SOX9 maintains human foetal lung tip progenitor state by enhancing WNT and RTK signalling

Abstract

The balance between self-renewal and differentiation in human foetal lung epithelial progenitors controls the size and function of the adult organ. Moreover, progenitor cell gene regulation networks are employed by both regenerating and malignant lung cells, where modulators of their effects could potentially be of therapeutic value. Details of the molecular networks controlling human lung progenitor self-renewal remain unknown. We performed the first CRISPRi screen in primary human lung organoids to identify transcription factors controlling progenitor self-renewal. We show that SOX9 promotes proliferation of lung progenitors and inhibits precocious airway differentiation. Moreover, by identifying direct transcriptional targets using Targeted DamID, we place SOX9 at the centre of a transcriptional network, which amplifies WNT and RTK signalling to stabilise the progenitor cell state. In addition, the proof-of-principle CRISPRi screen and Targeted DamID tools establish a new workflow for using primary human organoids to elucidate detailed functional mechanisms underlying normal development and disease.

Keywords: CRISPRi screen; ETVs; Lung organoids; SOX9; Targeted DamID.

Figure 1. A CRISPRi screen identified crucial factors regulating human foetal lung progenitor cell self‐renewal

1. Schematic of introducing inducible CRISPRi system into the organoid cells using serial lentiviral infection.

2. Schematic of the inducible CRISPRi system. The inducible CRISPRi system was controlled by both Dox inducible system at mRNA level and TMP stabilisation of DHFR destabilising domain at protein level to achieve a tight control of CRISPRi function as previously reported (Sun et al, 2021).

3. Representative flow cytometry results showing CD71, a cell surface marker, can be efficiently knocked‐down in the majority of organoid cells after 5 day of Dox and TMP treatment. N = 3 different organoid lines and 3 different gRNAs for CD71 were used. Mean ± SEM is labelled.

4. Workflow for a focussed library CRISPRi screen for transcription factors regulating human foetal lung tip progenitor cell self‐renewal. Parental organoid lines with inducible CRISPRi system were established by lentiviral transduction and expanded. A CRISPRi gRNA library of 300 gRNAs was packaged into lentivirus and was used to infect single cells from two independent parental inducible CRISPRi organoid lines with an infection efficiency of ~ 15%. TagRFP⁺EGFP⁺ double positive organoid cells were collected. A fraction of cells was frozen for analysis of gRNA abundance in the starting population. The rest were seeded in Matrigel, given 8 day to recover and grown into small organoid colonies before treatment with Dox and TMP. Organoids were cultured in self‐renewing medium with Dox and TMP for 2 week before harvest. Organoids were physically broken into pieces for passaging twice during this period. TagRFP^highEGFP⁺ and TagRFP^lowEGFP⁺ fractions were collected separately for downstream genomic DNA isolation and analysis.

5. Volcano plot summarising gRNA abundance changes for each target gene. Strong depletion hits (−log10(FDR) > 2 and log2(FC) < 0) dark blue; strong enrichment hits (−log10(FDR) > 2 and log2(FC) > 0) red; intermediate depletion hits (1 < −log10(FDR) < 2 and log2(FC) < 0) light blue; unchanged genes grey. FDR, false discovery rate.

6. Schematic of the experimental design for serial passaging assay to validate gene knock‐down effects on self‐renewal. ~ 3,000 TagRFP^highEGFP⁺ double positive cells were harvested for each condition and seeded in Matrigel. Organoid cells were given 8 day to recover and grown into small colonies before treating with Dox and TMP. Organoids were then maintained in Dox and TMP and serially passaged by breaking into pieces every 3–4 day.

7. Summary of serial passaging assay results for different knock‐downs. Three independent organoid lines each were transduced with two different gRNAs against the gene targets, making six organoid lines altogether.

8. Representative wide field images to show organoid growth of the indicated gene knock‐down organoids at the indicated passage number. Scale bars denote 1 mm.

Source data are available online for this figure.

Figure EV1. CRISPRi screen quality control

1. Expression levels heatmap of the selected transcription factors in the developing human foetal lung tip progenitor and stalk cells. Data from Nikolić et al (2017).

2. gRNA abundance distribution of the CRISPRi library after cloning into the plasmid vector. One gRNA targeting MYCN was missing; likely due to a gRNA synthesis issue.

