ROCK-nmMyoII, Notch and Neurog3 gene-dosage link epithelial morphogenesis with cell fate in the pancreatic endocrine-progenitor niche

In mouse pancreatogenesis, tissue growth, morphogenesis and differentiation must be coordinated, although how remains largely unresolved. We have reported that bipotent duct and endocrine progenitors exist within an epithelial niche called the ‘plexus state’ (Bankaitis et al., 2015), which comprises a web-like epithelium arising around E12.5 within an anlage of multipotent progenitors (MPC) (Kesavan et al., 2009; Hick et al., 2009; Villasenor et al., 2010). From E12.5-18.5, the plexus undergoes rapid expansion and remodeling, very different from classical branching morphogenesis, concurrent with a prolonged wave of endocrine-cell birth (Pan and Wright, 2011). The progressive replacement of plexus by an arbor-like ductal epithelium is associated with a great reduction in the production of new endocrine cells (Kopp et al., 2011; Bankaitis et al., 2015). Here, we investigate the coordination of plexus morphogenesis and endocrine cell-fate allocation.

Pancreatic endocrine cells derive from a Notch-responsive Sox9⁺ progenitor pool (Apelqvist et al., 1999; Kopp et al., 2011; Serup, 2012) via activation of the endocrine-lineage determinant neurogenin 3 (Neurog3; Gradwohl et al., 2000; Schwitzgebel et al., 2000; Gu et al., 2002). Neurog3 is necessary and sufficient for endocrine-cell birth (Gradwohl et al., 2000; Johansson et al., 2007). Cells with high-level Neurog3 (Neurog3^HI) exit the cell cycle (Miyatsuka et al., 2011), egress from the epithelium (Gouzhi et al., 2011) and cluster adjacent to the forming nascent islets of Langerhans. Lineage tracing in Neurog3-null and hypomorphic mice (Wang et al., 2010; Shih et al., 2012) showed that preventing acquisition of the Neurog3^HI state causes cells to remain intra-epithelial and to differentiate into mature acinar and ductal fates. The mechanisms regulating progression of epithelial progenitors through various Neurog3 expression states are poorly defined.

Non-autonomous feedback from Neurog3-expressing cells regulates numerous processes in forming duct versus endocrine tissues. Inactivating Neurog3 greatly expands low-level Neurog3 (Neurog3^LO) transcription throughout the Sox9⁺ epithelium and causes the loss of Neurog3^HI cells (Wang et al., 2010; Bechard et al., 2016). Neurog3 loss causes a widespread reduction in epithelial Notch-pathway activity (Magenheim et al., 2011; Shih et al., 2012; Qu et al., 2013). Thus, Neurog3^HI cells normally induce lateral inhibition of Neurog3 expression in surrounding Sox9⁺ progenitors to balance endocrine differentiation with progenitor maintenance (Afelik et al. 2012; Qu et al., 2013; Apelqvist et al., 1999; Murtaugh et al., 2003). Neurog3 deficiency causes a persistent reduction in Sox9⁺ cell replication, suggesting additional functions in supporting replicative growth of progenitor epithelium (Bankaitis et al., 2015). Neurog3 loss results in a dysmorphic plexus that is precociously transformed into more-mature epithelial duct and branched states, again suggesting that egressing endocrine cells modulate normal morphogenesis by maintaining their parental ‘plexus niche’ (Magenheim et al., 2011; Bankaitis et al., 2015). Collectively, these studies position Neurog3-expressing cells as a coordination nexus for progenitor maintenance, cell-fate allocation and plexus morphogenesis, to regulate growth and assembly of the ductal network with formation of a dispersed endocrine pancreas (Rieck et al., 2012; Bankaitis et al., 2015).

In epithelia, the molecular motor nmMyoII acts on filamentous-actin (F-actin) substrates in apical constriction, tissue folding, cell delamination and migration, and generation of inter- and intra-cellular tension (Gorfinkiel and Blanchard, 2011; Heisenberg and Bellaïche, 2013; Martin and Goldstein, 2014). Rho-associated kinase (ROCK) positively regulates nmMyoII activity (Wang et al., 2009), but also functions as a modulator of actin remodeling (Watanabe et al., 1999; Maekawa et al., 1999; Reinto and Ridley, 2003), affecting tissue and cell-shape change, cell polarization and migration (Li et al., 2005; Nakayama et al., 2005, 2008). ROCK-nmMyoII activities influence many complex morphogenetic and intracellular signaling processes (Amano et al., 2010). Rho-GTPase signaling, an upstream ROCK/nmMyoII regulator, influences neural-crest specification in frog embryos (Kim et al., 2015), and specific cell-fate decisions in mesenchymal cells (Sordella et al., 2003; McBeath et al., 2004). These reports clearly point to integration between morphogenesis signaling and the intrinsic gene-regulatory networks that drive cell fate.

We report here on the mechanistic interconnections between epithelial morphogenesis and fate in the pancreatic plexus niche. We use three-dimensional tissue reconstructions, gene-reporter analyses and genetic and pharmacological interventions to study interdependencies between Neurog3 gene activity, Notch signaling and ROCK nmMyoII-controlled epithelial-cell morphogenesis. We propose that sequential, dissociable steps in endocrine fate allocation are mediated by morphogenetic changes at an apical versus basal cell surface. Neurog3^LO expression initiates concurrently with apical surface contraction, and becomes upregulated as its apical surface becomes increasingly focalized and the cell begins basalward movement. Initial contraction, focalization and Neurog3 upregulation occur in the absence of Neurog3 protein, suggesting that endocrine specification and entry to commitment occur via epithelium-intrinsic inputs upstream or independent of Neurog3. nNmMyoII and ROCK oppositely regulate apical narrowing, focalization and basalward cell movement, and thus acquisition of the Neurog3^HI state. We propose a model in which a circuit comprising ROCK-nmMyoII, Neurog3 gene dosage and Notch signaling balances apportionment of endocrine cells from the plexus while enabling proper growth and morphogenesis of the pancreatic epithelium.

