Epithelial heparan sulfate regulates Sonic Hedgehog signaling in lung development

The tree-like structure of the mammalian lung is generated from branching morphogenesis, a reiterative process that is precisely regulated by numerous factors. How the cell surface and extra cellular matrix (ECM) molecules regulate this process is still poorly understood. Herein, we show that epithelial deletion of Heparan Sulfate (HS) synthetase Ext1 resulted in expanded branching tips and reduced branching number, associated with several mesenchymal developmental defects. We further demonstrate an expanded Fgf10 expression and increased FGF signaling activity in Ext1 mutant lungs, suggesting a cell non-autonomous mechanism. Consistent with this, we observed reduced levels of SHH signaling which is responsible for suppressing Fgf10 expression. Moreover, reactivating SHH signaling in mutant lungs rescued the tip dilation phenotype and attenuated FGF signaling. Importantly, the reduced SHH signaling activity did not appear to be caused by decreased Shh expression or protein stability; instead, biologically active form of SHH proteins were reduced in both the Ext1 mutant epithelium and surrounding wild type mesenchymal cells. Together, our study highlights the epithelial HS as a key player for dictating SHH signaling critical for lung morphogenesis.


Author summary
Defective development of the respiratory system leads to congenital or postnatal respiratory disorders. Although it is well established that lung morphogenesis is guided by interactions of multiple morphogen signaling pathways in epithelium and mesenchyme, how their regulatory functions are controlled by cell surface and extra cellular matrix (ECM) molecules is still largely unknown. Heparan Sulfate (HS) glycosaminoglycans are widely presented on the cell surface and in the extracellular matrix depending on the attached core proteins, and are thought to be important for transducing FGF signaling. The function of HS during lung development has not been studied in vivo. Here, using mouse genetics, we demonstrate that epithelial HS biosynthesis is critically required for lung PLOS

Introduction
The developmental program of the mammalian lung is a multi-step and fine-tuned process. In mouse, it starts from the emergences of lung primordium that expresses the transcriptional factor Nkx2.1 at embryonic day (E) 9.5 in the foregut endoderm [1,2]. The two primary lung buds then repeatedly undergo morphogenic changes to generate a tree-like structure until E16.5 [2]. This high hierarchy process, referred to as branching morphogenesis is essential to generate numerous airways and gas-exchanging units, and is critically regulated by interactions of signaling pathways in the epithelium and mesenchyme. Early buds formation is dependent on reciprocal interactions between the lung epithelium and mesenchyme mediated by distinct signalings. The distal signaling center is marked by the expression of fibroblast growth factor (FGF) 10 in the mesenchyme. FGF10 signaling serves as a chemoattractive cue for epithelium outgrowth [1,[3][4][5]. On the other hand, Sonic Hedgehog (SHH) produced in the epithelium is required for mesenchymal development and negatively regulates Fgf10 expression, leading to a focal Fgf10 expression pattern in the intertubular mesenchyme to facilitate the formation of new epithelial buds [3]. The SHH-FGF10 signaling, as well as other signaling pathways including WNTs and BMP pathways in the localized region act cooperatively to regulate the branching program [5][6][7]. Heparan sulfate glycosaminoglycans (HS-GAG) are presented as linear polysaccharide chains, and they are synthesized by the glycosytransferase exostosin (EXT) 1/2 complex that attaches alternating N-acetyl glucosamine (GlcNAc) and D-glucuronic acid (GlcA) units to the growing chain [8,9]. The HS chains covalently attach to core proteins to form Heparan Sulfate Proteoglycans (HSPGs). HS are presented both on the cell surface and in the extracellular matrix, depending on their core protein localizations [10,11]. Due to its conserved negatively charged sugar motif, HS chains exhibit a wide spectrum of binding affinity to growth factors demonstrated in invertebrates and vertebrates. In Drosophila, it has been shown that signaling activities of Hedgehog (Hh), FGFs and Wingless (Wg) are regulated by both HS chains and core proteins [11][12][13]. Various models have been proposed for HS involvement in morphogen movement, ligand distribution, and intracellular trafficking [11]. While mouse embryos lacking Ext1 failed to gastrulate [14], it is still incompletely understood how widely the proposed mechanisms are applicable to mammalian organogenesis.
Despite the critical roles of HS in the development of mammalian organs, especially those generated from branching morphogenesis [15][16][17][18][19][20], limited studies have unveiled the role of HS in the context of lung development. Mice deficient in N-deacetylase/N-sulfotransferase-1 (NDST1) had disrupted BMP signaling and displayed a lung atelectasis associated with differentiation defects [21]. Studies using ex-vivo organ culture demonstrated that O-sulfated HS in the developing lung epithelium is required for FGF10 binding [22,23]. However, mice lacking Hs6st1 display postnatal changes in lung with no major organogenesis defects [24]. These results indicated a spatial-temporal regulation of lung development by HS sulfation patterns. Despite the diverse mechanisms, there is still a lack of genetic evidence for understanding the role of HS in mammalian lung development. Therefore, a conditional model to deplete the HS chains in all HSPGs would provide new insights into the general roles of HS in mammalian lung development.
In the present study, using a Shh cre mouse line, we conditionally ablated Ext1 gene in the developing mouse lung epithelium. We demonstrated that epithelial HS is required for lung branching morphogenesis by maintaining SHH signaling activity. Decreased SHH signaling activity in mutant embryos leads to an expanded and elevated Fgf10 expression, which is likely to contribute to the dilated branching tips. We show that the production of biologically active form of SHH was reduced after loss of HS in the epithelium. Thus, these data highlight the crucial requirement of HS in SHH-producing epithelial cells to maintain SHH signaling activity, and reveal the importance of HS in lung morphogenesis.

