A distal lung organoid model to study interstitial lung disease, viral infection and human lung development

Abstract

Organoids have been an exciting advancement in stem cell research. Here we describe a strategy for directed differentiation of human pluripotent stem cells into distal lung organoids. This protocol recapitulates lung development by sequentially specifying human pluripotent stem cells to definitive endoderm, anterior foregut endoderm, ventral anterior foregut endoderm, lung bud organoids and finally lung organoids. The organoids take ~40 d to generate and can be maintained more than 180 d, while progressively maturing up to a stage consistent with the second trimester of human gestation. They are unique because of their branching morphology, the near absence of non-lung endodermal lineages, presence of mesenchyme and capacity to recapitulate interstitial lung diseases. This protocol can be performed by anyone familiar with cell culture techniques, is conducted in serum-free conditions and does not require lineage-specific reporters or enrichment steps. We also provide a protocol for the generation of single-cell suspensions for single-cell RNA sequencing.

Introduction

Organoids are used to study human development and to model disease^1,2. Among others, major progress has been made in the generation of lung organoids^3,4. The lung epithelium contains multiple cell types, including basal, ciliated, secretory, goblet and neuroendocrine cells in the airways, and alveolar epithelial type 1 (AT1) and surfactant-producing alveolar epithelial type 2 (AT2) cells in the alveoli, where gas exchange takes place^4,5. In addition, several novel transitional cell populations have been identified in the human distal lung respiratory airway or terminal respiratory bronchioles, very small airways in which alveoli empty and that contain more extensive alveoli at their distal tips^6–8.

The respiratory system originates from buds that arise on the ventral aspect of the anterior foregut endoderm (AFE) during embryonic stage (E9.5–E12.5 in the mouse, 4–7 post-conception weeks (pcw) in humans). These undergo a stereotyped branching process during the pseudoglandular stage (E12.5–E16.5 in the mouse, 5–17 pcw in humans). During the canalicular stage (E16.5–E17.5 in the mouse, 16–26 pcw in humans) cell cycle activity decreases and specialization of the airway epithelium occurs in the stalks, with the emergence of basal, goblet, club, ciliated and other cell types. This stage is followed by the saccular stage, where the canaliculi widen into sacculations that will give rise to primitive alveoli (E17.5 to birth in the mouse, 26–38 pcw in humans)^5,9–11. Expansion of alveolar number by further differentiation of immature saccules, alveolar maturation and secondary septation continue predominantly postnatally (alveolar stage, postnatal days 0–21 (d0-21) in mice, 38 pcw to the age of 21 years in humans)^11,12. Generating organoids that at least partially model human lung development requires recapitulating this process to the extent possible in vitro.

Since the first reports on their establishment from human pluripotent stem cells (hPSCs)^13,14, several publications showed the generation of lung organoids, either from fetal or adult lung^15–18 or from hPSCs^19–26. We developed a directed differentiation strategy of hPSCs (both embryonic stem cells (ESCs) and induced pluripotent stem cells (iPSCs)) into three-dimensional (3D) lung organoids that undergo a process similar to branching morphogenesis and that we have used to model fibrotic interstitial lung diseases (ILDs) and viral infection^20,27,28.

This protocol (Fig. 1a) begins by replicating early lung development through successive specification of definitive endoderm (DE) and AFE according to protocols we established previously (Fig. 1b, i–iii, d4.3, d6.3)^20,29,30. Subsequently, small organoids that are endowed with the expression patterns associated with lung buds in vivo (hence the name ‘lung bud organoids’ or LBOs, Fig. 1b, iv–vi, d8, d9, d15) are generated in suspension culture²⁰. These LBOs can give rise to lung and airway structures after transplantation under the kidney capsule of immunodeficient mice or can be embedded in Matrigel²⁰, where a process similar to branching morphogenesis ensues and organoids arise that expand more than 6 months (Fig. 1b, vii, viii, d42, d70). Genome-wide expression analysis and benchmarking against the Keygenes database³¹, which contains tissues from the first and second trimester of human gestation and of adult humans, revealed that organoids at d170 corresponded best to lung in the second trimester of human gestation²⁰. At these later stages of the cultures, AT2 cells arise at the tips. These contain lamellar bodies (LBs) by transmission electron microscopy and can take up and secrete fluorescently labeled surfactant protein B (SFTPB)²⁰, features specific to AT2 cells. Mesenchyme appears during the anteriorization and ventralization stages and becomes scarce after development in Matrigel²⁰. This may correspond to the attrition of mesenchyme during lung development³².

Fig. 1 |. Overview of the protocol.

a, Schematic representation of the distinct stages of the protocol. b, Representative brightfield images at indicated days of the protocol. Scale bars, 200 μm.

Applications of the method

The most unique feature of the organoids described here is the presence of mesenchyme that allows studies of epithelial–mesenchymal interactions and therefore recapitulation and modeling of at least some ILDs. ILDs are a heterogeneous group of entities with diverse or unknown etiologies affecting the lung interstitium³³. Development of pulmonary fibrosis (PF), most commonly associated with idiopathic pulmonary fibrosis (IPF), portends the worst prognosis^33,34. Injury to AT2 cells probably plays a critical role in the development of both familial and sporadic IPF, as mutations causing familial IPF specifically affect AT2 function, while depletion of AT2 cells in mouse models is associated with fibrosis³⁵. In addition, a fraction of patients with familial IPF have heterozygous mutations in telomerase components or other genes involved in telomere maintenance^36–43. Furthermore, several susceptibility loci have been identified that affect telomere length⁴⁴. As telomeropathy affects stem cell function and AT2 cells are facultative distal lung stem cells^4,45, these associations suggest a role for AT2 cells.

Finding therapeutic targets and novel drug treatments requires models to dissect pathogenetic mechanisms, identify biomarkers, discover drug targets and perform drug testing. It has been challenging to model ILDs in animal models^{34,35,46–48}. A complementary, more malleable, but also more reductionist avenue are lung organoids derived from hPSCs. We showed that PF caused by mutation in Hermansky–Pudlak syndrome (HPS) genes could be modeled²⁸. HPS is a genetic syndrome caused by abnormal biogenesis and trafficking of lysosome-related organelles and characterized by hypopigmentation, bleeding diathesis and in some but not all genotypes, PF, also known as HPS-associated interstitial pneumonia (HPSIP). Fibrosis in HPSIP is most probably linked to the fact that homeostasis of LBs in AT2 cells is abnormal and surfactant secretion is reduced, as LBs are also lysosome-related organelles⁴⁹. Of the nine genotypes, only three (HPS1, 2 and 4) are clinically associated with HPSIP. Accordingly, we found that only organoids derived from ESCs with mutation in HPS1, 2 or 4 showed fibrotic changes, whereas organoids with a mutation in HPS8, which is clinically not associated with fibrosis, did not²⁸. This model therefore recapitulates the clinical incidence of PF in HPS. By comparing genome-wide expression data from epithelial cells isolated from lung organoids generated from the fibrosis prone (HPS1^−/− and HPS2^−/−) and non-fibrosis-prone (HPS8^−/−) mutants, we furthermore identified interleukin-11 as a therapeutic target critical to fibrosis. This provides proof-of-principle evidence that pathogenetic mechanism and therapeutic targets can be identified using lung organoids with mutations associated with diseases²⁸.

The organoids can also be used to model viral infection. They can be infected by respiratory viruses, such as respiratory syncytial virus, parainfluenza and measles. We showed that the organoids are authentic infection models that do not promote the emergence of culture-adapted variants in the propagation of parainfluenza²⁷. In the case of respiratory syncytial virus, we could show that infected epithelial cells are dislodged into the lumen, as is observed in bronchiolitis caused by this virus⁴⁹.

The organoids can also be used for modeling human lung development. As an example, single-cell RNA sequencing (scRNAseq) and confocal analysis is presented in the ‘Anticipated results’ section, showing progressive loss of other endodermal fates present early in organoid development, and reciprocal acquisition of a distal lung fate. This model can therefore be used to examine relevant signaling pathways as well as transcriptional and epigenetic regulation of the progressive specification of distal lung. The fact that undifferentiated endoderm remains present in organoids early in development suggests that it is possible that by changing the culture conditions, other endodermal fates could be obtained, which may constitute another potential application of this model. The spontaneous formation of branching structures also suggest that this model could be used to study branching morphogenesis in the lung.

Comparison with other methods

Since the first reports on their establishment from hPSCs^13,14, several publications showed the generation of lung organoids, either from fetal or adult lung^15–18 or from hPSCs^19–26. In many cases, the structures are limited in size or contain, often by design, specific cell types (airway^23,26,50 or AT2 cells^19,24,51). The lung organoids described here present several unique features compared with others as a model of human distal lung development.

Branching morphogenesis

A first feature is that they undergo a process similar to branching morphogenesis and that the full potential to generate lung and airways is present in the early organoids. While differentiation is biased toward distal lung in the conditions we use in vitro (described here in this protocol), the early-stage lung organoids (LBOs in suspension) generate structures containing branching tubules with proximodistal specification when embedded in donor-derived mesenchyme after transplantation under the kidney capsule of immunodeficient mice²⁰. Another group (the protocol described in Miller et al.²²) also reported generation of 3D human lung organoids from hPSCs^13,14. While these smaller structures contained cells expressing markers of lung and airway¹⁴ and have some airway potential after subcutaneous xenografting in mice¹³, they did not show branching structures. However, these organoids were similar to our reported organoids, staged as early fetal lung²².

Pulmonary mesenchyme

Another notable feature of the distal lung organoids described here is the presence of mesenchyme, the origin of which is unclear. Mesenchyme could already be detected during the AFE stage of the protocol, remains present and relatively abundant during the LBO (suspension) stage, and becomes scarce, but still detectable after development to at least d60 in Matrigel²⁰. The mesenchymal cells can be isolated by cell sorting for EPCAM⁻THY⁺ cells²⁸. RNAseq²⁰ as well as scRNAseq (see ‘Anticipated results’) showed that this mesenchyme expresses genes associated with pulmonary mesenchyme (including WTN2B⁵² and TBX4 (ref. ⁵³)). The presence of mesenchyme is shared with the protocol of Miller et al., although they did not further characterize the mesenchyme²². The presence of mesenchyme in our organoids may explain what is probably the most unique application of these organoids, the capacity to recapitulate at least one ILDs: HPSIP. Together, these findings indicate that these distal lung organoids can represent an integrated model of distal lung development and disease.

Differentiation efficiency

A final distinguishing feature of the organoids described here is that any intervening purification steps are unnecessary. The in vitro generation of NKX2.1⁺ lung progenitors with high efficiency is a critical but challenging step. To resolve this issue, several groups have developed strategies based on lineage-specific reporters to isolate these before further differentiation and maturation. In two reports from the same laboratory, McCauley et al.⁵⁴ specified lung progenitors, and isolated NKX2.1^GFP+ cells at d15 for further culture in Matrigel 3D medium to yield proximal airway spheroids devoid of distal potential, while Jacob et al.¹⁹ resorted to a NKX2.1^GFPSFTPC^tdTomato reporter cell line to isolate NKX2.1^GFP+SFTPC^tdTomato+ cells for further culture in Matrigel to generate spheroids containing AT2 cells. While these protocols are advantageous when differentiation efficiency is modest, they only allow for studies using these specific reporter lines. To resolve this problem, surface markers were identified by which NKX2.1⁺ lung progenitors can be isolated from the cultures. Gotoh et al. reported carboxypeptidase M (CPM) as a specific marker of NKX2.1⁺ progenitors²⁴, which was then used by Konishi et al.²⁶ to select for NKX2.1⁺ cells. Hawkins et al. used differential expression of CD47 and CD26 to show that CD47^hiCD26^lo cells are enriched in NKX2.1⁺ lung progenitors⁵⁵. However, a rigorous head-to-head comparison with reporter lines has not been performed.

