1. Cancer Biology
  2. Cell Biology

EHMT2 methyltransferase governs cell identity in the lung and is required for KRAS G12D tumor development and propagation

  1. Ariel Pribluda  Is a corresponding author
  2. Anneleen Daemen
  3. Anthony Nelson Lima
  4. Xi Wang
  5. Marc Hafner
  6. Chungkee Poon
  7. Zora Modrusan
  8. Anand Kumar Katakam
  9. Oded Foreman
  10. Jefferey Eastham
  11. Jefferey Hung
  12. Benjamin Haley
  13. Julia T Garcia
  14. Erica L Jackson
  15. Melissa R Junttila  Is a corresponding author
  1. Department of Translational Oncology, Genentech, Inc, United States
  2. Department of Bioinformatics & Computational Biology, Genentech, Inc, United States
  3. Department of Immunology, Genentech, Inc, United States
  4. Department of Molecular Biology, Genentech, Inc, United States
  5. Department of Pathology, Genentech, Inc, United States
  6. Department of Genetics, Stanford University, United States
  7. Department of Discovery Oncology, Genentech, Inc, United States
https://doi.org/10.7554/eLife.57648
Share this article
Cite this article
  1. Ariel Pribluda
  2. Anneleen Daemen
  3. Anthony Nelson Lima
  4. Xi Wang
  5. Marc Hafner
  6. Chungkee Poon
  7. Zora Modrusan
  8. Anand Kumar Katakam
  9. Oded Foreman
  10. Jefferey Eastham
  11. Jefferey Hung
  12. Benjamin Haley
  13. Julia T Garcia
  14. Erica L Jackson
  15. Melissa R Junttila
(2022)
EHMT2 methyltransferase governs cell identity in the lung and is required for KRAS G12D tumor development and propagation
eLife 11:e57648.
https://doi.org/10.7554/eLife.57648
  2. Download BibTeX
  3. Download .RIS

Abstract

Lung development, integrity and repair rely on precise Wnt signaling, which is corrupted in diverse diseases, including cancer. Here, we discover that EHMT2 methyltransferase regulates Wnt signaling in the lung by controlling the transcriptional activity of chromatin-bound β-catenin, through a non-histone substrate in mouse lung. Inhibition of EHMT2 induces transcriptional, morphologic, and molecular changes consistent with alveolar type 2 (AT2) lineage commitment. Mechanistically, EHMT2 activity functions to support regenerative properties of KrasG12D tumors and normal AT2 cells—the predominant cell of origin of this cancer. Consequently, EHMT2 inhibition prevents KrasG12D lung adenocarcinoma (LUAD) tumor formation and propagation and disrupts normal AT2 cell differentiation. Consistent with these findings, low gene EHMT2 expression in human LUAD correlates with enhanced AT2 gene expression and improved prognosis. These data reveal EHMT2 as a critical regulator of Wnt signaling, implicating Ehmt2 as a potential target in lung cancer and other AT2-mediated lung pathologies.

Editor's evaluation

This is generally a well-designed and carefully conducted study that is likely to be of interest to many both inside and outside of the field of lung tumorigenesis and normal lung development and tissue homeostasis.

https://doi.org/10.7554/eLife.57648.sa0

Introduction

Lung cancer is the leading cause of cancer mortality in men and women, surpassing combined deaths from colon, prostate and breast cancer (https://www.cancer.org/, 2019). Approximately 40% of non-small cell carcinoma (NSCLC) is of the adenocarcinoma subtype. Genomic alterations of KRAS and TP53 mutations are the most frequent events within this tumor subtype (Cancer Genome Atlas Research, 2014). Considerable evidence indicates that alveolar type 2 (AT2) cells are the predominant cell of origin of lung adenocarcinoma (LUAD) (Desai et al., 2014; Mainardi et al., 2014; Sutherland et al., 2014; Xu et al., 2012).

Sustained tumor growth is maintained, in some instances, by a subset of plastic, stem-like cells, referred to as tumor-propagating cells (TPCs) (Batlle and Clevers, 2017; Beck and Blanpain, 2013). This rare cell subset is less sensitive to therapeutic intervention, and functionally contributes to tumor regrowth following treatment responses (Beck and Blanpain, 2013). TPCs are hypothesized to evolve extensively from the tumor-initiating event (Reya et al., 2001; Shackleton et al., 2009), yet the fidelity to its cell of origin remains undefined. TPCs in LUAD have been previously characterized and were shown to be required for KrasG12D tumor self-renewal (Zheng et al., 2013). Moreover, a TPC gene signature correlates with poor prognosis in non-small cell LUAD patients (Zheng et al., 2013). While these data implicate a critical function for TPCs in driving tumor progression, it is unknown how these rare cells maintain their stem-like properties.

