Amyloid-β-driven Alzheimer’s disease reshapes the colonic immune system in mice

SUMMARY

The “gut-brain axis” is an emerging target in Alzheimer’s disease (AD), although its immunological features remain poorly understood. Using single-cell RNA sequencing, coupled to extensive spectral-tuning flow cytometry validation of the colon immune compartment in the 5XFAD amyloid-β mouse model, we found several AD-associated changes including in B/plasma cell activity. Notably, levels of CXCR4⁺ antibody-secreting cells are reduced in 5XFAD colons. This change corresponds with accumulating CXCR4⁺ B cells and gut-specific IgA⁺ cells in the brain and dura mater, respectively. Consistently, a chemokine ligand for CXCR4, CXCL12, is expressed at higher levels in the 5XFAD brain and in in silico-analyzed human AD brain studies, supporting altered neuroimmune trafficking. An inulin prebiotic fiber diet could expand gut IgA⁺ cells, rescue peripheral T_reg levels, reduce dysbiosis, improve serum microbial metabolite levels, and attenuate overall AD-associated frailty. Our study reveals key aspects of the gut-brain axis and highlights potential targets against AD.

In brief

Makhijani et al. investigate the colonic immune landscape in the 5XFAD model, showing a reduction in CXCR4⁺ antibody-secreting cells and a concomitant increase in brain CXCL12. Dietary intervention with the prebiotic inulin protects against this immunophenotype and Alzheimer’s disease (AD) parameters like frailty, suggesting roles for intestinal immune cells in AD.

Graphical Abstract

INTRODUCTION

Late-onset Alzheimer’s disease (AD) primarily occurs in individuals over the age of 65 years, presents with symptoms of dementia and anxiety, and is more prevalent in females.¹ The neuropathology of AD involves an early pre-symptomatic period that correlates with extracellular deposition of amyloid-β (Aβ) followed by a clinical stage marked by the addition of intraneuronal aggregations of hyperphosphorylated tau protein.^2,3 The local innate immune response to Aβ drives microglia activation⁴ that can be protective at early stages.^5,6 The early adaptive immune response is also considered to be protective in AD.⁷ However, chronic activation of certain microglial subtypes promotes neurodegeneration by orchestrating synapse loss,⁸ blood-brain barrier (BBB) breakdown,^9,10 tau seeding,¹¹ and infiltration of cytotoxic T cells.¹²

The Aβ-associated pre-clinical phase of AD can persist up to 20 years.¹³ This stage has been the subject of intense study to better understand the underlying risk factors, identify early diagnosis biomarkers, and implement preventative interventions. Interestingly, this phase also correlates with metabolic and gastrointestinal complications. Patient studies have shown heightened prodromal weight loss,¹⁴ alterations in energy metabolism,^15,16 intestinal dysfunction,¹⁷ and changes in gut microbial composition,^18,19 compared with healthy aging individuals. An AD-associated loss of intestinal barrier integrity has also been suspected due to elevated levels of fecal proteins in patient serum.²⁰ A relationship between gut health and neuroinflammation has been noted in several neuropsychiatric conditions and after brain injury.^21,22

Termed the “gut-brain axis,” bidirectional gut-brain interactions drive both protective and deleterious effects in the brain in mouse models of various diseases. For instance, gut-microbiota-derived metabolites accumulating in the blood attenuate microglia activation via their ability to permeate the BBB.^23–25 Pre-existing gut dysbiosis promotes neuroinflammation via migratory activation of adaptive immune cells from the gut to the brain during ischemia.²⁶ In multiple sclerosis (MS), an autoimmune condition targeting myelin antigens, gut IgA⁺ plasma cell (PC) migration through the central nervous system (CNS) attenuates neuroinflammation via interleukin (IL)-10 production.²⁷ In fungal infections, gut-derived IgA⁺ cells accumulate in the meningeal border region and defend against pathogenicity.²⁸ AD-specific studies using germ-free mice show that microglia- and astrocyte-mediated neuroinflammation is connected to gut microbiome composition.^29,30 A direct translocation of gut microbes in the AD brain via the vagus nerves has been proposed,³¹ linked to the reported loss of gut barrier integrity.^32,33 In cases of autoimmunity and cancer, bacterial translocation to sites of disease has also been reported.^34,35

However, the nature of the gut immune response in AD, especially with respect to the abundant B lineage cells that partake in microbial homeostasis, is not fully understood. Here, we examined the overall landscape of colonic immune cells in AD using high-dimensional approaches. We found evidence of colon immune activation across both the cellular and humoral arms of adaptive immunity. Remarkably, levels of gut CXCR4^hi antibody-secreting cells (ASCs) were reduced in Aβ-AD modeled mice. Concurrently, we identified a chemokine axis where the AD brain produces increased levels of the CXCR4 cognate chemokine CXCL12, correlating with an increase in brain-associated B lineage cells expressing CXCR4. Additionally, dietary intervention with inulin fiber, known to support both gut health and cognition,^36–39 attenuated some AD parameters, including frailty associated with the disease, brain CXCL12 production, and CD45⁺ cells in the brain, while concomitantly boosting colonic IgA⁺ ASC and systemic T_reg levels. These benefits of inulin coincided with rescued dysbiosis and increased gut microbial-associated serum metabolites of the short-chain fatty acid (SCFA), bile acid, and indole families.

Overall, our findings uncovered key alterations to gut and brain immune cells in AD, as well as a likely chemokine axis shared by these cells. Results from our studies also suggest that targeting gut inflammatory responses via dietary interventions could promote both brain and gut health in AD.

RESULTS

Altered intestinal immune composition across the colonic immune landscape in Aβ-driven AD

To identify immunological changes to the gut-brain axis in Aβ-AD, we used the well-studied 5XFAD model. These mice co-express five familial human AD mutations, all driven by the Thy1 promoter, resulting in extracellular Aβ deposition and plaque formation in the brain starting at ~2 months of age and loss of pyramidal neurons by 9 months.⁴⁰ Neurodegenerative changes typically exhibit as both cognitive and motor impairments in the 5XFAD model and recapitulate aspects of human disease.^41,42 We observed phenotypes similar to those previously reported in this model.⁴³ Between 9 and 13 months of age, frailty increased in 5XFAD compared with wild-type (WT) age-matched female mice (Figure 1A) using a 31-point frailty index.⁴⁴ Frailty changes included a significant reduction in total body weight (Figure 1B), evident body tremors, and instances of diarrhea/constipation (Figure S1A). Similar rapid age-related decline was observed in male 5XFAD mice (Figures S1A and S1B). In both sexes, increasing anxiety was also observed via the open-field behavioral assay by 9–12 months of age in the 5XFAD group (Figure S1C), whereas the Y-maze assay, which measures spatial memory, showed limited correlation with disease (Figure S1D).

Figure 1. Single-cell sequencing reveals altered immune parameters in AD colons.

(A) Thirty-one-point frailty index scores comparing 5XFAD (square) with WT (circle) female mice at specified ages, n = 5 per group.

(B) Body weight of 12-month-old 5XFAD and WT mice, n = 10 per group. Error bars represent means ± SEM; pairwise comparisons were conducted using Student’s t test with a Welch’s correction; p values under 0.25 have been specified and denoted as ns or *p < 0.05 and **p < 0.01.

(C) Schematic workflow of scRNA-seq experiments.

(D) Volcano plot of differentially expressed genes (DEGs) across all immune cell types between WT and 5XFAD. DEGs (including log2 fold change [log2FC] and adj. p value [adj. p]) were calculated by the MAST method. Plotted adj. p cutoff is 0.05, and log2FC cutoff is 0.5.

(E) KEGG pathway scores of significantly (adj. p cutoff is 0.05) downregulated (left) and upregulated (right) DEGs evaluated using g:Profiler using an ordered query.

(F) UMAP plot of colon-derived CD45 immune cell single-cell transcriptomes (10× Genomics) pooled together as described in (C) from WT and 5XFAD mice. Cells were resolved into 17 distinct clusters.

(G) Quantification of relative abundance of each cluster of single immune cells by log2FC between WT and 5XFAD mice.

(H) UMAP of pseudotime trajectory analysis of WT and 5XFAD colon B cells. White circles represent the root node designated at naive B cells (cluster 0). Black lines depict branch pathways, where cells travel to a variety of predicted outcomes from intersecting points.

In this study, we focused on the colon because of its higher levels of bacterial colonization versus the small intestine⁴⁵ and as bacteria-derived metabolites are thought to affect neuroinflammation and vice versa.⁴⁶ Using H&E staining of colon Swiss rolls, 5XFAD mice showed no overt signs of acute or chronic colonic inflammation (Figure S2A). Colon weight-to-length ratios, used in colitis to measure overt gut inflammation,⁴⁷ also showed no significant differences between groups (Figure S2B). Changes in spleen weight mirrored body weight decreases (Figures 1B and S2B), and we observed no changes in splenic follicle number or T and B cell area in the 5XFAD mouse (Figure S2C). By immunofluorescence (IF) on tissue sections (Figure S3A) and enzyme-linked immunosorbent assay (ELISA) on colon lysates for Aβ (1–42) (Figure S3B), we found no measurable Aβ accumulation plaques in WT or 5XFAD colons, as previously reported.⁴⁸

For a deeper investigation of AD-associated compositional immune perturbations in the colon, we performed single-cell RNA sequencing (scRNA-seq) analysis on sorted CD45⁺ immune cells from enzyme-dissociated full-length colons from three 9-month-old female 5XFAD and three age-matched WT littermate control mice (schematic; Figure 1C). Since more AD patients are female,⁴⁹ and females more strongly manifest inflammatory aspects of the disease in the 5XFAD mouse model,⁵⁰ we focused scRNA-seq analysis on female mice. We used 9-month-old 5XFAD mice to study stable gut immune changes at a time point when neurodegenerative outcomes are well established⁴⁰ (schematic; Figure 1C). Using these experimental conditions, we created a single-cell map of colon immune cells that comprised 3,443 cells from 5XFAD and littermate WT control mice.

Across all immune populations in the colon, 5XFAD altered the expression of 561 genes, of which 550 genes had a significant adjusted p value cutoff of <0.05. The total list of significant differentially expressed genes (DEGs) is visualized using a volcano plot (Figure 1D), highlighting the most highly changed transcripts (|logFC| > 0.5). We observed pan-immune induction of ribosomal genes (Rps/40S family, Rpl/60S family, Eif family), mitochondrial genes (Cox7c, Tomm7), and immune activation genes (Lyn, Stat4, H2/histocompatibility family) in 5XFAD colon immune cells (Data S1; Figure S4A). Globally downregulated genes included Igha, Ighm, Ighd, Cd79b, Cd22, Irf8, and Ciita normally associated with B cells (Figures 1D and S4A). The DEGs with adjusted p values <0.05, separated into up- and downregulated genes by logFC value, were entered into g:Profiler as an ordered query (Data S2). Resulting Kyoto encyclopedia of genes and genomes (KEGG) pathways upregulated in the 5XFAD mice included neurodegenerative pathways (amyotrophic lateral sclerosis; Huntington’s, Parkinson’s, and prion diseases; and AD) and innate immune inflammation (ribosome, reactive oxygen species, oxidative phosphorylation, COVID-19) (Figure 1E, right). KEGG pathways downregulated in the 5XFAD mice included adaptive immune-associated genes, most significantly B cell receptor (BCR) and T cell receptor (TCR), as well as nuclear factor kappa-light-chain-enhancer of activated B cells (NF–κB signaling) (Figure 1E, left). Overall, perturbations in protein synthesis, metabolism, and immune activation reflect the manifestation of AD-associated pathology in gut immune cells.