The reasons why neither reporter cells nor cell sorting are required in our protocol are unclear. We speculate that some differences in our protocol may offer an explanation. We grow hPSCs in the presence of feeders, as we have found that passage in feeder-free conditions compromises future lung potential. Second, although commercial kits to generate DE have been recommended and are widely used⁵⁶, we found that, while DE generation was indeed efficient, the lung potential of the DE generated was suboptimal. We therefore adhere to a strict protocol to generate DE using reagents that have to be lot tested. We also generate DE in 3D embryoid bodies, and not in two-dimensional (2D) cultures. Finally, the LBO suspension culture step may select for cells specified to pulmonary epithelium and mesenchyme. The same may be true for the Matrigel step. Indeed, scRNAseq analysis suggests progressive acquisition of a distal lung at the expense of other endodermal fates over the first 80 d of the protocol (see ‘Anticipated results’).

A serum-free model

The published protocol that is the most closely related to the protocol described here was by Miller et al.²². However, the factors used in the latter differ from ours. In the specification of AFE they also use, in addition to BMP and TGF-β inhibition, Wnt agonism, FGF4 and Smoothened agonist. They do not include an extended period of suspension culture (which in the protocol described here generates LBOs). Furthermore, and importantly, they use fetal bovine serum (FBS) in their base medium, which we try to avoid as this is an uncharacterized reagent. Miller et al. do obtain proximal and some distal lung specification in the presence of serum, and in conditions similar to ours (but without BMP4 and FGF10) they obtain small spherical ‘bud tip’ organoids. The organoids we describe here do not develop in the presence of serum (Fig. 2).

Fig. 2 |. FBS-containing media does not maintain organoid cultures.

Tile scan of brightfield images showing the morphological changes of a d47 organoid switched to FBS-containing media. Scale bars, 1 mm.

Development and disease modeling

As mentioned before, the organoids of Miller et al. are much smaller and do not show branching. Commonalities between both types of organoids are the capacity to generate lung and airway structures after transplantation under the kidney capsule of immunodeficient mice, the presence of mesenchyme and the fact that repeated enrichment steps are not required. In contrast to the current protocol, that of Miller et al. has not been reported to be useful in modeling ILDs, and was not reported to be infectable by the viruses we used^20,27. Both models will probably serve complementary goals. The capacity to generate airway cells in the protocol of Miller et al., for example, may allow studies on airway development and proximodistal specification, whereas the organoids reported here will allow investigation of branching morphogenesis and alveolar specification, in addition to modeling ILDs.

We have also previously reported on a protocol in Collagen I (Col I) gels that allows further maturation of lung cells in 3D^21,57. This Col I protocol and the resulting organoids are distinct from the protocol described here. Rather than generating 3D LBOs in suspension culture, the Col I protocol continues lung progenitor specification in 2D, followed by plating into Col I gels. The Col I model is, in contrast to the organoids described here, permissive for the specification of all distal and airway lineages after retraction of CHIR99021 (CHIR), including cells compatible with AT1 cells. This finding is consistent with the observations of Kanagaki et al. in fibroblast-dependent alveolar organoids⁵⁸. However, retraction of CHIR in the distal lung organoids described here results in cell death and disintegration. In vitro, we were therefore not able to achieve induction of AT1 markers in lung organoids generated using the protocol described here. However, we do note that after transplantation under the kidney capsule, LBOs could generate cells expressing AT1 markers²⁰, suggesting that the potential exists, but that the correct in vitro conditions for AT1 generation have not yet been identified for these organoids.

In addition, while in most studies the maturity of the generated cells goes unreported or is equivalent to the second trimester of human gestation²¹, the cells generated by our Col I protocol exhibit transcriptomic profiles, ultrastructure and an array of lineage-specific markers that also partially match postnatal lungs²¹. Several mature cell types can be isolated using surface markers. Most notably, the Col I protocol allowed the generation NGFR⁺ basal cells, the stem cells of the postnatal airway⁵⁹, which can be isolated and further expanded using established techniques⁶⁰. On the other hand, and in contrast to the model described here, modeling fibrotic lung disease was not possible, and mesenchymal cells were not present. Both models are therefore complementary. The reason why both protocols behave differently is most likely linked to differences in the early stages of directed differentiation and the different extracellular matrices used to embed the organoids. The underlying mechanisms merit further investigation.

Limitations

Even after 6 months of culture in Matrigel, the organoid model matches the second trimester of human gestation on the basis of genome-wide expression analysis. The expression of SFTPC, a marker of mature AT2 cells occurs late in the culture (d170) and is not universal (Fig. 3a), although all other AT2 markers studied (SFTPB, ABCA3, LPCAT, NAPSA, SLC34A2) reproducibly and progressively increase and are consistently expressed by a majority of the cells by d80 (see ‘Anticipated results’). We have previously shown the presence of LBs and capacity to take up and release SFTPB in organoids >d120 old, indicating long-term maintenance of still immature AT2 cells²⁰. The combination of CHIR, FGF7 and dexamethasone, cAMP and IBMX has been reported to be necessary for the expansion and maintenance of so-called alveolospheres generated from hPSC-derived lung progenitors. Even here, however, maintaining SFTPC expression was challenging and required repeated sorting for a reporter, as well switching the culture conditions between presence and absence of CHIR^19,56. We also note that in the organoid protocol of Miller et al.²², SFTPC expression was rare.

Fig. 3 |. Maturation of organoids and selection of suspension organoids for embedding.

a, Induction of select AT2 markers in RNAseq studies (data from ref. ²⁰, mean ± s.e.m.). RPKM, reads per kilobase of transcript per million mapped reads. b, Representative images of d20 LBOs. The upper row shows organoids that we prefer not to pick due to the dense center and hollowing structures. The lower row shows ‘good’ organoids with folding epithelial structures and clear and transparent center.

Even though we observed a process reminiscent of branching morphogenesis, the branching is not well organized. The structures branch randomly in every direction and appeared disorganized, probably because neither directional cues nor physical constraints, such as a thoracic cavity, are provided. In addition, the exact nature and patterning of the mesenchyme present is unclear and the proportion declines along the culture time. The mesenchyme does not have the potential to generate endothelial cells in vivo after transplantation under the kidney capsule, although smooth muscle and occasional cartilage were present²⁰. Endothelial cells or blood vessels have also not been observed in vitro in our model. Furthermore, we could not achieve induction of AT1 markers in vitro, although AT1 markers were observed after engraftment in vivo²⁰. Terminal and architecturally fully appropriate maturation of organoids is still a challenge for the field in general. Finally, the use of Matrigel to embed LBOs and serum to inactive enzymatic activities in this model is still a barrier for translational research. The use of Matrigel might be replaced with synthetic hydrogels. However, when LBOs, generated using the protocol described here, were embedded in Col I gels, they did not show branching (Fig. 4). We do not know whether this is due to the molecular composition or the stiffness of Matrigel that allows branching morphogenesis of LBOs. Hence, the choice of 3D medium for embedding should take these two factors into consideration. The use of serum to inactivate enzymes in the protocol can be easily substituted with purified protease inhibitors.

Fig. 4 |. Morphology of lung organoids embedded in Col I gel.

Tile scan of a brightfield image showing the morphology of d50 organoids grown in the Col I gel. Scale bar, 1 mm.

Experimental design

A schematic of the protocol with examples of the typical morphologies at each stage is shown in Fig. 1. We maintain hPSCs on feeder cells. The protocol begins with mouse embryonic fibroblast (MEF) depletion (d1, Fig. 1b, i) followed by a stepwise, sequential induction of DE (d0–4, Fig. 1b, ii) and AFE (d5–6, Fig. 1b, iii) as previously described^20,61. hPSCs are passaged on Matrigel-coated plates to deplete residual MEFs for ~24 h (Steps 1–13). After MEF depletion, primitive streak and embryoid body induction is performed in embryoid bodies/primitive streak (EB/PS) medium in ultralow-attachment plates for 12–16 h (Steps 14–27). This is followed by switching to DE induction medium for another 76–80 h (d4; Steps 28–42). Essential in this step is exposure to high concentrations of Activin A. DE can be generated both in 2D adherent cultures and or in 3D as EBs in low-attachment plates. We found that lung potential is maintained the best in EBs. DE purity is verified by flow cytometry after staining for EPCAM, cKIT and CXCR4 (Steps 43–66)^20,61. It is essential that more than 90% of the cells express high levels of these three markers. Others have recommended the use of commercial kits to generate DE⁵⁶. We found that while generation of DE was efficient, the potential of these DE cells to generate lung and airway cells was more limited.

High-purity DE (>90%) is directed to AFE via inhibition of TGF-β, BMP and WNT signaling by SB 431542, NOGGIN and IWP2, respectively (Steps 67–89)^29,30,61. At the end of AFE induction (d6), cells are treated with ventralization medium in the presence of BMP4, FGF10, FGF7, all-trans retinoic acid (RA) and the glycogen synthase kinase 3β antagonist, CHIR for 48 h (Steps 90–92). At this time, formation of loosely adherent clumps can be observed (Fig. 1b, iv). Next, LBOs are generated (Steps 93–99). This is accomplished by resuspending clumps of cells at d8 and maintaining those in ultra-low-attachment plates in the presence of CHIR, FGF10, FGF7, BMP4 and RA, where they will form spheres consisting of folding epithelial sheets interspersed with mesenchymal cells (Fig. 1b, v, vi and Fig. 3b). The majority of the epithelial cells will express early, but not mature lung markers, and have an expression profile consistent with lung bud epithelium. The mesenchymal cells express markers consistent with pulmonary mesenchyme, and can make up 5–20% of the population by flow cytometry²⁰.

In the final stage (Steps 100–117), the LBOs are plated in Matrigel with CHIR, FGF10, FGF7, BMP4 and RA, where they undergo outward budding followed by a process similar to branching morphogenesis (Fig. 1b, vii, viii). Other alternatives can be used, such as Cultrex (R&D Systems), which in our hands gave similar results. We recommend growth factor-reduced Matrigel, as multiple factors at various concentrations may otherwise be present that could unpredictably affect organoid development, and therefore reproducibility. The mesenchyme is gradually lost after embedding in Matrigel. Approximately 90% of LBOs will yield rapidly expanding branching structures in which the fraction cells expressing a constellation of AT2 markers gradually increases, such that by d80, these form the majority of the cells (see ‘Anticipated results’). The organoids can be maintained in culture for more than 6 months. The entire protocol is conducted serum-free conditions.

We found that several medium constituents need to be lot tested. These include Activin A, N2 and B27 medium, and Knockout Serum Replacement (see ‘Reagents’). Lot testing involves comparative analysis for maximal efficiency of DE generation as evaluated by expression of EPCAM, cKIT and CXCR4 by flow cytometry. We found lot testing of Matrigel for organoid formation not necessary.

