Histone arginine methylation by Prmt5 is required for lung branching morphogenesis through repression of BMP signaling

Mammalian lung development consists of a series of remarkably elaborate branching events that drive the progression from simple lung buds to a stereotyped branching architecture to accomplish the vital function of gas exchange. In the mouse, the lung primordium arises from the ventral foregut around embryonic day (E)9.5 (Minoo et al., 1999; Wells and Melton, 1999). Once the primary lung buds have formed, they extend into the surrounding mesenchyme and begin the process of branching morphogenesis. Between E9.5 and E16.5, the primary buds generate a complex tree-like structure ending in thousands of terminal tubules through approximately 5000 times of stereotyped branching (Herriges and Morrisey, 2014; Metzger et al., 2008). The formation of a tree-like respiratory airway system is accompanied by the differentiation of epithelial cell types along the proximal-distal (P-D) axis (Domyan and Sun, 2011). The SRY-box-containing gene 2 (Sox2)-expressing proximal epithelial cells give rise to major conductive airways that are lined with ciliated, basal, neuroendocrine and secretory cells (Gontan et al., 2008; Morrisey and Hogan, 2010), whereas the Sox9/Id2-expressed distal region gives rise to the smaller bronchioles and gas exchange units that are lined with type I and type II alveolar epithelial cells (Jen et al., 1996; Liu and Hogan, 2002; Morrisey and Hogan, 2010; Rawlins, 2008).

Significant efforts have been made to identify the cellular and molecular mechanism controlling branching morphogenesis. Signaling factors, including Bmp4, Shh, Fgf10 and Wnt, play a key role in this process (Herriges et al., 2012; Herriges and Morrisey, 2014; Rock and Hogan, 2011). During branching, Fgf10 is dynamically expressed in mesenchymal clustered cells, inducing new bud formation and activation of Bmp4 and Shh pathways in the adjacent epithelium, which are responsible for the directional outgrowth of lung buds (Bellusci et al., 1997b; Lebeche et al., 1999). Both Bmp4 and Shh negatively inhibit Fgf10 activity to form a cleft, leading to bifurcation and two new bud points to finish a branching cycle (Bellusci et al., 1997a; Hyatt et al., 2002; Weaver et al., 2000). Inactivation of either Fgf10 or Shh signaling leads to complete abrogation of branching morphogenesis (Min et al., 1998; Pepicelli et al., 1998; Sekine et al., 1999), whereas inactivation of Bmp4 results in defects in branching morphogenesis and P-D airway patterning of the lung (Bellusci et al., 1996; Weaver et al., 1999). In contrast to the well-established roles of signaling factors during branching morphogenesis, relatively little is known about how these factors are regulated at transcriptional levels.

Protein arginine methyltransferase 5 (Prmt5) is a chromatin-modifying enzyme capable of catalyzing the symmetrical dimethylation of arginine residues on histone and non-histone substrates, and it plays a pivotal role in the regulation of diverse cellular processes ranging from transcription and RNA processing to signaling transduction, cell differentiation, apoptosis and organelle biosynthesis (Biggar and Li, 2015; Li et al., 2016; Stopa et al., 2015; Wang et al., 2007; Yue et al., 2013; Zhang et al., 2011; Zhou et al., 2010). Prmt5 is generally associated with transcriptional repression through methylation of histones H2A/H4R3 and H3R8 (Ancelin et al., 2006; Di Lorenzo and Bedford, 2011; Lee and Bedford, 2002; Tae et al., 2011; Tee et al., 2010; Zhang et al., 2015). In mice, Prmt5 is essential for maintaining the pluripotency of embryonic stem cells. Deletion of Prmt5 results in the downregulation of pluripotency transcription factors and causes embryonic lethality before implantation (Tee et al., 2010). Specifically deletion of Prmt5 in neural stem/progenitor cells leads to postnatal death in mice through reduced methylation of Sm proteins (Bezzi et al., 2013). Prmt5 is also expressed in primordial germ cell-like cells (PGCs) and directs histone arginine methylation in mouse germ cells. Inactivation of Prmt5 in PGCs results in germ cell loss during spermatogenesis (Wang et al., 2015a,b). Whether Prmt5 plays a role in lung development remains unclear.

