Abstract

BACKGROUND

   Glucagon-like peptide-2 (GLP-2) analogs are used clinically to enhance
   nutrient absorption in patients with short bowel syndrome (SBS);
   however, the precise mechanism remains unclear. To address this, the
   study aimed to clarify the dynamics of intestinal epithelial cells and
   immune cells in patients with SBS treated with GLP-2 analogs.

METHODS

   Five male patients diagnosed with SBS, all of whom received treatment
   with the GLP-2 analog teduglutide, were included in the study. We
   conducted longitudinal single-cell RNA sequencing (scRNA-Seq) analysis
   of intestinal tissue from patients with SBS over a year, integrating
   microbiome composition analysis.

RESULTS

   After treatment, the α-diversity of the gut microbiome increased,
   indicating a more varied microbial environment. ScRNA-Seq analysis
   revealed a reduction of T helper 2 cells and an increase in regulatory
   T cells, suggesting a shift toward an immunoregulatory intestinal
   environment. Additionally, nutrient-absorbing enterocyte-Top2 and
   middle clusters expanded, enhancing the absorption capacity, whereas
   major histocompatibility complex class I/II–expressing enterocyte-Top1
   cells declined, potentially modulating immune responses.

CONCLUSION

   The study findings indicate that GLP-2 analogs reshape intestinal
   immunity and microbiota, fostering a less inflammatory environment
   while promoting nutrient uptake efficiency. These insights offer a
   deeper understanding of the role of GLP-2 analogs in gut adaptation and
   provide a foundation for refining clinical strategies for SBS
   treatment.

FUNDING

   This work was supported by Sakaguchi Memorial Foundation, Grants-in-Aid
   from the Japanese Society for the Promotion of Science (JSPS)
   (21K18272, 23H03665, 23H02899, 23K27590, 25K22627, 23K08037), JST
   FOREST(21457195), and the Takeda Japan Medical Office Funded Research
   Grant 2022.

   Keywords: Clinical Research, Gastroenterology

   Keywords: Drug therapy
     __________________________________________________________________

   [90]graphic file with name jciinsight-10-194497-g061.jpg
     __________________________________________________________________

   GLP-2 analogs induce nutrient-absorbing enterocyte clusters and immune
   suppressive cells with icreased alpha-diversity of microbiota in small
   intestine.

Introduction

   Glucagon-like peptide-2 (GLP-2) is an intestinal hormone crucial for
   maintaining gut health and function. It is derived from the
   pro-glucagon gene, which produces multiple peptides, including GLP-1
   and GLP-2, primarily secreted by intestinal L cells ([91]1). GLP-2
   exerts its effects by binding to the GLP-2 receptor in subepithelial
   myofibroblasts and stimulating the expression and secretion of
   insulin-like growth factor-1 (IGF-1) ([92]2). Increased levels of IGF-1
   and IGF-2 are associated with gut growth, particularly in the mucosal
   and muscularis regions. IGF-1, in turn, activates IGF-1 receptor
   (IGF1R) in the intestinal epithelial cells, thereby promoting
   proliferation ([93]3–[94]6). GLP-2 stimulates intestinal epithelial
   growth and proliferation by activating intracellular signaling
   pathways, such as protein kinase A, protein kinase B, and extracellular
   signal–regulated kinase 1/2, thereby enhancing cell survival,
   proliferation, and nutrient absorption ([95]4, [96]6–[97]11). These
   pathways contribute to increased villus height, reduced apoptosis, and
   enhanced proliferation of crypt cells.

   GLP-2 plays multiple roles in digestive function: (a) It promotes
   intestinal mucosal growth by stimulating cell proliferation and
   inhibiting apoptosis, thereby increasing the surface area for nutrient
   absorption. (b) It maintains intestinal barrier integrity and prevents
   harmful bacterial invasion. (c) It prolongs intestinal transit time,
   thereby optimizing nutrient digestion and absorption ([98]12–[99]15).
   Recently, GLP-2 analogs have been clinically used for the treatment of
   short bowel syndrome (SBS) ([100]16–[101]21). However, the cellular and
   microbiome-level effects of GLP-2 analogs remain unclear. To address
   this knowledge gap, we conducted a longitudinal single-cell
   RNA-sequencing (scRNA-Seq) and 16S ribosomal RNA (16S rRNA) sequencing
   study to investigate how GLP-2 analogs influence intestinal cell
   populations and microbiome compositions over time. By employing these
   advanced techniques, we aimed to characterize the immunological and
   microbiological changes associated with GLP-2 treatment, providing
   deeper insights into its role in gut adaptation and its potential
   clinical implications for SBS management.

Results

GLP-2 analog therapy is effective for patients with SBS.

   This study enrolled 5 male patients diagnosed with SBS who were treated
   with the GLP-2 analog teduglutide ([102]Supplemental Table 1;
   supplemental material available online with this article;
   [103]https://doi.org/10.1172/jci.insight.194497DS1; cases 1–5).
   Clinical data, serum samples, intestinal tissue biopsies, and luminal
   content microbiota were collected at 3 time points: baseline (prior to
   treatment initiation, 0 months [0M]), 6 months after treatment (6M),
   and 12 months after treatment (12M). Intestinal microbial composition
   was assessed via 16S rRNA gene sequencing, while intestinal biopsies
   underwent histological evaluation and scRNA-Seq ([104]Figure 1A). All
   patients retained portions of the duodenum and jejunum, with complete
   resection of the ileum. The residual small intestinal lengths varied:
   case 1 had 30 cm of jejunum; case 2, 40 cm of duodenum; case 3, 35 cm
   of duodenum; case 4, 10 cm of jejunum; and case 5, 40 cm of jejunum.
   Teduglutide treatment was initiated between 6 months and 13 years
   following small intestinal resection. Prior to treatment, all patients
   required parenteral nutrition via central venous access due to
   inadequate oral intake.

Figure 1. GLP-2 analog is effective for patients with SBS.

