LATS1/2 suppress NFκB and aberrant EMT initiation to permit pancreatic progenitor differentiation

The Hippo pathway directs cell differentiation during organogenesis, in part by restricting proliferation. How Hippo signaling maintains a proliferation-differentiation balance in developing tissues via distinct molecular targets is only beginning to be understood. Our study makes the unexpected finding that Hippo suppresses nuclear factor kappa-light-chain-enhancer of activated B cells (NFκB) signaling in pancreatic progenitors to permit cell differentiation and epithelial morphogenesis. We find that pancreas-specific deletion of the large tumor suppressor kinases 1 and 2 (Lats1/2PanKO) from mouse progenitor epithelia results in failure to differentiate key pancreatic lineages: acinar, ductal, and endocrine. We carried out an unbiased transcriptome analysis to query differentiation defects in Lats1/2PanKO. This analysis revealed increased expression of NFκB activators, including the pantetheinase vanin1 (Vnn1). Using in vivo and ex vivo studies, we show that VNN1 activates a detrimental cascade of processes in Lats1/2PanKO epithelium, including (1) NFκB activation and (2) aberrant initiation of epithelial-mesenchymal transition (EMT), which together disrupt normal differentiation. We show that exogenous stimulation of VNN1 or NFκB can trigger this cascade in wild-type (WT) pancreatic progenitors. These findings reveal an unexpected requirement for active suppression of NFκB by LATS1/2 during pancreas development, which restrains a cell-autonomous deleterious transcriptional program and thereby allows epithelial differentiation.


Introduction
Hippo is a major regulator of the critical balance between progenitor cell renewal, proliferation, and differentiation during embryogenesis [1]. The Hippo pathway consists of a kinase cascade, including the upstream serine/threonine kinases 4 and 3 (or mammalian STE20-like protein kinases 1 and 2 [MST1/2]) [2,3] kinases 1 and 2 (LATS1/2), which in turn phosphorylate the transcriptional coactivators yesassociated protein 1 (YAP1) and WW domain containing transcription regulator 1 (WWTR1; or transcriptional coactivator with PDZ-binding motif [TAZ]). Phosphorylation by LATS1/2 sequesters pYAP1/pTAZ in the cytoplasm and thereby inhibits their transcriptional activities. Hippo pathway factors have been shown to govern cell fate decisions in many organs, including the pancreas [2,3]. Pancreatic progenitor epithelial cells give rise to endocrine, acinar, and ductal lineages, coinciding with branching and tube development [4][5][6]. At later stages of pancreas development, the transcriptional coactivator YAP1 was shown to promote ductal fate at the expense of endocrine and acinar cells [2,7,8]. Furthermore, YAP1 and TAZ are robustly activated in adult human pancreatic ductal adenocarcinoma (PDAC) and chronic pancreatitis [9,10], and YAP1 is required for progression of PDAC [11][12][13]. Together, these reports indicate that the Hippo pathway controls pancreatic progenitor differentiation and adult pancreatic cell proliferation; however, the downstream molecular mechanisms are still largely unclear.
Another conserved transcription factor (TF) pathway known to govern cell proliferation is nuclear factor kappa-light-chain-enhancer of activated B cells (NFκB) signaling [14]. NFκB and Hippo govern other similar cellular processes in addition to proliferation, such as EMT, and both are required for PDAC progression in mice [15][16][17][18]. Although reciprocal signaling between YAP1/TAZ and NFκB signaling has been reported in cell culture [19]-suggesting that these pathways may interact-little is known about mechanistic crosstalk between them during embryonic development or in adult tissues. Whether the Hippo pathway interacts with NFκB in maintaining a proliferation-differentiation balance is unknown.
Here, we report that the Hippo kinases LATS1/2 are required for suppression of NFκB in pancreatic progenitors to allow branching morphogenesis and differentiation of pancreatic cell lineages. Transcriptional profiling reveals hyperactive NFκB signaling when Lats1/2 are genetically ablated from embryonic pancreatic epithelium. Without LATS1/2 kinases, a burst of NFκB activator expression causes rampant cell-autonomous NFκB-dependent signaling, which overrides normal differentiation programs and initiates aberrant EMT. We show that this occurs via YAP1 hyperactivation and at least in part via inappropriate activation of the pantetheinase vanin1 (VNN1) and increased reactive oxygen species (ROS). Our findings identify a critical requirement for active cell-autonomous suppression of deleteriously elevated NFκB signaling by the Hippo pathway in pancreas progenitors, and we show that this suppression is required for proper developmental morphogenesis and differentiation.
In WT pancreas, we observed a progressive expansion of epithelial volume accompanied by complex branching morphogenesis between E10.75 and E12.5 (S1B Fig), as previously established [4,23,24]. By contrast, differences in organ morphology were already present in Lats1/ 2-deficient pancreas at E11.5, at which time we observed an abrupt, transient expansion in pancreatic volume (S1B Fig). Strikingly, the mutant pancreas did not initiate branch formation. By E13.5, WT pancreas volume rapidly surpassed that of Lats1/2 PanKO (S1C Fig). Whereas the WT pancreas developed into a characteristic branched gland, the mutant pancreas remained small and rounded, similar to the lung upon loss of Lats1/2 [25]. Together, these data indicate an early, transient increase in epithelial volume and subsequent morphogenetic failure following loss of Lats1/2.
At E15.5, the WT pancreas consisted of a branched, ramifying epithelial network intercalated with GCG + alpha cells (Fig 1K). By contrast, we observed defective epithelial architecture without discernible branches nor alpha cells in Lats1/2 PanKO pancreas ( Fig 1L). Together, these data point to concurrent failure of epithelial expansion, epithelial branching, and endocrine cell differentiation at the earliest stages of pancreatic differentiation.

LATS1/2 kinase suppression of YAP1/TAZ is required for pancreatic progenitor differentiation and subsequent morphogenesis
Because LATS1/2 kinases are known to phosphorylate the transcriptional coactivators YAP1 and TAZ, resulting in their exclusion from the nucleus, we next tested LATS1/2 activity by examining YAP1 and pYAP1 immunolocalization. In WT pancreas, both nuclear YAP1 and cytoplasmic pYAP1 were present (S1F and S1G Fig). By contrast, in the Lats1/2 PanKO pancreatic bud, increased nuclear YAP1 (arrowhead, S1H Fig) and total loss of cytoplasmic pYAP1 (asterisk, S1I Fig) were observed, indicating efficient deletion of Lats1/2 flox alleles by E10.5. One day later, we observed nuclear TAZ and YAP1 expression in a subset of WT epithelial cells at E11.5 to E12.0 (S1J and S1K Fig). Conversely, increased nuclear TAZ and YAP1 expression were detected in Lats1/2 PanKO pancreas at E11.5, further underscoring abrogation of LATS1/2 kinase function in the early Lats1/2-deficient pancreas (arrowheads, S1J and S1K Fig). Therefore, nuclear localization of YAP1 and TAZ, as well as loss of cytoplasmic YAP1/ TAZ, verify inactivity of LATS1/2 in mutant mice by E10. 5.

