Canonical Notch signaling in the developing lung is required for determination of arterial smooth muscle cells and selection of Clara versus ciliated cell fate

Lung development is orchestrated by complex mesenchymal-epithelial interactions that coordinate the temporal and spatial expression of multiple regulatory factors that are required for proper organ formation. Distinct populations of stem cells contribute to the epithelial and mesenchymal compartments (Mailleux et al., 2005; Perl et al., 2002; Rawlins et al., 2007; Rock et al., 2009). In the tracheal epithelium, the basal cell generates mucous cells, Clara (secretory) and ciliated cells (Hong et al., 2004; Rock et al., 2009). Smaller bronchi contain the latter two cell types and pulmonary neuroendocrine cells (PNECs). The distal-most airway, the alveolus, is lined with thin layers of flat Type I cells and cuboidal Type II cells (Kimura and Deutsch, 2007; Rawlins and Hogan, 2006). At the pseudoglandular stage (E11.5-16.5), during which most of airway branching morphogenesis takes place, it is thought that the terminal buds contain a population of multipotent epithelial progenitors (Perl et al., 2005). As the bronchial tree extends, descendants of these cells give rise to lineage-restricted progenitors that produce Clara and ciliated cells (and possibly other cell types) in the conducting airways (Cardoso and Lu, 2006; Perl et al., 2002).

The lung mesenchyme is comprised of multiple cell types, including connective tissue, endothelial cells, lymphatics, smooth muscle cells surrounding airways and blood vessels, myofibroblasts involved in septum formation, and cartilage-forming cells in the trachea. In addition, pleura-derived mesothelial cells cover the outer surface of the lung. The developmental origin of these cells is a matter of some dispute (Hall et al., 2000), with most cells believed to be derived from the splanchnic mesenchyme, whereas other cells (endothelial, smooth muscle) are believed to invade the lung as it expands (Cardoso and Lu, 2006; Galambos and Demello, 2007; Hall et al., 2000). Although bronchial smooth muscle cells (bSMCs) are derived from a distal lung mesenchyme lineage expressing fibroblast growth factor 10 (FGF10) (Mailleux et al., 2005), vSMCs are thought to derive from an invading population (Hall et al., 2000).

Evidence demonstrating the importance of Notch signaling in the developing respiratory system is rapidly growing. Mice that are genetically deficient in Hes1, a target of Notch signaling in several biological systems, show hyperplasia of PNECs and a decreased number of Clara cells, suggesting the bi-potential precursors of Clara and ciliated cells are separated from PNEC precursors via a Notch-mediated lateral inhibition feedback loop (Ito et al., 2000; Shan et al., 2007). As this manuscript was being prepared for publication, a role for Notch signaling as a suppressor of the ciliated cell fate was reported (Guseh et al., 2009; Tsao et al., 2009). Induced expression of a constitutively active Notch1 intracellular domain (N1ICD) in lung epithelial cells throughout development promoted mucous metaplasia and remarkably decreased the number of ciliated cells (Guseh et al., 2009). Conditional removal of Pofut1, a glycosyltransferase required for Notch signaling and possibly other cellular functions (Kopan and Ilagan, 2009), promoted ciliated cell expansion at the expense of Clara cells (Tsao et al., 2009). Interestingly, airway branching morphology was not altered by loss of Notch signaling in the epithelium, despite previous loss-of-function reports demonstrating enhanced branching when lung anlagen were cultured in the presence of γ-secretase inhibitors (GSI) (Tsao et al., 2008) or antisense Notch1 oligonucleotides (Kong et al., 2004). The discrepancy between in vivo and in vitro loss of function analyses might be explained by an unknown function for Notch signaling in lung mesenchyme.

Although these observations strongly suggest a role for Notch signaling in the developing lung, several caveats limit our ability to identify the cells in which Notch receptors function, and which specific receptor(s) contribute to lung organogenesis. Overexpression of N1ICD (Guseh et al., 2009) exposed the tissue to non-physiological levels of Notch pathway activation in both the level and duration of the signal. Moreover, given that Hes1 can respond to other signaling pathways (Yoshiura et al., 2007), notably FGF (Nakayama et al., 2008), its activation might not depend on Notch in every cellular context (Lee et al., 2007). To look at which specific cell types require Notch activity during lung morphogenesis, and to begin to assign functions to specific receptors, we examined the role of Notch signaling in different compartments throughout lung development. Given the dominant role suggested for Notch1, we wished to visualize the lineages derived from cells experiencing Notch1 activation. To map these lineages, we modified N1IP::CRE^LOW (Notch1 Intramembrane Proteolysis) (Vooijs et al., 2007) to generate the N1IP::CRE^HI knock-in mouse strain in which Cre activity was improved, thus increasing detection sensitivity. These experiments were followed by detection of N1ICD with epitope-specific antibodies to observe sites of Notch1 activity. Finally, genetic inactivation of the canonical Notch pathway in epithelia or jointly in the mesenchyme and mesothelium was achieved by removal of RBPjκ, the DNA binding partner of all four Notch receptors and a core component of canonical Notch signaling (Kopan and Ilagan, 2009); more specifically, RBPjκ is essential for Notch-mediated Hes1 activation. We uncovered a specific function for Notch signaling in the specification of the pulmonary vSMCs and in mesothelial epithelial-mesenchymal transition (EMT). We confirmed the function of Notch in selection of Clara or ciliated cell fate and extended these observations, demonstrating a lateral inhibitory role for Notch1 in this process and during Clara cell regeneration.