The cultures can be analyzed at any step of the differentiation using several approaches, including immunofluorescence (IF; Box 1), reverse transcription quantitative polymerase chain reaction (RT–qPCR; Box 1), electron microscopy, RNAseq and western blot, as we published previously^20,27,28. We provide an example of IF (Fig. 5, antibodies in Table 1) and RT–qPCR analysis (Fig. 6, primer sequences in Table 2). Additionally, the cultures can be analyzed by scRNAseq (Boxes 2 and 3). As the quality of the input single-cell suspension is critical to scRNAseq, we provide a protocol to generate single-cell suspension from organoids in Matrigel (Box 2). In the ‘Anticipated results’ section, we provide an example of scRNAseq analysis (Figs. 7 and 8).

Box 1 |. Standard methods to analyze lung organoids.

RT–qPCR ● Timing 4 h (hands on, 2 h)

Reagents

• Direct-zol RNA MicroPrep Kit (Zymo, cat. no. R2060)

• TRIzol (Fisher Scientific, cat. no. 15-596-018)

• Ethanol (Sigma-Aldrich, cat. no. E7023-500ML)

• qScript XLT cDNA SuperMix (Quantabio, cat. no. 95161–100)

• SYBR Green PCR Master Mix (Fisher Scientific, cat. no. 43–687–08)

• Nuclease free water (Ambion, cat. no. AM9937)

Disposables

• 1.5 ml RNase-/DNase-free Microcentrifuge tube (Nest Biotechnology, cat. no. 615601)

• 0.2 ml PCR tubes (Fisher Scientific, cat. no. NC9989922)

• 384-well reaction plate with barcode (Applied Biosystems, cat. no. 4309849)

Equipment

• Vortex mixer (Fisher Scientific, cat. no. 02215365)

• Centrifuge (Eppendorf, cat. no. 5424R)

• Thermal cycler T100 (Biorad, cat. no. 1861096)

• Quant Studio Real-time PCR System (Applied Biosystems, cat. no. A28567)

Procedure

1. Aspirate the medium from the insert with lung organoid and add 300 μl of Trizol to the insert using a P1000. Pipette up and down several times to break the Matrigel and collect the content of the insert. Transfer the organoid in trizol into a 1.5 ml tube and vortex until homogenized.

2. Centrifuge the sample at 16,000g for 5 min at 4 °C to remove the residual Matrigel. Take the supernatant and mix with equal amount of 100% ethanol and vortex thoroughly.

3. Transfer the mixture into a Zymo-Spin IC Column in a Collection Tube and centrifuge for 30 s at 16,000g. Transfer the column into a new collection tube and discard the flow through.

4. In an RNase-free tube, add 5 μl DNase I (6 U/μl), 35 μl DNA Digestion Buffer and mix by gentle inversion. Add the mix directly to the column matrix. Incubate at room temperature (20–30 °C) for 15 min.

5. Add 400 μl Direct-zol RNA PreWash to the column and centrifuge. Discard the flow through and repeat this step.

6. Add 700 μl RNA Wash Buffer to the column and centrifuge for 1 min to ensure complete removal of the wash buffer. Transfer the column carefully into an RNase-free tube.

7. To elute RNA, add 15 μl of DNase-/RNase-free water directly to the column matrix and centrifuge.

8. Reverse transcribe 100–1,000 ng of total RNA using the qScript XLT cDNA SuperMix. Combine following reagents in 0.2 ml microtubes:

   Component	Volume for 20 μl reaction
   qScript XLT cDNA SuperMix	4 μl
   RNA template	Variable
   RNase-/DNase-free water	Variable
   Total volume	20 μl

9. After sealing each reaction, vortex gently to mix contents. Centrifuge briefly to collect components at the bottom of the reaction tube.

10. Incubate as follows: 5 min at 25 °C, 60 min at 42 °C, 5 min at 85 °C, hold at 4 °C. Use first-strand product for RT–qPCR amplification.

11. For RT–qPCR, prepare technical triplicates of 15 μl reaction:

    Component	Volume for 15 μl reaction
    Sybr green	7.5 μl
    Forward primer (from 10 μM stock)	0.15 μl
    Reverse primer (from 10 μM stock)	0.15 μl
    cDNA	Variable
    Water	Variable
    Total volume	15 μl

12. Prepare samples in a 384-well plate and run for 40 cycles. Use primers listed in Table 1. Calculate relative gene expression on the basis of the average cycle (Ct) value, normalized to GAPDH as the internal control and reported as fold change (2(-ddCT)).

IF ● Timing 2 d (hands on, 6–8 h)

Reagents

• Paraformaldehyde 16% solution, EM grade (Electron Microscopy Sciences, cat. no. 15710)

• Glycine (Sigma-Aldrich, cat. no. G7126-100G)

• PBS without Ca²⁺ and Mg²⁺ (Cellgro, cat. no. MT21031CM)

• Triton X-100 (Sigma-Aldrich, cat. no. X100-100ML)

• Donkey serum (EMD Milipore, cat. no. S30-100ML)

• DAPI (Thermo Fisher, cat. no. D1306)

• Mounting reagent (OriGene Technologies, cat. no. E19-18)

• Dry ice

Disposables

• 24-Well insert (Falcon, cat. no. 353095)

• 24-Well plate (Falcon, cat. no. 353047)

• Cryo mold (Electron Microscopy Sciences, cat. no. 62534–15)

• Microtome blade MB35 Premier (Thermo Scientific, cat. no. 3050835)

• Adhesion microscope slides (Epredia, cat. no. 9991003)

• Coverslips (Electron Microscopy Sciences, cat. no. 72290–06)

Equipment

• Forceps (Dumont, cat. no. 11251–33)

• Cryotome (Leica, cat. no. CM 3050-S)

Reagent setup

• 0.2% PBST: to make 100 ml of PBST mix 0.2 ml of 100% Triton with 99.8 ml of PBS.

• 4% paraformaldehyde: to make 40 ml of 4% paraformaldehyde mix 10 ml of 16% paraformaldehyde and 30 ml of PBS.

• 50 mM glycine: for 2.5 M glycine stock solution mix 93.8 g of glycine powder in 500 ml of water. For 100 ml of 50 mM solution mix 2 ml of 2.5 M stock solution with 98 ml of water.

• 2% donkey serum in 0.2% PBST: for 100 ml of the solution mix 2 ml of 100% donkey serum with 98 ml of 0.2% PBST.

• 5% donkey serum in PBS: for 100 ml of the solution mix 5 ml of donkey serum with 95 ml of PBS.

Procedure

1. Aspirate the medium and remove the insert from the 24-well plate. Invert it and gently peel off the membrane from the insert using forceps.

2. Excise the lung organoid and embed in a cryo-mold on dry ice.

3. Cut frozen samples on a cryotome at the thickness of 10–12 μm and collect on adhesion microscope slides and air dry.

4. Fix the samples in 4% paraformaldehyde for 20 min, wash two times for 5 min in 50 mM glycine to inactivate the PFA, followed by washing in PBS.

5. Permeabilize samples for 10 min in 0.2% PBST (PBS + 0.2% Triton X-100) and block by incubating in PBS containing 5% donkey serum for 1 h, then incubate overnight in primary antibody in 0.2% Triton X-100 and 2% donkey serum (for primary antibodies and suggested dilutions, see Table 1).

6. The next day, wash the samples three times in PBS and 1% donkey serum and incubate with secondary antibody (1:200) for 1 h at the room temperature (for secondary antibodies, see Table 1). Stain nuclei with DAPI (1:500).

7. Mount sections with Mounting Reagent and coverslip. Slides are ready for the imaging when the mounting reagent dries properly.

Fig. 5 |. IF analysis of organoids at different stages.

a, Tile scan image at d8 (ventralization) stained for indicated markers. b, Tile scan images and higher magnification fields (insets) of culture LBOs at d20 of the culture protocol. c–e, Confocal tile scans of ~d60 organoids stained for the indicated markers, and higher magnification images (right) corresponding to numbered boxes. All scale bars in a–e, 500 μm, except for the insets of d (250 μm upper, 200 μm lower) and the insets of c and e (100 μm).

Table 1 |.

Antibodies

Primary antibodies for immunofluorescent staining
Name	Host species	Clone number	Manufacturer and cat. no.	RRID	Dilution factor
CDX2	Rabbit	EPR2764Y	Abcam (cat. no. ab76541)	AB_1523334	1:400
cCASP3	Rabbit	5A1E	Cell Signaling (cat. no. 9664S)	AB_2070042	1:400
COL4	Mouse	COL-94	Abcam (cat. no. ab6311)	AB_305414	1:100
EPCAM	Goat	Polyclonal	R&D Systems (cat. no. AF960)	AB_355745	1:400
FOXA1	Mouse	Q-6	Santa Cruz Biotechnology (cat. no. sc-101058)	AB_1124659	1:50
HT2-280	Mouse	TB-27	Terrace Biotech (cat. no. TB-27AHT2-280)	AB_2832931	1:100
hNUCL	Mouse	NM95	Abcam (cat. no. ab190710)	—	1:100
KI-67	Rabbit	D3B5	Cell Signaling (cat. no. 9129S)	AB_2687446	1:200
KRT13	Rabbit	EPR3671	Abcam (cat. no. ab92551)	AB_2134681	1:400
MUCIN1	Rabbit	EPR1023	Abcam (cat. no. ab109185)	AB_10862483	1:400
P63Α	Rabbit	D2K8X	Cell Signaling (cat. no. 13109)	AB_2637091	1:200
PROSPC	Rabbit	Polyclonal	Millipore Sigma (cat. no. AB3786)	AB_91588	1:200
SOX9	Goat	Polyclonal	R&D Systems (cat. no. AF3075-SP)	—	1:200
SPB	Rabbit	Polyclonal	Seven Hills Bioreagents (cat. no. WRAB-48604)	—	1:500
NKX2.1	Rabbit	Polyclonal	Seven Hills Bioreagents (cat. no. WRAB-1231)	AB_2832953	1:200
ZEB1	Rabbit	EPR17375	Abcam (cat. no. ab203829)	—	1:50
Name	Host species	Conjugate	Manufacturer and cat. no.	RRID	Dilution factor
Anti-Mouse IgG (H+L)	Donkey	AF- 488	Thermo Fisher Scientific (cat. no. A21202)	AB_141607	1:200
Anti-Goat IgG (H+L)	Donkey	AF- 488	Thermo Fisher Scientific (cat. no. A32814)	AB_2762838	1:200
Anti-Rabbit IgG (H+L)	Donkey	AF- 488j	Thermo Fisher Scientific (cat. no. A21206)	AB_2535792	1:200
Anti-Rat IgG (H+L)	Donkey	AF- 488	Thermo Fisher Scientific (cat. no. A21208)	AB_141709	1:200
Anti-mouse IgM Heavy Chain	Goat	AF- 488	Thermo Fisher Scientific (cat. no. A21042)	AB_141357	1:200
Anti-Mouse IgG (H+L)	Donkey	AF- 555	Thermo Fisher Scientific (cat. no. A31570)	AB_2536180	1:200
Anti-Goat IgG (H+L)	Donkey	AF- 555	Thermo Fisher Scientific (cat. no. A32816)	AB_2762839	1:200
Anti-Rabbit IgG (H+L)	Donkey	AF- 555	Thermo Fisher Scientific (cat. no. A31572)	AB_162543	1:200
Anti-Mouse IgG (H+L)	Donkey	AF- 647	Thermo Fisher Scientific (cat. no. A31571)	AB_162542	1:200
Anti-Goat IgG (H+L)	Donkey	AF- 647	Thermo Fisher Scientific (cat. no. A21447)	AB_141844	1:200

Fig. 6 |. RT–qPCR analysis of the core and tip areas of a lung organoid.

a, Upper image shows the morphology of organoids before isolation of core (dashed area) and tips. Bottom image shows the punched out core (left) and tip (right) of an organoid. b, RT–qPCR for select lung and non-lung endoderm markers in the centers relative to the periphery of the organoids (n = 3 independent experiments at ~d60, two-way ANOVA).

