To determine the roles of Lats1 and Lats2 in endocrine progenitor cells and pancreas development, both genes were inactivated in the Neurog3^Cre transgenic mouse strain. For tracing of Cre-recombinase activity, we also crossed the Rosa26^LSL-YFP line. The heterozygote mice (Neurog3^CreLats1^fl/+Lats2^fl/+Rosa26^LSL-YFP) and single knockout mice (Neurog3^CreLats1^fl/flLats2^fl/+Rosa26^LSL-YFP or Neurog3^CreLats1^fl/+Lats2^fl/flRosa26^LSL-YFP) from the initial F1 and F2 generations showed no abnormal phenotype. We, therefore, bred the single knockout mice with the Lats1&2 conditional allele line to get the Neurog3^CreLats1^fl/flLats2^fl/flRosa26^LSL-YFP mice (abbreviated as NL mice). The Lats1 single knockout littermates were used as control (Ctrl) (Figure 1—figure supplement 1A). NL mice were born in Mendelian ratios with no obvious differences from the littermate control. However, starting from 1 week of age, it became obvious that the NL mice were smaller in size than the control mice. By the time of postnatal day 19 (P19), the NL mice appeared smaller and weaker than their littermates (Figure 1—figure supplement 1B), with a much smaller pancreas (Figure 1—figure supplement 1C), and suffered from hypoglycemia. The majority of NL mice died before 3 weeks of age, exhibiting rear limb weakness or paralysis. Feeding the mice with 10% glucose (Xiao et al., 2010) or keeping them with their parents delayed, but did not prevent their death (data not shown), indicating that hypoglycemia only partially accounted for the early death. A few mice survived beyond 3 weeks with much smaller body sizes and infertility. These growth-related phenotypes could be attributed to the expression of Neurog3^Cre in the central nervous system (Song et al., 2010).

Histological analysis with Hematoxylin and eosin stain (H&E) staining showed that the acinar cells, ductal system, and islets could be clearly identified in control pancreases (Figure 1A), whereas remarkably disorganized structures were seen in the center of the NL pancreases including noticeable expansion of ductal cells and fewer acinar cells (Figure 1—figure supplement 1D). Immunohistochemical staining of Insulin (INS) revealed much smaller islets in NL pancreases in addition to structural disorganization, evident by immunofluorescent co-staining of Insulin and Glucagon (GCG) (Figure 1B). We further analyzed the mRNA expression level of lineage-specific genes in the P1 pancreas using quantitative RT-PCR (qPCR). We found that the mRNA expression of acinar-specific markers, Ptf1a, Amylase (Amy), and carboxypeptidase A1 (Cpa1), as well as endocrine-specific markers, Chromogranin A (ChrA), Insulin 1 (Ins1), and Insulin 2 (Ins2) were significantly lower in NL mice when compared with control mice (Ctrl, n = 4; NL, n = 4) (Figure 1C). In comparison, while the mRNA expression of two ductal-specific markers, Hnf1b and Sox9, had no significant change, Krt19 was significantly higher in NL pancreases (Figure 1C). To pinpoint the starting time of the observed defect in NL mice, we examined the E15.5 pancreas, when expression of Neurog3 reaches its peak. We did not observe any obvious differences between NL and control pancreases at E15.5 by H&E staining. The expansion of central ductal structure was noted in the NL pancreas at E16.5 which became even more obvious at P1 from H&E staining and KRT19 immunohistochemistry staining (Figure 1D). Together, these observations suggest that genetic disruption of Lats1&2 in Neurog3+ cells leads to acinar atrophy and ductal expansion in the postnatal pancreas, which are likely initiated in the developing embryonic pancreas.

