|Home | About | Journals | Submit | Contact Us | Français|
Deriving lung progenitors from patient-specific pluripotent cells is a key step in producing differentiated lung epithelium for disease modeling and transplantation. By mimicking the signaling events that occur during mouse lung development, we generated murine lung progenitors in a series of discrete steps. Definitive endoderm derived from mouse embryonic stem cells (ESCs) was converted into foregut endoderm, then into replicating Nkx2.1+ lung endoderm, and finally into multipotent embryonic lung progenitor and airway progenitor cells. We demonstrated that precisely-timed BMP, FGF, and WNT signaling are required for NKX2.1 induction. Mouse ESC-derived Nkx2.1+ progenitor cells formed respiratory epithelium (tracheospheres) when transplanted subcutaneously into mice. We then adapted this strategy to produce disease-specific lung progenitor cells from human Cystic Fibrosis induced pluripotent stem cells (iPSCs), creating a platform for dissecting human lung disease. These disease-specific human lung progenitors formed respiratory epithelium when subcutaneously engrafted into immunodeficient mice.
The discovery of embryonic stem cells (ESCs) and induced pluripotent stem cells (iPSCs) has resulted in an unprecedented opportunity to produce tissue-specific cell types that can be used in human disease modeling, drug screening, and patient-specific therapies. Generating disease-specific lung epithelial cells from human lung disease-specific iPSCs is particularly important because murine models of lung disease often do not phenocopy human lung disease. A prime example of the failure of mouse models to mimic human disease is the Cftr knockout mouse that does not display the Cystic Fibrosis (CF) disease-associated lung pathology observed in human patients (Snouwaert et al., 1992; Clarke et al., 1992; Guilbault et al., 2007). Extraordinary experiments have created new animal models of CF in both the ferret and the pig nearly 2 decades after the CFTR gene was cloned (Rogers et al., 2008; Stoltz et al., 2010; Sun et al., 2010; Kerem et al., 1989; Riordan et al., 1989). These heroic experiments allow the modeling of CF at a whole-animal level.
The work presented herein is an attempt to add a human model system to these remarkable prior efforts. Although it does not permit the analysis of the disease in the full animal context, the iPSC system will allow patient specificity. It is known that many patients with the same CFTR mutations do not have a similar clinical course (Kerem et al., 1990, 1992; Cutting, 2005). Some of this variability is thought to result from disease-modifying genes (Drumm et al., 2005; Haston and Hudson, 2005; Knowles, 2006). Using patient-specific iPSCs from patients such as these will permit the study of CF clinical variability in the laboratory without the need to identify these genes beforehand. Building on the remarkable recent success of the first drug to target the CFTR channel defect directly in patients with the rare G551D mutation (Accurso et al., 2010), it is clear how this iPSC-based system could serve as a platform for finding therapeutics for every patient with CF.
Although many iPSCs from patients with lung diseases are currently being produced, the major obstacle preventing the actual development of human lung disease models using these cells is the inability to convert them into lung progenitors and subsequently into differentiated pulmonary epithelial cell types. Several attempts to produce pulmonary epithelial cells from mouse and human ESCs have been made. Most investigators have focused on the generation of type II pneumocytes (Van Vranken et al., 2005; Samadikuchaksaraei et al., 2006; Wang et al., 2007). Fewer studies have targeted the differentiation of airway epithelial cells from pluripotent stem cells despite the fact that airway diseases such as asthma, CF, bronchitis, and bronchogenic carcinoma are, in aggregate, more prevalent than the diseases of the alveoli such as emphysema.
Prior attempts (Coraux et al., 2005; Van Haute et al., 2009) to generate functional pulmonary epithelium from ESCs were characterized by the stochastic production of limited numbers of cells and the generation of mixed cell populations, which contain undifferentiated pluripotent stem cells that carry a significant risk of teratoma formation after transplantation. Others have utilized incompletely defined media to induce respiratory cell differentiation (such as exposure of human ESCs to tumor cell extract; Roszell et al., 2009) or have used genetically modified pluripotent stem cells that were selected based upon the presence of a drug resistance gene (Wang et al., 2007). Unfortunately, genetic modulation raises the possibility of the introduction of deleterious mutations in the resulting cells.