In vivo fate mapping of cells that experienced Notch1 activation with N1IP::CRE^LOW allows identification of lineages in which Notch1 activity might be required (Vooijs et al., 2007). To enhance our ability to image such lineages in the lung, we generated N1IP::CRE^HI knock-in mice in which Cre recombinase [instead of Cre—6-Myc-Tag (Cre-6MT)] replaced the Notch1 intracellular domain. Ligand binding unfolds a negative regulatory domain, triggers ectodomain shedding and enables γ-secretase-mediated proteolysis of the Notch transmembrane domain. This leads to the release of Cre (Vooijs et al., 2007). When combined with a strain carrying a conditional reporter, Cre-mediated excision of a loxP-flanked ‘stop’ cassette constitutively activates reporter expression and indelibly marks cells that experienced Notch activation and all of their progeny. In N1IP::CRE^LOW, the inefficient Cre-6MT marks only a subset of cells (those experiencing moderate to high levels of sustained Notch activity, such as endothelium) (Vooijs et al., 2007), whereas the new N1IP::CRE^HI, R26R strain marked cells receiving moderate-to-low levels of Notch activity and therefore increased coverage of Notch1 activation patterns [a full description of this line will be provided elsewhere, but compare the lung image shown here and in Vooijs et al. (Vooijs et al., 2007)]. We used N1IP::CRE^HI, R26R mice to determine which lineages within the lung experienced Notch1 activation during development.

Scattered, β-galactosidase-labeled mesenchymal (Fig. 1A) and mesothelial cells (black arrowhead in Fig. 1A,B) were detected in the lung mesenchyme at E13.5 in N1IP::CRE^HI, R26R mice. As the lung matured, the number of these cells increased (Fig. 1A,C,D). To identify which mesenchymal cell types were derived from cells experiencing Notch1 activation, we co-immunostained tissue sections with anti-β-galactosidase and cell-type-specific antibodies (SM22α, PECAM, SMA). The N1IP::CRE^HI reporter abundantly marks the vascular plexus (Fig. 1I-N) and both arterial endothelial cells and vSMCs (gray arrowheads in Fig. 1O-Q). By contrast, β-galactosidase was not activated in bSMCs (Fig. 1G, white arrowheads in Fig. 1O-Q) or myofibroblasts located at the tip of the alveolar septum (white arrowheads in Fig. 1R-T). These data suggest that by E18.5, the descendents of cells experiencing Notch1 activation contributed extensively if not exclusively to endothelial and vSMC cells.

RBPjκ is ubiquitously expressed in lung mesenchyme (Fig. 2A, Fig. 3A). Dermo1-Cre (Yu et al., 2003) is expressed within the lung mesenchymal (supplementary material Fig. S1A-C) and mesothelial (see below) lineages; only a few endothelial cells are targeted by this strain and no expression is detected in the epithelium (Yin et al., 2008). To test whether canonical Notch signaling is necessary for mesenchymal lung development, we employed Dermo1-Cre (Yu et al., 2003) to delete floxed RBPjκ alleles from the mesenchymal and mesothelial lineages within the developing lung (Drm1-RKO mice). Drm1-RKO mice die within 24 hours due to a highly penetrant ventricular septal defect (VSD; supplementary material Fig. S2). Notably, Dermo1-Cre is expressed in the cardiac cushion tissue that is generated by endothelial-mesenchymal transition (EnMT) (Lavine et al., 2008; Timmerman et al., 2004). This indicates an unappreciated requirement for Notch signaling after EnMT has occurred. Importantly, Drm1-RKO pups filled their lungs with air and their breathing was not labored, consistent with normal surfactant expression and lung function in Drm1-RKO mice.

By E10.5, Dermo1-Cre had efficiently deleted RBPjκ from mesenchymal and mesothelial cells (Fig. 2B) (Yin et al., 2008), but Drm1-RKO lungs were morphologically indistinguishable from controls (Fig. 2C-H). Clara and Type II cells formed properly (supplementary material Fig. S3A-D), indicating that canonical Notch signaling does not contribute to the complex mesenchymal-epithelial feedback loops required for lung development (White et al., 2007; White et al., 2006; Yin et al., 2008). As expected, epithelial cells (Fig. 3C) and vascular cells retain RBPjκ protein (white arrowheads in Fig. 3D-I) within Drm1-RKO lungs. Because Dermo1-Cre is rarely active in endothelial lineages (supplementary material Fig. S3E-G) (Yu et al., 2003), these observations argue against conversion of lung mesenchyme, which is RBPjκ-depleted at E10.5, to endothelium (Stenmark and Abman, 2005).

To quantify the contribution of the Dermo1-Cre lineage to the vascular and bronchial SMC lineages, we counted cells double-positive for smooth muscle actin and β-galactosidase (SMA⁺, β-gal⁺) in arteries and airways of Dermo1-Cre, R26R embryos. SMA and SM22α are SMC markers, and β-galactosidase is a lineage tracer used to quantify the contribution of Drm1-RKO cells to these lineages. To quantify the contribution of RBPjκ-deficient (RBPjκ^−/−) cells to SMC, we stained for SM22α and RBPjκ proteins (Fig. 3J-L). SM22α⁺, β-gal⁺ cells and SM22α⁺, RBPjκ^−/− cells contributed equally to bSMC (Fig. 3M; 90% for both β-gal⁺ (gray arrowheads in supplementary material Fig. S1A-C) and RBPjκ^−/− cells (gray arrowheads in Fig. 3J-L). The absence of RBPjκ protein from most bSMCs indicated strongly that Notch signaling was dispensable for the execution of the bSMC differentiation program.