A prominent feature in polarized epithelial cells is a belt of filamentous actin (F-actin) circumscribing the sub-apical cell cortex (Martin and Goldstein, 2014). These belt-like structures (hereafter F-actin^BELT) are closely apposed to tight and adherens junctions, and are important in mediating remodeling processes such as apical constriction, tissue folding, cell intercalation and epithelial egress or extrusion, among others (Heisenberg and Bellaïche, 2013). To probe whether specific cell-shape changes are associated with duct versus endocrine cell-fate decisions, we compared the F-actin^BELT topologies in cells located within the endocrine progenitor-rich plexus, the endocrine progenitor-poor duct state and in cells expressing Neurog3 using an EGFP knock-in null allele (Lee et al., 2001). For the plexus, confocal z-stack reconstructions showed an intense multicellular meshwork of cortical F-actin closely associated with the Muc1⁺ apical lumen surface (Fig. 1A). The F-actin^BELT was located nearby the apical ZO1⁺ tight junctions, near the apical-most extent of Ecad⁺ adherens junctions, and apical with respect to the nuclei of all epithelial cells (Fig. S1A-H′ and data not shown), as expected for F-actin^BELT in most epithelia. Plotting the F-actin^BELT aspect ratio against F-actin^BELT perimeter length from many individual cells served as a metric of apical surface shape (Fig. S2A-B) and revealed a spectrum of apical surface shapes in Neurog3^+/+ and heterozygous-null Neurog3^EGFP/+ plexus (Fig. 1D). This spectrum in the plexus was similar throughout mid-to-late gestation, indicating that the overall mix of topologies is maintained during the major wave of cellular differentiation (Fig. 1E). Also in the plexus, numerous intense F-actin puncta were located directly at the Muc1⁺aPKC⁺ lumenal surface, in a highly focalized pattern (hereafter, F-actin^FOCAL), signifying tightly narrowed apical contacts or tethering points (Bechard et al., 2016) of egressing Neurog3-expressing cells (Fig. 1B,C; Fig. S3A-D). Essentially all Muc1⁺ cells expressing Neurog3^EGFP exhibited a narrowed F-actin^BELT or F-actin^FOCAL signal (Fig. 1F). In contrast, the F-actin^BELT of essentially all cells in the duct state exhibited a larger perimeter and decreased aspect ratio (Fig. 1G-J). Scoring E-cad⁺ cell boundaries near the apical surface of these cells confirmed a large apical domain (Fig. S4A-D) and cell-flattening phenotype consistent with ductal differentiation (Grapin-Botton, 2005). In duct-state epithelium, Neurog3-reporting cells and F-actin^FOCAL structures were reduced or absent compared with plexus, consistent with a large reduction in Neurog3⁺ cell numbers (Bankaitis et al., 2015). EGFP-expressing cells without Muc1⁺ apical lumen-contact were considered to have exited the epithelium as endocrine-committed cells, and were not investigated further. These analyses suggest that F-actin^BELT narrowing and F-actin^FOCAL formation are associated with activation of Neurog3 expression, while the larger F-actin^BELT shapes are associated with ductal or non-endocrine cell fates.

In E14.5 Neurog3^NULL (Neurog3^EGFP/EGFP) epithelium, Neurog3^HI (EGFP^HI) cells are lost, Neurog3^LO (EGFP^LO) cell numbers are increased, and the epithelial plexus is dysmorphic (Wang et al., 2010; Magenheim et al., 2011; Bankaitis et al., 2015; Bechard et al., 2016). Accordingly, we observed a dramatic loss of Muc1⁺F-actin^FOCAL structures, and a shift in apical F-actin^BELT shapes toward a more uniform circular morphology relative to Neurog3^+/+ and Neurog3^EGFP/+ samples (Fig. 2A-D,I). In addition to the previously reported ‘epithelial fattening’ observed in the Neurog3^NULL state, we also detected increased numbers of Sox9⁺ nuclei per unit of Muc1⁺ epithelium (Fig. S5A,B), consistent with a relative cell-crowding effect caused by the lack of egress of endocrine-committed cells. Although not establishing cause and effect, these data show that multiple morphological features of the plexus state, including apical F-actin^FOCAL structures formed in Neurog3^HI cells, are altered or lost in the absence of Neurog3 function.

We recently reported that the abnormal E14.5 Neurog3^EGFP/EGFP-null epithelium undergoes a delayed morphological compensation during late gestation (E16.5-18.5) to adopt an epithelial branched structure highly similar in gross morphology to controls (Bankaitis et al., 2015). This ‘corrective remodeling’ process provides a stringent model to assess whether Neurog3 promoter activity, which is reduced and expanded across the abnormal E14.5 Neurog3^EGFP/EGFP plexus, can be stimulated by Neurog3-independent cues that regulate the late-stage remodeling of the plexus to a more regular epithelial tree. Numerous inspections of E17.5 and E18.5 thick-sectioned and Muc1-labeled Neurog3^EGFP/EGFP pancreata confirmed epithelial remodeling to the grossly wild-type relative proportions of plexus, duct and branched states (Fig. S6A-D). The number of Sox9⁺ nuclei per linear unit of Muc1⁺ epithelium was comparable in E17.5 control and Neurog3^EGFP/EGFP plexus (Fig. S5B), in contrast to the increased cell density in E14.5 Neurog3^EGFP/EGFP plexus. Moreover, the spectrum of apical surface shapes within the E17.5 plexus of Neurog3^EGFP/EGFP samples was essentially indistinguishable from controls (Fig. 2E,G,J). Thus, multiple aspects of the abnormal morphogenesis in Neurog3^NULL tissue at E14.5 were corrected by late gestation. Surprisingly, such corrected plexus also displayed periodically spaced Neurog3^EGFP-HI cells (Fig. 2F,H), with narrowed F-actin^BELT structures at their apical surface (not shown). Although the EGFP^HI cell number in late-stage Neurog3^EGFP/EGFP plexus was reduced compared with control (quantified below), these data indicate that defects in Neurog3 transcriptional upregulation are substantially bypassed, independently of Neurog3 protein function, when the Neurog3^EGFP/EGFP plexus undergoes corrective remodeling. This effect was essentially specific for cells in the plexus state, because very few EGFP^HI cells were observed in non-plexus ductal or branched states, which is the case for normal ductal and branched state (Fig. 5C-D; Fig. S6A-D). These data suggest that morphogenetic cues deployed within a ‘normal’ remodeling plexus can profoundly influence Neurog3 levels, and that some initial steps in endocrine-cell commitment are initiated in part through Neurog3-independent pathways (Fig. 2K,L).

