Abstract

   Maternal pregestational diabetes mellitus (PGDM) can lead to adverse
   fetal outcomes, including lasting impacts on pancreatic development.
   However, the specific impacts of maternal PGDM on cellular functions
   and intercellular communication within the fetal pancreas remain poorly
   understood. Here, single‐nucleus RNA sequencing and spatial
   transcriptomics (ST) are employed to investigate cellular responses and
   spatial changes in the fetal pancreas (E16.5 and E18.5) under maternal
   PGDM conditions. The findings reveal significant cellular heterogeneity
   among acinar and beta cells, along with pronounced metabolic stress
   responses. More importantly, decreased insulin secretion is observed
   and accompanied by the compensatory increase of Pdx1, Nkx6.2, and Mafa,
   and substantial alterations in cell‐cell communication across multiple
   cell types. ST analysis further highlights enhanced spatial enrichment
   in cellular niches exposed to maternal PGDM. These findings provide
   valuable insights into the molecular mechanisms underlying fetal
   pancreatic response to maternal PGDM and offer a detailed
   spatiotemporal perspective on these processes.

   Keywords: fetal pancreas, pregestational diabetes mellitus,
   single‐nucleus RNA sequencing, spatial transcriptomics
     __________________________________________________________________

   Maternal pregestational diabetes mellitus (PGDM) induced metabolic
   stress in fetal mice pancreas, alongside significantly decreased
   insulin secretion, altered cell‐cell communication, and distinct
   changes in cellular niches.

   graphic file with name ADVS-12-2411126-g004.jpg

1. Introduction

   The prevalence of diabetes mellitus has steadily increased over the
   past few decades.^[ [30]^1 ^] Pregestational diabetes mellitus (PGDM)
   is a significant risk factor for complications during pregnancy and
   childbirth.^[ [31]^1 , [32]^2 ^] Although PGDM occurs less frequently
   in pregnant women (0.92%–1.18%), it is associated with a high risk of
   short‐ and long‐term adverse fetal and maternal outcomes, such as
   metabolic dysfunction and hypertension.^[ [33]^1 , [34]^2 , [35]^3 ^]
   Therefore, more in‐depth research into PGDM is urgently needed as a
   critical public health issue.

   The pancreas plays a vital role in regulating glucose homeostasis, and
   it must adapt to the metabolic stress imposed by pregnancy.^[ [36]^4 ^]
   This process involves hypertrophy and hyperplasia of pancreatic
   endocrine cells.^[ [37]^5 ^] Recent clinical studies^[ [38]^6 ^] have
   shown that women with elevated blood glucose levels during pregnancy
   experience worsened insulin sensitivity, potentially disrupting
   glucose‐insulin homeostasis in their offspring. Animal experiments have
   further indicated that fetal rats exposed to hyperglycemia in
   late‐pregnancy^[ [39]^7 ^] or throughout gestation^[ [40]^8 ^] exhibit
   decreased insulin sensitivity and impaired beta‐cell function compared
   to controls after birth. Fetal development is a critical period for
   pancreatic growth, with lasting effects on the individual's metabolic
   health throughout life.^[ [41]^9 ^] However, prior studies on the
   offspring of mothers with PGDM have largely focused on pancreatic
   function in adulthood, neglecting the effects during fetal development.
   Furthermore, these studies have lacked the single‐cell and spatial
   resolution to elucidate the specific effects of maternal hyperglycemia
   on the fetal pancreas.

   With the advent of technologies such as single‐cell RNA sequencing,^[
   [42]^9 , [43]^10 ^] single‐nucleus RNA sequencing (snRNA‐seq),^[
   [44]^11 ^] and spatial transcriptomics (ST),^[ [45]^12 ^] a deeper
   understanding of pancreas development in humans and mice has been
   achieved. Although animal models have limitations, they remain
   indispensable for fetal pancreatic research due to the ethical and
   practical challenges of conducting clinical trials in pregnant women.
   A recent study^[ [46]^10 ^] established a mouse islet atlas using
   diverse datasets, but little is known about the complex cellular
   responses to metabolic stress or intercellular communication changes
   within the fetal pancreas in the context of maternal PGDM.

   In this study, we performed snRNA‐seq and ST analysis to explore the
   effects of maternal PGDM on the fetal pancreas (E16.5 and E18.5). Our
   multi‐omics approach provided the following key findings: (1) we
   identified functional and metabolic heterogeneity among beta‐cells and
   acinar cells in both PGDM‐exposed group and control fetal mice; (2)
   snRNA‐seq revealed pronounced cellular responses to metabolic stress in
   the PGDM group, with transmission electron microscopy (TEM) confirming
   significant swelling of ribosomal and mitochondrial structures; (3) we
   observed decreased insulin secretion accompanied by the compensatory
   increase of Pdx1, Nkx6.2 and Mafa, and substantial alterations in
   cell‐cell communication across multiple cell types in the PGDM group;
   and (4) ST analysis demonstrated increased spatial enrichment in
   various cellular niches within the fetal pancreas exposed to maternal
   PGDM. These findings suggest that maternal PGDM profoundly impacts
   fetal pancreatic development and may provide valuable insights for
   preventing adverse metabolic health outcomes in offspring.

2. Results

2.1. Global Single‐Nucleus Transcriptome Map of the Fetal Pancreas (E16.5 and
E18.5)

   The data shown in Figure [47]S1 (Supporting Information) demonstrated
   that the moderate‐to‐severe mouse model of PGDM was successfully
   constructed (Figure [48]S1, Supporting Information).^[ [49]^2b ^] There
   was a significant decrease in the body weights of fetal mice and the
   numbers of live fetuses per litter in the PGDM group compared with the
   control group at E12.5, E14.5, E16.5, and E18.5 (Figure [50]S1,
   Supporting Information).

   To achieve high‐resolution molecular profiling of fetal pancreatic
   cells in response to maternal PGDM, we conducted snRNA‐seq on fetal
   pancreas from PGDM and control groups (Figure [51]1A). After quality
   control and filtration, 38237 single nuclei (E16.5) and 21657 single
   nuclei (E18.5) were included in the analysis (Figures [52]S2A–D and
   [53]S3, Supporting Information).

Figure 1.

   Figure 1
   [54]Open in a new tab

   Global single‐nucleus transcriptome map of fetal pancreas (E18.5). A)
   Schematic for the experimental design and workflow for data processing
   steps. B) Uniform manifold approximation and projection (UMAP)
   representation of the fetal pancreas (E18.5). C) Dotplot showing the
   expression levels of specific marker genes in each cell cluster at
   E18.5. D) Bar graph showing the cluster composition of the PGDM and
   control groups at E18.5 in this study. Abbreviations: E0.5, embryonic
   day E0.5; E16.5, embryonic day E16.5; E18.5, embryonic day E18.5.

   We identified 17 major cell types, using E18.5 snRNA‐seq data, based on
   gene expression markers: acinar_1 cells (Rbpjl and Sel1l), acinar_2
   cells (Cpb1, Amy2a1 and Pnliprp1), beta_1 cells (Trpm3, Kcnip1, and
   Robo2), beta_2 cells (Mlxipl and Pax6), alpha‐cells (Gcg and Nxph1),
   delta cells (Sst), ductal cells (Bicc1 and Erbb4), mesenchymal _1 cells
   (Col3a1, Vcan, and Adam12), mesenchymal_2 cells (Acta2, Actg2, and
   Col8a1), neuronal cells (Nrg3, Syt1, and Nrxn1), vascular endothelial
   cells (Flt1 and Eng), general endothelial cells (Vwf and Pecam1),
   lymphatic endothelial cells (Flt4 and Mmrn1), macrophages (F13a1 and
   Mrc1), T cells (Il7r, Il18r1, and Itga4), B cells (Blnk, Pou2af1, and
   Slamf6), and fibroblasts (Sulf1 and Wt1) (Figure [55]1B, C, and Table
   [56]S1, Supporting Information). We also identified 16 major cell types
   using E16.5 snRNA‐seq data (Figure [57]S3 and Table [58]S2, Supporting
   Information), and we found major cell types at different time points
   (E16.5 and E18.5) were consistent.