Wnt signaling is of critical importance in the lung, controlling tissue development, homeostasis and repair processes following lung damage. Wnt signaling is critical in the distal airway to maintain AT2 cell fate and function in the alveolar compartment. A subset of AT2 cells serve as adult tissue stem cells, replenishing themselves, as well as alveolar type 1 (AT1) cells, which are both required to maintain proper alveolar function. Deletion of β-catenin in AT2 cells leads to AT1 cell fate differentiation, demonstrating that β-catenin-mediated Wnt signaling is required for AT2 cell identity (Frank et al., 2016; Nabhan et al., 2018). Along the same lines, genetic models have also shown the forced expression of a stabilized form of β-catenin engenders aberrant cell fate in mouse lung (Pacheco-Pinedo et al., 2011). Together these data indicate that precise Wnt pathway activity is important for alveolar cell function, as well as, cell fate decisions. Furthermore, disruption of Wnt signaling equilibrium can have major consequences in some disease pathologies, including lung cancer (Nabhan et al., 2018; Tammela et al., 2017; Zacharias et al., 2018). What remains unclear is how Wnt signals are intrinsically regulated in cells to maintain or acquire facultative stem-like properties in both homeostatic and disease contexts. Understanding the mechanisms that underlie cell fate decisions may enable therapeutic approaches to enhance or prevent these processes for patient benefit.

In many biological contexts, cellular self-renewal and lineage fate commitment is controlled by epigenetic mechanisms, such as chromatin remodeling (Easwaran et al., 2014; Hemberger et al., 2009; Widschwendter et al., 2007). Recent work identified expression of the EHMT2 lysine methyltransferase (G9a) as a poor prognostic factor in LUAD (Chen et al., 2010; Huang et al., 2017). Interestingly, EHMT2 activity has been implicated in the regulation of cell identity during development primarily through its epigenetic histone methyltransferase activity (Chen et al., 2012; Epsztejn-Litman et al., 2008). While EHMT2 methyltransferase activity and its regulation of chromatin is well-described; the non-histone targets of epigenetic regulators are emerging as critical signaling mediators. Here, we investigated how EHMT2 methyltransferase functions to regulate cell fate in distinct cellular contexts relevant for lung tumor development, propagation and AT2 biology. We discover a mechanism of cell intrinsic Wnt signaling control that is governed by EHMT2 activity, establishing EHMT2 as a crucial arbitrator of cell fate gene expression in the lung. By discovering this alternative cell-intrinsic mechanism of governing Wnt signaling in cells, we are able to manipulate cell fate decisions to predictably limit the functionality of these cells to disable tumor propagation, development and AT2 trans-differentiation.

Results

EHMT2 activity is required for KrasG12D;Trp53 (KP) tumorsphere self-renewal

Given the association of EHMT2 expression and poor prognosis in LUAD (Chen et al., 2010; Huang et al., 2017), we sought to examine the expression and function of EHMT2 in primary murine KrasG12D;Trp53-/- (KP) tumors. Previous work established that KP tumor self-renewal was dependent on a TPC subset. Therefore, by utilizing the previously characterized surface markers CD24, ITGB4, and NOTCH (Zheng et al., 2013), we sorted the TPC population and evaluated EHMT2 protein expression. Using two distinct detection methods we observed a consistent and robust increase in EHMT2 protein expression (Figure 1A–C). Next, we evaluated the requirement of EHMT2 in self-renewal function of TPCs using ex vivo KP-derived organotypic cultures (i.e. tumorspheres), an established surrogate for measuring the in vivo regenerative capability of TPCs (Zheng et al., 2013). Pharmacological inhibition of EHMT2, using UNC0642 (Liu et al., 2013), resulted in a dose-dependent decrease of primary ex vivo KP-derived tumorsphere formation (Figure 1D). Pharmacological inhibition of EHMT2 in established tumorspheres led to a marked reduction in both histone H3 lysine 9 di- and tri-methylation marks (H3K9me2/3), consistent with potent EHMT2 inhibition (Collins and Cheng, 2010; Epsztejn-Litman et al., 2008; Shinkai and Tachibana, 2011; Figure 1—figure supplement 1). Both UNC0642 treatment or short hairpin RNA (shRNA)-mediated depletion of Ehmt2 similarly impaired secondary sphere formation (Figure 1F–I), albeit only shEHMT2.2 reached statistical significance, establishing a requirement for EHMT2 activity in TPC self-renewal.

Figure 1 with 1 supplement see all
Download asset Open asset
EHMT2 activity is required for KrasG12D;Trp53 (KP) tumorsphere self-renewal.