To take a closer look at immune cell subtype-specific changes, we used dimensionality reduction to identify 17 clusters of immune cells visualized by uniform manifold approximation and projection (UMAP) (Figure 1F). Based on the presence of canonical markers (Figures S4B and S4C), we identified one myeloid, eight B lineage, and eight T cell/innate lymphoid cell (ILC) clusters (Figure 1F). DEGs in individual clusters are shown in Figure S4A and Data S3. Minimal to no changes in proportions of T lineage clusters were observed in 5XFAD colons and innate cell clusters, including myeloid and ILC lineages, at the assessed resolution (Figures 1F and 1G). Most remarkably, we observed a near-complete reduction in the frequency of B lineage cluster 5 in the 5XFAD colons, with lesser changes in other B clusters (Figures 1F and 1G). We identified cluster 5 as a mixed cluster representing plasmablasts (PBs) and maturing PCs due to distinguishing transcriptional features lacking in the other seven B lineage clusters. The partial lack of Vpreb3 (surrogate light chain) and the presence of lambda light chain (Iglc2) transcript suggest that these cluster 5 cells have likely completed VDJ recombination⁵¹ (Figure S4C). Furthermore, the expression of Cxcr4, Fos, Fosb, Jun, Irf4, Irf8, and Blc11a and partially Spib, CD38, CD22, Ciita, Ms4a1, Cd79b, Pax5, Ighm, and Ighd suggests that cluster 5 is differentiating through developmental checkpoints toward a mature antibody-secreting PC⁵² (Figure S4C). A fully mature PC population appeared in nearby cluster 6 with Aicda transcript, responsible for class switch recombination, and Prdm1, Zbtb20, Xpb1, CD80, Igha, and Ighg2b mRNAs⁵² (Figure S4C). B cell trajectory prediction analysis conducted using Monocle3 showed that pseudotime starting at naive B cells (cluster 0) results in some cells differentiating into PCs (cluster 6), via several possible paths including a possible germinal center-like Myc⁺ B cell (cluster 7)⁵³ (Figures 1H, S5A, and S5B). PB cluster 5 cells were predicted to partially give rise to PC cluster 6, with developmental trajectories remaining consistent between WT and 5XFAD colon B cells (Figures 1H, S5A, and S5B). Although this did not rule out stress-induced apoptosis of any short-lived PBs within cluster 5, an abundance of migratory signatures⁵⁴ within this cluster and previous reports of neuroinflammation-driven gut-brain PC migration^27,55 support plausible colon PB emigration in the 5XFAD model.

Loss of CXCR4^hi gut PCs and systemic B cell activation in 5XFAD

The striking loss of cluster 5 B cells in 5XFAD colons led us to take a closer look at the transcriptional signatures in all PB/PC subsets. Collectively termed ASCs, cluster 5 (PB, or maturing PC) and cluster 6 (mature PC) showed transcription of several migratory receptors thought to respond to a variety of signals, including Cxcr4, Ackr3 (CXCR7), Tnfrsf13c (BAFFR), Gpr183, Cd38, S1pr1, Ffar1, Gpr174, and Itgb7 (Figures 2A and S4C). A similar expression signature was recently reported in Kaede mice including Cxcr4, Itgb7, Gpr174, and, transiently AP-1 factor Jun, which defined a gut-imprinted transcriptomic signature in colon B cell emigrants in the spleen under homeostatic conditions.⁵⁴ Isolated to the receding cluster 5, we also found Klf family genes (Klf2, Klf4, Klf5, Klf6) (Figures 2A and S4C), transcription factors that regulate B cell migratory programs⁵⁶ and IgA responses.⁵⁷ Under pathological conditions, gut-specific ASCs emigrate to sites of inflammation,^58,59 especially to the CNS.^27,60 As such, the loss of cluster 5, displaying a migratory signature, points toward possible colon ASC emigration in AD.

Figure 2. AD changes the B lineage landscape within the colon.

(A) UMAPs showing B cluster key (top left) and transcriptional changes in key migratory genes in WT and 5XFAD colon B cells (top right and bottom).

(B) Flow cytometric classification of B cell subsets pre-gated on CD45⁺CD19⁺ cells in 12-month-old female mouse colons (left). Representative histograms of CXCR4 expression (gMFI) per B cell subset in WT and 5xFAD as specified (middle), quantified by both frequency and cell number (right) in WT (circle) and 5XFAD (square) mice; n = 5 per group.

(C) Frequency of CXCR4⁺ cells per isotype within the B220⁻CD43⁻ (DN) population.

(D) Representative histograms of JUN expression (gMFI) per subset in WT and 5XFAD (left) quantified by both gated frequency and cell number (right); n = 5 per group.

(E) Frequency of JUN⁺ cells per isotype within the B220⁻CD43⁻ DN population.

(F) Frequency of naive, IgM, and class-switched B2 cells as a percentage of the total B220⁺CD43⁻ SP (B2 cells) population in colon, spleen, and blood.

Error bars represent means ± SEM; pairwise comparisons were conducted using Student’s t test with a Welch’s correction, p values under p = 0.25 have been specified and rest denoted as ns; *p < 0.05, **p < 0.01, ****p < 0.0001.

We next corroborated these transcriptional signatures with protein expression in colon B lineage cells using high-dimensional spectral tuning flow cytometry using full-length colon cell suspensions. Although flow cytometry markers offer less resolution than scRNA-seq, we were able to identify three major B cell subsets and three isotypes. First, we classified broad B cell subsets using B220 and CD43 expression within the CD19 immune (CD45⁺) compartment to capture B lineage heterogeneity within the colon (gating strategy; Figure S6). We used CD19⁺CD43⁺B220^lo/− (CD43 single positive [SP]) to mark B1 cells, although activated B cells and PCs may also exist in this population⁶¹ (Figure 2B). We used CD19⁺CD43⁻B220⁺ (B220SP) to represent the majority of B2 cells (Figure 2B), although some PCs could be B220^lo.⁶¹ Finally, because the traditional PC marker CD138 is sensitive to enzymatic tissue dissociation protocols, we used the CD43⁻B220⁻ (double negative [DN]) population to locate CD19⁺ maturing ASCs (Figure 2B). We suspect that, because the CD19-BCR complex is no longer required in mature PCs, the CD19⁺B220⁻CD43⁻ DN gate captures mainly early PCs but cannot rule out transitioning PBs. We used the term ASC to represent these multiple possibilities. Indeed, upon sub-gating, we found that DN cells were class switched (IgD⁻IgM⁻) or IgM⁺SP, with most of the class-switched cells being either IgA⁺ or IgM⁻IgA⁻, the latter likely representing IgG isotypes (Figures 2C and S6). Consistently, B220SP cells were largely naive (IgM⁺IgD⁺) or class switched (IgM⁺IgA⁻ or IgM⁻IgD⁻) (Figure S6). To measure PCs in the CD19⁻ compartment, we used terminal marker BLIMP1⁶² in the CD3⁻CD19⁻ population and found a significant reduction in the frequency of BLIMP1⁺ cells in the 5XFAD colon (Figure S7A). A trending decrease in IgA⁺ cells was observed by IF staining of colons (Figure S7B), and the percentage of colonic CXCR4⁺ cells in the total CD45⁺IgA⁺ population showed a significant decrease by flow cytometry (Figure S7C).

We then investigated CXCR4 protein expression in each B lineage subset and isotype, as per our scRNA-seq findings. The proportions of CXCR4^hi cells were significantly reduced within the CD43SP and DN compartments, with trending decreases in cell number in 5XFAD compared with WT colons (Figure 2B). Interestingly, the DN cells had the highest proportion and levels of CXCR4 expression compared with B220SP and CD43SP populations (Figure 2B). Within this DN group, the IgA isotype had the highest expression of CXCR4 compared with others (Figure 2C), suggesting a stronger ability to respond to CXCR4-mediated chemotaxis than other B lineage cells. Consistently, we found a significant decrease in the frequencies of CXCR4⁺ ASCs within both the IgA⁺ and IgM⁻IgA⁻IgD⁻ isotypes (Figure 2C) together with a trending decrease in their cell number in 5XFAD colons (Figure S7D). Other studies have shown that colon ASCs are largely IgA⁺ while expressing high levels of CXCR4⁺ and IL-10.⁶³ We also checked for the expression of JUN, found in developing PBs^64,65 and part of early CXCR4 signaling.^66,67 Consistent with CXCR4 staining, we found reduced JUN⁺ frequencies within both CD43SP and DN subsets (Figure 2D), as well as within the IgA⁺ and IgM⁻IgA⁻IgD⁻ isotype subsets (Figures 2E and S7E) in 5XFAD compared with WT colons. The location of CD19⁺JUN⁺ B cells⁶⁸ in the colon was validated by IF staining; they mapped to the outskirts of isolated lymphoid follicles (Figure S7F). Overall frequencies and cellularity of total B cells and of each B cell subset (CD43SP, B220SP, DN) were unchanged in AD colons by flow cytometry (Figure S7G), likely reflecting the reported rapid replenishment of colon B emigrants to maintain gut immune homeostasis.⁵⁴ Other immune subtypes in the colon showed limited changes with a trending increase in the frequency of total ILCs (Figure S7H), as also seen by scRNA-seq (Figure 1G), and trending decrease in total macrophage and dendritic cell (DC) frequencies (Figures S7I and S7J). Examining macrophage and DC subsets more specifically showed no major significant changes by frequency or cell number (Figures S7I and S7J).

As predicted by scRNA-seq B cell activation signatures, we also observed an increase in class-switched B220SP cells and a trending decrease in naive B220SP cells in 5XFAD colons (Figure 2F). By scRNA-seq, we observed activation signatures across all colon B cell clusters (0–7) (Figure S8). Transcriptionally upregulated genes included Stat4, Cd86, H2-Ab1, H2-D1, H2-Eb1, Lyn, and Cd86 associated with immune activation, and Eif4a1, Rps28, Rpl38, Uba52, Cox7c, Tomm7, and mt-Atp8 associated with translation by ribosomes and metabolic activity by mitochondria, representing pan-B cell dysfunction in the 5XFAD colon (Figure S8). By flow cytometry, an increased proportion of class-switched B2 cells in 5XFAD colons was mirrored in the blood, with trending increases in the small population of class-switched B2 cells in the spleen (Figures 2F and S9, gating strategies). With increased B2 class switching, an increased proportion of DN B cells (CD19⁺B220⁻CD43⁻) was observed in 5XFAD spleens with a trending increase in the blood compared with WT (Figures S10A and S10B), indicative of systemic B cell activation. Consistently, an increasing trend in the frequency of IgA⁺ cells was observed in blood; however, no differences were observed in the spleen, where the overall IgA frequencies were very low (Figure S10C). Trending increases in the expression of activation marker CD86 in splenic B2 cells and increased MHCII levels in blood B2 cells were observed in 5XFAD mice (Figures S10D and S10E). We also noted other markers of systemic B cell dysfunction, including increased marginal zone B cell proportions in 5XFAD spleens (Figure S10F), likely indicating increased antigen encounter.