Table 2 |.

Primers

Gene	Forward	Reverse
CAPN14	GCCATGCCTATACTCTCACAG	CTCCAGTCTCCTTTCCATTCC
CDX2	GCCAAGTGAAAACCAGGACG	CAGAGAGCCCCAGCGTG
GAPDH	AACTTTGGCATTGTGGAAGG	ACACATTGGGGGTAGGAACA
GATA4	CTCAGAAGGCAGAGAGTGTGT	CGGGAGGCGGACAGC
HHEX	GCGAGAGACAGGTCAAAACC	TTATTGCTTTGAGGGTTCTCCT
HNF6	GAGGATGTGGAAGTGGCTG	ACATCTGTGAAGACCAACCTG
HNF1B	CAGTCGGTTTTACAGCAAGTC	TGGATATTCGTCAAGGTGCTG
MNX1	GCACCAGTTCAAGCTCAAC	GCTGCGTTTCCATTTCATCC
MUC13	AGGAAGATGCTAATGGGAACTG	GAATGACAATGCCAGCGATG
PDX1	ATGAACGGCGAGGAGCAGTA	TGGGTCCTTGTAAAGCTGCG
PROX	AACATGCACTACAATAAAGCAAATGA	CAGGAATCTCTCTGGAACCTCAAA

Box 2 |. Preparation of single-cell suspensions for scRNAseq or flow cytometry ● Timing 1.5 h (hands on, 30 min).

Reagents

• Cell recovery reagent (Corning, cat. no. 354253)

• BSA fraction V (7.5%) (Gibco, cat. no. 15260–037

• IMDM (Gibco, cat. no. 12440–053)

• Dead cell removal kit (Miltenyi Biotec, cat. no. 130–090-101)

Disposables

• 24-Well insert (Falcon, cat. no. 353095)

• 24-Well plate (Falcon, cat. no. 353047)

• 15 ml centrifuge tube (Thermo Scientific, cat. no. 339650)

Equipment

• Centrifuge 5810 R (Eppendorf)

• Forceps (Dumont, cat. no. 11251–33)

Reagent setup

• Base medium: to make 100 ml of base medium, add 0.53 ml of 7.5% BSA to 99.47 ml of IMDM (final concentration of BSA 0.04% vol/vol). Sterile filter through a 0.22 μm filter. Store at 4 °C for 1 month.

Procedure

1. Remove the insert from the 24-well plate, invert it and gently peel off the membrane from the bottom of the insert using forceps.

2. Carefully place the insert with the membrane removed over the mouth of an open, empty 15 ml centrifuge tube.

3. Add 1 ml of cell recovery reagent to the insert such that the embedded organoid with Matrigel falls into the 15 ml centrifuge tube.

4. Close the 15 ml centrifuge tube and immediately place on ice to depolymerize the Matrigel.

5. Every 20 min, gently shake the tube to check if the Matrigel has dissolved and place it back on ice.

6. Once the Matrigel has dissolved, add 10 ml of wash medium to the tube. Time varies from ~45 min to 1 h depending on Matrigel thickness.

7. Centrifuge at 200g for 5 min.

8. Aspirate the wash medium and cell recovery reagent and add 1 ml of Accutase to the organoid and incubate at 37 °C.

   ▲ CRITICAL STEP Avoid using trypsin to dissociate the organoid as it is too harsh and the chances of losing fragile cell types is higher.

9. Gently shake the tube to disturb/dissociate the organoid every 5 min and replace the tube at 37 °C for 10–12 min. Time varies depending on the organoid size and density.

10. When the organoid is dissociated add 10 ml wash medium and centrifuge at 400g for 4 min.

11. Aspirate the wash medium and resuspend the pellet in base medium.

    ▲ CRITICAL STEP Check the base medium compatibility with the sequencing technology to be used for scRNAseq.

12. Remove dead cells from the cell suspension by using the MiltenyiBiotec dead cell removal kit.

13. Dilute the cell suspension to a concentration of 5,000 cells/μl for sequencing. For sequencing and analysis of scRNAseq data, see Box 3.

    ▲ CRITICAL STEP Cell suspension dilution varies depending on the sequencing vendor requirements.

Box 3. scRNAseq | analysis.

For our scRNAseq analysis (see ‘Anticipated results’ and Figs. 7 and 8) libraries were prepared and sequenced at the JP Sulzberger Columbia Genome Center High-Throughput Screening Center at Columbia University. Sequencing libraries were prepared using 10x Genomics Single Cell Gene Expression 3′ workflow and sequenced on the Illumina NovaSeq 6000. FASTQ files were processed using cell ranger version 5.0.1 for each sample and the reads were aligned to the GRCh38 reference genome.

Data were analyzed using the standardized Seurat pipeline V4 in R⁶⁵ and reciprocal principal component analysis-based integration (https://satijalab.org/seurat/articles/integration_rpca.html). The steps involved in the analysis include: quality control analysis, normalization, integration of samples, dimensionality reduction, clustering and visualization of the data⁶⁵. To begin, the Cell Ranger output count matrix was read into R and a Seurat object was created for each individual dataset. The four datasets were merged into one dataset and filtered on the basis of the number of unique molecular identifiers (nUMI ≥500), number of genes (nGenes ≥250) and mitochondrial abundance (<20%) for further analysis. Following filtering, the datasets were split and to account for batch effects, sequencing depth and to remove technical variability, normalization and variance stabilization was performed using the SCTransform (SCT) function in Seurat⁶⁶. Cell cycle scoring and mitochondrial abundance was calculated and regressed out along with the normalization process using SCTransform.

Integration is based identification of common features in different datasets that serves as anchors across these datasets, such that batch effects are attenuated⁶⁵. Merged analysis uses the processed data, but does correct those based on common anchors, and therefore does not correct for batch effects. However, as integration may be prone to overcorrection if differences in cellular composition between samples are large⁶⁷, as may be these case in the organoids described here at different time points, we present both integrated (Fig. 7) and merged (Fig. 8) analyses. Features to use when integrating the datasets were chosen using SelectIntegrationFeatures function. Principal component analysis for each dataset was done individually before identifying integration anchors. We then set the normalization method to ‘SCT’ and the reduction to robust principal component analysis while using the FindIntegrationAnchors function. The datasets were integrated using the integration anchors identified using IntegrateData function. Principal component analysis was used to identify the number of dimensions (nDims) for dimensionality reduction of the integrated dataset and visualization was performed using RunPCA. nDims was used to compute the shared nearest neighbor graphs using the k-nearest neighbor algorithm, to find relevant clusters and UMAP for visualization of the clusters was performed using RunUMAP function. Clusters were created with resolutions ranging from 0.4 to 0.8. We visualized and analyzed the different clusters on the basis of marker identification using FeaturePlot and found 0.4 resolution to produce the most biologically informative clusters for our study. We performed pseudotime analysis using scVelo⁶³ on the merged datasets.

Fig. 7 |. Integrated scRNAseq analysis.

a, UMAP clusters after integrated analysis of organoids collected at d25, d40, d60 and d80. b, Dot plots of dynamic expression of endodermal (FOXA1), lung (NKX2.1, CPM), distal lung (SFTPB, NAPSA, SLC34A2, LPCAT), distal airway (SCGB3A2), mid and hindgut (GATA4, CDX2), mesenchymal (THY1, COL1A2, ZEB1, TBX4, WNT2B), proliferating cell (Ki67) and housekeeping genes (ACTA2) genes over time. c, UMAP feature plots of genes shown in b.

Fig. 8 |. Merged scRNAseq analysis.

a, UMAP clusters after merged (as opposed to integrated) analysis of organoids collected at d25, d40, d60 and d80, showing expression of each marker on top of each graph. Clusters corresponding to each timepoint are shown in the upper left plot. b, scVelo trajectory analysis. Clusters are identified by colors. Pseudotime analysis was performed using scVelo on the merged datasets^63,64.

Expertise needed to implement the protocol

Any scientist with experience in cell culture can apply this protocol. The required equipment, except for the low-oxygen incubator and the picking hood, is standard to most cell culture facilities, and reagents can be purchased from standard scientific vendors.

Materials

Biological materials

• hPSCs: this protocol was developed and optimized for differentiation of RUES2 human ESCs (Rockefeller University Embryonic Stem Cell Line 2, National Institutes of Health (NIH) approval number NIHhESC-09-0013, registration number 0013; RRID: CVCL_B810, passages 17–28) and has also been shown to work with two iPSC lines generated using Sendai Virus and modified messenger RNA from human dermal fibroblasts (purchased from Mount Sinai Stem Cell Core facility, passages 17–26). We anticipate that this protocol could also be applied to other hESC and hiPSC lines, although some optimization might be required. This optimization is primarily geared at optimal DE induction as evaluated by flow cytometry for expression of EPCAM, CXCR4 and cKIT (see Steps 28–66) and may require adjusting density of hPSCs and feeders, as well as duration of DE induction (typically ranging from 72 to 96 h) ! CAUTION Research involving human ESCs must comply with state, institutional and funding agency regulations. An approval from the Human Embryonic and Human Pluripotent Stem Cell Committee at the investigator’s home institute is usually required. We obtained approval from the Columbia University Embryonic Stem Cell Research Oversight committee (H.-W.S.) and Mount Sinai Embryonic Stem Cell Research Oversight committee (Y.-W.C.). ! CAUTION Karyotyping is required every 6 months. ! CAUTION The cell lines used in your research should be regularly checked to ensure they are authentic and are not infected with mycoplasma.

• Pre-irradiated MEFs (GlobalStem, cat. no. GSC-6201G, RRID: CVCL_RB05) ! CAUTION Lot test recommended for best endoderm induction.

Reagents

Growth factors and small molecules

• Recombinant Human/Mouse/Rat Activin A protein (R&D Systems, cat. no. 338-AC) ▲ CRITICAL Lot test needed for best DE induction.

• RA (Tocris, cat. no. 0695) ! CAUTION Light sensitive.

• Recombinant Human BMP4 protein (R&D Systems, cat. no. 314-BP)

• CHIR 99021 (Tocris, cat. no. 4423)

• Recombinant Human FGF10 protein (R&D Systems, cat. no. 345-FG)

• Recombinant Human FGF2 protein (R&D Systems, cat. no. 233-FB)

• Recombinant Human KGF/FGF7 protein (R&D Systems, cat. no. 251-KG)

• IWP 2 (Tocris, cat. no. 3533)

• Recombinant Human Noggin protein (R&D Systems, cat. no. 6057-NG)

• SB 431542 (Tocris, cat. no. 1614)

• Y-27632 dihydrochloride (Tocris, cat. no. 1254)

Medium and supplements

• Bovine albumin fraction V (BSA) (7.5% solution, Gibco, cat. no. 15260037)

• l-Ascorbic acid (Sigma-Aldrich, cat. no. A4544)

• B27 (Gibco, cat. no. 17504044) ▲ CRITICAL Lot test needed for best DE induction.

• N2 (Gibco, cat. no. 17502048) ▲ CRITICAL Lot test needed for best DE induction.