   Next, we investigated the potential developmental connectivity among
   acinar cells, beta cells, and ductal cells at E16.5 and E18.5, with
   partition‐based graph abstraction (PAGA). The PAGA analysis represented
   that the acinar cells were highly connected to beta cells and ductal
   cells (Figure [59]S4A, Supporting Information). Subsequently, we
   performed pseudotime analysis with the Monocle2 R package to better
   understand the developmental relationship among acinar and beta cell
   populations. We found that acinar_2 cells (E16.5) and acinar_3 cells
   (E16.5) appeared first, followed by acinar_1 cells (E16.5), then
   acinar_2 cells (E18.5), and finally acinar_1 cell (E18.5); beta_2 cells
   (E16.5) appeared first, followed by beta_1 cell (E16.5), then beta_2
   cell (E18.5), and finally beta_1 cell (E18.5) (Figure [60]S4B,
   Supporting Information).

   In addition, we investigated the heterogeneity of acinar_1 cells,
   acinar_2 cells, beta_1 cells, and beta_2 cells by performing cluster
   composition analysis. This analysis revealed substantial differences in
   the composition of these cell types between the PGDM (E18.5) and
   control groups (E18.5), while the composition of other clusters
   remained relatively unaffected (Figure [61]1D). Notably, we observed
   significant metabolic heterogeneity within the acinar_1 cells, acinar_2
   cells, beta_1 cells, and beta_2 cells at E18.5 (Figure [62]S5,
   Supporting Information). Metabolic pathway analysis revealed
   differential expression of several pathways in these cell types,
   including glycolipid metabolism, glycolysis/gluconeogenesis, nitrogen
   metabolism, phenylalanine metabolism, steroid biosynthesis, sulfur
   metabolism, and taurine and hypotaurine metabolism (Figure [63]S6,
   Supporting Information).

   Further analysis showed that glycolysis/gluconeogenesis‐related
   pathways were significantly up‐regulated in multiple cell clusters,
   particularly acinar_1 cells, acinar_2 cells, beta_1 cells, and beta_2
   cells, in the PGDM group (E18.5) compared to the control group (E18.5)
   (Figure [64]S7, Supporting Information). This observation aligned with
   previous findings in diabetic mice (βV59 M), which exhibited high
   expression of glycolysis‐related genes despite reduced glucose
   utilization.^[ [65]^13 ^] In summary, our data revealed substantial
   heterogeneity in acinar cells and beta cells between the PGDM and
   control groups, particularly in their metabolic profiles.

2.2. Pronounced Cellular Response to Metabolic Stress Caused by Maternal PGDM
in the Fetal Pancreas at E16.5 and E18.5

   A previous study^[ [66]^14 ^] indicated that high glucose induced
   severe metabolic disorders in pancreatic cells. Therefore, we further
   explored the cellular response to metabolic stress in the fetal
   pancreas (Figure [67]2A). To accurately characterize the cell
   populations of acinar_1 cell, acinar_2 cell, beta_1 cell, and beta_2
   cell, we first presented seven major cell types from the E18.5 samples
   using uniform manifold approximation and projection (UMAP)
   (Figure [68]2B). A split‐by‐group UMAP clearly demonstrated that beta_2
   cells were enriched in the PGDM (E18.5) group, while acinar_2 cells and
   beta_1 cells were more prevalent in the control (E18.5) group
   (Figure [69]2C).

Figure 2.

   Figure 2
   [70]Open in a new tab

   Maternal PGDM caused increased cellular response to stress in the fetal
   pancreas at E16.5 and E18.5. A) Schematic diagram showing metabolic
   stress in the PGDM (left) and control (right) groups. B) UMAP plot
   showing seven cell populations in the PGDM (E18.5) and control (E18.5)
   groups. C) UMAP plot showing seven cell populations in the PGDM (E18.5,
   upper panel) and control (E18.5, lower panel) groups. D) Depiction of
   up‐regulated differentially expressed genes (DEGs) for the acinar_1
   cell, acinar_2 cell, beta_1 cell, and beta_2 cell populations in the
   PGDM (E18.5) group, compared to the control (E18.5) group. Red dots
   represented up‐regulated DEGs. E) ST data based on spots (bin 50,
   50 × 50 nanoballs aggregated) showing gene expression levels of Rpsa,
   Rps19, Rplp1, and Rpl14 in the PGDM (E16.5) and control (E16.5) groups.
   F) Top five enriched KEGG pathways of up‐regulated DEGs in the acinar_1
   and acinar_2 cell populations in the PGDM (E18.5) group, compared to
   the control (E18.5) group. G) Electron microscopy revealed impairments
   in the mitochondria and endoplasmic reticulum in the PGDM (E18.5) group
   compared to the control (E18.5) group. Abbreviations: ER, endoplasmic
   reticulum; M, mitochondria; N, nucleus; ST, spatial transcriptomics.

   Next, we identified differentially expressed genes (DEGs) that were
   up‐regulated between the PGDM (E18.5) and control (E18.5) groups
   across the four clusters mentioned above (Figure [71]2D). For example,
   the analysis of DEGs from the acinar_1 cell population revealed that
   genes such as Rpsa, Rps19, Gphn, Camk1d, and Cmss1 were significantly
   up‐regulated in the PGDM (E18.5) group compared to the control (E18.5)
   group (Figure [72]2D). Similarly, in the acinar_2 cell population,
   genes including Rpsa, Rplp1, Prss1, Prss3, Try4, Try5, Cela3b, Spink1,
   and Ctrb1 were expressed at significantly higher levels in the PGDM
   (E18.5) group compared to controls (Figure [73]2D). Analysis of DEGs
   from the beta_1 cell population demonstrated that the PGDM (E18.5)
   group had higher expression levels of Rpsa, Rpl14, Rps23, Ptger3, Aqp4,
   Slc5a10, Sorcs2, Sorbs1, Txnip, and Gpi1 than the control (E18.5) group
   (Figure [74]2D). Similarly, in the beta_2 cell population, genes such
   as Rpsa, Rpl14, Rpl5, Rps4x, Ptger3, Sorbs1, Txnip, Slc5a10, Sorcs2,
   and Scg2 were significantly up‐regulated in the PGDM (E18.5) group
   compared to controls (Figure [75]2D).

   Interestingly, many of the significantly up‐regulated DEGs, including
   Rpsa and Rpl14, were shared across multiple cell clusters (Figure
   [76]S8A, Supporting Information). These genes were primarily associated
   with ribosomal complexes (Figure [77]2D). Protein–protein interaction
   analysis revealed a tightly connected network among these DEGs (Figure
   [78]S8B, Supporting Information). In addition, ST results demonstrated
   notable differences in the expression of Rpsa, Rps19, Rplp1, and Rpl14
   between the PGDM (E16.5) and control (E16.5) groups (Figure [79]2D).
   These data provided support for the hypothesis that ribosomal genes
   might be affected by maternal PGDM at both E16.5 and E18.5. However,
   additional studies were necessary to further substantiate this
   perspective.

   To further elucidate the biological functions of these DEGs, we
   performed gene ontology (GO) enrichment analysis and kyoto encyclopedia
   of genes and genomes (KEGG) pathway analysis on the up‐regulated DEGs
   in the PGDM (E16.5 and E18.5) group compared to controls (E16.5 and
   E18.5). The GO enrichment analysis of acinar cells and beta cells at
   E18.5 highlighted the significant enrichment of genes involved in
   ribosome biogenesis, ribosomal subunit assembly, cytoplasmic
   translation, glucose 6−phosphate metabolic process, and glycolytic
   process through fructose‐6‐phosphate (Figure [80]S9, Supporting
   Information). The KEGG analysis at E18.5 suggested that significantly
   up‐regulated signaling pathways related to ribosome and protein
   digestion and absorption in the acinar_1 and acinar_2 cell populations
   (Figure [81]2F) and signaling pathways associated with ribosome in
   beta_1 cell and beta_2 cell populations (Figure [82]S10, Supporting
   Information). Similarly, results of GO enrichment analysis of acinar
   cells and beta cells at E16.5 showed that most of the significantly
   enriched terms were the regulation of peptidase activity and the
   regulation of morphogenesis of a branching structure (Figures [83]S11
   and [84]S12, Supporting Information). The top significantly enriched
   KEGG pathways of acinar cells and beta cells at E16.5 were protein
   digestion and absorption and pancreatic secretion (Figure [85]S13,
   Supporting Information). Together, the results of GO and KEGG analysis
   at E16.5 and E18.5 suggested that the up‐regulated signaling pathways
   in acinar cells and beta cells were correlated with cellular response
   to metabolic stress.