In contrast to bSMC, a striking requirement for RBPjκ was observed in vSMC. Whereas most vSMCs (81%) labeled with SM22α and β-galactosidase, indicating a robust contribution from the Dermo1-Cre lineage (Fig. 1H and Fig. 3M; white arrowheads in supplementary material Fig. S1A-C), only 15% of vSMC cells were SM22α⁺, RBPjκ^−/− (Fig. 3J-M); the rest contained RBPjκ protein and therefore must have arisen from outside the Dermo1-Cre lineage. This indicates that although canonical Notch signaling was not required for the execution of the bSMC differentiation program, Notch signaling promoted the selection of the vSMC fate. Finally, although the majority of endothelial cells appear to have been derived from outside the Dermo1-Cre lineage at E16.5, 21% of the endothelial cells (VEGFR2⁺) were β-galactosidase-positive and, thus, Dermo1-Cre derived (Fig. 3M; gray arrowheads in supplementary material Fig. S1D-F). Interestingly, the fraction of Dermo1-Cre-derived endothelial cells was reduced in Drm1-RKO lungs (Fig. 3M).

N1IP::CRE mice detect a population of cells, only some of which are engaged in Notch signaling. To look at which cells activated Notch1 within the lung mesenchyme, we used anti-N1ICD antibody to detect Notch1 activation. Double staining for N1ICD and SMA at E14.5 revealed that Notch1 receptor is activated in endothelial cells (Fig. 4A-C, asterisks), vSMCs and in peripheral mesenchymal cells (Fig. 4A-C, white arrowheads) but never in bSMCs (Fig. 4A-C, gray arrowheads). These observations complement the results obtained with the N1IP::CRE^HI (Fig. 1), indicating that Notch1 signaling is required for the arterial vSMC fate in developing mesenchyme and not in a general early precursor.

Platelet-derived growth factor receptor (PDGFR)-β is expressed in pericytes, the progenitor for vSMCs (Andrae et al., 2008), where it regulates migration, proliferation and differentiation into vSMC (Jin et al., 2008). We therefore measured the expression level of PDGFR-β in Drm1-RKO mesenchyme at E14.5. PDGFR-β was expressed in the wild-type mesenchyme (Fig. 4D-F) and in RBPjκ-positive cells located within Drm1-RKO mesenchyme (white arrowheads in Fig. 4G-I), but was absent from RBPjκ-negative cells, as judged by immunohistochemistry (gray arrowheads in Fig. 4G-I). Reduced PDGFR-β mRNA levels were confirmed by quantitative RT-PCR; expression of endothelial controls (VEGFR1 and 2) was not significantly altered in the same sample at E13.5 (Fig. 4J). Because Notch signaling might also be required for cell proliferation (Sakata et al., 2004; Wang et al., 2003), we assessed cell proliferation by immunohistochemical detection of phospho-histone H3 and observed no differences in the number of cells positive for phospho-histone H3 between Drm1-RKO and wild-type E14.5 lungs (Fig. 4K-M, white arrow points to a proliferating RBPjκ-negative cell).

During this analysis, we noticed that the overall numbers of vSMC in Drm1-RKO lungs did not change, perhaps explaining the normal lung morphology and function in Drm1-RKO mice. We therefore investigated the origin of the compensating RBPjκ-positive vSMC population. One possible source is the mesothelium, a specialized epithelium that covers the lung (Herrick and Mutsaers, 2004) in which some RBPjκ-positive cells are detected at E13.5 (gray arrowheads in Fig. 3G-I; supplementary material Fig. S4). The mesothelium can undergo EMT and contribute vSMC to the gut (Wilm et al., 2005) and the heart (Cai et al., 2008; Zhou et al., 2008). To assess mesothelial contribution to lung vSMC, we utilized Wt1-Cre Rosa^EYFP mice to fate map cells arising from this population (Wilm et al., 2005). Wilm's tumor protein 1 (Wt1) is a mesothelium marker. As seen in the gut, Wt1 expression was restricted to lung mesothelium throughout development (supplementary material Fig. S5A-C), but yellow fluorescent protein (YFP) expression was detected within the lung parenchyma and in endothelial cells three months after birth. YFP-positive cells were also observed in the mural wall, which consists of smooth muscle and pericytes. Importantly, YFP-positive vSMC cells were rarely seen in the lung (supplementary material Fig. S5D-I). Thus, we conclude that the mesothelial lineage does not constitute a major population of vSMC progenitors (see also Que et al., 2008).

Several investigators have reported that endothelial cells can transition into mesenchymal cells, a process called endothelial-mesenchymal transition (EnMT) (Arciniegas et al., 2007). Therefore, we tested whether endothelial cells contribute to pulmonary vSMCs using Tie1-Cre, R26R mice (supplementary material Fig. S6). Tie1-Cre is exclusively expressed in the endothelium, yet most vSMCs in the proximal pulmonary arteries were double-positive for β-galactosidase and SMA in Tie1-Cre mice, confirming that EnMT is responsible for forming most of the early (proximal) vSMCs. The contribution of the Tie1-Cre lineage to vSMC gradually declined; at the most distal positions, only an occasional vSMC was β-galactosidase-positive. Distal contribution from EnMT might have continued in Drm1-RKO mice, explaining the presence of RBPjκ-positive vSMC cells in these mice. However, we could not test this hypothesis directly because Tie1-Cre, Notch1^flox/flox embryos die at E9.5 due to the essential role of Notch1 in vascular development (Conlon et al., 1995; Huppert et al., 2000; Krebs et al., 2000), prior to lung bud formation (supplementary material Fig. S7) (Cheng, 2006).