To study how particular morphogenetic inputs could be linked to the progression of cells through Neurog3-expressing states, we focused on the morphological alterations occurring as epithelial (Muc1⁺) Neurog3 expression is activated and upregulated during endocrine-cell birth. In E14.5 Neurog3^EGFP/+ pancreata, we observed Muc1⁺EGFP^LO cells (6.5% of Muc1⁺EGFP⁺ cells, n=95; Fig. S7A,C) without a narrowed or focalized apical surface (Fig. 3A,A′). Another 19.3% of Muc1⁺EGFP⁺ cells showed a narrowed but not focalized apical surface, and all were EGFP^HI (Fig. 3B,B′). The majority (74.2%) of Muc1⁺EGFP⁺ cells exhibited clear apical focalization, and again all were EGFP^HI. The mean Neurog3^EGFP intensity of apically non-narrowed, narrowed and focalized Muc1⁺EGFP⁺ cells indicated a strong link between narrowing and focalization and the EGFP^HI state (Fig. 3E; Fig. S7C). The majority of Muc1⁺EGFP^HI cells with a focalized apical surface also showed a basalward displacement of the cell body – assessed as a displaced nuclear location relative to the Muc1⁺ epithelial plane – and indications of basal-surface ruffling and protrusive distortions shown in our previous publications to represent filopodial migratory activity at the basal leading edge of the cell (Fig. 3C,C′; Fig. S7A) (Bechard et al., 2016; Löf-Öhlin et al., 2017). Progressive basalward movement was associated with elongation of the Muc1⁺F-actin^FOCAL structures, and an eventual detachment from the lumenal surface (Fig. 3D,D′), signifying completion of cell egress from the epithelium. A Neurog3^BAC transgenic reporter line Neurog3^RG1 (hereafter Neurog3^RG) (Bechard et al., 2016) independently confirmed the association of apical narrowing, focalization and basalward cell displacement with acquisition of the Neurog3^HI state (Fig. 3F-I′; Fig. S7B-D′). These results suggest that Neurog3 upregulation within the plexus occurs concomitantly with a finely resolved sequence of events beginning with apical narrowing, then F-actin^FOCAL formation and basalward cell movement during endocrine-cell birth (Fig. 3E,J).

Although Neurog3^EGFP/EGFP (null) pancreata lack differentiated endocrine cells (Gradwohl et al., 2000; Shih et al., 2012), the late-stage correctively remodeling Neurog3^EGFP/EGFP plexus contained many Neurog3^EGFP-HI cells. To define the stage in endocrine-cell birth when Neurog3-deficient cells fail, we compared the progression of Neurog3^EGFP/+ and Neurog3^EGFP/EGFP Muc1⁺EGFP⁺ cells through apical narrowing, F-actin^FOCAL formation and basalward movement. In E14.5 Neurog3^EGFP/EGFP pancreata, cells exhibited a uniformly circular and narrowed apical surface compared with heterozygous controls, with essentially all being EGFP^LO (Fig. 3K,M). In the late-stage ‘corrected’ Neurog3^EGFP/EGFP plexus, however, 58.5% of all Muc1⁺EGFP^HI cells (n=91) were apically narrowed, versus 18.3% in the Neurog3^EGFP/+ plexus (n=82) (Fig. 3K-R), suggesting that Neurog3^NULL cells do not efficiently progress past apical narrowing (Fig. S8A). The lengths of F-actin^FOCAL structures in Neurog3-null Muc1⁺EGFP^HI cells were reduced relative to control (Fig. S8C,D), suggesting reduced basalward movement and consistent with indications of diminished protrusive distortions at the basal surface (Fig. S8C-E). Finally, a small proportion of Muc1⁺EGFP^HI Neurog3^EGFP/EGFP cells were non-apically narrowed and not basally displaced (15.3% of all Neurog3^EGFP/EGFP Muc1⁺EGFP^HI cells versus 2.3% in Neurog3^EGFP/+ controls; Fig. 3R-T). We interpret this population as cells that had previously entered the initial stages of apical narrowing, focalization and Neurog3 upregulation, but then failed in apical-surface detachment and returned to the epithelium. These data show that complete Neurog3 deficiency compromises, but does not completely block, cells in the plexus from entering the sequence of apical narrowing, focalization and basalward movement normally defined by the endocrine-committed Neurog3^HI state. Moreover, these data suggest that a program of morphogenetic regulation of apical narrowing, focalization and basalward cell movement can be effected independent of Neurog3 protein function (Fig. 3T).

The molecular motor nmMyoII, produced as IIa (Myh9), IIb (Myh10) or IIc (Myh14) isoforms, acts on F-actin substrates to drive processes such as apical constriction and cell delamination (Martin and Goldstein, 2014). The active phosphorylated form of nmMyoII light chain (p-nmMyoII) was detected at the apical aspect of pancreatic epithelial cells, in a pattern consistent with close localization to the F-actin^BELT network (Fig. S9A,B). The p-nmMyoII signal was detected on most but not all Muc1⁺ F-actin^FOCAL structures (Fig. S9C,D). Immunodetection of nmMyoIIa/IIb isoforms showed a substantial signal in the blood vessels and a lower pan-mesenchymal signal (Fig. S9K,L), consistent with our detecting p-nmMyoII in both tissues (data not shown). The available reagents did not detect nmMyoIIa/IIb in the pancreatic epithelium or in Neurog3^EGFP-expressing cells. Analysis of nmMyoIIC using a GFP fusion knock-in mouse strain (Ebrahim et al., 2013), however, showed a pattern surprisingly selective for the epithelium at E14.5, which became largely restricted to non-acinar epithelium by E17.5 (Fig. S9E-J), and a low nmMyoIIc-GFP signal in differentiated endocrine cells (data not shown). These data are consistent with roles for nmMyoII activity in mediating epithelial morphogenesis and endocrine differentiation in the plexus.