(A) Western blot analysis of EHMT2 in tumor-propagating cell (TPC) and non-TPC. Actin was used as loading control. (B) Flow cytometry analysis of EHMT2 in TPC and non-TPC (blue, TPC, red, non-TPC; gray, Isotype). (C) Quantification of EHMT2 geometric fluorescence intensity (gMFI) in (B) (n=2, mean ± SEM; two-tailed t-test, *p=0.05). (D) Tumorsphere formation of KP-derived primary cells seeded with increasing doses of EHMT2 inhibitor (n=2, mean ± SEM, One-way ANOVA with multiple testing, *p<0.005). (E) Western blot analysis showing reductions in H3K9me2/3 following EHMT2 inhibitor treatment (EHMT2i, EHMT2 inhibition) histone H3 was used as loading control. Ratio of H3K9me to H3 is depicted at the bottom of the Western blot. (F) Representative image of primary tumorspheres following secondary passaging in the absence of either vehicle control or EHMT2 inhibitor (EHMT2i, EHMT2 inhibition. Scale bar 100 μm). (G) Quantitation of tumorsphere growth after secondary passaging (n=5; mean ± SEM; two-tailed paired t-test, **p<0.005). (H) Relative qRT-PCR of Ehmt2 transcripts from primary tumorspheres, expressing either short hairpin RNA (shRNA) control (shControl), or shRNAs against Ehmt2 (shEhmt2. 1, shEhmt2. 2), (n=2; mean ± SEM; two-tailed paired t-test, *p<0.05). (I) Quantification of turbo RFP (tRFP)-positive tumorspheres following secondary passage of primary tumorspheres expressing control or Ehmt2 shRNAs. (shEhmt2.1, n=2; mean ± SD, shEhmt2.2, n=3; mean ± SD; two-tailed paired t-test, *p<0.05, shEHMT2.1 p=0.09).

Figure 1—source data 1

Western blot for Figure 1A showing G9a (EHMT2) in non-TPC vs TPC.

https://cdn.elifesciences.org/articles/57648/elife-57648-fig1-data1-v2.zip
Figure 1—source data 2

Western blot (right side) for Figure 1A showing ACTIN in non-TPC vs TPC.

https://cdn.elifesciences.org/articles/57648/elife-57648-fig1-data2-v2.zip

EHMT2 is required for in vivo tumor growth

Previous work has demonstrated that serial re-growth of KP tumors following orthotopic transplantation requires a sustained functional TPC population (Zheng et al., 2013). To evaluate the functional necessity of EHMT2 activity on in vivo tumor formation, we evaluated tumor formation following serial transplantation of orthotopically transplanted KP-derived primary cells harboring doxycycline (Dox)-inducible shRNAs. First, hairpin expression was induced in vivo for 13 days by doxycycline administration to mice with established primary lung tumors. Thereafter, shRNA-expressing cells were sorted from primary tumors and assessed for secondary tumorsphere formation ex vivo and tumor formation in vivo (Figure 2—figure supplement 1A-C). Sorted shEhmt2-expressing cells from primary recipients showed a significant decrease in both ex vivo tumorspheres and in vivo secondary tumor formation, establishing a role for EHMT2 in maintaining TPC stemness (Figure 2A–D). Continuous monitoring of secondary transplants in vivo revealed a substantial growth impairment in sh Ehmt2 -expressing tumors, which translated to a significant increase in overall survival (Figure 2E–F). Mice harboring shEhmt2-expressing tumors eventually succumb to tumor outgrowth; however, analysis of terminal tumors revealed re-expression of Ehmt2 transcript to a level equivalent to that of control tumors (Figure 2—figure supplement 2A,B). This data demonstrates that Ehmt2 expression is required for TPC-tumor growth. Taken together, these data indicate that EHMT2 activity in TPCs functions to maintain the self-renewal capacity of KP tumors.

Figure 2 with 2 supplements see all
Download asset Open asset
EHMT2 is required for in vivo tumor self-renewal.

(A) Ex vivo analysis of tumorsphere formation from primary orthotopic transplanted cells, expressing either short hairpin RNA (shRNA) control (shcontrol) or shRNAs against Ehmt2 (shEhmt2) (n=3 ± SD) Scale bar 500µM. (B) Quantification of tumorsphere formation in panel (A), two tailed t-test, ***p<0.0005 (C) Representative μ-CT images of shcontrol- or shEhmt2-expressing tumors (n=6) red circles depicting tumors. (D) Table comparing efficiencies of secondary passage in vivo and ex vivo from orthotopically-transplanted primary KP cells, expressing shcontrol (n=6) or shEhmt2 (n=6). (E) Overall tumor volume in secondary recipient mice orthotopically transplanted with KP cells from primary recipients, expressing either shRNAs targeting control (shcontrol) or Ehmt2 (shEhmt2.1) (n=6) tumor volume at end of study by treatment, Mann-Whitney test, *p<0.05. (F) Graph indicates survival of mice depicted in (E) (n=6 per group. Geha-Breslow-Wilcoxon test, **p<0.005).