LATS1/2 are required for pancreatic epithelial cell polarity and branching
Following Lats1/2 deletion, we observed a complete failure of epithelial branching. Previous studies have shown that proper evolution of pancreatic epithelial architecture is required for appropriate progenitor cell differentiation [24,28,30,31]. Therefore, we sought to characterize the onset of these early architectural defects by examining the microlumens that normally form and coalesce into a ductal plexus [30]. We performed MUC1 immunostains and found a marked progressive loss of MUC1 protein prior to the secondary transition in Lats1/2 PanKO (S4A Fig). Indeed, Muc1 mRNA was abolished by E13.5 in Lats1/2 PanKO (S4B Fig). Because these observations suggested loss of cell polarity, we examined epithelial apicobasal polarity and cell ultrastructure. Using classical cell apical polarity markers including protein kinase C iota (PKCI) and the Golgi protein golgin A2 (GOLGA2), we found a distinct displacement of apical surfaces, which was accompanied by a progressive hyperfusion of microlumens, an increase in lumen size, and a loss of the expected ductal plexus (S4C Fig).
We next assessed possible defects in cell shape. Measuring and comparing cell apical and basal membranes, we observed a widely variable, increased apical:basal ratio in the Lats1/ 2 PanKO (S4D Fig). We calculated this ratio based on an analytical method previously published to examine defects in bile duct lumenogenesis [32], in which variable or high apical:basal ratio measurements indicated failure of apical constriction. To assess apical constriction, which is critical for microlumen formation [4,31], we immunostained for phosphorylated myosin light chain 2 (pMYL2) and found its complete absence at the apical membrane of pancreatic epithelial cells (S4E Fig), suggesting the absence of expected rosette formation and loss of lumen size control. The apicobasal polarity abnormalities displayed in the Lats1/2 PanKO are summarized in the schematic in S4F Fig.

LATS1/2 kinases suppress EMT initiation
Throughout the Lats1/2 PanKO epithelia, we noted that the appearance and shape of mutant cells were altered and that integrity of the epithelium was ultimately lost. In contrast to typical cuboidal epithelial cells in the WT pancreas (Fig 1M), Lats1/2 PanKO cells showed spindleshaped cells reminiscent of mesenchyme ( Fig 1N). We therefore examined intermediate filament protein vimentin (VIM), which is normally enriched in mesenchymal cells [33]. WT pancreas epithelium completely lacks VIM at E11.5 (Fig 2A). By contrast, we observed ectopic high VIM expression in Lats1/2 PanKO pancreas (inset, Fig 2B), suggesting acquired mesenchymal characteristics soon after Lats1/2 deletion. We note that despite increased mesenchymal characteristics, the mutant epithelium retained relatively strong CDH1 expression, indicating that loss of Lats1/2 likely initiates an EMT but that this transition is not complete.

Loss of LATS1/2 kinases initiates an EMT transcriptional program and loss of apicobasal polarity in pancreatic epithelium
To investigate the molecular mechanisms disrupting pancreatic identity in the Lats1/2 PanKO pancreas, we carried out RNA sequencing (RNA-seq) of E11.0 mutant pancreata (Fig 3A), soon after Pdx1Cre early -driven deletion [21]. RNA-seq data were analyzed by comparing amplified total RNA from individual Lats1/2 PanKO (n = 3) and WT pancreas buds (n = 5). The validity of the RNA-seq data was confirmed, as the Hippo signaling pathway was identified within the top 10 up-regulated Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways ( Fig 3B). The most highly up-regulated pathway identified was the maturity onset diabetes of the young (MODY) gene set. YAP1 and TAZ transcriptional coactivators frequently bind TEA domain TFs (TEAD) to drive gene expression [7,41]. Differential expression of established TEAD target genes in RNA-seq data, summarized in S1 Table,  were counterstained with DAPI (blue). Scale = 25 μm. (K) Schematic illustrating partial EMT following Lats1/2 deletion, with acquisition of mesenchymal proteins including VIM, ACTA2, and TAGLN, normally absent from WT pancreas epithelium, and persistent CDH1. ACTA2, alpha smooth muscle actin; CDH1, E-cadherin; E, embryonic day; EMT, epithelial-mesenchymal transition; KRT19, cytokeratin 19; TAGLN, transgelin; VIM, vimentin; WT, wild type.
https://doi.org/10.1371/journal.pbio.3000382.g002 LATS1/2 constrain NFκB to permit pancreatic progenitor differentiation KEGG pathway analysis of our RNA-seq data also indicated abnormal up-regulation of cell adhesion molecule and tight junction pathways in the Lats1/2 PanKO pancreas (Fig 3B), likely a compensatory response to the observed defects in EMT, apicobasal polarity, and cell shape. Indeed, Hallmark pathway analysis confirmed a significant activation of EMT gene expression (S7B Fig). Loss of cell polarity and changes in cell shape are necessary steps in EMT [42]. Previous studies from our lab and many others have shown that polarity, cell movement, differentiation, and proliferation must be precisely controlled to coordinate proper pancreas organ formation, supporting the idea that Lats1/2 deletion hijacks these developmental mechanisms, ultimately impeding morphogenesis.

