docket No. 043846.00009 (PCT)

[0140] Analysis of human brain single-cell transcriptomic data further implicates LPS-associated signatures in the AD microglial landscape. Converging with previous observations that LPS immunoreactivity is predominantly co-localized with microglia, the analyses suggest that LPS-response signatures are detectable in microglia at early points along AD progression. Those LPS-responsive microglia may progressively acquire DAM-like traits as disease advances. Microglia co-expressing DAM and LPS signatures (DAM (+) LPS (+)) were more abundant than DAM microglia lacking LPS response (DAM (+) LPS (-)) across all disease stages, and showed additional shifts in immune and lipid-metabolic programs.

[0141] At the signaling level, MYD88, a key downstream adaptor of TLR4, was upregulated not only in LPS(+) microglia but also in DAM(+) LPS(-) microglia, consistent with TLR4 activation by multiple ligands including Ap. By contrast, TICAM1 (also known as TRIF), which mediates the MYD88-independent TLR4 signaling, was selectively upregulated in LPS (+) microglia, suggesting that LPS may engage a specific downstream TLR4 signaling distinct from the response to other AD-associated stimuli. Moreover, LPS(+) microglia showed higher scores for the TLR1 / TLR2 cascade, a pathway that can be activated by bacterial OMVs, supporting the contribution of bacterial factors to this microglial state.

[0142] Preferential LPS transmission in VSNs over VMNs would be attributed to their anatomical difference. VSNs terminate broadly across mucous and muscular layers and within the enteric nervous system (ENS), whereas VMNs are thought to synapse primarily with ENS neurons. Thus, LPS would enter VSNs directly through mucosal endings, whereas VMNs would receive LPS only indirectly via ENS; consequently, OMV entry would be favored in VSNs.

[0143] The present disclosure demonstrated that gut inflammatory cytokines in AD, especially TNF-a, activate VSNs and increase expression of intracellular motor proteins, thereby accelerating transport of OMV-bound LPS. Our previous work showed that inflammatory monocytes expand in AD-model mice and decline after fecal microbiota transfer. Considering that tissue-resident macrophages are the major source of TNF-a, we infer that AD-associated inflammatory macrophages in the gut secrete TNF-a, which in turn enhances VSN-mediated LPS delivery to the brain.

[0144] VSNs are known to express TLR4 and respond to LPS. Our transcriptomic profiling of LPS-positive VSNs revealed upregulation of TLR4-downstream pathways without a significant difference in TLR4 transcript abundance. In immune cells, TLR4 functions as a pattern-recognition receptor; cargo internalized via TLR4 is typically routed to lysosomes. Accordingly, OMVs taken up through TLR4 may be preferentially degraded, whereas OMVs internalized through alternative routes may undergo intracellular trafficking through theDocket No. 043846.00009 (PCT)

[0145] neuron. Moreover, TLR4 does not bind LPS directly; rather, MD-2 binds the lipid A moiety of LPS and then associates with TLR4. Thus, TLR4 / MD-2 complex is more likely to engage in the endocytosis of membrane-free LPS, whereas intact OMV uptake into VSNs probably proceeds largely via TLR4-independent mechanisms.

[0146] For the candidate of TLR4-independent mechanism, APP is a potential mediator of OMV-bound LPS delivery in VSN, supported by comprehensive approach of spatial transcriptomics and AlphaFold3-based protein co-folding analysis. APP is present at both presynaptic and postsynaptic neuronal compartments, where it modulates synaptic structure and function beyond its role as an Ap precursor. Recent evidence indicates prominent clathrin-independent uptake of APP into neurons. This raises the possibility that APP-dependent internalization may evade lysosomal degradation pathway, unlike TLR4-mediated endocytosis. Interestingly, CLSTN1 interacts with APP and couples it to axonal transport machinery through kinesin-1-associated adaptors. Thus, the present disclosure provides a working model in which APP binds to the outer core of OMV-associated LPS at the postsynaptic surface, then internalized via a clathrin-independent route and subsequently transported along axons via CLSTN1 -dependent transport. This model provides a mechanistic link between APP-mediated endocytosis and trafficking of LPS throughout VSN.

[0147] Compared to prior studies focusing mainly on the consequences of brain LPS deposition in AD, we delineated a gut-to-brain entry route for LPS and cross-validated it in vivo and in vitro. Also, whereas most studies have emphasized physiological roles of visceral sensory neurons and their responses to injury or inflammation, we highlight a pathological role of VSNs and characterize their disease-associated state at a single-cell level.

[0148] Technically, we established a novel three-compartment axis-on-a-chip that connects a peripheral organ to the central nervous system via sensory neurons, enabling controlled experiments to dissect gut-vagus-brain signaling. To our knowledge, we performed the first spatial transcriptomic analysis of visceral sensory neurons, integrating bacterial LPS detection with spatial gene expression to achieve detailed mapping.

[0149] Even though the present disclosure did not visualize AD-associated LPS transport in human tissue, it is confirmed that VSN-mediated LPS uptake on a human iPSC-derived gut-nerve-brain axis-on-a-chip, which helped addressing the technical challenges of sampling human nodose ganglia.

[0150] The data in the present disclosure provides that VSNs in the vagus nerve transport OMV-bound LPS from the gut to the brain in AD, which exacerbates amyloidopathy and neuroinflammation. In AD human brain, presence of LPS and bacterial OMV responsive microglia were expected even in early AD, which gains DAM signature among AD progression. Gut-derived TNF-a stimulates VSNs and upregulates expression of motorDocket No. 043846.00009 (PCT)

[0151] proteins, accelerating LPS transport. APP, especially the E2 domain, likely binds to LPS and promotes endocytosis into VSN. Together, these findings identify a modifiable gut-vagus-brain conduit for endotoxin trafficking that links intestinal dysbiosis to neuroinflammation and amyloid pathology early in AD (FIG. 30).

[0152] The present disclosure provides that a similar molecule of therapeutics can be delivered as found to an EV from gut to brain without going through BBB.

[0153] Embodiments of the present disclosure also include:

[0154] Embodiment 1. A composition comprising an extracellular vesicle (EV) and a therapeutic agent, wherein the EV is a bacterial EV and the therapeutic agent is a drug for treating a disease in the brain.

[0155] Embodiment 2. The composition of Embodiment 1, wherein the disease is selected from the group consisting of a neurodegenerative disorder, a brain tumor, a brain cancer, a neurological disorder, and a psychiatric disorder.

[0156] Embodiment 3. The composition of Embodiment 1, wherein the disease is selected from the group consisting of Alzheimer's Disease (AD), Parkinson's Disease (PD), Amyotrophic Lateral Sclerosis (ALS), Huntington's Disease (HD), Dementia with Lewy Bodies (DLB), Frontotemporal Dementia (FTD), Prion Diseases, Spinocerebellar Ataxia (SCA), spinal muscular atrophy (SMA), multiple system atrophy (MSA), progressive supranuclear palsy (PSP), chronic traumatic encephalopathy (CTE), cognitive dysfunction, motor dysfunction, and autonomic dysfunction.

[0157] Embodiment 4. The composition of Embodiment 1,

[0158] wherein the therapeutic agent is selected from the group consisting of Carmustine, Doxorubicin, Paclitaxel, Temozolomide, L-DOPA, Donepezil, Insulin, Valproic acid, Midazolam, Diazepam, Ketamine, Brexanolone, DNL310, viral vectors, genes, siRNA, peptide, and polysaccharides; or

[0159] wherein the therapeutic agent is a drug for treating a disease in the brain.

[0160] Embodiment 5. A method for delivering a therapeutic agent to brain using a composition comprising:

[0161] administering a composition of Embodiment 1 to a subject in need thereof, wherein the composition is delivered through visceral sensory neurons (VSN).Docket No. 043846.00009 (PCT)

[0162] Embodiment 6. The method of Embodiment 5, where in the composition is administered orally.

[0163] Embodiment 7. An organoid on chip model comprising at least two compartments axis for testing delivery of the composition of Embodiment 1 from gut to brain, wherein the compartments comprise, brain, gut, or nerve.

[0164] EXAMPLES

[0165] Those of ordinary skill in the art will realize that the following description is illustrative only and is not intended to be in any way limiting. Other aspects will readily suggest themselves to those skilled in the art having the benefit of this disclosure. Reference will now be made in detail to implementations of the example aspects as illustrated in the accompanying drawings. The same reference indicators will be used to the extent possible throughout the drawings and the following description to refer to the same or like items.

[0166] Example 1. Methods

[0167] Example 1-1. Mouse model used for animal experiments

[0168] For the animal experiments, the present disclosure utilized the ADLP mouse model, which was established in a previous study10. This model exhibits both Ap and tau pathology, making it a suitable representation of Alzheimer’s disease. The A pathology was introduced by inserting human APP and PSEN1 genes, each containing multiple mutations. Specifically, the APP gene harbors four mutations (K670N, M671L, 1716V, V717I), while the PSEN1 gene carries two mutations (M146L, L286V). These transgenes the present disclosure re placed under the control of the Thy1 promoter, ensuring their expression exclusively in the central nervous system (CNS).

[0169] For tau pathology, the human MAPT gene carrying the P301 L mutation was introduced. Like APP and PSEN1, this transgene was designed to be expressed only in the CNS using the PrP promoter. Two types of mice the present disclosure re used in the experiments: wild-type control mice without any genetic modifications and ADLPAPT mice carrying these mutations.

[0170] Example 1-2. Subdiaphragmatic vagotomy

[0171] For the posterior subdiaphragmatic vagotomy, mice the present disclosure re fasted from the day before surgery. Although the procedure could be performed without fasting, fasting was preferred to reduce intestinal movement, which could otherwise interrupt the surgical field.Docket No. 043846.00009 (PCT)

[0172] On the day of surgery, mice the present disclosure re anesthetized using inhalation anesthesia. The abdominal fur was completely shaved with an electric razor before making the incision. During anesthesia, if the mouse exhibited labor breathing, it indicates overexposure to anesthesia and respiratory suppression, requiring an adjustment of anesthesia intensity before proceeding with the surgery. The limbs the present disclosure re secured with tape, and a heating pad was placed beneath the mouse to prevent hypothermia during the procedure.

[0173] A 2 cm midline incision was made from just below the xiphoid process using forceps and surgical scissors, carefully avoiding damage to internal organs. The muscle and peritoneal layers the present disclosure re incised in the same manner to expose the abdominal cavity.

[0174] To maximize exposure of internal organs, four bulldog clamps the present disclosure re placed at the upper left, lower left, upper right, and lower right positions, securing the peritoneum and skin together. After exposure, the small intestine was carefully externalized from the abdominal cavity and covered with PBS-soaked gauze to prevent dehydration. Additional PBS-moistened gauze was strategically placed to optimize surgical field. Gauze was positioned behind the stomach to facilitate separation from the spleen and improve gastric exposure. Another piece was placed between the diaphragm and liver to shift the liver-spleen-stomach complex downward, ensuring clear visualization of the esophagus in the surgical field. Finally, gauze was placed under the median lobe of the liver to enhance gastric visibility.

[0175] A hemostat was used to stabilize the forestomach, the region of the stomach with the least blood flow, minimizing bleeding. The fibrous tissues between the stomach and spleen, as well as between the stomach and liver, the present disclosure re carefully dissected to expose the posterior vagal trunk. The stomach was then gently rotated leftward to fully reveal the posterior vagal trunk, which runs along the posterior aspect of the esophagus.