In the present study, we found that Prmt5 is highly expressed in the lung during active branching morphogenesis. Conditional inactivation of Prmt5 in lung epithelium results in halted branching morphogenesis and altered cell differentiation. The branching morphogenesis defects are associated with massive apoptosis and inhibited proliferation of lung epithelial cells, as a consequence of elevated and spatially disorganized expression of Bmp4, which is a direct target of Prmt5. These data demonstrate that lung branching morphogenesis is regulated by Prmt5 through transcriptional repression of BMP signaling, providing an intrinsic mechanism for histone methylation during embryonic lung development.

To address whether Prmt5 plays a role in lung development, we examined its spatial and temporal expression patterns by immunostaining of embryonic lung sections. We found that Prmt5 was expressed broadly throughout the lung epithelium and mesenchyme from E12.5 to E18.5, with both nuclear and cytoplasmic localization (Fig. 1A). Interestingly, Prmt5 expression was notably decreased from E14.5 to E16.5 (Fig. 1A); this decrease was further verified at the transcript level by qRT-PCR (Fig. 1B). These data imply that Prmt5 is required during embryonic lung development.

To further assess the role of Prmt5 during embryonic lung development, we conditionally inactivated Prmt5 in lung epithelium using a previously reported sonic hedgehog Cre line (Harfe et al., 2004; Harris et al., 2006). We bred the Prmt5^fl/fl allele with Shh^cre to generate Prmt5^fl/fl;Shh^cre/+ mice (herein referred to as Shh;Prmt5). Specifically Cre expression in lung epithelium was detected in an mTmG reporter mouse line at E12.5 (Fig. S1A); the efficiency of Cre-mediated epithelial Prmt5 deletion was examined by qRT-PCR, western blotting and immunostaining (Fig. 1C; Fig. S1B-E). Prmt5 expression and its catalyzed substrate H4R3sme2 were specifically absent in lung epithelium at E12.5, and their mesenchyme expression and H4R3me2 expression were not affected (Fig. 1C; Fig. S1D,E). Collectively, these experiments suggest that Prmt5 is specifically inactivated in lung epithelium.

Shh;Prmt5 mutants survived during embryogenesis but all died at birth from respiratory distress with a cyanotic phenotype (Fig. 1D; Table S1). Examination of lungs from E18.5 Shh;Prmt5 embryos revealed a disorganized architecture with the presence of multiple large cysts instead of well-formed lung lobes (Fig. 1E). A significant decrease in lung-to-body weight ratios was also observed (Fig. 1F; Fig. S2A), and an approximately threefold reduction in epithelial cells was consistently observed by FACS analysis in the Shh;Prmt5 mice (Fig. S1B). Lung cysts were easily seen by E14.5 in whole mounts of lungs isolated from Shh;Prmt5 mice, and these cysts became more apparent by E16.5 (Fig. S2B). However, an appropriate number of lung lobes was still formed in the Shh;Prmt5 mice (Fig. S2B). Hematoxylin and eosin (H&E)-stained lung sections revealed that these cyst-like structures formed as early as E12.5 and expanded in later development, resulting in several larger and rounder open spaces in Shh;Prmt5 lungs (Fig. S2C). These experiments suggest that the epithelial Prmt5 is essential for embryonic lung development to form a functional lung.

To evaluate whether Prmt5 regulates lung branching, we dissected lungs from control and Shh;Prmt5 mice and performed whole-mount E-cadherin immunostaining. Lungs from Shh;Prmt5 mice were not distinct from control at E11.5 (Fig. 2A). However, from E12.5 onwards, whereas stereotyped branching was actively underway in the control lungs, the Shh;Prmt5 lungs stopped new branch generation, and unbranched lung buds enlarged during development and resulted in a multi-cyst structure (Fig. 2A). Consistently, we observed significantly decreased branches with cystic structures in Shh;Prmt5 mice at E12.5 (Fig. 2B,C). We then examined the branching process by in vitro lung explant culture. Branching morphogenesis was halted, and no new branches were generated in Shh;Prmt5 lungs, whereas branching occurred normally in control mice (Fig. 2B,C). Therefore, these results demonstrate that the epithelial Prmt5 is essential for lung branching morphogenesis.