LATS1/2 kinases are required to suppress Vnn1 via YAP1 in the developing pancreas
To address what mechanisms might be driving these cellular defects, we analyzed genes altered in the Lats1/2 PanKO . The most highly up-regulated gene in Lats1/2 PanKO pancreas was Vnn1 (approximately 32-fold), which encodes a pantetheinase, a glycolipid-bound ectoenzyme that converts pantetheine into pantothenic acid and cysteamine [43]. While Vnn1 mRNA was undetectable in WT pancreas bud at E11.5 (asterisk, Fig 3C) and only low protein expression was detected by IF (Fig 3F), strong Vnn1 mRNA expression was observed in Lats1/ PanKO pancreas at E11.5 and E13.5 (Fig 3D and 3E). We also observed significantly increased VNN1 protein levels in Lats1/2 PanKO pancreas at E11.5 and E13.5 ( Fig 3G, S7C and S7D Fig). We asked whether YAP1/TAZ interaction with TEAD could drive Vnn1 expression. Indeed, motif analysis of chromatin immunoprecipitation sequencing (ChIP-seq) data show a TEAD binding site within 500 bp surrounding the Vnn1 transcription start site (TSS), suggesting a potential YAP1/TAZ/TEAD TF complex driving Vnn1 expression ( Fig 3H). These data strongly suggest that loss of Lats1/2 triggers aberrant VNN1 expression in the early pancreatic epithelium.
We asked whether Lats1/2 loss acted via YAP1 to induce VNN1 expression by testing the effect of constitutively active YAP1 expression on VNN1 activity. In E12.5 WT pancreas, VNN1 was detected at low levels at the apical membrane lining lumens ( Fig 3I). Like Lats1/2 depletion, YAP1 activation led to robust VNN1 expression in GFP + cells observed in Yap1 GFP pancreas at E12.5 (arrowheads, Fig 3J). VNN1 localized to the plasma membrane, while YAP1 GFP expression was nuclear. These findings suggest that aberrantly increased YAP1/TAZ activity leads to excess VNN1 expression in the cells that express YAP1 and that, normally, LATS1/2 maintain VNN1 at low levels via suppression of nuclear YAP1 in the developing pancreas. for RNA-seq of total RNA from individual WT (n = 5) and Lats1/2 PanKO (n = 3) embryos. Pancreatic bud is outlined (black). The heat map was generated using genes that were differentially expressed by at least 2-fold in Lats1/2 PanKO (mutant) compared to WT (control). Gene down-regulation and up-regulation are indicated in green and red, respectively. (B) Selected pathways from KEGG pathway analysis are shown. The x-axis indicates (-log 10 ) fold change, and the vertical red line designates significance, with a threshold of 1.5-fold change. Underlying data analysis can be found in S1 Data. (C, D) RNA ISH of Vnn1 antisense probe on WT and Lats1/2 PanKO pancreata (outlined in black) at E11.5 (n = 3 embryos per genotype). " � " denotes undetectable Vnn1 mRNA in WT pancreas epithelium. Scale = 50 μm. (E) Normalized Vnn1 mRNA expression was compared in Lats1/2 PanKO and WT pancreata at E13.5 (n = 3 embryos per genotype). Data are shown as mean ± SEM. Statistical significance was determined by Student t test ( �� p < 0.01). Underlying numerical values can be found in S1 Data. (F, G) Confocal images of VNN1 and CDH1 immunostains of WT and Lats1/2 PanKO pancreata at E11.5 (n = 3 embryos per genotype). Scale = 10 μm. (H) TEAD motif analysis in the Vnn1 promoter. (I, J) Confocal images of GFP, VNN1, and CDH1 immunostains of WT and Ptf1a-rtTA;Yap1-GFP (Yap1 GFP ) pancreata at E12.5 (n = 3 embryos per genotype). Arrowheads denote immunopositivity for both GFP and VNN1. Arrow points to cell with low GFP and no VNN1 expression. Scale = 20 μm. AU, arbitrary units; C, control; CDH1, E-cadherin; CTNNB1, catenin beta 1; E, embryonic day; GFP, Green Fluorescent Protein; ISH, in situ hybridization; KEGG, Kyoto Encyclopedia of Genes and Genomes; M, mutant; RNA-seq, RNA sequencing; TEAD, TEA domain transcription factor; TSS, transcription start site; Vnn1, vanin1; WT, wild type.