[0176] The posterior vagal trunk was identified and carefully severed using fine microscissors, ensuring no damage to surrounding esophageal tissue. In the sham operation, the nerve was only exposed without cutting.

[0177] After vagotomy, the intestines the present disclosure re repositioned in their original locations within the abdominal cavity. The gauze was removed, and the surgical site was inspected for bleeding. The peritoneal layer was first closed with absorbable sutures, spaced 5 mm apart. The muscle, subcutaneous, and skin layers the present disclosure re then closed together using non-absorbable nylon sutures, with 5-10 mm spacing to ensure uniform closure.Docket No. 043846.00009 (PCT)

[0178] Following surgery, the mouse was placed in a cage with a heating pad until it fully recovered from anesthesia. To prevent corneal injury, bedding materials (e.g., sawdust) the present disclosure re not used in the recovery cage.

[0179] Example 1-3. Extraction of mouse brain tissue

[0180] To extract brain tissue, mice the present disclosure re first anesthetized, and cardiac perfusion was performed by making an incision in the right atrium using fine micro-scissors while cold PBS was continuously infused into the left ventricle. After completing the perfusion, the brain was carefully extracted for further analysis.

[0181] For immunohistochemical staining, the extracted brain tissue was fixed in 4% paraformaldehyde (PFA) at 4°C for 24 hours. For other analyses, the brain was rapidly frozen using liquid nitrogen and stored at -80°C until further use.

[0182] Example 1-4. Maintenance of iPSCs

[0183] iPSCs the present disclosure re maintained on a 6-the present disclosure II cell culture plate coated with hESC-qualified Matrigel (Corning, 354277) and cultured in mTeSR™ Plus media (STEMCELL Technologies, 110-1130). When the cells reached approximately 80% confluency, they the present disclosure re detached using ReLeSR (STEMCELL Technologies, ST05872) and passaged at a 1:20 ratio. Only iPSCs with fewer than 30 passages the present disclosure re used.

[0184] After thawing frozen cells, at least one passage was performed before initiating VSGO differentiation. iPSC quality was monitored daily under a microscope, and only healthy cells the present disclosure re selectively maintained and passaged as needed. Bacterial or fungal contamination was checked daily via microscopic observation, and Mycoplasma contamination was periodically tested using MycoStrip (Invivogen, rep-mys-50).

[0185] Example 1-5. Induction of EpP Differentiation

[0186] EpP differentiation was performed using an Aggrewell™80024-the present disclosure II Plate (STEMCELL Technologies, 34815, hereinafter referred to as "Aggrewell"). Prior to differentiation induction, each well of the Aggrewell plate was lubricated with 500 pl of Anti-Adherence Rinsing Solution (STEMCELL Technologies, 07010). After lubrication, 2 ml of DMEM / F-12 (Gibco, 10565018) was added to each the present disclosure II and maintained until cell seeding.

[0187] To form embryoid bodies (EBs) within Aggrewell, iPSCs the present disclosure re first washed by adding 2 ml of Dulbecco’s phosphate buffered saline (DPBS, Gibco, 14190144) per the present disclosure II and then aspirating it to remove cell debris. Next, to detach the cells from the culture plate, 20 pM Y-27632 (STEMCELL Technologies, ST72304) andDocket No. 043846.00009 (PCT)

[0188] Accutase (Sigma, A6964) the present disclosure re added and incubated for 5 minutes. The detached cells the present disclosure re gently pipetted to achieve single-cell dissociation. The number of cells was then quantified to ensure 50,000 iPSCs per the present disclosure II, and the cells the present disclosure re resuspended in EB Formation Medium (STEMCELL Technologies, 05893, hereinafter referred to as "EB media") supplemented with 20 pM Y-27632 before being seeded into Agg rewell (-Day 2). Before cell seeding, the preexisting DMEM / F-12 in the wells was completely removed.

[0189] On Day -1, 1.5 ml of EB media without Y-27632 was replaced in each the present disclosure II. Until Day 7, the experiment was conducted according to previously established protocols.

[0190] On Day 0, the medium was changed to Chemically-defined Differentiation Medium (CDM), consisting of a 1:1 mixture of Ham's F12 GlutaMax (Gibco, 31765-035) and IMDM GlutaMax (Gibco, 31980-030), supplemented with Chemically-Defined Lipid (Gibco, 11905-031) 1X, BSA (Gibco, 15260037) 5 mg / ml, insulin (Sigma, I9278) 10 mg / ml, transferrin (Sigma, T8158) 20 mg / ml, 1 -thioglycerol (Sigma, M6145) 450 pM, and Normocin (Invivogen, Ant-nr-1) 100 pg / ml. Additionally, FGF-2 (R&D Systems, 233-FB-025) 4 ng / ml, SB-431542 10 pM, and BMP-4 (Peprotech, AF-120-05ET) 2.5 ng / ml the present disclosure re added. On Day 2, 1.5 ml of the 2 ml medium was replaced with fresh CDM containing the same components.

[0191] On Day 4, partial medium replacement was performed with CDM supplemented with FGF-250 ng / ml and LDN193189 (Tocris Bioscience, 6053) 200 nM, which was repeated on Days 6 and 7.

[0192] On Day 8, the medium was changed to Epibranchial Placode Maturation Medium (EPMM), which consists of a 1:1 mixture of Advanced DMEM / F-12 (Gibco, 12634010) and Neurobasal Medium (Gibco, 21103-049), supplemented with N2 supplement (Gibco, 17502-048) 0.5X, B27 without vitamin A supplement (Gibco, 12587-010) 0.5X, GlutaMax (Gibco, 35050-079) 1X, p-mercaptoethanol (Gibco, 21985-015) 0.1 mM, and Normocin 100 pg / ml. Additionally, FGF-3 (R&D Systems, 1206-F3-025 / CF) 50 ng / ml, FGF-10 (R&D Systems, 345-FG-025 / CF) 50 ng / ml, BMP-420 ng / ml, and IWR-1 (Sigma, I0161-5MG) 10 pM the present disclosure re added. The same medium was partially replaced daily until Day 14.

[0193] The composition of CDM is detailed in Table 1, and the composition of EPMM is detailed in Table 2

[0194] Table 1. Chemically-defined Differentiation Medium (CDM)

[0195] Component Supplier Catalog no Stock cone Final coneDocket No. 043846.00009 (PCT)

[0196] Ham’s F12 Glutamax Gibco 31765-035 - 49% (v / v) IMDM GlutaMax Gibco 31980-030 - 49% (v / v) Chemically-Defined

[0197] Gibco 11905-031 100x 1x

[0198] Lipid

[0199] BSA Gibco 15260037 75mg / ml 5mg / ml Insulin Sigma I9278 10mg / ml 7ug / ml Transferrin Sigma T8158 20mg / ml 15ug / ml

[0200] 1 -thioglycerol Sigma M6145 11.5M 450uM Normocin Invivogen Ant-nr-1 50mg / ml 100ug / ml

[0201] Table 2. Epibranchial Placode Maturation Medium (EPMM)

[0202] Component Supplier Catalog no Stock cone Final cone Adv DMEM / F12 Gibco 12634010 - 49% (v / v) Neurobasal Gibco 21103-049 - 49% (v / v)

[0203] N2 supplement Gibco 17502-048 100x 0.5x

[0204] B27-Vitamin A Gibco 12587-010 50x 0.5x GlutaMAX Gibco 35050-079 100x 1x Mercaptoethanol Gibco 21985-015 55mM 0.1mM Normocin Invivogen Ant-nr-1 50mg / ml 100ug / ml

[0205] Example 1-6. Differentiation of VSGOs

[0206] On Day 15, EpP aggregates cultured in Aggrewell™ the present disclosure re transferred to 1.5 ml tubes lubricated with 0.1% bovine serum albumin (BSA) in phosphate-buffered saline (PBS). Both the tubes and pipette tips the present disclosure re prelubricated with the BSA solution to minimize cell adhesion.

[0207] For VSGO culture, the EpP aggregates the present disclosure re directly seeded onto PLO / laminin-coated plates (Corning, 356231). In a 24-the well plate, 3-5 EpP aggregates the present disclosure re seeded per well, while in a 12-well plate, 7-10 aggregates the present disclosure re seeded per the present disclosure II.

[0208] After EpP seeding, nodose ganglion Differentiation Medium (NDM) was used for further culture. The NDM was supplemented with 20 pM Y-27632, 10 pM IWR-1, 10 pM DAPT (Sigma, D5942), 20 ng / ml BMP-4, 50 ng / ml FGF-3, 50 ng / ml FGF-10, 50 ng / ml BDNF (Peprotech, 450-02), 50 ng / ml GDNF (Peprotech, 450-10), and 20 ng / ml NGF-p (Peprotech, 450-01).Docket No. 043846.00009 (PCT)

[0209] On the following day, the medium was completely replaced with the same NDM composition, except without Y-27632. Afterward, the culture medium was partially changed every 2-3 days:

[0210] 12-well plate: 500 pl of the total 1.5 ml per well was replaced.

[0211] 24-well plate: 300 pl of the total 1 ml per well was replaced.

[0212] Additionally, 2 pg / ml laminin (Sigma, L2020) was added every second medium change.

[0213] The composition of NDM is detailed in Table 3.

[0214] Table 3. The Composition of Nodose Ganglion Differentiation Medium (NDM) Component Supplier Catalog no Stock cone Final cone Neurobasal Gibco 21103-049 - 99% (v / v)

[0215] N2 supplement Gibco 17502-048 100x 1x

[0216] B27-Vitamin A Gibco 12587-010 50x 1x

[0217] GlutaMAX Gibco 35050-079 100x 1x Mercaptoethanol Gibco 21985-015 55mM 0.1mM Normocin Invivogen Ant-nr-1 50mg / ml 100ug / ml

[0218] Example 1-7. Differentiation of HCO

[0219] The differentiation of human colon organoids (HCOs) was performed using a modified version of a previously established protocol. In brief, 200,000 iPSCs the present disclosure re evenly seeded per well onto a 6-the present disclosure II cell culture plate coated with hESC-qualified Matrigel. The cells the present disclosure re maintained in mTeSR+ medium, with media changes every two days until they reached 80% confluency.

[0220] Once the cells reached 80% confluency, the medium was replaced with RPMI 1640 medium (Gibco, 11875093) supplemented with 2 mM GlutaMAX (Gibco, 35050079), 100 ng / ml Activin A (Cell Guidance Systems, GFH6), 3 pM CHIR99021 (Tocris Bioscience, 4423), and 100 ll / rnl Penicillin-Streptomycin (Pen-Strep, Sigma, P4333) (Day 1).

[0221] On Days 2 and 3, the medium was replaced with RPMI 1640 medium containing 2 mM GlutaMAX, 100 ng / ml Activin A, 0.2% BSA (Gibco, 15260037), and 100 ll / rnl Pen-Strep.

[0222] From Day 4 to Day 7, daily medium changes the present disclosure re performed using RPMI 1640 medium supplemented with 2 mM GlutaMAX, 1X B27 supplement (Gibco, 12587-010), 3 pM CHIR99021, 500 ng / ml FGF-4 (Peprotech, 110-31), and 100 U / ml Pen-Strep.

[0223] From Day 8 onward, the medium was changed every two days to Advanced DMEM / F12 (Gibco, 12634010) supplemented with 2 mM GlutaMAX, 1X B27 supplement, 3Docket No. 043846.00009 (PCT)

[0224] pM CHIR99021, 100 ng / ml EGF (Peprotech, AF-100-15), 300 nM LDN193189, and 100 U / ml Pen-Strep.