   Secondary validation using TEM confirmed the presence of dilated
   cisternae in the endoplasmic reticulum, swollen mitochondria, and
   decreased mitochondrial cristae in the acinar cells of the PGDM (E18.5)
   group compared to controls (E18.5) (Figure [86]2G). In conclusion, we
   observed that the cellular response to metabolic stress was
   significantly affected in the fetal pancreas of the PGDM group at both
   E16.5 and E18.5 compared to controls.

2.3. The Impairment of Beta Cell Caused by Maternal PGDM in the Fetal
Pancreas at E16.5 and E18.5

   Transcription factor (TFs) played a crucial role in pancreas
   development and endocrine differentiation. Thus, we analyzed the
   activities of regulons within our snRNA‐seq data using single‐cell
   regulatory network interference and clustering (SCENIC) analysis. We
   first identified cell‐type‐specific regulons for each cluster,
   including Gata4(+), Nr5a2(+), Etv5(+), Atf6(+), and Thrb(+) in beta_1
   cells, and Jund(+), Ybx1(+), Mafb(+), Junb(+), and Cebpb(+) in beta_2
   cells (Figure [87]S14A, Supporting Information). Next, we explored the
   impact of maternal PGDM on each cell population by identifying the
   up‐regulation of cell‐type‐specific regulons in the PGDM group compared
   with controls (Figures [88]3A and [89]S14B, Supporting Information).
   Notably, Nkx6.2(+), Mafb(+), Mafa(+), and Pdx1(+), which were
   predominantly expressed in both beta_1 and beta_2 clusters, were
   significantly up‐regulated in the PGDM group compared to controls
   (Figure [90]3A). More specifically, the PGDM group demonstrated higher
   activity of these regulons in both beta_1 and beta_2 clusters
   (Figure [91]3B).

Figure 3.

   Figure 3
   [92]Open in a new tab

   Decreased insulin secretion accompanied by the compensatory increase of
   Pdx1, Nkx6.2, and Mafa in the PGDM group compared with the control
   group. A) SCENIC analysis revealed significant changes in the
   expression levels of regulons in the PGDM group (E18.5) within each
   cluster compared with the control group (E18.5). B) UMAP plots showing
   the regulon activity (“on‐red,” and “off‐gray”) of Nkx6.2(+), Mafb(+),
   Mafa(+), and Pdx1(+) in the PGDM group (E18.5) and controls (E18.5).
   C,D) Representative immunofluorescence images of fetal pancreas tissue
   sections at E18.5 for PDX1 (green), insulin (red), and nuclear stain
   DAPI (blue) in the PGDM group (C) and controls (D). E) Statistical
   analysis of immunofluorescence results. n  =  5/group (≈200
   β‐cells/group were quantified). F) MS/MS spectrum of Ins1 using
   timsTOF_HT mass spectrometer G) Quantification of Ins1 proteins using
   quantitative proteomics data of fetal pancreas in the PGDM group (E16.5
   and E18.5) and controls (E16.5 and E18.5). H) CellChat analysis showed
   decreased insulin signaling networks based on ligand‐receptor pairing
   in the PGDM group (E16.5), as compared with the control group (E16.5).
   I) Schematic showing regulons identified using SCENIC in the PGDM and
   control group. Abbreviations: SCENIC, single‐cell regulatory network
   inference and clustering; TF, transcription factor.

   Consistent with this observation, the PGDM group exhibited elevated
   gene expression levels of Pdx1, Nkx6.1, Nkx6.2, Mafb, and Mafa compared
   to the control group (Figure [93]S14C, Supporting Information). This
   finding was further supported by immunofluorescence (IF) staining
   results, which demonstrated significantly increased expression of the
   PDX1 protein in the PGDM group relative to the control group
   (Figure [94]3C–E).

   To further strengthen these results, we performed quantitative
   proteomic analysis of fetal pancreas at E16.5 and E18.5
   (Figure [95]3F). we first carried out KEGG pathway analysis to better
   understand the systemic effects of maternal PGDM on the fetal pancreas.
   We found that the top 10 significantly up‐regulated KEGG pathways in
   the PGDM group (E16.5 and E18.5) included one carbon pool by folate,
   spliceosome, steroid hormone biosynthesis, protein digestion and
   absorption, hippo signaling pathway, and PI3K−Akt signaling pathway,
   which provided a new direction for us to further understand the effects
   of maternal PGDM on fetal pancreas (Figure [96]S15, Supporting
   Information).

   Notably, the results of proteomic analysis showed that the protein
   expression level of Ins1 in the PGDM group (E16.5) was significantly
   lower (40.0% decrease, P = 0.01) than that in the control group (E16.5)
   (Figure [97]3G). A decreased Ins1 expression trend (39.4% decrease, P =
   0.07) was observed in the PGDM group (E18.5), even if statistical
   significance was not reached (Figure [98]3G). Furthermore, CellChat
   analysis revealed decreased insulin signaling networks based on
   ligand‐receptor pairing in the PGDM group (E16.5), as compared with the
   control group (E16.5) (Figure [99]3H). These results were consistent
   with the findings of reduced birth weight in the offspring of the PGDM
   group, compared to the control group (Figure [100]S1, Supporting
   Information).

   In conclusion, although multiple TFs including Pdx1, Mafa, and Nkx6.2
   exhibited increased activity in the PGDM group compared to the control
   group to compensate for insulin synthesis (Figure [101]3I), the
   impairment of endoplasmic reticulum and mitochondria (Figure [102]2G)
   caused by maternal PGDM resulted in compromised insulin secretion and
   beta cells.

2.4. The Correlation Between Maternal Metabolites and Fetal Pancreas
Development in the Presence of Maternal PGDM (E18.5)

   In addition, we performed a nontargeted metabolomic analysis of plasma
   samples from pregnant mice at E18.5, to enhance the understanding of
   the connection between maternal metabolism and fetal pancreatic
   responses. The metabolomic results showed significantly upregulated
   differential metabolites involved in glucose and lipid metabolism in
   pregnant mice with PGDM, such as Tagatose, Erythrose, D‐Allose,
   Fructose, Alpha‐d‐xylopyranose, D‐Glycero‐D‐gulo‐heptose, Oleoyl
   tyrosine, and N‐octanoylsphingosine 1‐phosphate (Figure [103]S16,
   Supporting Information). Correlation analysis revealed strong
   correlations between maternal metabolites involved in glucose
   metabolism and differentially expressed proteins that regulated
   glucolipid metabolism in the pancreas of offspring, such as Erythrose
   and OSTC, D−Allose and HMCS2 (Figure [104]S17, Supporting Information).
   In summary, these results offer additional mechanistic insights into
   the impact of maternal PGDM on fetal pancreas development.

2.5. Altered Intercellular Interactions Between Acinar Cells and Multiple
Cell Clusters in the Presence of Maternal PGDM (E16.5 and E18.5)

   Recent evidence has highlighted the crosstalk between acinar cells and
   endocrine cells (such as beta cells) in the context of diabetes.^[
   [105]^15 ^] Based on this, we next investigated cellular interactions
   between acinar cells and endocrine cells during maternal PGDM (Figure
   [106]4A). Initially, we observed that Zbtb2(+), Irf3(+), Meox1(+),
   Sox12(+), and Sox11(+) were significantly upregulated in the acinar_2
   cell population in the PGDM group compared to controls
   (Figure [107]3A). SCENIC analysis further revealed that
   genes regulated by Irf3(+), such as Try4, Prss1, Spink1, and Gm5771,
   were predominantly enriched in the acinar_2 cluster (Figure [108]4B).
   Consistent with this, we found that the PGDM group (E18.5) exhibited
   significantly higher expression levels of Try4, Try5, Prss2, and
   Gm10334 within the acinar_2‐cell population compared to the control
   group (E18.5) (Figure [109]4C).