• β-Mercaptoethanol (Sigma-Aldrich, cat. no. M6250)

• GlutaMAX (Gibco, cat. no. 35050061)

• Knockout serum replacement (Gibco, cat. no. 10828028) MEM Non-Essential Amino Acids Solution (Gibco, cat. no. 11140050) ▲ CRITICAL Lot test needed for best DE induction.

• Monothioglycerol (MTG) (Sigma-Aldrich, cat. no. M6145)

• Penicillin–streptomycin (10,000 U/ml) (Gibco, cat. no. 15140122)

• Primocin (InvivoGen, cat. no. ant-pm-2)

• Stem cell qualified FBS (Atlanta Biologicals, cat. no. S10250)

• Human fibronectin protein, CF (R&D Systems, cat. no. 1918-FN)

• Growth factor reduced Matrigel (Corning, cat. no. 354230) or Cultrex 3-D Matrix RFG basement membrane extract (R&D System, cat. no. 344501001)

• Ham’s F12 (Cellgro, cat. no. 10-080-CV)

• Iscove’s modified Dulbecco’s medium (IMDM, Cellgro, cat. no. 10-016-CV)

• Dulbecco’s modified Eagle medium (DMEM) and Ham’s F12, 50/50 mix (Cellgro, cat. no. 10-092-CV)

• Gelatin from bovine skin (Sigma-Aldrich, cat. no. G9391)

• Dulbecco’s phosphate-buffered saline (DPBS) without Ca²⁺ and Mg²⁺ (Cellgro, cat. no. MT21031CM)

• Cell culture grade water (Corning, cat. no. 25-055-CVC)

Enzymes

• 0.05% Trypsin/ethylenediaminetetraacetic acid (EDTA) (Gibco, cat. no. 25300120)

• Accutase/EDTA (Innovative Cell Technologies, cat. no. AT104)

• Dispase (Corning, cat. no. 354235)

Other reagents

• cKIT-PE antibody (BioLegend, cat. no. 313204)

• CXCR4-APC antibody (BioLegend, cat. no. 306510)

• CD184 (CXCR4) MicroBead Kit, human (Miltenyi Biotec, cat. no. 130–100–070)

• Dimethyl sulfoxide (DMSO) (Fisher, cat. no. BP231-100)

Equipment

• 10 cm² tissue-culture dish (BD Falcon, cat. no. 353003)

• 15 ml conical tube (BD Falcon, cat. no. 352097)

• 50 ml conical tube (BD Falcon, cat. no. 352098)

• Posi-Click 1.7 ml microcentrifuge tube (Denville, cat. no. C2170)

• 24-Well transwell insert (BD Falcon, cat. no. 8770) ▲ CRITICAL Inserts from BD Falcon hold more medium in the inserts than other brands tested. This will result in better organoid morphology.

• 24-Well flat-bottom tissue culture-treated plate (Falcon, cat. no. 353047)

• 24-Well flat-bottom not treated cell culture plate (Falcon, cat. no. 351147)

• Six-well ultralow-attachment plate (Costar, cat. no. 3471)

• Six-well flat-bottom tissue culture-treated plate (Falcon, cat. no. 353046)

• 96-Well U-bottom non-tissue culture-treated plate (Corning, cat. no. 351177)

• P10 barrier tips (Denville Scientific, cat. no. P1096-FR)

• P20 barrier tips (Denville Scientific, cat. no. P1121)

• P200 barrier tips (Denville Scientific, cat. no. P1122)

• P1000 barrier tips (Denville Scientific, cat. no. P1126)

• Serological pipets, individually wrapped, 5 ml (Fisher, cat. no. 13-678-11D)

• Serological pipets, individually wrapped, 10 ml (Fisher, cat. no. 13-678-11E)

• Serological pipets, individually wrapped, 25 ml (Fisher, cat. no. 13-678-11)

• Disposable sterile bottle-top filters (Corning, cat. no. 431118)

• Disposable sterile bottles, 250 ml (Corning, cat. no. 430281)

• Disposable sterile bottles, 500 ml (Corning, cat. no. 430282)

• Disposable sterile bottles, 1,000 ml (Corning, cat. no. 430518)

• Fluorescence-activated cell sorting (FACS) tube (BD Falcon, 352008 or Corning, cat. no. 352008)

• Round-bottom polystyrene test tubes with cell strainer snap cap (BD Falcon, cat. no. 352235 or Corning, cat. no. 352235)

• Tissue culture hood

• Normoxic incubator (95% air/5% CO₂/37 °C)

• Hypoxic incubator (5% O₂/5% CO₂/37 °C)

• Centrifuge

• Hemocytometer

• Picking hood

• Microscopes: dissecting microscope: Nikon SMZ1500; laser scanning confocal microscope: Leica TCS SP8 Stellaris (40× oil immersion objective); Leica DMi1 Inverted Phase Contrast Microscope (Hi Plan 4×, Phase Contrast Hi Plan 10× Ph1 and 20× Ph1) ▲ CRITICAL Similar microscopes from other manufacturers can be used.

• Pipettes

• Pipet aid

• FlowJo flow cytometry software (https://www.flowjo.com/solutions/flowjo (RRID: SCR_008520))

Reagent setup

Growth factors and small molecules

FGF2

Reconstitute at 20 μg/ml in sterile 0.1% BSA/DPBS (wt/vol). Aliquot 100 μl to 500 μl in microcentrifuge tubes. Aliquots can be stored at −20 °C to −70 °C for up to 3 months.

FGF10

Reconstitute at 10 μg/ml in sterile 0.1% BSA/DPBS (wt/vol). Aliquot 500 μl in microcentrifuge tubes. Aliquots can be stored at −20 °C to −70 °C for up to 3 months.

FGF7

Reconstitute at 10 μg/ml in sterile 0.1% BSA/DPBS (wt/vol). Aliquot 500 μl in microcentrifuge tubes. Aliquots can be stored at −20 °C to −70 °C for up to 3 months.

BMP4

Reconstitute at 10 μg/ml in sterile 0.1% BSA/DPBS (wt/vol). Aliquot 500 μl in microcentrifuge tubes. Aliquots can be stored at −20 °C to −70 °C for up to 3 months.

CHIR 99021

Reconstitute at 3 mM in DMSO. Aliquot 100 μl in microcentrifuge tubes. Aliquots can be stored at −20 °C to −70 °C for up to 6 months.

All-trans RA

Reconstitute at 50 μM in DMSO. Aliquot 100 μl in microcentrifuge tubes. Aliquots can be stored at −20 °C to −70 °C for up to 3 months.

IWP2

Reconstitute at 1 mM in DMSO. Aliquot 100 μl in microcentrifuge tubes. Aliquots can be stored at −20 °C to −70 °C for up to 6 months.

Noggin

Reconstitute at 100 μg/ml in sterile 0.1% BSA/DPBS (wt/vol). Aliquot 100 μl in microcentrifuge tubes. Aliquots can be stored at −20 °C to −70 °C for up to 3 months.

SB431542

Reconstitute at 10 mM in DMSO. Aliquot 100 μl in microcentrifuge tubes. Aliquots can be stored at −20 °C to −70 °C for up to 6 months.

Y-27632

Reconstitute at 10 mM in DMSO. Aliquot 100 μl in microcentrifuge tubes. Aliquots can be stored at −20 °C to −70 °C for up to 6 months.

Activin A

Reconstitute at 100 μg/ml in sterile 0.1% BSA/DPBS (wt/vol). Aliquot 100 μl in microcentrifuge tubes. Aliquots can be stored at −20 °C to −70 °C for up to 3 months.

Gelatin

Prepare a 0.1% (wt/vol) solution by dissolving gelatin in tissue culture grade water. Heat the gelatin solution to boiling for 5–10 min. After cooling down, filter the gelatin solution through a 0.22 μm filter to make it sterile. The gelatin solution can be kept at 4 °C for up to 3 months.

Fibronectin

Make 100 μl aliquots and store them at 4 °C for up to 6 months.

l-Ascorbic acid

Reconstitute at 50 mg/ml in sterile cell culture grade water (wt/vol). Filter the solution through a 0.22 μm filter to make it sterile. Make 100 μl aliquots and store at −20 °C to −70 °C for up to 3 months. ! CAUTION Discard unused solution. Always use a freshly thawed aliquot for culture.

Media

MEF medium

Make 500 ml MEF medium by combining the reagents as detailed in the table below. Sterilize by filtering through a 0.22 μm filter. MEF medium can be stored at 4 °C for up to 1 month.

Reagent to add	Stock concentration	Volume to add	Final concentration
IMDM		415 ml
Stem cell-qualified FBS		75 ml	15%
MTG	11.5 M	20 μl	0.46 mM
GlutaMAX	100×	5 ml	1×
Penicillin–streptomycin	100×	5 ml	1×

Stop medium

Make 500 ml Stop medium by combining reagents as detailed in the table below. Sterilize by filtering through a 0.22 μm filter. Stop medium can be stored at 4 °C for up to 1 month.

Reagent to add	Stock concentration	Volume to add	Final concentration
IMDM		465 ml
FBS		25 ml	5%
GlutaMAX	100×	5 ml	1×
Penicillin–streptomycin	100×	5 ml	1×

hPSC maintenance medium

Make 500 ml of hPSC maintenance medium by combining reagents as detailed in the table below. Sterilize by filtering through a 0.22 μm filter. hPSC maintenance medium can be stored at 4 °C for up to 1 month. ▲ CRITICAL Supplement FGF2 to a final concentration of 20 ng/ml right before use.

Reagent to add	Stock concentration	Volume to add	Final concentration
DMEM/F12		389 ml
Knockout serum		100 ml	20%
MEM-nonessential amino acids		5 ml
β-Mercaptoethanol	14.3 M	3.5 μl	0.1 mM
Primocin	50 mg/ml	1 ml	100 μg/ml
GlutaMAX		5 ml
Supplement with FGF2 before use
FGF2	10 μg/ml	2 μl	20 ng/ml

Serum-free differentiation (SFD) medium

Make 1,000 ml of SFD medium by combining reagents as detailed in the table below. Sterilize by filtering through a 0.22 μm filter. SFD medium can be store at 4 °C for up to 1 month. On the day of use, supplement SFD medium with L-ascorbic acid and MTG to a final concentration of 50 μg/ml (1 μl/ml) and 0.45 mM (0.039 μl/ml), respectively, to make complete SFD medium.

Reagent to add	Stock concentration	Volume to add	Final concentration
IMDM		725 ml
Ham’s F12		242.5 ml
N2		5 ml
B27		10 ml
7.5% BSA	7.5%	7.5 ml
Penicillin–streptomycin	100×	10 ml	1×
To make complete SFD medium, add the following to SFD
Reagent to add	Stock concentration	Volume to add per ml	Final concentration
GlutaMAX	100×	10 μl	1×
Ascorbic acid	50 mg/ml	1 μl	50 μg/ml
MTG	11.5M	0.039 μl	0.45 mM

EB/PS medium

Make 12 ml of EB/PS medium, sufficient for a six-well plate, by adding reagents as detailed in the table below to complete SFD medium. Prepare the medium fresh on the day of use.

Complete SFD medium supplemented with	Stock concentration	Volume to add per ml	Final concentration
Y-27632	10 mM	1 μl	10 μM
BMP4	10 μg/ml	0.3 μl	3 ng/ml

Endoderm induction medium

Make 13 ml of endoderm induction medium, sufficient for a six-well plate, by adding reagents as detailed in the table below to complete SFD medium. Prepare the medium fresh on the day of use.