To assess potential function(s) of nmMyoII in apical-surface narrowing and focalization within the plexus, we explanted plexus-rich segments of E15.5 dorsal pancreata (Fig. 4A), and cultured them with or without the nmMyoII inhibitor blebbistatin (BBS) (Straight et al., 2003), which directly interferes with intrinsic GTPase motor activity (Kovacs et al., 2004) to inhibit all nmMyoII-dependent processes. With vehicle treatment over 36 h of culture, the plexus maintained gross morphological features similar to those observed at equivalent in vivo stages. Conversely, BBS inhibition caused a broad, rapid and reproducible transformation of the plexus into an abnormal ‘duct-like’ state that exhibited increased lumen diameter and flattened epithelial cell morphologies (Fig. 4B-G). Within these transformed duct-like states, F-actin^BELT perimeters became enlarged compared with those in untreated explants (Fig. 4H-L). Much of any remaining plexus exhibited distended lumens that were prominent at the ‘nodes’ where epithelial segments of the plexus intersected, and acinar lumens were also distended (Fig. S10A-F). The BBS-mediated effects on epithelial morphology were dose dependent and reversible. An 18 h BBS washout led to restoration of normal F-actin^BELT morphologies (Fig. S11A-H), indicating a robust recovery in the 25-50 µM dose-range of live tissue with differentiating endocrine cells (not shown). These BBS doses and time frames of exposure that caused minimal cell death in the explants (not shown) were consistent with the notion that the effects on epithelial morphology were likely the outcome of acute nmMyoII inhibition. The effects of BBS on plexus morphology were not caused by loss of epithelial integrity or apical polarity, because aPKC, the primary cilium marker γ-tubulin, and E-cad maintained their respective apical and basolateral localizations during the experimental timeframe (Fig. S12A-F). Concomitant with BBS-induced duct-like transformation, there was a marked decrease in the number of Neurog3⁺ cells (Fig. 4C,F, quantified below). In rare regions where plexus remained, large numbers of both Neurog3⁺ cells still existed, suggesting that BBS had not accessed all areas of the explant volume equivalently, and that in such areas endocrine-cell birth could progress forward. Additionally, although the immunofluorescent signal for Sox9 was somewhat reduced in BBS-treated explants, there was no change in epithelial Pdx1 or Prox1 immunodetection intensity (Fig. S13A-H), suggesting that the reduced Neurog3⁺ cell numbers did not result from loss of pancreatic epithelial identity (Westmoreland et al., 2012). These experiments targeting nmMyoII activity, although not cell-type specific in nature, suggest functions for nmMyoII in limiting apical surface size, in maintaining the plexus morphology and in mediating the production of Neurog3-expressing populations within the plexus (Fig. 4M).

ROCK-nmMyoII pathway activity has been shown to regulate apical narrowing, cell migration and cell-rear retraction (Riento and Ridley, 2003; Worthylake et al., 2001), and our data presented here and that of others (Gouzhi et al., 2011) suggest that these processes are essential for endocrine-cell birth. To test how separate parts of the ROCK-nmMyoII pathway influence epithelial morphogenesis and endocrine-cell birth, we tested whether pharmacological inhibitors of ROCK and nmMyoII block or augment apical narrowing, focalization and basalward cell displacement in Neurog3-expressing cells. Consistent with our results above, BBS-treated Neurog3^RG explants showed reduced numbers of Neurog3^RG+ cells along the distended Muc1⁺ lumens (Fig. 5A-D; Fig. S14A-E; quantified below). Among the Neurog3^RG+ cells present in BBS-treated explants, quantification of non-apically narrowed, apically narrowed and basally displaced Neurog3^RG+ cells indicated a distinct shift toward a non-apically narrowed and non-displaced positioned state (Fig. 5G,G′). The Muc1⁺ F-actin^FOCAL structures observed showed a decreased length compared with controls (Fig. 5H,I). These results fit the notion that nmMyoII activity is crucial for apical narrowing, focalization and basalward cell movement during endocrine-cell birth.