Our overall strategy is to employ a stepwise differentiation approach that mimics the timing and coordination of the signaling pathways that guide lung development. Since there is limited data concerning the embryonic signaling pathways that regulate human lung organogenesis, we hypothesized that human lung organogenesis is, in part, analogous to mouse organogenesis. Therefore, knowledge of mouse lung development can serve as a roadmap to guide human iPSC differentiation strategies. Mechanisms that regulate mouse embryonic endoderm regionalization, lung specification, and subsequent progenitor patterning and growth have been well studied (reviewed in Morrisey and Hogan, 2010). The onset of lung specification within the endoderm is accompanied by Nkx2.1 expression and the downregulation of Sox2 along the dorsal-ventral axis of the gut tube (Lazzaro et al., 1991; Minoo et al., 1999; Que et al., 2009). Later, Sox2 expression again increases in the area of the future trachea, bronchus, and bronchioles. The Nkx2.1+Sox2+ cells in this region of the embryonic lung are airway progenitor cells that give rise to the mature airway epithelium (Que et al., 2009). In contrast, Sox9, FoxP2, and ID2 are expressed in the distal embryonic lung bud tip and mark a multi-potent embryonic lung progenitor population (Perl et al., 2005; Shu et al., 2007; Rawlins et al., 2009) capable of producing all of the cell types of the airway and alveoli. We report the differentiation of these two early progenitor cell populations from mouse and human pluripotent stem cells.
Because Nkx2.1 is the earliest marker of lung endoderm distinguishing it from the rest of the foregut endoderm, we have focused on producing Nkx2.1+ lung endoderm that is devoid of Nkx2.1+ thyroid endoderm and neuronal progenitors. A recent report using human ESCs demonstrated that a differentiation strategy based on mimicking mouse gut organogenesis led to the production of anterior foregut cells and Nkx2.1+ cells with efficiencies up to 30% (Green et al., 2011). However, these Nkx2.1+ cells were not demonstrated to be composed of purely lung progenitors. TUJ1 (a marker of neuronal tissue) and PAX8 (a marker of thyroid tissue) expression was not evaluated at the single-cell level by antibody costaining with an Nkx2.1 antibody. Development of a novel strategy to produce Nkx2.1+ progenitors that reflect only lung and not thyroid or brain differentiation is therefore needed (Watanabe et al., 2005), and the technique must be applicable to human disease-specific iPSCs in order for the technique to become widely relevant to the study of human lung disease.
Building on the aforementioned studies, we report an efficient and consistent step-wise differentiation method to generate definitive endoderm (DE), foregut endoderm, early lung endoderm, multipotent Nkx2.1+ embryonic lung progenitor cells, and airway progenitor cells beginning with mouse ESCs and subsequently with lung disease-specific human iPSCs. We show that a high dose of Activin, transient WNT activation, and staged BMP4 inhibition converts ESCs into DE with high efficiencies. Further inhibition of TGFβ alone is sufficient to “anteriorize” endoderm into Sox2+ foregut endoderm. We show that this anteriorization of the endoderm to foregut endoderm enhances the ultimate differentiation of Nkx2.1+ cells at later steps of the strategy. BMP4, FGF2, and WNT are each necessary for induction of the Nkx2.1+ immature lung progenitors. These early progenitors can mature into Nkx2.1+Sox2+ proximal progenitor cells and Nkx2.1+p63+ airway basal stem cells in vitro, and can differentiate into airway epithelium. A modification of this strategy was used to produce human disease-specific immature airway epithelium derived from CF iPSCs, creating a new platform for dissecting human lung disease.
A recent report described a monolayer-based strategy using synergistic activation of Nodal and Wnt-β-catenin signaling as well as staged BMP4 inhibition to direct mouse ESCs toward an endoderm fate with high efficiency (Sherwood et al., 2011). Here, we adapted this technology to generate DE from mouse ESCs (Figure S1A available online). This protocol resulted in a remarkable 80%–90% of cells being converted into DE based on the dual expression of the endoderm transcription factors FOXA2 and SOX17 (Figure S1B). Next, we asked if the newly generated DE cells have the competence to generate Nkx2.1+ cells. After culturing them for 2 days in serum-free medium containing BMP4, FGF2, and GSK3iXV (a WNT agonist) (Figure 1A), we observed that less than 1% of the cells were Nkx2.1+ (Figure 1A). However, more than 60% of the cells were Cdx2+, suggesting that the majority of the cells were specified to a hindgut fate (Figure 1A). Because the endoderm cells at this stage did not efficiently differentiate into Nkx2.1+ cells, we sought to test if an “anteriorization” step would facilitate lung fate