EHMT2 preserves TPC function by preventing AT2 differentiation

To elucidate the mechanistic basis of EHMT2 in maintaining tumor self-renewal, we characterized the phenotypic impact of pharmacological inhibition of EHMT2 in tumorspheres. EHMT2 inhibition resulted in significant reductions in BrdU-labeled cells (fivefold) and expression of cleaved caspase-3 (>10 fold) (Figure 3—figure supplement 1A-D). Reduced proliferation and cell death were previously associated with cell differentiation (Domen and Weissman, 1999; Ruijtenberg and van den Heuvel, 2016), therefore, we further explored cell fate changes as a possible treatment outcome. Since LUAD arises predominantly from the distal alveolar compartment (Mainardi et al., 2014; Sutherland et al., 2014; Xu et al., 2012), we quantified established gene signatures pertaining to distal alveolar cell types from RNA sequencing data derived from EHMT2 i-treated tumorspheres (Treutlein et al., 2014). The AT2 gene signature was significantly increased following EHMT2 inhibition, in contrast to other cell lineages (Figure 3A). Protein expression of surfactant protein C (SPC), a canonical AT2 marker was also significantly upregulated in EHMT2 -inhibited tumorspheres, likely due to differentiated progenitors (Figure 3—figure supplement 2). Concomitant with an increase in SPC+, we observed an increase in CD74, an additional cell surface marker characterizing AT2 cells (Lee et al., 2013), showing an increase of the double-positive population (1.8-fold) (Figure 3B and C). Moreover, we observed a significant increase in the transcript levels of multiple surfactants in both EHMT2-inhibited and EHMT2-depleted (shEhmt2) tumorspheres (Figure 3—figure supplement 3A.B), indicating increased/enhanced AT2-like cell fate features when EHMT2 activity is impaired. Importantly, this AT2-like conversion was confirmed in Ehmt2-depleted tumor cells from secondary passage in vivo (Figure 3—figure supplement 4). Furthermore, evidence supporting cell fate transition was observed when transmission electron microscopy (TEM) of EHMT2-inhibited tumorspheres revealed a significant increase in lamellar bodies (Balis and Conen, 1964), which are distinct specialized structures responsible for the storage and release of surfactants and serve as a morphometric readout for AT2 cells (Figure 3D and E). Together these results demonstrate that EHMT2 inhibition triggers an enhanced/reinforced AT2-like cell state in KP-derived tumorspheres.

Figure 3 with 6 supplements see all
Download asset Open asset
EHMT2 preserves TPC function by preventing AT2 differentiation.

(A) Graphs show enrichment analyses of distinct alveolar cell-lineage gene signatures in transcriptomes generated from KP-derived primary tumorspheres following EHMT2 inhibition (EHMT2i) vs vehicle control (control) (n=4; mean Z-score ± SEM, two-tailed paired t-test, *p<0.05), each paired with immunofluorescence (IF) micrographs of representative canonical marker from their respective cell lineage. Scale bar 100µM (See Figure (S3E) for quantitation of IF). (B) Representative flow cytometry of cells derived from primary tumorspheres treated with either vehicle control (control) or EHMT2 inhibitor (EHMT2i) for 5 days and immuno-stained for alveolar type 2 (AT2) markers surfactant protein C (SPC) and CD74. (C) Quantification of the SPC-CD74 double positive (DP) population depicted in (B). (n=4; mean ± SEM, two-tailed paired t-test, *p<0.05). (D) Representative transmission electron microscopy (TEM) image of cells extracted from primary tumorspheres, treated as in (B). (Upper panel, scale bar 2 μm; lower panel, respective insets in the upper panel, scale bar 0.5 μm). (N, nucleus; yellow arrows, lamellar bodies). (E) Quantification of TEM in (D) (n=2; mean ± SEM, two-tailed paired t-test, *p<0.05). (F) Representative flow cytometry of tumor-propagating cells (TPCs) sorted after EHMT2 inhibitor (EHMT2i)- or vehicle control-treatment of primary tumorspheres and immuno-stained for AT2 markers SPC and CD74. (G) Quantification of (F), showing fold-change in EHMT2i/control ratio of AT2 markers SPC and CD74 (n=2; mean ± SEM, One-way ANOVA with Tukey’s multiple comparison test, **p<0.005). (H) Relative expression of Sftpc and Cd74 transcripts in EHMT2i vs control; TPC and non-TPC, respectively. (n=3; mean ± SEM, two-tailed paired t-test, *p<0.05). (I) Spearman’s rank correlation analysis between orthogonal human AT2 gene signatures and EHMT2 transcript in 546 human lung adenocarcinomas (LUAD). (n=546, linear regression analysis, ***p<0.0001, r=−0.4).