We also used an inducible gene knockout approach (Sauer, 1998; Tichelaar et al., 2000) to selectively abrogate Prmt5 activity at different lung developmental stages. Prmt5^fl/fl;SPC-rtTA;TetO-Cre (herein referred to as SPC;Prmt5) triple-transgenic mice were generated by maintaining pregnant dams on doxycycline from E6.5 onwards. Immunohistochemical analysis showed specifically Cre expression and Prmt5 deletion in the epithelium of E14.5 SPC;Prmt5 lungs (Fig. S3A). Most SPC;Prmt5 mice died of respiratory distress shortly after birth. However, some SPC;Prmt5 mice (13 of 83 from 10 litters, SPC;Prmt5 mice with doxycycline vs. 21 of 81 from 9 litters, SPC;Prmt5 mice without doxycycline) survived to adulthood with multiple large cysts at the distal epithelial branch (Fig. S3B). Histologic analysis showed that the SPC;Prmt5 mice exhibited cyst-like structures in the distal regions of the lung at E14.5 and E18.5 (Fig. S3C), similar to what has been observed in Shh;Prmt5 mice. Interestingly, enlarged airspaces were observed in mature lungs of SPC;Prmt5 mice that were maintained on doxycycline from E12.5 onwards, whereas no morphologic abnormalities were observed in the lungs of SPC;Prmt5 mice that were maintained on doxycycline from E14.5 onwards (Fig. S3C). Quantitative analysis of in vitro cultured lungs showed a significant reduction in branching in SPC;Prmt5 mice compared with controls, whereas no significant differences were observed at the start of culture, E12.5 (Fig. S3D,E). Collectively, these data further demonstrate that epithelial Prmt5 is necessary for lung branching morphogenesis.

To assess further the molecular effects of Prmt5 inactivation on lung branching morphogenesis, we performed whole-mount immunofluorescence co-staining to examine the expression of a distal marker Sox9 and a proximal marker Sox2 (Gontan et al., 2008; Perl et al., 2005; Que et al., 2007). In Shh;Prmt5 mice at E11.5, Sox2 and Sox9 expressions were comparable to controls (Fig. 3A). However, although the Sox2⁺ region aggressively extended accompanied by branching morphogenesis in later developmental stages, it remained largely unchanged in Shh;Prmt5 lungs (Fig. 3A,B). Conversely, although Sox9 was expressed in the distal epithelium in the control lung, it was expanded to most of the lung epithelium in the Shh;Prmt5 mice (Fig. 3A,B). Consistently, clearly reduced Sox2 and expanded Sox9 expression areas were detected in the E14.5 Shh;Prmt5 lung sections (Fig. 3C). These changes were further supported by qRT-PCR results from FACS-purified lung epithelium showing decreased Sox2 and increased Sox9 mRNA expressions in Shh;Prmt5 mutant lungs compared with controls (Fig. 3D). Collectively, these observations reveal that the dilated branch tips express Sox9 and the few branch stalks express Sox2 in Shh;Prmt5 lungs.

Next, we investigated whether loss of Prmt5 affects epithelial cell differentiation. All analyzed early-airway epithelial markers, including Sox2, Sox9, Nkx2.1, Sftpc, SSEA1 and E-cadherin, expressed in the E12.5 Shh;Prmt5 lungs (Fig. 4A). Expression of α-smooth muscle actin (α-SMA), a marker of airway smooth muscle differentiation, appeared to be well maintained around conducting airways in Shh;Prmt5 lungs at E12.5 and E14.5 (Fig. 4A). The generation of distal epithelial cell types, including alveolar type II cells (Abca3⁺, Sftpc⁺, Sftpd⁺) and type I (Ager⁺, Aqp5⁺) cells, were observed in cystic epithelium of E18.5 Shh;Prmt5 mice (Fig. 4B). However, neither ciliated (Ac-tub⁺) nor neuroendocrine cells (CGRP⁺) were detected in the proximal airways of Shh;Prmt5 lungs at E18.5 (Fig. 4B). These observations indicate that disorganization of the lung structure is associated with defective epithelial specification in the proximal airways of Shh;Prmt5 mice.