Results
Epithelial loss of Ext1 disrupts HS synthesis and lung morphogenesis EXT1 protein is essential for the biosynthesis of HS chains [8]. To investigate the role of HS in lung development, we specifically ablated Ext1 gene in the developing lung epithelium, which was achieved by a Shh cre mediated recombination. Ext1 mutant mice carry two conditional Ext1 flox alleles and a Shh cre allele (Ext1 f/f ; Shh cre ). The Cre recombinase gene in Shh cre is inserted into the coding region of Shh, resulting in mutation of one Shh allele, the Ext1 f/+ ; Shh cre mice were therefore used as controls to eliminate the influence of Shh haploinsufficiency. Mutant mice were born with respiratory defects and died of breath failure shortly after birth. Analyzing the gross morphology and histology revealed that the mutant lungs were highly hypoplastic and displayed reduced branching numbers and enlarged airways at E12.5 and E14.5, with resultant cyst-like architecture in E18.5 lung (Fig 1A-1L, S1A and S1B Fig). Notably, although the lung development could be initiated, the domain branching was severely disrupted: the four right lobes were established as early as E12.5 in control lungs, while mutant lungs displayed a delayed lobe specification, and the accessory lobe was absent in later developmental stages (Fig 1A-1F). We also observed that only two lung lobes were formed in a small proportion of the mutant lungs which were most severely affected (S1C and S1D Fig). Loss of Ext1 resulted in a failure in separation of esophagus and trachea (S1E-S1H Fig). Moreover, we observed a malformation of cartilage rings in mutant lungs (Fig 1M-1P), suggesting a cell non-autonomous requirement of Ext1 in epithelium for mesenchymal development.
We next analyzed the HS distribution by immunofluorescent staining of lung sections using two common specific antibodies against HS, 3G10 and 10E4. The 3G10 antibody recognizes the neo-epitope after heparinase III digestion, which gives signals where HS is presented without providing information of HS structure. Therefore, it served as a general marker of HS. The 10E4 antibody recognizes the native HS chains of HSPG with N-sulfates and low content of O-sulfates. Consistent with previous studies [22,23], we observed that 3G10 and 10E4 signals are ubiquitously presented in the developing lung epithelium and mesenchyme. Strong 3G10 signal was found in the basement membrane where 10E4 signal is not presented (Fig 2A, 2C, 2E and 2G), indicating that HS in basement membrane is highly O-sulfated. We found that nearly all the HS 3G10 staining in the epithelial cell surface and basement membrane (Fig 2A-2D) and the HS 10E4 staining in the epithelial cell surface (Fig 2E-2H) were abolished when Ext1 is depleted in the developing lung epithelium. These results further confirmed that deletion of Ext1 could efficiently disrupt HS biosynthesis and that epithelial HS is critically required for lung morphogenesis.
Epithelial HS is required for mesenchymal cell development but dispensable for epithelial cell proximal-distal patterning and differentiation We next tested whether epithelial HS regulates lung patterning and cell differentiation. Although functional epithelial cells have not differentiated during branching morphogenesis, those progenitor cells can be classified into SOX2+ and SOX9+ cells that specify the proximal and distal lineages, respectively. To examine whether epithelial HS regulates lung patterning, E12.5 lungs were analyzed by the SOX2/SOX9 immunostaining. Our results showed no apparent changes in proximal-distal patterning as indicated by the SOX2 and the SOX9 marked region despite the severe defects of branching in mutant lungs (Fig 3A and 3B). Notably, SOX9 + mesenchymal cells that develop into future cartilage around the trachea appeared to be reduced (Fig 3A and 3B, inserts). In E14.5 control lungs, SOX9+ cells formed regular arranged nodes along the trachea, whereas the SOX9+ cells around the trachea in mutant lung seemed to be more dispersed and the nodes were irregularly arranged (Fig 3C and 3D, arrows), which coincides with the malformation of cartilage ring observed at E18.5 (Fig 1M and 1P). Alpha smooth muscle actin (α-SMA) staining, representing airway smooth muscle differentiation, was lining the whole developing airways and extended to the stalk of the growing lung buds in control lungs, however, in mutant lungs, its expression was restricted to a more proximal region (Fig 3E and 3F [25]. To test whether ablation of HS in epithelium could have an impact on epithelial cell differentiation, we detected markers of major cell types of lung epithelium at E18.5. Both the type I and type II alveolar cells were presented in the cyst-like structure of the mutant lungs (Fig 3G and 3H). The club cells and ciliated cells were also presented in the cyst-like structures (Fig 3I and 3J). Basal cells and mucus cells in the trachea seem to be normally differentiated (Fig 3K-3N). These results suggest that HS in epithelium is not required for its proximal-distal patterning and differentiation but exerts a cell nonautonomous regulatory function for some aspects of the mesenchymal development.