3. Pearson correlation of gRNA abundance between different samples indicated in axes. Between two independent CRISPRi parental lines 3 day after lentiviral transduction. R = 0.98 indicated great consistency of lentiviral transduction.

4. Pearson correlation of gRNA abundance between technical replicates (Rep1 and Rep2). Great consistency was observed between TagRFP^high and TagRFP^low populations.

5. Pearson correlation of gRNA abundance between biological replicates. A lower correlation was observed reflecting the variation of human tissue samples.

Data information: Orange circles in (C–E) represent non‐targeting control gRNAs.

Figure EV2. Validation of the CRISPRi screen results

1. qRT–PCR results showing the targeted genes (IRF6, MYBL2 and ZBTB7B) were efficiently knocked down by the inducible CRISPRi system using the gRNAs selected from the CRISPRi gRNA library.

2. Representative EdU staining images of non‐targeting gRNA control and IRF6 or MYBL2 knock‐down experiments.

3. Quantification of the percentage of EdU⁺ cells in each of three parental organoid lines used with non‐targeting control, IRF6 knock‐down and MYBL2 knock‐down. n = 1,649, 1,705, 3,548 cells were scored for NT controls. n = 2,517, 950, and 1,313 cells were scored for IRF6 gRNAs. n = 1,098 and 1,306 cells were scored for MYBL2 gRNAs.

4. qRT–PCR results showing ARID5B was not knocked down by the inducible CRISPRi system using the gRNAs selected from the CRISPRi gRNA library.

Data information: Error bars: mean ± SEM. Statistical analysis was using the two‐tailed paired t‐test. P‐values are reported as follows: *P < 0.05, **P < 0.01, ***P < 0.001 and n.s. non‐significant. N = 3 organoid lines (biological replicates) used for each panel. Source data are available online for this figure.

Figure 2. SOX9 regulates human foetal lung progenitor cell self‐renewal

1. Human foetal lung tip organoids faithfully express in vivo tip progenitor markers, SOX2 and SOX9. Upper panel: SOX2/SOX9 dual expression in human foetal lungs. Lower panel: SOX2/SOX9 in human foetal lung tip organoids. SOX9 green, SOX2 red, E‐Cadherin white.

2. qRT–PCR showing that SOX9 gRNAs effectively knocked‐down SOX9 transcript levels using inducible CRISPRi after 5 day of Dox and TMP treatment. Three different organoid lines and two different SOX9 gRNAs were used. Error bars: mean ± SEM. Two‐sided Student's t‐test with equal variance ***P < 0.001.

3. SOX9 was knocked down at the protein level using the inducible CRISPRi system after 4 day of Dox/TMP treatment.

4. SOX9 knock‐down organoids at Passage #4 (P4).

5. Summary of the serial passaging assay results for SOX9 knock‐downs. Three independent organoid lines each were transduced with two different gRNAs against SOX9, making six different organoid lines altogether. For serial passaging assay, organoids were maintained in Dox and TMP and broken into pieces every 3–4 day.

6. Schematic of RNA‐seq to identify SOX9 downstream targets. Two independent inducible CRISPRi parental organoid lines were transduced with 2 different non‐targeting control gRNAs, two different SOX9 gRNAs and two different SOX2 gRNAs, respectively. Organoids were supplemented with Dox/TMP for 5 day before harvesting for RNA‐seq.

7. Venn Diagram showing the overlapping number of differentially expressed (DE) genes in SOX9 knock‐down organoids from different parental lines.

8. All SOX9 positively and negatively regulated genes were used to score against an scRNA‐Seq dataset from 9 to 11 pcw human foetal lung epithelium (left panel, preprint: He et al, 2022). SOX9 negatively regulated genes were primarily enriched in secretory lineage populations (middle panel). SOX9 positively regulated genes were primarily enriched in tip progenitor cells (right panel).

Data information: Scale bars denote 50 μm (A, C) and 100 μm (D). Source data are available online for this figure.

Figure EV3. SOX2 and SOX9 knock‐down resulted in different transcriptome changes

1. Unsupervised hierarchical clustering of non‐targeting control, SOX2 knock‐down and SOX9 knock‐down RNA‐Seq results.

2. Venn diagram showing minimal overlap of differentially expressed genes after SOX2 knock‐down in two different parental organoid lines. Overlapping DE genes were labelled in boxes.

