A genetic screen reveals FREM2 as a mediator of pulmonary hypoplasia

To identify regulators of pulmonary hypoplasia (PH), we conducted a large-scale forward genetic screen in mice using N-ethyl-N-nitrosourea (ENU) mutagenesis. One of the recessive mutant pups identified in this screen exhibits hemorrhage in regions around the eyes, along with blood-filled blisters and digit formation defects in the hindlimbs (Fig. 1a). These mutant pups also display smaller lungs with partial fusion among the right cranial (RCr) lobe, the right middle (RMd) lobe and the right caudal (RCd) lobe (Fig. 1b-f). The lung edge in mutant pups appear thickened and more rounded compared with that in the controls. Mutant animals are born at a ratio slightly lower than the Mendelian ratio, indicating that this mutation may cause embryonic lethality for some mice. However, most mutant pups die within 24 h of birth. To identify the phenotype-causing mutation, we performed whole-exome sequencing of G4 genomic DNA samples and identified Frem2, which encodes an ECM protein, as a candidate gene (Fig. 1g). Next, we carried out genetic linkage analysis by genotyping 156 G4, G5 and G6 mutant animals and found complete linkage between the lung phenotypes and the Frem2  allele (Fig. 1h). The identified allele carries a missense mutation (c.7229 G > A) in a codon corresponding to a highly conserved amino acid that is predicted to cause an amino acid substitution in the transmembrane domain (p.C2410Y) (Fig. 1i). We then carried out a complementation test by crossing mice carrying the ENU-induced Frem2 allele (Frem2 ) with mice carrying a Frem2 deletion allele (Supplementary Fig. 1a, b), and found that complementation did not occur in Frem2 double-heterozygotes (Fig. 1j-m), indicating that loss of Frem2 function is likely to be responsible for the lung phenotypes observed in Frem2  mice. To further investigate the role for Frem2 in lung development, we analyzed lung formation in Frem2 mice. Frem2 mice exhibited reduced Frem2 mRNA levels (Supplementary Fig. 2a) and smaller lungs (Supplementary Fig. 2b, c) with partial fusion of the right lung lobes (Supplementary Fig. 2 d), and most of these pups die within 24 h of birth. These phenotypes were similar to those of Frem2  animals. These mutant pups also show hemorrhage around the eyes, blood-filled blisters in the hindlimbs and syndactyly (Supplementary Fig. 2e).

PH may be secondary to congenital diaphragmatic hernia (CDH). CDH is a developmental defect characterized by incomplete closure of the diaphragm and herniation of fetal abdominal organs such as the stomach, intestines and/or the liver into the thoracic cavity that causes pulmonary hypoplasia, postnatal pulmonary hypertension due to vascular remodeling and cardiac dysfunction. Frem2 (Frem2) mice on a mixed CAST/EiJ/B6 background have been reported to exhibit CDH. We thus examined diaphragm formation and found that approximately 20% of Frem2  mice display CDH (Supplementary Fig. 3a, b) without the presence of abdominal contents in the thoracic cavity (Fig. 1f). These results suggest that primary defects in lung formation, rather than its lung compression, cause the reduction in the lung size in Frem2  mice.

To examine the formation of the lungs in more detail, we performed a systematic analysis of lung development. Frem2  mice display no obvious differences in lung size, branching pattern or lobe separation compared with those of wild-type (WT) mice at E13 (Supplementary Fig. 4a). Starting at E14.5, Frem2  lungs appeared smaller than those of their WT counterparts, and this phenotype became more pronounced at later stages (Supplementary Fig. 4b, c). Interestingly, Frem2  mice exhibited shortened distal lung buds without significant changes in the number of lung branching tips at E15 (Supplementary Fig. 4d-f). Notably, Frem2  mice also exhibited reduced cell proliferation rate in the epithelium compared with WT mice at E13.5 and E14.5 (Supplementary Fig. 4g-j). These results indicate that Frem2 may direct lung growth by promoting lung bud elongation. Collectively, these data suggest that loss of Frem2 function cause severe PH characterized by a dramatically reduced lung size.