Taken together, the reduced amounts of migratory gene-expressing colon ASCs and increased B2 cell class switching and inflammatory markers throughout the periphery suggest whole-body B cell dysfunction in 5XFAD, with the possibility of ASC emigration from the colon. The colon B activation observed here may be linked to a dysbiosis phenotype,^32,69,70 and the peripheral B cell immunological changes may be correlated with the “leaky gut”³¹ and changing blood BCR repertoire previously observed in AD.⁷¹

AD CNS contains CXCL12-producing cells, coupled to increased CXCR4⁺ B cells, and IgA⁺ ASCs that recognize gut commensal antigens

Next, we sought to better understand the effect of systemic inflammation, including neuroinflammation, correlating with the observed gut ASC phenotype. There was a transcriptional downregulation of CXCL12-binding receptors; its cognate receptor, Cxcr4, and sequestration receptor, Ackr3 (CXCR7), were found in cluster 5 (PB) and cluster 6 (PC), respectively (Figure 2A). Although CXCR4 is also expected to bind to MIF and CXCL14, it preferentially responds to CXCL12 chemotaxis.^72,73 Additionally, B cell migration has been linked to CXCR4-mediated trafficking⁷⁴ specifically in the brain^75,76 with CXCR4 expression reported in dura mater IgA⁺ cells.²⁸ Therefore, we predicted that this phenotype exists in AD and is orchestrated by a CXCL12 signal. To study this effect, we conducted high-dimensional spectral tuning flow cytometry analysis on dissociated mouse brains from 5XFAD mice (gating strategies; Figure S11). We first confirmed that myeloid cells accumulate in the 5XFAD brain at later stages.⁷⁷ We observed an overall increase in the number of myeloid cells in the AD brain marked by CD11b⁺CD11c⁺ within live CD45^med cells (Figure 3A) and an increase in intracellular CXCL12 expressed by those cells (Figure 3B). Using in silico re-analysis of a human single-nucleus RNA-seq (snRNA-seq) dataset published by Zhou et al.,⁷⁸ we identified significant increases in CXCL12 transcript in a subset of AD microglia versus healthy aging in prefrontal cortex tissue, with minor changes in endothelial and astrocyte CXCL12 (Figures 3C and S12A). By IF tissue staining we found that a subset of 5XFAD myeloid cells (CD45^med) expressed CXCL12, specifically in the dentate gyrus region compared with WT (Figure 3D). IF also revealed increased CXCL12 expression on CD45⁻ astrocyte-shaped cells, with enrichment seen in subregions of the hippocampal formation (Figure 3D). WT brains appeared to only produce CXCL12 in rounded structures, morphologically consistent with endothelial cell vasculature (Figure 3D). By flow cytometry, we also saw a decrease in brain endothelial CD31⁺ cell frequencies and reduced CXCL12 production by endothelial cells in 5XFAD brains (Figures S12B and S12C). In the colon, we found unchanging CXCL12 expression in fibroblasts and epithelial cells—a major source of colon CXCL12⁷⁹—and a significant increase in myeloid (CD11b) CXCL12 in 5XFAD colons (Figure S12D), whereas each population’s frequency and cellularity remained unchanged (Figure S12E). Also, a trending increase in total adaptive immune cell numbers (CD45^hi) was seen in 5XFAD brains (Figure S12F), with no changes in overall CD19 B cells and an increase in CD3 T cell frequencies (Figure S12G), neither of which was a source of CXCL12.

Figure 3. Increased CXCL12 production by AD glial cells correlates with brain infiltration of CXCR4⁺ B cells and gut-specific IgA⁺ cells in the dura mater.

(A) Representative plots of flow cytometric assessment of myeloid lineage cells in the brain parenchyma (left). Frequencies and total cell numbers of microglia marked using CD45^medCD11b⁺CD11c⁺ population in WT (circle) and 5xFAD (square) 12-month-old female mice; n = 5 per group.

(B) Representative histograms (left) and quantification (right) of CXCL12 gMFI within the microglia population.

(C) UMAP of re-analyzed human scRNA-seq data from Zhou et al. (left) with CXCL12 transcript levels (center) and violin plots showing p values measured by one-tailed rank-sum test (right) from healthy and AD prefrontal cortex tissue.

(D) Representative immunofluorescence staining of brain hippocampal regions with intracellular CD45 (red) and CXCL12 (green) antibodies in WT (top) and 5XFAD (bottom) 12-month-old female mice. CXCL12 expression in endothelial structures (gray arrow), CD45⁻ astrocyte-shaped cells (yellow arrow), and CD45⁺ microglia (white arrow) are highlighted. Scale bars: 200 μM.

(E) Frequencies of CD19⁺ B cell subsets evaluated by B220 and CD43 expression using flow cytometry analysis of brain and dura mater cells from WT (circle) and 5xFAD (square) 12-month-old female mice; n = 5 per group.

(F) Frequency of total IgA⁺ (left) and IgA⁺CXCR4⁺ (right) cells within the brain parenchyma of WT (circle) and 5XFAD (square) 12-month-old female mice measured using flow cytometry; n = 5 per group.

(G) Flow cytometric assessment of frequencies of CXCR4⁺ B cells in the brain using CD19⁺ subsets.

(H) Flow cytometric assessment of frequencies of CD19⁺B220⁺CD43⁻ SP B2 subsets.

(I) UMAP of re-analyzed mouse brain CD45 scRNA-seq analysis from Su et al. (top left) with Cxcr4 transcript levels (left) and violin plots showing p values measured by one-tailed rank-sum test (right) in 8-month-old female WT and 5XFAD mice.

(J) Commensal ELISpot assessment of number of IgA and IgG ASCs in the dura mater of 9-month-old female mice; n = 8 per group.

Error bars represent means ± SEM; pairwise comparisons were conducted using Student’s t test with a Welch’s correction; p values under p = 0.2 have been specified and rest denoted as ns. *p < 0.05, **p < 0.01, ****p < 0.0001.

Overall, our findings showed a significant increase in CXCL12 expression in the 5XFAD brain myeloid cells, a suspected aberrant expression pattern in brain endothelial cells, and an increase in colon myeloid CXCL12, pointing to a pathology likely impacting trafficking of CXCR4⁺ cells in the colon and brain. The loss of CXCR4⁺ IgA⁺ ASCs in the colon was not observed in 5XFAD female mice at 5 months of age (Figures S13A–S13F), and the aberrant expression of brain CXCL12 was trending upward (Figures S13G–S13M). In male 5XFAD mice, by 9 months, both significant loss of frequencies of gut CXCR4⁺ IgA⁺ ASCs and increased brain myeloid CXCL12 were observed (Figures S14A–S14L). This suggests that, whereas brain myeloid activation may start early, the observed gut phenotype is a function of late-stage accumulation and is consistent in both sexes.

Next, we sought to investigate B cells and ASCs within the brain and brain border regions. We found increases in B220⁻CD43⁻ (DN) CD19⁺ B cell frequencies within the AD brain and dura mater (Figure 3E), as well as a trending increase in frequency of IgA⁺ and IgA⁺CXCR4⁺ ASCs expected to be of mucosal origin (Figure 3F). Within the 5XFAD brain parenchyma, flow cytometry analysis showed increased frequencies of CXCR4⁺ B cells, including in both B220SP and DN subsets (Figure 3G). We also observed a near-significant increase in class-switched B220SP B2 cells in the brain, along with a decrease in naive B cells (Figure 3H). Data mining of 5XFAD brain CD45⁺ cell scRNA-seq data from Su et al. (Figures 3I and S15A) and Keren-Shaul et al.⁴ showed that brain-infiltrating B and T lineage cells also express Cxcr4 mRNA at higher levels compared with WT (Figure S15B), consistent with our findings.

Given the reduction of colon CXCR4⁺ B cells and the presence of such cells within the CNS in the 5XFAD mice, we next determined if any CNS-infiltrating B cells and/or ASCs are potentially linked to the gut. Thus, we performed a gut commensal enzyme-linked immunospot (ELISpot) assay, which identifies ASCs that recognize gut-derived commensal antigens²⁷ in the dura mater of mice. Remarkably, we found a significant increase in IgA⁺ and a near-significant increase in IgG⁺ gut-specific ASCs in the 5XFAD dura mater (Figure 3J). Adoptive transfer of colon immune cells from donor CD45.1 mice into 5XFAD via intraperitoneal (i.p.) injection allowed us to mimic gut-brain migration as previously established using labeled macrophages⁸⁰ (schematic; Figure S16A). By flow cytometry (gating strategies; Figures S16B–S16D), we found that IgA⁺ ASCs compared with CD3⁺ T cells preferentially travel into the 5XFAD brain with a smaller proportion of IgA⁺ immune cells found in the dura mater (Figures S16E and S16F). Additionally, we hypothesized that inhibition of CXCR4 using AMD3100 would block gut-brain IgA⁺ cell migration via CXCL12 chemotaxis in advanced-stage 5XFAD mice⁸¹ (schematic; Figure S17A). As expected, we saw a trending reduction in IgA⁺CXCR4⁺ cell frequencies in the brain; this change was coupled to a significant increase in the frequency of IgA⁺CXCR4⁺ cells in the colon (Figures S17B and S17C) representing increased retention of locally derived gut IgA. Along with trending increases in IgA⁺ frequencies in the blood (Figure S17C), we found near-significant increases in IgA levels in the serum by ELISA with no corresponding changes in serum IgG1 levels or fecal Ig concentrations upon AMD3100 treatment (Figure S17D). The retention of colon-derived CXCR4⁺IgA⁺ cells and increased serum IgA levels during AMD3100 treatment may represent cells that could have migrated to higher concentrations of CXCL12 observed in the brain. Induction of brain CXCL12 expression; brain enrichment of CXCR4⁺ B cells and IgA⁺ cells in the dura, including commensal-specific IgA⁺ cells; and CXCR4-dependent migration of gut IgA cells to the brain may implicate the long-range action of this chemokine on the gut-brain axis.