Complete SFD medium supplemented with	Stock concentration	Volume to add per ml	Final concentration
Y-27632	10 mM	1 μl	10 μM
BMP4	10 μg/ml	0.05 μl	0.5 ng/ml
FGF2	10 μg/ml	0.25 μl	2.5 ng/ml
Activin A	100 μg/ml	1 μl	100 ng/ml

Anteriorization medium-1

Make 13 ml of anteriorization medium-1, sufficient for a six-well plate, by adding reagents as detailed in the table below to complete SFD medium. Prepare the medium fresh on the day of use.

Complete SFD medium supplemented with	Stock concentration	Volume to add per ml	Final concentration
Noggin	100 μg/ml	1 μl	100 ng/ml
SB431542	10 mM	1 μl	10 μM

Anteriorization medium-2

Make 12 ml of anteriorization medium-2, sufficient for a six-well plate, by supplementing reagents as detailed in the table below to complete SFD medium. Prepare the medium fresh on the day of use.

Complete SFD medium supplemented with	Stock concentration	Volume to add per ml	Final concentration
SB431542	10 mM	1 μl	10 μM
IWP2	1 mM	1 μl	1 μM

Ventralization/branching medium

Make 12 ml of ventralization/branching medium, sufficient for a six-well plate, by adding reagents as detailed in the table below to complete SFD medium. The prepared medium can be stored at 4 °C up to 3 d.

Complete SFD medium supplemented with	Stock concentration	Volume to add per ml	Final concentration
CHIR99021	3 mM	1 μl	3 μM
FGF10	10 μg/ml	1 μl	10 ng/ml
FGF7 (KGF)	10 μg/ml	1 μl	10 ng/ml
BMP4	10 μg/ml	1 μl	10 ng/ml
All-trans RA	0.5 mM	0.1 μl	50 nM

Plating irradiated MEFs for hPSC culture

Precoat six-well flat bottom tissue culture-treated plates with 2 ml 0.1% gelatin made with cell culture grade water for 15 min at room temperature (20–25 °C). Thaw a frozen vial of pre-irradiated MEFs containing ~2 × 10⁶ cells in a water bath at 37 °C. Add the thawed MEFs to 10 ml wash medium and centrifuge at 400g for 4 min. Aspirate the wash medium and add 1 ml of MEF medium. Count the MEFs and seed at a density of 17,000–25,000 cells/cm² in 2 ml of MEF media in a normoxic incubator overnight. The irradiated MEFs should be plated 1 d before plating of hPSCs to allow attachment and spreading of the MEFs producing sufficient extracellular matrix to prevent premature differentiation.

hPSC culture

Maintain hPSCs on irradiated MEFs. Culture cells in hPSC maintenance medium. Change medium daily. Passage hPSCs with Accutase/EDTA and replate at a dilution of 1:48. Maintain cultures in a humidified normoxic incubator. Further details covering how to properly maintain an hPSC culture can be found in ‘ES Cell International Pte Ltd: Methodology Manual Human Embryonic Stem Cell Culture 2005’ and ‘Harvard Stem Cell Institute (HSCI) StemBook Protocols for pluripotent cells, URL: http://www.stembook.org/protocols/pluripotent-cells’ and in our previously published protocols^57,61. Lines are karyotyped and verified for mycoplasma contamination using PCR every 6 months.

Thawing and aliquoting Matrigel

Thaw growth factor-reduced Matrigel on ice at 4 °C overnight. For aliquots used for organoid embedding, prechill Posi-Click 1.7 ml microcentrifuge tubes on ice for 15 min and aliquot 0.5 ml of thawed Matrigel to each tube on ice. For aliquots used for coating of culture plates, prechill 15 ml conical tubes on ice for 15 min and aliquot 1 ml of thawed Matrigel to each tube on ice. Aliquots can be stored at −20 °C for up to 6 months. CRITICAL Keep all parts contacting Matrigel cold to avoid polymerization.

Coating dishes with Matrigel for MEF depletion

Thaw 1 ml Matrigel and add to 29 ml of cold IMDM to a 50 ml prechilled conical tube to obtain a final concentration of 3.3% (vol/vol) and keep on ice. Make sure the diluted Matrigel is mixed well. Immediately transfer the 10 ml diluted Matrigel/IMDM solution to the 10 cm² tissue culture dish. Matrigel-coated dishes can be used after 3 h of incubation at room temperature or stored at 4 °C for up to 2 weeks. Keep the dishes flat in the refrigerator and avoid drying out of the Matrigel coating solution.

Fibronectin plates

Prepare fibronectin-coated six-well plates by diluting fibronectin to 4 μg/ml in DPBS. Add 2 ml fibronectin/DPBS solution to each well and incubate the plates in a normoxic incubator for at least 30 min or 4 °C overnight. Make sure the fibronectin coating covers the entire plate (~12 ml of a six-well plate).

Procedure

▲ CRITICAL The following procedures describe how to generate one six-well plate EBs from hPSCs.

MEF depletion on Matrigel (d1) ● Timing 18–24 h (hands on, 20 min)

• 1Prepare one Matrigel-coated dish as described in ‘Reagent setup.’
• 2Dissociate two wells of hPSCs (from a six-well plate, 90–95% confluent, corresponding to 5–7 × 10⁶ cells; see ‘Reagent setup’) by aspirating the hPSC maintenance medium from the wells, followed by adding 1 ml per well Accutase and incubate in a normoxic incubator for 2–3 min.
• 3Verify under a microscope that MEFs have detached from the plate, then aspirate the Accutase.
• 4Neutralize the enzyme by adding 2 ml Stop medium to each well.
• 5Gently flush the cells off the well by pipetting up and down.
• 6Transfer the cell mixture to a 15 ml conical tube.
• 7Pellet the dissociated cells by centrifugation at 400g for 4 min.
• 8Aspirate as much of the supernatant containing enzyme and Stop medium as possible.
• 9Resuspend the cells with 10–12 ml hPSC maintenance medium.
• 10Aspirate the Matrigel-coating solution from the dish (from Step 1).
• 11Plate the cells in the Matrigel-coated dish.
• 12Gently rock the dish with a side-to-side motion a few times to ensure the cells are evenly distributed.
• 13Incubate the cells in a normoxic incubator overnight.

Embryoid body formation/primitive streak induction (d0) ● Timing 12–16 h (hands on, 15 min)

• 14On d0, prepare the EB/PS medium as described in ‘Reagent setup’.

  ▲ CRITICAL STEP Prepare the medium fresh on the day of use.

• 15Remove the hPSC maintenance medium from the Matrigel-coated dish.
• 16Add 3 ml of cold (4 °C) trypsin/EDTA to the dish.
• 17Incubate the dish for 1–1.5 min in a normoxic incubator.
• 18Aspirate the trypsin/EDTA solution.
• 19Neutralize the enzymes by adding 10 ml wash medium.
• 20Gently flush the cells off the dish by pipetting up and down using a 10 ml serological pipet.

  ? TROUBLESHOOTING

• 21Transfer the cell mixture to a 15 ml conical tube.
• 22Pellet the dissociated cells by centrifugation at 400g for 4 min.
• 23Aspirate as much of the supernatant containing enzyme and Stop medium as possible.
• 24Resuspend the cells with 12.5 ml EB/PS medium.
• 25Distribute 2 ml per well of the cell mixture to a six-well ultralow-attachment plate.
• 26Gently rock the plate with a side-to-side motion a few times to ensure even distribution of cells and avoid aggregation of cells before placing the plate back in an incubator.
• 27Place the ultralow-attachment plate in a hypoxic incubator for 12–16 h to allow EB formation.

Endoderm induction (d1-4) ● Timing ~3 d (hands on, 20 min)

• 28Prepare the endoderm induction medium as described in ‘Reagent setup’.

  ▲ CRITICAL STEP Prepare the medium fresh on the day of use.

• 29After 12–16 h of EB formation, gently collect all EBs from the ultralow-attachment plate to a 15 ml conical tube.

  ? TROUBLESHOOTING

• 30After EB collection, add 1 ml per well of the endoderm induction medium to the ultralow-attachment plate that previously contained the EBs.

  ▲ CRITICAL STEP Endoderm induction medium is added to the empty wells to prevent them from drying out, which promotes cellular attachment.

• 31Allow the EBs to settle down for 5 min or centrifuge the conical tube at 130g for 1 min to pellet down the EBs.
• 32Aspirate the EB/PS medium.
• 33Gently resuspend the EBs with 6.5 ml endoderm induction medium using a 5 ml serological pipet.
• 34Distribute them 1 ml per well of the EB mixture equally back to the low-attachment plate.

  ▲ CRITICAL STEP Only take 1ml of the EB mixture each time. Gently resuspend the EBs each time before taking the EB mixture for each well to ensure similar EB numbers are distributed to each well.

• 35Rock the plate with a side-to-side motion to ensure that EBs are evenly distributed.
• 36Return the plate to a 5% CO₂/5% O₂ incubator.
• 37On d2, add 1 ml fresh Endoderm induction medium to each well (in total 3 ml per well).

  ▲ CRITICAL STEP Prepare the medium fresh on the day of use.

• 38Rock the plate with a side-to-side motion to ensure that EBs are evenly distributed.
• 39Return the plate to a 5% CO₂/5% O₂ incubator.
• 40On d3, add 2 ml fresh Endoderm induction medium to each well (in total: 5 ml medium per well).

  ▲ CRITICAL STEP Prepare the medium fresh on the day of use.

• 41Rock the plate in a side-to-side motion to ensure that EBs are evenly distributed.
• 42Return the plate to a 5% CO₂/5% O₂ incubator.

DE yield examination (d4.1-4.3) ● Timing hands on, 1.5 h

▲ CRITICAL On d4.1-4.3 (74.5–79.5 h after exposure of EBs to Activin A), verify DE yield by flow cytometric analysis of CXCR4 and cKIT expression. Always verify the DE yield is >90% before continuing to the anteriorization stage. Different hPSC lines might have different optimal DE induction times.

• 43Prepare 25 ml of complete SFD medium as described in the ‘Reagent setup’.
• 44Swirl the plate slowly to make the EBs concentrate in the middle of the wells.
• 45Gently collect half of a single well of EBs (EBs are typically concentrated in the middle of the well) to a 15 ml conical tube by a P1000 pipette. Return the plates (containing 5.5 wells of EBs) back to a hypoxic incubator until later use.
• 46Allow the collected EBs to settle down for 5 min or centrifuge the conical tube at 130g for 1 min to pellet down the EBs.
• 47Aspirate the medium.
• 48Add 1 ml of cold trypsin/EDTA to the conical tube.
• 49Gently tap the tube to swirl the EBs in the trypsin/EDTA solution.

  ▲ CRITICAL STEP EBs with good endoderm yield usually start to dissociate after 2–2.5 min. Do not digest the EBs for more than 4 min.