specification. Snoeck and colleagues reported that Noggin (BMP inhibitor) synergized with SB431542 (a TGFβ inhibitor) to suppress a posterior endoderm fate (Cdx2+) in favor of an anterior endoderm fate (Sox2+) (Green et al., 2011). Because we had already incorporated BMP inhibition during DE generation (Figure S1A), we asked whether continuous BMP inhibition is necessary for anteriorization or if TGFβ inhibition alone was sufficient for anterior patterning. Hence, we treated endoderm with combinations of Activin (TGFβ agonist), A-83-01 (TGFβ antagonist), BMP4, and Dorsomorphin (BMP4 antagonist) for 2 days (Figure 1B). We then compared the number of FoxA2+Sox2+ anterior endoderm cells to the total number of FoxA2+ endoderm cells (Figure 1B). We also exposed the treated cells to the BMP4/FGF2/WNT agonist cocktail for another 2 days to examine their competence to generate Nkx2.1+ cells (Figure 1B). Our results revealed that TGFβ inhibition itself was sufficient to increase FoxA2+Sox2+ anterior endoderm (40%–55%, compared to <0.1% in medium alone; Figures 1B and 1C) and enhance the competence of endoderm cells to form Nkx2.1+ cells (~7%–13% as compared to <1% without regionalization; Figures 1B and 1D). We observed that continuous Activin treatment resulted in rare FoxA2+Sox2+ cells and few Nkx2.1+ cells (Figure 1B), suggesting that the optimal duration of Activin exposure was important. We also found that additional BMP4 inhibition was not required for the generation of FoxA2+Sox2+ cells and ultimately may result in slightly lower percentages of Nkx2.1+ cells (Figure 1B). We also observed less neuroectodermal marker TUJ1 if BMP4 was present between days 5 to 7 (data not shown), in agreement with the finding that BMP4 suppresses neuronal commitment and promotes nonneural lineage differentiation from ESCs (Zhang et al., 2010). FGF and WNT signaling have been shown to maintain hindgut identity and actively repress foregut endoderm differentiation (Wells and Melton, 2000; Ameri et al., 2010). Therefore, we also tested if their inhibition would enhance anteriorization together with a TGFβ inhibitor. Subsequent experiments revealed that an FGF antagonist (PD173074) induced cell death, while a WNT antagonist (IWR-1) did not increase the FoxA2+Sox2+ population (data not shown).
Nkx2.1 is not a specific marker of the lung, and its expression is also found in the thyroid and ventral forebrain (Figure S2A). Therefore, it was important to determine the identity of the Nkx2.1+ cells generated in our culture system. Unfortunately, there are no reliable markers of the lung lineage at embryonic stage E9 that are not expressed in the brain or thyroid. Surfactant protein C (Sp-C), which is the most specific marker for lung epithelial progenitors, is not detected until E10–E11 (Wert et al., 1993). SOX9 and FOXP2 are indeed coexpressed with NKX2.1 in distal lung multipotent epithelial progenitor cells (Figure S3B) and are not found in brain or newly specified thyroid (Figure S2B and data not shown), but their expression is not evident until after E11–E12. SOX2, on the other hand, is expressed in the E9 lung endoderm (Figure S2A) and later in airway epithelial progenitors (Figures S3A and S3B), but SOX2 can also be found in NKX2.1-expressing cells in ventral forebrain (Figures S2A and S2C). Thus, at the earliest stage of lung endoderm specification (E8.75–E9), none of these markers (SPC, SOX2, SOX9, or FOXP2) can be used to reliably and uniquely identify lung cells. In contrast, TUJ1 expression in Nkx2.1+ cells in ventral forebrain begins at E8.0 and continues to be present thereafter in ventral forebrain and not in thyroid or lung (Figures S2B and 2C), making Tuj1 expression a specific indicator of neuronal cell identity within the NKX2.1 domain. PAX8 expression is detected in the primordial thyroid at E8.75 at the time of thyroid specification (Figure S2B), but not in the lung or brain. Therefore, PAX8 can be used as a specific indicator of thyroid cell identity in the NKX2.1 domain (Figures S2B and S2C). In summary, we conclude that at the early specification stage, we can regard Nkx2.1+/Tuj1−/Pax8− cells as likely to be of lung cell lineage. Accordingly, we interrogated our ESC-derived Nkx2.1+ cells for TUJ1 and PAX8 expression to exclude neuronal and thyroid identity (Figure 2A). As expected for any Nkx2.1+ cell in the embryo (thyroid, lung, and ventral forebrain), all of our ex vivo differentiated Nkx2.1+ cells costained with FOXA2 (Figure 2A). We detected a small subset of Tuj1+ cells in the cultures, but these neuronal cells did not overlap with the Nkx2.1+ cells (Figure 2A). We did not detect any Pax8+ cells in our culture (Figure 2A; Figure S2B and data not shown). Therefore, this suggests that the Nkx2.1+ cells that we differentiate in vitro represent lung endoderm cells. These Nkx2.1+ cells are proliferative, with more than half expressing Ki67 (Figure 2B).