Inulin fiber improves features linked to AD and associated gut and brain parameters in 5XFAD mice

Given the increased accumulation of DN B cells and IgA⁺ ASCs expressing gut migratory markers in the brain, coupled to a reduction of CXCR4⁺ B cells in the colon of 5XFAD mice, we next tried to target this axis by boosting potentially protective gut IgA⁺ ASC levels in the colon. During homeostasis, intestinal IgA⁺ ASCs produce IL-10 in addition to opsonizing commensal antigens, acting to maintain microbial homeostasis in the gut.⁸² IgA ASC migration into the CNS is thought to attenuate neuroinflammation,^27,28 whereas its loss in the colon might promote gut dysbiosis and an eventual loss of regulatory IgA responses. We, therefore, predicted that boosting IgA responses in the gut would simultaneously reverse the leaky gut phenotype predicted in AD intestines^32,33 and dampen neuroinflammation in AD. Fiber diet interventions are thought to boost systemic T_regs levels via gut microbiome-derived metabolites like SCFAs,^83,84 bile acids,^85,86 and indole-3-proprionic acid (IPA).⁸⁷ In turn, T_regs are thought to boost IgA levels via their action on B cell isotype switching.^88,89 Inulin specifically has been shown to promote such type 2 immune responses via microbial metabolite production.³⁶

Given their link to gut IgA, we first investigated tolerogenic transcriptional signatures in T_regs in the baseline 5XFAD colon using our scRNA-seq data. The thymus-derived HELIOS⁺FOXP3⁺ T_reg subcluster showed lower levels of Il10, Tgfb, and Ctla4 mRNA, whereas microbiome-derived HELIOS⁻ T_regs were transcriptionally unchanged compared with WT⁹⁰ (Figure S18A). Although we found no major changes in total T cells or T_regs within the colon by flow cytometry (Figures S18B and S18C), T_reg frequencies were markedly reduced in the spleen, brain, and blood (Figures S18D–S18F). Interestingly, colon T_regs were partially CXCR4⁺, and these cells showed a significant reduction in CXCR4 expression in 5XFAD colons (Figure S18G). These T_regs may also migrate toward a CXCL12 gradient in the brain in a loop to attenuate neuroinflammatory gliosis.^91,92 Assessment of microbiome modulation of T_regs is discussed below.

To test the effect of fiber diet interventions on AD pathology, we fed mice with control (4.7% cellulose), cellulose-enriched (17%), or inulin-enriched (17%) diets from 8 weeks of age until the end of the study at 15 months (schematic; Figure 4A). Because cellulose is an inert insoluble fiber, in contrast to inulin, this was used as an additional control diet as performed previously.⁹³ An enlarged cecum in inulin-fed mice, compared with both cellulose and control diets, was observed (Figure S19A), consistent with the expected microbial metabolism of these fibers. Compared with control and cellulose-fed mice, inulin-fed mice showed significant improvements in frailty in both WT and 5XFAD groups, including marked changes in body weight (Figures 4B, 4C, S19B; statistics: Data S4). Via the open-field behavioral assay, significant or near-significant increases in anxiety were observed in control and cellulose-fed 5XFAD compared with WT mice, which were not significant within inulin diet conditions (Figure S19C, left). Nonetheless, the overall beneficial effect of inulin in this behavioral assay was minimal, which may be confounded by age-related motor deficits by 15 months of age.⁹⁴ No changes were detectable via the Y-maze assay (Figure S19C, right). Quantification of Aβ plaques via immunohistochemistry (IHC) in the hippocampus and cortex were conducted using antibodies against two different Aβ epitopes and showed promising but inconsistent reductions in the magnitude and significance of Aβ burden in inulin-fed mice (Figures S20A–S20D). However, inulin diet did show significantly reduced immune cell (CD45⁺) frequencies in the brain (Figures 4D and S19D). In inulin-, but not control or cellulose-fed mice, we observed an induction of IgA⁺ cells in both WT and 5XFAD conditions in both the colon and the spleen (Figures 4E and 4F). As expected, these IgA⁺ cells were CXCR4⁺ (Figure S19E). We also observed an induction of spleen and colon T_reg frequencies in the inulin-fed but, not in cellulose or control, diet conditions (Figures 4F and 4G).⁹⁵ Numbers of RORγt⁺ T_regs, representing the microbiome-induced pT_reg population, were increased in inulin-fed colons compared with both other diets (Figure 4H). Interestingly, the inulin diet also resulted in a trending reduction in CXCL12 levels in 5XFAD brains as measured by IHC, co-localizing with IBA1⁺ microglia, CD45⁺, and other cells (Figures 4I and S19F), implying both local and systemic effects.

Figure 4. Inulin fiber rescues various AD parameters while boosting colon IgA and T_reg levels and dampening gliosis-associated CXCL12 in the brain.

(A) Schematic of dietary intervention study.

(B) Thirty-one-point frailty index score of mice at 12 months.

(C) Body weight of WT and 5XFAD 12-month-old female mice on control (circle), cellulose (square), and inulin diets (triangle); n = 4–5 per group.

(D) Quantification of cellular CD45 staining by IHC normalized to total area; n = 3 per group.

(E–H) Flow cytometric quantification of colonic and splenic immune cell frequencies of IgA⁺, total FOXP3 ⁺ T_regs, and FOXP3⁺RORgt⁺ T_regs at endpoint as indicated; n = 4–5 per group. Bar graph data represent means ± SEM. Statistical analysis was conducted using two-way ANOVA and Tukey post hoc test, and p values under 0.25 have been shown. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

(I) Quantification of CXCL12⁺ cells in the brain by IHC normalized to total area; n = 3 per group. Bar graph data represent means ± SEM and pairwise Student’s t test with Welch’s correction was used for statistical analysis.

(J) Fecal 16S rRNA sequencing showing the top 30 microbial families in each diet condition as indicated.

Microbiome 16S sequencing analysis showed clear separation of dietary intervention by principal-component analysis (Figure S21A). It identified most alpha diversity in inulin-fed feces by the Simpson index (Figure S21A). The Bifidobacteriaceae family (Figure 4J), part of the Actinobacteria phyla (Figure S21B), and the Erysipelotrichaceae family (Figure 4J) were among the most supported by the inulin diet. A trending expansion of the Bifidobacterium genus was observed in control versus inulin-fed 5XFAD mice and was nearly lost within the cellulose diet groups (Figure S21C), whereas Akkermansia, reduced in 5XFAD versus WT control mice, remained unchanged between diets (Figure S21C). Because Bifidobacterium produces anti-inflammatory metabolites that impact the gut barrier, gut immune composition, and cognition,^38,96–98 we conducted targeted metabolomics on serum from diet-fed mice. We observed induction of SCFAs like propionic acid and valeric acid, bile acids like ursodeoxycholic acid and muricholic acid isomers, as well as IPA in inulin compared with control diet-fed mice (Figures S22A–S22C), all of which have been shown to drive the observed T_reg induction in 5XFAD mice,⁹⁹ which has been linked to the aforementioned IgA isotype switching.

Thus, the inulin diet may be one intervention that can modulate gut microbial composition and associated gut microbial metabolites; boost regulatory immunity within the large intestine, including T_regs and IgA⁺ ASCs; reduce activated chemokine migratory axes within the brain; protect against frailty; and potentially alleviate some neurodegenerative manifestations of AD.

DISCUSSION

Changes in gut immunity and microbial homeostasis have been reported in neuroinflammation and brain injury.^21,22,100 However, the nature of gut-brain crosstalk in AD is not well understood. Here, we examined the gut immune landscape in AD using a pre-clinical Aβ-AD model, the 5XFAD mouse. At the global level, we found pan-immune transcriptional changes in the colon, including in genes involved in mitochondrial stress and immune activation. Specifically, in B lineage cells of the colon, we saw a reduction of CXCR4^hi ASCs and increased B2 class switching in both female and male mice, collectively reflecting gut B cell dysfunction in the context of AD. Our results also indicate that whereas brain CXCL12 myeloid dysfunction starts at the early stages (5 months), the gut CXCR4⁺ ASC pathology manifests later (12 months), possibly due to gut B cell-replenishing mechanisms. Because humoral immunity is a key player in maintaining gut microbial homeostasis,¹⁰¹ we suspect that these B lineage changes are closely linked to observed dysbiotic microbiome shifts.^102,103

Although the migration of gut ASCs into the brain and brain border regions is emerging as a novel component modulating neurological disease,^27,28 its potential involvement in AD has not been fully explored. In our colony of 5XFAD mice, we found a striking loss of colonic ASCs with a migratory signature. This migratory signature suggests an ability to respond to various signals, including CXCR4/7 ligands, S1P, BAFFR ligands, FFAR1-binding medium-chain fatty acids, and GPR183-binding oxysterols. Some of these have been linked to intestinal IgA homing¹⁰⁴ and resilience in AD. For example, GPR183-mediated migration¹⁰⁵ may be affected by APOE2-mediated oxysterol production by the brain.¹⁰⁶ Here, we focused on CXCR4 as we observed transcriptional downregulation of Cxcr4 as well as CXCR7 (Ackr3) in gut ASCs, a factor that increases CXCR4 responsiveness to CXCL12 in B cells.⁸⁴ Furthermore, strong CXCR4 expression has been reported on IgA⁺ cells found in the meningeal dura mater.²⁸

Within the 5XFAD brain and colon, we found a significant increase in CXCL12 levels, the cognate binding partner for CXCR4. Increased CXCL12 levels were observed in 5XFAD brain myeloid cells, in both sexes and trending at an earlier stage of 5 months of age in females. We also suspect CXCL12 production in other glial cells like astrocytes at the hippocampal formation, which requires more direct investigation. The accompanying reduction in endothelial CXCL12 may reflect the reported degeneration of the blood-cerebrospinal fluid (CSF) barrier in AD¹⁰⁷ and mirror CXCL12 polarization-mediated trafficking of cells into the brain parenchyma as suggested by MS studies.¹⁰⁸ The smaller CXCL12 induction in colon myeloid cells, albeit accompanied with a trending decrease in myeloid cell frequency, may represent CXCL12 dysfunction in AD. Although CXCL12 expression has been observed in other models in the colon¹⁰⁹ and bone marrow,¹¹⁰ we expect heightened brain expression of CXCL12 in 5XFAD to partially redirect CXCR4 trafficking, including potentially circulating gut immune cells, into the AD brain parenchyma from the periphery. Already recognized as a key mediator of AD-associated neuroinflammation, gliosis may also activate a CXCL12-CXCR4 migratory axis signaling gut ASC migration into the brain and border regions. Using commensal ELISpot assays, we confirmed increased levels of gut-specific IgA⁺ ASCs in the 5XFAD dura mater. We also found increased CXCR4⁺ B cells within the 5XFAD brain parenchyma. CNS-infiltrating B cells have been linked to AD pathophysiology.¹¹¹ The role of gut IgA ASCs in the CNS is likely multifactorial, including potential protective or compensatory effects on inflammation.^27,28 The CXCL12-CXCR4 axis has been implicated in B cell development⁷⁴ and ASC trafficking toward inflammatory sites,^110,112 as well as in facilitating crosstalk between dura fibroblasts and immature B cell progenitors in the skull bone marrow.^75,76 Additionally, we found increased CXCR4 expression in other immune cell subtypes in the brain, suggesting that CXCL12-mediated recruitment is not likely limited to gut-derived ASCs and can include myeloid and T lineage cells. Interestingly, changes in CXCL12 expression in α-synuclein-mediated Parkinson’s disease¹¹³ and MS¹⁰⁸ have been reported, suggesting this axis may be a key feature shared by several neuroinflammatory conditions.