In the developing pancreas, expression of Neurog3 marks the beginning of endocrine cell differentiation and reaches its peak at E15.5. However, as described before, apparent defects in pancreatic morphology were not observed through H&E staining until E16.5. Therefore, subsequent analysis was focused on the E16.5 pancreas. We monitored Cre-recombinase activity with Rosa26^LSL-YFP allele. Yellow fluorescent protein (YFP) expression was used as an indicator for Lats1&2 deletion as no viable antibodies were available for staining of LATS1&2. We also performed immunofluorescent staining of pancreatic and duodenal homeobox 1 (PDX1), an early pancreatic progenitor marker and β-cell-specific transcriptional factor. YFP+ cells formed cell clusters and were co-stained with PDX1 in control pancreases (Figure 2A). In contrast, in NL pancreases, the majority of YFP+ cells did not form clusters, but instead formed a distinct layer next to ductal cells. In addition, there were also much fewer PDX1 and YFP double-positive cells in NL pancreases, as compared with the control (Figure 2A), suggesting that most endocrine progenitors in the NL pancreas do not develop into β-cells at E16.5. Presence of a few double-positive cells in NL pancreases could be due to incomplete deletion of genes. At birth, control pancreases showed co-staining of YFP and PDX1 double-positive cells in the center region of islets and YFP single-positive cells in the periphery region of islets, which is consistent with the normal structure of islet of Langerhans where β-cells localize to the center of islets while α-cells localize to the periphery. In contrast, NL pancreases had much fewer PDX1-positive cells which were found in small unorganized cell clusters, while the majority of YFP+ cells did not express PDX1 (Figure 2—figure supplement 1). To further delineate the defect of Lats1/2 null endocrine progenitor cells, we performed immunostaining on the transcription factors Islet 1 (ISL1) and NK2 homeobox 2 (NKX2.2), which are expressed after Neurog3 in endocrine pancreas development (Arda et al., 2013). We found that ISL1 was widely expressed in YFP+ cells in control pancreases, but only expressed in a few small YFP+ cell clusters in NL pancreases, in which the majority of YFP+ cells did not express ISL1 (Figure 2B). A similar expression pattern was also observed for NKX2.2 (Figure 2C). Thus, these data indicate that loss of Lats1&2 leads Neurog3+ endocrine progenitor cells to halt the differentiation program almost immediately and prevents them from further endocrine lineage progression.

During the endocrine formation process, several morphogenetic events coincide with cell differentiation including epithelial-to-mesenchymal transition (EMT) and delamination of differentiating endocrine cells during pancreatic development (Gouzi et al., 2011; Rukstalis and Habener, 2009). The new model proposed by Sharon et al., 2019 suggests that differentiating endocrine precursor cells undergo ‘leaving the cord’ or ‘delamination’ process with lower E-cadherin (CDH1) expression. How Neurog3 mediates CDH1 downregulation and facilitates ‘leaving the cord’ is unclear. We observed that YFP+ cells were connected to epithelial cords but expressed lower CDH1 than the cord epithelial cells in control E16.5 pancreases (Figure 3A). YFP+ cells in the NL pancreas were connected to the epithelial cords and formed buds or sheaths similar to YFP+ cells in the control. However, they express CDH1 at a level as high as of that in the epithelial cord (Figure 3A). With respect to the few escapees as previously described in PDX1 expression, the NL pancreas, too, contained a few escapees marked by lower CDH1 expression. Interestingly, we found that the YFP+ cells in NL pancreases did not stay as ‘budding’ or ‘sheath’ as development continued, and returned to single epithelial cells at P1 (Figure 3—figure supplement 1). These data suggest that an active Hippo pathway is required for downregulation of CDH1 expression in Neurog3-expressing progenitor cells, but is dispensable for ‘leaving the cord’ or ‘delamination’.

We have observed that the differentiation of Lats1&2 null endocrine progenitor cells was blocked by lack of expression of ISL1 and NKX2.2 transcription factors. To expand on this, we interrogated the question of whether these progenitor cells returned to the ductal cell fate. To do so, we performed immunofluorescent staining and found that, similarly to the control pancreas, there was no overlap of YFP and SOX9 expression in the NL pancreas at E16.5 or P1 (Figure 3B). In addition, we examined another known ductal cell marker, cytokeratin 19 (KRT19) in the NL pancreas. We observed that KRT19 was not present in YFP+ cells in control pancreases but was highly expressed in YFP+ cells of NL pancreases at E16.5 and P1 (Figure 3C). In fact, at E16.5, the expression of KRT19 in YFP+ cells of NL pancreases was even higher than the ductal cells, which later became similar to ductal cells at the P1 stage (Figure 3C). These data suggest that an early effect of Lats1&2 deletion in Neurog3+ cells is to activate KRT19 expression, but not SOX9 expression, further indicating that KRT19 expression is not controlled by SOX9, but instead by YAP1.