Figure 4.

   Figure 4
   [110]Open in a new tab

   Cell–cell interaction analysis between acinar cells and multiple cell
   types in the presence of maternal PGDM (E18.5). A) Schematic of the
   intercellular communication in the PGDM and control groups. B) Heatmap
   showing the expression profile of candidate target genes regulated by
   Irf3 in each cell population based on snRNA‐seq data (E18.5). C) Violin
   plots showing the expression of Try5, Prss2, Try10, Pard3, Gm10334,
   Try4, and F2r across cell populations among the PGDM (E18.5) and
   control (E18.5) groups, based on snRNA‐seq data (E18.5). D) Heatmap
   highlighting differential cell interaction strength. Red represented
   increased interaction strength in the PGDM group (E18.5), and blue
   represented decreased interaction strength in the PGDM group (E18.5),
   based on snRNA‐seq data (E18.5). E,F) ST data based on spots (bin 10,
   10 × 10 nanoballs aggregated) showing spatial distribution of acinar_1
   cells, acinar_2 cells, alpha cells, beta_1 cells, beta_2 cells, delta
   cells, ductal cells, and neuronal cells in the PGDM (E18.5) and control
   (E18.5) groups. G) CellChat analysis showed that the ligand‐receptor
   pairing signaling network based on snRNA‐seq data was enhanced in the
   PGDM group (E18.5) compared to the control group (E18.5). H) Comparison
   of the relative information flow of each signaling pathway between PGDM
   (E18.5) and control (E18.5) groups in acinar_1 and acinar_2 cells,
   based on snRNA‐seq data (E18.5). Significantly enriched signaling
   pathways in the PGDM group were colored in red text, while signaling
   pathways significantly enriched in the control group were colored in
   blue. I) Circle plots illustrating the alterations of PARs signaling
   pathway network among the PGDM group (E18.5) and the controls (E18.5),
   based on snRNA‐seq data (E18.5). Abbreviations: snRNA‐seq,
   single‐nucleus RNA sequencing; PARs, protease‐activated receptors.

   Remarkably, the CellChat analysis demonstrated that interactions
   between acinar_2 cells and beta_2 cells, delta cells, ductal cells,
   fibroblast cells, lymphatic endothelial cells, and mesenchymal_2 cells
   were notably enhanced in the PGDM group, with increased interaction
   strength compared to the control group (Figure [111]4D). The ST data
   confirmed this by showing the spatial proximity between acinar_2 cells,
   alpha cells, beta_1 cells, beta_2 cells, delta cells, and ductal cells
   (Figure [112]4E,F). To further investigate the mechanisms underlying
   the interactions of acinar_2 cells with alpha, beta_2, delta, and
   ductal cells, we performed ligand‐receptor analysis using CellChat at
   E16.5 and E18.5. The analysis revealed significant up‐regulation of
   signaling interactions, including Try5‐Pard3, Try4‐Pard3, Prss2‐Pard3,
   and Gm10334‐Pard3, in the PGDM group (E18.5) (Figure [113]4G). And
   similar results were further corroborated by snRNA‐seq data (E16.5) and
   ST data (E16.5) (Figures [114]S18 and [115]S19, Supporting
   Information), suggesting that these interactions might mediate
   communication between acinar_2 cells and other cell types under the
   PGDM condition (E16.5 and E18.5) (Figure [116]4G). For example, we
   conducted the CellChat analysis using snRNA‐seq data (E16.5 and E18.5)
   and found that Try5‐Pard3 exhibited significantly up‐regulated
   signaling interactions in the PGDM group compared with the control
   group. Remarkably, acinar_2 cell population expressed significantly
   higher levels of Try5 in the PGDM group compared to controls
   (Figure [117]4C), and Try5 was specifically expressed in the acinar_2
   cell population. Furthermore, ST analysis revealed that Try5 and Pard3
   were spatially adjacent (Figure [118]S20, Supporting Information).
   Collectively, we believed the consistency of the findings strengthened
   our conclusions.

   In addition, significant signaling pathways between the PGDM and
   control groups were identified based on overall information flow within
   inferred signaling networks, particularly the protease‐activated
   receptors (PARs) signaling pathway (Figure [119]S19B, Supporting
   Information). Further analysis revealed that the PARs signaling pathway
   was highly active among acinar_1 cells and acinar_2 cells in the PGDM
   group (Figure [120]4H). Notably, the PARs signaling pathway network was
   significantly enhanced between the acinar_2 cluster and multiple cell
   clusters in the PGDM group compared to the control group
   (Figure [121]4I and Figure [122]S18A, Supporting Information).

   Overall, these findings suggested that the PARs pathway might play a
   critical role in mediating cell–cell interactions between acinar_2
   cells and various cell clusters (such as beta_1 cells, beta_2 cells,
   delta cells, and ductal cells) under maternal PGDM conditions (E16.5
   and E18.5).

2.6. Enhanced Intercellular Interactions Between Beta Cells and Endothelial
Cells (ECs) in the Presence of Maternal PGDM

   A previous study^[ [123]^16 ^] demonstrated a close physiological
   relationship between blood vessels and islets, with the blood vessel
   endothelium playing a critical instructive role in the development and
   differentiation of pancreatic endocrine cells. Building on this, we
   used snRNA‐seq data to investigate potential interactions between beta
   cells and ECs (Figure [124]5A). We observed that in the PGDM group, the
   increased interactions between beta_2 cells and other cell types were
   predominantly concentrated on ECs, in contrast to the control group
   (Figure [125]5B). Spatial distribution images corroborated this by
   showing the proximity of beta_2 cells to general ECs, lymphatic ECs,
   and vascular ECs (Figure [126]5C,D).

Figure 5.

   Figure 5
   [127]Open in a new tab

   Intercellular communication networks between beta cells and endothelial
   cells. A) A schematic model showing the intercellular communication
   between beta cells and endothelial cells. B) A Sankey diagram showing
   the upregulated interaction strength (PGDM vs control) from beta cells
   to other cell types in the cell–cell communication network, based on
   snRNA‐seq data (E18.5). C,D) ST data based on spots (bin 10, 10 × 10
   nanoballs aggregated) showing spatial distribution of beta_2 cell, GEC,
   LEC, and VEC in the PGDM (E18.5) and control (E18.5) groups. E) The dot
   plot, based on snRNA‐seq data (E18.5), displaying increased
   interactions between beta_2 cells, GEC, LEC, and VECs based on
   ligand‐receptor pairing in the PGDM group (E18.5) compared to the
   control group (E18.5). F‐H) Comparative relative information flow of
   each signaling pathway in beta_2 cells (F), VEC (G), and GEC (H) was
   analyzed on the basis of the snRNA‐seq data (E18.5) between PGDM and
   control groups. Significantly enriched signaling pathways in the PGDM
   group were colored in red text, and signaling pathways with blue text
   were significantly enriched in the control group. I) The violin plot,
   derived from snRNA‐seq data (E18.5), showing the gene expression levels
   for the indicated genes in beta cells and endothelial cells. Red
   represented the PGDM group; blue represented the control group.
   Abbreviation: GEC, general endothelial cell; LEC, lymphatic endothelial
   cell; VEC, vascular endothelial cell.

   To further understand the mechanisms behind these interactions, we
   compared ligand‐receptor interaction pairs between the PGDM and control
   groups, focusing on the beta_2 cells and ECs. Notably, the
   ligand–receptor pairs Vegfa‐Vegfr2[kdr], Vegfa‐Vegfr1[flt1],
   Ncam1‐Fgfr1, Mpzl1‐Mpzl1, and Dlk1‐Notch (Notch1, Notch2, and Notch4)
   were significantly up‐regulated in the PGDM group (Figure [128]5E).
   Information flow analysis further highlighted the activity of the MPZ,
   NCAM, and NOTCH pathways between beta_2 cells and ECs in the PGDM group
   compared to the control group (Figure [129]5F–H). Additionally, the
   PGDM group exhibited elevated expression of Vegfa, Fgfr1, Dlk1, Notch1,
   Notch2, and Notch4 compared to the controls (Figure [130]5I). Moreover,
   ST analysis showed that Vegfa and Kdr, and Dlk and Notch2 were
   spatially adjacent (Figure [131]S21, Supporting Information). These
   findings suggested that VEGF and Notch pathways were likely key
   mediators of the enhanced interactions between beta cells and ECs in
   the PGDM group.