In the mouse, Sox2 is rapidly downregulated in the foregut pre-lung endoderm just as Nkx2.1 expression initiates in this same prelung endoderm during the process of lung specification at E9 (Figure S3A). Thus the very earliest lung endoderm cells are Nkx2.1+ and Sox2-low cells. Soon thereafter, sustained high-level SOX2 expression is detected in the proximal airway epithelial progenitors of the future trachea, bronchus, and bronchioles during the process of branching morphogenesis (Figures S3A and S3B). Airway progenitors are therefore Nkx2.1+Sox2+. In contrast, SOX9 and FOXP2 are expressed exclusively in the distal tip multipotent lung progenitor cells (Figure S3B), making SOX9 and FOXP2 markers that uniquely identify a distinct population of multipotent embryonic lung progenitor cells within the NKX2.1 domain because they are not present in the proximal airway progenitors or the thyroid or brain as mentioned above. Therefore, we checked if our ESC-derived Nkx2.1+ cells at day 9 contained both proximal airway (Nkx2.1+Sox2+) and distal multipotent (Nkx2.1+Sox9+; or Nkx2.1+FoxP2+) progenitor cells. We counted the Nkx2.1+ cells generated from 10 independent experiments. The average proportion of airway progenitor cells (Nkx2.1+Sox2+) was 1.8% ± 0.8% out of the total Nkx2.1+ cells, while the proportions of distal multipotent progenitor cells, Nkx2.1+Sox9+ or Nkx2.1+FoxP2+, were 4.8% ± 2.5% and 6.6% ± 3.1% out of total Nkx2.1+ cells, respectively. The remaining Nkx2.1+ cells (not thyroid or neuron are cells with minimal SOX2 expression and may reflect early prelung endoderm. Thus, the majority of Nkx2.1+ cells at day 9 likely represent an early lung endoderm that has not been specified to an airway progenitor or a distal tip multipotent lung progenitor population. Figure 2C shows representative fields of cells enriched in Nkx2.1+Sox2+, Nkx2.1+Sox9+, and Nkx2.1+FoxP2+ cells. In aggregate, these results imply that ESC-derived Nkx2.1+ cells at day 9 were immature, with most lacking proximal and distal progenitor cell markers, likely representing early lung endoderm cells. Although rare, the existence of Nkx2.1+Sox2+, Nkx2.1+Sox9+, and Nkx2.1+FoxP2+ cells suggested that the ESC-derived Nkx2.1+ lung endoderm cells were differentiating into progenitor cells of the airway and multipotent lung progenitor cells that could later be differentiated into mature epithelial cells.
The expression of lung and other anterior cell fate lineage genes including FoxN1 (thymus), Pax8 (thyroid), and Pax9 and Tbx1 (pharyngeal pouch endoderm) was analyzed by real-time qPCR (Figure S4A). As expected, the results showed that Sox2 was downregulated as pluripotent cells differentiated into DE. Later, Sox2 levels increased in accord with the expectation that anterior endoderm expresses Sox2. Nkx2.1, Sox9, and FoxP2 expression was minimal in ESCs, DE, and anteriorized endoderm, but was greatly increased (10–20 fold) after stimulation with the FGF2/WNT/BMP4 induction cocktail, consistent with previous results demonstrating these lung-specific marker combinations by antibody staining (Figure 2C). Similarly, FoxN1, Tbx1, Pax8, and Pax9 expression was low in ESCs, DE and anteriorized endoderm. However, in growth-factor-induced anterior foregut cells, we observed a modest increase in Pax9 and Tbx1 expression, whereas FoxN1 and Pax8 expression was still very minimal. Despite the increase in Pax9 and Tbx1 gene expression, we did not detect any reproducible staining for these proteins by immunofluorescence (data not shown). All of these results are consistent with our finding that a combination of FGF, BMP4, and WNT signaling predominantly drives the differentiation of lung-specific Nkx2.1+ progenitor cells from anterior endoderm.
Based on mouse genetic studies, BMP4, FGF2, and WNT signaling are each required to specify lung endoderm from the anterior foregut (Domyan et al., 2011; Goss et al., 2009; Serls et al., 2005). Therefore, we tested if we could recapitulate these findings in our in vitro mouse ESC differentiation system. To do this, we exposed the anteriorized endoderm cells to combinations of BMP4, FGF2, and WNT agonists and antagonists (Figures 3A and 3B). We found that BMP4 alone could induce NKX2.1 expression. In contrast, FGF2 or/and GSKiXV without BMP4 was not sufficient for NKX2.1 induction, suggesting that BMP4 signaling is required for lung specification (Figures 3B and 3C). Interestingly, in the presence of FGF and WNT antagonists, BMP4 could no longer induce Nkx2.1+ cells. This result implies an endogenous secretion of FGF and WNT ligands. It also suggests that FGF and WNT activation are needed for BMP4-dependent NKX2.1 specification (Figures 3B and 3C). Despite presumptive endogenous FGF signaling, an increase in FGF2 concentration upregulated Nkx2.1+ cell percentage, which is in agreement with the finding that high doses of FGF2 promote lung fate from foregut (Figures 3B and 3C; Serls et al., 2005). Compared to FGF2 signaling, exaggerated WNT activation was detrimental, with a decrease in Nkx2.1+ cells (Figures 3B and 3C) and an increase in hindgut Cdx2+ cells (data not shown), consistent with the fact that WNT signaling is known to posteriorize gut to hindgut prior to its effect on the lung (Sherwood et al., 2011).