Following gene knockouts of Lats1&2, we observed that NL mice died before 3 weeks of age with a much smaller pancreas. Histological analysis showed acinar atrophy and disorganized ductal expansion in the NL pancreas, which are frequently associated with pancreatitis (Lerch and Gorelick, 2013). Our previous study involving acinar-specific deletion of Lats1&2 (using Ptf1a^CreER model) has found that loss of Lats1&2 first activates pancreatic stellate cells (PSCs) (Liu et al., 2019). We wondered whether a similar mechanism happened in the NL pancreas. Immunostaining showed that there was a significant increase of Vimentin-positive mesenchymal cells in the NL pancreas at E16.5 (Figure 4A). Those expanded mesenchymal cells were closely associated with YFP+ Lats1&2 null cells (Figure 4—figure supplement 1A). These mesenchymal cells were positive for activated PSC marker alpha smooth muscle actin (ACTA2) (Figure 4—figure supplement 1B). Next, we determined whether there was immune cell infiltration in the NL pancreas. We observed CD45+ immune cells with large cell bodies, similar to the size of epithelial cells, in the E16.5 pancreas. We then stained tissue with macrophage marker F4/80 and found that these CD45+ cells were macrophages. The number of macrophages was increased in NL mice at E16.5, especially in the ductal area (Figure 4B, D), and further dramatically increased at P1 in the whole pancreas (Figure 4C, D). Overall, quantification analysis showed that NL pancreases contained a significantly greater density of macrophages at E16.5 and P1 in comparison to control pancreases at E16.5 and P1 in both the acinar and ductal regions of the pancreas (E16.5: Ctrl, n = 3; NL, n = 3. P1: Ctrl, n = 4; NL, n = 3) (Figure 4D). Together, these results indicate that loss of Lats1&2 in Neurog3+ endocrine progenitor cells activates PSCs in the developing pancreas, and recruits immune cells, continuously.

Loss of Lats1&2 in Neurog3-expressing endocrine progenitor cells blocks the endocrine lineage from further differentiation. We hypothesized that uncontrolled downstream Hippo effectors, YAP1 and TAZ, mediate this phenotype. We observed that YAP1-positive cells had increased CTNNB1 and KRT19 expression in control and NL pancreases at E16.5 (Figure 5A), which is consistent with our RT-qPCR experiments measuring relative mRNA expression level (Figure 1C). It was observed that YFP+ cells had no YAP1 expression in control pancreases at E16.5 (Figure 5A), suggesting YAP1 protein expression predominated in the epithelial cord region, which is consistent with previous studies (Gao et al., 2013; Duvall et al., 2022). However, YAP1 expression was much higher in NL pancreases, with many YFP+ cells co-localizing with positive YAP1 nuclei staining despite a few escaped YFP+ cells lacking YAP1 expression (Figure 5—figure supplement 1A). We observed a similar YAP1 staining pattern in P1 pancreases of NL mice. We further co-stained Neurog3 and YAP1 on E16.5 pancreases and found that, in the control pancreas, YAP1 was expressed in Neurog3+ cells located in the epithelial cord but was sequestered into the cytoplasm (Figure 5B). No YAP1 expression was detected in the cytoplasm or the nucleus in the Neurog3+ cells after they left the epithelial cord (Figure 5B). In comparison, much fewer Neurog3+ cells, but more YAP1+ cells, were observed in the NL pancreas, overall. In addition, YAP1 co-localized with Neurog3 in the nuclei of cells residing in the epithelial cord of NL pancreases (Figure 5B). Using qPCR, we then further quantified mRNA expression of the YAP1 targets that we previously identified (Luo et al., 2016). We found that the mRNA expressions levels of Ankrd1, Cyr61, and TGFb1 had no significant changes, but Ctgf, SPP1, Cxcl10, and Cxcl16 were significantly higher in the NL pancreas at P1 compared to control pancreas (Ctrl, n = 4; NL, n = 4) (Figure 5C). Immunofluorescent staining showed that SPP1 protein was expressed in ductal cells in control pancreases, but was also positive in YFP+ cells in NL pancreas at E16.5 and P1 (Figure 5—figure supplement 1B). Together, these data demonstrate the increased expression of YAP1 and its target genes in Lats1&2 null cells.