[0225] After Day 20, during each medium change, floating cell spheroids in the medium the present disclosure re carefully collected using a pipette, mixed with Matrigel, and plated into either 24-well cell culture plates or microfluidic chips. For 24-well plates, 40 pl of Matrigel containing spheroids was seeded per the present disclosure II. For microfluidic chips, 10 pl of Matrigel containing spheroids was seeded per chip.

[0226] Following seeding, the medium was replaced with the previously described Advanced DMEM / F12 medium containing growth factors, and medium changes continued every two days until Day 30.

[0227] Finally, three days before the final experiment, CHIR99021 was removed from the Advanced DMEM / F12 growth factor medium, and media changes continued without CHIR99021 until the experiment was initiated.

[0228] Example 1-8. Differentiation of HBOs

[0229] The differentiation of human brain organoids (HBOs) was conducted using a previously established protocol.

[0230] To initiate HBO differentiation, embryoid bodies (EBs) the present disclosure re first formed in Aggrewell, following the same method as in EpP differentiation. However, the seeding density was increased to 2,000,000 iPSCs per the present disclosure II. On the day of seeding, the cells the present disclosure re cultured in EB media supplemented with 20 pM Y-27632. On the following day, which is Day -1, the medium was replaced with EB media without Y-27632.

[0231] On Day 0, the medium was switched to iPSC media supplemented with 10 pM Dorsomorphin (Merck, P5499) and 10 pM SB-431542. The iPSC media consisted of DMEM / F12 with GlutaMAX (Gibco, 10565-018), 20% KnockOut Serum Replacement (Gibco, A3181501), 1% MEM Non-Essential Amino Acids Solution (Gibco, 11140050), 0.1 mM Mercaptoethanol, and 100 U / ml Penicillin with 100 pg / ml Streptomycin. This medium was changed daily until Day 3.

[0232] On Day 4, the EBs the present disclosure re transferred from Aggrewell to a Petri dish, ensuring that they remained separate and did not aggregate with each other. On Day 5, the individual EBs the present disclosure re transferred into a 96-the present disclosure II ultra-low-attachment microplate (Corning, 7007). From this stage onward, the medium was switched to Neurobasal-A Medium (Gibco, 10888-022) supplemented with B-27 minus vitamin A, 100 U / ml Penicillin, 100 pg / ml Streptomycin, GlutaMAX, and 0.5% (v / v) Matrigel Basement Membrane Matrix (Corning, 354234). Growth factors the present disclosure re added at different time points according to the following schedule.Docket No. 043846.00009 (PCT)

[0233] From Day 5 to Day 13, 20 ng / ml EGF and 20 ng / ml FGF-2 the present disclosure re added to the medium, and daily medium changes the present disclosure re performed. From Day 14 to Day 22, the same medium composition was used, but media changes the present disclosure re performed every two days instead of daily. From Day 23 to Day 40, EGF and FGF-2 the present disclosure re replaced with 20 ng / ml BDNF and 20 ng / ml Neurotrophin-3 (NT-3, Peprotech, 450-03), and the medium was changed every two days. After Day 40, the medium was changed every four days, and no additional growth factors the present disclosure re included.

[0234] Example 1-9. Immunocytochemistry of VSGOs

[0235] VSGOs were cultured on PLO / laminin-coated slide glasses within a 24-well cell culture plate. For immunocytochemistry, the culture medium was removed, and the samples the present disclosure re washed three times with PBS. The samples the present disclosure re then fixed overnight at 4°C with 4% paraformaldehyde (PFA) in PBS. After fixation, the wells the present disclosure re washed three times with PBS.

[0236] The blocking-permeabilization solution (BPS) was prepared by mixing 0.3% Triton X-100, 5% Normal horse serum (Vector, S-2000-20), and 0.5 mg / ml BSA in PBS. The samples the present disclosure re incubated in BPS overnight at room temperature. After blocking, the samples the present disclosure re incubated at 4°C for 72 hours with primary antibodies diluted in BPS. Following primary antibody incubation, the wells were washed three times with PBS-T (PBS containing 0.3% Triton X-100). The secondary antibody and DAPI the present disclosure re then diluted in PBS containing either 5% Normal horse serum or 5% Normal goat serum (Vector, S-1000), and the samples the present disclosure re incubated at 4°C for 72 hours. After secondary antibody incubation, the wells the present disclosure re washed four times with PBS-T. The samples the present disclosure re then mounted onto slide glasses and prepared for fluorescence imaging.

[0237] Example 1-10. Immunocytochemistry in a Microfluidic Chip

[0238] For immunocytochemistry within the microfluidic chip, collagen components the present disclosure re removed by first removing the culture medium and injecting 50 pl of Cell Recovery Solution (Corning, 354253) into each chamber. The chips the present disclosure re then incubated on an orbital shaker at 4°C for 2 hours. After incubation, the Cell Recovery Solution was removed, and the chambers the present disclosure re washed three times with PBS. Each washing step involved incubation in PBS at room temperature on a shaker for 10-15 minutes. To optimize the cleaning process, different volumes of PBS the present disclosure re injected into the VSGO and HCO chambers to create a pressureDocket No. 043846.00009 (PCT)

[0239] gradient. When 50 l of PBS was injected into the VSGO chamber, 20 pl was injected into the HCO chamber and vice versa.

[0240] Following the washing steps, 4% paraformaldehyde (PFA) was injected into the chambers at 50 pl for the VSGO chamber and 20 pl for the HCO chamber, and the chips the present disclosure re incubated at 4°C on an orbital shaker for 24 hours. After fixation, the samples the present disclosure re washed three times with PBS while maintaining the pressure gradient, with 1-hour intervals between each wash.

[0241] After washing, 50 pl of BPS was added to the VSGO chamber and 20 pl to the HCO chamber, and the chips the present disclosure re incubated at 4°C on an orbital shaker for 24 hours. The BPS was then removed and replaced with primary antibody-containing BPS at 50 pl for the VSGO chamber and 20 pl for the HCO chamber, followed by incubation at 4°C for 72 hours on an orbital shaker. After primary antibody incubation, the samples the present disclosure re washed three times with PBS-T while maintaining the pressure gradient, with 1-hour intervals between each wash.

[0242] For secondary antibody staining, the samples the present disclosure re incubated with PBS containing either 5% Normal horse serum or 5% Normal goat serum (Vector, S-1000), along with the secondary antibody and DAPI at 4°C for 72 hours on an orbital shaker. After secondary antibody incubation, the samples the present disclosure re washed three times at room temperature with 1-hour intervals. The final washing step was performed at 4°C for 24 hours on an orbital shaker, after which fluorescence imaging was performed.

[0243] Example 1-11. Immunohistochemical Staining

[0244] For immunohistochemical staining, fixed brain tissues the present disclosure re incubated in 30% sucrose solution for 72 hours to facilitate dehydration. After dehydration, tissues the present disclosure re stored at -80°C until further use. For staining, the brain tissues the present disclosure re sectioned into 25 pm-thick slices and processed for immunohistochemical analysis.

[0245] The tissue sections the present disclosure re placed in Transwell plates containing PBS and washed three times with PBS. For A and tau staining, additional processing was performed by incubating the tissues in 70% formic acid for 20 minutes to enhance staining efficiency. Other staining procedures the present disclosure re carried out without formic acid treatment. The sections the present disclosure re then incubated in blocking-permeabilization solution (BPS), as described in the immunocytochemistry protocol, at room temperature on a shaker for 1 hour. After blocking, the solution was replaced with primary antibody diluted in BPS, and the tissues the present disclosure re incubated for 20 hours at 4°C on a shaker.Docket No. 043846.00009 (PCT)

[0246] Following primary antibody incubation, the sections the present disclosure re washed three times with PBS at room temperature. The solution was then replaced with PBS containing 5% Normal horse serum or 5% Normal goat serum (Vector, S-1000) along with the secondary antibody, and the tissues the present disclosure re incubated for 1 hour at room temperature. After secondary antibody incubation, the sections the present disclosure re washed three more times with PBS before replacing the solution with DAPI diluted at 1:5000 in PBS. The sections the present disclosure re incubated in DAPI solution for 15 minutes at room temperature on a shaker, followed by an additional three PBS washes. Finally, the stained tissues the present disclosure re mounted onto slide glasses, covered with coverslips, and prepared for imaging.

[0247] For staining, the following primary antibodies were used in addition to those previously mentioned in the immunocytochemistry protocol:

[0248] AT8 (Thermo Fisher, MN 1020, Mouse, 1:200)

[0249] Biotinylated 4G8 (Biolegend, 800704, streptavidin, 1:700)

[0250] Iba1 (SYSY, 234308, Guinea pig, 1:500)

[0251] GFAP (Invitrogen, 130300, Rat, 1:1000)

[0252] LPS (2D7 / 1) (Abeam, ab35654, Mouse, 1:100)

[0253] LPS (C6) (Invitrogen, MA541631, Mouse, 1:100)

[0254] D54D2 (Cell Signaling, 8243S, Rabbit, 1:1000)

[0255] Example 1-12. Fluorescence Imaging

[0256] Fluorescence imaging was conducted using the Andor BC43 Benchtop Confocal Microscope (Andor, Belfast, UK). 3D reconstruction and rendering were performed using Imaris software (Imaris for Neuroscientists, Oxford Instruments, version 10.0).

[0257] Example 1-13. Ap ELISA

[0258] For the enzyme-linked immunosorbent assay (ELISA) of Ap, the Amyloid beta (1-40) Human ELISA Kit (IBL, 27713) and the Amyloid beta (1-42) Human ELISA Kit (IBL, 27711) the present disclosure re used.

[0259] The tissue samples the present disclosure re first weighed using a precision microbalance, and PBS containing Protease Inhibitor Cocktail (Sigma, P8340) at a 1:100 dilution was added at four times the tissue weight. The samples the present disclosure re then homogenized using a tissue grinder. After homogenization, the samples the present disclosure re centrifuged at 13,000 rpm at 4°C for 5 minutes, followed by ultrasonication.

[0260] The total protein concentration was quantified, and the homogenates the present disclosure re adjusted to a final protein concentration of 1 mg / ml by addingDocket No. 043846.00009 (PCT)

[0261] Radioimmunoprecipitation Assay (RIPA) buffer (iNtRON Bio, IBS-BR002) to obtain 100 pl of sample solution. The samples the present disclosure re then subjected to ultracentrifugation at 100,000 g at 4°C for 1 hour. After centrifugation, the supernatant was collected in a new tube and used for ELISA to quantify RIPA-soluble Ap.

[0262] For RIPA-insoluble Ap quantification, the pellet was resuspended in 400 pl of 70% formic acid and pipetted thoroughly to dissolve the aggregates. The sample was then subjected to a second ultracentrifugation at 100,000 g at 4°C for 1 hour. The resulting supernatant was collected and used for ELISA to measure RIPA-insoluble Ap.

[0263] The ELISA procedure was conducted following the manufacturer’s instructions provided with the respective kits.

[0264] Example 1-14. Statistical Analysis

[0265] Statistical analyses the present disclosure re performed using MedCalc 20.113 (MedCalc Software, Ostend, Belgium). The Kolmogorov-Smirnov test was conducted to assess normality.

[0266] For parametric variable comparisons, either the Student’s t-test or analysis of variance (ANOVA) with Tukey’s post hoc test was used. For nonparametric variable comparisons, either the Mann-Whitney U test or the Kruskal-Wallis test with Conover’s post hoc test was applied.