Blocking ROCK activity with Y27632 (Uehata et al., 1997) is known to alter cell migration and also cause defects in cell-rear retraction in some cell types, with such changes attributed, at least in part, to ROCK-mediated modulation of nmMyoII activity, actin-filament stabilization and disassembly of integrin-signaling complexes (Riento and Ridley, 2003; Worthylake et al., 2001). In pancreatic explants, ROCK inhibition led to several effects on tissue and cell morphology that were noticeably distinct from those under nmMyoII inhibition. Y27632-treated explants showed no evidence of plexus transformation into ‘duct-like’ states, and there was no decrease in the number of Muc1⁺ Neurog3^RG+ cells exiting from the epithelium (Fig. 5E-G,G″). Upon close inspection, and consistent with known functions for ROCK in regulating cell migration, reporter-positive Neurog3^RG+ cells (we did not specifically distinguish between reporter-high versus low states in these drug-treatment conditions) scored in Y27632-treated explants displayed increased basalward movement relative to control Neurog3^RG+ cells (Fig. S14A,F-I). Treatment with Y27632 increased the number of Neurog3^RG+ cells with, as well as length of, F-actin^FOCAL structures at their apical surface (Fig. 5E,F,I) suggesting a block in detachment of the apical-cell rear, even with the exaggerated basalward movement of these cells. We confirmed a cell-rear detachment defect in Neurog3^RG+ cells using EdU pulse-chase methods. Briefly, replicating progenitor cells incorporate EdU at S phase, and over a 17-24 h period a proportion of them differentiate and become epithelially detached (Muc1⁻) (Bankaitis et al., 2015). Defective rear detachment caused by Y27632 treatment should thus cause an accumulation of EdU⁺ Muc1⁺ Neurog3^RG+ cells. One hour after EdU injection into pregnant dams, E13.5 pancreas was explanted and cultured for 18 h with or without Y27632. Confocal z-stack volumes of EdU pulse-labeled explants were scored for attached (Muc1⁺) or detached (Muc1⁻) EdU⁺ Neurog3^RG+ states. In vehicle controls, 63.4±3.1% of EdU⁺ Neurog3^RG+ cells were Muc1⁺, 30.6±3.9% were Muc1⁻ and 7.0±1.0% were non-categorizable (Fig. 5J-L). In Y27632-treated explants, 86.6±3.1% of EdU⁺ Neurog3^RG+ cells were Muc1⁺, 4.8±2.9% Muc1⁻ and 8.6±2.7% were non-categorizable (Fig. 5J,M,N). These data are consistent with two functions for ROCK in the birth of endocrine cells: first in limiting basalward movement, and then in directing cell-rear detachment to allow epithelial exit. Finally, short-term (6 h) time-lapse videos of pancreatic explants cultured on fibronectin, with or without BBS or Y27632, confirmed that nmMyoII inhibition reduced, whereas ROCK inhibition augmented, basalward movement of Neurog3^RG+ cells (Fig. S15A-O). Thus, ROCK and nmMyoII coordinate different aspects of endocrine-cell delamination by regulating apical narrowing, focalization, basalward movement and detachment of the rear of the cell from its focalized apical contact.

Because nmMyoII inhibition blocked behaviors associated with endocrine commitment and reduced the numbers of Neurog3⁺ cells, we hypothesized that nmMyoII activity is required for acquisition of the Neurog3-expressing state. We therefore quantified how BBS affected the number of Muc1⁺Neurog3⁺ cells formed in explants, as a metric for the BBS effect on all Neurog3-expressing states. Notably, because the processes of apical narrowing and focalization were perturbed under BBS-treated conditions, we lacked key criteria necessary for quantifying changes in the numbers of Neurog3^LO versus Neurog3^HI cells, and thus do not report on such changes here. Comparison of the (Neurog3⁺Muc1⁺DAPI⁺)/(Neurog3-negative Muc1⁺DAPI⁺) cell fraction in E15.5 explants treated with BBS or vehicle showed a near 50% reduction in all Neurog3⁺ cells under nmMyoII inhibition (Fig. 6A-B′,E). We next tested the effect of nmMyoII inhibition in the presence or absence of the small-molecule Notch inhibitor DBZ (Milano et al., 2004), which causes ‘default’ derepression of Neurog3 and acquisition of the Neurog3⁺ state (Serup, 2012; Bankaitis et al., 2015). Consistent with this effect, DBZ-only explants showed ∼2.5-fold increase in total (Neurog3⁺Muc1⁺DAPI⁺)/(Neurog3-negative Muc1⁺DAPI⁺) cells over vehicle alone (Fig. 6C,C′,E). In BBS/DBZ-treated explants, Neurog3⁺ cell numbers increased slightly over vehicle alone, but were greatly reduced compared with DBZ alone (Fig. 6D-E). These results indicate that nmMyoII inhibition interferes with processes controlling Neurog3 transcriptional activation or upregulation, or both, even when in the context of induced escape from Notch repression, suggesting that endocrine-cell birth requires permissive inputs supplied by nmMyoII activity.

Apical focalization and initiation of basalward movement are associated with the Neurog3^HI state in normal and Neurog3-null epithelium (Fig. 3). Because ROCK inhibition augmented apical focalization and basalward movement in Neurog3-expressing cells (Fig. 5), we hypothesized that ROCK normally functions to limit endocrine-cell birth. We first tested whether Y27632 treatment can alleviate the defective Neurog3 transcriptional upregulation observed in early-stage Neurog3^EGFP/EGFP (null) explants. In Neurog3^EGFP/EGFP explants, Neurog3 is broadly activated at low levels, but Y27632 did not increase the number of Neurog3^HI cells (Fig. 6F,G). Thus, Y27632 treatment is not able to promote Neurog3 upregulation in the absence of Neurog3 protein function. We next tested whether ROCK inhibition could raise the level of Neurog3 expression in cells already expressing low levels of functional Neurog3 protein, by exposing explants from a strong hypomorphic condition, Neurog3^EGFP/Flox (Wang et al., 2010), to Y27632. Neurog3 hypomorphs are much more prone to maintaining cells in the Neurog3^LO state than are wild-type or heterozygous-null controls in which most cells progress forward to the Neurog3^HI state (Wang et al., 2010; Bechard et al., 2016). Accordingly, vehicle-treated Neurog3^EGFP/Flox tissues had significantly increased Muc1⁺Sox9⁺EGFP^LO cell numbers over vehicle-treated Neurog3^EGFP/+ explants, and much reduced numbers of endocrine-committed Muc1⁺Sox9^-EGFP^HI cells (Fig. 6J versus M). The Y27632-treated Neurog3^EGFP/Flox explants, however, showed near-complete restoration of Muc1⁺Sox9⁻EGFP^HI numbers towards those seen in vehicle-treated Neurog3^EGFP/+ and in 20 nM DBZ-treated Neurog3^EGFP/Flox explants (Fig. 6K-Q). Therefore, we conclude that ROCK activity normally gates the transition from Neurog3^LO to Neurog3^HI states, and ROCK inhibition can bypass the defects in Neurog3 upregulation seen under Neurog3 hypomorphism. Interestingly, Y27632-treated Neurog3^EGFP/+ explants exhibited only a slight increase in the total Muc1⁺EGFP⁺/Muc1⁺EGFP⁻Sox9⁺ cell number relative to vehicle, indicating that Y27632 does not cause augmentation of Neurog3 activation or upregulation when a normal number of Neurog3-expressing cells are present (Fig. 6M,N,R).