N1IP::CRE^HI labeled mesothelial cells (Fig. 1A-F), which prompted us to examine whether Notch signaling regulated EMT in this population. We used the organ culture explant system previously described (Wilm et al., 2005) to address this question. Briefly, we labeled surface mesothelial cells in cultured embryonic lungs at E14.5 with a fluorescent chemical (CCSFE; 5-(and-6)-carboxy-2,7-dichlorofluorescein diacetate succinimidyl ester) for 2 hours (Fig. 5A). Following washout, the lungs were incubated for 48 hours to examine the location of CCSFE-labeled mesothelial cells. Some CCSFE-labeled cells were detected in the mesenchyme at the end of the chase (Fig. 5B), indicating that lung mesothelial cells can undergo EMT ex-vivo. Transforming growth factor β (TGFβ), a general inducer for EMT, increased the number of CCSFE-labeled cells migrating into the mesenchyme, and the TGFRII inhibitor SD208 blocked EMT (Fig. 5C-E). To determine whether Notch signaling participates in EMT and/or mesothelial migration, we repeated the organ culture with embryonic lungs in the presence of the γ-secretase inhibitor DAPT (N-[N-(3,5-difluorophenacetyl)-L-alanyl]-S-phenylglycine t-butyl ester). γ-secretase is required for Notch cleavage and therefore for pathway activity. DAPT treatment reduced mesothelial migration in a dose-dependent manner (Fig. 5F-I). Addition of TGFβ restored migration in the presence of DAPT, indicating that in the presence of strong TGFβ signals, Notch signaling was dispensable for EMT (Fig. 5I).

To confirm that mesothelial EMT could be enhanced by Notch signaling, we locally activated Notch signaling within the lung mesothelial cells in organ culture using a recombinant Cre-recombinase fused to the HIV-TAT peptide (TAT-Cre) (Shaw et al., 2008; Xu et al., 2008). To induce constitutive Notch activity in labeled mesothelial cells, we cultured embryonic lungs from E14.5 Rosa^N1ICD-GFP (Murtaugh et al., 2003) or Rosa^EYFP mice (in which Cre activity will delete a loxP-flanked stop segment, resulting in expression of N1ICD::GFP or YFP, respectively) with 5 μM TAT-Cre for 5 hours. Following cellar uptake of TAT-Cre on the surface of the lungs and subsequent washout, growth media was replaced and the lung organ cultures were maintained for an additional 3-day period. Mesothelial EMT and migration were assessed by immunohistochemistry with anti-GFP antibody (Fig. 5J-L). Whereas 37% of mesothelial cells underwent EMT in control ROSA-YFP cultures, 77% of cells expressing N1ICD::GFP underwent EMT. Furthermore, 5% of the N1ICD cells migrated more than 50 μm inwards, whereas control YFP⁺ cells were never detected that deep (white arrowheads in Fig. 5K,L). These results indicate that Notch activation is sufficient to induce EMT in the mesothelium and that it accelerated the mobility of mesothelium-derived cells.

By P21, when lung epithelial development is nearly complete, all of the lineages derived from cells experiencing Notch1 activation (as marked by N1IP::CRE^HI, R26R) have been marked. Histological analysis of intact lungs identified β-galactosidase-positive cells as airway epithelial cells (Fig. 6A,B). Immunohistochemistry for CC10 (Clara cells; Fig. 6C), acetylated tubulin (ciliated cells; Fig. 6D), and calcitonin gene related peptide (CGRP) (PNECs; Fig. 6E) determined that β-galactosidase-positive cells differentiated predominately into Clara cells. A few ciliated cells were also labeled (see below). To obtain a three-dimensional (3D) image of how the epithelial lineages were distributed in the bronchial tree, we manually removed alveolar capillary cells that are robustly labeled with β-galactosidase, obscuring the epithelium. β-galactosidase-positive cells appeared throughout the conducting airway (Fig. 6F), reaching the highest density in the distal conducting airways (Fig. 6G-I).

To examine the role of Notch signaling in lung epithelia, we generated SHH-Cre, RBPjκ^flox/flox (Shh-RKO) mice (Harfe et al., 2004; Harris et al., 2006). Staining for RBPjκ in developing Shh-RKO lungs confirmed that RBPjκ was absent from lung epithelial cells but its expression remained intact in all other pulmonary lineages (supplementary material Fig. S8A-D). Unlike Pfout1-deficient mice that survive to weaning (Tsao et al., 2009), Shh-RKO mutant mice die at birth from an undetermined cause, apparently unrelated to the lung because breathing appeared normal, mice were not cyanotic and no morphological pulmonary defects were observed. Shh-RKO reproduced the phenotypes seen with loss of Pofut1 (Tsao et al., 2009), namely, expansion of Foxj1-positive ciliated cells at the expense of Clara cells throughout the entire lung epithelium (supplementary material Fig. S8I-O). The stem cell population still gave rise to normal alveolar epithelial cell types (Type II and Type I cells) in Shh-RKO lungs, indicating that the defect was restricted to the Clara and ciliated lineages (supplementary material Fig. S8P-S). Collectively, these data suggest that Notch signaling functions during lung development in a bi-potential progenitor to either induce the Clara cell fate or to permit Clara cell differentiation by blocking a default ciliated fate. To differentiate between these possibilities, we examined ciliated cell production during the pseudoglandular stage, when ciliated cells and Clara cells are determined from an epithelial progenitor cell population residing at the branch tip among proSP-C-positive cells (Fig. 7A). At E16.5, Foxj1 (which marks ciliated cells) was observed only within the proximal airway in a ‘salt and pepper’ pattern (Fig. 7A, asterisk), suggesting that determination of ciliated cell fate occurs in a transitional zone located between distal stem cells and proximal differentiated daughters. By contrast, nearly all the proximal cells expressed Foxj1 in RBPjκ-deficient epithelium (Fig. 7B, asterisk). The salt and pepper distribution of Clara and ciliated cells further suggests a lateral inhibition mechanism involving activation of Notch signaling in neighbors of nascent ciliated cells at the transition zone (Fig. 7E). To test whether lateral interaction among decedents led to Notch1 activation, we used an anti-N1ICD antibody to detect activated Notch1. Double staining for N1ICD and Foxj1 at E16.5 revealed a mutually exclusive distribution of these markers (Fig. 7C,D). Coupled with N1IP::Cre labeling (Fig. 6), these data indicate that neighbors of Foxj1-positive cells, but not the distal progenitors, activate Notch signaling. Cells experiencing Notch1 activation become Clara cells. The few β-galactosidase-positive ciliated cells we saw were either derived from a labeled Clara cell (Perl et al., 2002) or indicate that Notch signaling is activated again during formation of a ciliated cell from a Clara cell.