To address the mechanism of lung defects in Shh;Prmt5 mice, microarray and RNA-sequencing (RNA-seq) analyses were performed. We found that the expressions of some cell cycle regulators were changed in Prmt5 mutant lungs (Fig. S4A, Table S2). p21 (Cdkn1a), a known direct target of Prmt5 (Zhang et al., 2015), was also upregulated in Prmt5 mutant mice (Fig. S4A, Table S2). We then determined whether cell apoptosis and proliferation were affected in Shh;Prmt5 lungs. By TUNEL assay and cleaved caspase-3 (CC3) immunostaining, we observed significantly increased apoptosis in the epithelium of Shh;Prmt5 lungs (Fig. 5A,C; Fig. S4B). Interestingly, this increase in apoptosis seemed to spatially restrict in particular epithelial regions. We then determined the percentage of proximal (Sox2⁺ and CC3⁺) and distal (Sox9⁺ and CC3⁺) apoptosis in the epithelium of control and Shh;Prmt5 lungs. We found that proximal epithelial apoptosis was significantly higher than distal epithelial apoptosis in Shh;Prmt5 lungs at both E11.5 (Fig. S4C,D, Movie 1) and E12.5 (Fig. 5B,D). We then analyzed epithelium proliferation by Ki67 immunostaining. The numbers of Ki67-positive cells were significantly decreased in total epithelium of Shh;Prmt5 lungs (Fig. 5A,C). The reduced proliferation of lung epithelium was also demonstrated by pH3 immunostaining, a marker for cells in late G2 and mitosis (Fig. 5A,C). No significant changes were observed in proliferation between proximal and distal epithelium in Shh;Prmt5 lung tissues (Fig. 5B,D; Fig. S4C,D), and both apoptosis and proliferation were not significantly affected in the mesenchyme of Shh;Prmt5 lungs (Fig. 5C; Fig. S4D). Thus, Prmt5 controls epithelial apoptosis and proliferation during embryonic lung development.

Previous studies have shown that the cell cycle inhibitor p21 is a Prmt5 substrate in regulating cell proliferation and apoptosis in several organs (Hu et al., 2015; Zhang et al., 2015), raising the possibility that p21 regulates apoptosis and proliferation in the lung epithelium. Indeed, we detected an increased p21 mRNA expression in Prmt5 mutant lungs by microarray, RNA-seq and qRT-PCR analysis (Table S2, Figs S4A and S5A). We then expected that knockout p21 in the Shh;Prmt5 background would rescue lung developmental defects in Shh;Prmt5 mice. However, branching defects and changes of epithelial apoptosis and proliferation were not affected by p21 inactivation (Fig. S5B-D), suggesting that the function of Prmt5 in regulating cell apoptosis and proliferation during embryonic lung development is p21-independent.

To assess further the molecular mechanism that underlies branching defects in Prmt5 mutants, we examined the expression of multiple pathways known to regulate lung branching morphogenesis, including Bmp4, Shh, Wnt and Fgf10 (Herriges and Morrisey, 2014). In situ hybridization showed that Bmp4 expression was elevated and expanded in the epithelium of Shh;Prmt5 lungs instead of being expressed at the distal tip of the branching airways, as observed in the control lungs (Fig. 6A) and as has been previously reported (Alanis et al., 2014; Bellusci et al., 1996; Weaver et al., 1999). Consistent with the in situ results, significantly increased Bmp4 expression was also detected by qRT-PCR (Fig. 6B). In contrast, the expression of other checked genes, including Fgf10, Shh, Wnt2, Wnt7b, Fgfr2b and Axin2, were not significantly altered (Fig. 6A,B). Many BMP signaling-related genes were also differentially expressed in Prmt5 mutant, in both microarray and RNA-seq analyses (Table S2). Several described downstream effectors of canonical BMP signaling (Holleville et al., 2003; Hollnagel et al., 1999; Shen et al., 2014) were also significantly upregulated in Shh;Prmt5 lungs, whereas the expression of other BMP family members was not affected (Fig. 6C). Moreover, the phosphorylated Smad1/5/9, which is a direct readout of canonical BMP signaling (Moustakas and Heldin, 2009), was strongly expressed in the pulmonary epithelium and the mesenchyme surrounding the epithelium in Shh;Prmt5 lungs from E11.5 to E14.5 (Fig. 6D). In addition, western blot analysis detected increased p-Smad1/5/9 expression (Fig. 6E). These results demonstrate that canonical BMP-Smad1/5/9 signaling is activated in Shh;Prmt5 mice. We further determined whether Bmp4 is a direct target of Prmt5 by chromatin immunoprecipitation (ChIP) assays, using Prmt5 and H4R3sme2 antibodies. We observed that Prmt5 and H4R3sme2 were able to bind to the promoter region of Bmp4 at E12.5 (Fig. 6F-H). In contrast, Prmt5 and H4R3sme2 binding was not observed in an unrelated intergenic region and at 2 megabases upstream of the Bmp4 locus (Fig. 6F-H). Together, the results of these experiments suggest that Prmt5 and H4R3sme2 directly regulate Bmp4 transcription in the developing lung.