2.7. Spatial Profiling Identified Cellular Niche Changes in Response to
Maternal PGDM

   A previous study^[ [132]^17 ^] found that beta cells tend to cluster in
   specific regions, forming insulin secretion hotspots. Additionally,
   gene expression levels in acinar cells, such as Try5, had been shown to
   vary based on their distance from the islets.^[ [133]^15a ^] This
   evidence highlighted the importance of maintaining a functional
   cellular niche for pancreatic biology. In this context, we performed
   neighborhood enrichment analysis to explore the impact of maternal PGDM
   on local cellular niches. To improve resolution, we annotated ST data
   using snRNA‐seq data with anchor‐based integration method^[ [134]^18 ^]
   (Figures [135]6A and [136]S22, Supporting Information). Next, we
   constructed a spatial neighbor graph, assigning a fixed number of
   nearest neighbors to each spot within the ST data
   (Figure [137]6B,D,F,H). The neighborhood enrichment analysis of ST data
   at E16.5 revealed a tight clustering of cells within each cluster in
   the PGDM group (Figure [138]6C), compared to a more dispersed
   distribution observed in the control group (Figure [139]6E). Similar
   results were obtained using ST data from E18.5 (Figure [140]6G,I),
   indicating that PGDM exposure might result in altered spatial
   organization within the fetal pancreas. Overall, these data suggested
   that maternal hyperglycemia might significantly impact the local
   cellular niche of the fetal pancreas.

Figure 6.

   Figure 6
   [141]Open in a new tab

   Spatial cellular niches in the fetal pancreas are affected by the
   maternal PGDM. A) Schematic representation of the cellular niches
   analysis workflow. First, cell types in stereo‐seq chips based on spots
   (bin 20, 20 × 20 nanoballs aggregated) were annotated with snRNA‐seq
   data by using “TransferData” function. Next, the spatial neighbor graph
   was computed to identify the cellular niches. Finally, the spatial
   neighborhood enrichment analysis was performed, and the enrichment
   score (Z‐score) was calculated. B, D, F, H) Representative spatial
   neighbor graph in the PGDM group (B) and control group (D) at E16.5,
   and the PGDM group (F) and control group (H) at E18.5. C, E, G, I)
   Spatial neighborhood enrichment analysis of the PGDM group (C) and
   control group (E) at E16.5, and the PGDM group (G) and control group
   (I) at E18.5.

3. Discussion

   Previous studies have demonstrated that pancreatic development is
   highly dependent on the interactions between islets and non‐endocrine
   islet cells, such as acinar cells,^[ [142]^19 ^] ductal cells,^[
   [143]^20 ^] blood vessels,^[ [144]^16 ^] and sympathetic innervation.^[
   [145]^21 ^] The pancreas, being a highly organized structure, is
   particularly vulnerable to disruptions in glucose metabolism during
   embryo development.^[ [146]^22 ^] However, little is known about the
   effects of maternal PGDM on the fetal pancreas, especially concerning
   interactions between various pancreatic cell types under maternal PGDM
   exposure. Utilizing snRNA‐seq data and ST analysis has enabled a more
   comprehensive understanding of these effects.

   In our study, we successfully established a PGDM animal model to
   evaluate the effects of maternal PGDM on the fetal pancreas. We
   employed snRNA‐seq data (E16.5 and E18.5) and ST analysis (E16.5 and
   E18.5) to explore the underlying molecular mechanisms. We identified
   the cellular heterogeneity in acinar and beta cell populations based on
   snRNA‐seq data (E16.5 and E18.5), each displaying unique DEGs, KEGG
   pathways, TFs, and metabolic pathways. Additionally, maternal PGDM
   induced significant metabolic stress in the fetal pancreas,
   characterized by dilated endoplasmic reticulum cisternae and swollen
   mitochondria. Remarkably, our results showed that multiple TFs,
   including Pdx1, Mafa, and Nkx6.2 were more active in the PGDM group
   than in the controls to compensate for insulin synthesis; however,
   maternal PGDM‐induced damage to the endoplasmic reticulum and
   mitochondria impaired insulin secretion and beta cells. Furthermore,
   the local cellular niche of the fetal pancreas and the intercellular
   communications between islets and non‐endocrine cells were
   significantly altered in the PGDM group compared to the controls.

   For the first time, we utilized snRNA‐seq technology to profile the
   transcriptomes of individual fetal pancreatic cells in the PGDM and
   control groups, thus providing valuable insights into the cellular
   heterogeneity within the fetal pancreas under maternal PGDM conditions.
   We observed heterogeneity in both acinar cells and beta cells in the
   PGDM and control groups, underscoring the functional diversity within
   these cell types. The heterogeneity of acinar cells had previously been
   reported in both healthy individuals and adult diabetic (db/db) mice.^[
   [147]^11 , [148]^15 ^] For instance, one study^[ [149]^11a ^]
   identified three different acinar subpopulations in healthy
   individuals–secretory acinar cells, acinar cells with high levels of
   regenerating protein, and idling acinar cells. Another study^[
   [150]^15a ^] demonstrated the presence of two distinct acinar cells in
   adult db/db mice–peri‐islet acinar cells, located close to the islet,
   and tele‐islet acinar cells, located farther from the islet–with
   differing morphology and function.

   Research into beta cell heterogeneity in non‐diabetic donors and adult
   db/db mice had also been documented.^[ [151]^11 , [152]^23 ^] A recent
   study^[ [153]^11c ^] identified three distinct beta‐cell subtypes–β1,
   β2, and β3 cells–with varying biological functions in non‐diabetic
   individuals. Another study^[ [154]^23 ^] revealed that adult db/db mice
   (aged 6 and 10 weeks) displayed eight distinct beta cell subsets, each
   differing in morphology and functions. Collectively, these findings
   supported our results, highlighting the complexity, heterogeneity, and
   adaptability of acinar cells and beta cells. Further analysis was
   required to explore these differences.

   We also noted that glycolysis/gluconeogenesis‐related pathways were
   significantly up‐regulated in multiple cell clusters, particularly
   acinar_1 cells, acinar_2 cells, beta_1 cells, and beta_2 cells, in the
   PGDM group (E18.5) compared to the control group (E18.5). Mitochondrial
   dysfunction might be responsible for the metabolic changes.
   Mitochondria were the principal sites for oxidative phosphorylation
   (OXPHOS) to produce ATP. Under maternal PGDM, mitochondrial function
   was impaired, causing the disruption in the OXPHOS process.
   Consequently, beta cells depended on glycolysis (the so‐called
   Warburg‐like effect) to compensate for the energy shortage for their
   rapid growth, proliferation, and insulin secretion.^[ [155]^24 ^]
   However, the long‐term activation of the Warburg effect produced more
   reactive oxygen species, leading to the dysfunction in beta cells and
   impairment of glucose‐stimulated insulin secretion.^[ [156]^25 ^]

   Our study also showed that maternal PGDM could disrupt the structure of
   endoplasmic reticulum and mitochondria compared to controls, reflecting
   the presence of endoplasmic reticulum stress in the fetal pancreas of
   PGDM group. These findings were supported by previous studies.^[
   [157]^26 ^] For instance, one study^[ [158]^26a ^] reported
   significantly increased DNA fragmentation in islet cells under high
   glucose conditions, which was mitigated by chemical inhibition of
   endoplasmic reticulum stress. Similarly, another study^[ [159]^26c ^]
   using a gestational diabetes mellitus (GDM) animal model demonstrated
   elevated expression of 18 ribosomal genes in islets from GDM offspring,
   coinciding with endoplasmic reticulum stress. Interestingly, serum from
   exercising mice and individuals had been shown to protect beta cells
   from ER stress‐induced apoptosis,^[ [160]^27 ^] and physical exercise
   had been shown to restore vascular homeostasis by reducing endoplasmic
   reticulum stress in diabetic mice.^[ [161]^28 ^] While physical
   activity during pregnancy was an intuitive strategy for reducing
   endoplasmic reticulum stress in the fetal pancreas, high‐quality
   clinical evidence supporting its efficacy was lacking. Current Chinese
   guidelines recommend exercise interventions for managing GDM,^[
   [162]^29 ^] and further studies are needed to determine whether such
   interventions could improve beta cell function by reducing endoplasmic
   reticulum stress in fetal pancreatic cells.