LATS1/2 kinases suppress a pro-oxidant program in the developing pancreas
Because Vnn1 was the most highly up-regulated gene in the Lats1/2 PanKO and was absent from early WT progenitors, we reasoned that aberrant VNN1 might drive abnormalities downstream of Lats1/2 deletion. VNN1 and its pantetheinase product cysteamine are known to promote oxidative stress [44]. Consistent with this idea, further RNA-seq analysis revealed that multiple Gene Ontology (GO) processes regulating ROS metabolism and biosynthesis were significantly increased in Lats1/2 PanKO total RNA compared to WT (S8A Fig).
As proof of principle that the Lats1/2 i-PanKO recapitulates that of the Lats1/2 PanKO pancreas, we performed mosaic deletion of Lats1/2 flox alleles in the Lats1/2 i-PanKO system. Deletion was induced using one-quarter of the normal tmx dosage at E8.5 only, and pancreata were examined at E13.5. While cytoplasmic and nuclear YAP1 expression was observed in WT littermates, robust nuclear YAP1 immunostaining was observed in Lats1/2 mutant clones (Lats1/ 2 mosaic i-PanKO ) at E13.5 (S8B Fig). A TAGLN + artery was observed in WT pancreas section (arrow), while the clone exhibited robust TAGLN immunopositivity, colocalizing with the high YAP1 + cells (arrowhead, S8B Fig). These data support the idea that the Lats1/2 i-PanKO system indeed recapitulates that of Lats1/2 PanKO .
To pinpoint the earliest time point at which oxidative stress was triggered following Lats1/2 deletion, pancreata were explanted at E10.5, and CellROX activation was quantified 2 days later (since pancreatic explants grow more slowly in ex vivo culture compared to in vivo development [48]). WT and Lats1/2 PanKO pancreas explants exhibited similarly low levels of Cell-ROX fluorescence (S8C and S8D Fig), indicating that oxidative stress is triggered later than E11.5 (approximately, given cultured explant). Therefore, oxidative stress is temporally downstream of Vnn1 up-regulation, which occurs at E11.0.
To determine whether the ROS increase was primarily in Lats1/2 PanKO epithelial cells or the result of increased immune cell infiltration, as seen in disease contexts [49], we asked whether immune cells invaded the mutant tissue, which could in turn spur the oxidative stress response [50]. To test this idea, we assessed leukocyte infiltration in the Lats1/2 PanKO pancreas prior to morphogenetic failure. We found no infiltrating protein tyrosine phosphatase receptor type C (PTPRC) + leukocytes within either WT or Lats1/2 PanKO pancreas epithelium at E11.5 (S8E and S8F Fig). We found only a few PTPRC + leukocytes outside the Lats1/2 PanKO bud epithelial periphery (arrows, S8E Fig), suggesting initiation of a cell-autonomous epithelial cell response at E11.5. By E14.5, however, leukocyte infiltrate could be detected within the epithelium and was significantly increased in Lats1/2 PanKO pancreas relative to WT (S8G and S8H Fig). These results showed that the oxidative stress response preceded any immune infiltration, indicating that pancreatic defects in Lats1/2 PanKO are not secondarily caused by recruitment of inflammatory cells and that initial increased ROS levels are epithelial cell autonomous. Studies have shown that VNN1 increases oxidative stress through its product cysteamine [51]. To mimic a cellular environment high in VNN1 [52], independent of LATS1/2 kinase activity, we treated WT pancreas epithelium with cysteamine and asked whether this affected redox homeostasis in WT pancreata. Such a finding would suggest that ROS response detected upon loss of LATS1/2 may largely be mediated by elevated VNN1 and its product cysteamine. Indeed, we detected robust increase in ROS following cysteamine treatment of WT pancreas epithelium (Fig 4E and 4F). We confirmed the specificity of the assay by treatment with the antioxidant N-acetyl-cysteine (NAC) as a control (Fig 4G and 4H). Quantification of the ROS levels showed an 81-fold increase upon cysteamine treatment ( Fig 4I). Thus, this experimentally elevated VNN1-cysteamine signaling promotes a pro-oxidative state in WT pancreas, mimicking the effect of depleting LATS1/2.
To determine whether aberrant ROS signaling was causing partial EMT, we treated E12.5 explants with an antioxidant to inhibit ROS and measured TAGLN expression. TAGLN was negligibly expressed within WT pancreatic explants, with or without NAC (Fig 4J and 4K). As expected, TAGLN + pixels were robustly increased in Lats1/2 i-PanKO explants ( Fig 4L) compared to WT. NAC, however, did not down-regulate TAGLN expression in Lats1/2 i-PanKO explants ( Fig 4M and 4N), indicating that 16 hour antioxidant treatment alone does not rescue the mesenchymal phenotype in Lats1/2-deficient pancreata. Overall, these data strongly suggest that LATS1/2 normally suppress VNN1-cysteamine signaling and the downstream prooxidative gene program in pancreatic progenitors during normal homeostasis and development but that either parallel pathways drive EMT (TAGLN expression) or short-term NAC treatment is insufficient to block these deleterious effects.
https://doi.org/10.1371/journal.pbio.3000382.g004 LATS1/2 constrain NFκB to permit pancreatic progenitor differentiation Noting that both WT and mutant pNFκB1 + cells were enlarged and rounded (arrowheads, Fig 5A and 5B), which are hallmarks of epithelial mitotic cells [54], we asked whether pNFκB1 + cells were proliferating. Indeed, approximately 49% of pNFκB1 + progenitor cells were pHH3 + in both WT and Lats1/2 PanKO pancreata at E14.5 (arrowheads, Fig 5D and 5E). This confirmed that NFκB1 was activated in proliferating pancreatic cells during development, and this relationship was maintained in the Lats1/2 PanKO . Yet the overall proportions of pHH3 + epithelial cells were higher in Lats1/2 PanKO pancreata compared to WT at E10. 75 Because pNFκB1 activity persisted, though aberrantly increased cell proliferation did not, we next addressed why the Lats1/2 PanKO pancreas failed to expand in size as development progressed. We examined cell death by cleaved caspase3 (CASP3) immunostaining, because NFκB signaling can promote proliferation or cell death depending on the cellular context [55]. Few CASP3 + epithelial cells were observed in WT pancreas at E11.5-E14.5, and there was no significant difference in the proportion of CASP3 + epithelial cells in Lats1/2 PanKO pancreas at E11.5 (S9E Fig), a time point when the Lats1/2 PanKO pancreatic bud was enlarged relative to WT. Beginning at E12.5, however, the rate of cell death increased significantly in Lats1/2-deficient pancreas compared to WT, and this rate continued to increase from E13.5 to E14.5 (S9E and S9F Fig). In addition, because the mutant bud at these stages failed to branch, and pancreatic vasculature is normally closely entwined with WT branching epithelium [56] (arrows, S10A Fig), we predicted that vascularization would fail in Lats1/2 PanKO . Indeed, endothelial cells positive for platelet and endothelial cell adhesion molecule 1 (PECAM1) and endomucin (PE + ) failed to associate at all with Lats1/2 PanKO CDH1 + cells at E13.5 (arrowheads, S10A Fig).
Because there was a dearth of blood vessel intercalation, we tested whether mutant cells at the center of the pancreas were experiencing hypoxia. Indeed, by E15.5 we observed increased hypoxia inducible factor 1 subunit alpha (HIF1A) immunopositivity in the Lats1/2 PanKO relative to WT (S10B Fig), confirming increased hypoxia. Together, these cell death, absent vascularization, and hypoxic phenotypes occurring at E12.5 and later likely explain the failed bud expansion following Lats1/2 deficiency.
https://doi.org/10.1371/journal.pbio.3000382.g005 LATS1/2 constrain NFκB to permit pancreatic progenitor differentiation NFκB activity in pancreas cells. Indeed, H 2 O 2 provoked RELA phosphorylation and nuclear translocation in WT pancreas explants, thereby mimicking the loss of Lats1/2 (arrowheads, Fig  5G, 5H, and 5I). These findings show that an oxidative state activates NFκB, even in WT pancreas. Overall, these data reinforce the model that LATS1/2 kinases are central factors in suppressing VNN1 activity, which in turn increases both oxidative stress and NFκB activity ( Fig  5J). Together, these data validate the hypothesis that LATS1/2 kinases normally restrict cellintrinsic and deleterious excessive NFκB signaling in the developing pancreas.

NFκB is required to initiate pancreatic EMT downstream of Lats1/2 deletion
Our findings suggest that NFκB hyperactivity is a major mechanism driving epithelial defects in Lats1/2 PanKO pancreas. Therefore, we next asked whether NFκB activity is required or sufficient to induce the phenotypic abnormalities of Lats1/2 PanKO . To test for this requirement, we inhibited NFκB in Lats1/2 PanKO and examined the effects on EMT initiation. Indeed, treatment of the Lats1/2 PanKO epithelium with the NFκB inhibitor anatabine suppressed ectopic VIM expression, suggesting that NFκB is required for EMT initiation downstream of Lats1/2 deletion (Fig 6A, 6B, 6C, 6D and 6E). (N.B. anatabine treatment also lessened endogenous mesenchymal VIM expression; asterisks, Fig 6C and 6D). We measured the expression of transcriptional targets Nfkbia and Mmp9 following anatabine treatment of WT pancreatic explants and found significant down-regulation of both NFκB target genes Nfkbia and Mmp9 (S10D Fig), thereby confirming the efficacy of NFκB inhibition ex vivo. These data show that Lats1/2 depletion leads to EMT initiation (as measured by ectopic VIM expression) via activated NFκB.
We asked whether NFκB inhibition is sufficient to rescue the Lats1/2 mutant phenotype. We measured endocrine cell differentiation in pancreatic tissue, which was explanted at E11.5 and treated with anatabine for 2 days. As expected, INS+/GCG+ endocrine cells were decreased in vehicle-treated Lats1/2 PanKO explants compared to WT (S10E Fig). Surprisingly, NFκB inhibition did not affect endocrine differentiation in Lats1/2-deficient endocrine cells (S10E and S10F Fig), indicating that blocking NFκB signaling does not rescue the differentiation phenotype in Lats1/2 PanKO pancreas.
Lastly, to test for sufficiency of NFκB activity in initiating EMT, we treated WT E10.5 pancreas explants with the NFκB activator phorbol 12-myristate 13-acetate (PMA) [59,60]. In WT controls, VIM was never expressed in epithelial cells but was readily observed in the mesenchyme surrounding the epithelium in vehicle-treated explants (Fig 6F and 6F'). By contrast, exogenous NFκB activation by PMA treatment triggered robust VIM expression in the WT CDH1 + pancreas epithelium (Fig 6G), with a striking 39-fold increase in colocalization area (white, Fig 6G' and 6H). Overall, VIM expression within the progenitor epithelium was significantly higher following exogenous NFκB activation (Fig 6I and 6J), visualized by slice view through vehicle- (Fig 6K) and PMA-treated explants (arrowheads, Fig 6L). NFκB activation also increased VIM in mesenchymal cells surrounding epithelial branches, where VIM was normally expressed albeit at lower levels (S10G Fig). Exogenous NFκB activation similarly triggered expression of another mesenchymal protein, TAGLN, normally absent from WT pancreas epithelium (S10H Fig). Together, these data confirm that hyperactivated NFκB is sufficient to initiate abnormal EMT in pancreatic progenitors (Fig 6M).
Collectively, these findings show that LATS1/2 provide an endogenous protective mechanism to suppress NFκB signaling and to maintain epithelial integrity by suppressing EMT pathways. Triggering NFκB or removing protective LATS1/2 suppression of NFκB activity leads to deleterious progression toward EMT and hinders developmental epithelial morphogenesis. This work underscores the role of the Hippo pathway acting as a rheostat to control cell-intrinsic NFκB signaling in pancreatic epithelium.