Previous reports suggested that BMP signaling plays an important role in both lung branching morphogenesis and P-D epithelium specification (Bellusci et al., 1996; Wang et al., 2013b; Weaver et al., 1999). To determine whether the activation of BMP-Smad1/5/9 signaling in the Prmt5 mutant contributes to the lung defects described above, we inhibited BMP signaling by using the BMP antagonist Noggin (Geng et al., 2011) in an in vitro lung culture system. Administration of Noggin in culture was effective in reducing the expression of the canonical BMP signaling pathway readout p-Smad1/5/9, as determined by western blotting and immunostaining (Fig. S6A,B). Furthermore, Shh;Prmt5 lungs cultured in control medium retained their dilated epithelial tip phenotype and evaluated p-Smad1/5/9 level, recapitulating the in vivo phenotype (Fig. 7A,B; Fig. S6B), suggesting that this in vitro lung culture system would allow us to examine how BMP signaling contributes to the defects in Shh;Prmt5 lungs. Treatment of Shh;Prmt5 lung explants with Noggin led to a significantly increased tip number, accompanied by smaller epithelial tip size, with morphology similar to that observed in control lungs (Fig. 7A,B). Furthermore, in Noggin-treated Shh;Prmt5 lungs, the Sox2 expression domain extended farther from the point where the main bronchi split than it did in Noggin-free culture conditions, and more Sox2-expressing secondary branches were also detected (Fig. 7A,B). On the cellular level, epithelial apoptosis and proliferation were rescued significantly in Shh;Prmt5 lung explants treated with Noggin (Fig. 7C,D). However, no ciliated cells (Ac-tub⁺) were observed in the proximal regions of Noggin-treated Shh;Prmt5 lungs (Fig. S6C), suggesting that other Prmt5 effectors instead of BMP signaling were involved for proximal epithelium specification. In summary, these studies suggest that the activated canonical BMP signaling significantly contributes to the lung branching morphogenesis defects in the Shh;Prmt5 lungs.

A previous study identified Hdac1/2, an epigenetic regulator, as a repressor of Bmp4 expression through H3K9ac during embryonic lung development (Wang et al., 2013b), raising the possibility that Prmt5 and Hdac1/2 form a complex and that both H3K9ac and H4R3sme2 are required for Bmp4 repression in Prmt5 mutant. To address this, we first examined the expression of Hdac1/2 and its downstream target genes by qRT-PCR. Except for p21, no significant differences were found between control and Shh;Prmt5 lungs (Fig. S7A). Western blotting confirmed that Hdac1/2 and H3K9ac expressions were not affected by Prmt5 knockout (Fig. S7B). In co-immunoprecipitation experiments, no detectable interactions between Prmt5 and Hdac1/2 were observed (Fig. S7C). Moreover, we detected a specific binding of Prmt5/H4R3sme2 with H4R3sme2 but not for H3K9ac (Fig. S7D), a histone mark that was modified by Hdac1/2 to repress Bmp4 expression (Wang et al., 2013b). In summary, these experiments indicate that Prmt5 and Hdac1/2 function independently to repress Bmp4 expression during lung development.

In the present study, we provide convincing evidence that protein arginine methyltransferase Prmt5 is required for lung branching morphogenesis through repression of BMP signaling. We propose that activated canonical BMP-Smad1/5/9 signaling resulting in a halted branch cycle, accompanied by massive apoptosis and reduced proliferation in lung epithelium of Prmt5 mutants, which eventually forms a lung with fewer and large cystic tips (Fig. 8). We identified an intriguing epigenetic mechanism whereby Prmt5-mediated H4R3sme2 transcriptionally represses Bmp4 expression to regulate lung branching morphogenesis.

Our results indicate an important role for histone methylation in lung branching morphogenesis. First, experiments on both Shh;Prmt5 and SPC;Prmt5 mice demonstrated that epithelial Prmt5 is required during lung branching morphogenesis. Second, the temporal expression pattern of Prmt5 is correlated with its crucial role in branching morphogenesis, as higher Prmt5 expression was detected before E16.5, which is when the lung buds undergo active branching to generate a tree-like network of airways (Herriges and Morrisey, 2014; Metzger et al., 2008). Finally, the critical role of Prmt5 in branching morphogenesis is further supported by the results from SPC;Prmt5 mice, in which deletion of Prmt5 before or after E14.5 resulted in cyst-like lung structures or no morphologic abnormalities, respectively. To our knowledge, this is the first study to show that lung branching morphogenesis is regulated by a protein methylation writer. Before our present investigation, previous studies suggested that the chromatin-remodeling factor Hdac1/2 represses Bmp4 expression through H3K9 acetylation (Wang et al., 2013b). Of note, our experiments indicate that these two histone modifications, H4R3 dimethylation and H3K9 acetylation, function independently for Bmp4 repression during early lung development.