To test whether ciliated cells required canonical Notch signaling subsequent to their formation, we generated FOXJ1-Cre, RBPjκ^flox/flox mutant mice in which RBPjκ was deleted after the ciliated fate was determined. These mice were viable and fertile and displayed no change in the frequency of ciliated or Clara cells (supplementary material Fig. S9). Examination of SP-C-rtTA, (tetO)₇-Cre, Rosa^N1ICD-GFP triple transgenic mice revealed a reduction in ciliated cells, establishing that N1ICD inhibits ciliated cell fate acquisition [data not shown and Guseh et al. (Guseh et al., 2009)]. These observations expand the data reported earlier (Tsao et al., 2009) and confirm that Pofut1 deletion expanded the ciliated epithelium by eliminating Notch1 signals.

To determine whether Notch1 was reactivated during Clara cell regeneration, we used N1IP::CRE^LOW, R26R mice (Vooijs et al., 2007) that label Clara cells infrequently compared to N1IP::CRE^HI (compare Fig. 8A,B and Fig. 6F,G). A few β-galactosidase-positive cell clusters could be discerned within the conducting airway epithelia at P14 (Fig. 8A,B). Histological analysis revealed that each cluster included one to eight cells (Fig. 8C,D), and that these cells were positive for CC10 (Clara cell marker) but negative for CGRP (PNEC marker). Next, we injured the airway epithelium in N1IP::CRE^LOW, R26R mice with a single injection of naphthalene and examined β-galactosidase staining patterns throughout the regeneration process. Naphthalene toxicity induced apoptotic death of most Clara cells within conducting airways by 3 days. Following this, lung epithelial stem cells initiate a regeneration program, which is nearly complete 14 days after a single exposure to naphthalene (Plopper et al., 1992; Rawlins et al., 2007). Following injury, all epithelial β-galactosidase-positive cells disappeared (Fig. 8J), and regeneration was initiated from β-galactosidase-negative cells. β-galactosidase-positive cells began to reappear 5 days post-injury (Fig. 8N) and increased in number within clusters as the epithelium recovered (Fig. 8K,L). Furthermore, as judged by the timing of β-galactosidase and CC10 expression at 5, 7 and 21 days post-injury, Notch1 activation preceded Clara cell differentiation during regeneration (Fig. 8M-P, Table 1).

To test whether Clara cells required Notch for their maintenance, we generated triple-transgenic CCSP-rtTA, (tetO)₇-Cre, RBPjκ^flox/flox mice. Although some lung toxicity and mosaicism was reported in these mice (Sisson et al., 2006), mosaic deletion patterns of RBPjκ would predict that no RBPjκ^−/− Clara cells would be found 6 weeks after doxycycline (DOX) administration if RBPjκ played an important role in Clara cell maintenance. When DOX was added to the diet beginning at P31, many RBPjκ-null Clara cells were detected by double staining (supplementary material Fig. S10) indicating that they were not replaced by cells expressing RBPjκ. Collectively, these results demonstrate that Notch1 activation did not mark stem cells or their transient-amplifying daughters. Instead, it was activated in cells during the final stages of differentiation, where it might regulate mucous production [data not shown and Guseh et al. (Guseh et al., 2009)]. Notch signaling might not be required for Clara cell differentiation or maintenance, but this conclusion is confounded by the low turnover rates of the adult lung.

A role for Notch upstream of Hes1 in regulating PNEC fate selection has been proposed (Ito et al., 2000). Pofut1-deficient mice appear to confirm this suggestion because they contain an increased number of PNECs (Tsao et al., 2009). We thus examined the distribution of PNECs in Shh-RKO mice within the lung epithelia using immunohistochemistry. Although the average number and size of PNEC clusters was slightly increased in RBPjκ-deficient epithelia (Fig. 9A,C,E), this increase was similar to the increase reported in Hes1 heterozygote lungs (Ito et al., 2000). Whereas Hes1 expression was essentially abolished in the epithelia of Pofut1-deficient lungs (Tsao et al., 2009), Hes1 in RBPjκ-deficient epithelia was not dramatically reduced at either the protein or mRNA levels as late as E18.5 (Fig. 9B,D,F). Thus, Hes1 expression in lung epithelial cells does not rely on canonical Notch signaling.