BMP2, BMP4, and BMP7 are by far the most studied members of the BMP family. Both BMP2 and BMP4 signal through the type I receptor (ALK3), whereas BMP7 binds to a separate type I receptor (ALK2) (reviewed by von Bubnoff and Cho, 2001; Chen et al., 2004; Sieber et al., 2009; Miyazono et al., 2010). We compared the effects of BMP2 and BMP7 to BMP4 on induction of Nkx2.1+ cells from the anterior endoderm. Our results showed that BMP2 was less efficient than BMP4 at inducing Nkx2.1+ cells, while BMP7 (10 ng/ml) had no effect (Figure 4A). Even with an increased concentration (>100 ng/ml), BMP7 still failed to generate Nkx2.1+ cells (data not shown). This suggests that BMP7 signaling via the ALK2 receptor is not required for NKX2.1 induction. In canonical BMP2/4 signaling, BMP2/4 binds to BMP receptor I/II complex, leading to phosphorylation of Smad1/5/8, followed by formation of heteromeric complexes with Smad4. These complexes translocate to the nucleus and activate expression of target genes (von Bubnoff and Cho, 2001; Chen et al., 2004; Sieber et al., 2009; Miyazono et al., 2010). Besides Smad1/5/8-mediated transcription, BMP-induced receptor complexes can activate the mitogen-activated protein kinase (MAPK) pathway via ERK, JNK, or p38 (Kozawa et al., 2002). Using Dorsomorphin (a pSmad1/5/8 inhibitor) and PD98059 (a MAPKK/ERK inhibitor), we observed that Dorsomorphin completely abrogated generation of Nkx2.1+ cells, while PD98059 only partially decreased the Nkx2.1+ cell proportion (Figure 4A). We also observed significant cell death in the presence of PD98059, so we could not exclude the possibility that the decrease in Nkx2.1+ cells was due to apoptosis. In addition, Smad and MAPK-mediated pathways can be integrated (Aubin et al., 2004), and the decrease in Nkx2.1+ cells in the presence of PD98059 might also be due to a downregulation of Smad-dependent signaling. Overall, we conclude that Smad-dependent BMP2/4 signaling cascade is necessary for NKX2.1 specification from foregut cells (Figure 4B).
As we demonstrated above, most ESC-derived Nkx2.1+ cells at day 9 were immature lung endoderm cells. We therefore continued to differentiate these cells in order to produce more mature Nkx2.1+Sox2+ embryonic airway progenitors. We selectively applied signaling factors to generate airway stalk progenitors from immature lung Nkx2.1+ cells (Figures 5A and 5B). Multipotent Nkx2.1+Sox9+ tip progenitors replicate at the lung bud tip and their progeny are laid down during the pseudo-glandular phase of branching morphogenesis (E11 to E16.5) to constitute the primitive airway tree and differentiate into Nkx2.1+Sox2+ airway progenitors. BMP4 is expressed distally and BMP7 is expressed around the airway, and distal FGF2 and FGF10 are replaced by proximal FGF7 (Weaver et al., 1999; Bellusci et al., 1996; Lebeche et al., 1999; Powell et al., 1998; Morrisey and Hogan, 2010), while canonical WNT signaling is inhibited in the proximal stalk progenitors and present in the distal tip (Shu et al., 2005; Okubo and Hogan, 2004). Similarly, airway stalk cells are exposed to inhibitory Noggin while the uncommitted tip cells are exposed to BMP4 (Weaver et al., 1999; Que et al., 2006). Retinoic acid (RA) concentration is relatively higher in the stalk region than at the tips (Malpel et al., 2000). We therefore switched cells at day 9 to a “proximal induction” medium prepared with RA-supplemented B27, BMP7, FGF7, IWR-1 (WNT antagonist), and Noggin (Figure 5C, left panel). After 2–3 days, we observed an increase in Nkx2.1+ Sox2+ airway progenitor cells. The proportion was ~10% of the total number of Nkx2.1+ cells, as compared to 1%–2% before treatment. In separate experiments in which individual growth factors were removed, we found that WNT antagonism has the most pronounced effect on Nkx2.1+Sox2+ cell production. Addition of Dorsomorphin (BMP antagonist) to replace Noggin at this stage had little effect on the number of Nkx2.1+Sox2+ cells. Interestingly, the small molecule PD98059 (a MAPKK/ERK inhibitor) enhanced generation of Nkx2.1+Sox2+ airway progenitor cells to up to 18% of the total number of Nkx2.1+ cells (Figure 5C, right panel). This suggests that airway progenitors may be formed after inhibition of a MAPKK/ERK-related pathway. Most notably, we started to detect a small fraction of Nkx2.1+p63+ cells (about 1%–4% of the total number of Nkx2.1+ cells) at day 11–12 (Figure 5D). An increase in the duration of culture in the proximal induction medium from 2 to 5 or more days resulted in higher proportions of Nkx2.1+Sox2+ and Nkx2.1+p63+ cells (data not shown). Although the production of Nkx2.1+Sox2+ and Nkx2.1+p63+ cells still needs to be optimized, our results demonstrate that ESC-derived Nkx2.1+ cells can be differentiated into proximal airway progenitors and conducting airway basal cells.