Discussion
In this study, we show a critical requirement for active restriction of YAP1/TAZ and NFκB pathways in pancreas to allow lineage differentiation and morphogenesis (Fig 7). We find total failure of endocrine, acinar, and ductal cell lineage development upon genetic deletion of Lats1 and Lats2. In other organs like kidney, liver, and heart, YAP1/TAZ activation leads to expansion and/or maintenance of organ progenitors, often resulting in increased organ size [1,[61][62][63]. By contrast, we find that high YAP1/TAZ activity in the early Lats1/2-depleted pancreas initially expands the pancreas progenitor pool but quickly blocks pancreas morphogenesis by arresting branching and lineage differentiation. We show that high YAP1/TAZ levels stimulate NFκB activators like Vnn1, leading to loss of epithelial integrity and abnormal transition to a partially mesenchymal state. In humans, elevated VNN1 is associated with pathological conditions, including colitis and systemic sclerosis [64][65][66]; in mice, it stimulates the inflammatory response in experimental models of colitis. Consistent with these models, aberrant cell-autonomous VNN1 expression creates a strikingly pro-oxidant environment in Lats1/2 PanKO , which we find triggers cell-autonomous NFκB hyperactivation and EMT initiation, as evidenced by the inappropriate acquisition of mesenchymal cell characteristics and loss of epithelial integrity. We propose that WT pancreatic cells are normally poised to respond to ROS and NFκB cues by initiating EMT, a crucial first step of cancer progression [67]. Our findings identify a strict requirement for LATS1/2 kinases to control of NFκB and thereby coordinate morphogenesis and permit pancreatic lineage differentiation.
A recent study by Semb and colleagues has shown that YAP1/TEAD in bipotent pancreatic progenitors promotes pro-ductal Notch signaling, which in turn inhibits Neurog3 expression [28]. They further show that YAP1 promotes Notch1 and Hes1 and that a YAP1/TEAD/Hes family bHLH TF 1 (HES1) complex forms in pancreatic progenitor epithelium, suggesting a dual function for YAP1 in promoting ductal and restricting endocrine fate. This work builds on the established role of Notch1 in inhibiting endocrinogenesis [68][69][70][71][72][73][74][75] and advances our understanding of cell fate in the developing pancreas. While we observed increased Notch1 ligand Jagged1 (Jag1) expression in Lats1/2 PanKO at E13.5 (Fig 1J), indicating Notch activation downstream of YAP1/TAZ, we find that the Lats1/2 mutant cells lose pancreatic identity before or during the secondary transition. We believe that the differences in phenotype reported here and by Mamidi and colleagues result from the mouse model we use whereby the Lats1/2-deficient pancreas has constitutive activation of both YAP1 and TAZ, as shown in . (E) The percent area covered by VIM + pixels within the pancreatic epithelium (above a set threshold) was quantified in anatabine-and vehicle-treated Lats1/2 PanKO and WT explants. Data are shown as mean ± SEM. Statistical significance was determined by Student t test ( � p < 0.05; �� p < 0.01; ��� p < 0.001; ���� p < 0.0001). (F, G) Compressed Z stack confocal images of CDH1 and VIM immunostains of WT pancreas explants treated with 10 μg/mL PMA (n = 3 explants per treatment). Masked VIM (VIM within CDH1 + progenitor surface volume) was used to exclude mesenchymal VIM. Scale = 50 μm. (F',G') Colocalization (white) was calculated above common thresholds and visualized using Imaris. (H) The percent area of VIM +CDH1 colocalization pixels was quantified using ImageJ. (I, J) Mean VIM + pixel intensity and the percent area of VIM + pixels (above a common threshold) within WT pancreatic epithelium were quantified using ImageJ. Underlying numerical values can be found in S1 Data. LATS1/2 constrain NFκB to permit pancreatic progenitor differentiation S1H and S1I Fig. By contrast, Mamidi and others utilized a constitutively activated YAP1 transgenic mouse model [28], which retains endogenous TAZ. Our activated YAP1 pancreas model [2] also does not manipulate endogenous TAZ expression, which may help to explain the low proportion of mesenchymal TAGLN + cells in Yap1 GFP pancreas. Similarly, we observed no increase in the proportion of pNFκB1 + cells in the Yap1 GFP pancreas, relative to WT. Therefore, we conclude that activated YAP1 is sufficient to repress endocrine progenitor cell fate alone but that dual hyperactivation of both YAP1 and TAZ suppresses all 3 pancreatic lineages (acinar, ductal, and endocrine) and induces aberrant EMT in the developing pancreas as a result of inappropriate NFκB activation.
In addition, work from the Melton group demonstrated the role of YAP1 in guiding the directed in vitro differentiation of human pluripotent stem cells towards a beta cell fate [29]. During intermediate steps of their beta-cell-directed differentiation protocol, YAP1 is required to maintain pancreatic progenitors. Conversely, to achieve mature beta cell differentiation, YAP1 must later be inhibited in vitro. These findings agree with our model that YAP1 activity must be kept at bay for normal lineage differentiation and morphogenesis.
It is notable that the Lats1/2 PanKO differs dramatically from the Mst1/2 PanKO . Multiple groups reported pancreas defects following deletion of Hippo kinases Mst1/2 (Mst1/2 PanKO ) [2,3]. Deleting Mst1/2 from the Pdx1Cre lineage led to smaller pancreata that retained all 3 pancreatic lineages, albeit with defective islet, ductal, and acinar architecture. Specifically, the Mst1/2-deficient pancreatic acinar cells lost their identity, with abnormally up-regulated Sox9 and Hes1, both ductal-specific markers. These data align with YAP1 restriction of endocrine identity reported by others [3,8,28] and confirmed in this report, showing that endocrine progenitors (NEUROG3 and NKX6.1 + cells in the early pancreas) were the first lineage lost in Lats1/2-deficient pancreata. It is notable that the Lats1/2 deletion provoked a more severe phenotype compared to that of Mst1/2. This phenotypic difference is likely due to additional, noncanonical input upstream of LATS1/2, such as CDK1 or Aurora A [76][77][78][79]. While the Mst1/2deficient pancreas have obvious defects in differentiation upon increased nuclear YAP1 [2,3], we propose that residual LATS1/2 in these mutants, phosphorylated by Kibra or other activators, could indirectly quell potential activation of NFκB and other mechanisms driving EMT in this model. Alternatively, because LATS1/2 kinases are known to phosphorylate other targets in addition to YAP1 and TAZ [77], it is possible that proper regulation of these noncanonical targets could yield the relatively mild phenotype in developing Mst1/2 PanKO pancreata. Similarly, some of the Lats1/2 PanKO defects are potentially not YAP1/TAZ dependent. Future studies are needed to determine the extent of LATS1/2 activation in Mst1/2 PanKO pancreata and untangle potential roles for noncanonical LATS1/2 targets in WT pancreatic progenitors.