   [105]Figure 1
   [106]Open in a new tab

   (A) Study design: Five male patients with short bowel syndrome (SBS)
   were enrolled and treated with the glucagon-like peptide-2 (GLP-2)
   analog teduglutide. Serum samples, intestinal tissue biopsies, luminal
   microbiome samples, and clinical records were collected at 3 time
   points: before treatment (0 months, 0M), 6 months after initiation
   (6M), and 12 months after initiation (12M). Intestinal microbiomes were
   analyzed using 16S ribosomal RNA (16S rRNA) sequencing, and intestinal
   tissue samples underwent histological examination and single-cell RNA
   sequencing (scRNA-Seq). (B and D) Body weight (B) and stool output (D):
   Each line represents changes in body weight or stool weight for an
   individual patient across the 3 time points. (C) Parenteral fluid
   volume: Each dot represents the volume of parenteral fluid administered
   to an individual patient across the 3 time points; bars indicate the
   mean ± SD. (E) The eicosatrienoic acid/arachidonic acid (T/T) ratio in
   blood: The left panel shows changes in the T/T ratio at the 3 time
   points. The right panel compares the T/T ratio before treatment (0M)
   with posttreatment (either 6M or 12M, with the later time point
   selected). Each dot represents an individual sample; bars indicate the
   mean ± SD. Values are log-transformed. (F) Blood biomarkers: Changes in
   blood levels of lipase and citrulline at the 3 time points. (G)
   Intestinal villi length: Changes in the total villus length (sum of
   villus height and crypt depth) at the 3 time points based on
   histological analysis. Each dot represents an individual sample; bars
   indicate the mean ± SD. Due to missing crypt depth data for case 4 at
   6M, this data point was excluded. Statistical analyses included the
   1-tailed paired t tests (B–F) and the Mann-Whitney test (G). IE,
   intestinal epithelium; LP, lamina propria; N/A, not available due to
   missing crypt depth data.

   After 12 months of teduglutide therapy, 4 of the 5 patients exhibited
   weight gain, with the exception of case 2 ([107]Figure 1B). Parenteral
   fluid requirements decreased in all cases, and total stool volume was
   reduced, indicating improved intestinal function ([108]Figure 1, C and
   D). Serum biochemical analysis revealed a posttreatment decline in the
   eicosatrienoic acid/arachidonic acid (T/T) ratio ([109]Figure 1E).
   Concurrently, levels of citrulline and lipase, which are biomarkers of
   intestinal absorptive capacity, increased, consistent with enhanced
   nutrient uptake ([110]Figure 1F). Histologic assessment demonstrated
   increased villus and crypt height over the 12-month period ([111]Figure
   1G; [112]Supplemental Figure 1, A–E; and [113]Supplemental Table 2),
   supporting the role of GLP-2 analogs in promoting mucosal adaptation.
   Collectively, these data suggest that teduglutide treatment fosters
   mucosal growth, improves nutrient absorption, and contributes to
   nutritional rehabilitation in patients with SBS.

GLP-2 analog therapy enhances microbial diversity and functional capacity
without uniform taxonomic shifts in SBS.

   We analyzed the composition of the intestinal microbiota at 0M, 6M, and
   12M using 16S rRNA gene sequencing. The gut microbiota exhibited
   dynamic alterations throughout the treatment period. Principal
   coordinate analysis based on Bray-Curtis dissimilarity and weighted and
   unweighted UniFrac plot revealed marked interindividual variation, with
   no consistent clustering pattern across time points ([114]Figure 2A and
   [115]Supplemental Figure 2A). The α-diversity, measured by observed
   operational taxonomic units, was significantly elevated at 12M compared
   with 0M and 6M, indicating increased microbial richness following
   treatment ([116]Figure 2B).

Figure 2. GLP-2 analog therapy enhances microbial diversity and functional
capacity without uniform taxonomic shifts in SBS.

   [117]Figure 2
   [118]Open in a new tab

   (A) β-diversity analysis: Principal coordinate analysis (PCoA) based on
   the Bray-Curtis dissimilarity index to assess gut microbiota
   differences among samples. (B) α-diversity analysis: Comparison of
   microbial diversity at 0 months (0M), 6M, and 12M based on observed
   operational taxonomic units (OTUs). The box shows the interquartile
   range (IQR), with the line inside indicating the median. The whiskers
   extend to the smallest and largest values within 1.5 times the IQR.
   Each dot represents an individual sample. (C) Microbial composition:
   Relative abundance of gut microbiota in each sample, represented by
   bacterial genus and categorized by phylum. (D) Differentially abundant
   taxa: Log2 fold change (LFC) of differentially abundant genera
   identified through Analysis of Compositions of Microbiomes with Bias
   Correction 2 (AMCOM-BC2). Genera with LFC > 1 were positively enriched,
   while those with LFC < –1 were negatively enriched. (E) MetaCyc
   functional pathways: Predicted MetaCyc functional pathways that showed
   significant temporal differences by Friedman test and exhibited
   increased abundance at 12M compared with 0M and/or 6M are shown. The
   x-axis represents the –Log10 (P value) from statistical comparisons of
   pathway abundances. Statistical analyses included permutational
   multivariate 2-way ANOVA (A), the Mann-Whitney test (B), and Friedman
   test (E).

   The microbiota composition patterns varied across patients. Cases 1 and
   3 exhibited a dominance of Proteobacteria at 0M and 6M, which decreased
   at 12M. Cases 2 and 4 initially harbored Firmicutes-dominant profiles,
   with a posttreatment increase in Bacteroides. Case 5 demonstrated
   transient Proteobacteria enrichment at 6M, followed by a shift in
   microbial community structure at 12M ([119]Figure 2C). Differential
   abundance analysis comparing 0M and 12M identified increased
   representation of TM7x, Porphyromonas, Neisseria, Peptostreptococcus,
   Lachnoanerobaculum, Parvimonas, and Gemella following treatment.
   Conversely, Limosilactobacillus, Saccharopolyspora, Romboutsia,
   Lachnospiraceae, HT002, Roseburia, Reyranella, A2, and Aquabacterium
   declined at 12M ([120]Figure 2D).