Using IgA commensal ELISpot, gut CD45.1 cell adoptive transfer, and small-molecule CXCR4 inhibition with AMD3100, we provide evidence that gut IgA⁺ immune cells can migrate to the brain and that migration can be modulated by targeting the CXCL12-CXCR4 axis. Although our CXCR4 inhibition involved targeting the gut and periphery using i.p. administration, this inhibition via intrathecal or intracisternal administration may further elucidate mechanisms of this axis. Other related targeting like CXCR7 blockade¹¹⁴ or CXCL12 (e.g., NOX-A12^115,116) would additionally validate gut-brain migration and determine the neuroprotective role of ASCs. Further experimentation with this axis may also inform potential therapeutic targets and/or adverse reactions to B cell targeting in AD.

Another finding of our work is the presence of multiple markers of potentially low-grade systemic inflammation within the blood and spleen of 5XFAD mice. We found increased B2 cell class switching throughout the periphery and a systemic reduction of T_reg levels at the investigated time points (9–12 months of age). A similar B cell phenotype has been reported in human peripheral blood mononuclear cell (PBMC) studies, implicating a change in BCR repertoire with AD pathology.⁷¹ Although some BCRs are expected to generate Aβ auto-antibodies,¹¹⁷ it remains to be seen whether these BCRs are specific to gut microbial antigens, including microbial amyloids¹¹⁸ in AD, accumulating in the periphery due to a leaky gut.^20,31,32

To determine whether rescuing inflammatory phenotypes within the AD gut could constitute a potential therapeutic, we utilized the dietary supplement, inulin. Inulin is a soluble prebiotic fiber and, in our study, altered serum levels of bioactive metabolites, including SCFAs,¹¹⁹ bile acids,³⁶ and IPA,³⁹ correlating to an increase in Actinobacteria phyla and Bifidobacteriaceae family species among some others. We also saw an increase in Erysipelotrichaceae, which has been implicated in IgA coating¹²⁰ and sex-specific differences in other AD models via SCFA production.¹²¹ Bifidobacteria is a well-studied probiotic shown to produce the abovementioned metabolites^122,123 with a known impact on T ¹²⁴ and consequences for IgA isotype switching.⁸⁸ In our model, consumption of dietary inulin resulted in protection against various features of AD, matching inulin studies performed in the APOE AD model, which showed protection³⁹ as well as in human cognition.³⁸ Within the local colon immune system, we found an increase in T_reg frequencies in 5XFAD mice systemically and restored pT_reg levels in the colon upon inulin feeding, as previously reported.⁸³ Additionally, we found increased gut IgA⁺ cell levels in inulin-fed mice, likely driven by T_reg-mediated effects on IgA isotype switching.^{88,89,125,126} Because these impacts were not seen in cellulose diet-fed mice, results suggest that microbial fermentation of inulin leads to modulation of gut T_regs and IgA⁺ cells. These IgA⁺ cells were CXCR4⁺ that, as alluded to earlier, may migrate toward CXCL12 signals from the AD brain. Interestingly, we also observed a trending reduction in brain CXCL12 production upon inulin feeding, predicted to be at least partially driven by systemic SCFA accumulation and reduced brain inflammation.^23,24 SCFAs accumulating in the periphery can act directly on brain microglia via SCFA transporters^24,127 and TREM2 activation.¹²⁸ The impact of inulin-derived metabolites on astrocyte CXCL12 expression is not fully understood. The anti-inflammatory effects of inulin-derived metabolites may also include gut epithelial remodeling,⁹³ induction of regulatory ILC2s in the gut,³⁶ and protection via brain T ¹²⁹ that likely acted in concert in our model. The protection conferred by inulin in the gut may also attenuate systemic inflammation driven by leaky gut antigens.⁹³ Cellulose showed trending improvements in the frequency of CD45 cells in the brain, suggesting a potentially independent mechanism, which warrants future investigation. Overall, inulin diet rescued some parameters of AD, likely via multiple mechanisms acting on the gut, the brain, and the peripheral immune cells.

Our data support a model whereby AD facilitates gut and systemic inflammatory alterations in association with increased CXCL12 production from CNS-resident myeloid and glial cells. More work is needed to determine the relative contributions of local versus systemic and gut-related inducers of glial CXCL12. Increased CXCL12 in the CNS, in turn, may promote migration of immune cells, including B cells with inflammatory capacity, as well as IgA⁺ gut ASCs with the potential to dampen or promote inflammation. The reduction of B cell subsets from the colon potentially compromises local intestinal defenses, facilitating reported gut microbial changes linked to AD progression.^32,70,102 These processes are targetable by dietary supplements, including soluble fiber that inhibits the chemokine cascade and restores gut and systemic immune regulatory cells, resulting in the dampening of aspects of AD disease progression. Our investigation provides a better understanding of colon immune responses in AD and reveals key links to neuroinflammation. Through this investigation, we provide further evidence for the importance of the gut-brain axis, including the role of gut immune cells, and reveal possible dietary and therapeutic interventions to target inflammation in AD.

Limitations of the study

Although gut ASCs are expected to attenuate inflammatory responses to commensal species within the gut, the protective nature of brain and dura mater migrant gut ASCs remains unclear and requires further study. In autoimmunity, IgA⁺ ASCs produce IL-10 to attenuate neuroinflammation.²⁷ In AD, intravenous immunoglobulin delivered with a BBB-penetrating treatment suggests that antibodies have a protective effect.¹³⁰ Depletion of immature and mature B cells using biologics also appears to be protective in neurological conditions, including AD.^111,131 Temporal signals acting on heterogeneous B cell subsets differentially may account for these observations. The BCR repertoire in gut ASCs may also drive differential inflammatory effects upon brain migration. Gut IgA⁺ ASCs have isoform-dependent heterogeneity in the engagement of Fcα receptors that can differentially modulate inflammation in humans.¹³² This implicates, for example, Trichomonas levels¹³³ and previous Salmonella encounters¹³⁴ in certain individuals and their baseline IgA repertoire. The role of BCRs against microbial amyloids,¹¹⁸ as well as brain Aβ, encountered via gut-CSF encounters³⁹ is also poorly understood. Studying brain migratory gut-derived ASC antigen specificities would allow determination of whether and which ASCs are protective.

Our experimental model and approach also posed some inherent limitations. Although we did not observe measurable Aβ plaques in the 5XFAD colon, the Thy1 promoter is expressed in varying degrees in enteric neurons, immune cells like T cells, and hematopoietic stem cells. Therefore, we cannot rule out that low levels of Aβ expression could have impacted our observed phenotype during the disease process. Corroborating our Aβ-AD results in other AD mouse models would confirm this phenotype and inform the role of MAPT, APOE, and TREM2 genes on gut-brain AD pathology and prebiotic interventions. As well, the impact of regional microbial diversity and resulting immune changes throughout the GI tract,¹³⁵ including the mouth and small intestine, requires further study to provide a full picture of gut-neuroimmune interactions in AD. Additionally, as our scRNA analysis appeared enriched in adaptive cells and informed our focus on B lineage phenotypes, the migratory and inflammatory role of colon myeloid cells in 5XFAD may also require more in-depth study beyond our flow cytometry workup.

The route of migration of gut immune cells into the brain in AD remains enigmatic. Possible routes include the BBB, the meningeal lymphatics, and/or the vagus nerve.¹³⁶ Although we find trending increases in DN B cells and IgA⁺ cells in 5XFAD blood, it is unclear whether these are mainly driven by the systemic accumulation of gut bacterial antigens via a leaky gut,^31,32 their migration via the bloodstream,¹³⁷ or some combination of the two. Gut-specific IgA⁺ ASCs within the dura mater imply that lymph-CSF barrier dysfunction might be at play, already associated with AD.¹³⁸ Immune cells within human AD CSF also show reduced CXCR4 expression,¹³⁹ suggesting possible infiltration into tissue, although early reductions reported in CSF CXCL12 levels in AD suggest a temporal effect.¹⁴⁰

Our diet study was conducted using a different dose (~17%) and longer duration (13 months) of prebiotic fiber treatments than in other models^36,39 to assess its impact on long-acting changes on the 5XFAD gut immune system, neuroinflammation, and disease features. Of note, we observed an inconsistency in the magnitude and significance of improvement in Aβ plaques using two different antibodies against Aβ, and thus, more studies are needed to definitely draw any conclusions on the capacity of inulin to reduce Aβ burden in the brain. Because such protection via behavioral assays and Aβ plaque data was only modest or questionable in our model, partially influenced by aging processes, we believe that studies on the impact of varying durations and doses of inulin are required to inform the most impactful level of inulin treatment. This could also be coupled to measurement of localized Aβ plaque levels¹⁴¹ by different antibodies, a larger cohort, and behavioral assays including the Morris water maze that measure hippocampal learning and memory behavior and are well established in AD mouse models.^39,142

Here, we only studied the Aβ component of the disease pathophysiology with analyses from a single mouse model of AD. Given the importance of the gut microbiota in AD, it will be important to validate gut immunological signatures across different models and in different facilities. Consistently, because immune responses to extracellular Aβ and intracellular tau are expected to be different,¹⁴³ it is unclear whether these phenotypes are relevant to tau-driven late stages of AD. Finally, although we have provided evidence for some of our conclusions through use of publicly available human data, immunophenotyping of matched human brain and colon tissue would improve the translational conclusions of our work.

RESOURCE AVAILABILITY

Lead contact

Requests for further information, materials, and data should be directed to and will be fulfilled by the lead contact, Daniel Winer (dwiner@buckinstitute.org).

Materials availability

This study did not generate any new materials.

Data and code availability

• The scRNA-seq and 16S rRNA sequencing data have been deposited at the Gene Expression Omnibus under accession GEO: GSE255819 and GSE305697, respectively.

• This paper does not report original code.

• All other raw data are available from the lead contact upon request.

STAR★METHODS

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS

Mice

C57BL/6J (000664) and 5XFAD hemizygous (008730) (B6.Cg-Tg(APPSwFILon, PSEN1*M146L*L28V)6799Vas/Mmjax, MMRRC strain #034848)⁴⁰ mice were purchased from The Jackson Laboratory. The 5XFAD-derived WT littermates were generated via crossing of hemizygous mice and were housed together after weaning and aged until specified harvest dates. Female mice used for the study were age-matched littermates except those used for comparative flow cytometry validation, frailty and diet studies where non-littermates were also used. Male mice used in the comparative studies were age-matched littermates except comparative flow cytometry validation where non-littermates were also used. Mice were maintained under group-housed, pathogen-free, temperature and 12 h light-dark cycle-controlled environments at the Buck Institute for Research in Aging, Toronto Medical Discovery Tower, or Krembil Brain Institute’s vivarium facilities. The animal facility at the Buck Institute for Research on Aging is accredited by AAALAC International (Unit Number 001070). All protocols and procedures described herein were approved by the Buck’s Institutional Animal Use Committee.

METHOD DETAILS

Immune cell isolation

Mice were euthanized using CO₂ inhalation prior to collecting colons, blood, and spleens. After euthanization, blood was collected via cardiac puncture into K2-EDTA tubes (BD, 367855). Spleens were filtered through 70 μm cell strainers to generate single cell suspensions. Blood and spleens were subjected to hemolysis. Full-length colons were dissected from below the cecum and above the rectum, to include the distal, middle, and proximal colon, averaging 7.5 inches in length. Colons were processed using a lamina propria dissociation kit (Miltenyi Biotec, 130–097-410) with the gentleMACS Dissociator (Miltenyi Biotec, 130–093-235). Prior to brain collection, mice were additionally transcardially perfused with 10–20 mL of cold PBS. Mouse whole brains were processed using a brain dissociation kit (Miltenyi Biotec, 130–107-677) with the gentleMACS Octo Dissociator (Miltenyi Biotec, 130–096-427). All cell suspensions were filtered through a 40 μm cell strainer prior to downstream application.