Nkx2.1+ lung progenitor cells derived from pluripotent stem cells must possess the capacity to generate functional respiratory epithelium to be useful. Therefore, we tested our mouse ESC-derived Nkx2.1+ progenitors for their capacity to form mature respiratory epithelium by subcutaneous engraftment (Figure 6). Because we cannot separate committed Nkx2.1+Sox2+ airway progenitor cells from other cells, our assay evaluated the ability of Nkx2.1+ cells to differentiate within a mixed cell population. 20,000–50,000 cells were suspended in 50% Matrigel and injected under the skin of immunodeficient mice. Twenty to thirty days after injection, engrafted tissues were excised for examination. We observed that many epithelial spheres formed within grafts and that some of these spheres contained Nkx2.1-expressing cells. In some Nkx2.1+ spheres, we detected markers of mature airway epithelial cells (Figure 6), including Sox2+ proximal airway epithelial cells (Figure 6B), p63+ basal stem cells, CC10+ Clara cells, FoxJ1+ ciliated cells, and Muc5ac+ mucin-secreting cells (Figures 6C and 6D). Triple immunofluorescence staining with confocal imaging demonstrated spheres that contain more than one marker of mature airway epithelium. Such mature airway epithelial markers have never been detected in ESC-derived teratomas. Previously published basal stem cell differentiation using a 3D sphere-forming assay produced ciliated and basal cells, but not Clara cells (Rock et al., 2009). We did not detect any type I and type II pneumocyte markers such as PRO-SPC, PRO-SPA, and AQUAPORIN5 despite having distal lung multipotent cells (Nkx2.1+Sox9+ and Nkx2.1+FoxP2+) in the initial cell mixture. Overall, we concluded that ESC-derived Nkx2.1+ cell-containing populations are capable of airway epithelial cell differentiation.
Next, we examined whether a similar stepwise differentiation approach could be used to generate lung airway progenitors from CF disease-specific human iPSCs (Figure 7A). All differentiation steps were performed in xeno-free conditions. Furthermore, two of the CF iPSC lines were generated using modified RNAs (RiPSC) rather than virally encoded reprogramming factors (Warren et al., 2010) and thus were genetically unmodified, another advantage for possible future clinical use. CF iPSCs were maintained on Geltrex-coated plates in complete mTeSR1 medium. High yields of DE were obtained after treatment for 3–4 days in RPMI-1640 medium in the presence of Activin (100 ng/ml) and PI3 kinase inhibitor LY294002 (5 μM). More than 85%–90% of cells coexpressed the transcription factors SOX17 and FOXA2 at day 4, demonstrating very efficient production of DE from a RiPSC line that is compound heterozygous for CFTR mutant alleles ΔF508 and G551D (Figure 7B). This protocol was also successfully applied to other human pluripotent cells including three CF iPSC lines homozygous for the ΔF508 allele, a wild-type human BJ RiPSC line (Warren et al., 2010), and the HUES-3 and HUES-9 human ESC lines (data not shown).
By using anteriorization conditions similar to those used by Green et al., we were able to induce SOX2 expression in DE cells with efficiencies of up to 50%–60% after 4 days of treatment with A-83-01 (a TGFβ antagonist; Figure 7C). We found that Noggin (a BMP4 antagonist) was actually not necessary for this anteriorization. Similar combinations of growth factors and agonists (BMP4, FGF2, and GSK3iXV) identified in our studies on mouse ESCs could be used to generate Nkx2.1+ cells with efficiencies of 10%–30% (Figure 7D). These Nkx2.1+ cells were negative for TUJ1 and PAX8 (Figure 7D), demonstrating that they were not of neuronal or thyroid identity. In addition, subpopulations of these Nkx2.1+ cells were also positive for SOX2 and SOX9 (Figure 7D). These markers suggest the presence of both committed airway progenitors (Nkx2.1+Sox2.1 cells) and multipotent embryonic lung progenitors (Nkx2.1+Sox9+ cells).