That ROCK inhibition does not substantially increase Neurog3^HI cell numbers in Neurog3^EGFP/+ explants, but does in Neurog3 hypomorphs (Fig. 6Q,R), implies that ROCK inhibition cannot promote Neurog3^HI cell numbers if large numbers of Neurog3^HI cells are already present. We proposed that this inability resulted from Notch-mediated lateral inhibition from the Neurog3^HI cell population. To test this idea, we treated heterozygous Neurog3^EGFP/+ explants with DBZ and Y27632 together, and compared total cell numbers (Muc1⁺EGFP⁺/Muc1⁺EGFP⁻Sox9⁺) with those after vehicle, Y27632 or DBZ treatment. DBZ/Y27632 co-treated Neurog3^EGFP/+ explants showed a large increase in total (Muc1⁺EGFP⁺/Muc1⁺EGFP⁻Sox9⁺) cells over DBZ or Y27632 alone (Fig. 6M-P,R). These data suggest that the failure of Y27632 to promote robust Neurog3 upregulation in Neurog3^EGFP/+ explants is circumvented by inhibiting Notch activity, and thus that ROCK inhibition stimulates Neurog3 activation and upregulation provided that cells escape Notch repression. Notably, the majority of Muc1⁺EGFP⁺ cells in DBZ-treated Neurog3^EGFP/+ explants remained Sox9⁺ (Fig. S16B), suggesting that pharmacological Notch inhibition derepresses Neurog3 expression but interferes with the normal process of Sox9 downregulation normally associated with Neurog3 upregulation (Shih et al., 2012). Because this effect is not seen in the context of Y27632 alone, we propose it likely that ROCK and Notch activity regulate endocrine differentiation through separate mechanisms or at different stages of Neurog3 activation and upregulation. Like Y27632 alone, DBZ alone had no apparent effect on the number of cells upregulating EGFP-reporter in homozygous null explants, nor did DBZ and Y27632 in combination (Fig. 6F-I), indicating that the ability of DBZ or Y27632, both individually and in combination, to cause Neurog3 upregulation is dependent on the presence of at least low levels of Neurog3 protein. These results suggest that, provided Neurog3 protein is available, ROCK activity limits the normal transition of Neurog3^LO cells to Neurog3^HI endocrine-committed cells.

The results presented above, together with evidence from other reports, suggest that Neurog3 gene dose, Notch and ROCK-nmMyoII work together in a feedback control circuit that coordinates endocrine fat allocation, progenitor growth and epithelial plexus morphogenesis (Wang et al., 2010; Magenheim et al., 2011; Bankaitis et al., 2015). We were interested in the degree of robustness of this circuit, and its ability to yield, under various degrees of challenge, compensatory functions in driving the linked processes of morphogenesis and endocrine differentiation. We reasoned that the late-stage corrective remodeling of an initially (E14.5) abnormal Neurog3^EGFP/EGFP plexus, occurring concurrent with a partial restoration of Neurog3^HI cell number (Fig. 2), reflects the presence but only partial success of a compensatory response. We proposed that the failure to fully restore Neurog3^HI cell numbers resulted from the absence of feedback signals normally derived from endocrine-committed Neurog3 protein-producing cells during their departure from the plexus epithelium. We therefore tested whether a hypomorphic level of Neurog3 represented a sufficient feedback stimulus to confer a full compensatory response linking conversion of the plexus into an epithelial structure with a full restoration of Neurog3^HI endocrine-committing cells. We quantified the number of apically narrowed Muc1⁺EGFP^HI cells in the plexus (Muc1⁺EGFP^HI cells as a proportion of Sox9⁺ cells) at E13.5, E16.5 and E18.5 to compare endocrine-cell birth as a function of plexus remodeling in Neurog3^EGFP/+, Neurog3^EGFP/Flox and Neurog3^EGFP/EGFP states. Neurog3^EGFP/+ pancreata showed robust endocrine differentiation in all E13.5, E16.5 and E18.5 samples. The Neurog3^EGFP/EGFP-null pancreata again showed that the early and pervasive defects in Neurog3 upregulation were progressively but only partially rescued by E16.5 and E18.5, confirming that Neurog3-dependent processes induce the normal rate of Neurog3^HI cell birth (Fig. 7A-F). In the Neurog3^EGFP/Flox hypomorph, remarkably, although very low numbers existed at early stages, there was a complete restoration of the number of Neurog3^HI cells by E18.5 (Fig. 7A-J). Thus, even a strongly reduced level of Neurog3 is sufficient to engage, albeit late, an adaptive feedback control circuit that links endocrine differentiation and plexus morphogenesis to establish and maintain the balance between cell differentiation and epithelial remodeling.

Our results focus on a new ‘global’ understanding of the functional integration between programs of Neurog3-driven endocrine-fate determination, Notch-mediated progenitor maintenance, and ROCK/nmMyoII-mediated epithelial morphogenesis. We describe distinct ways in which the ROCK/nmMyoII pathway mediate a dissociable sequence of epithelial egression events linked to the acquisition of pre-endocrine-committed and endocrine-committed Neurog3-expressing states within the dynamically remodeling epithelium. The main findings of this paper are presented schematically in Fig. S17. We propose that a feedback circuit comprising ROCK/nmMyoII, Notch and Neurog3-dependent influences bestows self-organizing feedback properties that ensure the correct balance between progenitor growth, differentiation and morphogenesis.