   It had been well‐established that Pdx1(+), Nkx6.2(+), Mafa(+), and
   Mafb(+) were critical TFs for beta cell development and maturation.^[
   [163]^30 ^] Our data indicated that these TFs were predominantly
   expressed in beta cells and were significantly upregulated in the PGDM
   group compared to controls. As mentioned previously, distinct
   mitochondrial and endoplasmic reticulum damage were observed in the
   PGDM group. Within the endoplasmic reticulum, the native structure of
   insulin was formed. The close collaboration between the endoplasmic
   reticulum and mitochondria was essential for insulin secretion.^[
   [164]^31 ^] Maternal PGDM‐induced damage to the endoplasmic reticulum
   and mitochondria impaired insulin secretion and the functions of beta
   cells. Therefore, insufficient insulin secretion might trigger the
   compensatory upregulation of Pdx1 to increase insulin production, since
   insulin is undeniably vital for the development of all organs and
   overall fetal growth of the fetus.^[ [165]^9b ^] In addition, a
   previous study reported that the offspring of GDM mice were born with
   prematurely developed beta cells, largely due to the increased
   expression of PDX1 in the GDM group.^[ [166]^22 ^] Collectively, our
   findings suggested that maternal PGDM might result in the impairment of
   beta cells in offspring at birth, potentially increasing their risk of
   metabolic diseases, such as diabetes, in adulthood.^[ [167]^32 ^]

   Increasing evidence has shown that cell–cell communication between
   exocrine and endocrine cells is essential for fetal pancreatic
   development.^[ [168]^12 , [169]^15 , [170]^33 ^] We found that the PGDM
   group (E18.5) expressed higher levels of Try4, Try5, Prss2, and Gm10334
   within the acinar_2‐cell population compared to the controls (E18.5),
   which might relate to islet expansion.^[ [171]^15a ^] Previous
   studies^[ [172]^15a ^] also demonstrated significantly higher mRNA
   levels of Try5 in “peri‐islet” acinar cells, located near the islet,
   compared to “tele‐islet” acinar cells, which were further from the
   islet, in adult db/db mice. This zonated gene expression signature was
   not observed in age‐matched control mice. Furthermore, the same study^[
   [173]^15a ^] reported significant increases in the expression levels of
   Try4 (4.4‐fold), Try5 (8.5‐fold), Try10 (9.2‐fold), Prss1 (1162‐fold),
   Prss2 (1.07‐fold), and Prss3 (338‐fold) in db/db mice compared to
   controls. These findings supported our observations.

   By analyzing ligand‐receptor pair communication, we also found a
   significant increase in interactions between acinar_2 cells and beta_2
   cells via the Try5‐Pard3 and Prss2‐Pard3 signaling pathways in the PGDM
   group (E16.5 and E18.5) compared to the control group (E16.5 and
   E18.5). Additionally, we observed enrichment of the PARs pathway in the
   PGDM group (E16.5 and E18.5) but not in controls (E16.5 and E18.5). It
   had been shown that PARs (PAR1‐4) were important mediators of
   acinar‐beta‐cell crosstalk and played a role in maintaining beta‐cell
   homeostasis.^[ [174]^15b ^] These findings extended prior research,
   indicating that the Try5‐Pard3, Try4‐Pard3, and Prss2‐Pard3 signaling
   pathways were critical to acinar‐beta‐cell crosstalk. Further
   elucidation of these crosstalk mechanisms was crucial for understanding
   alterations in fetal pancreatic cells in response to maternal PGDM.

   Though ECs were known to provide essential developmental signals
   supporting beta‐cell development,^[ [175]^16 ^] the mechanisms
   underlying beta cell‐EC interactions in fetuses exposed to maternal
   PGDM remained poorly understood. In our study, we discovered
   upregulated expression of ligand–receptor pairs between beta_2 cells
   and ECs in the PGDM group (E18.5) compared to controls (E18.5),
   particularly, the VEGFA‐VEGFR pair. Tosti et al.^[ [176]^11a ^]
   reported that ECs were located closest to pancreatic islets among
   pancreatic non‐epithelial cells in healthy individuals, suggesting that
   ECs played a critical role in beta‐cell proliferation and regeneration
   in a VEGFA‐dependent manner.^[ [177]^34 ^] After VEGFA induction, ECs
   proliferated in mouse islets, recruited circulating monocytes that
   differentiated into macrophages,^[ [178]^34 ^] and activated various
   growth factors and cytokines that promoted beta‐cell regeneration.^[
   [179]^34 ^] These findings suggested that further research was needed
   to determine the role of the VEGFA‐VEGFR signaling pathway in beta
   cell‐EC interactions in maternal PGDM.

   Maintaining niche homeostasis was essential for pancreatic development,
   yet the impact of maternal PGDM on cellular niches remained
   unidentified. One study^[ [180]^11a ^] quantified islet distribution
   during pancreas morphogenesis in healthy human tissue, reporting that
   the minimum distance between islet centroids is 400 µm. Another study^[
   [181]^35 ^] used a multi‐spectral imaging mass cytometry approach to
   analyze cellular niches in non‐diabetic (ND, n = 13) and type 2
   diabetes mellitus (T2D, n = 16) populations. In T2D pancreases, the
   study^[ [182]^35 ^] found that certain cell types, including alpha
   cells, beta cells, delta cells, ductal cells, macrophages, and CD8 T
   cells, exhibited closer spatial proximity than in ND pancreases. Our
   neighborhood enrichment analysis of ST data showed tight distribution
   within each cluster in the PGDM group, though further investigation was
   needed to determine the significance of these findings.

   There were several limitations to this study. First, although we
   integrated multi‐omics data to analyze the effects of intrauterine
   hyperglycemia on the fetal pancreas during critical developmental
   periods (E16.5 and E18.5), we did not analyze the pancreas at earlier
   stages (E8.5‐14.5) due to the small dimensions of the pancreas at these
   stages, which made multi‐omics analysis challenging. For example, even
   at E18.5, the pancreas was only 5 mm in length (Figure [183]1A).
   Second, pancreatic exocrine tissues contained high levels of hydrolytic
   enzyme activities,^[ [184]^11a ^] making the samples used for
   multi‐omics analysis demanding. The relatively small number of samples
   necessitated further validation of our findings. We utilized IF and TEM
   to independently confirm the multi‐omics data. Third, the impact of
   maternal PGDM on ribosomal genes needed to be interpreted with caution,
   as these ribosomal genes were not ideal markers for distinguishing
   cellular differences. Further studies were required to validate our
   findings at the single‐cell level, such as single‐cell proteomics.
   Additionally, we induced severe hyperglycemia using STZ intraperitoneal
   injection^[ [185]^2b ^] because severe hyperglycemia posed significant
   risks to the fetus. However, the effects of varying degrees of
   pregestational hyperglycemia (e.g., moderate vs severe) on pancreatic
   cells remained unclear.^[ [186]^36 ^] Further investigation was
   warranted. Finally, ethical and practical constraints limited clinical
   trials in pregnant women, making animal models the primary strategy for
   studying fetal pancreas development during pregnancy. Future studies
   with more human fetal pancreas samples were needed to authenticate our
   findings and expand our understanding of fetal pancreas development in
   maternal PGDM.

4. Conclusion

   Taken together, for the first time, we conducted a comprehensive
   investigation of the effects of intrauterine hyperglycemia on fetal
   pancreas using integrated snRNA‐seq and ST approaches. We observed
   pronounced cellular responses to metabolic stress, significantly
   decreased insulin secretion accompanied by the compensatory increase of
   Pdx1, Nkx6.2 and Mafa in PGDM offspring at birth, altered cell‐cell
   communications among multiple cell types, and distinct changes in
   cellular niches in the PGDM group compared to controls. These findings
   provide a foundation for further exploration of intrauterine
   hyperglycemia's effects on fetal pancreas and have implications for
   future PGDM‐related research and treatment strategies aimed at
   mitigating fetal pancreas injury during pregnancy.