As EHMT2 activity is critical for the self-renewal of tumorspheres and its expression is selectively enriched in TPCs, we interrogated whether the observed cell fate changes occur within the TPC fraction. EHMT2 inhibition in primary tumorspheres caused a reduction in TPCs, (Figure 3—figure supplement 5), consistent with their reduced stemness. Notably, EHMT2-inhibited TPCs displayed a statistically significant increase (fourfold) in the AT2 surface markers, SPC and CD74, compared to matched controls (Figure 3F and G). Consistently, Sftpc and Cd74 transcripts increased exclusively in TPCs (Figure 3H). Together, these data indicate that EHMT2 inhibition leads to an induced AT2-like cell state within the TPC subset, thereby reducing their regenerative capacity by a mechanism similar in features to differentiation. To assess whether the relationship between EHMT2 activity and cell state extends to human tumors, we used clinical adenocarcinoma specimens and assessed the association between EHMT2 transcript and cell lineage gene signatures (Treutlein et al., 2014) in a panel of 546 LUADs (Cancer Genome Atlas Research, 2014). Consistent with our murine data, EHMT2 gene expression negatively correlates with the AT2 cell gene signature (Figure 3—figure supplement 6), supporting the concept that EHMT2 activity impairs differentiation. Taken together, the data indicate that EHMT2 activity represses an alveolar differentiation program in murine LUAD as a means to preserve stem-like properties that enable tumor self-renewal.

Wnt activation impairs TPC self-renewal and induces AT2 cell lineage marker expression

Previous work indicates that EHMT2 can suppress promiscuous transcription by regulating chromatin structure through the positioning of repressive H3K9 methylation marks in a context-dependent manner (Chen et al., 2012; Kim et al., 2017; Zylicz et al., 2018). We performed an assay for transposase-accessible chromatin using sequencing (ATAC-seq) to characterize chromatin accessibility in EHMT2 inhibitor-treated, tumorsphere-derived TPCs. Surprisingly, very limited changes in chromatin accessibility were observed in TPCs upon EHMT2 inhibition (Figure 4—figure supplement 1A,B), indicating that EHMT2 inhibition does not induce widespread chromatin remodeling: 11 promoter and 320 non-promoter sites became more accessible following EHMT2 inhibition, whereas 3 promoter and 54 non-promoter sites were less accessible (FDR <0.05, fold-change >1.5). The promoter sites were not disproportionately enriched for any pathway, and could not account for the EHMT2-induced phenotype.

In the absence of EHMT2 i-induced chromatin accessibility changes, we investigated signaling pathways that influence alveolar cell fate decisions in TPCs. A subset of AT2 cells in normal lung act as tissue stem cells, and Wnt signaling is critical for the maintenance of their stem cell identity (Nabhan et al., 2018; Zacharias et al., 2018). We reasoned that EHMT2 could regulate Wnt signaling as a means to maintain stemness and prevent differentiation in TPCs. Consistent with this line of reasoning, EHMT2 inhibition resulted in a statistically significant increase in Axin2 expression only within the TPC subset (Figure 4A), implicating a functional link between EHMT2 activity and Wnt signaling.

Figure 4 with 4 supplements see all
Download asset Open asset
Wnt activation impairs TPC self-renewal and induces AT2 cell lineage marker expression.

(A) Relative expression of Axin2 transcripts in tumor-propagating cells (TPCs) and non-TPC following EHMT2 inhibition (EHMT2i) vs vehicle control (control) (n=4 mean± SEM; two-tailed paired t-test, *p<0.05). (B) Representative micrographs of primary tumorspheres passaged to single cells following 5 days of the GSK3b inhibitor, CHIR (5 days CHIR) at indicated doses vs vehicle control (control) and assessed for secondary sphere formation (scale bar 100 μm). (C) Quantification of sphere formation experiments as represented in (b) (n=4 mean ± SEM; two-tailed paired t-test, **p<0.005). (D) Relative expression of Axin2 transcripts in primary tumorspheres (n=5 mean ± SEM, two-tailed paired t-test, *p<0.05). (E) Relative expression of alveolar type 2 (AT2) markers in tumorspheres (n=5 mean ± SEM, two-tailed paired t-test, *p<0.05). (F) Flow cytometry for surfactant protein C (SPC) in primary tumorspheres, treated with two doses of GSK3β inhibitor (CHIR) for 5 days vs vehicle control (control). (G) Relative expression of TCF4-containing AT2 markers in TPCs, treated with EHMT2i vs control (n=6, mean ± SEM, two-tailed paired t-test, *p<0.05).