   To infer microbial functional potential, we performed PICRUSt2-based
   pathway prediction. The posttreatment (12M) microbiota tended to be
   enriched in several metabolic pathways, including nitrate reduction,
   pyrimidine deoxyribonucleotides de novo biosynthesis, O-antigen
   building blocks biosynthesis, UDP-N-acetylmuramoyl-pentapeptide
   biosynthesis II, octane oxidation, d-mannuronate biosynthesis,
   guanosine ribonucleotides de novo biosynthesis, superpathway of
   pyrimidine nucleobases salvage, and succinate fermentation to
   butanoate, collectively supporting enhanced adenosine and short-chain
   fatty acid (SCFA) metabolism ([121]Figure 2E). In the succinate
   fermentation to butanoate pathway, the enzyme commission entries for
   4-hydroxybutanoyl-CoA dehydratase, 4-hydroxybutyrate dehydrogenase, and
   succinate-semialdehyde dehydrogenase tended to be enriched at 12M
   ([122]Supplemental Figure 2B). In summary, GLP-2 analog treatment did
   not promote uniform expansion of specific bacterial taxa but was
   associated with increased microbial diversity and enrichment of
   SCFA-related metabolic pathways, suggesting a shift toward a more
   metabolically active and potentially beneficial microbiota.

Single-cell transcriptomics reveals epithelial expansion and mucosal
remodeling during GLP-2 therapy.

   To characterize cellular alterations in the small intestine during
   GLP-2 analog therapy, we conducted scRNA-Seq on intestinal biopsy
   samples collected at 0M, 6M, and 12M from cases 1, 2, 3, and 5. Case 4
   was excluded because of dialysis-related fluctuations in dietary
   intake, which could confound quantitative assessments. Additionally,
   12M samples from cases 2 and 5 were not available.

   Unsupervised clustering of transcriptomic data identified 19 distinct
   cell populations ([123]Figure 3A). Epithelial cells were defined by
   high EPCAM expression, while immune cell populations, including CD4^+
   and CD8^+ T cells, were identified based on CD3E and PTPRC expression
   patterns ([124]Figure 3B). Relative cell type abundances were
   quantified across the 3 time points ([125]Figure 3C). Overall, the
   proportions of epithelial cells, B cells, and plasma cells were similar
   between 0M and 6M ([126]Supplemental Figure 2, A and B). However, in
   cases 1 and 3, the proportion of epithelial cells increased markedly at
   12M compared with baseline ([127]Figure 3D and [128]Supplemental Figure
   3, A and B). Notably, the frequencies of other immune populations,
   including total T and B cells, remained relatively stable throughout
   the treatment period. These findings indicate that prolonged GLP-2
   treatment is associated with an expansion of the epithelial cell
   compartment in the small intestine, consistent with enhanced mucosal
   remodeling and regeneration.

Figure 3. Single-cell transcriptomics reveals epithelial expansion and
mucosal remodeling during GLP-2 therapy.

   [129]Figure 3
   [130]Open in a new tab

   (A) Uniform manifold approximation and projection (UMAP) visualization
   of small intestine mucosal cells after removing dead cells. Data from 4
   patients (0 months [0M], 6M, and 12M) include a total of 37,314 cells:
   case 1 (17,510 cells), case 2 (8,067 cells), case 3 (8,201 cells), and
   case 5 (3,536 cells). Cells were clustered into 19 groups, color-coded
   based on metadata. (B) Dot plot showing expression levels of key genes
   in each cluster. Dot color represents scaled expression, and size
   indicates the percentage of cells expressing each gene. (C)
   Visualization of cell distributions at 0M (10,187 cells), 6M (14,095
   cells), and 12M (13,032 cells). (D) Changes in cell cluster proportions
   over time (0M, 6M, 12M).

GLP-2 therapy modulates the intestinal immune cell landscape with selective
changes in T cell subsets.

   To delineate changes in the immune cell landscape during GLP-2 analog
   therapy, we identified immune cells by PTPRC expression and further
   resolved them into 16 distinct clusters using unsupervised clustering
   ([131]Figure 4A). CD4^+ T cell populations were distributed across the
   CD4T-1, CD4T-2, regulatory T cell (Treg), and Th17 resident memory
   clusters, while CD8^+ T cells were classified into CD8T, CD8T resident
   memory, effector CD8T-1, effector CD8T-2, cycling T, and general T cell
   clusters ([132]Figure 4B). Innate immune subsets were also identified,
   including cytotoxic cells, T cell receptor γδ^+ (TCRγδ^+) cells, NK
   cells, and group 3 ILCs, characterized by the expression of TRDC,
   NCAM1, and RORC, respectively. Of note, cycling T cells expressed
   elevated levels of MKI67, indicative of proliferative activity
   ([133]Supplemental Figure 4A). We next assessed the distribution of
   immune clusters across patients and time points ([134]Supplemental
   Figure 4, B–E). The relative proportions of TCRγδ^+ cells, ILC3s, and
   NK cells remained largely stable between 0M and 6M ([135]Figure 4, C
   and D, and [136]Supplemental Figure 4, D and E). However, a trend
   toward increased ILC3s’ abundance was observed at 6M compared with
   baseline. Given the limited changes among innate subsets, our
   subsequent analyses focused on the temporal dynamics of CD4^+ and CD8^+
   T cell populations, which may be more responsive to GLP-2–mediated
   mucosal modulation.

Figure 4. GLP-2 therapy modulates the intestinal immune cell landscape with
selective changes in T cell subsets.

   [137]Figure 4
   [138]Open in a new tab

   (A) Uniform manifold approximation and projection (UMAP) visualization
   of immune cells identified using PTPRC, with B cells, plasma cells, and
   myeloid cells excluded. Data from 4 patients at 0 months (0M), 6M, and
   12M include a total of 17,736 cells: case 1 (7,769 cells), case 2
   (3,915 cells), case 3 (4,161 cells), and case 5 (1,891 cells). Cells
   were clustered into 16 groups, color-coded by metadata. (B) Dot plot
   showing expression levels of key genes in immune cell clusters. (C)
   Cell distributions at 0M (4,937 cells), 6M (6,653 cells), and 12M
   (6,146 cells). (D) Changes in immune cell clusters at 0M, 6M, and 12M.
   ILC, innate lymphoid cell; NK, natural killer.

GLP-2 analog treatment selectively expands Tregs and activates
growth-associated signaling pathways in the CD4^+ compartment.