To examine genome-wide transcriptional changes in NL mutant mice, we performed single-nucleus RNA sequencing (snRNA-seq) on E17.5 pancreata from control and NL embryos. After initial data processing and removal of low-quality cells (see Methods), we retained 1784 cells from control and 2198 cells from NL mice. Unsupervised clustering and uniform manifold approximation and projection (UMAP) analysis revealed nine distinct cell types represented in both the control and NL pancreas tissue (Figure 6A, B, Figure 6—figure supplement 1A). We annotated these cell populations using hallmark cell-type-specific genes and gene expression profiles (Figure 6—figure supplement 1B). Two independent NL mutant embryos showed similar pattern, demonstrating the reproducibility of snRNA-seq analysis (Figure 6—figure supplement 1C, D). When we compared the percentage of different cell types in control versus NL mice, we observed that the ductal cell population expanded at the expense of endocrine lineage in NL mice, confirming our histology findings (Figure 6C). In addition, the NL pancreas displayed elevated numbers of immune cells and myofibroblasts (Figure 6C). Immune cell cluster is comprised of three sub-populations, including a Il7r⁺ Il18r1⁺ T cell cluster and two Mrc1⁺ macrophage clusters (Figure 6D, E). Looking at additional macrophage marker genes showed that NL pancreas contained more Ly86⁺Lgmn⁺Mertk⁺ macrophages. These results are in line with the previous findings that Yap1 activation may contribute to macrophage recruitment (Liu et al., 2019; Yang et al., 2020). Analysis of fibroblast and myofibroblast combined populations identified a total of four subclusters, including two myofibroblast clusters, a fibroblast cluster, as well as a mesothelial cluster (Figure 6F). Interestingly, the two myofibroblast clusters have distinctive distribution between control and NL pancreas. Specifically, the control pancreas was associated with more Ctnna2⁺Ncam1⁺ myofibroblast, whereas NL pancreas was associated with more Hmcn2⁺Dgkb⁺ myofibroblast (Figure 6G). Gene Set Enrichment Analysis (GSEA) showed that myofibroblast 2 of NL pancreas enriched genes for extracellular matrix organization, collagen formation and ECM–receptor interaction (Figure 6—figure supplement 1E). These data suggested that the loss of Lats1&2 in endocrine progenitor cells re-shaped the tissue microenvironment.

To further characterize the differences between control and NL pancreas, we analyzed the acinar population. Three sub-clusters were identified, including a normal acinar cluster – Acinar 1, with expression of various digestive enzymes and acinar transcription factors (Cel, Pnliprp1, Cpb1, Rbpjl, and Mecom) (Figure 6H, I). This population was mainly found in control pancreas. In comparison, Acinar 2, existed in both control and NL pancreas, was found to highly express Dlc1 but lower digestive enzymes especially Cel, suggesting an immature acinar state. Interestingly, Acinar 3 were mainly found in NL pancreas and express low level of digestive enzymes.

Analysis on ductal cells revealed a total of four sub-clusters, which were all associated with substantial Onecut2, Sox9, Etv6, and Cdh1 expression, confirming their ductal lineage (Figure 6J, K). Control pancreas mainly comprised of Ductal 1, while NL pancreas has more Ductal 2 and 3, with Ductal 4 uniquely to NL pancreas (Figure 6J). Among these sub-clusters, Ductal 2 had the highest Prom1 progenitor marker expression (Figure 6K). Interestingly, Ductal 3 showed high expression of Notch1, Notch3, and Dclk1, as well as endocrine cell markers Isl1 and Pax6. Similar to our staining results, we found Krt19 and Ctnnb1 to be highly expressed in Ductal 2, 3 and 4 cells. YAP and Wwtr1 were also high in these populations, however, their target gene Ctgf mainly expressed in Ductal 3 and 4 with highest expression in Ductal 4 of NL pancreas. Interestingly, Sema3e, Csf1, and Pdgfb had similar expression pattern as Yap target, which may indicate a potential YAP1 regulation (Figure 6K). In addition, differential gene expression analysis and K-means clustering revealed distinct gene expression patterns among the four ductal sub-populations (Figure 6—figure supplement 1F). Gene Ontology analysis of the Ductal 4 highly expressed genes showed enriched migration, wound healing, GTPase-mediated signaling, cell matrix adhesion, TGFβ signaling, and Hippo pathway (Figure 6—figure supplement 1G). These data suggest that loss of Lats1&2 expression leads Neurog3-expressing cells differentiate to a new ductal-like population.

Analysis of pancreatic cells only (acinar, ductal, and endocrine) also revealed three acinar sub-populations, four ductal sub-populations as well as endocrine α- and β-cell populations (Figure 6L). To study the dynamic processes of cellular development and differentiation, we performed single-cell trajectory analysis. Because the gene expression data suggested that Ductal 2 population has progenitor cell signature in developing pancreas, we set Ductal 2 population as root and traced the possible differentiation path of ductal cells into acinar, α- and β-cells using pseudotime analysis (Figure 6M). The Ductal 2 path split into endocrine lineage and into Ductal 4, a population specific to NL pancreas. By checking cell cycle status, NL pancreas generally has more cells in S and G2M phases (Figure 6—figure supplement 1H). Specifically, Acinar 1, the major population in control pancreas, has fewer cells in S/G2M than Acinar 2 and 3, the major ones in NL pancreas (Figure 6N). It was also shown that a large number of proliferating ductal cells were present in NL pancreas (Figure 6N, Figure 6—figure supplement 1H). Altogether our data suggest that Lats1&2 deletion results in high YAP1 and TAZ which block further differentiation of ductal lineage, resulting in fewer endocrine cells.