5. Experimental Section

PGDM Mouse Model

   C57BL/6J mice were purchased from Beijing Weitong Lihua Experimental
   Animal Technology Co. Ltd. (Beijing, China). All mice were acclimatized
   to laboratory conditions (22–26 °C with a 12‐h light/dark cycle) for
   one week before the experiment. The Animal Ethics Committee of Peking
   University First Hospital approved all animal procedures (Approval No.
   J2023052). For the experimental group (PGDM group, Figure [187]1A),
   8–9–week–old female mice were administered an intraperitoneal injection
   of streptozotocin (STZ, # S0130, Sigma–Aldrich) at a dose of 80 mg/kg
   body weight/day for three consecutive days, as previously described.^[
   [188]^37 ^] The STZ was freshly dissolved in sodium citrate buffer
   (0.1 mol L^−1, pH 4.5, Solarbio). Control group mice (Figure [189]1A),
   also 8–9–weeks–old, were injected intraperitoneally with sodium citrate
   buffer (0.1 mol L^−1, pH 4.5, sterile, Solarbio). One week after the
   final injection, random glucose levels were measured using an Accu‐chek
   Performa glucometer (Roche, Germany). The PGDM model was confirmed
   when two consecutive random blood glucose readings exceeded
   14 mmol L^−1. After confirming successful PGDM induction, female mice
   were housed with male mice for overnight mating (female: male  =  2:1).
   The presence of a vaginal plug the following morning was considered
   embryonic day (E) 0.5. Pregnant mice were sacrificed on embryonic days
   E12.5, E14.5, E16.5, and E18.5 (Figure [190]S1, Supporting
   Information), and fetal pancreas tissue was rapidly dissected using a
   stereomicroscope (Leica).

Sample Preparation of snRNA‐seq

   Fetal pancreas samples were collected from PGDM and control groups
   (Figure [191]S2, Supporting Information). Nuclei were isolated from
   snap‐frozen tissue according to the manufacturer's protocols. Briefly,
   the frozen fetal pancreas was homogenized in homogenization buffer.
   After homogenization, filtration, and concentration, nuclei were
   separated and collected.

snRNA‐seq Library Preparation and Sequencing

   snRNA‐seq libraries were prepared using the Kit v3.1 (#1 000 268; 10×
   Genomics) following the manufacturer's instructions. Briefly, frozen
   fetal pancreas tissues were ground using a tissue grinder, and the
   nuclei were extracted using lysis buffer. The nuclei suspension was
   subjected to filtration (30 µm), washing, and density gradient
   centrifugation. The purified nuclei were then mixed with a
   reverse‐transcription reagent mix (#1 000 123; 10× Genomics) and loaded
   onto a microfluidic chip for library construction. Libraries were
   sequenced on a NovaSeq platform (Illumina). Data quality metrics were
   summarized in Table [192]S3 (Supporting Information).

snRNA‐seq Data Processing

   Raw sequencing data were aligned to the mouse genome (mm10) and
   processed using Cell Ranger (version 7.1.0). The resulting expression
   matrix was analyzed using the Seurat package (version 4.4.0) in R
   (version 4.3.2). Low‐quality nuclei were filtered out based on the
   following criteria: nFeature_RNA > 200, nFeature_RNA < 4000, percent.mt
   < 20, and nCount_RNA < 30 000.^[ [193]^38 ^] The filtered data were
   normalized using the “SCTransform” function in Seurat. Batch effects
   between samples were corrected using Harmony (version 1.2.0) with
   default parameters in Seurat “RunHarmony” function. Clustering was then
   performed using the “FindClusters” function with a resolution parameter
   set to 0.5. Dimensionality reduction was carried out using the
   “RunUMAP” function with dimensions 1:20. The final snRNA‐seq library
   contained 21657 nuclei (E18.5) and 38237 nuclei (E16.5), representing
   all major fetal pancreatic cell types.

Identification of DEGs

   DEGs analysis between the PGDM and control groups was performed within
   each single‐nucleus cluster using the “FindMarkers” function in Seurat,
   based on the Wilcoxon rank‐sum test. DEGs were defined as genes with a
   | log2 fold change (FC)| > 0.25 and an adjusted P‐value < 0.05.

GO Enrichment Analysis

   Differential expression analysis between PGDM and control groups was
   carried out with the R Seurat package using the “FindMarkers” function.
   For each cell cluster, the top 50 upregulated genes (avg_log2FC > 1 and
   P_val_adj < 0.05) were filtered as the selected DEGs between the PGDM
   and control groups. Subsequent GO enrichment analysis between PGDM and
   control groups based on the selected DEGs was conducted using the
   clusterProfiler package (version 4.6.2).^[ [194]^39 ^]

KEGG Pathway Enrichment Analysis

   KEGG pathway enrichment analysis between PGDM and control groups within
   each cluster was conducted using the clusterProfiler package (version
   4.6.2).^[ [195]^39 ^] KEGG terms with Benjamini‐Hochberg adjusted
   P‐values < 0.05 were considered significantly enriched.

Spatial Transcriptomics

   Reagents were obtained from the STOmics Gene Expression kit (#101KT114,
   BGI, Shenzhen, China), and the procedure was performed according to the
   manufacturer's instructions.

Sample Preparation and Tissue Permeabilization

   Fetal pancreas tissues were embedded in OCT and flash‐frozen in liquid
   nitrogen. Cryosections (10 µm thick) were then mounted onto the surface
   of Stereo‐seq chips. The chips were incubated at 37 °C for 5 min and
   then fixed in cold methanol (−20 °C) for 30 min. Subsequently,
   permeabilization was performed using PR enzyme (#F1‐230828, BGI,
   Shenzhen, China).

cDNA Library Preparation for Sequencing

   Following tissue permeabilization, reverse transcription mix
   (#101KT114, BGI, Shenzhen, China) was added to synthesize cDNA. After
   reverse transcription, tissues were removed, and cDNA release mix
   (#101KT114, BGI, Shenzhen, China) was applied to release the cDNA from
   the chip surface. The supernatant containing cDNA was purified using
   DNA clean beads (#N411‐02, Vazyme, Nanjing, China) and transferred into
   1.5 mL EP tubes for polymerase chain reaction (PCR) amplification.

   The amplified PCR products were then fragmented using Fragmentation Mix
   (#101KL114, BGI, Shenzhen, China), and spatial libraries were
   constructed with the library kit (#101KL114, BGI, Shenzhen, China).
   Finally, the libraries were sequenced on the BGI‐T7 sequencing platform
   (MGI, Shenzhen, China). Data quality for ST data was shown in Table
   [196]S4 (Supporting Information).

ST Data Processing

   Raw FASTQ files were processed using the SAW software
   ([197]https://github.com/STOmics/SAW) and converted into gene
   expression matrices for further analysis. ST data were analyzed using
   the Seurat package (version 4.4.0) in R (version 4.3.2). Briefly, the
   count matrices were normalized and scaled using the “SCTransform”
   method. Dimensionality reduction was then performed via the “RunPCA”
   function. Clustering and visualizations of bins were conducted with the
   “FindNeighbours”, “FindClusters”, and “RunUMAP” functions. Relative
   gene expression features were visualized using the “SpatialFeaturePlot”
   function in Seurat (version 4.4.0).

   Stereo‐seq chips had a spot diameter of ≈220 nm, with a subcellular
   resolution of 500 nm between spots.^[ [198]^40 ^] The bin sizes were as
   follows: bin 10 (10 × 10 nanoballs were aggregated) measured 5 µm ×
   5 µm, bin 20 (20 × 20 nanoballs were aggregated) measured 10 µm ×
   10 µm, and bin 50 (50 × 50 nanoballs were aggregated) measured 25 µm ×
   25 µm.^[ [199]^41 ^] Bins 10, 20, and 50 were selected based on the
   specific research objectives.^[ [200]^40 ^] The specific bin size used
   was reported in the corresponding figure legend.