Single-cell RNA sequencing analysis

Cell sorting and sequencing

Single-cell suspensions of colons from 9-month-old littermate WT and 5XFAD female mice were prepared as described above. Cells were stained with LIVE/DEAD Fixable Blue cell stain (Thermofisher, L23105) for viability assessment and blocked with FcBlock with CD16/32 (Biolegend, 101320). Cells were then stained with anti-mouse CD45 (Biolegend, 30-F11). Live CD45⁺ cells were sorted using the FACS Aria sorter at the Buck Institute for Aging Research flow cytometry core facility and collected into RPMI-1640 (Wisent) with 10% FBS (Wisent). Samples were processed using a Chromium Next GEM Single Cell 3′ GEM, Library & Gel Bead Kit v3.1. The resulting libraries were sequenced at UC Davis Genome Center Bioinformatics Core using Illumina NovaSeq 6000 sequencing. The resulting sequencing data were processed using Cell Ranger (6.0.1).

Data processing

The Seurat (v4.0.2) workflow was used to filter and down sample the data; SCTransform was used to normalize and scale, followed by integration, clustering (res 0.9) and plotting using Seurat. LogNormalize was used with a resolution of 0.5 for re-analysis of Su et al. and Zhou et al. data. Differentially expressed genes (DEGs) were obtained utilizing the MAST method in the Seurat function FindMarkers. The adjusted (adj.) p values were based on the Bonferroni correction for false discovery.

Trajectory analysis

To confirm the putative B cell developmental pathways in the colon, cell trajectory analysis was conducted using the R package, Monocle 3 (v1.3.4).¹⁴⁵ The standard Monocle 3 pipeline with default settings was used without pruning for the learn_graph function. Finally, the order_cells function was run using Naive B cells (Cluster 0) as the root of the trajectory.¹⁴⁵

Flow cytometry

Single cell suspensions of specified mouse tissues and blood were stained using the LIVE/DEAD Fixable Blue cell stain (Thermofisher, L23105) for viability assessment and blocked with CD16/32 based FcBlock (Biolegend, 101320) for 20 min at 4°C. Cells were further stained with fluorophore-conjugated antibodies for 30 min at 4°C in the dark. The following antibodies were used CD11b-BUV563 (BD Biosciences, clone: ICRF44), CD11b-BV650 (Biolegend, Clone: M1/70), EpCAM-BV711 (Biolegend, Clone: G8.8), CD19-APCCy7 (Biolegend, Clone: 6D5), IgD-BUV605 (Biolegend, Clone: 11–26c-2a), IgM-PeCy7 (Biolegend, Clone: Rmm-1), CD11c-efluor450 (Invitrogen, Clone: N41B), CD45.2-AF700 (Biolegend, Clone: 104), CD127-BV711 (Biolegend, Clone: A7R34), B220-PEFire810 (Biolegend, Clone: RA3–6B2), CD45-BUV805 (BD Bioscience, Clone: 30-f11), CD23-APCCy7 (Biolegend, Clone: B3B4), CD138-BV650 (Biolegend, Clone: 281–2), CD31-APCCy7 (Biolegend, Clone: MEC13.3), LPAM-1(Integrin α4β7)-PE (Biolegend, Clone: DATK32), CXCR4-BV421 (Biolegend, Clone: L276F12), CD86-AF700 (Biolegend, Clone:GL-1), CD86-PeDazzle594 (Biolegend, Clone:GL-1), IgM BUV615 (BD, Clone: R6–60.2), CD21-BV421 (Biolegend, Clone: 7E9), MHCII (I-A/I-E)-BUV395 (BD, 2G9), CD43-PeCy5 (Biolegend, Clone: 1B11), CD11c-BUV737 (BD, Clone: HL3), KLRG1-BV480 (BD, Clone: 2F1), GL7-PerCpCy5.5 (Biolegend, Clone: GL7), gp38-PerCpCy5.5 (Biolegend, Clone: 8.1.1), CD3-APCFire810 (Biolegend, Clone: 17A2), IgM-BV510 (BD, Clone: II/41), CD45-BUV737 (BD, Clone: 30-f11), F4/80-BV480 (BD, Clone: T45–2342), CD19-BV570 (Biolegend, Clone: 6D5). This was followed by fixation and permeabilization using either the FOXP3 staining buffer set (eBioscience, 00–5523-00) or the Cytofix/Cytoperm kit (BD, 554714). Intracellular staining was done using the following antibodies: BLIMP1-PE (Invitrogen, Clone: 5E7), CXCL12-PE (R&D Biosystems, Clone: IC350P), IgA-BV786 (BD, Clone: C10–1), c-JUN-AF647 (Cell Signaling Technologies, Clone: 60A8), CXCR4-BUV737 (BD, Clone; 2B11), FOXP3-AF488 (Biolegend, Clone: 150D), RORγt-BV421 (BD, Clone: Q31–378), RORγt-PercP-efluor710 (Invitrogen, Clone: AFKJS-9). Stained cells were analyzed on the Cytek Aurora at the Buck Institute’s Flow Cytometry Core facility within 24 h. All raw data were unmixed using SpectroFlo (v3.0) and analyzed using FlowJo (v10.8.1) software.

Frailty scoring

The health span of mice was assessed using a 31-term frailty index to identify humane interventions and endpoints. Non-invasive clinical assessment of 31 physical manifestations of age-related deficiencies and scored based on severity was conducted using previously published recommendations.⁴⁴ Scoring was performed on female and male mice between 3 and 13 months of age, as specified.

Immunohistochemistry (IHC) and immunofluorescence (IF) microscopy

Colon Swiss rolls and perfused whole brains were fixed with 10% buffered formalin for 24–48 h and embedded into paraffin blocks. Formalin-fixed paraffin embedded (FFPE) tissues were sectioned into 5 μm slices and mounted onto slides. Following xylene-based deparaffinization, heat-induced antigen retrieval was performed at pH 6.0 (Biocare Medical, Rodent Decloaker) prior to hematoxylin and eosin (H&E) or immunofluorescence staining. The following primary antibodies were used: rabbit anti-mouse IgA (NSJ Bio-reagents R20169), mouse anti-mouse CXCL12 (R&D, MAB350), rat anti-mouse CD45 (eBioscience, 14–0451-82), rat anti-mouse CD19 (Invitrogen, 13–0194-82), mouse anti-mouse amyloid beta (Clone 6E10, Biolegend, 803004), and rabbit anti-mouse Iba1 (Fujifilm, 019–19741). Appropriate secondaries were used in Alexa Fluor 488, 647 or 555 as specified. The following conjugated antibodies were used: rabbit anti-mouse c-JUN AF647 (Cell Signaling Technology, 40502) and rat anti-mouse B220-PE (Biolegend, 103207). Mouse-on-mouse blocking reagents (Vector Laboratories) were used where required. After DAPI (Sigma) staining and mounting (ProLong Gold Antifade Mountant, ThermoFisher, P36930), the Zeiss Axioscan 7 microscope slide scanner was used to collect 20X images. Images were processed using Zen Blue lite (Zeiss) software for background normalization and Fiji¹⁴⁴ for quantification.

Spleens were placed in a histology tray, embedded in OCT (Fisher Healthcare), and frozen in 2-methylbutane cooled on dry ice. Trays were wrapped in aluminum foil and stored at −80°C until further use. Frozen tissues were subsequently sectioned with a cryostat to 5–6 μm in preparation for acetone fixation and staining. Defrosted and PBS-re-hydrated tissues were subjected to FcReceptor blocking (BD Pharmingen, 553141) and primary antibody staining with rat-anti-mouse antibodies. Data depicted were obtained by staining with B220-AF647 (Biolegend, 103226) and CD4-PE (Biolegend, 100512). After DAPI (Sigma) staining and mounting (ProLong Diamond, ThermoFisher, P36965), spleen slides were visualized by microscopy using the Zeiss AxioObserver with Zen Blue (Zeiss) software. Images were processed using Fiji.¹⁴⁴

Behavioral assays

Open field test

Mice were placed individually in total darkness into Coulbourn Instruments TruScan Photobeam Sensor E63–12 activity cages (26 cm. length × 26 cm. width × 39 cm. height, Harvard Biosciences, Holliston, MA). Cages were equipped with rows of Coulborn TruScan Photobeam Linc infrared photocells (Harvard Bioscience, Holliston, MA) interfaced with TruScan 2.070 tracking software (Harvard Bioscience, Holliston, MA). After a 15-min adaptation period in the testing room, open field activity was recorded for 20 min. Recorded beam breaks were used to measure total path lengths in the margin versus the center of the cage, along with active times and rearing events (stereotypy).

Y-maze

Mice were placed in the center of a small Y-maze (each arm length: 35 cm), and spontaneous alternation was recorded in a single continuous 10-min trial using EthoVision XT 15.0 tracking software (Noldus Technology, Leesburg, VA). Each of the three arms was designated a letter A–C, and entries into the arms were recorded. The percent of spontaneous alternation was calculated over the total number of entries possible.

Dietary interventions

Fiber diet studies were conducted on age-matched female WT and 5XFAD mice fed ad libitum from 8 weeks to 15 months of age with the following diets. Mice were fed cellulose diet (D13081109i, Research Diets Inc.) containing 16.6% cellulose and no inulin; or inulin diet (D13081108i, Research Diets Inc.) containing 17.3% inulin and no cellulose; or a control diet (D12450Ji, Research diet) containing 4.7% cellulose comparable to standard chow.

Commensal enzyme-linked immunosorbent spot assay (ELISpot)

Dura mater was scored from the skull and dissociated with Collagenase P and DNase I in 2% FBS for 20 min at 37°C. After neutralization with 2 mL PBS, cells were counted to perform the assay. The ELISpot assay was conducted as previously described.²⁷ Briefly, membrane plates (Millipore Sigma, MSIPS4W10) were coated with 100 μL/well of allogenous heat-killed fecal matter (1mg/mL) diluted 1:10 and placed at 4°C overnight. Plates were then blocked with 100 μL/well of 10% FBS/RPMI (Sigma) for 2 h at 37°C. Starting with 1×10⁶ cells, 200 μL/well of single-cell suspension solutions in FBS/RPMI were loaded onto the plate in serial 2-fold dilutions and left overnight at 37°C. Cells were discarded the next morning, and plates were washed with 0.1% Tween 20 in PBS. Subsequently, HRP-conjugated IgA and AP-conjugated IgG detection antibodies were added and incubated for 2 h at 37°C. Plates were developed using AEC Peroxidase (for HRP-conjugated Ab) (Vector Laboratories, Cat.# SK-4200) or Vector Blue (for AP-conjugated Ab) (Vector Laboratories, cat. #SK-5300) substrates to detect IgA and IgG ASCs, respectively. After drying, spots were counted using a light microscope and quantified with respect to the original cell concentration.