Since Frem2  mice exhibit smaller lungs, we sought to determine whether FREM2 is required for lung growth in humans. We recruited five patients with unilateral or bilateral cryptophthalmos, a typical phenotype caused by FREM2 mutations, and their parents (Fig. 2a-e). We conducted whole-exome sequencing of genomic DNA samples to identify potential mutations in FREM2. Detected mutations were further validated by targeted Sanger sequencing (Fig. 2f-j). Four patients carry the same homozygous missense mutation (c.6499 C > T (p.Arg2167Trp)) in FREM2, and their parents are heterozygotes (Fig. 2f-i). The p.Arg2167Trp mutation is located in the fourth of five consecutive Calx-β domains; in this mutant protein, a highly conserved arginine is replaced with a tryptophan at position 2167 (Arg2167Trp) (Fig. 2k). The fifth patient carries compound heterozygous mutations in FREM2 (c.15delG; c.6499 C > T (p.Arg2167Trp)) (Fig. 2e, j). The deletion mutation (c.15delG) is predicted to cause a frameshift and results in nonsense-mediated mRNA decay of FREM2. Next, we examined lung size using high-resolution computed tomography (HRCT) and 3D reconstruction, a method for estimating lung dimensions. Interestingly, the patients with FREM2 mutations display varying degrees of PH, characterized by smaller lungs, and no obvious changes in body mass index (BMI) compared with age- and sex-matched healthy controls (Fig. 2l-r and Supplementary Tables 1-3). Notably, the patient carrying the compound heterozygous mutations (c.15delG; c.6499 C > T (p.Arg2167Trp)) exhibits a more severe lung phenotype (Fig. 2e, j, p). These results in humans are consistent with the findings in mice, suggesting that FREM2 dysfunction contributes to symptoms of PH.

We then examined the spatio-temporal expression patterns of Frem2/FREM2 in developing mouse lungs. FREM2 appeared to localize at the membrane of the pulmonary epithelial cells, mesenchymal cells in the lung parenchyma, and mesothelial cells during embryonic development (Fig. 3a, b). FREM2 can be detected in mesothelial cells labeled with PDPN as early as E12.5 and was also highly expressed from E16.5 to E18.5 (Fig. 3c). Frem2 mRNA is expressed in the lungs from E11.5 to P0 (Fig. 3d). Frem2 mRNA is highly expressed in mesothelial cells labeled with Wt1, with relatively lower expression in epithelial cells and other mesenchymal cells in E16.5 lungs (Fig. 3e, f). Meanwhile, FREM2 is identified as a marker for mesothelium in humans from A Census of the Lung: CellCards from LungMAP. We found that FREM2 was also highly expressed in the mesothelium of human infant lungs (Fig. 3g and Supplementary Table 4).

Changes in cell shape are associated with the cell polarization state. We therefore examined mesothelial cell polarity by analyzing the localization of the Golgi apparatus relative to the cell nucleus, using the cis-Golgi matrix marker GM130, a widely used method to determine cell polarity in different cell types. In WT mesothelial cells, the GM130-labeled Golgi apparatus exhibits a ribbon-like morphology and localized preferentially by the long edges of the nucleus (Fig. 4e, f). However, in Frem2  mesothelial cells, the Golgi apparatus exhibits a more compact structure with changed distribution (Fig. 4e, f). Frem2  lungs exhibit no obvious changes in the morphology or distribution of the Golgi apparatus in the airway epithelium (Supplementary Fig. 7a, b). We further tested changes in the polarization of mesothelial cells using another polarity marker, ZO-1. The apical side of WT mesothelial cells exhibit clear ZO-1 localization. In contrast, Frem2  mesothelial cells exhibit decreased apical localization of ZO-1 (Fig. 4g, h and Supplementary Fig. 7c, d). This effect on mesothelial cell polarity appeared specific, as we did not observe significant differences between WT and mutant animals in cell proliferation (Supplementary Fig. 8a-d) or in cell apoptosis (Supplementary Fig. 8e, f) in the lung parenchyma or in the mesothelium. These data suggest that Frem2 is required for the polarization of mesothelial cells toward a spindle shape during lung mesothelium formation.

Next, we examined whether disruption of mesothelial cell polarity can inhibit lung growth. Vangl2, a key component of the planar cell polarity pathway expressed in lung cells, including mesothelial cells, is essential for cell polarization during tissue formation. To disrupt mesothelial cell polarization, we inhibited Vangl2 by using morpholino-mediated gene knockdown, a method that has been used in lung development studies. After ex vivo treatment with two Vangl2 vivo-morpholinos (vivo-morpholino-1 and vivo-morpholino-2) for 48 h, E16.5 lungs exhibit significantly reduced organ size (Fig. 4i-k and Supplementary Fig. 9). We chose vivo-morpholino-1 for the following studies, and found that this treatment led to abnormally rounded mesothelial cells (Fig. 4l, m) and altered mesothelial cell polarity (Fig. 4n, o). These data indicate that mesothelial cell polarization is required for lung growth.