To further demonstrate that YAP1 and TAZ are the downstream effectors of Lats1&2 in endocrine lineage development, we performed gene knockouts of Yap1 and Wwtr1 in addition to Lats1&2 in Neurog3+ endocrine progenitor cells. To this end, we intercrossed Neurog3^CreLats1^fl/flLats2^fl/+ with Yap1^fl/flWwtr1^fl/fl mice for several generations to obtain the following gene knockout mice models: Neurog3^CreLats1^fl/flLats2^fl/+Yap1^fl/flWwtr1^fl/fl (named as NTY where Lats2 is heterozygous and Yap1/Wwtr1 are null), Neurog3^CreLats1^fl/flLats2^fl/flYap1^fl/+ Wwtr1^fl/fl (named as NLT where Yap1 is heterozygous but Wwtr1 is null), and Neurog3^CreLats1^fl/flLats2^fl/flYap1^fl/flWwtr1^fl/fl (named as NLTY with quadruple null Lats1&2, Yap1, and Wwtr1). First, we found that NTY mice were normal with no defect in the pancreas, suggesting that YAP1/TAZ are no longer needed in Neurog3-expressing cells for endocrine lineage development. Because NL mice showed an obvious defect in the postnatal P1 pancreas, we focused on and analyzed the P1 pancreases of NLT and NLTY mice. Histological analysis showed that, unlike NL pancreases, NLTY pancreases were similar to the control pancreas. However, NLT pancreases still showed some abnormality in the ductal network region, suggesting incomplete rescue when YAP1 expression remained (Figure 7—figure supplement 1A). We further analyzed endocrine cells by staining YFP, INS, and Somatostatin (SST), and observed that only a small fraction of YFP+ cells, scattered throughout the NL pancreas, were INS- and SST-positive (Figure 7A and Figure 7—figure supplement 1B). However, similar to control pancreases, most YFP+ cells in NLT and NLTY pancreases were INS- and SST-positive, suggesting that reduced levels of YAP1/TAZ rescued the endocrine defects of loss of Lats1&2. However, we still observed some ductal-like YFP+ cells in NLT pancreases, suggesting incomplete rescue compared to NLTY pancreases (Figure 7A). Furthermore, we analyzed immune cell infiltration and found that unlike NL pancreas, there were much fewer macrophages in both NLT and NLTY pancreases (Figures 7B, 4C). In addition, quantification analysis showed that overall macrophage density is slightly elevated in NLT and NLTY pancreases compared to control pancreases, but is not statistically significantly (Ctrl, n = 4; NL, n = 3; NLT, n = 3; NLTY, n = 3) (Figure 7C). Taken together, our results demonstrate that YAP1/TAZ are the downstream effectors of Lats1&2 during endocrine differentiation, and a tight control of their activities is required for normal pancreatic development.

The established role of Hippo in regulating cell proliferation prompted us to investigate whether inactivation of Lats1&2 affects pancreatic β-cell proliferation and function. To address this, we deleted Lats1&2 in pancreatic β-cells using the mouse-insulin-promoter-1-driven CreER (Ins1^CreER, named as ML). We first tested whether Lats1&2 are necessary for pancreatic β-cell function at the adult stage (Figure 8A). The mice were subjected to 5-day Tamoxifen (TAM) injection at 180 mg/kg/day, and pancreases were collected 4 weeks after TAM injection for histological analysis. Normal pancreatic architecture and similar endocrine cell mass were observed between control and ML pancreases through H&E staining (Figure 8B). To confirm the β-cell-specific Lats1&2 deletion efficiency, we performed western blot analysis and were unable to detect LATS1&2 protein in islets of ML mice (Figure 8C). To further test the influence of Lats1&2 deficiency on β-cell function, we performed glucose tolerance tests (GTT). Results indicate that ML mice displayed a normal GTT profile when compared to control mice (Ctrl, n = 6; ML, n = 5) (Figure 8D).

Next, we tested whether Lats1&2 is required for embryonic β-cells. To this end, we injected the pregnant mice carrying Ins1^CreERLats1&2^fl/flRosa26^YFP offspring with Tamoxifen at E12 (Figure 8E). The Tamoxifen treatment induced the majority Insulin+ β-cells to express YFP at P1, showing an efficient gene deletion by this procedure (Figure 8F). However, the Lats1&2 deletion did not affect pancreatic architecture and endocrine pancreas morphology (Figure 8F). Together, these findings suggest that Lats1&2 are dispensable for pancreas β-cell proliferation and function.