[0267] Since MedCalc 20.113 does not provide exact p-values for Conover’s post hoc test, scikit-posthocs (v0.9.0) was additionally used to obtain accurate values. For correlation analysis, the Kendall rank correlation coefficient test was conducted.

[0268] All reported p-values the present disclosure re two-sided. Data visualization and statistical result representation the present disclosure re performed using GraphPad Prism 8 (GraphPad Software, CA, USA).

[0269] Example 2. Bacterial extracellular vesicles propagate to the brain via visceral sensory nerve, not via blood.

[0270] To determine the entry pathway of naturally occurring bacterial extracellular vesicles into the brain, a specific marker capable of selectively tracking these vesicles is required. According to previous studies, lipopolysaccharide (LPS), observed in the outer membrane of Gram-negative bacteria, is known to enter cells only when bound to extracellular vesicles. Additionally, it has been reported that only LPS bound to extracellular vesicles can be transported to the brain via the vagus nerve. Based on this evidence, the present disclosure aimed to clearly identify the route through which LPS reaches the brain and to elucidate the mechanism by which bacterially derived extracellular vesicles are delivered to the brain.Docket No. 043846.00009 (PCT)

[0271] To investigate this, subdiaphragmatic vagotomy was performed in an Alzheimer’s disease (AD) mouse model, and the amount of LPS in the brain was examined to determine whether vagotomy resulted in a reduction in LPS accumulation (FIG. 1). Considering previous reports that LPS deposition is increased in the brains of AD patients, the present disclosure confirmed that hippocampal LPS accumulation was also elevated in the AD mouse model. Interestingly, this accumulation was significantly reduced following vagotomy, suggesting a vagus nerve-dependent route for LPS entry into the brain.

[0272] To further investigate the pathway by which LPS is transported to the brain, immunohistochemical staining was performed. Quantification was conducted in the medulla, specifically at a location where the dorsal motor nucleus of the vagus nerve (DMVN) is most clearly visible. This specific region was chosen because it allows simultaneous observation of the area postrema (AP), DMVN, and nucleus tractus solitarius (NTS). The caudal portion of the medial NTS (mcNTS), in particular, is known to receive sensory signals originating from the gut. The AP is a region lacking the blood-brain barrier (BBB), and the DMVN contains the cell bodies of vagal motor neurons (VMNs) that communicate with the gut. By analyzing this region, the present disclosure could effectively compare the three pathways by which gut-derived LPS could reach the brain, namely via the bloodstream, VMNs, or visceral sensory neurons (VSNs).

[0273] Immunohistochemical staining for LPS was performed in this area. To precisely locate the DMVN, additional choline acetyltransferase (CHAT) staining was conducted. Using optical and fluorescence microscopy, the exact locations of AP, DMVN, mcNTS, and lateral caudal NTS (IcNTS) the present disclosure re identified in Alzheimer’s disease model (AS) mice, and LPS accumulation in these regions was quantified and compared. Notably, LPS deposition in the mcNTS was statistically significantly higher than in other regions, indicating a preferential accumulation in this area (FIG. 2).

[0274] To further validate these findings, LPS accumulation in bilateral NTS regions was compared between AS and vagotomized AD model (AV) mice. In AS mice, no significant difference in LPS accumulation was observed between the left and right NTS. However, in AV mice, LPS levels in the NTS on the vagotomized (right) side the present disclosure re significantly lower than those observed in the left NTS, confirming a statistically significant reduction in LPS accumulation on the side where the vagus nerve was severed (FIG. 3).

[0275] In addition, the present disclosure performed immunostaining of the nodose ganglion to confirm that LPS could propagate to the brain via VSN. In WT mice, LPS accumulation in PHOX2B-positive neurons was minimal, suggesting limited entry of gut-derived bacterial components into VSN under normal conditions (FIG. 4a). In contrast, AD mice exhibited significantly increased LPS accumulation within PHOX2B-positive neurons, indicating that LPS is actively taken up by VSNs in disease conditions (FIG. 4b).Docket No. 043846.00009 (PCT)

[0276] Interestingly, LPS signals in the present disclosure are not uniformly present in all VSNs but the present disclosure only employed a subset of PHOX2B-positive neurons, suggesting that only specific neuronal subtypes or functionally distinct VSNs are involved in the uptake and transport of bacterial extracellular vesicles. This selective accumulation implies that certain VSNs are more susceptible to LPS entry.

[0277] To further explore this phenomenon, the present disclosure quantified the proportion of PHOX2B-positive neurons that also displayed LPS signals in WT and AD mice (FIG. 4c). While a trend toward increased LPS-positive VSNs was observed in AD mice compared to WT controls, this difference did not reach statistical significance.

[0278] Through these experiments, the present disclosure demonstrated that bacterial extracellular vesicles are transported to the brain specifically via the visceral sensory neurons of the vagus nerve.

[0279] Example 3. Modeling in vitro screening platform of bacterial extracellular vesicle transport

[0280] The previous findings demonstrated that bacterial EVs reach the brain not via the bloodstream but through visceral sensory neurons (VSNs). Based on this, the present disclosure aimed to develop an in vitro screening platform using iPSC-derived visceral sensory ganglion organoids (VSGOs) to investigate the conditions under which bacterial extracellular vesicles are more efficiently transported to the brain.

[0281] First, the present disclosure established in vitro gut-nerve-brain axis-on-a-chip model utilizing microfluidic devices. The present disclosure connected VSGO with human brain organoids (HBO) and human colon organoids (HCO) (FIG. 5a).

[0282] Next, the present disclosure extracted wild type mice (ADLPwr) driven fecal EVs and treated them into the colon chamber (FIG. 5b). Immunostaining showed that LPS signals the present disclosure re observed in the brain side neurite gel channel, indicating that bacterial-EV bound LPS could propagate through VSGO to brain.

[0283] The above results successfully recapitulate that VSN mediated bacterial EV transport among in vitro systems.

[0284] Example 4. TNF-a enhances neuronal activation and facilitates the transport of bacterial extracellular vesicles

[0285] Previous in vivo study have shown that treating the gut with proinflammatory cytokines, including TNF-a, can activate VSN. Additionally, gut inflammation is the present disclosure Il-documented in AD mouse models. Epidemiological studies have also indicated that inflammatory bowel disease, in which TNF-a plays a key role in pathophysiology,Docket No. 043846.00009 (PCT)

[0286] significantly increases the future risk of developing AD. The present disclosure hypothesize that increased TNF-a in the gut during AD could activate VSNs, ultimately enhancing the propagation of gut microbiome-derived EVs from the gut to the brain through activitydependent transport.

[0287] To investigate whether inflammatory conditions influence neuronal activity and bacterial EV transport, the present disclosure performed live calcium imaging in the gut-nerve-brain axis on a chip following TNF-a stimulation (FIG. 6a). Calcium imaging in VSGOs revealed that neuronal activity increased following TNF-a treatment compared to baseline conditions (FIG. 6b, 6c). Quantitative analysis showed a gradual and sustained increase in calcium fluorescence intensity (AF / F) over time, indicating that TNF-a exposure led to prolonged neuronal activation (FIG. 6d). These findings suggest that inflammatory stimuli can enhance the excitability of visceral sensory neurons, modulating the transmission of gut-derived signals to the brain.

[0288] Next, the present disclosure investigated whether TNF-a induced neuronal activation affects bacterial EV transport using Atto-488 tagged E. coli EV (FIG. 7a). EVs the present disclosure re treated to the colon chamber under two conditions: EV-only and EV with TNF-a treatment. Fluorescence intensity measurements in the media of brain chamber showed that bacterial EV transport was significantly enhanced in the presence of TNF-a (FIG. 7b).

[0289] Immunocytochemistry confirmed that Alexa-488-labeled EVs the present disclosure re colocalized with TUJI-positive neuronal structures (red), with a higher accumulation observed in the TNF-a treated condition (FIG. 7c, 7d and 7e).

[0290] Together, these results demonstrate that inflammatory conditions enhance neuronal excitability and promote the transport of bacterial extracellular vesicles along the gut-nerve-brain axis, providing further evidence that inflammation could amplify gut-derived bacterial component propagation to the brain.

[0291] Example 5. Analysis of the Results

[0292] The present disclosure provides VSN as a novel and efficient shortcut for drug and EV transport to the brain, bypassing the BBB. The results demonstrate that bacterial EVs originating from the gut propagate to the brain through the VSN rather than the bloodstream, establishing a direct gut-nerve-brain axis for molecular transport. The result provides a new paradigm for CNS drug delivery, particularly in diseases where BBB penetration remains a major challenge.

[0293] The present disclosure provides that the role of the vagus nerve in transmitting pathological proteins from the gut to the brain in neurodegenerative disorders by the direct evidence that VSN carries bacterial EVs directly from gut to brain, underscoring the VSN’s broader role as a conduit for molecular transport. The in vivo experiments confirm that LPS-Docket No. 043846.00009 (PCT)

[0294] carrying EVs fail to accumulate in the brain following vagotomy, ruling out passive diffusion through the BBB as a primary mechanism. Furthermore, in vitro gut-nerve-brain axis-on-a-chip successfully recapitulate VSN-mediated transport, further validating this mechanism.

[0295] Additionally, the present disclosure identified TNF-a as a key modulator of VSN activity, where pro-inflammatory conditions significantly enhanced both neuronal excitability and bacterial EV transport. The results demonstrate that TNF-a stimulation increases neuronal activity in VSNs and enhances EV transmission to the brain, suggesting that inflammation amplifies gut-derived molecular transport via the VSN. These findings not only highlight the inflammatory modulation of VSN transport but also suggest therapeutic strategies targeting gut inflammation to mitigate pathological molecule entry into the brain.

[0296] The implications of the findings in the present disclosure extend beyond bacterial EV transport. Given the present disclosure established role of EVs in intercellular communication and drug delivery, the results suggest that harnessing the VSN as a natural delivery conduit can provide a highly efficient platform for CNS-targeted therapies. By engineering drug-loaded EVs with VSN-specific uptake markers, the present disclosure provides a precision drug delivery system via gut-nerve-brain axis for targeted CNS therapy.

[0297] In conclusion, the above data establishes the VSN-mediated gut-nerve-brain axis as a novel pathway for bacterial EV transport and drug delivery. These findings pave the way for therapeutic interventions targeting VSNs to enhance CNS drug delivery strategies.

[0298] Example 6. Posterior subphrenic vagotomy

[0299] Mice were fasted from the day before surgery to minimize intestinal peristalsis that can impair the operative view. On the day of surgery, mice were anesthetized with inhalational isoflurane gas anesthesia. If animals exhibited gasping or panting under inhalation, the anesthetic level was reduced to avoid respiratory suppression. The abdominal fur over the intended incision was removed with an electric clipper. Limbs were secured with tape and a heating pad was placed under the animal to prevent intraoperative hypothermia.

[0300] Using forceps and scissors, we incised the skin and subcutaneous fat from just caudal to the xiphoid process for ~2 cm. With care to avoid visceral injury, we continued the incision through the muscle and peritoneum. Bulldog clamps were placed through the full thickness of the abdominal wall (peritoneum to skin) at the left-upper, left-lower, right-upper and right-lower quadrants to maximize exposure. The small intestine was gently exteriorized to improve visualization and covered with PBS-moistened gauze to prevent desiccation. Additional PBS-moistened gauze was placed to displace the liver and optimize exposure of the stomach as follows: (1) behind the stomach (to separate it from the spleen and reveal the posterior aspect), (2) between the diaphragm and liver (to pull the liver-stomach-spleenDocket No. 043846.00009 (PCT)

[0301] complex caudally and bring the esophageal region into view), and (3) beneath the median lobe of the liver (to expose the stomach).