To examine which cell type defects are responsible for the lung growth phenotype in Frem2  mice, we generated a Frem2  allele (Supplementary Fig. 1c-e) and specifically deleted Frem2 using the Wt1CreER line, which can induce efficient recombination in the lung mesothelium as early as E11.5. We conditionally deleted Frem2 in the lung mesothelium by intraperitoneal tamoxifen injection at E9.5 (Fig. 5a, b). Mice with mesothelial deletion of Frem2 exhibit smaller lungs (Fig. 5c, d) and alterations in mesothelial cell shape (Fig. 5e, f) and polarity (Fig. 5g-j), similar to Frem2  animals. To determine the effect of FREM2 on WT1 mesothelial cell-derived cells, we performed lineage tracing analysis. We did not observe obvious differences in the number of WT1 mesothelial cell-derived cells in the lung parenchyma in Wt1CreER;Frem2 ;mT/mG mice compared with Wt1CreER;Frem2 ;mT/mG mice (Supplementary Fig. 10), indicating that FREM2 does not affect cell differentiation from mesothelial cells during lung development. We also deleted Frem2 using the Twist2 line, which induces efficient recombination in the lung mesenchyme and mesothelium as early as E10.5.. Twist2;Frem2  mice exhibit hemorrhage in regions around the eyes (Supplementary Fig. 11a), fused and smaller lungs (Supplementary Fig. 11b-e) and altered mesothelial cell shapes (Supplementary Fig. 11f, g) compared with controls. However, epithelial cell-specific deletion of Frem2 (Nkx2.1;Frem2 ) do not lead to obvious differences in lung size (Supplementary Fig. 11h-j). Notably, adult Twist2;Frem2  mice, but not adult Nkx2.1;Frem2  mice exhibit a lower forced vital capacity (Supplementary Fig. 11k, l), indicating reduced lung volumes in the smaller lungs. Collectively, these data suggest that the function of Frem2 is specifically required in the mesothelium for lung growth and function.

Since other internal organs, such as the heart and stomach, are also enveloped by a mesothelium (called the epicardium in the heart), we sought to determine whether Frem2 is also required for the growth of these organs. We found that Frem2  mice display smaller hearts than their WT siblings (Supplementary Fig. 12a--). We also analyzed the distribution of FREM2 in the developing heart and observed that FREM2 locates around the epicardium at E16.5 (Supplementary Fig. 12d). Frem2  hearts display a more rounded epicardial cell shape (Supplementary Fig. 12e, f) and altered epicardial cell polarity (Supplementary Fig. 12g, h). In addition, FREM2 is highly expressed around the stomach mesothelium at E18.5 (Supplementary Fig. 13a). Frem2  stomachs also exhibit a reduced size (Supplementary Fig. 13b, c), a more rounded mesothelial cell shape (Supplementary Fig. 13d) and altered mesothelial cell polarity (Supplementary Fig. 13e). Collectively, these data suggest that FREM2-mediated mesothelial cell polarity plays the primary role in organ growth.

To investigate the molecular mechanisms underlying Frem2-mediated lung growth, we examined ECM protein formation and levels. Mesothelial cells are predominantly attached to the basement membrane, which is underlain by an ECM layer containing elastic fibers and dense bundles of fibrillar collagen that can be produced by mesothelial cells. Since elastin provides tissues and organs with elasticity to promote organ formation, we hypothesized that Frem2 deficiency might lead to structural defects in Elastin, which caused defects in pleural mesothelium formation and lung growth. To test this hypothesis, we analyzed Elastin structure and levels. The elastic fibers around WT mesothelial cells are relatively continuous at E13.5 and E16.5 (Fig. 6a, b and Supplementary Fig. 14a, b). In contrast, fragmented elastic fibers are observed in Frem2  lungs (Fig. 6a, b and Supplementary Fig. 14a, b) and Wt1CreER;Frem2  lungs (Fig. 6c, d). However, Frem2  lungs exhibit no significant changes in the protein levels of Elastin, Fibronectin or Collagen IV (a basement membrane component) (Supplementary Fig. 14c, d), indicating that FREM2 is required for elastic fiber formation but not for Elastin protein expression. Next, we examined elastic fibers in the epicardium and the stomach mesothelium. Frem2  mice also exhibit disrupted elastic fibers around the epicardium (Supplementary Fig. 12i) and the stomach mesothelium (Supplementary Fig. 13f). Thus, we hypothesized that disruption of elastic fibers might phenocopy the Frem2 deficiency-induced PH and defects in pleural mesothelium formation. Interestingly, after treatment for 8 h or 48 h with 2 mU/ml elastase, the lung mesothelium became difficult to be separated from the underlying tissue compared with controls (Fig. 6e). After treatment for 48 h with 2 mU/ml elastase, E16.5 lungs exhibit a significant reduction in size (Fig. 6f-h) and alterations in mesothelial cell shape (Supplementary Fig. 15a, b) and polarity (Fig. 6i, j) compared with controls. These results suggest that FREM2-mediated elastin formation is essential for mesothelial polarization and lung growth.