Since EHMT2 inhibition led to Axin2 upregulation whilst impairing TPC self-renewal and inducing an AT2-like cell fate, we assessed whether Wnt activation alone was sufficient to achieve these outcomes. Indeed, pharmacological activation of Wnt signaling using two doses of the GSK3β inhibitor CHIR99021 (Figure 4B and Figure 4C) impaired TPC stemness as illustrated by reduced tumorsphere self-renewal, consistent with EHMT2 inhibition (Figure 4B–F). Moreover, corresponding Wnt pathway activation and transcriptional activation of AT2 marker genes was observed at secondary passage (Figure 4E), with a concordant dose-dependent increase in SPC surface expression (Figure 4F). To better understand the link between Wnt pathway activation and selective increase of AT2 cell lineage gene expression, we performed an unbiased analysis of transcription factor binding motifs within the promoters of cell lineage signature genes (Treutlein et al., 2014). We found that the Tcf4 motif ranked eighth among 264 tested motifs for AT2 genes, while ranking much lower for other cell lineages, supporting the concept that AT2 cell fate is directly linked to a Wnt-driven signaling process (Figure 4—figure supplement 2). Moreover, 4 out of 7 Tcf4-containing AT2 genes are significantly induced in TPCs upon EHMT2 inhibition (Figure 4—figure supplement 3), further supporting the concept that the EHMT2 -inhibitor phenotype is Wnt-mediated. Notably, the Tcf4-containing AT2 genes that are induced upon EHMT2 inhibition display accessible chromatin configurations at their promoters irrespective of treatment (Figure 4—figure supplement 4), suggesting that these genes are poised to respond to Wnt-mediated signals and therefore would not require chromatin accessibility changes to enable gene expression (Zacharias et al., 2018).

EHMT2 restrains Tcf4 gene transcription by repressing chromatin bound β-catenin through RUVBL2

Given the convincing link between inhibition of EHMT2 activity and Wnt-mediated AT2 gene expression, we sought to elucidate the mechanistic basis for Tcf4-mediated gene transcription. Since we observed only a limited change in chromatin accessibility in response to EHMT2 inhibition, we reasoned that a non-histone substrate may be controlling this process. Previous reports have shown that EHMT2 -dependent methylation of the non-histone substrate RUVBL2 (REPTIN, TIP48, and TIP49b) can repress HIF1α-mediated transcription (Lee et al., 2010). RUVBL2 has also been shown to antagonize β-catenin activity (Bauer et al., 2000; Mao and Houry, 2017). We reasoned that the non-histone substrate, RUVBL2, might function to repress β-catenin activity through a EHMT2 -mediated mechanism. First, we confirmed the RUVBL2 and β-catenin interaction in a human NSCLC cell line (Figure 5A and B). Immunoprecipitation of RUVBL2 showed its ability to interact with β-catenin and HIF1-α proteins in hypoxic and normoxic conditions (Figure 5A), consistent with previously reported results (Lee et al., 2010). Reciprocally, immunoprecipitation of β-catenin showed an interaction with RUVBL2. However, the RUVBL2-β-catenin interaction was reduced exclusively in the EHMT2 -inhibited context (Figure 5B), indicating that the RUVBL2-β-catenin interaction requires EHMT2 activity, analogous to that of the RUVBL2-HIF-1α (Lee et al., 2010). To better visualize the impact of EHMT2 inhibition on the RUVBL2-β-catenin interaction with the relevant TPC subset, we performed proximity ligation assay (PLA). Indeed, PLA confirmed the RUVBL2-β-catenin interaction in TPCs and demonstrated a significant loss of signal in the presence of EHMT2 I, providing additional support for the requirement of EHMT2 activity to maintain the RUVBL2-β-catenin interaction (Figure 5C and D).

Figure 5
Download asset Open asset
EHMT2 restrains Tcf4-mediated gene transcription by repressing chromatin bound β-catenin through RUVBL2.

(A) Western blot demonstrating expression of HIF1-α and β-catenin in A549 lysates immunoprecipitated with RUVBL2 antibody. Cells were cultured under hypoxic conditions (1% O2) vs control (ambient O2). (B) Western blot demonstrating co-immunoprecipitation of RUVBL2 in A549 lysates co-immunoprecipitated with β-catenin antibody. Cells were treated with EHMT2 inhibitor vs control. (C) Proximity ligation assay in EHMT2i-treated vs vehicle-treated tumor-propagating cells (TPCs) (red, RUVBL2-β-catenin proximity ligation; blue, DAPI counterstain). Insets show a magnification of the red signal in nuclei of vehicle-treated TPCs (scale 10 μm). (D) Quantification of normalized nuclei with a positive signal (n=2; mean ± SEM, two-tailed paired t-test, **p<0.05). (E) Cytoplasmic (cyt), nuclear (nuc) and chromatin (chro) subcellular fractionation of EHMT2i-treated (as indicated) tumorspheres compared to control. Histone H4 and Tubulin are loading controls of chromatin and cytoplasmic fractions, respectively. (F) Chromatin immunoprecipitation using RUVBL2 antibody followed by qPCR (ChIP-qPCR) of areas flanking a Tcf4 binding sites in promoters of the alveolar type 2 (AT2) genes Etv5, Slc34a2 and Lamp3 (n=2 mean ± SEM; Etv5, Slc34a2, two-tailed paired t-test, *p<0.05). (G) Representative qPCR of Tcf4 binding motif within promoters of AT2 genes.