   Single-cell transcriptomic profiling of CD4^+ T cells identified 9
   distinct clusters based on gene expression signatures ([139]Figure 5A).
   Naive CD4^+ T cells were marked by CCR7 expression, while Tregs were
   defined by high FOXP3 expression. Th2 and Th17 cells were characterized
   by GATA3 and RORC expression, respectively. Th1 subsets were split into
   2 clusters: Th1-CXCR3 and Th1-TBX21, based on predominant expression of
   CXCR3 and TBX21. Additional CD4^+ clusters (CD4-1, CD4-2, CD4-3)
   included a subset with high SLC4A10 and KIT expression ([140]Figure 5B
   and [141]Supplemental Figure 5A).

Figure 5. GLP-2 analog treatment selectively expands Tregs and activates
growth-associated signaling pathways in the CD4^+ compartment.

   [142]Figure 5
   [143]Open in a new tab

   (A) Uniform manifold approximation and projection (UMAP) visualization
   of CD4^+ T cells identified using CD4. Data from 4 patients at 0 months
   (0M), 6M, and 12M include a total of 7,093 cells: case 1 (2,990 cells),
   case 2 (1,697 cells), case 3 (1,703 cells), and case 5 (703 cells).
   Cells were clustered into 9 groups, color-coded by metadata. (B)
   Expression patterns of key genes in CD4^+ T cell clusters. Each dot
   represents a gene, where color intensity indicates the scaled
   expression level in each cluster, and dot size represents the
   percentage of cells expressing the gene. (C) Cell distributions at 0M
   (1,772 cells), 6M (2,916 cells), and 12M (2,405 cells). (D) Proportion
   of each cluster in CD4^+ T cells at 0M, 6M, and 12M. (E) The percentage
   of Th1-CXCR3 cells, Th1-TBX21 cells, T regulatory cells (Tregs), Th17
   cells, and Th2 cells at 0M, 6M, and 12M. (F) Comparison of
   differentially expressed gene (DEG) sets related to Gene Ontology (GO)
   Biological Process (BP) across 0M, 6M, and 12M using compareCluster.
   Significantly enriched GO terms (y-axis) with P < 0.01 and q < 0.05 are
   visualized as dot plots. (G) Comparison of DEG sets related to
   biological pathways based on WikiPathways (WP) across 0M, 6M, and 12M
   using compareCluster. Significantly enriched pathways (y-axis) with P <
   0.05 and q < 0.2 are visualized as dot plots. (F and G) Numbers below
   the column name represent the number of genes related to GOBP (F) or WP
   (G) for each cluster. Dot size represents the gene ratio (proportion of
   DEGs), and color indicates the adjusted P value. The 2-tailed paired
   Student’s t test was used to compare cluster proportions (E).
   compareCluster analysis was performed using Over-Representation
   Analysis (ORA) based on a hypergeometric test. Multiple-testing
   correction was applied using the Benjamini-Hochberg (BH) method to
   control the false discovery rate (FDR) (F and G).

   Quantitative analysis revealed a consistent increase in the proportion
   of Tregs within the CD4^+ population across patients during treatment,
   while Th2 cells declined over the same period ([144]Figure 5, C–E, and
   [145]Supplemental Figure 5, B and C). In contrast, the frequencies of
   Th1-CXCR3, Th1-TBX21, and Th17 cells remained relatively unchanged.
   Gene Ontology (GO) enrichment analysis revealed upregulation of
   pathways involved in vascular endothelial growth factor (VEGF)
   signaling, TCR signaling, and epidermal growth factor (EGF) signaling
   at both 6M and 12M ([146]Figure 5, F and G). We then performed separate
   GO enrichment analyses for Treg and Th2 cell populations
   ([147]Supplemental Figure 5, D and E). Prior to treatment, Th2 cells
   were enriched for pathways related to adaptive immune response and
   lymphocyte-mediated immunity. In contrast, Tregs posttreatment showed
   enrichment for pathways such as positive regulation of tolerance
   induction and regulation of CD4^+ αβ T cell proliferation. These data
   suggest that GLP-2 therapy enhances immune tolerance pathways in Tregs
   while dampening immune activation in Th2 cells.

Diverse CD8^+ T cell subset dynamics reflect patient-specific responses to
GLP-2 analog therapy.

   Within the CD8^+ T cell compartment, 13 transcriptionally distinct
   clusters were identified ([148]Figure 6A). These included resident
   memory T cells (RMTs), defined by high CD69 and ITGAE expression,
   spanning RMT-1, RMT-2, RMT-IFNG, RMT-GZMB, RMT-KLRC1, RMT-KLRC4,
   RMT-ADRB1, RMT-REG4, and RMT-GSTM2 subsets. Additionally, clusters
   characterized by low CD69 and high TGFB1 expression were grouped as
   TGFB1-like populations, including TGFB1, TGFB1-GZMK, TGFB1-GZMB, and
   TGFB1-KLRC4 ([149]Figure 6B and [150]Supplemental Figure 6A).

Figure 6. Diverse CD8^+ T cell subset dynamics reflect patient-specific
responses to GLP-2 analog therapy.

   [151]Figure 6
   [152]Open in a new tab

   (A) Uniform manifold approximation and projection (UMAP) visualization
   of CD8^+ T cells identified using CD8A and CD8B markers. Data from 4
   patients at 0 months (0M), 6M, and 12M include a total of 5,904 cells:
   case 1 (2,486 cells), case 2 (1,300 cells), case 3 (1,472 cells), and
   case 5 (646 cells). Cells were clustered into 13 groups, color-coded by
   metadata. (B) Expression patterns of key genes in CD8^+ T cell
   clusters. Each dot represents a gene, where color intensity indicates
   the scaled expression level in each cluster, and dot size represents
   the percentage of cells expressing the gene. (C) Cell distributions at
   0M (1,744 cells), 6M (2,185 cells), and 12M (1,975 cells). (D)
   Proportion of each cluster in CD8^+ T cells at 0M, 6M, and 12M. (E)
   Comparison of differentially expressed gene (DEG) sets related to Gene
   Ontology (GO) Biological Process (BP) across 0M, 6M, and 12M using
   compareCluster. Significantly enriched GO terms (y-axis) with P < 0.01
   and q < 0.05 are visualized as dot plots. (F) Comparison of DEG sets
   related to biological pathways based on WikiPathways (WP) across 0M,
   6M, and 12M using compareCluster. Significantly enriched pathways
   (y-axis) with P < 0.05 and q < 0.2 are visualized as dot plots. (E and
   F) Numbers below the column name represent the number of genes related
   to GOBP (E) or WP (F) for each cluster. Dot size represents the gene
   ratio (proportion of DEGs), and color indicates the adjusted P value.
   compareCluster analysis was performed using Over-Representation
   Analysis (ORA) based on a hypergeometric test. Multiple testing
   correction was applied using the Benjamini-Hochberg (BH) method to
   control the false discovery rate (FDR) (E and F).