3. qPCR of selected DE genes from SOX9 RNA‐seq data following SOX9 knock‐down in a further 2 independent organoid lines. Cells harvested 5 days after knock‐down. Error bars: mean ± SEM. Statistical analysis was using the two‐tailed paired t‐test. P‐values are reported as follows: **P < 0.01, ***P < 0.001. N = 3 bio‐replicates (Organoid line BRC2174 with two different NT gRNAs and two different SOX9 gRNAs, and Organoid line BRC2136 with 1 NT gRNA and 1 SOX9 gRNA) were used.

4. Sashimi plot to visualise splicing junction of NT control and SOX9 KD. Upper panel: Sashimi plot was used to visualise splicing junction information in non‐targeting gRNA control and SOX9 knock‐down groups. Junctional reads between intron #1 and exon #2 were only observed in SOX9 knock‐down groups and not in non‐targeting gRNA control groups. Lower panel: major SOX9 domains in relation to the SOX9 genomic locus. Exon #1 contains DIM and part of the HMG domain. DIM, dimerization domain. HMG, high‐mobility group domain. PQA, proline‐glutamine‐alanine repeats domain. TA, transactivation domain.

5. Heatmap of gene expression from representative GO terms: cell division and small molecule metabolism together with gene expression of upregulated non‐lung lineage genes. OL, organoid line.

6. Selected GO enrichment in DE genes after SOX9 knock‐down.

Source data are available online for this figure.

Figure 3. Targeted DamID (TaDa) identified SOX9 directly regulated targets

1. Schematic of SOX9 TaDa lentiviral construct design. SOX9 is fused with E.coli DNA adenine methylase (Dam). The fusion protein binds to SOX9 binding targets and methylates adenine in the sequence GATC. Methylated GATC is specifically recognised and cleaved by DpnI restriction enzyme. TaDa is designed to produce an extremely low level of Dam‐Sox9 fusion protein: an mNeonGreen open reading frame (ORF1) was placed in front of Dam‐SOX9 fusion (ORF2). The two ORFs were separated by two stop codons and a frame‐shift C (represented by 2 black diamonds), such that the Dam‐SOX9 fusion protein is translated at very low levels after rare translational re‐initiation events.

2. Venn Diagram showing the overlap between DE genes identified in the SOX9 knock‐down RNA‐seq experiment and genes annotated from SOX9 TaDa peaks.

3. Heatmap showing expression level of all 171 SOX9 directly regulated genes across nontargeting control and SOX9 knock‐down organoid lines.

4. CD44 protein expression in SOX9⁺ tip progenitor cells. SOX9 green, SOX2 red, CD44 white.

5. LGR5 mRNA expression enrichment in SOX9 expressing tip progenitor cells. SOX9 red, LGR5 cyan.

6. Tip progenitor cells are of high WNT signalling activity. NOTUM yellow, SOX9 red.

7. Design of constitutively activated β‐catenin overexpression lentiviral construct.

8. qRT‐PCR showing rescue of SOX9 transcription after constitutively activated β‐catenin overexpression in organoids cultured without WNT activators. N = 3 different organoid lines. Error bars: mean ± SEM.

9. WB showing SOX9 protein was rescued after constitutively activated β‐catenin overexpression in organoids cultured without WNT activators.

Data information: Scale bars = 50 μm in all panels. Source data are available online for this figure.

Figure EV4. SOX9 directly activates tip cell genes and represses secretory cell genes

1. Summary of enriched TF binding motifs in SOX9 TaDa peaks. The SOX motif was enriched, indicating the SOX9 TaDa faithfully identified SOX9 binding sites across the genome.

2. SOX9 direct transcriptional target enrichment in human foetal lung scRNA‐seq data. All SOX9 direct transcriptional targets were used for scoring. Similar to Fig 2H, SOX9 directly activated targets were enriched in tip progenitor cells (left panel), whereas SOX9 directly repressed targets were enriched in secretory cell lineages (right panel).

3. SHH was co‐expressed with SOX9 in human foetal lung tip progenitor cells. SOX9 in yellow and SHH in red. No‐probe controls are shown in the right panel.