[0302] The forestomach, which has relatively sparse vasculature, was gently grasped with a hemostat to stabilize the stomach. Fibrous tissues between spleen and stomach and between liver and stomach were carefully dissected to minimize organ injury, and the stomach was rotated to the left to expose the posterior vagus nerve coursing along the dorsal esophagus. The posterior vagus nerve was identified and, with attention to avoid damaging the esophageal wall, transected using micro-scissors. For sham operations, the nerve was visualized but not cut.

[0303] Viscera were returned to the peritoneal cavity, the intraperitoneal gauze was removed, and hemostasis was confirmed. The peritoneum was closed first with absorbable sutures at 5 mm spacing. The abdominal wall (muscle and subcutaneous layers) and skin were then closed together with nylon sutures at uniform 5-10 mm spacing. Postoperatively, mice were placed in a warmed recovery cage until fully awake. To avoid corneal injury, animals were recovered in an empty cage without bedding. All expreiments were approved by the Institutional Biosafety Committee (IBC) of Seoul National University. IACUC number for animal experiments: SNU-240221-3-5. At least 2 cohorts were used for analysis.

[0304] Example 7. Fluoro-gold retrograde tracing

[0305] To verify completeness of the posterior vagotomy, we used Fluoro-Gold (Fluorochrome) as a retrograde tracer. 2 mg ml-1of Fluoro-Gold dissolved in PBS was administered at 0.4 ml per mouse. On day 5 after injection, mice were euthanized and brains were collected. Brains were split at the midline in the sagittal plane, and Fluoro-Gold-positive neurons were counted in the dorsal motor nucleus of the vagus nerve (DMVN) in the left and right medulla to compare signals between sides.

[0306] Example 8. CCK-8 assay

[0307] To assess CCK-8 responsiveness after posterior vagotomy or sham surgery, we performed CCK-8 assay >1 month after surgery (post-operative recovery). Mice were fasted for 24 h before testing, then singly housed in small bedding-free cages to reduce distraction and environmental confounders, and allowed to acclimate for ~2 h. CCK-8 (Tocris Bioscience, 1166) diluted in PBS was administered intraperitoneally at 12.5 pg kg-1.

[0308] Immediately after injection, standard chow pellets were made available, and food intake was quantified for 30 min by measuring the change in pellet mass.

[0309] Example 9. Harvesting Mouse tissuesDocket No. 043846.00009 (PCT)

[0310] Anesthetized mice underwent transcardial perfusion by puncturing the right atrium with micro-scissors and continuously infusing cold PBS into the left ventricle. Brains (including brainstem) and colons were then dissected. For immunohistochemistry, tissues were post-fixed in 4% paraformaldehyde (PFA) at 4 °C for 24 h; for other assays, tissues were snap-frozen in liquid nitrogen and stored at -80 °C.

[0311] Nodose ganglia was harvested by following the previously described protocol55.

[0312] Briefly, cervix was truncated between the first and second cervical segments. After extracting the brain, skull base was exposed. Occipital bone was carefully cut along the midline and then separated from the temporal bone. Nodose ganglia were exposed and dissected along the vagus nerve, which comes down from the jugular foramen.

[0313] Example 10. Immunohistochemistry of hippocampus and brainstem

[0314] Fixed tissues were dehydrated in 30% sucrose for 72 h at 4 °C and stored at -80 °C. For staining, brains were sectioned at 25 pm. Sections were placed on Transwell inserts containing PBS and washed three times in PBS. For Ap and tau staining, antigen retrieval was performed in 70% formic acid for 20 min; other targets did not receive formic acid treatment. Sections were then incubated at room temperature for 1 h in blocking and permeabilization solution (BPS; 0.3% TritonX-100, 5% normal horse serum (Vector, S-2000-20) and 0.5 mg ml-1BSA mixed in PBS), followed by primary antibodies diluted in BPS overnight at 4 °C on a shaker. After three PBS washes at room temperature, sections were incubated for 1 h at room temperature with secondary antibodies diluted in PBS containing 5% normal horse serum or 5% normal goat serum (Vector, S-1000). After three PBS washes, sections were incubated with DAPI (1:5,000 in PBS) for 15 min at room temperature, washed three additional times in PBS, mounted on slides and coverslipped.

[0315] Primary antibodies included the following: AT8 (Thermo Fisher, MN 1020; mouse; 1:200), AT180 (Thermo Fisher, MN1040; rabbit; 1:200), biotinylated 4G8 (BioLegend, 800704; detection with streptavidin; 1:700), ChAT (Merck, AB144P; goat; 1:200), Iba1 (Synaptic Systems, 234308; guinea pig; 1:500), GFAP (Invitrogen, 13-0300; rat; 1:1,000), LPS (clone 2D7 / 1; Abeam, ab35654; mouse; 1:100), LPS (clone C6; Invitrogen, MA541631; mouse; 1:100), lipid A (Thermo Fisher, PA1-73178 ; goat; 1:100) and D54D2 (Cell Signaling Technology, 8243S; rabbit; 1:1,000). No data was excluded in quantified immunostaining.

[0316] Example 11. Immunohistochemistry of nodose ganglia

[0317] Fixed ganglia were dehydrated in 30% sucrose for 72 h at 4 °C. The ganglia were mounted on cryomold (Sakura, cat. no. 4557) and covered with OCT compound (Sakura, cat. no. 4583), then were stored at -80 °C. For staining, ganglia were sectioned at 10pm. Sections were immediately placed on slide glass and placed overnight at -20 °C. ImmEdgeDocket No. 043846.00009 (PCT)

[0318] pen (Vector, cat. no. H-4000) carefully washed three times with PBS then incubated at room temperature for 30min with 0.3% TritonX-100 in PBS (PBS-T), followed by blocking solution (5% normal horse serum or goat serum in PBS-T) at room temperature for 1h, then incubated with primary antibodies diluted in blocking solution at 4 °C overnight. After three washes with PBS-T, slides were incubated with secondary antibodies diluted in blocking solution for 1 h at room temperature. After two PBS-T and two PBS washes, slides were incubated with DAPI (1:5,000 in PBS) for 5 min at room temperature, washed three additional times with PBS, and then coverslipped.

[0319] Primary antibodies included those listed for immunohistochemistry and the following: Phox2B (Abeam, ab183741; rabbit; 1:100), LPS (clone 2D7 / 1; Abeam, ab35654; mouse; 1:100) and Tujl (Novus biologicals, NB100-1612; chicken; 1:2000)

[0320] Example 12. Analysis of human postmortem brain sequencing data

[0321] To investigate the transcriptomic profiles of human microglia in Alzheimer's disease, we utilized a publicly available dataset previously reported by the SEA-AD10. We selectively extracted microglial clusters and performed downstream analysis in Scanpy (v1.11.5) with no additional quality control. We used Decoupler (v2.1.1) for LPS score calculation, pseudobulk analysis and Gene set analysis, and scvi-tools (v1.4.0) was used for semi-supervised clustering utilizing single-cell annotation using variational inference (SCANVI)12’1356.

[0322] To calculate the LPS response score, we utilized a gene signature derived from a previously reported transcriptomic dataset of induced microglia treated with either LPS or saline. The scoring was performed using the Univariate Linear Model (ULM) algorithm11 12. Significant DEGs were identified based on an absolute Iog2 fold change > 0.5 and an adjusted p-value < 0.05. For ULM scoring, the reported test statistics ('stat') of these filtered DEGs were utilized as weights. To ensure robustness against extreme values, these weights were clipped at the 5th and 95th percentiles prior to the final score calculation.

[0323] To identify overlapping DEGs between the LPS-responsive and DAM signatures, we integrated the LPS DEG set with the DAM DEGs identified from our single-nucleus data. Specifically, we aggregated the single-nucleus profiles into donor-by-brain-region pseudobulk profiles and performed differential expression analysis using pyDESeq2, controlling for sex and brain region as covariates. Genes were then prioritized based on their test statistics: for concordant upregulated directions (LPS+ 1 DAM+), genes were ranked in descending order of their 'stat' values, while for concordant downregulated directions (LPS- 1 DAM-), they were ranked in ascending order. It was then calculated an integrated rank for each intersection by summing the individual ranks. The top 20 genes from each intersection were subsequently selected as seed markers to guide semi-supervised clustering via scAN VI.Docket No. 043846.00009 (PCT)

[0324] To quantify gene set activity across groups, we aggregated single-nucleus data into pseudobulk profiles stratified by donor, brain region, and LPS / DAM-associated cluster. Following quality control and loglp-transformation, Gene Set Variation Analysis (GSVA) was performed using the Reactome Pathways 2024 database. To identify cluster-specific pathway enrichment, GSVA scores for each cluster were compared against all remaining samples using a Welch’s t-test. Statistical significance was reported as FDR-adjusted q-values to account for multiple testing.

[0325] Quality control was performed to mitigate noise in the pseudobulk datasets. At the sample level, profiles failing to meet the minimum thresholds of 10 cells or 1,000 counts were discarded. Gene filtering was implemented based on both abundance and prevalence: genes required at least 10 counts per sample and an aggregate sum of 15 counts.

[0326] Additionally, to ensure subtype-specific and donor-consistent representation, genes were filtered based on a minimum detection rate of 30% per subtype and presence in at least 10% of the total samples (FIG. 9).

[0327] Example 13. Amyloid beta ELISA

[0328] For enzyme-linked immunosorbent assays (ELISAs) of amyloid-p (Ap), we used the Amyloid beta (1-40) Human ELISA kit (IBL, 27713) and the Amyloid beta (1-42) Human ELISA kit (IBL, 27711). Tissue pieces were weighed on a microbalance and homogenized in PBS containing protease inhibitor cocktail (Sigma, P8340; 1:100) at a volume equal to 4* tissue weight using a tissue grinder. Homogenates were centrifuged at 13,000 rpm for 5 min at 4 °C, followed by sonication. Total protein concentration was measured, and aliquots were adjusted with radioimmunoprecipitation assay (RIPA) buffer (iNtRON Biotechnology, IBS-BR002) to 1 mg ml-1protein in a final volume of 100 pl. Samples were then ultracentrifuged at 100,000 g for 1 hour at 4 °C. The supernatant was assayed by ELISA to quantify RIPA-soluble Ap. The pellet was resuspended in 400 pl of 70% formic acid, triturated, and ultracentrifuged again at 100,000 g for 1 hour at 4 °C. The resulting supernatant was used to quantify RIPA-insoluble Ap by ELISA. All ELISAs were performed according to the manufacturers’ manuals. We performed Grubb’s test (Alpha threshold 0.05) to exclude outliers.