Epithelial HS controls SHH-FGF10 signaling
Previous studies have demonstrated that increased FGF signaling leads to abnormal branching and increased tip size [26,27], which shares striking resemblances with our Ext1 mutant lungs, suggesting a possible involvement of aberrant FGF signaling in the defective sac-like lungs. Analysis of the expression of epithelial and mesenchymal housekeeping genes Cdh1 and Vimentin revealed no changes in Ext1 f/f ; Shh cre mutant lungs (S4 Fig), suggesting the proportion of epithelial cells versus mesenchymal cells was not significantly altered, confirming the validity of using Gapdh expression as an internal control. Indeed, qPCR revealed a 2-fold increase of Fgf10 expression in mutant lungs compared to controls ( Fig 4A). Concurrently, FGF10 targets, Spry2, Dusp6, Etv4 and Bmp4, were upregulated ( Fig 4A). Etv5, another target of FGF signaling, was decreased ( Fig 4A), similar to what was observed in the Yy1 mutant lungs [27], suggesting that epithelial HS may have pleiotropic effects on other signaling pathways. Particularly, whole mount and section RNA ISH revealed that Fgf10 expression was uniformly high in the surrounding mesenchyme of mutant lungs, distinct from the split localization between branching tubes in control lungs (Fig 4B-4E). Expression of phosphorylated-ERK (pERK), a well-established indicator of FGF signaling activity in lungs [28,29], exhibited an expanded pattern in the dilated epithelium with much stronger fluorescent intensity compared to controls (Fig 4F and 4G),consistent with the expanded pattern of Fgf10. The expression of Fgf9 and Fgfrb2 were unchanged ( Fig 4A).
Since we conditionally deleted Ext1 in the epithelium, the cell non-autonomous effect on Fgf10 suggests that epithelial loss of HS may have affected factors that are produced by epithelium to regulate Fgf10 expression in mesenchyme. SHH ligand secreted from lung epithelium is essential to direct the mesenchymal development. Previous studies showed that loss of Shh not only resulted in branching defects, but also caused defects of trachea cartilage and airway myofibroblast [30,31]. All of these defects were observed in our Ext1 mutant lungs. Notably, Fgf10 is reported to be negatively regulated by SHH signaling [3]. Our data indicate that HS might regulate a hierarchical signaling pathway by modulating SHH signaling activity. However, no significant change in the expression level of Shh was detected by qPCR( Fig 4H). ISH analysis confirmed the qPCR results and showed no apparent changes in its expression level and pattern (Fig 4I-4L). In contrast, Gli1, Ptch1 and Hhip1 were all reduced as shown by qPCR and/or ISH experiments (Fig 4H and 4M-4P), reflecting a decreased SHH signaling activity. We also introduced a Gli1 lacZ reporter allele into the Ext1 mutant. β-galactosidase (βgal) staining revealed a decreased Gli1 expression in mutant mesenchyme, confirming a reduction in SHH signaling activity (Fig 4Q and 4R).  FGF9 is also expressed in the developing lung epithelium, and epithelial specific overexpression of Fgf9 causes dilation of airways and ectopic Fgf10 expression [32]. Moreover, it was demonstrated that FGF diffusion is restricted by GAGs in source cells [15]. The unchanged Fgf9 expression did not exclude the possibility of FGF9 over-release into the mesenchyme, which may potentially contributes to the upregulated and expanded Fgf10 expression. However, FGF9 distribution appears to be normal in Ext1 f/f ; Shh cre mutant lungs as revealed by immunofluorescent staining (S5A-S5D Fig), and consistently, the mesenchymal expression of pERK was not significantly altered in Ext1 f/f ; Shh cre mutant lungs (Fig 4F and 4G). These results indicate that Fgf9 mediated signaling pathway is not responsible for this phenotype.
Together, these data uncovered a critical regulatory role of epithelial HS in directing epithelial-mesenchymal interaction during lung branching by proper control of SHH-FGF10 signaling.