In this study, we used genetic analyses to decipher where Notch signaling acts during lung development. In addition, the new N1IP^HI::CRE, R26R mice permitted a higher resolution mapping of the lineages experiencing at least one round of Notch1 activation. Conditional gene targeting was complemented by organ culture experiments to confirm the role of activated Notch1. With these tools, we report a function for Notch signaling in vSMC specification and in regulating mesothelial EMT and migration rates, and confirm recent observations describing Notch signaling as a suppresser of ciliated cell fate that permits or induces secretory cell formation (Guseh et al., 2009; Tsao et al., 2009). We extend these observations by providing evidence for direct Notch1 involvement in this process. Notch activation in regenerating Clara cells could reflect regulation of mucous secretion (Guseh et al., 2009). Finally, we report interesting differences between mice lacking RBPjκ and Pofut1 (longevity, role in Hes1 expression) and report a function for Notch signaling after EMT occurred in the heart. This might reflect an impact of Pofut1 on lung development via non-canonical Notch activity or additional substrates. However, RBPjκ is dedicated to canonical Notch signaling in the lung, where RBPjκ-like protein (RBPL) will compensate for Notch-independent RBPjκ activity such as interaction with Pitf1a (Beres et al., 2006).

Ablation of Notch signaling in lung mesenchyme and mesothelium with the Dermo1-Cre transgene resulted in no obvious morphological alterations. Published reports (Proweller et al., 2007) established that canonical Notch signaling was not required for the maintenance of the SMC cell fates; indeed, bSMC formed in the absence of RBPjκ as efficiently as they did in its presence. By contrast, selection of the vSMC fate, or migration of SMCs arising outside of the arteries, was severely impaired. vSMC were strongly labeled in N1IP::CRE^HI, R26R mice, implying that in addition to its established role in promoting endothelial cell differentiation into a vascular network (Gridley, 2007), Notch1 activation was required within vSMC precursors. A possible explanation for the involvement of Notch in arterial vSMC differentiation has been proposed based on observations that three SMC marker genes (SM-MHC, SMA and PDGFR-β) responded to Notch activation (Doi et al., 2006; Jin et al., 2008; Noseda et al., 2006). We find that expression of PDGFR-β is significantly reduced in RBPjκ-deficient cells, confirming that Notch signaling is required cell-autonomously for PDGFR-β expression (Jin et al., 2008). Because PDGF signaling plays a crucial role in the recruitment of pericytes and vSMC progenitors during vasculogenesis (Andrae et al., 2008), loss of RBPjκ (and hence, PDGFR-β) might impair vSMC recruitment. Jagged1-expressing peripheral endothelial cells activate Notch signaling on pericytes, promoting SMC differentiation (High et al., 2008; Liu et al., 2009). Therefore, Notch signaling is required both for recruitment and differentiation in vSMC. Accordingly, we detected cells that experienced Notch1 activation in mesenchymal cells surrounding the lung vasculature. Bronchial SMC differentiation depends on FGF10 (Mailleux et al., 2005). Therefore, Notch and FGF10 might separate vascular and bronchial SMC progenitors, respectively, from mesenchymal stem cells. Importantly, in the absence of RBPjκ in vivo, vSMCs still formed outside the Dermo1-Cre lineage, presumably to compensate for the loss of Notch signaling, obscuring a potential phenotype observed with GSI or antisense oligonucleotides in organ culture.

Because Dermo1-Cre is active in the mesenchyme and in the mesothelium, and transdifferentiation of mesothelial cells via EMT was described in the heart (Cai et al., 2008; Wada et al., 2003; Zhou et al., 2008), the gut (Wilm et al., 2005) and the liver (Ijpenberg et al., 2007), we used Wt1-Cre, R26YFP mice to determine whether vSMC were derived, in a Notch dependent manner, from the mesothelium. Although YFP-positive cells were observed in vascular endothelium and the mural wall, we concluded that the mesothelial lineage (Wt1, Dermo1-positive) did not contribute significantly to the vSMC population under normal conditions (see also Que et al., 2008). Vital dye pulse-chase experiments, constitutive Notch1 activation, and inhibitor studies in lung organ cultures identified a role for Notch signaling alongside TGFβ in mesothelial EMT. Thus, if mesothelial cells were involved in vSMC rescue, in vitro inhibition of Notch signaling would greatly reduce their ability to rescue. In addition, Notch activation enhanced migration of mesothelium-derived cells; this finding might have important implications for understanding the aggressive metastatic nature of malignant mesothelioma because elevated Notch signaling has been observed in malignant human mesothelial cells (Graziani et al., 2008).

Another potential source for vSMCs in Dermo1-Cre, RBPjκ^flox/flox mice is endothelial cells, which remained positive for Notch signaling and have been known to undergo EnMT in the lung (Arciniegas et al., 2007). However, due to the early lethality associated with endothelial-specific loss of Notch1, we could not demonstrate a contribution of EnMT to vSMC in Dermo1-Cre, Tie1-Cre, RBPjκ^flox/flox animals. The origin of the rescuing cells, therefore, remains speculative, awaiting creation of an endothelial-specific Flp-recombinase-based reporter. Importantly, Notch signaling is necessary to promote endocardial EnMT during formation of cardiac valves (Timmerman et al., 2004), and activation of Notch signaling is sufficient to induce EnMT in vitro (Noseda et al., 2004).