   While no single CD8^+ subset exhibited consistent enrichment across all
   cases, patient-specific changes were observed. For instance, RMT-KLRC4
   increased in case 1, TGFB1 in case 2, RMT-IFNG in case 3, and
   TGFB1-GZMK in case 5 ([153]Supplemental Figure 6, B and C). Notably,
   RMT-KLRC4 showed a general increase at both 6M and 12M across patients
   ([154]Figure 6, C and D). GO analysis and WikiPathway analysis of CD8^+
   T cells also revealed enrichment in VEGF, EGF, TCR, TGF-β, and insulin
   signaling pathways following treatment ([155]Figure 6, E and F).
   Collectively, these findings highlight a treatment-associated increase
   in Treg abundance within the CD4^+ compartment, while CD8^+ T cell
   dynamics were more heterogeneous. Nonetheless, both populations
   exhibited convergent upregulation of signaling pathways associated with
   tissue remodeling, immune modulation, and homeostasis.

GLP-2 therapy promotes nutrient-absorptive epithelial lineages while
suppressing immune-associated enterocyte subtypes.

   We analyzed epithelial cells to assess changes following GLP-2
   treatment and identified 7 distinct enterocyte clusters based on gene
   expression and zonation-specific markers ([156]22, [157]23)
   ([158]Figure 7, A–C, and [159]Supplemental Figure 7, A–C).
   Enterocyte-Top1 and -Top2 clusters, localized to the villus tip, were
   defined by high expression of ALPI, APOC3, APOA4, APOB, and SELENOP.
   Enterocyte-middle and enterocyte-bottom clusters, positioned in the
   mid- and basal villus, were marked by DGAT1, RBP2, and GSTA1. A
   crypt-associated cluster, enriched in GPX2 and REG1A, was designated as
   the crypt cluster. Secretory progenitor cells were defined by HEXIM1,
   SOX4, CSKMT, and DMBT1, while the tuft cell cluster expressed PLCG2,
   TRPM5, and IGF1R.

Figure 7. GLP-2 therapy promotes nutrient-absorptive epithelial lineages
while suppressing immune-associated enterocyte subtypes.

   [160]Figure 7
   [161]Open in a new tab

   (A) Uniform manifold approximation and projection (UMAP) embedding of
   epithelial cells identified by EPCAM expression from the total cell
   population. Cells from all 4 patients across 3 time points (0 months
   [0M], 6M, and 12M) are included, with a total of 6,636 cells (5,112
   from case 1, 244 from case 2, 936 from case 3, and 344 from case 5).
   Cells are clustered into 7 groups, colored by metadata as indicated in
   the figure. (B) Expression patterns of key marker genes for each
   cluster. Each dot represents a gene, where color saturation indicates
   scaled expression levels, and dot size represents the percentage of
   cells expressing that gene. (C) UMAP embedding of epithelial cells
   across 3 time points: 0M, 6M, and 12M (505 cells at 0M, 2,076 cells at
   6M, and 4,055 cells at 12M). (D) Proportions of epithelial cell
   clusters at 0M, 6M, and 12M. (E) Heatmap of Z-score normalized gene
   expression for key intestinal mucosal pathways across clusters. Columns
   represent clusters (indicated by color and label), and rows represent
   genes categorized by pathway. To compare gene expression between
   clusters, the Wilcoxon rank-sum test was used.

   As expected, total epithelial cell numbers increased following GLP-2
   treatment. Analysis of relative cluster proportions revealed a decrease
   in enterocyte-Top1 and an increase in enterocyte-Top2 posttreatment
   ([162]Figure 7, C and D, and [163]Supplemental Figure 8, A and B). The
   enterocyte-middle and -bottom clusters remained stable, whereas crypt
   and tuft cell clusters declined. Patient-specific analysis showed
   increased enterocyte-Top2 and middle clusters in cases 2, 3, and 5 but
   not in case 1.

   GO analysis indicated that enterocyte-Top1 was enriched in
   immune-related pathways, including T cell activation, leukocyte
   adhesion, and TCR signaling ([164]Supplemental Figure 8C). In contrast,
   enterocyte-Top2 was associated with pathways related to nutrient
   processing, such as lipid transport, glycerolipid metabolism, and
   intestinal absorption, suggesting distinct functional specializations
   ([165]Supplemental Figure 8D). To assess these roles more directly, we
   evaluated functional gene expression across clusters ([166]Figure 7E
   and [167]Supplemental Figure 9). Barrier integrity and
   nutrient-processing genes — covering ion transport, amino acid and
   glucose transport, cholesterol and fatty acid metabolism, digestive
   enzymes, and bile acid and vitamin transport — were enriched in
   enterocyte-Top2 and middle clusters. Conversely, MHC class I/II genes
   were predominantly expressed in enterocyte-Top1, supporting its role in
   immune modulation. The crypt cluster showed enrichment in glutathione
   metabolism, butanoate metabolism, cell proliferation, and antimicrobial
   responses. Notably, VEGFA and EGFR were enriched in enterocyte-Top2 and
   middle clusters, whereas IGF1R was specific to tuft cells. These
   findings suggest that the enterocyte-Top1 cluster is primarily involved
   in immune responses and MHC class I and II expression, whereas
   enterocyte-Top2 and middle clusters are more specialized for nutrient
   absorption, including amino acid, glucose, and fatty acid transport, as
   well as cholesterol transport. The crypt cluster was enriched in
   antimicrobial responses and metabolic pathways, such as glutathione and
   butanoate metabolism.

Longitudinal gene expression profiling uncovers coordinated enhancement of
absorptive function and barrier integrity in response to GLP-2.