Because deletion of Lats1/2 led to dramatic changes in cell shape and to loss of epithelial integrity, in this study we interrogated EMT mechanisms. Classical EMT was originally described as a unidirectional transition in which cells delaminate from a classical epithelium, and CDH1 expression is replaced by that of VIM [80]. Focused interest on the role of EMT contribution during metastasis and tumorigenesis has led to a growing consensus that EMT is not a binary process; instead, cells undergoing EMT exhibit a continuum of epithelial, mesenchymal, or hybrid epithelial-mesenchymal characteristics [81]. Tumor cells display a similar range of intermediate epithelial-mesenchymal features during tumor budding, the first stage of metastasis [82]. Therefore, elucidating the mechanisms that regulate and prevent EMT could yield insights for therapeutically abrogating or even blocking tumorigenesis. To that end, data presented here delineate a novel mechanism that influences on EMT. In embryonic Lats1/2 PanKO pancreatic epithelium, we observe inappropriate acquisition of this mesenchymal program. Surprisingly, persistent epithelial CDH1 expression coincided with expression of VIM, TAGLN, and ACTA2. Furthermore, we found an absence of some classical EMT TFs, with neither SNAI2 nor ZEB1 detectable by IF (although Twist2 was up-regulated). These observations suggest that EMT is initiated but is not complete upon loss of Lats1/2, and its initiation is not dependent on classical EMT mechanisms like suppression of CDH1 and promotion of SNAI2 [67]. Therefore, the Lats1/2 PanKO pancreas provides a novel developmental model for study of EMT initiation and hybrid (partial) epithelial/mesenchymal cell fate [83].
Further implicated in aberrant EMT in Lats1/2 PanKO pancreas progenitors is a group of highly up-regulated factors: the NFκB activators Vnn1 and Lurap1l, as well as NFκB downstream targets. Together, their expression suggests an inappropriate boost of NFκB activity. Changes in Vnn1 expression downstream of YAP1 have been reported in studies making use of microarray and ChIP-seq screens [2,84]; however, the influence of VNN1 products and downstream effectors on pancreatic differentiation has not been reported. Intriguingly, motif analysis of ChIP-seq data shows a TEAD binding sites within 500 bp surrounding the Vnn1 TSS, suggesting a potential YAP1/TAZ/TEAD TF complex driving Vnn1 expression. Excess VNN1 promotes inflammation and worsens prognoses in human diseases like colitis and systemic sclerosis [64][65][66]. In mice, the VNN1 inhibitor RR6 has a prohibitively short half-life (approximately 80 minutes) [85,86], making it too short-lived to assess in Lats1/2 PanKO pancreas explants. In future studies, deleting Vnn1 from Lats1/2 PanKO pancreata could determine the extent to which VNN1 drives the Lats1/2 PanKO induction of EMT program.
Perhaps the most striking finding in this study is that aberrant EMT initiation is triggered by inappropriately activated YAP1/TAZ and NFκB. Our experiments show that Lats1/2 deletion-induced EMT initiation can be countered with the NFκB inhibitor anatabine, which we show suppresses VIM expression in Lats1/2 PanKO cells. This finding aligns with the presence of an NFκB Response Element within the VIM promoter [87]. It also raises many questions regarding crosstalk between Hippo and NFκB signaling, which remains largely uncharacterized. Recent work showed that loss of a single allele of Nr5a2 leads to a pre-inflammatory state with cell-autonomous up-regulation of cytokines and other stress response genes [88], revealing a relationship between regulators of pancreatic development and NFκB-dependent inflammatory pathways. This study points to the importance of homeostatic transcriptional suppression of inflammatory pathways, thereby supporting the mechanism identified in the present study.
We report that NFκB is phosphorylated and activated in mitotic pancreatic progenitor cells, consistent with reports that NFκB controls cell proliferation during branching morphogenesis of other organs, e.g., the embryonic mammary gland [89]. A previous study reported that embryonic inhibition of NFκB activity in the Pdx1 lineage yielded viable adult mice [90]. Although Norlin and colleagues did not report on cell proliferation per se in pancreata lacking NFκB activity, this report suggests that NFκB activation may not be required for cell proliferation (at least in the Pdx1 lineage) in the embryonic pancreas. Distinct, parallel pathways promoting EMT or proliferation may exist downstream of Lats1/2 deletion in the pancreas. The fact that NFκB is not required to control pancreatic cell proliferation reinforces our model that it is instead the restraint of NFκB activity-the suppression of NFκB signaling-which is critical to regulate cellular proliferation during pancreas formation. Importantly, recent analysis of murine lung bacterial pneumonia has revealed the importance of YAP1/TAZ in controlling the dynamic NFκB injury response, required to allow alveolar regeneration [91]. This study underscores the importance of the YAP1/TAZ-NFκB signaling axis in endodermal organ development and disease.
In conclusion, our studies demonstrate how suppressing intrinsic deleterious genetic programs is a critical aspect of normal organogenesis. Inappropriately activated YAP1/TAZ or elevated NFκB derail cellular identity in the pancreas by triggering mesenchymal gene expression and EMT initiation. Without Lats1/2, VNN1 is up-regulated, which in turn activates NFκB and EMT by disrupting redox homeostasis. Our data suggest that NFκB-induced partial EMT is a primary driver of Lats1/2 PanKO defects. We report the novel finding that Hippo signaling in the pancreas ordinarily suppresses these events, thereby allowing normal morphogenesis and consequent acinar, endocrine, and ductal cell differentiation. Our study underscores the exquisite and essential control that LATS1/2 kinases exert to suppress NFκB cell intrinsically in normal pancreatic progenitors to maintain cell identity. Intrinsic redox and NFκB mechanisms can be ectopically activated in WT pancreatic cells, highlighting the clinical significance of elucidating both the protective function of LATS1/2 kinases and the YAP1/TAZ-NFκB signaling axis.