We identified Bmp4 as a direct target of transcriptional repression by Prmt5 through H4R3 symmetric dimethylation (H4R3sme2). This finding is consistent with the concept that the histone mark H4R3sme2 suppresses gene transcription (Tae et al., 2011; Wang et al., 2007; Xu et al., 2010; Zhang et al., 2011). Bmp4 is an evolutionarily conserved, secreted molecule that is implicated in lung initiation (Domyan et al., 2011) and embryonic development (Hogan, 1996). Excessive BMP signaling, achieved by either overexpressing Bmp4 in lung epithelium (Bellusci et al., 1996) or knocking out the endogenous Bmp4 antagonist Gremlin (Michos et al., 2004), resulted in abnormal fetal lung formation. Our study provides mechanism evidence on how Bmp4 is regulated at the transcriptional level. However, our discovery of elevated canonical BMP-Smad1/5/9 signaling in Shh;Prmt5 lungs is complementary to the finding of a prior study by Norrie et al., which found an increased Bmp4 expression by loss of Prmt5 in limb progenitor cells (Norrie et al., 2016). The authors reported that a non-canonical Bmp4-p38 pathway, instead of a canonical BMP-Smad1/5/9 pathway, plays a dominant role (Norrie et al., 2016). We speculate that the disparity in these two mechanisms could be due to tissue- or cell-specific effects, given that Prmt5 was reported to play distinct roles in different organ contexts (Bezzi et al., 2013; Norrie et al., 2016; Wang et al., 2015b; Zhang et al., 2015). Another possible explanation is that Prmt5 has a more profound role in development through regulating canonical BMP-Smad1/5/9 signaling, as the authors stated that there might be a mild or transient increased canonical Bmp4 activity that was not detectable in their experiments (Norrie et al., 2016).

Our study indicates that BMP signaling regulates epithelial apoptosis and proliferation during embryonic lung development. Specifically, elevated BMP-Smad1/5/9 signaling results in massive apoptosis and inhibited proliferation of lung epithelium, which lowers the number of epithelial cells and could contribute to the halted branching formation and few Sox2⁺ stalks in the Prmt5 mutant. In agreement with this, significantly increased epithelial apoptosis was observed in the shortened Sox2⁺ proximal regions of Shh;Prmt5 lungs. Notably, our discoveries of partial rescued Sox2⁺ stalks and no rescue of cell differentiation defects in Noggin-treated Prmt5 lungs are consistent with the proposed requirement for BMP signaling, which is required for P-D cell fate specification of lung endoderm (Weaver et al., 1999) but not for Clara/ciliated cell differentiation (Tadokoro et al., 2016). However, we think that there are several possible, non-exclusive explanations for the partial rescued Sox2⁺ regions in the Prmt5 mutant. One is the cell non-autonomous effects of Noggin in the in vitro cultured explants (Geng et al., 2011; Lüdtke et al., 2016). Alternatively, other lung-expressed BMP family ligands, Bmp5 and Bmp7 (Bellusci et al., 1996; King et al., 1994), may also contribute to the partial rescue phenotype. Although the transcriptional levels of Bmp5 and Bmp7 were not changed in Shh;Prmt5 mice, we cannot exclude the possibility that signaling by these BMP ligands is blocked by Noggin treatment. Additionally, genes related to cell junction/adhesion and some transcription factors in regulation of lung development were consistently altered in Prmt5 mutant lungs, as demonstrated by microarray and RNA-seq data, suggesting that these factors were also involved. Further studies are needed to address and evaluate these possibilities.