• 50When EBs are completely dissociated (no visible clumps) or after 4 min, neutralize the enzymes with 10 ml stop medium.
• 51Take 25 μl of cell mixture and mix with Trypan Blue, then count the cell number using a hemocytometer.
• 52Pellet the dissociated cells by centrifugation at 400g for 4 min.
• 53Aspirate the Stop medium.
• 54Resuspend the cells in complete SFD medium on the basis of the cell counts: 100 μl of complete SFD medium per million cells.
• 55Add CXCR4 (1:100) and cKIT (1:100) antibodies that have been verified to be suitable for flow cytometric analysis.
• 56Stain the cells on the basis of the manufacturer’s protocol (BioLegend, for example, in this protocol).
• 57After the staining procedure is done, add 5–10 ml of complete SFD medium to the conical tube.
• 58Pellet down the dissociated cells by centrifugation at 400g for 4 min.
• 59Aspirate the supernatant.
• 60Resuspend the cells in 500 μl of complete SFD medium.
• 61Filter the cell mixture through a cell strainer cap (mesh size: 35 μm) attached to a FACS tube.
• 62Add 4 ml of complete SFD medium to the FACS tube.
• 63Pellet down the filtered cells by centrifugation at 400g for 4 min.
• 64Aspirate the supernatant.
• 65Resuspend the cells in 300 μl of complete SFD medium.
• 66Determine the endoderm yield by determining the CXCR4 and cKIT double positive population via a flow cytometric analyzer. Continue differentiation only with EBs that have an endoderm yield that is >90%.

? TROUBLESHOOTING

Anteriorization (d5-6) ● Timing ~2 d (hands on, 2 h)

• 67Prepare fibronectin-coated six-well plates as described in ‘Reagent setup’.
• 68Prepare 15 ml of complete SFD medium as described in ‘Reagent setup’.
• 69Prepare 50 ml of Anteriorization medium-1 as described in ‘Reagent setup’.

  ▲ CRITICAL STEP Prepare the medium fresh on the day of use.

• 70Swirl the ultralow-attachment plate of EBs (from Step 45) slowly to make the EBs concentrate in the middle of the wells.
• 71Gently collect all the remaining EBs from the plate and pool in a 15 ml conical tube by a P1000 pipette.
• 72Allow the EBs to settle down for 5 min or centrifuge the conical tube at 130g for 1 min to pellet down the EBs.
• 73Aspirate the medium.
• 74Dissociate the EBs into single cells with 3 ml of trypsin.
• 75Gently tap the tube to swirl the EBs in the trypsin/EDTA solution.
• 76Neutralize the enzymes with 10 ml Stop medium.
• 77Pellet the dissociated cells by centrifugation at 400g for 4 min.
• 78Aspirate the Stop medium.
• 79Resuspend the cells in 10 ml of complete SFD medium.
• 80Take 25 μl of cell mixture and mix with Trypan Blue and count the cell number using a hemocytometer.
• 81Pellet the cells by centrifugation at 400g for 4 min.
• 82Aspirate the complete SFD medium.
• 83Resuspend the cells in Anteriorization medium-1 (7.5 × 10⁵ cells per 2 ml of Anteriorization medium-1).

  ? TROUBLESHOOTING

• 84Add 2 ml of cell mixture to each well of the fibronectin-coated six-well plate (from Step 67).

  ? TROUBLESHOOTING

• 85Rock the plate in a side-to-side motion to ensure that cells are evenly distributed.

  ? TROUBLESHOOTING

• 86Return the plate to a normoxic incubator.

  ? TROUBLESHOOTING

• 87On d6, after 24 h (±2 h) after adding the Anteriorization medium-1, prepare an appropriate amount of the Anteriorization medium-2 (2 ml per well) as described in ‘Reagent setup’.

  ▲ CRITICAL STEP Prepare the medium fresh on the day of use.

  ? TROUBLESHOOTING

• 88Aspirate Anteriorization medium-1 and replace with Anteriorization medium-2.

  ? TROUBLESHOOTING

• 89Return the plates to a normoxic incubator.

? TROUBLESHOOTING

Ventralization and LBO formation (d6-20/25) ● Timing ~14–20 d (hands on, 30 min)

• 90On d6, 24 h (± 2 h) after adding the Anteriorization medium-2, prepare an appropriate amount of the ventralization medium/branching medium (2 ml per well) as described in ‘Reagent setup’.

  ▲ CRITICAL STEP Prepare the medium fresh on the day of use.

  ? TROUBLESHOOTING

• 91Replace the Anteriorization medium-2 with ventralization medium/branching medium (2 ml per well).

  ? TROUBLESHOOTING

• 92Return the plates to a normoxic incubator.

  ? TROUBLESHOOTING

• 93On d8, 48 h later, prepare an appropriate amount of the ventralization medium/branching medium (2 ml per well) as described in ‘Reagent setup’.

  ? TROUBLESHOOTING

• 94Aspirate all the old ventralization medium/branching medium and add 2 ml fresh ventralization medium/branching medium to each well.

  ? TROUBLESHOOTING

• 95Suspend the organoids by gently pipetting up and down with P1000 tips.

  ? TROUBLESHOOTING

• 96Transfer the suspended organoids to six-well ultralow-attachment plates (one well to one well).

  ? TROUBLESHOOTING

• 97Rock the plate with a side-to-side motion to ensure that cells are evenly distributed.

  ? TROUBLESHOOTING

• 98Return the plate to a normoxic incubator.

  ? TROUBLESHOOTING

• 99On d10, feed the organoids by tilting the plate and allowing the organoids to sink to the bottom edge. Remove the old medium while avoiding touching the organoids. Add 2 ml freshly prepared ventralization medium/branching medium to each well. Feed the organoids with freshly prepared medium on the day of use every other day.

? TROUBLESHOOTING

Branching organoid (d20/25–end of experiment) ● Timing ~25–150 d (hands on, 2 h)

• 100Examine organoids daily under a microscope between d20 and d25. When the desired stage is reached, proceed with embedding.

  ? TROUBLESHOOTING

• 101One night before embedding, thaw the desired amount of Matrigel (150 μl/insert) as described in ‘Reagent setup’.

  ▲ CRITICAL STEP Keep Matrigel on ice during the entire procedure to avoid polymerization.

  ? TROUBLESHOOTING

• 102On the day of use, prepare 50 ml of the ventralization medium/branching medium as described in ‘Reagent setup’.

  ? TROUBLESHOOTING

• 103Add 100 μl per well of the ventralization medium/branching medium to a 96-well U-bottom non-tissue culture-treated plate.

  ? TROUBLESHOOTING

• 104Select the organoids with folding structures under a microscope in a picking hood. One six-well plate should contain hundreds to thousands of organoids. Organoids with folding epithelial structures will gradually form and the majority should adopt this morphology after d15 (Figs. 1b, vi and 3b).

  ? TROUBLESHOOTING

• 105Typically, one to four organoids will go on to be plated in each 24-well insert. First, put this desired number of organoids per insert into each well of a 96-well U-bottom non-tissue culture-treated plate and set aside.

  CRITICAL STEP Putting too many organoids in the same insert will inhibit their growth in the long run.

  ? TROUBLESHOOTING

• 106Place 24-well inserts into non-tissue culture-treated plates.

  ? TROUBLESHOOTING

• 107Layer 50 μl of 100% cold Matrigel into the bottom of each insert.

  ▲ CRITICAL STEP The Matrigel should be distributed evenly to cover the entire surface of the insert. If not, tap the plate to spread the Matrigel before it polymerizes.

  ? TROUBLESHOOTING

• 108Wait 5 min or until the Matrigel has solidified.

  ? TROUBLESHOOTING

• 109Gently remove as much of the ventralization medium/branching medium as possible from the 96-well plate (from Step 105), one well at a time.

  ? TROUBLESHOOTING

• 110Use a P1000 tip to quickly take ~30–50 μl of 100% cold Matrigel.

  ? TROUBLESHOOTING

• 111Mix the organoids with the cold Matrigel gently to avoid creating bubbles.

  ? TROUBLESHOOTING

• 112Pick up the organoids and immediately but slowly put the organoid–Matrigel mixture in the center of an insert.

  ? TROUBLESHOOTING

• 113Wait 5 min for the Matrigel to solidify to secure the organoids in the center of the insert.

  ? TROUBLESHOOTING

• 114Add another 50 μl of 100% cold Matrigel to the insert to create a Matrigel sandwich.

  ? TROUBLESHOOTING

• 115Put the plates in a normoxic incubator for 10 min to make sure all Matrigel has solidified.

  ? TROUBLESHOOTING

• 116Add 500 μl/insert of ventralization medium/branching medium to the inserts.

  ? TROUBLESHOOTING

• 117Add another 500 μl per well of ventralization medium/branching medium into the wells.

  ? TROUBLESHOOTING

• 118Return the plate to a normoxic incubator.
• 119Replace the medium every 2–3 d until the cultures reach the desired timepoints of experiments. Medium is replaced both in the insert and in the well. To aspirate medium from the top (insert), tilt the plate so that one can aspirate without disturbing the Matrigel.
• 120When the desired timepoint is reached, standard analyses can be performed (see Box 1 for RT–qPCR and IF and Boxes 2 and 3 for scRNAseq).

Timing

Steps 1–13, MEF depletion (d1): 18–24 h (hands on, 20 min)

Steps 14–27, EB formation (d0): 12–16 h (hands on, 15 min)

Steps 28–42, DE induction (d1-4): ~3 d (hands on, 20 min)

Steps 43–66, DE yield test (d4.1-4.3): 1.5 h (hands on, 1.5 h)

Steps 67–89, Anteriorization (d5-6): ~2 d (hands on, 2 h)

Steps 90–99, Ventralization/LBO formation (d6-20/25): ~14–20 d (hands on, 30 min)

Steps 100–120, Branching organoid (~d20-25): ~25–150 d (hands on, 2 h every 2–3 d)

Troubleshooting

Troubleshooting advice can be found in Table 3 and Fig. 9.

Table 3 |.