Activation of SHH signaling partially rescues the branching phenotype in mutant lungs
To test if the decreased SHH signaling activity in mutant lungs contributes to the branching defects, we manipulated SHH signaling in an explant culture system by using Smoothened agonist (SAG). We chose a final concentration of 10nM, which was close to the dose used in a previously published paper [28], Ptch1 ISH confirmed that administration of SAG at this concentration could efficiently increase SHH signaling during the 48h of culture (S6A and S6B  Fig). Control lungs treated with SAG did not exhibit significant alternations in branching number (S6C Fig) and epithelial lumen size as previously reported (Fig 5A, 5B, 5E, 5F, 5I, 5J and 5M). The mutant lungs cultured in vitro have the similar branching defects as what we observed in vivo, resulting in significantly dilated branching tips after culturing for 48hrs ( Fig  5C,5G and 5K). However, SAG treatment efficiently restricted the tip size of the mutant lungs to a level that was comparable to the control lungs, although the loss of branching number was not rescued (Fig 5C, 5D, 5G, 5H, 5K, 5L and 5M). We also measured Fgf10 expression in the cultured lungs, data from ISH analysis confirmed an attenuation of Fgf10 expression in Ext1 f/f ; Shh cre explants after SAG treatment (Fig 5N-5Q). Consistently, the epithelial pERK level in the branching tips of SAG treated Ext1 f/f ; Shh cre lungs was decreased, suggesting an attenuation of FGF signaling (Fig 5R-5U). These data argue that epithelial HS is required to maintain the normal epithelium lumen size and FGF signaling, at least partly through maintaining SHH signaling activity.
The lack of rescue of branching number by SAG treatment suggested that epithelial inactivation of Ext1 may have pleiotropic effects on other signaling pathways that affect lung branching. Previous studies have shown that WNT signaling, both canonical and non-canonical, mediates the effects of multiple factors on lung branching [33][34][35][36][37][38][39], and given the critical regulatory function of HS on WNT/β-catenin signaling in gut [40], we then tested whether WNT signaling is involved in our model. QPCR and ISH studies suggest that Wnt2 expression was not significantly changed, while the expression of Wnt7b was reduced(S7A- S7E Fig) in  [41].Epithelial F-actin distribution, which is also under the control of WNT/PCP pathway, was comparable between control and mutant lungs (S7J and S7K Fig). These results indicate that noncanonical WNT signaling was overall normal in Ext1 f/f ; Shh cre lungs. To test whether the decreased canonical WNT signaling accounted for some of the branching phenotype, we treated lung explants with LiCl at 10mM, a previously used dose that could efficiently activate canonical WNT signaling in explant culture system [37].However, LiCl treatment was insufficient to rescue the branching defects of mutant lung explants (S7L-S7S Fig). These findings indicate that epithelial HS directly or indirectly modulates canonical WNT signaling in developing lung, however, the reduction of canonical WNT signaling may only play a minor role in branching or synergize with other dysregulated signaling pathways to contribute to some aspects of the developmental defects of Ext1 mutant lungs.