On the basis of our experiments, we can thus propose a speculative model explaining how GSI and Notch1 antisense oligonucleotides impact branching morphogenesis, yet two genetic models of global Notch loss [this study and Tsao et al. (Tsao et al., 2009)] did not reproduce the branching phenotype. Endothelial cells deficient in Notch signaling display branching and tube formation defects (Gridley, 2007; Hellstrom et al., 2007). In our opinion, the profound effects of GSI and antisense nucleotides on development and branching of lung anlagen growing in organ culture can thus be attributed to the disruption of vascular network formation, failed recruitment of vSMC, failed compensation by EnMT, or some combination of these. Indeed, DAPT-treated Tie2-GFP lung rudiment cultures show extensive migration and deficient vasculogenesis, whereas airway branching continues at an enhanced rate (Robert Mecham, Washington University, St Louis, MO, personal communication). We conclude that the negative effects of global Notch inhibition are most probably a reflection of losing the vascular endothelial network and/or its associated SMCs, which must therefore negatively regulate branching morphogenesis and positively contribute to maintaining distal fates.

Guseh and colleagues have reported that misexpression of a constitutively active Notch1 fragment with a mosaic SPC-Cre transgene causes mucous metaplasia of the airway and decreases the number of ciliated cells. In addition, this non-physiological and persistent activation generated alveolar cysts (Guseh et al., 2009). They interpreted this to suggest that Notch signaling suppresses alveolar development. In contrast to these observations, loss-of-function analysis of RBPjκ, a core component of canonical Notch signaling (this study), or Pofut1 (Tsao et al., 2009), which might be required for both canonical and non-canonical functions, did not support a role for physiological Notch signing within lung epithelial cells in regulating alveolar morphogenesis [(Morimoto and Kopan, 2009; Tsao et al., 2009) and data not shown]. RBPjκ- or Pofut1-null epithelium did not display alterations in branching morphology and contained normal alveoli. We concur with the conclusion that loss of Notch signaling leads to loss of Clara cells and provide evidence that Notch1 is involved. We describe a distal-to-proximal transition zone in which ciliated cells induce Notch activation in their neighbors, inhibiting them from selecting the same fate and permitting development of Clara cells. Finally, tracing the lineage of cells experiencing Notch1 activation indicates that these cells overwhelmingly assume the Clara cell fate (with a few ciliated cells labeled secondarily). Interestingly, Notch1 was activated again during epithelial regeneration following pharmacological injury. Notably, Notch1 activation was not involved in maintenance of the epithelial stem cells. Several previous reports show that Wnt signaling promotes proliferation of the airway epithelial stem cells early during regeneration (Reynolds et al., 2008; Zhang et al., 2008). Induction of Notch ligand by Wnt activity (Estrach et al., 2006) might trigger Notch activation during Clara cell regeneration.

Notch signaling was reported previously to be involved in a similar lateral inhibitory process, in which ciliated cells inhibit their neighbors from assuming the same fate. In the zebrafish pronephros, transporting epithelia and multiciliated cells (MCCs) form in a salt and pepper pattern (Liu et al., 2007). It has been shown that zebrafish Jagged2 expression in presumptive MCCs induced activation of zebrafish Notch3 in neighboring cells, blocking MCC fate and driving the alternative transporting epithelial cell fate. In addition, a Hes1-related protein was involved (Ma et al., 2007). In Xenopus, ciliated cells express Delta ligands to activate Notch signaling (and Hes-related proteins), inhibiting the selection of ciliated cells by neighboring epidermal cells (Deblandre et al., 1999).

Although Hes1 is a well-characterized Notch target gene in some cells, it can be regulated by other pathways (Nakayama et al., 2008; Yoshiura et al., 2007). Accordingly, it has been reported that Hes1 and Pofut1 regulate PNEC foci number as well as size (Ito et al., 2000; Tsao et al., 2009), yet RBPjκ-deficient mice retain Hes1 expression, and the number of PNECs was decreased only to the intermediate degree seen in Hes1 heterozygotes, not nulls. These data suggests that Hes1 expression might be controlled by upstream signals to which Pofut1 (but not RBPjκ) contributes. This could either imply involvement of other pathways (Nakayama et al., 2008; Yoshiura et al., 2007) or non-canonical Notch signaling.

To visualize N1IP-CRE activity, the tracheas of 2- to 3-week-old N1IP-CRE, R26R mice were filled with 0.2% GAD fixative (0.2% glutaraldehyde, 2 mM MgCl₂ in PBS) before isolation. After removal of the lungs from the thorax, both were further fixed in 0.2% GAD fixative for 1 hour at room temperature. The fixed lungs were washed (in 2 mM MgCl₂, 0.1% Tween-20, 0.05% dextrin in PBS) three times for 5 minutes. After washing, lungs were filled with X-gal solution (2 mM MgCl₂, 35 mM potassium ferrocyanide, 35 mM potassium ferricyanide, 1 mg/ml X-gal, 0.02% NP-40, 0.01% Na deoxycholate in PBS). The X-gal-filled lungs were submerged in X-gal solution and incubated for 12-24 hours at 4°C in the dark, then post-fixed with 4% paraformaldehyde overnight at 4°C.

β-galactosidase-expressing fetal lungs were dissected in ice-cold PBS, then fixed with 0.5% GAD fixative (0.5% glutaraldehyde, 2 mM MgCl₂, 0.1% Tween-20 in PBS) for 1 hour at 4°C. The fixed lungs were washed in solution three times for 5 minutes each before equilibration with 30% sucrose in PBS. After embedding in OCT compound (Sakura), frozen sections (5-6 mm) were generated and either stored at −20°C or incubated in X-gal solution for 3-6 hours at 37°C. The stained sections were counterstained with Nuclear Fast Red (Vector Laboratory).

Fetal lungs were dissected and fixed in 4% paraformaldehyde for 1 hour to overnight at 4°C, embedded in paraffin or OCT (for frozen sections) and sectioned at 6-7 μm. 4% formaldehyde was used for the detection of β-galactosidase. Sections were rehydrated and treated with 0.3% hydrogen peroxide in MeOH for 10 minutes before staining. The antibodies and conditions used for individual immunohistochemical analyses are described in supplementary material Table S1.