Ethics statement
All animal experiments were performed in accordance with the Guide for the Care and Use of Laboratory Animals and the Animal Welfare Act, as well as protocols approved by the University of Texas Southwestern Medical Center Institutional Animal Care and Use Committee (IACUC; approval number 2017-102243, approval date 26 September 2017). All animals were observed daily, and appropriate care was provided by the veterinary staff of the UTSW Animal Resource Center, which is fully accredited by the Association for Assessment and Accreditation of Laboratory Care, International (Unit Number 000673) and by the NIH Office of Laboratory Animal Welfare (Assurance Number D16-00296). Dams were euthanized via IACUC-approved humane methods, using carbon dioxide asphyxiation and secondary cervical dislocation.

Mice
E10.0-E18.5 embryos and postnatal mice (Mus musculus) were dissected in phosphate-buffered saline (PBS) and fixed in 4% paraformaldehyde (PFA)/PBS at 4˚C for 3 hours for section and whole-mount IF or overnight (O/N) for RNA ISH. Tissue was then washed, dehydrated to 70% ethanol (EtOH), and stored at −20˚C.

Section IF
For paraffin sectioning, tissues were embedded as described previously with some modifications [30]. Briefly, tissues were dehydrated to 100% EtOH (2 × 30 minutes), washed in xylene (2 × 10 minutes), and then incubated in 1:1 xylene:paraffin for 30 minutes at 65˚C. After incubating for at least 4 hours at 65˚C with paraffin replacement approximately every hour, tissues were embedded in paraffin and sectioned at 10 μm.
Paraffin sections were baked at 60˚C for 10 minutes and de-paraffinized in xylene and rehydrated through EtOH to PBS. All washes were performed using PBS. After washing, tissue sections were permeabilized using 0.3% Triton-X/PBS for 10 minutes. Antigen retrieval was performed under pressure using acidic buffer A (Electron Microscopy Sciences, Hatfield, PA). For standard immunostaining, sections were washed, blocked in CAS block (Invitrogen, Carlsbad, CA), and incubated in primary antibodies (diluted in CAS Block, S2 Table) using hybridization chambers O/N at 4˚C. The next day, sections were washed and incubated in Alexa Fluor series antibodies (Invitrogen, Carlsbad, CA; 1:200) for 2 hours at room temperature (RT). Sections were then washed, incubated in 4',6-diamidino-2-phenylindole dihydrochloride (DAPI) nuclear stain (Sigma Aldrich, St. Louis, MO) (1:10,000), rinsed, and mounted in Prolong Gold Anti-fade reagent with DAPI (Invitrogen, Carlsbad, CA). Images were obtained using Nikon A1R confocal or LSM710 Meta Zeiss confocal microscopes. Unless stated otherwise, representative images from 3 experiments are shown.
Tyramide signal amplification immunostaining was performed per manufacturer's protocol (Life Technologies, Carlsbad, CA). Briefly, after antigen retrieval, sections were washed, blocked using tyramide blocking reagent for at least 1 hour at RT, and incubated in primary antibodies (diluted in tyramide block, S2 Table)

Whole-mount IF
Whole-mount immunostaining was performed as described previously [31]. Briefly, after fixation, tissues were washed, dehydrated to 100% methanol (MeOH), and stored at −20˚C. Prior to immunostaining, tissues were bleached in 4:1:1 MeOH:DMSO:H 2 O 2 solution for 2 hours at RT, placed in fresh MeOH for 1 hour, and then rehydrated through MeOH series to PBS. Block solution was prepared using 0.5% blocking reagent powder (Perkin Elmer, Waltham, MA) in 1M Tris (pH 7.5) and heated at 65˚C to dissolve. Tissues were blocked for 2 hours at RT and then incubated in primary antibodies (S2 Table) O/N at 4˚C. The next day, tissues were washed for 5-8 hours, refreshing PBS approximately every hour, and then incubated in secondary (Alexa Fluor) antibodies (diluted 1:200 in Blocking Reagent) O/N at 4˚C. Tissues were then washed in PBS 3 times for 20 minutes, dehydrated to 100% MeOH, equilibrated in a 1:2 benzyl alcohol:benzyl benzoate (BABB) solution, and mounted in BABB using 1.6 mm concavity slides (Electron Microscopy Sciences, Hatfield, PA). Images of whole-mount immunostaining were obtained using a LSM710 Meta Zeiss confocal microscope.

Total RNA-seq
RNA was isolated from pancreas buds at E11.0 (39-40 somites). Embryos were dissected on ice in cold PBS. Mesenchyme was manually removed using forceps, and pancreas epithelia were homogenized using a 20-gauge needle in RLT buffer (Qiagen, Hilden, Germany) containing 1% β-mercaptoethanol (Fisher Scientific, Hampton, NH). The amount of 35 ng total RNA was isolated from each pancreas bud using RNeasy Microkit (Qiagen, Hilden, Germany) following manufacturer instructions. RNA sample quantity and quality were measured using Bioanalyzer with RNA 6000 Pico Kit (Agilent Technologies, Santa Clara, CA), and samples with RIN < 8.8 were excluded. cDNA libraries were constructed by using KAPA Stranded RNA-seq Kit with RiboErase (HMR) (KR1151-v3.15, Kapa Biosystems, Pleasanton, CA). Amplified cDNA was validated and quantified on an Agilent Bioanalyzer with the High Sensitivity DNA chip. The purified libraries were normalized, pooled together, denatured, and diluted at final concentration of 1.8 pM. Furthermore, 1.3 mL of diluted pool was used to perform cluster generation, followed by 2 × 75 bp sequencing on NextSeq500 (Illumina, San Diego, CA). From each sample, we obtained about 31 million reads.

Gene set enrichment analysis
Gene set enrichment analyses using multiple algorithms were performed using the EGSEA [97] package. Specifically, the EGSEA analysis included the camera, safe, gage, padog, plage, zscore, gsva, ssgsea, globaltest, ora, and fry algorithms. Gene sets from the h, c2, and c5 collections of the Molecular Signatures Database [98] (MSigDB) and the KEGG [99] pathways included in the EGSEAdata package were indexed. p-Values were combined using Wilkinson's method followed by application of the Benjamini-Hochberg [100] algorithm to control the false discovery rate (FDR). Results were ordered hierarchically by vote rank, with a bin width of 5, followed by FDR.