Epithelial HS is required for the production of biologically active form of SHH in the developing lung Hh-HS interactions have been extensively studied in vivo in Drosophila. It has been proposed that HSPGs in both the receiving cell and producing cell is required for Hh signaling [42][43][44], we asked whether HS plays conserved roles to orchestrate SHH signaling during lung development in vertebrates. First, we examined the denatured protein extracts from E14.5 lungs by western blot using a polyclonal antibody that recognizes SHH N-terminus. As shown in Fig  6A, no detectable change in SHH protein was observed, suggesting that protein stability was not affected in mutant lungs. It was previously shown that SHH shedding regulated by ECM factors such as a disintegrin and metalloprotease (ADAM) directs processing of the N terminal lipidpeptide, which exposes the zinc coordination sites (5E1 epitope) to reach the full inductive potential of the ligand [45]. And in vitro studies suggest that cell surface HS in Hh-producing cells spatiotemporally participates in this biological process [42,46,47]. To test whether HS is required for the spatial distribution of SHH protein in lung epithelium and mesenchyme, we stained lung sections using the monoclonal 5E1 SHH antibody. To mark the epithelium, costaining with a GFP antibody was performed, as a Gfp reporter gene was also inserted into the Shh locus along with Cre in the Shh cre allele, the GFP signaling is therefore consistent with Shh mRNA [48]. We observed that SHH 5E1 signal presented as large particles both in the epithelium and the surrounding mesenchyme in control lungs; although SHH 5E1 signals could be found in the mesenchyme, both the intensity and the number of SHH particles were prominently reduced in mutant lungs (Fig 6B-6G). We then obtained Shh null lungs by crossing the Shh cre/+ mice to generate Shh cre/cre embryos. There was no detectable 5E1 staining in Shh cre/cre lungs(S8A- S8D Fig), confirming the specificity of these cluster signaling. Because the 5E1 antibody recognizes the zinc coordinate site required for its receptor binding. [49], in mutant lungs, less 5E1 reactive SHH puncta reflected a low level of functional SHH protein clusters. Intriguingly, the apical-basal distribution of SHH particles in epithelium was also severely disturbed after epithelial ablation of HS: SHH were prone to be enriched in the basal side on the epithelium-mesenchyme interface (Fig 6B and 6C), where HS 3G10 signaling is most strongly presented (Fig 2A and 2C); whereas the basal distribution was not observed in mutant lungs (Fig 6E and 6F). Finally, we performed an immunoprecipitation (IP) assay with the E14.5 lung homogenates. Consistent with the above data, we found that 5E1 immunoprecipitated SHH protein was significantly reduced (Fig 6H), which confirms the reduction of 5E1 immunoreactive SHH protein. These results suggested a possibility that the HS chains in the epithelial cell surface and basement membrane possibly serve as an important SHH repository network for its processing and later transportation, assisting to maintain a proper concentration of biologically active SHH in the surrounding mesenchyme.