[0329] Example 14. Nodose ganglia nuclei extraction and snRNA-seq

[0330] Nodose ganglia were collected from seven ADLP(WT) and five ADLP(APT) mice and snap-frozen. For nuclei extraction from frozen tissue, samples were thawed at 4 °C and kept on ice. Small tissue fragments contained in cryovials were transferred into pre-cooled 1.5 ml tubes by adding 100 pl of homogenization buffer (250 mM sucrose, 25 mM KCI, 5 mMDocket No. 043846.00009 (PCT)

[0331] MgCI2, 10 mM Tris-HCI pH 8.0, 1 pM DTT, 0.10% (v / v) Triton X-100, 0.2 U pl’1SUPERasedn™ RNase Inhibitor (Thermo Fisher Scientific, cat. no. AM2694) in nuclease-free water) to each vial and gently rinsing the tissue off the vial wall and tip. Pooled tissues were transferred into a pre-cooled low-bind tube and mechanically disrupted on ice with a chilled pestle. An additional 900 pl of homogenization buffer was added, and the suspension was incubated on ice for 5 min to release nuclei. The homogenate was passed through a 40 pm Flowmi cell strainer to remove large debris and centrifuged at 500g for 5 min at 4 °C to pellet the nuclei. The supernatant was discarded, and the pellet was resuspended in 1 ,000 pl of homogenization buffer, followed by a second centrifugation (500g, 5 minutes, 4 °C). The final pellet was resuspended in PBS containing 1% (w / v) BSAand 0.2 II pl-1RNase inhibitor. The number and integrity of nuclei were assessed by YOYO-1 iodide (Thermo Fisher Scientific) staining under fluorescence microscopy. Each sample was loaded into one lane of a 10x Genomics Chromium Next GEM-X 3' chip at a concentration adjusted to nuclei counts, according to the manufacturer’s instructions. Libraries were sequenced on a NovaSeq system (Illumina).

[0332] Example 15. Analysis of Nodose ganglia snRNA-seq

[0333] Single-nucleus RNA sequencing (snRNA-seq) analysis was performed using the computing infrastructure of the Genomic Medicine Institute Research Service Center.

[0334] Sequencing reads were aligned to the mouse reference genome (GRCm39-2024-A) using Cell Ranger (v9.0.1). To remove ambient RNAs from droplet-based sequencing outputs, CellBender (vO.2.2) was applied to the raw count matrices. Low-quality cells were filtered out according to the quality-control criteria described in FIG. 11a, b. Potential doublets were identified and removed for each library using the Scrublet implementation available in Scanpy (v1.10.4). A subset of 3,000 highly variable genes was used as pivots for integration in the scVI model (v1.3.2), with each sample condition modeled as an independent batch. The following parameters were applied for model training: n_top_genes=3000, subset=True, layer-counts', flavor='seurat_v3', and batch_key='file'. The scVI model was trained using default options (n_layers=1, n_latent=10, gene_likelihood='zinb', and dropout_rate=0.1). High-precision matrix computation (torch. set_float32_matmul_precision('high')) was enabled to ensure numerical stability and reproducibility of latent representations. Training was performed for a maximum of 400 epochs, with validation every five epochs and early stopping applied to prevent overfitting. A neighborhood graph was constructed from the scVI latent representation, followed by computation of a Uniform Manifold Approximation and Projection (UMAP) embedding. Clusters were then identified using unsupervised Leiden clustering. Gene set enrichment analysis (GSEA) was conducted using GSEApy (v1.1.9),Docket No. 043846.00009 (PCT)

[0335] and transcription factor activity analysis was performed using the CollecTRI resource through the decoupler package (v2.1.1).

[0336] Example 16. Maintenance of human iPS cells

[0337] Human iPS cells were maintained in mTeSR Plus media (Stemcell Technologies, cat. no. 110-1130) on hESC-qualified Matrigel (Corning, cat. no. 354277)-coated six-well plates. Cells were passaged upon reaching 80% confluency using ReLeSR (Stemcell Technologies, cat. no. ST05872), and split at a 1:20 ratio. E3 (hPSCreg UIOi002-A) iPS cell line was used in this study.

[0338] Example 17. Differentiation of VSGOs

[0339] Differentiation of VSGOs was performed with a little modification from our previously published protocol5. Briefly, human iPS cells were washed twice with 2ml per well of DPBS (Gibco, cat. no. 14190144) to remove cellular debris. The cells were then treated with Accutase (Sigma, cat. no. A6964) containing 20pM Y-27632 (Stemcell Technologies, cat. no. ST72304), incubated for 5 min and gently pipetted to achieve a single-cell suspension. After that, the cells were resuspended in 2ml per well of AggreWell EB formation medium (Stemcell Technologies, cat. no. 05893, hereinafter referred to as EB medium) containing 20 pM Y-27632, then distributed equally as 1 x 105cells per well into the StemFIT (MICROFIT, cat. no. H2951000) pre-rinsed with 2ml of anti-adherence rinsing solution (Stemcell Technologies, cat. no. 07010) (day -2). At day 0, the medium was partially changed (2ml / well) with chemically defined differentiation medium (1:1 mixture of Ham’s F12 Glutamax (Gibco, cat. no. 31765-035) and IMDM Glutamax (Gibco, cat. no. 31980-030) supplemented with 1x chemically defined lipid (Gibco, cat. no. 11905-031), 5 mg ml-1of BSA (Gibco, cat. no. 15260037), 10 mg ml-1of insulin (Sigma, cat. no. I9278), 20 mg ml-1of transferrin (Sigma, cat. no. T8158), 450 pM of 1-thioglycerol (Sigma, cat. no. M6145) and 100 pg ml-1of normocin (Invivogen, cat. no. Ant-nr-1)), supplemented with 6 ng ml-1of FGF-2 (R&D systems, cat. no. 233-FB-025), 15 pM of SB431542 and 3.75 ng ml’1of BMP4 (Peprotech, cat. no. AF-120-05ET). On day 2, the medium was partially changed with chemically defined differentiation medium supplemented with supplemented with 4 ng ml-1of FGF-2, 10 pM of SB431542 and 2.5 ng ml-1of BMP4. On day 4, the medium was partially changed with chemically defined differentiation medium supplemented with 75 ng ml-1of FGF-2, 300 nM of LDN193189 (Tocris Bioscience, cat. no. 6053). On day, 6, the medium was partially changed with chemically defined differentiation medium supplemented with 50 ng ml-1of FGF-2, 200 nM of LDN193189. From day 8, the medium was partially exchanged every two days with Epibranchial placode maturation medium (EPMM, 1:1 mixture of Advanced DMEM / F12 (Gibco, cat. no. 12634010) and Neurobasal medium (Gibco, cat. no.Docket No. 043846.00009 (PCT)

[0340] 21103-049) supplemented with 0.5* N2 supplement (Gibco, cat. no. 17502-048), 0.5* B27 without vitamin A supplement (Gibco, cat. no. 12587-010), 1* GlutaMax (Gibco, cat. no. 35050079), 0.1 mM mercaptoethanol (Gibco, cat. no. 21985-015) and 100 pg ml-1 of Normocin) supplemented with 50 ng ml-1of FGF3 (R&D systems, cat. no. 1206-F3-025 / CF), 50 ng ml-1of FGF10 (R&D systems, cat. no. 345-FG-025 / CF), 20 ng ml-1of BMP4 and 10 pM of IWR1 (Sigma, cat. no. I0161-5MG) until day 14.

[0341] Example 18. Differentiation of BOs

[0342] Human brain organoids were generated from human iPS cells by following previously described protocol31. Briefly, human iPS cells were washed twice with 2ml per well of DPBS to remove cellular debris. The cells were then treated with Accutase containing 20pM Y-27632, incubated for 5 min and gently pipetted to achieve a single-cell suspension. After that, the cells were resuspended in 2ml per well of AggreWell EB formation medium, and plated in AggreWell 80024 well plates (STEMCELL Technologies, cat. no. 34815) pre-rinsed with 1ml of anti-adherence rinsing solution (day 0). On the following day, the medium was changed to EB formation medium without Y-27632. From day 2 to day 5, the medium was partially changed with DMEM / F-12 supplemented with Glutamax, 20% KnockOut Serum Replacement (Gibco, cat. no. 10828010), 1% MEM Non-Essential Amino Acids Solution (Gibco, cat. no. 11140-050), 0.1 mM 2-mercaptoethanol, 100 U / mL penicillin, 100 pg / mL streptomycin, 10 pM dorsomorphin (Merck, cat. no. P5499), and 10 pM SB-431542. On day 6, embryoid bodies were collected and transferred to individual wells of a 96-well ultra-low-attachment microplate (Corning, cat. no. 7007). These organoids were cultured in Neurobasal medium supplemented with B27 without vitamin A supplement, 100 U / mL penicillin, 100 pg / mL streptomycin, Ix GlutaMAX, and 0.5% (v / v) Matrigel Basement Membrane Matrix (Corning, cat. no. 354234), supplemented with 20 ng ml’1epidermal growth factor (EGF; Peprotech, cat. no. AF-100-15) and 20 ng ml’1FGF-2. Fresh neural differentiation medium was replaced daily until day 15, then every two days between day 16 and 24. From day 25 to 42, Neurobasal medium supplemented with B27 without vitamin A supplement, 100 U / mL penicillin, 100 pg / mL streptomycin,, Ix GlutaMAX, and 0.5% (v / v) Matrigel Basement Membrane Matrix, supplemented with 20 ng ml’1brain-derived neurotrophic factor (BDNF; Peprotech, cat. no. 450-02) and 20 ng ml’1neurotrophin-3 (NT-3; Peprotech, cat. no. 450-03), was partially changed every two days. After day 43, the brain organoids were cultured in Neurobasal medium supplemented with B27 without vitamin A supplement, 100 U / mL penicillin, 100 pg / mL streptomycin, Ix GlutaMAX, and 0.5% (v / v) Matrigel Basement Membrane Matrix, and the medium was renewed every four days.

[0343] Example 19. Differentiation of COsDocket No. 043846.00009 (PCT)

[0344] Differentiation of CO followed a previously described protocol5. Briefly, dissociated iPS cells were seeded 2.0 * 105 cells per well in hESC-qualified Matrigel-coated six-well culture plate. Until the cell reached 80% confluency, the medium was exchanged for mTeSR+ medium every other day. When iPS cells reached confluency, we changed the medium to RPMI 1640 medium (Gibco, cat. no. 11875093) containing 2 mM GlutaMAX supplement, 100 ng ml-1Activin A (Cell guidance systems, cat. no. GFH6), 3 pM CHIR99021 (Tocris Bioscience, cat. no. 4423) and 100 U ml-1Pen-strep (Sigma, cat. no. P4333) (day 1). On days 2 and 3, we changed medium daily for RPMI1640 medium containing 2 mM GlutaMAX supplement, 100 ng ml-1 Activin A and 0.2% BSA (Gibco, cat. no. 15260037) and 100U ml-1Pen-strep. At days 4-7, medium was changed daily for RPMI 1640 medium containing 2 mM GlutaMAX supplement, 1* B27 supplement (Gibco, cat. no. 12587-010), 3 pM CHIR99021, 500 ng ml-1FGF-4 (Peprotech, cat. no. 110-31) and 100 U ml-1Pen-strep. After day 8, the medium was changed every other day for Advanced DMEM / F12 (Gibco, cat. no. 12634010) with 2 mM GlutaMAX supplement, 1* B27 supplement, 3 pM CHIR99021, 100 ng ml-1EGF, 300 nM LDN193189 and 100 U ml-1Pen-strep. After day 20, we collected floating sphere gently by pipet, resuspend them in Matrigel, and seeded at24-well culture plate or microfluidics organoid chamber. For 24-well culture plate, Matrigel was treated 40 pl per well and 5 pl per chamber for microfluidic devices. After seeding, the medium was changed every other day with Advanced DMEM / F12 based medium as described before. Last, 3 days before harvest, the medium was replaced with Advanced DMEM / F12 based medium with previously described growth factors, except CHIR99021.