Figure 5—source data 1

Western blot for Figure 5E showing G9a (EHMT2) expression in subcellular fractionation of EHMT2i-treated tumorspheres compared to control.

https://cdn.elifesciences.org/articles/57648/elife-57648-fig5-data1-v2.zip
Figure 5—source data 2

Western blot for Figure 5E showing TUBULIN expression in middle membrane (upper band) in subcellular fractionation of EHMT2i-treated tumorspheres compared to control.

https://cdn.elifesciences.org/articles/57648/elife-57648-fig5-data2-v2.zip
Figure 5—source data 3

Western blot for Figure 5E showing histone H4 expression in the bottom membrane and beta catenin expression (upper membrane) in subcellular fractionation of EHMT2i-treated tumorspheres compared to control.

https://cdn.elifesciences.org/articles/57648/elife-57648-fig5-data3-v2.zip

In order to gain spatial insight into how EHMT2 inhibition impacts the relationship between EHMT2, RUVBL2, β-catenin and chromatin, we performed subcellular fractionation of tumorspheres. While EHMT2 inhibition showed loss of both EHMT2 and RUVBL2 proteins from the chromatin fraction, equal amounts of β-catenin remain chromatin-bound (Figure 5E). The sustained levels of chromatin-bound β-catenin following EHMT2 inhibition, suggests a critical role for EHMT2 -mediated RUVBL2 regulation that occurs at the chromatin (Figure 5E). Notably, the cytoplasmic fraction of controls confirms the presence of both β-catenin and RUVBL2, consistent with the cytoplasmic signal observed in the TPC PLA control (Figure 5C–E). Given that we established that RUVBL2 interacts with β-catenin and RUVBL2 abundance is reduced in the chromatin fraction when EHMT2 activity is inhibited, we tested whether RUVBL2 chromatin occupancy is changed specifically on Tcf4-containing AT2 genes in TPCs. We observed over 90% reduction in RUVBL2 promoter occupancy within Tcf4 elements of AT2 genes Slc34a2 and Etv5 (Figure 5F and G). Taken together these results indicate that EHMT2 directly controls Tcf4-containing AT2 gene expression through RUVBL2-mediated repression of β-catenin transcription.

EHMT2 controls Wnt signaling and differentiation within AT2 cells

We next explored whether EHMT2 functions similarly to regulate β-catenin activity in normal AT2 cells. Pharmacological inhibition of EHMT2 in vivo for 6 days demonstrated a robust induction of Axin2 protein expression in primary distal alveolar cells using flow cytometry (Figure 6—figure supplement 1A,B). To assess the impact of EHMT2 inhibitor-mediated Wnt induction on AT2 cell fate, we derived primary AT2 cells from adult murine lung and determined the ability of AT2 progenitors to form alveospheres. Ex vivo culturing of primary AT2 cells leads to the formation of alveospheres with cells resembling both AT2 and AT1 cell fates (Barkauskas et al., 2013; Desai et al., 2014). In contrast to KP tumorspheres, inhibition of EHMT2 activity did not impair ex vivo alveosphere formation; however, the resulting spheres were significantly smaller relative to controls (Figure 6B). During alveosphere formation and expansion, emerging cells express transcriptional and surface markers consistent with an AT2 to AT1 cell differentiation (Barkauskas et al., 2013; Zacharias et al., 2018). By analyzing expression of surface markers indicative of these fates (Desai et al., 2014), we observed that EHMT2 inhibition significantly reduced the proportion of AT1 cells (SPC-PDPN+) from 67.2 to 24.4% (Figure 6—figure supplement 2). Interestingly, while the proportion of AT2 cells (SPC +PDPN-) remained unchanged, we observed a marked increase in double positive cells (SPC +PDPN + ) from 15 to 62%. Together these data indicate that EHMT2 is required for differentiation of the AT2 progenitor pool and for complete and proper AT1 differentiation. Consistent with our previous results in TPCs, EHMT2 inhibition resulted in enhanced Wnt signaling, reflected by increased Axin2 expression, as well as Lgr4 and Lgr5, two prominent Wnt signaling pathway components (de Lau et al., 2011; Figure 6D). The observed increase in Tcf4-containing AT2 marker gene expression supports increased Wnt-mediated activity in primary AT2 cells as in the TPC context (Figure 6E). In contrast to the TPC subset, we observed significant changes in the proportion of AT1 cells, consistent with the reduced plasticity observed in AT2 cells. These data demonstrate that EHMT2-mediated regulation of Wnt signaling and cell fate is also observable in untransformed, primary AT2 cells. To test whether EHMT2 loss is similarly required for the differentiation of AT2 cells in vivo, we deleted EHMT2 in the alveolar compartment by intratracheal administration of an adeno-associated virus encoding Cre (AAV9-Cre)(Nabhan et al., 2018). This approach allowed us to identify the EHMT2-targeted cells using a tdTomato reporter and confirm reduced EHMT2 expression from this EHMT2 -deleted population (Figure 6—figure supplement 3). TdT was expressed exclusively in the alveolar space and colocalized preferentially with SPC-expressing cells (Figure 6—figure supplement 4). To further assess how EHMT2 deletion impacts AT2 differentiation, we subsequently injured the lung using hyperoxia to promote alveolar repair (Figure 6—figure supplement 5). EHMT2 loss in the alveolar compartment showed a significant 2.5-fold increase in double positive (SPC + PDPN + ) cells, in just 4 days following injury consistent with our ex vivo alveosphere findings (Figure 6—figure supplement 6F). Studies evaluating the impact of EHMT2 loss at later stages of the tissue repair process will contribute further to our understanding of its role in epithelial cell fate decisions in the lung. All together these data demonstrate that EHMT2 functions as a cell intrinsic mechanism to directly control cell fate decisions in the context of primary, untransformed AT2 cells.