E14.5 lungs were collected from timed pregnant CD1 wild-type or N1IP::CRE, R26R mice and labeled with 40 μM CCFSE (5-(and-6)-carboxy-2,7-dichlorofluorescein diacetate, succinimidyl ester; Molecular Probes) in DMEM containing 10% FBS, 1 mM L-glutamine and 1 mM penicillin-streptomycin for 2 hours at 37°C, 5% CO₂. The CCFSE was prepared as a 20 mM stock solution in DMSO. In some experiments, samples were treated with 5 μM recombinant Tat-Cre protein for 5 hours at 37°C, 5% CO₂. Subsequently, the labeled lungs were washed and cultured on a filter (cell culture insert, 1.0 μm pore size P.E.T. track-etched membrane; Falcon) in medium without CCFSE or Tat-Cre for 48 hours at 37°C, 5% CO₂. TGFβ (2 μg/ml; R&D Systems), SD208 (5 mM; Calbiochem) or DAPT (1 mM; Calbiochem) was added to culture medium to appropriate final concentrations (described in each figure legend). To prepare frozen blocks, at least three whole lungs at end time points were fixed using fresh 4% formaldehyde for 1 hour at 4°C followed by equilibration with 30% sucrose in PBS. Sections (6 μm) were cut and mounted with Vectamount with DAPI (Vector Laboratories) or stained with anti-GFP-antibody (see supplementary material Table S1) for Tat-Cre experiments. Using a Carl Zeiss Axio imager Z1 microscope, the fluorescence of CCFSE was observed with the EGFP filter under 400× magnification. To quantify mesothelial EMT, ten photographs of GFP-positive cells were taken for each genotype under 400× view.

TAT-Cre recombinant protein was expressed and purified according to previously described protocols (Peitz et al., 2002). Briefly, Escherichia coli TUNER(DE3)pLacI (Novagen) containing pTriEx(Novagen)-His-TAT-NLS-Cre was cultured with LB medium containing 100 mg/ml ampicillin and 34 mg/ml chloramphenicol at 37°C with shaking until the OD₆₀₀ was 0.9-1.0. Expression was induced with 0.5 mM isopropylthiogalactoside for 4 hours. Cells were harvested, resuspended in phosphate buffer (50 mM Na₂HPO₄, 5 mM Tris pH 7.8, 500 mM NaCl) containing lysozyme (Sigma) and Benzonase (Novagen) for lysis. TAT-Cre protein was purified from the supernatant with Ni-NTA matrix (Qiagen). The matrix was washed extensively with phosphate buffer containing 20 mM imidizole and eluted with phosphate buffer containing 250 mM imidazole. Protein-containing fractions were pooled and dialyzed against 600 mM NaCl, 20 mM HEPES, and subsequently against 600 mM NaCl, 20 mM HEPES, 50% glycerol to concentrate.

RBPjκ^flox/flox mice were kindly provided by Tasuku Honjo, Kyoto University, Kyoto, Japan (Tanigaki et al., 2002). Dermo1-Cre mice were generated by David Ornitz at Washington University, St Louis, MO (Yu et al., 2003). FOXJ1-Cre mice were a kind gift from Michael Holtzman, Washington University, St Louis, MO (Zhang et al., 2007). These mice were maintained on the CD1 background. Dermo1-Cre mice were mated to RBPjκ^flox/flox mice to generate heterozygous Dermo1-Cre, RBPjκ^+/flox mice, which were then crossed to RBPjκ^flox/flox mice to create Dermo1-Cre, RBPjκ^flox/flox conditional knockout animals. FOXJ1-Cre, RBPjκ^flox/flox mice were generated in the same way.

The N1IP::CRE (Vooijs et al., 2007), Wt1-Cre (Wilm et al., 2005) and CCSP-rtTA, (Tet)7O-Cre (Perl et al., 2002) mice were described previously. SHH-Cre is commercially available at The Jackson Laboratory. All animal procedures were performed according to NIH guidelines and maintained in the animal facility under Washington University animal care regulations.

Embryonic lung mRNA was isolated using the RNeasy kit (Qiagen) following the manufacturer's instructions. cDNA was synthesized using the SuperScript II first-strand cDNA synthesis kit (Invitrogen). Quantitative RT-PCR was performed on an ABI 7500 machine using Power SYBR Green for Foxj1, CC10, Hes1 and Gapdh. Amplification and analyses were performed according to the manufacturer's instructions. All reactions were normalized to Gapdh. Results were plotted as relative expression compared with control, where control was scaled to 1.

The authors thank Adrian Shifren, David Ornitz, Rober Mecham and Scott Boyle (from Washington University, St Louis, MO) for careful reading of the manuscript. We also thank David Ornitz for the Dermo1-Cre mice, Tasuku Honjo for the conditional RBPjκ^flox/flox mice and Steven Brody for anti-FoxJ1 antibody. This work was supported by Washington University and by NIH grants P50 CA094056 (David Piwnica-Worms) and RO1 DK066408 (R.K.). M.M. was supported by a Toyobo Biotechnology Foundation Long-term Research Grant, the Japanese Society for the Promotion of Science and The Kanae Foundation for the Promotion of Medical Science. We wish to thank Patricia Gonzalez-DeWhitt (Eli Lilly and Company, Indianapolis, IN) for the generous gift of the rabbit anti-VLLS antibody. Deposited in PMC for release after 12 months.