Many organs of higher organisms are heavily branched structures that arise by an apparently similar process of branching morphogenesis, which is intricately regulated by a conserved network of signaling factors (Affolter et al., 2009). Fgf10 signaling is reported to direct the outgrowth of epithelium in lung (Bellusci et al., 1997b), prostate (Donjacour et al., 2003; Wilhelm and Koopman, 2006), salivary gland (Makarenkova et al., 2009) and pancreas (Bhushan et al., 2001; Pulkkinen et al., 2003), whereas Shh signaling is upregulated by Fgf10 in the lung (Abler et al., 2009; Herriges et al., 2015) and prostate (Pu et al., 2004) during branching. Bmp4 signaling inhibits branching morphogenesis in lung (Bellusci et al., 1996; Wang et al., 2013b), kidney (Chi et al., 2011), prostate (Keil et al., 2015; Lamm et al., 2001) and pancreas (Ahnfelt-Rønne et al., 2010). Whether the Prmt5-BMP-Smad1/5/9 regulatory mechanism identified in this study is conserved during branching morphogenesis in other organs and species needs to be addressed in studies on tissue-specific Prmt5 knockout mice.

The Prmt5 knockout first mouse line, Prmt5^{tm2a(EUCOMM)Wtsi}, was obtained from the European Conditional Mouse Mutagenesis Program (EUCOMM; http://www.knockoutmouse.org). Prmt5^fl/fl allele was generated as previously described (Wang et al., 2015a,b). The lung epithelium-specific Prmt5 knockout mice (Shh;Prmt5) were generated by crossing Prmt5^fl/fl mice with Shh^cre/+ mice (Harfe et al., 2004; Harris et al., 2006), which were kindly gifted by Prof. Lin Xinhua (Fudan University). The inducible lung epithelial-specific Prmt5 knockout mice (SPC;Prmt5) were generated by crossing Prmt5^fl/fl mice with SPC-rtTA mice (Tichelaar et al., 2000) and TetO-Cre mice (Sauer, 1998). Administration of doxycycline started from different gestation stages to the endpoint of the experiment by feeding the pregnant mice with 625 mg/kg of doxycycline-containing food. Rosa-mTmG reporter mice and p21 knockout mice were kindly gifted by Prof. John Speakman [Institute of Genetics and Developmental Biology-Chinese Academy of Sciences (IGDB-CAS)] and Prof. Yuan Weiping (Chinese Academy of Medical Sciences), respectively. The embryos used in these experiments were harvested from time-mated females; noon of the day on which a vaginal plug was detected was considered E0.5. All mice used in this study were bred in the C57BL/6 strain background, and genotyped by genomic DNA PCR using primers listed in Table S3. Mice were housed in a specific pathogen-free condition with a 12 h light/12 h dark cycle in a temperature- and humidity-controlled environment. All animal work was performed in accordance with the guidelines set by the animal welfare committees of the IGDB-CAS.

Lungs from control and Prmt5 mutant embryos were harvested at either E11.5 or E12.5 as indicated. Lungs were placed on a Nuclepore track-etched membrane (Whatman, 110614) and cultured at the air/liquid interface with DMEM-F12 (Invitrogen, C11330500BT) and 10% fetal bovine serum (FBS, Invitrogen, 16000-044). Noggin (Excell, CB055; 500 ng/ml) was added to inhibit BMP signaling in the Noggin rescue experiment. Lungs were cultured in an incubator at 37°C and 5% CO₂ for the indicated time before harvest and analysis. All lung explant culture experiments were repeated three to five times with embryos from at least three pregnant mothers.

Immunofluorescence staining was performed following a previously published protocol (Liu et al., 2017). Briefly, 8-μm-thick frozen sections were pretreated with 0.5% Triton X-100 in PBS for 15 min at room temperature. After blocking, sections were incubated with primary antibodies overnight at 4°C followed by 2 h of incubation with appropriate secondary antibodies and counterstained with DAPI. The antibodies used are listed in Table S4. Images were acquired using a Leica TCS SP8 confocal microscope and a Nikon Ti-E&N-STORM super resolution microscope system. For quantification, images were obtained from a minimum of five mice per genotype. Cells were counted in at least three different areas of each image for each mouse. Three-dimensional reconstitution images were obtained using Imaris software (Bitplane) and assembled using reconstruction software (Fiala, 2005) from z-stacks of sections.