ROCK/nmMyoII-dependent processes mediate a step-wise cell-egression sequence for progression into and through Neurog3^LO and Neurog3^HI states. Inhibiting nmMyoII activity interferes with apical narrowing and focalization, and basalward cell movement, and limits formation of Neurog3⁺ cells. Similar deficits in nmMyoII-inhibited explants occurred even when endocrine cells were overproduced when lifting Notch-mediated repression. Because our experiments did not discern cell-autonomous effects of nmMyoII inhibition, future dissection of its function specifically within Neurog3-expressing cells will likely be informative. ROCK inhibition augmented apical narrowing, focalization and basalward cell movement, and efficiently promoted progression of Neurog3^LO into Neurog3^HI cells in Neurog3 hypomorphs, but not in heterozygous Neurog3^EGFP/+ explants with much more robust Notch-mediated inhibition. A pro-endocrine modulation by ROCK inhibition occurred in heterozygous explants only with Notch co-inhibition, implying that, in normal tissue, ROCK de-repression promotes Neurog3^LO-to-HI transitions up to a ‘set point’ where an appropriate number of Neurog3^HI cells exert lateral inhibition as a dominant patterning determinant. Thus, we suggest that ROCK and nmMyoII are important factors in mediating timely and spatially localized cellular responses to signals that both drive (Neurog3 as a cell-autonomous instructor) and inhibit (via Notch) endocrine-fate acquisition. Intimate connections between epithelial morphogenesis programs and intrinsic cell-fate gene-regulatory networks are supported by additional studies, such as when Cdc42 inactivation led to failed progenitor-cell apicobasal polarization, block of epithelial tubulogenesis and severe endocrine-cell depletion (Kesavan et al., 2009). This phenotype resulted at least in part from improper positioning of cells within environments conducive for endocrine differentiation, as discussed below. Moreover, perturbation in Pdx1⁺ epithelial cells of CELSR proteins, components of the Wnt/PCP (planar cell polarity) signaling orthogonal to apico-basal polarity, caused large reductions in Neurog3⁺ cells (Cortijo et al., 2012). Other reports connect morphogenesis and differentiation via RhoA (Petzold et al., 2013; Azizoglu et al., 2017) and Rac signaling (Löf-Öhlin et al., 2017). An important link between Rac, apical domain size control and endocrine cell-fate acquisition occurs in mice and humans (Löf-Öhlin et al., 2017). It will be important in the future to dissect how morphogenetic pathways mediate recognition and processing of epithelial cues, possibly under precise local spatial control, that control progenitor growth and cell fate. Targeting such activities could alter the growth and differentiation properties of pancreas-differentiated iPS/ES cells in vitro.

Stem-cell niche microenvironments are commonly asymmetric, e.g. showing pro-maintenance or pro-differentiation cues at basal or apical aspects of a cell or tissue (Li and Xie, 2005). Theoretically, controlling the timing and exposure to such separated cues could regulate which cells undergo cell-fate allocation, and when. Particularly relevant, therefore, are the associations drawn here, in both Neurog3^EGFP/+ and Neurog3^EGFP/EGFP pancreata, between apical surface size/shape, basalward cell movement within and away from the epithelium, and Neurog3 expression. Cells in all genetic conditions studied here, including the Neurog3 protein null, only underwent Neurog3 upregulation upon narrowing and focalization of their apical surface (F-actin^FOCAL) and basalward cell displacement. We suggest, therefore, that apical-basal cell-body positioning within the epithelial plane controls, at least in part, acquisition of distinct Neurog3-expressing states upstream of Neurog3 protein function. This finding seems conceptually aligned with zebrafish retinal cell behavior, wherein neurogenesis is coupled to interkinetic nuclear migration along an apical-to-basal Notch gradient (Del Bene et al., 2008). In addition, in the zebrafish retina, apical domain size has been linked to modulation of Notch signaling (Clark et al., 2012). In the fish and mouse pancreas, high, medium and low Notch signaling impart duct, bipotent progenitor and endocrine fates, respectively (Ninov et al., 2012; Shih et al., 2012), associated in our studies with large, intermediate or narrowed/focalized apical surfaces. Because ROCK and nmMyoII inhibition showed opposite effects on the derivation of Neurog3⁺ cells, in the presence and absence of Notch inhibition, we suspect that regulation of apical surface size and basalward cell movement is a fundamental function by which ROCK and nmMyoII mediate cell fates.

Collectively, our data support previous studies suggesting that differentiating endocrine-cell progeny perform a ‘feedback support role’ for the epithelial plexus niche, and extend the range of functions for Neurog3 beyond its roles as an endocrine-lineage determinant and initiator of Notch-mediated lateral inhibition. The picture emerging is that Neurog3 influences numerous developmental processes in the plexus niche, including driving robust progenitor replication (Bankaitis et al., 2015), non-cell-autonomously refining gene-expression patterning (Wang et al., 2010; Bechard et al., 2016), cell-autonomously determining endocrine-cell fate (Gradwohl et al., 2000; Gu et al., 2002) and regulating tissue morphogenesis (Magenheim et al., 2011; Bankaitis et al., 2015). We show the first evidence that Neurog3-dependent non-autonomous mechanisms stimulate the production of new endocrine cells from the plexus (Fig. 7). We propose that a feedback function for Neurog3⁺ cells drives, or is a strong supportive influence in, the systematic integrative sustenance of epithelial growth and morphogenesis that is balanced with the apportionment of duct versus endocrine cell types. Future analyses might lead to understanding how transcription-factor determinants feed into the morphogenetic programs to build precise micro-environments, and their coordination to ensure timely placement of the right types and numbers of lineages throughout organogenesis. Under Pdx1 deficiency, early dorsal pancreatic bud evagination, cell polarization and microlumen formation all occur (Marty-Santos and Cleaver, 2016; Villasenor et al., 2010) but there is a complete failure in subsequent plexus assembly. Moreover, tissue-specific inactivation of transcription factors – such as HNF1b (De Vas et al., 2015), HNF6 (Pierreux et al., 2006) or Prox1 (Wang et al., 2005; Westmoreland et al., 2012) – all cause epithelial malformations concurrent with defective endocrine lineage allocation. Although these phenotypes were attributed largely to direct effects on genes such as Neurog3, equally consistent is the idea that an altered epithelial architecture precludes efficient spatial deployment of signals that drive endocrine differentiation and epithelial morphogenesis programs. Building systems-like network models linking epithelial morphogenesis, gene transcription and intracellular signaling as principal components for a larger ‘niche-framework’ will be important.