Actomyosin forces can regulate cell shape and promote organ formation. We hypothesized that compromised actomyosin contractility might also lead to defects in mesothelial cell shape and polarity, and a reduced lung size. After treatment for 48 h with Y27632, a ROCK inhibitor that can diminish actomyosin contractility, E16.5 lungs displayed a significant reduction in lung size (Fig. 6k-m), reduced apical mesothelial enrichment of phosphorylated myosin light chain (pMLC) (Supplementary Fig. 16a, b), a marker of actomyosin tension, unaltered total MLC levels (Supplementary Fig. 16c, d), and alterations in mesothelial cell shape (Supplementary Fig. 15c, d) and polarity (Fig. 6n, o) compared with controls. Notably, Y27632 treatment also results in disruption of elastic fibers (Fig. 6p, q), while Frem2  lungs and elastase-treated lungs also exhibit decreased apical enrichment of pMLC and unaltered total MLC levels in the mesothelium (Fig. 7a-d and Supplementary Fig. 16e-h). Interestingly, all of these phenotypes are also observed in Vangl2 morpholino-treated lungs (Fig. 7e-j). We further examined how Vangl2 affects elastin fiber formation, and we examined the expression of genes involved in elastin deposition and assembly. After ex vivo treatment with Vangl2 morpholinos for 48 h, E16.5 lungs exhibit significantly decreased levels of Fbn1, Fbln4, Fbln5, Lox and Loxl1 but not Fbn2 (Supplementary Fig. 17). These results suggest that Vangl2 is required for elastin deposition and assembly. In addition, intraperitoneal injection of Y27632 also leads to reduced lung size (Supplementary Fig.18). Since several pathways, including Wnt, FGF, hedgehog, and BMP signalings, are involved in the regulation of lung growth, we examined for changes in these pathways. We found that Frem2  lungs exhibited no significant changes in Wnt signaling genes (Wnt2, Wnt3a, Wnt5a, Wnt7b, Axin2 or Ccnd1) (Supplementary Fig. 19a), FGF signaling genes (Fgf9, Fgf10, Fgfr1c, Fgfr2c, Spry1or Spry2) (Supplementary Fig. 19b), hedgehog signaling genes (Shh, Smo, Gli1, Gli2 or Gli3) (Supplementary Fig. 19c) or Bmp4 (Supplementary Fig. 19 d), indicating that Frem2 signals via a Wnt, FGF, hedgehog or BMP independent pathway to direct lung growth. Next, we examined hearts and stomachs and found that Frem2  hearts and stomachs also display decreased apical enrichment of pMLC in the epicardium (Supplementary Fig. 20a) and mesothelium (Supplementary Fig. 20b), respectively. In addition, elastase, Y27632 or Vangl2 morpholino treatment also leads to decreased sizes of the heart (Supplementary Fig. 21a-f) and stomach (Supplementary Fig. 21g-l). Altogether, these data suggest that ROCK signal, elastic fibers and mesothelial cell polarity interact to drive lung and other internal organ growth (Fig. 7k).

We sought to further understand how FREM2 deficiency leads to impaired elastic fiber formation around the mesothelium. Our observation of short elastic fibers prompted us to measure the levels of MMP2 and MMP9, critical Elastin-cleaving enzymes. Frem2  lungs exhibit significantly increased levels of MMP2 in the mesothelium (Fig. 8a). Patients with FREM2 mutations also exhibit increased MMP2 levels in the serum compared with healthy controls (Fig. 8b). However, Mmp9 levels appear unaffected in Frem2  lung mesothelium compared with WT (Supplementary Fig. 22). Consistent with the above findings, knockdown of FREM2 in MeT-5A human pleural mesothelial cells led to increased levels of MMP2 but not MMP9 (Fig. 8c-f).