4. LEF1 and WIF1 were co‐expressed with SOX9 in human foetal lung tip progenitor cells. SOX9 in red and LEF1 (left panel) and WIF1 (right panel) in yellow.

5. Lentiviral construct design for overexpressing SOX9 in human foetal lung progenitor cells.

6. Representative images showing organoid morphology does not change after 3 days of SOX9 overexpression. SOX9 overexpressed organoid indicated with arrow.

7. qRT–PCR results showing that 3 days of SOX9 overexpression led to ETV5 and MYCN transcription being significantly upregulated, however, ETV4 and CFTR were not changed. N = 4 organoid lines (bio‐replicates) were used. Error bars: mean ± SEM. Two‐tailed Student's t‐tests were performed. P‐values are reported as follows: *P < 0.05; ***P < 0.001.

Data information: Scale bars denote 50 μm (C, D) and 100 μm (F). Source data are available online for this figure.

Figure 4. SOX9 directly regulates ETV4 and ETV5 expression, thereby enhancing FGF signalling

1. A, B

   HCR images showing ETV5 (A) and ETV4 (B) expression in SOX9 expressing tip progenitors. ETV4/5 red, SOX9 cyan.

2. C

   qRT–PCR showing ETV4 and ETV5 dual knock‐down in tip organoids. N = 3 different organoid lines and 2 different ETV4 and ETV5 gRNA combinations were used. Error bars: mean ± SEM. Two‐tailed paired t‐test: **P < 0.01; ***P < 0.001.

3. D

   Summary of the serial passaging assay results for ETV4; ETV5 dual knock‐downs. Three independent organoid lines each were transduced with 2 different ETV4; ETV5 gRNA combinations, making 6 different organoid lines altogether. Organoid cells were given 8 days to recover after plating and grown into small colonies before treating with Dox and TMP. Organoids were then maintained in Dox and TMP and serially passaged by breaking into pieces every 3–4 day.

4. E

   Representative images of organoids grown in self‐renewing medium (left) and self‐renewing medium without RTK ligands (EGF, FGF7 and FGF10) (right).

5. F

   qRT–PCR of ETV4, ETV5 and SOX9 expression level in organoids in self‐renewing medium and self‐renewing medium without RTK ligands. N = 3 different organoid lines used. Error bars: mean ± SEM. Two‐tailed paired T‐test: **P < 0.01; n.s. nonsignificant.

6. G

   Venn diagrams showing overlap of SOX9‐ETV5 TaDa peak regions (upper) and overlapping genes from TaDa peak annotations (lower).

7. H

   Examples of SOX9 and ETV5 genomic occupancies showing co‐regulation. At the CD44 locus (upper), SOX9 binds to an intron, whereas ETV5 primarily occupied the promoter. By contrast, for the LGR5 gene (lower), SOX9 and ETV5 both showed occupancy at the promoter.

8. I

   Schematic model of SOX9 regulation of tip progenitor cell self‐renewal.

Data information: Scale bars denote 50 μm (A, B) and 100 μm (E). Source data are available online for this figure.

Figure EV5. Identification of direct ETV5 binding targets

1. qRT–PCR showing SOX9 transcription 6 day after FGF10 supplementation (500 ng/ml), or FGF10 supplementation and removal of WNT activators. N = 4 different organoid lines. Error bars: mean ± SEM. Statistical analysis was using the two‐tailed paired t‐test. P‐values are reported as follows: **P < 0.01; ***P < 0.001; n.s. nonsignificant.

2. Pearson correlation of SOX9, ETV4 and ETV5 TaDa. ETV4 and ETV5 TaDa exhibited great consistency.

3. Motifs enriched in ETV5 TaDa peaks. The ETS binding motif was highly enriched.

4. Genomic occupancy annotated features for SOX9 and ETV5 peaks.

5. Venn diagram showing overlap of differentially expressed genes after ETV4 and ETV5 double knock‐down in two different parental organoid lines. Overlapping DE genes related to cell division by GO analysis are labelled in the box.

6. Heatmap showing expression level of all 42 DE genes after ETV4; ETV5 double knock‐down across different organoid lines. Directly regulated genes are marked by asterisks.

Source data are available online for this figure.