   We next examined transcriptomic changes over the treatment period.
   Differential gene expression analysis showed significant upregulation
   of genes involved in differentiation (KRT20, BNIP5), antiapoptotic
   signaling (BIRC3), barrier function (MUC13, MUC17), immune regulation
   (CD55), nutrient processing (GLS, TMC5), cell movement (MYH14), and
   tight junction (ACTN4) at 6M and 12M compared with 0M ([168]Figure 8, A
   and B, and [169]Supplemental Figure 10, A and B). GO analysis revealed
   that antigen receptor-mediated signaling was enriched at baseline (0M),
   while nutrient response pathways predominated at 12M ([170]Figure 8C).
   WikiPathway analysis supported these trends, with T cell activation
   enriched at 0M and 6M and both VEGFA/VEGFR2 and EGF/EGFR signaling
   enriched at 6M and 12M ([171]Figure 8D).

Figure 8. Longitudinal gene expression profiling uncovers coordinated
enhancement of absorptive function and barrier integrity in response to
GLP-2.

   [172]Figure 8
   [173]Open in a new tab

   (A) Volcano plot showing differentially expressed genes (DEGs) in
   epithelial cells at 0 months (0M) versus 6M. The x-axis represents Log2
   fold change (LFC), and the y-axis represents –Log10 (P value). Vertical
   dashed lines indicate a threshold of |LFC| > 0.58, and the horizontal
   dashed line indicates a P value cutoff of 5 × 10^–10. (B) Gene Ontology
   (GO) pathway enrichment analysis of DEGs between 0M and 6M. Genes with
   LFC > 0.58 that were upregulated at 6M were analyzed. Dot plots display
   significantly enriched biological processes (BPs) (P < 0.01, q < 0.05),
   with dot size indicating gene count and color representing the adjusted
   P value. (C) GOBP enrichment analysis of DEGs across 0M, 6M, and 12M
   using compareCluster. (D) Differential gene expression analysis of
   biological pathways based on WikiPathways (WP) across 0M, 6M, and 12M
   using compareCluster. (C and D) Numbers below the column name represent
   the number of genes related to GOBP (C) or WP (D) for each cluster. Dot
   plots show significantly enriched GO terms (P < 0.01, q < 0.05) (C) or
   pathways (P < 0.05, q < 0.2) (D), with dot size representing the gene
   ratio (proportion of DEGs) and color indicating the adjusted P value.
   (E) Heatmap of Z-score normalized gene expression for key intestinal
   mucosal pathways across time points. Columns represent clusters
   (indicated by color and label), and rows represent genes categorized by
   pathway. GO analysis and compareCluster analysis were performed using
   Over-Representation Analysis (ORA) based on a hypergeometric test
   (B–D). Multiple-testing correction was performed using the
   Benjamini-Hochberg (BH) method to control the false discovery rate
   (FDR) (C and D). To compare gene expression between clusters, the
   Wilcoxon rank-sum test was used (E).

   To validate these functional shifts, we reassessed expression of key
   gene sets over time ([174]Figure 8E and [175]Supplemental Figure 11).
   Barrier and nutrient absorption genes increased at 6M and 12M, while
   immune-related genes, including MHC class I and II, were downregulated.
   Collectively, these findings indicate that GLP-2 analog treatment
   promotes the expansion of enterocyte-Top2 and middle clusters
   specialized in nutrient uptake, while reducing enterocyte-Top1 cells
   associated with immune activity.

   In addition, we analyzed gene expression changes within the
   enterocyte-Top2 population over the treatment course ([176]Figure 9A).
   MUC13, GCNT3 (involved in mucin regulation and barrier function), APOA4
   (lipid absorption), PTPRR (link to immunosuppressive signaling), XKR9
   (apoptosis-related gene), and PCLO and RASAL2 (negative regulators of
   RAS signaling) gene expressions were upregulated at 12M after GLP-2
   agonist treatment ([177]Figure 9, B and C). These genes were
   specifically and highly expressed within the enterocyte-Top2
   population. GO enrichment analysis of this population revealed
   increased activity in pathways related to intestinal absorption,
   cell-cell junction assembly, and intestine cholesterol absorption at
   12M after treatment ([178]Figure 9D). Consistently, KEGG pathway
   analysis demonstrated enrichment in pathways associated with tight
   junction, fat digestion and absorption, and glycerophospholipid
   metabolism ([179]Figure 9E). These data support the conclusion that
   GLP-2 analog treatment not only promotes expansion of enterocyte-Top2
   population but also enhances their functional gene programs related to
   nutrient uptake and epithelial barrier integrity. Furthermore, we
   examined SCFA-induced gene expression in epithelial cells, particularly
   within the enterocyte-Top2 cluster. SCFA is known to induce
   glycerophospholipid metabolism, which maintains membrane integrity and
   barrier function, so we further examined this pathway. Expression
   levels of key genes in glycerophospholipid pathway, such as GPAT3,
   LPCAT3, PLA2G4A, and CHKA, were progressively upregulated throughout
   treatment ([180]Figure 9F). Notably, these genes, particularly LPCAT3,
   CHKA, and PCYT1A, were enriched in enterocyte-Top2 and middle clusters
   ([181]Figure 9, G and H). Longitudinal analysis verified sustained
   upregulation of these genes over the treatment period ([182]Figure 9H).
   These findings support a model in which increased SCFA level enhances
   glycerophospholipid metabolism, reinforcing membrane stability and
   tight junction integrity in enterocyte-Top2 cluster, thus linking
   microbiome metabolism to epithelial function.

Figure 9. Longitudinal gene expression analysis of enterocyte-Top2 cluster
reveals functional reprogramming in response to GLP-2.