Discussion
Although it is well understood that multiple morphogens regulate interconnected transcriptional network to control lung development, how the cell surface and ECM molecules are involved in these morphogenic processes is still incompletely understood. HS has been reported to regulate branching morphogenesis of multiple organs such as mammary gland,

Fig 6. Epithelial-derived HS is required for the production of biologically active form of SHH in the developing lung.
(A) Representative image of western blot analysis of SHH protein in E14.5 lungs. The protein level of SHH was not changed in Ext1 f/f ; Shh cre mutant lungs compared to control lungs. (B-G) Immunoflouorescent staining for SHH(5E1) and GFP in E12.5 lungs. The GFP expression is consistent with the Shh mRNA. The Ext1 f/f ; Shh cre mutant lungs presented less SHH particles in both the epithelium and mesenchyme. SHH particles were prone to localize on the basal side of the epithelium in control lungs, whereas the basal-localization of SHH was reduced in Ext1 f/f ; Shh cre mutant lungs.(H) Immunoprecipitation(IP) of E14.5 lung homogenates using 5E1 monoclonal antibody revealed reduced 5E1-immunoreactive SHH was produced in Ext1 mutant lungs compared with control lungs. The longer exposure of the 5E1 IP band was shown below the boxed area. C, Control; K, Ext1 mutant. (I) Model for the regulatory role of epithelial HS in lung branching morphogenesis. During normal lung branching, SHH produced by epithelium signals to mesenchyme to ensure a focal Fgf10 expression pattern, epithelial HS is required for the production of biologically active form (5E1 immunoreactive) of SHH ligand, possibly by controlling higher order processing. Loss of epithelial HS leads to a reduced production of biologically active form of SHH, thus causing a reduced SHH signaling submandibular gland, and lacrimal gland [16][17][18][19][20][50][51][52]. In a number of contexts, loss of HS leads to disorders of FGF signal transduction. Thus, we had initially speculated that loss of HS in epithelial cells could lead to an FGF loss of function phenotype during lung development. To our surprise, an FGF gain of function phenotype was observed. We demonstrate that lung epithelial derived HS, presented both on the cell surface and basement membrane ECM, is critical for the maintenance of SHH signaling activity, thereby exerting an indispensable role in branching morphogenesis. Ablation of HS synthetase Ext1 in the epithelium resulted in strikingly reduced levels of biologically active (5E1-immunoreactive) SHH particles both on the epithelium and mesenchyme and subsequently reduced signaling activity in the receiving cells, causing an unrestricted Fgf10 expression, which ultimately led to expanded branching tips and reduced branching number (Fig 6I). Our study highlights the critical functions of HS during lung development.
In Ext1 conditional knockout lungs, we observed an apparent decline of cell proliferation of the epithelial and mesenchymal cells; in addition, epithelial differentiation was largely unaffected, whereas the airway myofibroblast and cartilage development were severely disrupted, suggesting there is a cell non-autonomous requirement of the epithelial HS for multiple aspects of the lung development. Lung development is critically regulated by interactions of signaling molecules between epithelium and mesenchyme, among which SHH is one of the most studied molecules. SHH, produced by epithelium cells, signals to the adjacent mesenchyme cells to support mesenchyme cell development and regulate the expression of growth factors such as Fgf10 and Wnt2 [31]. FGF10 in turn signals to the epithelium, which is critically required for domain branching, bifurcation and epithelium extension [1]. Shh null mutant failed to form a functional lung, as a result of branching halt [31]. Recent studies showed conditional knockout of transcriptional factors, such as ETV and YY1 in lung epithelium reduced Shh expression, which subsequently broke the balance between SHH and FGF signaling, resulted in enlarged tips and decreased branching number [27,28]. These studies suggested that a fine-tuning of SHH-FGF signaling is essential for lung branching morphogenesis. Our data showed that epithelial deletion of HS caused similar branching defects as described above, including enlarged tip bud size, decreased branching number, and expanded Fgf10 expression. We further provided evidence that SHH signaling was reduced in mutant lungs, highlighting an essential requirement for HS in dictating the SHH-FGF10 signaling cascade.
Prior to our study, several studies have examined the function of HS in lung morphogenesis. The O-sulfated HS in the lung epithelium was proven to be required for FGF10-FGFR2b binding and signaling in an organ culture system [22,23]. In support of this notion, genetic ablation of Hs6st in lacrimal gland disrupted FGF10-FGFR2b signaling [16]. However, mice lacking Hs6st seemingly did not bear lung branching defects [24]. In Ext1 mutant lungs, we observed an increased FGF signaling, as exemplified by increased expression of multiple FGF targets and pERK level, which stands in contrast to the previous notion that HS is required for FGF signaling transduction, suggesting that the overabundance of FGF10 ligand in the local environment may compensate the decreased FGF10-FGFR2b interaction, or an HS-independent mechanism that facilitates FGF10-FGFR2b binding may exist in the lung epithelium when HS is removed.