[0345] Example 20. Fabrication of 3-compartment axis-on-a-chip

[0346] Methods for fabrication of 3-compartment microfluidic chips were performed as described in our previous study, with several modifications. Briefly, microchannel chips were produced by curing polydimethylsiloxane solution (PDMS; Dow Chemical, Sylgard 184) over a silicon wafer containing 150 pm (W) X 150 pm (H), microfluidic channel. Cured PDMS was detached and each VSGO, HCO, and HBO chamber were punched using 1 mm, 2 mm, and 3 mm biopsy punches, keeping 2 mm spacing between punches. The media chips were produced by curing PDMS over mold without microchannel. Cured PDMS was detached and media chambers for VSGO, CO, and BO were punched using 4 mm biopsy punches, aligned with the corresponding organoid chambers. Microchannel chips, media chips, and glass coverslip were sterilized at 120 °C for 30 min and then dried at 80 °C dry oven overnight. Dried microchannel chips were bonded together with glass coverslip using an oxygen plasma treatment (FEMTO Science), and immediately the media chips were bonded on the top of the microchannel chips making final assembled chips. DDWwas injected into the entire chips, and chips were stored in room temperature before experiments.Docket No. 043846.00009 (PCT)

[0347] Example 21. Cell culture in 3-compartment axis-on-a-chip

[0348] ECM hydrogel solution for neurite bundle sprout was prepared by mixing 1.5 mg ml’1Collagen type 1 solution (Corning, 354236), 30% Matrigel Growth Factor Reduced Basement Matrix (Corning, 356231), NaOH (1N), and PBS. Stored microfluidic chips were sterilized by exposing ultraviolet (UV) light for over 15 minutes, and all DDW inside the microfluidic chips were suctioned and air dried at room temperature for about 10 min. 5 l of ECM hydrogel solution was injected through the VSGO chamber by pipet, making sure that both entire microchannels and the part of BO and CO chambers were covered by solution. Promptly, D14 VSGO was loaded to the VSGO chamber before gelation of the hydrogel. Microfluidic chips containing hydrogel and VSGO were moved inside the humid chamber and incubated at 37 °C for 30 min. Then each media reservoirs were filled with 40 pl of VSGO medium; Neurobasal medium (Gibco, 21103-049), supplemented with 0.5X N2 supplement, 0.5X B27 without vitamin A supplement, IXGlutaMAX, 0.1 mM mercaptoethanol and 100 pg ml-1 of Normocin, supplemented with 50 ng ml’1of FGF3, 50 ng ml’1of FGF10, 20 ng ml’1of BMP4, 10 pM of IWR1, 10 pM of DAPT (Sigma, D5942), 50 ng ml’1of BDNF (Peprotech, 450-02), 50 ng ml’1of GDNF (Peprotech, 450-10), and 20 ng ml’1of NGF-b (Peprotech, 450-01). Medium was exchanged every 3-4 days until VSGO D28. On day 28 of VSGO, medium inside the reservoirs were removed and then CO and BO were seeded into each organoid chamber, with following addition of 5 pl of hydrogel solution. After gelation for 30 minutes at 37 °C, 60 pl of BO medium, 40 pl of VSGO medium, and 20 pl of CO medium were injected in to each media reservoir. Media change was done every 24 hours during tri-culture period and various assays were done at D32 ~ D38 of VSGO. The entire procedures of culture cell in 3-compartment microfluidic chips are explained in fig. S5. All experiments were repeated for at least 2 batches for analysis.

[0349] Example 22. Electrophysiology of axis-on-a-chip

[0350] Slight modifications were made to the 3-compartment axis-on-a-chip for electrophysiological assays. Media chamber for BO was punched to be closer to the media chamber for VSGO, to reveal the surface of BO. On day 28 of VSGO, CO and BO were seeded into each organoid chamber, without following addition of 5 pl of hydrogel solution into the BO chamber to prevent intervention of hydrogel during recording.

[0351] Electrophysiological recording was done at D32 ~ D38 of VSGO.

[0352] To evaluate the activity of brain organoids in response to neural signals transmitted from the stimulated colon through the VSGO within the connected organoid-on-a-chip system, MEMS neural probes integrated with 16- or32-channel platinum microelectrodes were employed. Neural activity was monitored inside an incubator using a compact, custom-Docket No. 043846.00009 (PCT)

[0353] built microdrive system consisting of a chamber for organoid-on-a-chip placement, a microdrive mechanism for precise vertical positioning of the neural probe, and an acrylic enclosure designed to minimize medium evaporation. After securing the neural probe to the microdrive with two screws, the organoid-on-a-chip was positioned beneath the probe, and the probe was gradually inserted into the organoid over approximately 10 seconds using the microdrive. The organoid with the inserted probe was then enclosed within the acrylic chamber to maintain stable environmental conditions during recording. Electrical signals obtained from the black Pt-coated microelectrodes were amplified and digitized using an RHD2132 amplifier interfaced with an RHD2000 Evaluation System (sampling rate: 30 kS / s per channel; 0.1 Hz high-pass and 7.5 kHz low-pass filters; 16-bit ADC). Each spontaneous recording was acquired for at least 5 minutes. Without removing the inserted probe, a high concentration of glucose(20mM) was subsequently introduced into the colon compartment, and neural activity was recorded for an equivalent duration. In addition, neural activation in response to bile acid (1:100 diluted) stimulation was also examined

[0354] Raw multichannel recordings (30 kHz) were down-sampled to 1 kHz, and power-line interference was removed using a 60 Hz notch filter. Band-limited power (1-100 Hz) was quantified within sliding windows (2.0 s length, 1.0 s step). For each window, the power spectral density (PSD) was estimated using Welch’s method, and the band power was obtained by trapezoidal integration of the PSD across frequency. This yielded a time series of band power per channel for each condition. The resulting traces were smoothed using a one-dimensional Gaussian filter (o = 1 sample; step size = 1 s), followed by logtransformation [ln(P + £), £ = 1 x io-6] to stabilize variance. Pairwise signal coherence was then calculated between channels, and coherence values were averaged across 1-100 Hz to obtain a single band-limited coherence estimate for each channel pair.

[0355] Example 23. Live-cell calcium imaging using axis-on-a-chip

[0356] Calcium imaging for axis-on-a-chip was performed using the Fluo-4 Direct Calcium Assay Kit (Invitrogen). Axis-on-a-chips on day 32 of VSGO, connected with CO were incubated with a 1:1 mixture of reagent and VSGO medium in a humidified 5% CO2 incubator for 2 hours and then washed twice with medium. Fluorescent signals were captured every 2,000 microseconds for 5 minutes in a time series. First, baseline images were captured for spontaneous activity, and then 4 pl of TNF-alpha (Gibco, AF-300-01A) was treated in an CO chamber to reach a final concentration of 50 ng / ml. All measurements of fluorescence intensity were corrected by comparing background signal intensity, and the magnitude of the ratio change (AF) was normalized to the baseline ratio for each region of interest (AF / F0 ratio, fold-change).Docket No. 043846.00009 (PCT)

[0357] Example 24. Immunocytochemistry of axis-on-a-chip

[0358] For immunocytochemistry experiments, we removed media and washed with PBS three times, then treated with cell recovery solution (Corning, cat. no. 354253) for 3 hours. After removing cell recovery solution, microfluidic chips were washed with PBS three times before fixation with 4% paraformaldehyde in PBS overnight at 4°C. After fixation, chips were washed with PBS three times. Blocking and permeabilization solution (BPS; composition as in the corresponding immunohistochemistry section) was treated for overnight at room temperature. Primary antibody mixed in BPS was then added for 72 hours at 4°C. As indicated in Example 10, the primary antibodies used included the following: AT8 (Thermo Fisher, MN1020; mouse; 1:200), AT180 (Thermo Fisher, MN1040; rabbit; 1:200), biotinylated 4G8 (BioLegend, 800704; detection with streptavidin; 1:700), ChAT (Merck, AB144P; goat; 1:200), Iba1 (Synaptic Systems, 234308; guinea pig; 1:500), GFAP (Invitrogen, 13-0300; rat; 1:1,000), LPS (clone 2D7 / 1; Abeam, ab35654; mouse; 1:100), LPS (clone C6; Invitrogen, MA541631; mouse; 1:100), lipid A (Thermo Fisher, PA1-73178 ; goat; 1:100) or D54D2 (Cell Signaling Technology, 8243S; rabbit; 1:1,000). The chips were then washed with PBS containing 0.3% TritonX-100 (PBS-T) three times, and secondary antibody with 4', 6-diamidino-2-phenylindole (DAPI) mixed in 5% normal goat serum (Vector, cat. no. S-1000) with PBS-T for 72 hours at 4°C. The chips were then washed with PBS three times and were finally filled with PBS for confocal imaging.

[0359] Additional primary antibodies are DSG2 (Abeam, ab96761; rabbit; 1:500) orTujl (Abeam, ab78078; mouse; 1:1000).

[0360] Example 25. Fluorescent imaging and analysis

[0361] Fluorescent imaging was conducted using an Andor BC43 Benchtop Confocal Microscope (Andor); 3D reconstructions and renderings were performed using Imaris software (Imaris for Neuroscientists, Oxford Instruments, v.10.0).

[0362] Example 26. Alexa-488 labeling of E. coli OMV

[0363] Alexa-488 labeled OMV were prepared as described previously. Briefly, E.coli OMV (240 g) mixed with 50 pM of Alexa Fluor™ 488 NHS Ester (Invitrogen, Carlsbad, CA, USA) were incubated at room temperature for 1 hour. Ethanolamine (final 5 mM; Sigma-Aldrich, St. Louis, MO, USA) was used for terminate the reaction. Residual Alexa-488 dye was removed utilizing a size-exclusion spun column Lfrom ExoLutE® Conditioned Medium Exosome Isolation Kit (SL Bigen, Incheon, Republic of Korea).

[0364] Example 27. Preparation of outer membrane vesicles (OMV) from Escherichia coliDocket No. 043846.00009 (PCT)

[0365] Outer membrane vesicles (OMV) were isolated from an E. coli strain of intestinal origin using a modified protocol based on ultracentrifugation. Briefly, the bacteria cultured in lysogeny broth (LB) at 37°C for 16 hours were pelleted at 6,000 x g, for 20 min at 4°C. To remove the remaining debris, the supernatant was filtrated with a 0.45 pm pore-sized filter (SPL Life Sciences, Pocheon, Republic of Korea). The filtrate was concentrated with a 100 kDa cutoff hollow fiber cartridge (GE Healthcare Bio-Sciences Corp, Marlborough, IL, USA). OMV were then pelleted by ultracentrifugation at 150,000 x g for 3 hours at 4°C using a Type45 Ti rotor (Beckman Coulter, Brea, CA, USA). Subsequently, the pellet was resuspended in 50% iodixanol solution (50% iodixanol, 150 mM NaCI, 20 mM HEPES, pH 7.4) and subjected to buoyant density gradient centrifugation across sequential layers of 10%, 40% and 50% iodixanol. Following ultracentrifugation at 200,000 xg for 2 hours at 4°C using a SW55 Ti rotor (Beckman Coulter), the fraction containing OMV was collected from the interface between 10% and 40% layers. The protein concentration of OMV was quantified using Bradford assay (Bio-Rad, Hercules, CA, USA), and the samples were split into aliquots and frozen at -80°C before use. OMV propagation assay using axis-on-a-chip.