All animal study protocols were approved by the UT Health San Antonio Animal Care and Use Committee. Neurog3^Cre mice (stock number: 005667), MIP^CreER (stock number: 024709), and Rosa26^LSL-YFP mice (stock number: 006148) were obtained from The Jackson Laboratory. Neurog3^Cre mice were kindly provided by Dr. Andrew Leiter and Dr. Seung Kim. Lats1^fl/fl and Lats2^fl/fl mice were kindly provided by Dr. Randy L. Johnson. Yap1^fl/fl and Wwtr1^fl/fl mice were kindly provided by Dr. Eric N. Olson. We generated (1) Neurog3^CreRosa26^LSL-YFPLats1^fl/flLats2^fl/+ mice (as Control), (2) Neurog3^CreRosa26^LSL-YFPLats1^fl/flLats2^fl/fl mice (NL mice), (3) Neurog3^CreRosa26^LSL-YFPLats1^fl/flLats2^fl/+Yap1^fl/fWwtr1^fl/fl mice (NTY mice), (4) Neurog3^CreRosa26^LSL-YFPLats1^fl/flLats2^fl/flYap1^fl/+Wwtr1^fl/fl mice (NLT mice), (5) Neurog3^CreRosa26^LSL-YFPLats1^fl/flLats2^fl/flYap1^fl/flWwtr1^fl/fl mice (NLTY mice), and (6) MIP^CreERRosa26^LSL-YFPLats1^fl/flLats2^fl/fl mice (ML mice). All offspring were genotyped by PCR of genomic DNA from the toe with primers specific for the Neurog3^Cre, MIP^CreER, Rosa26^LSL-YFP, Lats1, Lats2, Yap1, and Wwtr1 loci. For the timed mating experiment, male mice were introduced to the cage in the afternoon and removed in the morning on the second day, which was considered gestational day 0.5 (E0.5). The pregnant mice were then used to harvest the embryos at the indicated time. For the tamoxifen-mediated Lats1&2 deletion in adult β-cells, 6- to 8-week-old mice were intraperitoneal (i.p.) injected with 200 mg/kg tamoxifen (TAM, Sigma-Aldrich, T5648) for 5 consecutive days. For the early β-cell Lats1&2 deletion, female Lats1&2^fl/fl mice and male mice carrying MIP^CreER; Lats1&2^fl/fl were timed-mated and after 12 days, the pregnant mice were injected with 100 mg/kg tamoxifen once (George et al., 2015). PCR was used for validation of knockout alleles (Table 1).

Pancreases harvested from animals were fixed in 4% paraformaldehyde overnight and then submitted to the Histology Core of the University of Texas Health Science center at San Antonio. The tissues were immersed in serial dilutions of ethanol and xylene and then embedded in the paraffin. The blocks were then cut into 5 µm sections with a microtome. H&E staining was performed at the Histology Core. For Picrosirius red staining, the section slides were first deparaffinized with xylene and washed in serial dilutions of ethanol solutions. Then, the slides were stained with Hematoxylin, 0.1% Fast Green, and Pico-Sirius Red solution (Polysciences, 24901) as previously described (Trinder et al., 2016).

For immunofluorescent staining, tissues were deparaffinized, rehydrated, and submerged in 200°C heated R-Universal Epitope Recovery Buffer solution (Electron Microscopy Sciences, Hatfield, PA, 62719-20) for 30 min and then let cool at room temperature for 25 min. Sections were permeabilized using 0.3% tris-buffered saline with Triton (TBST) (0.3% Triton X-100, Acros Organics, Fair Lawn, NJ, 21568-2500) and 0.025% phosphate-buffered saline with Tween20 (PBST) (0.025% Tween20, Fisher Bioreagents, Fair Lawn, NJ, BP337-500) for 4 min each. Sections were subsequently blocked with 10% donkey serum in 0.025% PBST for 35 min at room temperature. Sections were then incubated with primary antibodies diluted in 10% donkey serum in 1× phosphate-buffered saline (PBS) at 4°C overnight. Then, sections were incubated with fluorescent-tagged Alexa Fluor secondary antibodies (1:250, Jackson ImmunoResearch, West Grove, PA) diluted in 10% donkey serum in 1× PBS for 1 hr at room temperature. Additionally, sections were incubated with diamidino-2-phenylindole dihydrochloride (DAPI) (1:1000, Invitrogen, Carlsbad, CA, P36935) for 4 min at room temperature. Finally, sections were covered with a drop of VectaShield Vibrance Antifade Mounting Medium (Vector Laboratories, Inc, Burlingame, CA, H-1700). All images were captured using Microsystems DMI6000 B microscope and software (Leica Microsystems, Buffulo Grove, IL) and Zeiss LSM510 confocal microscopes. All primary and secondary antibodies used are listed inTables 2 and 3, respectively.