Our data indicate that the branching defects in Ext1 mutant lungs mimicked those with increased FGF signaling [26][27][28]53], this is likely resulted from an increased and expanded Fgf10 expression in the mesenchyme. Thus, we speculated that signaling that acts upstream of and an elevated and expanded Fgf10 expression. Decreased SHH signaling and unrestricted FGF10 signaling lead to dilated branching tips and decreased branching number. Scale bar: B and E, 20μm.
https://doi.org/10.1371/journal.pgen.1006992.g006 FGF10 might be altered. Indeed, we have detected a retard of SHH signaling in Ext1 mutants, without obvious changes in its mRNA level and protein stability, indicating that HS-SHH interaction in the producing cell maintains the signaling activity possibly by a higher order processing of SHH. The current knowledge on HS-Hh interaction is mostly obtained from in vitro experiments and in vivo experiments on lower organisms. In ttv (Ext1 in Drosophila) null Drosophila embryo, Hh failed to move to its target cells [54]. Hh also failed to move through Glypican-mutant cells [44]. Biochemical studies suggested that the multimerization of vertebrate SHH induces long range signaling activity [55,56], and SHH multimer formation and release is dependent on HS [42,45,46]. In particular, it was previously shown that nanoscale oligomerized Hh interacts with HSPGs to form the densely packed multimers, which is essential to present itself on cell surface or in lipoproteins for long range signaling [43]. Recent evidence indicates that HS in the heteroprotein hubs can attract sheddases and sheddase activators such as ADAM to help SHH multimers to undergo N-terminal processing, a procedure required to reveal its zinc coordinate sites and results in its functional activation. [42,[45][46][47]. In support of this model, after abrogating HS synthesis in lung epithelium, we observed a dramatic reduction of the 5E1-immunoreactive SHH particles both in the epithelium and mesenchyme, with a reduction of SHH signaling activity. We speculate that abrogation of epithelial HS changed ECM dynamics required for SHH processing, this notion is supported by a recent finding showing compromised SHH signaling in the umbilical cord of ADAMTS9 mutant mice [57]. Observations in Ext1 null embryos found that less IHH(5E1) protein was associated with endodermal cell surface although its protein level was not changed, similar to our finding [14]. In this study, we observed strong 3G10 epitope of HS in the basement membrane (Fig 2A and 2C). Previous biochemical analysis demonstrated that a higher sulfation level of HS chains leads to a higher affinity to SHH protein [58]. This could possibly explain why we observed a strong signal of SHH on the basal side of the epithelium, because the embryonic lung basement membrane HS are highly O-sulfated rather than N-sulfated [23]. In the context of developing ventral spinal cord, Sulf1, an enzyme that reduces the 6-O-sulfation of HS and is expressed in SHH producing cells, was required for the production of biologically active form of SHH [59][60][61], demonstrating that SHH signaling is precisely regulated by the dynamic change of HS sulfation pattern. Additional genetic evidence is required to understand the role of HSPGs and HS modification enzymes in SHH signaling more conclusively in the developing lung. Moreover, it would also be interesting to investigate the potential role of mesenchymal HS during lung development.
Our results support that epithelial HS is required for the production of fully activated SHH ligand to ensure normal lung branching. Although the tip dilation phenotype and hyperactivation of epithelial FGF signaling in mutant lungs can be attenuated by SAG, the lack of rescue of branching number suggests that there must be additional signaling pathways that act coordinately with SHH-FGF10 signaling to contribute to the phenotype. Alternation of multiple signalings have been reported to have similar defects in Ext1 mutant lungs. For example, Fgf9 overexpression in lung epithelium induces upregulation of Fgf10, and causes epithelium expansion [32]; inactivation of Frizzled 2 results in dilated branching tips through modulation of non-canonical WNT signaling [35]. However, we did not find evidence of changes in FGF9 signaling or non-canonical WNT signaling. Intriguingly, Ext1 mutant lungs exhibited a dramatically decreased Wnt7b expression and canonical WNT signaling. Despite the failure of LiCl to rescue the branching, the decreased canonical WNT signaling may work synergistically with other dysregulated pathways to affect lung morphogenesis in Ext1 mutant lungs, further genetic study is required to analyze this issue. In addition to the disruption of external cues required for lung branching, loss of HS may potentially result in changes of ECM composition, which may cause intrinsic cellular defects that affect lung branching. Proper deposition of Collagen and Integrin is required for the migration activity of lung epithelial [62][63][64], and a recent finding in C. elegan provide evidence of the requirement of HS for neural cell migration [65]. Whether HS interacts with these factors to contribute to the lung branching phenotype needs to be further explored.
In conclusion, our study demonstrates epithelial derived HS is critical in maintaining SHH signaling activity, thus dictates the epithelial-mesenchymal interaction during lung development. As Hh signaling is proposed to be involved in lung homeostasis and tumorigenesis [66,67], it is possible that HS is involved in the pathogenesis of many lung diseases such as cancer, fibrosis and COPD, in addition to its essential role during organogenesis.