   [183]Figure 9
   [184]Open in a new tab

   (A) Volcano plot showing differentially expressed genes (DEGs) in
   enterocyte-Top2 cluster at 0 months (0M) versus 12M. The x-axis
   represents Log2 fold change (LFC), and the y-axis represents –Log10 (P
   value). Vertical dashed lines indicate a threshold of |LFC| > 0.58, and
   the horizontal dashed line indicates a P value cutoff of 0.05. (B and
   C) Heatmap of Z-score normalized gene expression for DEGs shown in A,
   across time points (B) and clusters (C). (D and E) CompareCluster
   analysis of DEGs in enterocyte-Top2 cluster across 0M, 6M, and 12M.
   Rows represent Gene Ontology (GO) Biological Process (BP) (D) and Kyoto
   Encyclopedia of Genes and Genomes (KEGG) pathways (E). Dot plots show
   significantly enriched GO terms or pathways (P < 0.01, q < 0.05), with
   dot size indicating the gene ratio (proportion of DEGs) and color
   indicating the adjusted P value. Numbers below the column name indicate
   the number of genes related to the pathways for each cluster. (F) Gene
   expression related to glycerophospholipid metabolism in enterocyte-Top2
   cluster across time points. Each dot represents a gene, color intensity
   indicates the scaled expression level in each cluster, and dot size
   represents the percentage of cells expressing the gene. (G and H)
   Heatmap of Z-score normalized expression of genes related to
   glycerophospholipid metabolism across time points (G) and clusters (H).
   compareCluster analysis was performed using Over-Representation
   Analysis (ORA) based on a hypergeometric test (D and E).
   Multiple-testing correction was performed using the Benjamini-Hochberg
   (BH) method to control the false discovery rate (FDR) (D and E). To
   compare gene expression between clusters, the Wilcoxon rank-sum test
   was used (B, C, G, and H).

   To further explore potential mediators of enterocyte-Top2 cell
   expansion, we conducted ligand–receptor interaction analysis using our
   single-cell transcriptomics dataset ([185]Figure 10, A–C). At 0M, key
   outgoing signaling pathways included MIF, GALECTIN, MK, CCL, and CXCL,
   while incoming signaling was generally weak but exhibited connectivity
   among immune cell subsets. By 6M, additional outgoing pathways, such as
   TGF-β, ANNEXIN, PARs, GRN, and SEMA3, emerged alongside those observed
   at baseline, indicating progressive complexity in the outgoing
   signaling landscape. Concurrently, incoming signaling strength
   increased, particularly within regulatory and memory T cell
   populations. At 12M, the intercellular signaling network further
   expanded, with dominant outgoing pathways including TGF-β, VEGF, PARs,
   KIT, RANKL, and CD70. The overall increase in incoming signaling
   strength at this time point supports a robust enhancement of
   intercellular communication following GLP-2 analog treatment. Several
   signaling pathways, MIF, GALECTIN, MK, CXCL, VEGF, CCL, VISFATIN, IL16,
   TNF, BAFF, LIGHT, FASLG, TWEAK, CD40, CSF, BAG, and IL1, were
   consistently detected across all time points, suggesting they
   participate in maintaining basal immune–epithelial interactions. In
   contrast, TGF-β, PARs, KIT, SEMA3, CD70, and RANKL were uniquely
   upregulated at 12M, indicating the emergence of specific biological
   processes, including tissue remodeling (TGF-β), neuro-immune signaling
   (SEMA3), and enhanced adaptive immunity (CD70, RANKL).

Figure 10. Changes in ligand–receptor interactions between clusters reveal
enhanced intercellular communication following GLP-2 analog treatment.

   [186]Figure 10
   [187]Open in a new tab

   (A–C) Heatmaps showing outgoing signaling patterns (left panels) and
   incoming signaling patterns (right panels) for each cluster at 0 months
   (0M) (A), 6M (B), and 12M (C). Columns represent cell clusters
   (indicated by color and label), and rows represent signaling pathways.
   Relative interaction strength is shown on the scale bars.

   To further dissect cellular crosstalk, we examined interactions between
   immune and nonimmune cell populations, including epithelial cells and
   CD34^+ cells. Immune cells broadly upregulated TGF-β signaling as
   senders, while CD34^+ cells showed corresponding receptor expression,
   suggesting they act as primary receivers (detected at 12M) ([188]Figure
   11A). Additionally, SEMA3 (detected at 12M), produced by epithelial
   cells, showed strong correlation with CD34^+ cells, as did upregulated
   VEGF and VISFATIN signaling pathways ([189]Figure 11B and
   [190]Supplemental Figure 12, A and B). These interactions are
   consistent with a GLP-2–induced program supporting vascular
   regeneration. At 12M, epithelial cells also upregulated outgoing
   signals, such as TWEAK and PARs (detected at 12M), toward immune
   populations, supporting roles in barrier maintenance and tissue repair
   ([191]Figure 11, C and D). Overall, our data suggest that GLP-2 analog
   treatment enhances communication between epithelial, immune, and CD34^+
   cells, which foster an immune-suppressive, tissue-protective
   environment that reinforces intestinal barrier integrity.

Figure 11. GLP-2 analog treatment enhances communication between epithelial,
immune, and CD34^+ cells.

   [192]Figure 11
   [193]Open in a new tab

   (A–D) Circle plots showing intercellular signaling networks. The TGF-β
   (A), SEMA3 (B), and PARs (D) signaling pathways were newly detected at
   12 months (12M). The TWEAK signaling pathway (C) was upregulated at 12M
   compared with 0M. Arrow colors indicate the source (sender) clusters of
   the signaling, and arrow widths represent the relative signaling
   strength.

Discussion

   GLP-2 analogs are clinically effective in promoting intestinal
   adaptation in patients with SBS, reducing dependence on parenteral
   nutrition. While their trophic effects on the intestinal mucosa are
   well established ([194]17, [195]18), the mechanisms by which GLP-2
   therapy modulates epithelial cell differentiation and intestinal immune
   compartments in humans have remained largely undefined. In this study,
   we combined longitudinal single-cell transcriptomics with microbiome
   profiling to elucidate the cellular and molecular remodeling events
   induced by GLP-2 analog treatment in SBS. Our findings offer
   mechanistic insights into how GLP-2 promotes nutrient absorption and
   orchestrates immune adaptation in the human intestine.

   Clinically, patients showed improvements in nutritional status,
   including weight gain, reduced parenteral support, and increased levels
   of citrulline and lipase, consistent with enhanced absorptive function.
   These improvements were accompanied by histological increases in villus
   height and crypt depth, reinforcing the known trophic effects of GLP-2
   on mucosa. These physiological adaptations aligned with decreased stool
   output and reduced need for parenteral nutrition, underscoring the
   therapeutic impact of GLP-2 analog in SBS management.