Pancreases from the animal were homogenized in Trizol (Invitrogen, 15596026) using a probe sonicator from Qsonica (25%, 15 s ON, 15 s OFF) immediately after harvest. RNA extraction and reverse transcription real-time PCR (RT-qPCR) were performed as previously described (; Qin et al., 2018). Relative expression of mRNA was calculated based on the ^ΔΔCt method. The primers for quantitative real-time PCR are listed in Table 1.

Islets from Lats1&2^fl/fl and Ins1^CreER; Lats1&2^fl/fl mice were isolated as described (Qin et al., 2018). After 3 hr of recovery, 50–60 islets were handpicked into low retention tubes (Eppendorf, Z666548). Total protein was extracted by incubating the islets with Laemmli buffer at 95°C for 5 min and homogenized for another 5 min in a sonication device (Diagenode, B01060001) at 4°C. Protein concentration was determined by the BCA method (Thermo Fischer Scientific, 23225) and 5–10 µg of total protein were loaded onto the 10% sodium dodecyl sulfate–polyacrylamide gel electrophoresis gel for electrophoresis. All antibody information and working concentrations are shown inTables 2 and 3.

Blood glucose levels of the animal were determined by a glucometer (BIONIME GS550). The GTT was performed on overnight-fasted 5-month post-tamoxifen Lats1&2^fl/fl and Ins1^CreER; Lats1&2^fl/fl mice. 1.5 g/kg glucose was intraperitoneal injected into animals and blood glucose levels were monitored at indicated time points in the figures.

All macrophage quantification was performed using the free, open-source program, Fiji (Schindelin et al., 2012). First, immunofluorescent staining of DAPI and F4/80 macrophage marker was performed on each pancreas section. Four randomly chosen, representative areas of each mouse pancreas were captured at ×20 objective magnification. Each fluorescent channel of the chosen areas was saved as separate images. Each of these separate channel images were then overlayed using the Image Calculator Tool in Fiji in order to visualize DAPI and F4/80 signal, together. Once combined, using the Channels Tool, each channel can be overlayed or removed as desired. The number of pixels per micrometer was calculated and defined using the Straight Line Tool and Measure Tool in Fiji. This allows for the total area of any given selection to be measured in micrometers squared. Using the DAPI image channel and Threshold tool, a threshold limit was set to 15 for the minimum and 255 for the maximum, and total DAPI-positive area (in micrometers squared) within this threshold parameter was calculated for each embryonic pancreas region. Then, each F4/80+ DAPI+ cell was hand counted within each respective region using the Multi-point tool in Fiji. Finally, the total number of counted F4/80+ macrophage cells was divided by the measured DAPI+ area. This yields the calculated macrophage density of total number of F4/80+ macrophage per micrometer squared in regions of the pancreas. For quantification of total macrophage density per image, the calculated density of each acinar- and ductal-cell region per image were averaged together. All images were captured using Microsystems DMI6000 B microscope and software (Leica Microsystems, Buffalo Grove, IL).

All quantification results in this study except bioinformatic analysis described below were presented as the mean ± standard error of the mean. Statistical analysis was performed by a two-tailed Student’s t-test. All p-values <0.05 were considered statistically significant. Sample size was not calculated before the experiment. The sample size was determined based on the availability of generated conditional knockout mice to provide at least three biological replicates for a sufficient statistical power.

Mouse E17.5 pancreas was flash frozen in liquid nitrogen and minced to smaller chunks on ice with sterile micro-dissection scissors. The finely minced tissues were transferred to a pre-chilled 2 ml Dounce homogenizer containing 2 ml cold 1× Homogenization Buffer (HB) which contained 0.26 M sucrose, 0.03 M KCl, 0.01 M Mg₂Cl, 0.02 M Tricine–KOH pH 7.8, and Ultra-Pure Water (Thermo Fisher), with the following added on the day of tissue processing – 1 mM Dithiothreitol (DTT), 0.5 mM Spermidine (Sigma-Aldrich), 0.15 mM Spermine (Sigma-Aldrich), 0.3% NP40, 1× Roche Complete Protease Inhibitor Cocktail, 1 U/μl RNase Inhibitor (Fisher Scientific). The tissue was homogenized on ice by moving a ‘loose’ pestle up and down 10–15 times to disrupt tissue structure, followed with a ‘tight’ pestle to achieve a final homogenate. This homogenate was then strained through a 70-μm Flowmi strainer into a 2-ml DNA/RNA LoBind tube and centrifuged at 4°C at 350 rcf for 5 min. After the centrifugation, the supernatant was discarded, and the pellet was washed with 1 ml cold PBS, passed through a 40-μm Flowmi strainer, and centrifuged again. For density gradient separation, the isolated nuclei were resuspended in 400 μl of 1× HB and combined with an equal volume of 50% chilled Iodixanol solution. Subsequent layers of 30% and 40% Iodixanol solutions, diluted in 1× HB, were gently layered under the initial mixture to create a gradient. In the final step, the tube with the nuclei was centrifuged using a pre-cooled bucket-rotor centrifuge at 3000 rcf for 20 min at 4°C. After centrifugation, the top layer was carefully removed, without disrupting the ‘fuzzy’ white ring of nuclei. A narrow-bore pipette tip was used to collect no more than 200 μl of the nuclei, which were then mixed with an equal volume of 1× HB. Nuclei concentration was determined using a Countess II FL automated cell counter (Thermo Fisher Scientific), adjusting to a final concentration optimal for 10x Genomics Chromium library preparation.