Troubleshooting table

Step	Problem	Possible reason	Solution
20	Rapid reattachment of cells after adding serum-containing media to deactivate the trypsin	Step performed too slowly, or cells were not completely detached	Be quick but gentle. Leave those cells that do not detach from the dish. Do not force the cells to detach from the dish, this will damage cells and lead to larger aggregation of EBs, which will lead to lower endoderm yield
29	Large clusters and dead cells appeared around the EBs	Vigorous pipetting or cells were unevenly distributed when placed in incubator for EB formation	Gentle pipetting can help to remove dead cells. For large clusters, collect them separately and pipette up and down with P1000 tip a few times to break them into smaller clusters
66	Low endoderm yield	Some hPSC lines do not generated endoderm efficiently	Try multiple lines and use lines that generally yield DE of >90% purity as determined by CXCR4/cKIT/EPCAM staining. Alternatively, CXCR4 microbeads can be used to enrich the endodermal population on d4. DE cells do not survive well after sorting
		Vigorous pipetting during dissociation	We noticed that pipetting during dissociation of EBs for flow cytometric analysis can lead for loss of CXCR4 expression
		Suboptimal reagents	Critical reagents for DE induction are MEFs used for maintenance of hPSCs, N2/B27 and Activin A. These need to be batch-tested for DE induction using a line that reliably generates DE
		Inadequate hPSC maintenance	DE is the most problematic germ layer to generate. Verify pluripotency of hPSCs by staining for OCT4 and SOX2 (~100% of the cells should be positive). Even if the cells express proper pluripotency markers, the optimal density matters. We typically plate RUES2 cells at a density of 7,000–9,000 cells/cm2 and hiPSCs at 10,000–15,000 cells/cm2. If excessive cell death is observed, 10 μM Y-27632 (ROCK inhibitor) to hPSC maintenance media for the first 24 h following plating can be tried. MEFs for hPSC maintenance are plated at a density of 17,000–23,000 cells/cm2. However, optimal density can vary from batch to batch and needs to be tested using DE induction as the end point
83–99	Inefficient or abnormal LBO formation. The majority of LBOs should contain folding layers of epithelial cells. For examples, see Fig. 1b (d15) and Fig. 3b	Defective DE	See Troubleshooting for Step 66 above
		Over-trypsinization of EBs	Over-trypsinization of EBs will lead to extensive cell death the next day. Always use cold trypsin
		Defective anteriorization or ventralization	This can be verified by IF (for illustration, see Fig. 5 and Anticipated results for description), and by the morphology of the cultures at d8-10 (Fig. 9). In addition to defective DE, this can be caused by the cell density in Step 83. The optimal seeding density for this step is 7.5 × 105 cells per well (Fig. 9a, left). If the initial cell seeding density is lower than 7.5 × 105 cells per well, smaller and fewer organoids will be observed on d8. These small organoids tend to become hollow spheres faster in culture. If the culture is initiated with many more than 7.5 × 105 cells per well, organoid formation might be less obvious on d8. In this case, the organoids can still be resuspended and proceed to Ventralization and LBO formation (Fig. 9a, right). These densities may have to be adjusted for different hPSCs lines
		Small LBOs at d10 (Step 99)	Organoids might be small on d10. If they do not sink to the bottom, add 1 ml of fresh media directly to the well and wait for two more days before replacing the media
		Exhaustion of the culture media	LBOs are highly proliferative between d8 and d20 (Fig. 9b and Supplementary Fig. 1). If the media turn yellow quickly, more fresh ventralization media/branching media can be fed to each well
100–117	Inefficient or abnormal formation of branching organoids in Matrigel	Defective DE and/or LBO generation	See Troubleshooting for Steps 83–99
		Timing of LBO plating in Matrigel	The optimal timing to embed organoid is between d20 and d25. When culturing organoids in suspension for too long, they will become hollow spheroids and eventually burst and reform smaller cell clumps (Fig. 9c). These small cell clumps will not generate branching organoids
		Picking of LBOs	Although no clear correlation could be found between LBO morphology and subsequent quality of Matrigel organoid generation, it is best to avoid picking very dense LBOs that lack the folding structures, or LBOs with cysts
		Rapid polymerization of Matrigel	We suggest embedding organoids one insert at a time to avoid Matrigel solidification in U-bottom 96-well plate before transferring to 24-well inserts (Steps 107–112)
		Exhaustion of the culture media	With multiple organoids in one insert, more ventralization media/branching media can be fed to each insert, especially in later-stage organoid cultures

Fig. 9 |. Troubleshooting of LBO formation.

a, Effect of different plating densities on the morphology of the ventralized AFE stage (d8), before lifting the cells to generate LBOs. b, Proliferation as measured by staining for Ki67 and 7-AAD at different stages of the protocol indicated on top of the panels. Single-cell suspension at d35 generated as described in Box 2. c, Abortive organoids generated from LBOs cultured in suspension for more than 25 d. Scale bars in c, 250 μm.

Anticipated results

Brightfield images of the sequential stages of differentiation are shown in Fig. 1b. Using this differentiation protocol, self-organizing cell clumps will form in 2 d after switching to the ventralization medium/branching medium (Fig. 1b, iv). Some of the organoids will detach from the monolayer but most of them will remain loosely attached to the plate. The number of organoids formed in each well varies from hundreds to thousands. The cell clumps might look damaged right after suspension, but they will self-organize/repair themselves the next day. Cell clumps appear homogeneous at the beginning and folding structures (Fig. 1b, v) will appear 2–3 d after suspension. Folding structures will become more complex with time (Fig. 1b, vi). Budding of organoids happens within 1 week after embedding in Matrigel and obvious branching can be observed after 2–3 weeks (Fig. 1b, vii). Formation of self-organizing cell clumps might not be obvious in some iPS lines. However, they will form self-organizing cell clumps after suspension and branch out after embedding in Matrigel. These organoids can be maintained for more than 6 months and will fill the entire well they are cultured in.

Organoids can be analyzed using standard approaches, including RT–qPCR (Box 1), IF (Box 1), flow cytometry and RNAseq as published²⁰, as well as scRNAseq (Boxes 2 and 3). We use mRNA expression predominantly to compare organoids of different genotypes or grown in different conditions to each other. We prefer to benchmark development by morphology and IF, as published^20,28 rather than mRNA expression by RT–qPCR or bulk RNAseq. mRNA determination typically uses a baseline control and reports fold changes compared with this control. In the developmental trajectory of the organoids, these would be LBOs or hPSCs. However, the latter do not express mature differentiation markers, leading to large and highly variable fold changes. For example, Fig. 3a shows the mRNA fold changes of ABCA3 and surfactant genes when comparing more mature organoids (d170) with immature organoids (d15 and d25)²⁰. Since immature organoids do not express or express very little mature markers, the fold changes of these markers are high and very variable due to the infinitesimally small denominator.

Examples of IF studies and anticipated results at consecutive stages, using d20 and d60 as representative examples, are shown in Fig. 5. During the AFE stage, cells grow in islands containing mostly SOX2⁺ cells, although clusters of cells expressing the hindgut markers CDX2⁺ are present within these islands (Fig. 5a, d8, left). The islands of epithelial cells are interspersed with ZEB1⁺ mesenchymal cells (Fig. 5a, d8, right). In the suspension stage at d20, LBOs consist of folding and invaginating epithelial cells surrounded by mesenchymal cells (ZEB1⁺), while large sections of most LBOs express NKX2.1 (Fig. 5b, d20). After development in Matrigel, NKX2.1 and EPCAM are expressed in the entire periphery of the organoids from ~d40 on (Fig. 5c), indicating nearly full commitment to the lung lineage. We also discovered at this stage (images shown are d60) small hollow tubular structures in the dense centers of the organoids that expressed EPCAM but not NKX2.1 (Fig. 5c). These cells express FOXA1 indicating that they are endodermal (Fig. 5d). Most proliferating cells, as detected by staining for Ki67 are in the periphery (Fig. 5d, inset 1), although rare proliferating cells were also observed in the central endodermal tubules (Fig. 5d, inset 2, arrow). Surrounding these structures in the center are cells, identified by the presence of 4′,6-diamidino-2-phenylindole (DAPI) (Fig. 5e), that did not stain with any lung or airway markers, or with mesenchymal markers (CD90, ZEB1, PDGFRA, PDGFRB) (not shown). However, staining for cleaved CASPASE3 (cCASP3) reveals that, whereas apoptotic cells are rare to absent in the tips (Fig. 5e, inset 1), the cells in the center outside the tubules were largely apoptotic (Fig. 5e, inset 2). These may be remnants of the mesenchymal cells present in the LBOs.

As scRNAseq is now a standard analysis method, we provide scRNAseq data of 25-, 40-, 60- and 80-d-old organoids that confirm the IF data, and also provide a more granular view of differentiation. This analysis revealed ten clusters (Fig. 7a), with, as expected, a progressive gain of distal lung markers and disappearance of mesenchymal and other endodermal markers (Fig. 7b). While many clusters could not be assigned to specific cell types and probably represent transitional states, as is to be expected in differentiating, fetal-stage organoids, some clusters could be identified (Fig. 7c). Cluster 8 consisted of only proliferating cells. However, we note that Ki67⁺ cells were also present throughout, and their frequency decreased over time. All clusters except cluster 9 expressed FOXA1 (Fig. 7c) and EPCAM (not shown), and are therefore endodermal, consistent with our published IF studies. Cluster 9 are mesenchymal cells whose abundance and proliferation (as evaluated by Ki67 expression) declines over time, again consistent with our previously published studies using flow cytometry²⁰. These expressed, in addition to classical mesenchymal markers such as THY1, ACTA2, ZEB1 and COL1A2, also TBX4 and WNT2B. TBX4 is expressed in pulmonary mesenchyme⁵³, whereas WNT2B is expressed in the endoderm surrounding ventral AFE and essential for induction of a pulmonary fate in this area⁵². The mesenchyme is therefore consistent with pulmonary or anterior foregut mesenchyme. We identified mesenchyme expressing these markers previously in RNAseq studies of LBOs²⁰. The lung marker NKX2.1 (also known as TTF1) and the lung progenitor marker, CPM²⁴, are first only observed in clusters 0 and 1, and then progressively spread to the other clusters such that at d80 the organoids are entirely NKX2.1⁺CPM⁺. The widespread expression of NKX2.1 confirms our IF studies (Fig. 5c and ref. ²⁰). Furthermore, at d60 and d80, progressive induction of the distal, predominantly AT2-associated markers (NAPSA, SFTPB, SFTA3, LPCAT1, ABCA3) is observed (Fig. 7c). Low levels of expression of these markers are already evident in cluster 1 at d40 as well (Fig. 7c). However, no SFTPC was detected, consistent with the fact that full AT2 maturation occurs late in the development of the organoids and is highly variable²⁰, indicative of slowly progressive development. This is consistent with the reanalysis of our published RNAseq data in Fig. 3a, where even at d170 expression of SFTPC was not present in all organoids, and was lower than that of the other surfactant proteins and of ABCA3 (ref. ²⁰). We note that cells in cluster 1 also expressed the club cell marker, SCGB3A2, at d25, with expression becoming more diffuse later on. We previously identified small populations of SCGB3A2 in the organoids by IF²⁰. It is possible that the SCGB3A2⁺ cells are precursors of the recently identified SCGB3A2⁺ cells in the most distal airways that may function as transitional intermediates between AT1 and AT2 cells^6–8. Basal cells, defined by expression of p63 and NGFR and mucin-producing cells were not detected (not shown). When a merged (as opposed to integrated)⁶² analysis of the data was performed, each timepoint formed a separate major cluster, at least in part due to batch effects. Nevertheless, the decrease in endodermal markers and the increase in distal markers with time was evident (Fig. 8). scVelo trajectory analysis⁶³ at each timepoint indicated trajectories that ended in the population expressing distal lung markers (NAPSA, SFTPB, SFTA3, LPCAT1, ABCA3) (Fig. 8b). Taken together, the data show progressive acquisition of a distal lung at the expense of alternative endodermal fates and attrition of mesoderm throughout the development of the organoids.

In alveolospheres, stomach and intestinal markers were identified by scRNAseq⁵⁶. Furthermore, from our own previous analysis of RNAseq data, a variable and modest match with human intestine was observed²⁰. These data suggest that other endodermal cells may be present during directed differentiation into lung. We therefore verified the scRNAseq data from the organoids for such markers. Indeed, the small clusters 4 and 5 expressed mid- and hindgut markers, such as CDX2 and GATA4 (ref. ⁶⁴), the expression of which progressively wanes at later stages. However, intestinal and stomach markers could not be detected by IF. To verify whether the NKX2.1⁻FOXA1⁺ cells in the centers of the organoids might expressed other endodermal RNAs but not protein, we punched out the centers of the organoids with a 1,000 μl pipette tip and compared the expression of select endodermal genes in centers and tips by RT–qPCR (Fig. 6). The centers of the organoids were indeed consistently enriched for mRNAs encoding endodermal markers such as CDX2 (midgut and hindgut) and PDX1 (pancreas), whereas some other markers, such as MUC13 and GATA4 (stomach) were more variably enriched. Small populations of cells with other endodermal potentials may therefore be present in the centers of the organoids, at least initially.

Data availability

scRNAseq data are available in the Gene Expression Omnibus database (Gene Expression Omnibus Submission (GSE215825), access code Urihimewhncrbql). The RNAseq datasets that support the findings of this study are from ref. ²⁰, and available from the Sequence Read Archive. The Sequence Read Archive accession number for d25 LBOs sequencing is SRP073749 and SRR4295269 for d170 LBOs. Source data are provided with this paper.