Enzyme-linked immunosorbent assay (ELISA)

Tissue Aβ ELISA

Colons from 12-month female mice were normalized by weight (1:20; 50 mg/1000 μL) and lysed using T-PER buffer (ThermoFisher Scientific, Cat# 78510) which was supplemented with SIGMA FAST, EDTA-Free protease inhibitor cocktail (Millipore Sigma, Cat# S8830–2TAB). The LEGEND MAX Human Amyloid β (1–42) ELISA Kit (Biolegend, Cat#448707) was used to quantify Amyloid β using a mouse brain lysate positive control.

Fecal and serum Ig ELISA

Fecal pellets were resuspended in 1X PBS at 1 mg/μL and centrifuged at 16,000 × g. Blood from cardiac puncture was coagulated and spun down at 16,000 × g to collect serum. Supernatants were used as samples for ELISA. Unlabelled mouse IgA (0106–01, Southern Biotechnologies) or IgG1 (Southern Biotechnologies, Cat#0102–01) were used as standards. ELISA plates (Thermo Scientific, Cat#430341) were coated with a 1:1000 dilution of anti-mouse Ig capture antibody (Southern Biotechnologies, Cat#1010–01) and incubated overnight at 4°C. Plates were blocked with 1% BSA in 1X PBS (100 μL/well) for 2 h at 37°C. Samples and standards (50μL/well) were added to the plate and incubated for 2 h at 37°C. Biotin-conjugated anti-mouse secondary detection antibodies (50 μL/well) for IgA (Southern Biotechnologies, Cat#1040–08) or IgG1 (Southern Biotechnologies, Cat#1070–08) were then added at a 1:5000 dilution and incubated for 1 h at 37°C. HRP-conjugated streptavidin (R&D Systems, Cat#890803) was then added at a 1:200 dilution and incubated for 30 min at 37°C. To develop plates, 50 μL/well of TMB substrate (Invitrogen, Cat#00–4201-56) was added to wells and reactions were stopped with 50 μL/well of 1N H₂SO₄. All dilutions were all done in 0.1% BSA in 1X PBS. Washes were conducted with 0.01% Tween 20 in 1X PBS. Plates were read using the SpectraMax i3 (Molecular Devices). Absorbance at 570 nm was subtracted from absorbance at 450 nm and this difference was used to plot a standard curve from which sample concentrations were interpolated.

16S fecal microbiome sequencing

Fecal pellets from 15-month-old mice were subject to the Quick-DNA Fecal/Soil Microbe Microprep kit (Zymo Research. Cat. #D6012). Isolates were sent for a 16S rRNA sequencing service (Novogene, Cat.# PE250). Raw reads were processed through a previously published R pipeline employing DADA2.¹⁴⁶ Taxonomic assignment was conducted using the latest RDP training set (v18)¹⁴⁷ and results were integrated into a Phyloseq object.¹⁴⁸ The Simpson index alpha diversity assessment was conducted using the Kruskal-Wallis test with Dunn’s post-hoc analyses (B-H correction). Beta diversity was evaluated using PCoA of Bray-Curtis dissimilarity and differential abundance across experimental conditions and diets.

Targeted serum metabolomics analysis

Sample preparation

Serum samples (30 μL) were used for targeted metabolomics analysis. SCFAs were analyzed using a derivatization-based approach with internal standards (13C-acetic acid, d7-butyric acid), followed by protein precipitation with acetonitrile. The supernatant was derivatized with 3-nitrophenylhydrazine (3-NPH) and 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide (EDC) in acetonitrile, incubated at 40°C for 30 min at 400 rpm. For bile acids and tryptophan-related metabolites, extraction was performed using 80% methanol containing an internal standard (d4-cholic acid). Samples were vortexed and incubated (−20°C, 20 min). These and derivatized samples were dried, reconstituted in acetonitrile:water (4:1), centrifuged (18,000 × g, 10 min, 4°C), and analyzed by LC-MS.

LC-MS analysis

Targeted metabolomics was performed using a Thermo Q Exactive Orbitrap mass spectrometer coupled to a Vanquish UHPLC system (Thermo Scientific, USA). SCFAs were analyzed in positive ion mode using a Waters Atlantis Premier BEH C18 AX column (1.7 μm, 2.1 mm × 100 mm) with a 30-min gradient of water/acetonitrile (0.1% formic acid) at 0.2 mL/min (55°C). Bile acids and tryptophan-related metabolites were analyzed in negative and positive ion modes, respectively, using an Accucore Vanquish C18+ column (1.5 μm, 2.1 mm × 100 mm) with a 30-min gradient at 0.3 mL/min (40°C).

Data acquisition and analysis

Data were acquired using targeted single ion monitoring-parallel reaction monitoring (tSIM-PRM) with a resolution of 70,000 (MS1) and 17,500 (PRM). MS parameters included an AGC target of 1e6 (MS1) and 1e5 (PRM), IT of 100 ms, and scan range of 50–500 m/z. PRM fragmentation used stepped normalized collision energy (20, 40, 65). Metabolites were identified by matching retention times, m/z values, and characteristic fragments with reference standards and databases (Sigma and HMDB). Peak areas were processed using Thermo Quan Browser for relative quantification.

Gut-brain migration assays

Adoptive transfer

Colons from 8 to 10-week-old female CD45.1 JAXBoy mice (Jackson Laboratory, C57BL/6J-Ptprc^em6Lutzy/J, 033076) were processed as described above and subjected to dead cell removal (Miltenyi Biotec, Cat #130–090-101). These donor cell preparations were immunophenotyped by flow cytometry. Pooling 3–4 mouse colons provided 4X10⁶ cells and injected using intraperitoneal (I.P.) injection into 14- to 18-month-old female 5XFAD recipient mice. Following 48 h for migration via I.P., as described in Haney et al.,⁸⁰ recipient mouse brains and dura mater were collected for analysis by flow cytometry.

CXCR4 inhibition

AMD3100 (Sigma, Cat# 239820) or PBS was administered to 7–9-month-old female 5XFAD mice on days 1, 3, 5, 8, 9, and 10 via intraperitoneal (I.P) injection. As previously described,⁸¹ a dose of 1 mg/kg/day and concentration of AMD3100 100 μg/mL in 1X PBS was used. Mice were sacrificed by CO₂ asphyxiation 3 to 4 hours after injection on day 10 and feces, blood, brain and colon tissues were collected for analysis by flow cytometry and ELISA.

QUANTIFICATION AND STATISTICAL ANALYSIS

All data are shown as the mean ± standard error of the mean (SEM) and analyzed using GraphPad Prism Software (v9.4.0) except 16S microbial sequencing data. Pairwise statistical difference between two groups was determined using a two-tailed Welch’s t test (i.e., assuming unequal standard deviations). A two-way analysis of variance (ANOVA) was used to assess the effects of diet and genotype on the specified immune cell populations and other measures. If a significant interaction or main effect was observed, a Tukey’s post hoc test was performed to determine pairwise differences between groups. The statistical analysis used for 16S microbiome data involved either Shapiro-Wilk test (for normal distribution) or the Kruskal-Wallis test, followed by Dunn’s post-hoc pairwise comparisons with Benjamini-Hochberg correction. For all analyses, statistical significance was set at p = 0.05, with * denoting p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001 and not significant denoted as ns. All statistical details and n numbers are described in the figure legends.

Supplementary Material

SUPPLEMENTAL INFORMATION

Supplemental information can be found online at https://doi.org/10.1016/j.celrep.2025.116109.

KEY RESOURCES TABLE.