The Visualization of the Spatial Distribution for Specific Cell Type

   To visualize the spatial distribution of specific cells of the ST data
   (E18.5), spots were annotated based on snRNA‐seq data in Seurat. And
   then the first top 150 markers in each cell type from snRNA‐seq data
   were calculated by the “FindAllMarkers” function as cell‐type‐specific
   signature genes. Finally, the cell type signature score was calculated
   for spots (bin 10) by using UCell package and visualized.^[ [201]^42 ^]

Cell‐Cell Communication Analysis

   Cell‐cell communication networks were analyzed through ligand–receptor
   interactions using the CellChat R package (version 1.6.1) based on
   snRNA‐seq data.^[ [202]^43 ^] First, the “rankNet” function in CellChat
   was used to compare information flow between the PGDM and control
   groups across major clusters. Next, significant intercellular
   communications between cell populations were predicted for each
   signaling pathway using the “computeCommunProbPathway” function.
   Dominant senders and receivers in the intercellular communication
   network were then identified.

Investigation of Transcription Factor Activity

   To identify differentially regulated regulons (i.e., TF and their
   putative target genes) between PGDM and control groups within each
   cluster, the pySCENIC analysis pipeline (version 0.12.1) was used.^[
   [203]^44 ^] The input matrix for pySCENIC consisted of gene expression
   counts for all cells. Briefly, gene co‐expression network analysis,
   TF‐motif enrichment, and regulon activity were performed using the
   “pyscenic grn”, “pyscenic ctx”, and “pyscenic aucell” functions,
   respectively. The Wilcoxon rank‐sum test was used to identify specific
   regulons that were differentially active between PGDM and control
   groups within each cluster. Binarized regulon activity was then
   projected onto UMAP trajectories.

Evaluation of Metabolic Heterogeneity at Single‐Nucleus Resolution

   Single‐nucleus metabolic activity was quantified using the scMetabolism
   (version 0.2.1) R package.^[ [204]^45 ^] The metabolic gene set
   signature scores, based on KEGG pathways, were evaluated using the
   “AUCell” method. These signature scores were then imported into the
   Seurat S4 object for visualization.

PAGA Trajectory Analysis

   PAGA^[ [205]^46 ^] was applied to assess the global connectivity
   topology among the selected cell clusters. Two distinct snRNA‐seq data
   (E16.5 and E18.5) were first integrated and corrected the batch effects
   with “sce.pp.harmony_integrate” function.^[ [206]^47 ^] The
   “scanpy.tl.paga” function was applied then to construct the PAGA graph
   with default settings.

Pseudo‐Time Analysis

   The Monocle2 (version 2.28.0) R package^[ [207]^48 ^] was applied to
   analyze pseudo‐time trajectories to predict developmental processes of
   acinar cells, beta cells and ductal cells. Briefly, the input to
   Monocle2 was created from the UMI count matrix of the highly variable
   genes using the “newCellDataSet” function. The “DDRTree” method was
   applied in the “reduceDimension” function to reduce dimensions, and
   “plot_cell_trajectory” function was performed for visualization.

Cell‐Type Niche Analysis

   A spatial graph was computed for each sample using the spatial
   coordinates of bin 20 to identify neighboring spots. Two spots were
   connected by an edge if they were adjacent in the spatial map. The
   spatial neighbor graphs were generated using the
   “sq.gr.spatial_neighbors” function, and neighborhood enrichment was
   analyzed using the “sq.gr.nhood_enrichment” function.^[ [208]^49 ^]

TEM

   TEM was performed on fetal pancreas samples to examine ultrastructural
   details. The samples were fixed in 2.5% glutaraldehyde, dehydrated in
   ethanol, embedded in resin, sectioned, and counterstained. The
   ultrastructure of the pancreas tissues was observed using a TEM (TECNAI
   G2 20 TWIN, FEI, Czech Republic) at magnifications of 4400x, 10000x,
   and 25500x.

Quantitative Proteomics

   To assess the robustness of the SnRNA‐seq data, quantitative proteomic
   analysis of fetal pancreas samples (E16.5 and E18.5) was performed.
   Briefly, the main analysis steps were described below. The protein was
   extracted from fetal pancreas (E16.5 and E18.5), and DIA quantitative
   proteomics analysis was performed by Beijing Qinglian Biotech Co.,Ltd
   (Beijing, China). MS spectra were acquired with mass range
   300–1500 m z^−1 with a resolution of 60000 FWHM (1222 m z^−1). Peptides
   were measured by a timsTOF_HT mass spectrometer (Bruker, Bremen,
   Germany). Then, data were processed by database searching against
   uniprot Mus musculus UP000000589 proteome database (54822 target
   sequences downloaded on 2024‐03‐07) using Spectronaut software
   (Biognosys, version 19.0). Lastly, based on the differentially
   expressed proteins (fold change cutoff = 1.2 and P‐value < 0.05)
   between PGDM and control (E16.5 and E18.5) groups, the biological
   functions of differentially expressed proteins were evaluated by using
   KEGG pathway analysis.

Metabolomics Analysis

   Plasma samples from 12 pregnant mice at E18.5 were prepared, and 180 µL
   of supernatant was transferred for LC‐MS analysis after centrifugation.
   The untargeted metabolomics analysis was performed by Metware
   Biotechnology Co., Ltd (Wuhan, China). The MS analysis was operated in
   the positive ion mode and negative ion mode using full scan analysis
   over m/z 75–1000 at 35 000 resolution. Differential metabolites between
   PGDM and control (E18.5) groups were identified by VIP score (VIP > 1)
   and P‐value (P‐value < 0.05, Student's t test).

Tissue Immunofluorescence (IF)

   Fetal pancreas samples (4 µm thick) were baked for 60 min at 60 °C,
   then dewaxed and rehydrated. Antigen retrieval was performed in sodium
   citrate buffer (pH  =  6.0, #ZLI‐9064, Zhongshan Golden Bridge,
   Beijing, China) under high pressure. The sections were then incubated
   in endogenous peroxidase blocking solution (#P0100B, Beyotime,
   Shanghai, China) for 15 min at room temperature, followed by blocking
   with 5% BSA for 30 min at room temperature. The samples were incubated
   overnight at 4 °C with primary antibodies against Pdx1 (#5679, RRID:
   AB_10 706 174, 1:200 dilution, Cell Signaling Technology, USA) and
   insulin (#GB12335, RRID: AB_2 923 360, 1: 1500 dilution, Servicebio,
   Wuhan, China). The next day, sections were incubated with 488‐labeled
   goat anti‐rabbit IgG (#A0423, RRID: AB_2 891 323, 1: 600 dilution,
   Beyotime, Shanghai, China) and Cy3‐labeled goat anti‐mouse IgG (#A0521,
   RRID: AB_2 923 334, 1: 600 dilution, Beyotime, Shanghai, China).
   Finally, sections were stained with 4,6‐diamidino‐2‐phenylindole (DAPI,
   #C1002, 1: 1000 dilution, Beyotime, Shanghai, China) for 5 min at room
   temperature, and immunofluorescent images were captured using a
   fluorescence microscope (DMi 8; Leica, Microsystems, Germany).

Statistical analysis

   Data were presented as mean ± SD. Statistical significance was assessed
   using the Wilcoxon rank‐sum test for nonparametric data and the t test
   for parametric data. A P value of less than 0.05 was considered
   statistically significant. All statistical analysis were conducted
   using R (version 4.3.2), unless otherwise specified.

Conflict of Interest

   The authors declare no conflict of interest.

Author Contributions

   Y.D. and S.W. contributed equally to this work. Y.D. and H.Y. conceived
   and designed the research. Y.D. and S.W. performed the experiments.
   Y.D. and Z.Y. performed the analysis of snRNA‐seq and spatial
   transcriptomics data. Y.D. and H.Y. wrote the manuscript. All authors
   helped revise the manuscript.

Supporting information

   Supporting Information
   [209]ADVS-12-2411126-s001.pdf^ (38.1MB, pdf)

Acknowledgements