We also quantified the expression of various anterior cell fate genes by qPCR (Figure S4) and observed very similar gene expression patterns to those seen in mouse ESC differentiation. SOX2 was downregulated after differentiation into DE and then increased as expected after anteriorization. We again observed dramatic upregulation of NKX2.1, SOX9, and FOXP2 expression (20–30 fold) in anteriorized endoderm cells after FGF2/BMP4/WNT induction. The expression of other anterior cell fate genes including PAX8 (thyroid endoderm), PAX9 (pharyngeal endoderm), and TBX1 (pharyngeal endoderm anterior to lung/esophagus) was only modestly upregulated without any expression of FOXN1 (thymus endoderm) (Figure S5). In aggregate, the differentiation spectrum of the endoderm was remarkably similar in the mouse and human platforms.
Finally, human RiPSC-derived Nkx2.1+ mixed cell populations were subcutaneously engrafted (Figure 7E). Many spheres formed in engrafted tissues after 30 days under the skin of immunodeficient recipient mice. In Nkx2.1+ spheres, some of the Nkx2.1+ cells coexpressed p63, suggesting that these Nkx2.1+ cells had matured into airway basal stem cells (Figure 7E). Further differentiation into ciliated cells, Clara cells, and mucin-secreting goblet cells was not detected, indicating that further optimization of differentiation conditions is required.
We report a stepwise developmentally guided strategy to differentiate murine pluripotent stem cells into lung multipotent progenitors and airway progenitors that produced respiratory epithelium when engrafted subcutaneously into a mouse. We then adapted our strategy to produce disease-specific lung progenitor cells from human CF iPSCs and other human pluripotent stem cell lines, thus creating a platform for dissecting human lung disease.
Additionally, we show that pluripotent stem cell systems can be used as a discovery tool to investigate steps in lung development. We were able to independently confirm recent mouse developmental genetics findings demonstrating that BMP4 is essential for lung specification, as well as similar studies that show that canonical WNT signaling is necessary for lung bud specification (Domyan et al., 2011; Goss et al., 2009). We also demonstrate that FGF signaling in addition to BMP and WNT signaling is required in the current culture system for NKX2.1 induction, an observation that has proven difficult to define in murine genetic systems due to the presence of multiple FGFs in the vicinity of the foregut lung anlage. Mechanistic dissection of the BMP4 signaling effect also revealed an observation that BMP4 signaling occurs through a Smad-dependent pathway and that pharmacologic Smad modulation enhances Nkx2.1+ lung cell differentiation. We have also shown that the carefully timed staged antagonism or agonism of the same pathway is essential to drive optimal differentiation. The premature addition of WNT signals that are later necessary for Nkx2.1+ lung cell induction at the endoderm stage, for example, regionalizes endoderm to a hindgut intestinal epithelium and prevents the differentiation of Nkx2.1+ cells (Spence et al., 2011; Cao et al., 2011; Sherwood et al., 2011). Thus, not only combinations of signaling cascades, but also their precise timing, are essential to recapitulate in vivo differentiation.
Although we report differentiation of Nkx2.1+Sox2+ proximal airway progenitors and Nkx2.1+p63+ double-positive airway stem cells derived from both human and mouse pluripotent stem cells, the proportions of Nkx2.1+ lung and airway progenitors are low. This might reflect an intrinsic bias in the differentiation of our iPSCs. It is also possible that other pathways cooperate with FGF, BMP, and WNT signaling cascades to effect Nkx2.1+ endoderm differentiation or, conversely, that other pathways need to be inhibited. Zaret and colleagues have suggested that the anterior gut contains an epigenetic prepattern that regulates the ability of gut endoderm to differentiate into a particular tissue despite the presence of the correct growth factors (Zaret et al., 2008). Perhaps many of the iPSC-derived foregut endoderm cells do not possess the epigenetic prepattern that would allow them to differentiate into Nkx2.1 cells. Additionally, it must be determined whether Nkx2.1+ progenitor cells expanded in culture undergo modifications that permit the expression of thyroid or neuronal genes. It will be important to investigate each of these possibilities so that we can rationally increase the efficiency of the differentiation protocols described herein by overcoming the relevant mechanistic hurdles. Ultimately, we will need protocols that will be effective for any human iPSC line and reagents to be able to sort pure populations of human cells in the absence of genetic reporters or selectable drug resistance genes. Our strategy can then in principle be applied to any human lung disorder.
Finally, conditions must be found to affect the maturation of iPSC-derived lung epithelium ex vivo. To study an airways disease like CF and to screen for small molecule therapeutics using iPSC-derived cells, we will require an Nkx2.1+ epithelium that contains all of the appropriate human cellular constituents. That same epithelium must display functional coordinated ciliary beating, epithelial secretory function (including regulated CFTR-mediated chloride ion transport and mucus secretion), and an airway stem cell-driven capacity for epithelial regeneration.