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
TruStain FcX PLUS (Anti-Mouse CD16/32)	Biolegend	Cat#156603; RRID: AB_2783138
Alexa Fluor® 700 Mouse Anti-Mouse CD45.1 (Clone A20)	Biolegend	Cat#110724; RRID: AB_493732
BUV805 Rat Anti-Mouse CD45.2 (Clone 104)	BD Biosciences	Cat#741957; RRID: AB_2871265
BUV805 Rat Anti-Mouse CD45 (Clone 30-F11)	BD Biosciences	Cat#748370; RRID: AB_2872789
Alexa Fluor® 700 Rat Anti-Mouse CD45 (Clone 30-F11)	Biolegend	Cat#103128; RRID: AB_493715
BUV737 Rat Anti-Mouse CD45 (Clone 30-F11)	BD Biosciences	Cat#568344; RRID: AB_2872790
BUV563 Rat Anti-CD11b (Clone M1/70)	BD Biosciences	Cat#741242; RRID: AB_2870793
Brilliant Violet 650™ Rat Anti-CD11b (Clone M1/70)	Biolegend	Cat#101259; RRID: AB_11125575
FITC Rat Anti-Mouse F4/80 (Clone BM8)	Biolegend	Cat#123107; RRID: AB_893500
BV480 Rat Anti-Mouse F4/80 (Clone T45–2342)	BD Biosciences	Cat#565635; RRID: AB_2739313
BV650 Rat Anti-Mouse F4/80 (Clone BM8)	Biolegend	Cat#123149; RRID: AB_2564589
eFluor™ 450 Hamster Anti-Mouse CD11c (Clone N418)	eBioscience	Cat#148–0114-82; RRID: AB_1548654
BUV737 Hamster Anti-Mouse CD11c (Clone HL3)	BD Biosciences	Cat#612797; RRID: AB_2870124
APC/Cyanine7 Rat Anti-Mouse CD19 (Clone 6D5)	Biolegend	Cat#115530; RRID: AB_830706
Brilliant Violet 570™ Rat Anti-Mouse CD19 (Clone 6D5)	Biolegend	Cat#115535; RRID: AB_10933260
BB700 Rat Anti-Mouse CD19 (Clone 1D3)	BD Biosciences	Cat#566411; RRID: AB_2744315
Pacific Blue™ Rat anti-Mouse CD19 (Clone 6D5)	Biolegend	Cat#115523; RRID: AB_439718
PE/Fire™ 810 Rat anti-CD45R/B220 (Clone RA3-6B2)	Biolegend	Cat#103287; RRID: AB_2894692
APC Rat anti-CD45R/B220 (Clone RA3-6B2)	Biolegend	Cat#103211; RRID: AB_312996
APC/Fire™ 810 Rat Anti-Mouse CD3 (Clone 17A2)	Biolegend	Cat#100268; RRID: AB_2876392
Brilliant Violet 785™ Rat Anti-Mouse CD8a (Clone 53–6.7)	Biolegend	Cat#100750; RRID: AB_11218801
PE/Cyanine7 Rat Anti-Mouse CD4 Antibody RM4-5 (Clone RM4-5)	Biolegend	Cat#100528; RRID: AB_312729
Brilliant Violet 605™ Rat Anti-Mouse IgD (Clone 11-26c.2a)	Biolegend	Cat#405727; RRID: AB_2562887
BUV395 Rat Anti-Mouse IgD (Clone 11-26c.2a)	BD Biosciences	Cat#564274; RRID: AB_2738723
PE/Cyanine7 Rat Anti-Mouse IgM (Clone RMM-1)	Biolegend	Cat#406514; RRID: AB_10642031
BV510 Rat Anti-Mouse IgM (Clone II/41)	BD Biosciences	Cat#743324; RRID: AB_2741425
BUV615 Rat Anti-Mouse IgM (Clone R6-60.2)	BD Biosciences	Cat#752320; RRID: AB_2875837
PE/Cyanine5 Rat Anti-Mouse CD43 Activation-Associated Glycoform (Clone 1B11)	Biolegend	Cat#121216; RRID: AB_528811
APC/Cyanine7 Rat Anti-Mouse CD23 (Clone B3B4)	Biolegend	Cat#121216; RRID: AB_528811
Brilliant Violet 421™ Rat Anti-Mouse CD21/CD35 (CR2/ CR1) (Clone 7E9)	Biolegend	Cat#123421; RRID: AB_2650891
PerCP/Cyanine5.5 Rat Anti-GL7 Antigen (Clone GL7)	Biolegend	Cat#144610; RRID: AB_2562978
PerCP/Cyanine5.5 Rat Anti-Mouse CD138 (Syndecan-1) (Clone 281–2)	Biolegend	Cat#142510; RRID: AB_2561600
Brilliant Violet 650™ Rat Anti-Mouse CD138 (Syndecan-1) (Clone 281–2)	Biolegend	Cat#142518; RRID: AB_2650927
BUV395 Rat Anti-Mouse I-A/I-E (Clone 2G9)	BD Biosciences	Cat#743876; RRID: AB_2741827
BUV496 Rat Anti-Mouse I-A/I-E (Clone 2G9)	BD Biosciences	Cat#750171; RRID: AB_2874376
Alexa Fluor® 700 Rat Anti-Mouse CD86 (Clone GL-1)	Biolegend	Cat#105024; RRID: AB_493720
PE/Dazzle™ Rat Anti-Mouse CD86 (Clone GL-1)	Biolegend	Cat#105042; RRID: AB_2566409
BV786 Rat Anti-Mouse IgA (Clone C10-1)	BD Biosciences	Cat#743298; RRID: AB_2741409
PE Rat Anti-Mouse IgA (Clone 11-44-2)	Southern Biotech	Cat#1165-09L; RRID: AB_2794659
Brilliant Violet 421™ Rat Anti-Mouse CD184/CXCR4 (Clone L276F12)	Biolegend	Cat#146511; RRID: AB_2562788
Super Bright™ 600 Rat Anti-Mouse CD184/CXCR4 (Clone 2B11)	eBioscience	Cat#63-9991-82; RRID: AB_2688143
BUV737 Rat Anti-Mouse CD184/CXCR4 (Clone 2B11)	BD Biosciences	Cat#63-9991-82; RRID: AB_2871133
PerCP/Cyanine5.5 Hamster Anti-Mouse Gp38/Podoplanin (Clone 8.1.1)	Biolegend	Cat#127422; RRID: AB_2814015
APC/Cyanine7 Rat Anti-Mouse CD31 (Clone MEC13.3)	Biolegend	Cat#102534; RRID: AB_2860595
Brilliant Violet 711™ Rat Anti-Mouse CD127/IL-7Rα (Clone A7R34)	Biolegend	Cat#135035; RRID: AB_2564577
Brilliant Violet 711™ Rat Anti-Mouse Ep-CAM/CD326 (Clone G8.8)	Biolegend	Cat#118233; RRID: AB_2632775
BV480 Hamster Anti-Mouse KLRG1 (Clone 2F1)	BD Biosciences	Cat#746353; RRID: AB_2743672
Alexa Fluor® 647 Rabbit Anti-Mouse c-JUN (Clone 60A8)	Cell Signaling Technology	Cat#40502S; RRID: AB_2909794
Alexa Fluor® 488 Mouse Anti-FOXP3 (Clone 150D)	Biolegend	Cat#320012; RRID: AB_439747
PerCP-eFluor™ 710 Rat Anti-ROR gamma(t) (Clone AFKJS-9)	eBioscience	Cat#46-6981-82; RRID: AB_10717956
BV421 Rat Anti-ROR gamma(t) (Clone Q31-378-9)	BD Biosciences	Cat#562894; RRID: AB_2687545
PE Mouse Anti-CXCL12/SDF-1 (Clone 79018)	RnD Systems	Cat#IC350P; RRID: AB_3655504
Brilliant Violet 605™ Mouse Anti-Mouse CX3CR1 (Clone SA011F11)	Biolegend	Cat#149027; RRID: AB_2565937
PE/Cyanine5 Mouse Anti-Mouse CD64 (FcγRI) (Clone X54-5/7.1)	Biolegend	Cat#139331; RRID: AB_2922467
PE/Dazzle™ 594 Mouse Anti-XCR1 (Clone ZET)	Biolegend	Cat#148234; RRID: AB_2924477
Alexa Fluor® 700 Rat Anti-Mouse CD172a (SIRPα) (Clone P84)	Biolegend	Cat#144022; RRID: AB_2650812
Alexa Fluor® 647 Hamster Anti-Mouse CD103 (Clone 2E7)	Biolegend	Cat#121409; RRID: AB_535951
Brilliant Violet 785™ 700 Rat Anti-Mouse CD206/MMR (Clone C068C2)	Biolegend	Cat#141729; RRID: AB_2565823
Anti-Human β-Amyloid, 17–24 Antibody (Clone 4G8)	Biolegend	Cat# 800709; RRID: AB_2565326
Anti-Human β-Amyloid, 1–16 Antibody (Clone 6E10)	Biolegend	Cat# 803001; RRID: AB_2564653
Anti-CXCL12/SDF-1 (Clone 79018)	RnD Systems	Cat# MAB350-SP; RRID: AB_3655504
Anti-IBA1	Fujifilm	Cat# 019-19741; RRID: AB_839504
Anti-Mouse CD45 (Clone 30-F11)	Biolegend	Cat#103103; RRID: AB_493715
Chemicals, peptides, and recombinant proteins
CXCR4 Antagonist I, AMD3100	Sigma-Aldrich	CAS 155148-31-5; Cat#239820
Collagenase Type I	Millipore Sigma	Cat#SCR103
DNase I, bovine pancreas	Millipore Sigma	Cat#260913
TruBlack	Biotium	Cat#23014
DAPI	Millipore Sigma	Cat#D9542
ProLong Gold Antifade Mountant	ThermoFisher Scientific	Cat#P36930
Rodent Decloacker, 10X	Biocare Medical	Cat#RD913
N-(3-Dimethylaminopropyl)-N′-ethylcarbodiimide hydrochloride	Sigma-Aldrich	Cat#E6383
3-Nitrophenylhydrazine hydrochloride	Sigma-Aldrich	Cat#N21804
Butyric acid	Sigma-Aldrich	Cat#19215
2-Methylbutyric acid	Sigma-Aldrich	Cat#49659
Valeric acid	Sigma-Aldrich	Cat#75054
Isobutyric acid	Sigma-Aldrich	Cat#46935-U
Propionic acid	Sigma-Aldrich	Cat#94425
Taurocholic acid sodium salt hydrate	Sigma-Aldrich	Cat#T4009
Cholic acid	Sigma-Aldrich	Cat#C1129
Glycocholic acid hydrate	Sigma-Aldrich	Cat#G2878
Indole-3-propionic acid	Sigma-Aldrich	Cat#57400
Butyric acid (D7, 98%)	Cambridge Isotope Laboratories	Cat#DLM-1508
Pyridine	Sigma-Aldrich	Cat#270407
L-Tryptophan	Sigma-Aldrich	Cat#T8941
Cholic acid (2,2,4,4-D7, 98%)	Cambridge Isotope Laboratories	Cat#DLM-2611
Water	Honeywell Burdick & Jackson	Cat#LC365
Acetonitrile	Honeywell Burdick & Jackson	Cat#LC015
Methanol	Honeywell Burdick & Jackson	Cat#LC230
Ammonium formate	Sigma-Aldrich	Cat#70221
Formic acid	Fisher Scientific	Cat#A117
Critical commercial assays
LIVE/DEAD™ Fixable Blue Dead Cell Stain Kit, for UV excitation	Invitrogen	Cat#L23105
Lamina Propria Dissociation Kit, mouse	Miltenyi Biotec	Cat#130-097-410
Adult Brain Dissociation Kit, mouse and rat	Miltenyi Biotec	Cat#130-107-677
Dead Cell Removal Kit	Miltenyi Biotec	Cat#130-090-101
FOXP3 staining buffer set	eBioscience	Cat#00-5523-00
Cytofix/Cytoperm kit	BD Biosciences	Cat#554714
Deposited data
scRNAseq: raw and analyzed colon immune cells	This study	GEO: GSE255819
16S sequencing: fecal microbiome	This study	GEO: GSE305697
scRNAseq: CD45 5XFAD brain	Su et al.7	GEO: GSE207702
scRNAseq: 5XFAD brain	Keren-Shaul et al.4	GEO: GSE98971
snRNAseq: Hluman dorsolateral pre-frontal cortex	Zhou et al.78	AD Knowledge Portal
Experimental models: Organisms/strains
Mouse: WT C57BL/6J	The Jackson Laboratory	RRID: IMSR_JAX:000664
Mouse: 5XFAD (B6.Cg-Tg(APPSwFlLon, PSEN1*M146L*L286V)6799Vas/Mmjax_	The Jackson Laboratory	RRID: MMRRC_034848-JAX
Mouse: CD45.1 JaxBoy (C57BU6J-Ptprcem6Lutzy / J)	The Jackson Laboratory	RRID: IMSR_JAX:033076
Software and algorithms
Flow jo (v10.6.2)	FlowJo, LLC	N/A
Fiji/ImageJ	Schindelin et al.144	N/A
ZenBlue Lite	ZEISS	N/A
Prism (v9.4.0)	GraphPad	N/A
RStudio (2024.04.2 + 764)	Posit Software	N/A
SpectroFlo (v3.0)	Cytek Biosciences	N/A
Xcalibur Data Acquisition and Interpretation Software	ThermoFisher Scientific	N/A
Other
Mouse diet: Irradiated cellulose diet (16.6% cellulose and no i nulin)	Research Diets Inc.	Cat#D13081109i
Mouse diet: Irradiated inulin diet (17.3% inulin and no cellulose)	Research Diets Inc.	Cat#D13081108i
Mouse diet: Irradiated control diet (4.7% cellulose and no inulin)	Research Diets Inc.	Cat#D12450Ji
Vibratome VT10000	Leica	RRID: SCR_016495
Axioscan Microscope Slide Scanner v.7	ZEISS	N/A
Aurora	Cytek Biosciences	N/A
TQExactive Orbitrap MS instrument with Vanquish LC system	ThermoFisher Scientific	N/A

Highlights.

• AD is associated with altered immune parameters in the colons of 5XFAD mice

• 5XFAD mice show reduced CXCR4⁺ ASCs in the colon and increased CXCL12 expression in the brain

• CXCR4⁺ B cells and gut-specific IgA⁺ ASCs increase in the 5XFAD brain or dura mater

• Inulin diet modulates the gut microbiome and gut immune cells and attenuates frailty in AD