The p-Smad1/5/9 staining was performed with the Tyramide Signal Amplification (TSA) system (Perkin Elmer LAS, NEL702001KT) following a previously described protocol (Mou et al., 2016). Briefly, 8-μm-thick frozen sections were pretreated with 0.5% Triton X-100 in PBS for 15 min, and endogenous peroxidases were blocked by incubating with 3% H₂O₂ for 30 min. The slides were incubated with p-Smad1/5/9 antibody at 4°C overnight. The next day, the slides were incubated with biotin-conjugated donkey anti-rabbit IgG in 5% BSA/0.2% PBS-Triton X-100 for 2 h. After washing with 0.2% PBS-Triton X-100, the slides were incubated with ABC Elite reagent (Vector, PK-6100) for 2 h and then incubated with TSA working solution for 15 min.

Whole-mount staining was performed following a previously published protocol (Metzger et al., 2008; Tang et al., 2011). Briefly, lungs isolated from indicated developmental stages were fixed in 4% paraformaldehyde for 1 h at 4°C and then dehydrated with methanol. Lungs were incubated with E-cadherin or Sox2/Sox9 antibodies (Table S4) overnight at 4°C, followed by 2 h of incubation with appropriate secondary antibodies. Then, the stained lungs were visualized by a Leica MZ16F stereomicroscope.

For histologic analysis, lungs isolated from indicated developmental stages were fixed in 4% (wt/vol) paraformaldehyde (Sigma-Aldrich, P6148) overnight at 4°C and then processed to paraffin (Thermo, 8330) embedding and sectioned at 5 µm. For immunohistochemical staining, sections were deparaffinized and rehydrated. Antigen retrieval was performed using the microwave method. After blocking, sections were incubated with diluted primary antibodies (Table S4) in 0.2% PBS-Triton X-100 overnight at 4°C followed by 1 h of incubation with selected secondary antibodies and counterstained with hematoxylin. H&E staining was performed as described previously (Ying et al., 2015).

RNA was prepared using TRIzol reagent (Ambion, 15596) and reverse transcribed with a FastQuant RT Kit (TIANGEN, KR106) according to the manufacturer's protocol. PCR was quantified using SYBR Green in an Agilent Technologies StrataGene Mx3000P system. At least three biological replicates were performed per genotype and three technical replicates were performed for every independent experiment. The mRNA levels were normalized to internal β-actin expression. The primers used are listed in Table S3.

For qRT-PCR on purified mGFP⁺ cells, Prmt5^fl/fl;Shh^cre/+;mTmG embryonic lungs were homogenized, and mGFP⁺ and mTomato⁺ cell populations were sorted by flow cytometry (BD FACS Aria II).

In situ hybridization on 15-µm frozen sections was performed essentially as described previously (Wang et al., 2013a; Zhang et al., 2014). For each marker, at least three independent specimens were analyzed. TUNEL assay was assessed on 8-μm-thick frozen sections by using an In Situ Cell Death Detection Kit (Roche, 11684795910), according to the manufacturer's instructions.

For microarray analysis, RNA was isolated from E14.5 lungs of three control and three SPC;Prmt5 embryos. Microarray analysis was performed according to the protocol described previously (Zheng et al., 2013).

For RNA sequencing, total RNA was extracted from E14.5 control and Shh;Prmt5 lungs. For each sample, seven lungs were collected and pooled. Sequencing libraries were generated using a NEBNext^® Ultra™ RNA Library Prep Kit and sequenced on an Illumina HiSeq platform. The 125 bp/150 bp paired-end reads were generated and reads were aligned to the mouse reference genome (mm10) using TopHat v. 2.0.12. HTSeq v. 0.6.1 was used to count the read numbers, and then the fragments per kilobase million (FPKM) value of each gene was calculated based on the length of the gene and read count mapped to this gene. Differential expression analysis was performed using the DESeq R package (1.18.0). The full-sequence dataset was deposited in the GEO database (accession no. GSE108934).

Approximately 20 lungs from E12.5 wild-type mice were isolated and pooled in a ChIP assay as described previously (Ying et al., 2015). Primers for quantitative PCR detection of Bmp4 chromatin regions are listed in Table S3. Western blotting was performed as described previously (Ying et al., 2015).

All experiments were repeated at least three times, and all data are presented as means±s.e.m. Statistical calculations were analyzed by GraphPad Prism version 6 (GraphPad Software, San Diego, CA). A two-tailed unpaired Student's t-test was used for the comparison between two experimental groups. For experiments with more than two groups, one-way ANOVA was performed with appropriate multiple comparisons as described in the figure legends. All statistics are representative of biological replicates. *P<0.05, **P<0.01, n.s., not significant.

We thank Shuangbo Kong and Haibin Wang for their input in the in situ hybridization experiment.