Base calling was performed using RTA 3.4.4, demultiplexing was performed using cellranger-arc v2.0.1 (Bcl2fastq 2.20.0), and alignment was performed using cellranger-arc v2.0.1 (BWA 0.7.17-r1188). Sequenced reads were aligned to a 10x Genomics provided mouse reference sequence (refdata-cellranger-arc-mm10-2020-A-2.0.0). The subsequent analysis was performed using Seurat v.5.0.1 in R software v.4.2.0. A quality control step was performed in merged control and mutant samples to remove low-quality cells, including possible cell debris (nFeature <200), cell doublets (nFeature >10,000), and cells with excessive cellular stress (>5% mitochondrial gene expression). The merged/filtered Seurat object was then subjected to a series of standard Seurat commands including ‘NormalizeData’, ‘FindVariableFeatures’, ‘ScaleData’, ‘RunPCA’, ‘FindNeighbors’, ‘FindClusters’, and ‘RunUMAP’, with 20 PCA dimensions (determined by elbow plot with ndims = 50) and a resolution of 0.2, in order to identify different cell populations present in the dataset. ‘FindAllMarkers’ function was performed to identify marker gene expression in each cell population identified in the previous step (only.pos = TRUE, min.pct = 0.25, logfc.threshold = 0.25). Each cell population was then assigned to a specific cell type by manually reviewing their corresponding marker gene expression with prior knowledge. Individual gene expression was visualized using ‘VlnPlot’, ‘FeaturePlot’, and ‘DotPlot’ functions in Seurat with default setting. The cell cycle status of the cells was determined using ‘CellCycleScoring’ command in Seurat with default setting. To analyze each individual cell type, the merged Seurat object was subset to smaller objects only containing cells with specific cell identities. The subset Seurat objects were then subject to similar analysis for re-clustering (including ‘FindVariableFeatures’, ‘ScaleData’, ‘RunPCA’, ‘FindNeighbors’, ‘FindClusters’, and ‘RunUMAP’) followed by further analysis. GSEA as well as individual plotting was performed using ClusterProfiler v4.4.4 in R software. In order to compensate the significant difference in reading depth between Myofibroblast 1 and 2 clusters, fold change of gene expression was calculated using the averaged gene expression of all the cells in each cluster, instead of the raw read counts. In addition, only genes with >25% expression in each cluster were included in the GSEA analysis. A significant enrichment was considered with multiple-test adjusted p-value <0.05.

To identify differentially expressed genes (DEGs) among the four Ductal sub-clusters, ‘FindMarkers’ Seurat function was performed to compare each pair of the four sub-clusters. The resulting genes were filtered to identify DEGs defined as log2 fold change >1.5 or <−1.5, and adjusted p < 0.01. A pseudobulk sequencing dataset for each Ductal sub-cluster was created by manually aggregating the gene read counts of all cells in each cluster. The pseudobulk dataset was then passed to DESeq2 v.1.36.0 and ComplexHeatmap v.2.13.2, to perform K-means clustering in order to identify expression patterns of the DEGs across the four Ductal sub-clusters. After identification of K-means clustering, the information was passed back into Seurat to plot gene expression heatmap at single-cell level. Overrepresentation analysis for lists of genes was performed using the clusterProfiler v.4.4.4 package in R with default settings. A significant enrichment was considered with multiple-test adjusted p-value <0.05.

Single-cell trajectory analysis was performed using Monocle3 v.1.3.4 in R software with default setting. The subset Seurat object containing all acinar, ductal, and endocrine cells were passed into Monocle3 to infer their cell-type transition states, by setting Ductal 2 population as the root.