NOD-SCID IL2Rgamma null (NOD/SCIDIl2rg−/−) immunodeficient mice were purchased from Jackson Laboratory. Animal procedures were performed in accordance with Massachusetts General Hospital (MGH) and national guidelines and regulations and were approved by the Institutional Animal Care and Use Committee (IACUC) for MGH. The use of human CF induced-pluripotent cell lines and embryonic cell lines (HUES-3 and HUES-9) was reviewed and approved by the Embryonic Stem Cell Research Oversight committee (ESCRO) and IRB at MGH.
Mouse DE was generated as described in Supplemental Experimental Procedures. To anteriorize endoderm, on and after day 5, the cells were split and reseeded on to the plates precoated with 804G-conditioned medium and fed with D0 medium + 0.5–1 μM A8301 (CalBiochem, 616454) for 2 days. Then the medium was rinsed with D0 medium two times and switched to D0 medium supplemented with 50 ng/ml BMP4, 100 ng/ml FGF2 (GIBCO, PHG0026), and 5–10 nM GSK3iXV for another 2–3 days. To generate Nkx2.1+Sox2+ proximal progenitor cells, the cells were rinsed with D0 medium two times and switched to D0 medium containing RA-supplemented B27, 50 ng/ml BMP7, 50 ng/ml FGF7, 100 nM IWR-1, and 1–2 μM PD98059 for 2 days or longer.
Human iPSCs were maintained on Geltrex-coated plates in complete mTeSR1 medium (Stemcell). High yields of DE progenitors were obtained after treatment for 3–4 days in RPMI-1640 medium in the presence of 2% B27 supplement minus vitamin A (GIBCO, 12587-010), Activin A (Peprotech, 100 ng/ml) and PI3 Kinase inhibitor LY294002 (5 μM). To generate foregut endoderm cells, DE was treated for 4 days with RPMI-1640 medium containing 2% B27 and 500 nM A-83-01 (TGFβ antagonist) with or without Noggin (BMP4 antagonist, 100 ng/ml) or Dorsomorphin (2–5 μM). After that, the cells were exposed for 4 days or longer to RPMI-1640 medium containing 2% B27, 50 ng/ml BMP4, 100 ng/ml FGF2, and 5 nM GSK3iXV for NKX2.1 induction from endoderm cells.
At each differentiation step, the cells were fixed with fresh paraformaldehyde (4%) for 15 min at room temperature, rinsed in PBS, washed with PBS + 0.2% Triton X-100, and incubated with the primary antibodies at 4°C overnight (>16 hr) diluted in PBS + 1% BSA. Following incubation, the cells were rinsed four times in PBS + 0.2% Triton X-100, and incubated with secondary antibodies at room temperature for 2 hr. The images were visualized using an Olympus IX71 inverted fluorescence microscope or a Nikon A1 Confocal Laser microscope. The primary antibodies used are summarized in Table S2. The secondary antibodies were purchased from Invitrogen (Alexa Fluor 549 and Alexa Fluor 488). The quantification was performed by counting at least five random fields at 20× magnification and calculating the average and standard deviation.
The embryos or lungs at desired embryonic stages were dissected out and fixed at 4°C for 6 hr or overnight with fresh 4% PFA. Then tissues were rinsed with PBS for 3 min (5 min per time) and incubated in 30% sucrose in PBS at 4°C overnight. The tissues were soaked in OCT for 1 hr and then frozen in OCT for cryosectioning at 7 μm thickness. The slides were stained with the primary and secondary antibodies as described above.
Anywhere from 2× 105 to 5× 105 cells were suspended in 200 μl 1:1 mixture of Growth Factor Reduced Matrigel (BD Biosciences, 354230) and Advanced DMEM and were injected subcutaneously into NOD/SCIDIl2rg−/− null mice. The tissues were harvested and examined after 20–30 days, fixed overnight in cold 4% paraformaldehyde, rinsed two times with PBS, soaked in 30% sucrose for 2–3 hr, embedded in OCT, and sectioned at 7 μm. The slides were examined for airway epithelial cells based on NKX2.1 expression. The differentiation of ESC-derived Nkx2.1+ cells into basal stem cells was based on expression of P63 and CK5; goblets cells, based on expression of MUC5AC; Clara cells, based on expression of CCSP; and ciliated cells, based on the expression of FOXJ1.
For more information on how to culture these cells, please refer to the Rajagopal Lab website: http://www.massgeneral.org/regenmed/staff/Rajagopallab. We intend to deposit the reported cells in the Harvard Stem Cell Institute iPS Cell Core Facility (please see http://www.hsci.harvard.edu/ipscore/node/3).
The work in this manuscript was supported by a Harvard Stem Cell Institute Seed Grant and an NIH-NHLBI Early Career Research New Faculty (P30) award to J.R. We thank Drs. William Anderson and Darrell Kotton for reading this manuscript and for their critical comments. We also wish to extend our thanks to all of the members of the Rajagopal Laboratory for their reading of the text, their constructive criticisms and comments, and valuable discussion and support.