Motif analysis
A motif library was generated by concatenating the Tead1, Tead2, Tead3, and Tead4 motif files from the HOMER [101] motif database (http://homer.ucsd.edu/homer/). The genomic region extending from 400 bp upstream to 100 bp downstream of the Vnn1 TSS was scanned using findMotifs.pl from the HOMER tool suite. Sequence logos were generated from position weight matrices using the ggseqlogo [102] Bioconductor R package (http://bioconductor.org). Annotation information, including motif locations, was plotted along genomic coordinates with the Gviz [103] package.
ISH was performed as previously described [110]. Briefly, fixed E11.0 embryos were dehydrated, embedded in paraffin, and sectioned as described above. Paraffin sections (10 μm thick) were de-paraffinized in xylene, rehydrated, treated with 15 μg/mL proteinase K (Sigma Aldrich, St. Louis, MO) for 15 minutes at RT, washed in PBS, post-fixed in 4% PFA/PBS for 15 minutes at RT, washed, and incubated in pre-hybridization solution for 30 to 45 minutes at RT. Probe hybridization (1 μg/mL) was performed O/N at 65˚C. The next day, slides were washed in 0.2X SSC and then MBST, followed by blocking in 2% blocking reagent (Roche, Basel, Switzerland) for 2 hours at RT. Sections were incubated in Anti-Digoxigenin (1:4,000, Roche, Basel, Switzerland) antibody O/N at 4˚C. The next day, slides were washed in MBST followed by NTMT. Substrate color reaction using BM purple (Roche, Basel, Switzerland) was performed at 37˚C for 4-24 hours. Slides were then post-fixed in 4% PFA/PBS for approximately 1 hour, washed, dehydrated, and mounted using Permount (Fisher Scientific, Hampton, NH). Images were obtained using Zeiss Axiovert 200 M inverted microscope and DP-70 camera (Olympus, Shinjuku, Japan).

Pancreas explant cultures
Pancreata were explanted as described previously [30]. Briefly, WT and Lats1/2 PanKO pancreata were explanted at E10.5-E11.5 onto fibronectin-coated plastic dishes in complete DMEM (American Type Culture Collection [ATCC], Manassas, VA), containing 1% penicillin/streptomycin, 10% fetal bovine serum (ATCC, Manassas, VA), and 10 μg/ml gentamicin (Thermo-Fisher Scientific, Waltham, MA). The next day, explants were treated in complete medium with anatabine [112] (Cayman Chemicals, Ann Arbor, MI), cysteamine [52], PMA [59,60], NAC [113] (Sigma Aldrich, St. Louis, MO), or H 2 O 2 [114] (Fisher Scientific, Hampton, NH). Experimental cysteamine treatment at 4 mM was utilized based on a previous dose response study, which showed that 7 mM cysteamine induced cell death and that 2 mM cysteamine failed to induce ROS [52]. See S5 Table for chemical information. EtOH (vehicle for anatabine), dimethyl sulfoxide (DMSO, vehicle for SN50 [115] and PMA), and PBS (vehicle for cysteamine) were used as controls. Media were replenished every 1 to 2 days. Explants were fixed after 2 to 4 days. Explant immunostaining was carried out as previously described [30]. Briefly, explants were fixed in 4% PFA/PBS at RT for 15 minutes, washed in PBS, permeabilized in 0.3% Triton-X/PBS for 1 hour, washed in PBS, and blocked for 1 hour in CAS Block (Invitrogen, Carlsbad, CA). Primary antibodies (S2 Table) were incubated O/N at 4˚C. For the second day, explants were washed in PBS and then incubated in secondary antibodies O/N at 4˚C. Explants were washed the next day and mounted flat on a glass slide with Prolong Gold Anti-Fade (Invitrogen, Carlsbad, CA). Images were obtained using Nikon A1R confocal or LSM710 Meta Zeiss confocal microscopes. Representative compressed Z stack images from 3 independent experiments are shown.

CellROX
CellROX Green Reagent (ThermoFisher Scientific, Waltham, MA) oxidative stress assay was performed as described [113], according to the manufacturer's instructions. Briefly, CellROX Green Reagent was added to explant media at a final concentration of 5 μM and incubated at 37˚C for 30 minutes. (This and all subsequent steps were performed in the dark.) Pancreas explants were then washed, fixed in 4% PFA/PBS at RT for 15 minutes, washed, permeabilized in 0.3% Triton-X/PBS for 1 hour at RT, washed, blocked in CAS block (Invitrogen, Carlsbad, CA) for 1 hour at RT, incubated in primary antibodies for 2 hours at RT, washed, incubated in secondary antibodies O/N at 4˚C, washed, and mounted as described above. Explants were imaged using Nikon A1R confocal microscope within 24 hours of CellROX incorporation assay.
The amount of 30 μL of lysis buffer was prepared per sample (lysis buffer: 10 mL PBS + 10 μg/ml aprotinin [final] + 10 μg/ml leupeptin [final] + 10 μg/ml pepstatin [final] + 1 phos-STOP tablet from Roche, Basel, Switzerland). All dissection steps were performed in PBS in a Sylgard dish placed in a larger Petri dish with ice, under the dissecting microscope. Stomachs along with the pancreas were dissected and transferred to a clean dish on ice. Once all stomachs were dissected, when available, reporter expression was used to determine the genotype (Lats1/2 PanKO versus all others as control). Otherwise, stomachs were kept on ice until genotyping was complete. When fewer than 4 stomachs were obtained per genotype, pancreata were dissected and snap-frozen for later use.
Pancreata were dissected away from the stomach, and remaining stomach or spleen pieces were removed (note that the pancreatic mesenchyme was not removed). Pancreata were transferred to a clean dish on ice using a glass pipette. Once all pancreata were dissected, samegenotype pancreata were transferred into a small Sylgard dish along with a drop of PBS using a glass pipette; 20 μL of lysis buffer was placed next to the PBS, and forceps were used to quickly transfer the pancreata from PBS into the lysis buffer. Each pancreas was excised into 4-5 pieces as fast as possible, and the pieces were transferred using a pipette (2-20 μL) into a microcentrifuge tube with 10 μL lysis buffer on ice (total volume = 30 μL). Once all genotypes were complete, homogenization was carried out by pipetting up and down the contents of each tube, until no chunks remained. Then, 3 μL of 10% Triton-X (prediluted in PBS) was added into each tube, and the contents were mixed by pipetting. Samples were then frozen at −80˚C to pellet the cellular debris for 20 minutes (up to O/N). The tubes were centrifuged at 10,000g for 5 minutes, and the supernatant was transferred into a fresh tube on ice.