[0366] Axis-on-a-chips on day 32 of VSGO, connected with both CO and BO, and Alexa-488 E.coli OMV were used for OMV propagation assay. Briefly, OMV was treated to the CO chamber of the axis-on-a-chip every 24 hours for 3 days, and signal intensity from BO chamber was measured every after 24 hours of treatment. At first, all medium of axis-on-a-chips were removed and then 20 pl of OMV solution diluted 1:100 in CO medium with or without 50 ng / ml of TNF-alpha was treated to CO reservoir. VSGO and BO reservoirs were changed with fresh medium. After 16 hours of treatment, medium of BO chamber (NB) was changed with 60 pl of Opti-MEM Reduced Serum Medium (Opti; Gibco, 31985-070) to minimize interference in fluorescence signal measurement. After 24 hours of treatment (8 hours after Opti change), 30 pl of Opti from each BO chamber was harvested and transferred to Black 96well (SPL, 30296). Signal intensity of OMV-tagged Alexa-488 fluorescence was measured using Microplate reader. All measurements of fluorescence intensity were normalized to the measurements of OMV only group.

[0367] Example 28. Quantified propagation assay

[0368] For the quantified propagation assay, confocal images were captured specifically in the neurite bundle area. The quantification process involved dividing the volume of ATTO-488 signal colocalized with neurite signals by the total neurite volume. Only puncta that overlapped with more than 50% of the neurite volume were included in the analysis.Docket No. 043846.00009 (PCT)

[0369] Example 29. PRV-GFP retrograde transmission assay

[0370] PRV-CAG-EGFP virus (Brainvta, cat. no. P03001) was used for the retrograde transmission assay. COs cultured on Matrigel domes in 24-well plates were treated with 600 l per well of cell recovery solution and incubated on a shaker for 2 hours at 4 °C to release the organoids from the domes. The released COs were transferred to BSA-coated 15 ml conical tubes using BSA-coated wide-bore pipette tips. To remove debris and remaining Matrigel, organoids were washed twice by adding 2 ml of DPBS, followed by centrifugation at 300g for 3 minutes. For viral infection, COs were incubated with PRV-CAG-EGFP virus at a multiplicity of infection of 0.5 overnight at 37 °C. Due to the difficulty in obtaining accurate single-cell counts for organoids, the multiplicity of infection was estimated by dissociating and counting single cells from a separate well of the same batch. Following overnight viral incubation, the infected organoids were washed twice with DPBS, collected by centrifugation and then seeded onto the CO chamber of the axis-on-a-chip in which VSGO neurite bundle has sprouted. After 72 hours of co-culture, entire axis-on-a-chip was immunostained and fluorescence images were taken for analysis.

[0371] Example 30. Sample preparation of Nodose ganglia Visium HD spatial transcriptomics

[0372] Formalin-fixed paraffin-embedded (FFPE) tissue blocks were sectioned at a thickness of 5 pm and mounted onto standard glass slides. Deparaffinization, immunofluorescence (IF) staining, and imaging were performed following the Visium HD FFPE Tissue Preparation Handbook (CG000684). For immunofluorescence staining, tissue sections were incubated with two primary antibodies: rabbit anti-Tuj1 (Abeam, catalog no. ab18207; dilution 1:2000) and mouse anti-LPS (Abeam, catalog no. ab35654; dilution 1:100). After washing, sections were incubated with Alexa Fluor® 488-conjugated goat antirabbit IgG (dilution 1:500) and Alexa Fluor® 555-conjugated goat anti-mouse IgG (dilution 1:500) as secondary antibodies. Fluorescent images were acquired using an EOVS M7000 microscope at 40* magnification. Subsequent probe hybridization, ligation, slide preparation, probe release, extension, library construction, and sequencing were carried out according to the Visium HD Spatial Gene Expression Reagent Kits User Guide (CG000685). Sequencing was performed on an Illumina NextSeq 2000 platform using paired-end reads (43 cycles for Read 1, 10 cycles for i7, 10 cycles for i5, and 50 cycles for Read 2).

[0373] Example 31. Analysis of Nodose ganglia spatial transcriptomics

[0374] Spatial transcriptomics (Visium HD) analysis was performed using the computing infrastructure of the Genomic Medicine Institute Research Service Center. Sequencing reads were aligned to the mouse reference genome GRCm39-2024-A provided by 10x GenomicsDocket No. 043846.00009 (PCT)

[0375] using Space Ranger (v4.0.1). For quality control, bins with no detected genes and genes not detected in any bin were initially removed. Cell binning was conducted using the Bin2cell (vO.3.3) method to aggregate transcript counts within spatially adjacent 2um level bins. For image segmentation, StarDist (model “2D_versatile_fluo”) was applied to the DAPI channel, and nuclei were mapped to spatial bins by area overlap. During the cell mapping procedure, low-quality cells were excluded according to the quality-control criteria described in FIG. 26a and b. Doublet-like cells arising from overlapping bins during the binning step were identified and additionally removed for each library using Scrublet as implemented in Scanpy (vl.10.4). To overcome the inherent sparsity of spatial transcriptomic data and enable reliable cell-type annotation, we performed supervised annotations for each library based on canonical cell-type markers previously established from our snRNA-seq dataset. Cell types were assigned when canonical markers exhibited detectable expression (raw counts greater than zero) as follows: adipocytes (Lipe and PlinT), skeletal muscle cells (C m), endothelial cells (Pecaml and Emcri), smooth muscle cells (Myh1T), nodose ganglion (Phox2b), jugular ganglion (Prdm12), satellite glia (Acsbgl and Ednrb), Schwann cells (Mpz and Ncmap), fibroblasts (Den) and immune cells (Ptprc and Cd44). Cells lacking detectable expression of these canonical markers were temporarily categorized as “Other” and subsequently subjected to semi-supervised annotation to refine their cell-type identities.

[0376] To further resolve undefined cell types, we applied a semi-supervised clustering method using ResolVI. We selected 3,000 highly variable genes as anchors and trained the model with the following parameters: n_hidden = 32, njatent = 10, njayers = 2, dropout_rate = 0.05, dispersion = gene, genejikelihood = nb, n_neighbors = 10. After annotating full set of cell types for each library, we performed semi-supervised integration using scANVI13(scvi-tools) using n_hidden = 128, njatent = 30, njayers = 2, dropout_rate = 0.2, dispersion = gene, genejikelihood = zinb and latent_distribution = normal; the AD nodose ganglia dataset served as the reference. A neighborhood graph was constructed from the scANVI latent space and used to compute a LIMAP embedding. For vagal sensory neuron subtype classification, we downloaded a prior dataset and used it as a subtype reference. We trained a CellTypist callsifier(v1.7.1) using this reference dataset, first applying the model to our snRNA-seq data and subsequently transferring the predicted labels to spatial data. We jointly integrated the three nodose ganglia datasets, includinga previous published reference scRNA-seq, as well as our snRNA-seq and ST datasets using CellHint (v1.0.0) to achieve cross-modal alignment. Gene set enrichment analysis was performed with GSEApy (v1.1.9), and significance thresholds and additional statistics are reported in the figure legends and Data S1.Docket No. 043846.00009 (PCT)

[0377] Example 32. Gene set scoring

[0378] We computed per-cell gene-set scores using scanpy.tl.score_genes (Scanpy vl.10.4). For each signature, the score equals the mean expression of the signature genes minus the mean expression of a matched control set, where control genes are randomly sampled from a gene pool within expression bins to match the signature’s expression distribution. Unless noted otherwise, the gene pool comprised all expressed genes after standard preprocessing, control set size was 50 genes per bin, and 50 expression bins were used. Scores were rescaled to 0-1 across cells for visualization.

[0379] Example 33. Protein-LPS complex structure prediction analysis

[0380] We predicted protein-ligand complex structures using AlphaFold3, providing the selected protein structure as a template for each prediction. For structure prediction, we chose the extracellular domain sequences of the candidate proteins APP (E1 and E2 domains modeled each), CLSTN1, TIMP2, NGFR (1-188, TNFR domain) and ANXA2. We provided experimentally determined structures forAPP-E1, APP-E2, ANXA2, and TIMP2 with PDB IDs 3KTM, 3UMK, 2HYV and 1GXD_C, respectively. For NGFR, we provided the AlphaFold2 model of the protein as a template, and the AF2 structure of the protein was modeled by using the structure with PDB ID 3BUK_C as a template. For CLSTN1, we used the deposited AlphaFold2 database model of the protein and used the extracellular region 29-859 only as a template. For the LPS ligand, we selected the R1 outer core region from the reported E.coli R1 LPS structure. We generated a smiles string file of the LPS R1 outer core structure utilizing open babel.

[0381] We evaluated AF3 protein-ligand complex structure predictions with interchain predicted TM-score (ipTM) and the minimum of the mean predicted aligned error (pAE) between chains. We generated pAE plots using matplotlib and used PyMOL3 for the visual analysis and image generation of APP E2 domain-LPS R1 outer core complex structure.

[0382] Protein sequences, template structure information and the LPS ligand in smiles format used for the predictions, and theAF3 prediction evaluation metrics (pIDDT, ipTM, pAE) are provided in FIG. 28.

[0383] Example 34. Statistical analysis

[0384] Statistical analyses unrelated to snRNA-seq and spatial transcriptomics were performed in MedCalc v20.113 (MedCalc Software). Normality was assessed with the Kolmogorov-Smirnov test. For comparisons of normally distributed data, we used Student’s t-test when variances could be assumed equal, and Welch’s t-test when variances were unequal; for multi-group comparisons with unequal variances, we used the Brown-Forsythe analysis of variance (ANOVA) followed by Tamhane’s T2 post-hoc test. For paired, normallyDocket No. 043846.00009 (PCT)

[0385] distributed samples, we used the paired t-test. For non-parametric data, we used the Mann-Whitney II test for two groups and the Kruskal-Wallis test with Conover’s post-hoc procedure for multiple groups. For paired, non-normally distributed samples, we used the Friedman test with Conover’s post-hoc procedure. Exact post-hoc P values were additionally computed with scikit-posthocs (v0.9.0). All tests were two-sided. Statistical plots and summary graphics were generated in GraphPad Prism v8 (GraphPad Software).

Claims

Docket No. 043846.00009 (PCT)CLAIMS1. A composition comprising an extracellular vesicle (EV) and a therapeutic agent, wherein the EV is a bacterial EV and the therapeutic agent is a drug for treating a disease in the brain.

2. The composition of claim 1 , wherein the disease is selected from the group consisting of a neurodegenerative disorder, a brain tumor, a brain cancer, a neurological disorder, and a psychiatric disorder.

3. The composition of claim 1 , wherein the disease is selected from the group consisting of Alzheimer's Disease (AD), Parkinson's Disease (PD), Amyotrophic Lateral Sclerosis (ALS), Huntington's Disease (HD), Dementia with Lewy Bodies (DLB), Frontotemporal Dementia (FTD), Prion Diseases, Spinocerebellar Ataxia (SCA), spinal muscular atrophy (SMA), multiple system atrophy (MSA), progressive supranuclear palsy (PSP), chronic traumatic encephalopathy (CTE), cognitive dysfunction, motor dysfunction, and autonomic dysfunction.

4. The composition of claim 1 , wherein the therapeutic agent is selected from the group consisting of Carmustine, Doxorubicin, Paclitaxel, Temozolomide, L-DOPA, Donepezil, Insulin, Valproic acid, Midazolam, Diazepam, Ketamine, Brexanolone, DNL310, viral vectors, genes, siRNA, peptide, and polysaccharides.

5. A method for delivering therapeutics to brain using a composition comprising:administering a composition of claim 1 to a subject in need thereof,wherein the composition is delivered through visceral sensory neurons (VSN).

6. The method of claim 5, where in the composition is administered orally.

7. An organoid on chip model comprising at least two compartments axis for testing delivery of the composition of claim 1 from gut to brain, wherein the compartments comprises, brain, gut, or nerve.

Images