   Despite clinical improvements, microbiome responses to GLP-2 analog
   were heterogeneous across the patients, echoing prior findings on
   interindividual variability in upper gastrointestinal microbiome
   composition ([196]24–[197]27). Microbial analysis revealed increased
   α-diversity over time, although taxonomic changes were patient
   specific. Functional inference, however, indicated upregulation of
   pathways related to SCFA biosynthesis, including acetate and butanoate.
   These metabolites are known to enhance epithelial barrier integrity and
   regulate immune responses, suggesting that GLP-2 drives a functional,
   rather than compositional, restructuring of the microbiome. SCFA
   signaling has also been linked to induction of glycerophospholipid
   metabolism, a pathway we found upregulated in absorptive enterocyte
   clusters. These findings support a model in which GLP-2–induced
   microbial changes promote epithelial membrane stability and tight
   junction integrity via SCFA-mediated signaling.

   Immunologically, we observed an increase in Tregs and a concurrent
   decrease in Th2 cells across patients. This immunological shift likely
   reflects improved epithelial barrier function and reduced antigen
   exposure ([198]28–[199]30), as well as increased microbial production
   of immunomodulatory metabolites, such as SCFA ([200]31–[201]39). While
   GLP-2 receptors are not expressed in immune cells ([202]40), these
   changes are plausibly mediated by downstream effects of epithelial and
   stromal remodeling ([203]41, [204]42). Notably, ILC3s, which play a
   critical role in maintaining epithelial homeostasis through cytokines
   such as IL-22 ([205]43–[206]54), showed a subtle but consistent
   increase during the treatment. Together, these findings suggest that
   GLP-2 analog had indirect but significant effects on mucosal immunity,
   positioning the hormone as modulator of both structural and
   immunological adaptation in the gut.

   Concurrently, scRNA-Seq of intestinal biopsies revealed that GLP-2
   analog treatment promotes a shift toward absorptive phenotypes.
   Specifically, we observed expansion of enterocyte-Top2 and middle
   clusters enriched for genes involved in lipids, amino acids, and
   vitamin transport along with upregulation of barrier function markers
   ([207]22, [208]23). In contrast, enterocyte-Top1 cells characterized by
   high MHC class I/II expression declined suggesting a shift away from
   epithelium–immune interactions toward a more absorptive epithelial
   phenotype ([209]55, [210]56). Importantly, GLP-2 receptors were not
   expressed in these absorptive enterocytes, implying that trophic
   effects are mediated through paracrine signaling from GLP-2–expressing
   subepithelial myofibroblasts or neural cells via IGF-1 and EGF
   pathways, consistent with prior murine studies ([211]2, [212]3, [213]5,
   [214]11, [215]57–[216]59). Indeed, IGF1R was notably enriched in tuft
   cells ([217]60), while EGFR expression was prominent in enterocyte-Top2
   clusters, consistent with previous reports ([218]61–[219]64). Our
   findings suggest that GLP-2 analogs may enhance the development of
   enterocyte-Top2 clusters through EGFR signaling.

   This study has limitations. The sample size was small, was restricted
   to male patients, and lacked certain stromal and innate immune
   populations, such as fibroblasts, mesenchymal cells, and granulocytes.
   Additionally, ligand–receptor and pathway activity were inferred
   computationally and will require experimental validation. Nonetheless,
   the longitudinal single-cell approach provides a view of epithelial,
   immune, and microbial remodeling in human SBS during GLP-2 therapy.
   Despite these constraints, our longitudinal design and integrative
   single-cell approach offer a comprehensive view of the cellular
   ecosystem remodeling that occurs during GLP-2 therapy in humans.

   In conclusion, GLP-2 analog treatment in SBS orchestrated a coordinated
   remodeling of the intestinal ecosystem, promoting absorptive epithelial
   differentiation, regulatory immune adaptation, and SCFA-associated
   microbial function. These findings provide mechanistic insights into
   GLP-2–driven intestinal rehabilitation and may inform therapeutic
   strategies for malabsorptive and inflammatory intestinal disorders.

Methods

   See the [220]Supplemental Methods for more information.

Sex as a biological variable.

   Our study involved 5 male participants due to the limited availability
   of eligible patients.

   However, the underlying mechanisms targeted in GLP-2 analog are
   conserved in female and used clinically in both male and female
   individuals, and thus the findings are expected to be relevant to both
   sexes.

Statistics.

   All graphs show mean ± SD unless otherwise stated. P values less than
   0.05 were considered significant.

Study approval.

   All human samples were obtained from individuals providing written
   informed consent following protocols approved by the Keio University
   IRB.

Data availability.

   All data are available in the main text or the supplemental material.
   Individual data points for each figure are available in the
   [221]Supporting Data Values file.

   16S-rRNA-Seq data are available on the NCBI BioProject with the entry
   number PRJDB20353.

   ScRNA-Seq data are available on the NCBI GEO with the entry number
   [222]GSE294205.

Author contributions

   YK, Y Yamada, and TS conceptualized and designed the study.
   Investigation was performed by YK, AC, AT, MH, AS, IK, YA, and SS
   (human tissue preparation and analysis) and MS, AK, T Kumagai, HD, YS,
   YI, KS, SY, YM, HK, M Kano, Mototoshi Kato, Y Yoshimatsu, KT, KO, T
   Kanai, NH, Motohiko Kato, T Kumagai and T Kuroda (clinical data
   analysis). YK and HT performed pathological data analysis. ScRNA-Seq
   and microbiome analyses, as well as data visualization, were conducted
   by YK and KM. Funding acquisition was carried out by Y Yamada and TS.
   AF, Y Yamada, and TS supervised the study. YK, KM, and TS wrote the
   original draft. Review and editing were performed by JM, CCH, AF, Y
   Yamada, and TS. All authors reviewed and approved the manuscript.

Supplementary Material

   Supplemental data
   [223]jciinsight-10-194497-s057.pdf^ (50.7MB, pdf)
   ICMJE disclosure forms
   [224]jciinsight-10-194497-s058.pdf^ (2.7MB, pdf)
   Supporting data values
   [225]jciinsight-10-194497-s059.xls^ (710.5KB, xls)

Acknowledgments