Ethics statement
Animal experiments were performed following the Institutional guidelines and approved by the Institutional Animal Care and Use Committee of Institute of Zoology, Chinese Academy of Sciences(approve number: AEI-09-012014).

Animals
Mouse lines for Ext1 f/f , Gli lacZ , Shh cre were described elsewhere [48,68,69]. All lines were maintained on a mixed background. Mice were housed in a normal 12 light/12 dark cycle in the SPF (Specific-Pathogen-Free) animal facility of Institute of Zoology, Chinese Academy of Sciences, Beijing, China. Shh cre line was used to delete Ext1 specifically in lung epithelial cells during lung morphogenesis. Mutant mice (Ext1 f/f ; Shh cre ) were generated by crossing Ext1 f/+ ; Shh cre male with Ext1 f/f female. Embryos were harvested and analyzed from time-mated females at indicated times (noon of the next day after mating counted as 0.5 day of gestation). Ext1 f/+ ; Shh cre littermates were referred as controls.

Histology and immunofluorescent staining
Embryos were harvested at indicated times and fixed in 4% paraformaldehyde at 4˚C overnight, followed by dehydrated in a series of ethanol and embedded in paraffin for sectioning. For immunohistochemistry staining, deparaffinized and rehydrated paraffin sections (5μm) were microwaved in sodium citrate buffer (pH6.0). The primary antibodies were listed in S1 Table, fluorescent conjugated secondary antibodies were all from Jackson Immunoresearch. For Immunofluorescent staining of SOX2, SHH(5E1) [70], pERK, FGF9, P63,10E4 and 3G10, ABC immunoperoxidase detection systems (Vector PK-6200 Laboratories) was used to introduce HRP. The Cyanine 3 Tyramide Signal detection was performed according to the guidelines of TSA amplification system (PerkinElmer). For 3G10 staining, sections were pretreated by heparanase III (400 mU/ul, Sigma) at 37˚C for 2 h to expose neo-epitope site. TUNEL staining was performed using the DeadEnd™ Fluorometric TUNEL System (Promega) according to manufacturer's protocol. F-actin staining was performed on 12μm-cryosections using Alexafluor546 conjugated phalloidin (Invitrogen). All the slides were imaged on a Zeiss LSM 780 laser-scanning confocal microscope under the same laser exposure between control and mutant lungs.

Alcian blue cartilage staining
Respiratory tracts were dissected and stained with alcian blue solution as previously described [27]. The tracheas were rinsed with PBS and immersed in 70% glycerol for image.
RNA in situ hybridization (ISH) cDNA fragments amplified from mouse lung cDNA were used as probe templates. The primers were listed in S2 Table. RNA riboprobes were produced according to the instruction of DIG RNA Labeling Kit (Roche). Sense riboprobes transcribed by T3 RNA polymerase were used as negative control. ISH on paraffin sections was performed as previously described [37]. Whole mount in situ hybridization was carried out using an optimized protocol [71].

Organ culture
Lungs were harvested from E11.5 embryos in DMEM/F12 (Corning), and then placed onto Nucleopore Track-Etch membranes (8μm Whatman) to expand lungs at the air/medium interface. Ex vivo lungs were cultured in DMEM/F12 with 100U penicillin and 100μg streptomycin (Thermo Fisher Scientific) at 37˚C in 5% CO2. SHH signaling activator, smoothened agonist SAG (10nM, SelleckChem) was added into culture media. DMSO was used as a vehicle control. Bright field images were captured 48 hours after SAG administration, and quantification was performed using Image J software.

Quantitative RT-PCR (qRT-PCR)
RNA isolation with Trizol (Invitrogen) was performed according to manufacturer's instructions. First strand cDNA was reverse transcribed using GoScript Reverse Transcription System (Promega). Quantitative PCR was performed using GoTaq qPCR Master Mix (Promega) and amplification was quantified by CFX-96 system (Bio-rad). Expression levels were normalized to Gapdh. At least three biological replicates are performed for each experiment. Primer sequences used are listed in S2 Table. Immunoprecipitation (IP) and western blot Lungs were dissected from Embryos in PBS, and homogenized using Dounce tissue grinder in RIPA lysis buffer (Beyotime) containing protease inhibitor and phosphatase inhibitor (Roche). For each IP, approximately 80mg pooled lung tissues were homogenized in 500μl lysis buffer, lung lysates were centrifuged at 15000g for 20min,supernantants were collected. 40μl Protein A/G plus agrose beads(Santa Cruz biotech, Inc) and 5μg 5E1 antibody (DSHB) was used for each IP reaction. After rotation at 4˚C overnight, the beads were washed, reconstituted and boiled in 30ul 1X SDS loading buffer. For western blot analysis of tissue lysates, approximately 80μg total protein was reduced in 4X SDS loading buffer and subjected to 12% SDS-PAGE. Immunobloting was performed on PVDF membrane. Membranes were incubated with antibodies against SHH (goat polyclonal, AF464, RD Systems Inc., 1:300),β-actin(mouse monoclonal, CW0096M, Cwbiotech, 1: 2000) and p-JNK(rabbit polyclonal, 9251, Cell signaling technology,Inc., 1:1000) at 4˚C overnight. HRP conjugated anti-mouse, anti-rabbit and anti-goat secondary antibodies (Jackson Immunoresearch) were used and peroxidase activity was detected by SuperSignal West Pico Chemiluminescent Substrate (Thermo Scientific). Chemiluminescent signal was detected and pictures were captured by GE ImageQuant LAS 4000. Quantification was performed using the Image J software.

Statistical analysis
Data are shown as means ± SEM. For comparative analysis, Unpaired Student's two-tailed ttest and graphs were achieved using GraphPad Prism6. p<0.05 was considered statistically significant.