We propose that cooperative engagement of adaptive morphogenetic regulation with Neurog3-dependent processes confers robustness on endocrine differentiation, affording resilience even when genetic insults cripple endocrine differentiation (Bankaitis et al., 2015; Fig. 7). In the case of Neurog3^EGFP/EGFP-null epithelium, that the morphologically abnormal plexus state is populated very broadly with Neurog3^LO cells indicates that signals triggering Neurog3 transcriptional activation and specifying the endocrine-biased state are extensive over the epithelium. That adaptive restoration of Neurog3^HI cells occurs only during late-stage ‘corrective remodeling’ (in association with a specific sequence of cell-shape changes) suggests that localized morphological alterations in the plexus epithelium exert central control over endocrine commitment via Neurog3 upregulation. Importantly, the restoration of Neurog3^HI cells is only moderate in the absence of Neurog3 protein (Fig. 7), which we interpret as indicating that plexus morphogenesis normally synergizes with Neurog3-dependent processes to instruct appropriate numbers of endocrine cells to begin exit from the growing, dynamically remodeling epithelium. For example, even with a low level of Neurog3 protein function, the strongly hypomorphic epithelium could essentially completely restore Neurog3^HI cell numbers back from the massive mid-gestation reduction. We feel that such cross-regulatory feedback between non-autonomous morphogenetic determinants and Neurog3 represents a large-scale ‘community effect’ (Gurdon, 1988) to modulate progenitor flux toward the various epithelially derived lineages. We propose that such a community effect is likely influenced by biomechanical (e.g. intra/intercellular tension) and biochemical factors (Cebola et al., 2015; Marinari et al., 2012), some derived directly from endocrine cells engaged in a prolonged process of epithelial egression. Because cell departure is associated with fate acquisition in developing systems across phyla, linking this process to the architecture and behavior of the parental epithelium could reflect a broad process of organ-system self-organization.

Neurog3^EGFP [B6.129.S-Neurog3^{tm1(EGFP)Khk/Mmcd}] knock-in (Lee et al., 2001) and Neurog3^FLOX [Neurog3^{tm1(cre/ERT)Ggu}] mice (Wang et al., 2008) were from Guoqiang Gu (Vanderbilt University, Nashville, TN, USA). Neurog3^EGFP/+ and Neurog3^EGFP/EGFP pancreata were assigned definitively via EGFP-fluorescence level and lack of EGFP⁺ islet clusters, all readily observable under epifluorescence and agreeing with PCR genotyping and immunolabeling confirmation of genotype, as expected for a coding-region deletion allele. Tables S1 and S2 describe Neurog3^EGFP/EGFP and Neurog3^Flox/+ genotyping, and Neurog3^EGFP/Flox PCR genotyping followed Wang et al. (2010). Neurog3^RG1 mice were genotyped as described (Bechard et al., 2016). nmMyoIIC^GFP mice were a kind gift from Robert Adelstein (National Heart, Lung, and Blood Institute, Bethesda, MD, USA). Wild-type mice were of mixed genetic background. All protocols were approved by Vanderbilt University IACUC.

Embryonic pancreas was fixed in paraformaldehyde (4%, 4°C for 4-6 h, Fisher 04042-500) and for cryosectioning was sucrose equilibrated (30%, 4°C overnight, VWR 0335-12KG) and embedded in OCT (Tissue-Tek, Sakura). A Leica CM3050 S produced thick sections of 30-40 µm or 10 µm thin sections. Table S1 lists primary antibodies incubated overnight at 4°C in PBS containing 0.1% Triton X100, 0.5% BSA and 0.01% sodium deoxycholate. All secondary antibodies were from Jackson ImmunoResearch Laboratories, used at 1:500. Imaging was carried out using a Zeiss Apotome or confocal Zeiss LSM 510 META microscope. Confocal z-stack volumes were acquired at 40× magnification with 3D reconstructions using Imaris software. To measure apical F-actin cortices, 2D sections (1.5-2.0 µm sectional planes) from 3D volumes were assessed for whole F-actin^BELTS visible along the Muc1⁺ lumen surface, perimeter:aspect ratio was determined by line or polygon-trace functions in Imaris.

Distinguishing plexus and non-plexus states was carried out as described (Bankaitis et al., 2015). Fluorescence intensity measurements used mean pixel intensity in ImageJ. Sox9 and Neurog3 numbers were counted manually. For Neurog3⁺ or EGFP⁺ explant analysis, cell numbers in each state were summed from individual plexus volumes (six or more for each sample, ∼30% of total explant), averaged within each sample, and averaged between all samples. Measurements of morphological structures and nuclear densities were made using line or polygon-trace functions in Imaris. Nuclei were defined by H2B-mCherry signal in Neurog3^RG samples or Sox9 immunopositivity in Neurog3^EGFP/+ and Neurog3^EGFP/EGFP samples. Low-level Sox9 immunopositivity in cells initiating epithelial egression in Neurog3^EGFP/+ samples was digitally enhanced (brightness and contrast) in Image J. All Student's t-tests were two-tailed.

Explants were cultured at 37°C in 5% CO₂ at the air-liquid interface on trans-well supports (Costar 3422) in DMEM (no phenol red) with 10% FBS, penicillin/streptomycin (10 U/ml each) and 1% L-glutamine (Shih and Sander, 2014). Blebbistatin [Abcam, S(-) form, CAS, 856925-71-8], Y27632 (Selleckchem, S1049) and DBZ (Cayman Chemical, 209984-56-4) or vehicle (0.1% DMSO final) were added. Explants were fixed (for 2 h in 4% paraformaldehyde) for analysis.

We thank Drs Matt Tyska, Roland Stein and Guoqiang Gu (Vanderbilt University), and various Gu and Wright lab members for helpful discussions. Supporting data were gathered in the Vanderbilt Cell Imaging Shared Resource using imaging scholarships from the Vanderbilt University Medical Center Digestive Disease Research Center and Diabetes Research and Training Center and the Vanderbilt-Ingram Cancer Center, supported by the National Institutes of Health.