Figure 6 with 7 supplements see all
Download asset Open asset
EHMT2 controls Wnt signaling and in AT2 cells and impairs tumor initiation.

(A) Scatter and contour plots demonstrating intracellular staining of AXIN2 in CD24 negative (CD24-) epithelial cells sorted from EHMT2-inhibited mice compared to control (n=6 for each group). The blue contour plot shows gating on the CD24- cell population. (B) Sphere number and size in alveospheres after treatment with EHMT2 inhibitor (EHMT2i) vs control. (n=10 ± SEM, two-tailed paired t-test, ***p<0.01). (C) Cell-lineage marker analysis of alveospheres, treated with either vehicle control or EHMT2i. Each panel shows a graph and a pie-chart depicting epithelial percentages of the AT1 marker podoplanin (PDPN) and the alveolar type 2 (AT2) marker surfactant protein C (SPC). (n=2 ± SEM), (D) Relative expression of Axin2, Lgr4 and Lgr5 transcripts (n=6–9 ± SEM, two-tailed paired t-test, **p<0.005, ***p<0.0001). (E) Relative expression of tcf4-containing AT2 transcript markers (n=6 ± SEM; two-tailed paired t-test, **p<0.05, ****p<0.0001). (F) Percentage of SPC +PDPN + double-positive cells out of SPC +in wildtype (control) (n=5) and Ehmt2fl/fl (n=5) groups, 4 days post hyperoxic (75% O2) injury. Quantitation represented as per-mouse, two-tailed paired t-test, *p<0.05 (G) Survival of KrasG12D;Trp53 (KP) (n=16), KrasG12D;Trp53;Ehmt2fl/f (KPE) (n=10), KrasG12D (K) (n=14) and KrasG12D;Ehmt2fl/f (KE) (n=9). Gehan-Breslow-Wilcoxon test, (*p<0.05).

The impaired regenerative capacity of AT2 cells observed by EHMT2 loss, together with the Wnt signaling effects impelled us to examine the outcome of EHMT2 deletion in Kras-dependent tumor initiation–an event that was previously shown to require cell fate alterations in other tissue contexts (Shibata et al., 2018). Additionally, genetic cooperativity between Ras and Wnt signaling pathways has been reported (Juan et al., 2014; Pacheco-Pinedo et al., 2011) and linked to cell fate effects. In this case, mutant beta catenin within Scgb1A1 + cells caused a distal cell fate change and enhanced tumor formation within Scgb1A1 + expressing cells (Pacheco-Pinedo et al., 2011). To assess the impact on tumor initiation we deleted EHMT2 using a conditional allele of the Ehmt2 gene (Ehmt2 fl/fl) together with conditional KrasG12D in the absence or presence of Trp53 loss (KPE: Kraslsl.G12D/wt: Trp53fl/fl; Ehmt2fl/fl and KE: Kraslsl.G12D/wt; Ehmt2fl/fl) KPE mice showed a striking reduction in tumor formation and significant decrease in tumor burden in comparison to KP mice, which translated to a significant increase in overall survival (Figure 6—figure supplement 7A,B). Of note, the observed increase in overall survival in both KPE and KE mice was independent of p53 status. These results demonstrate that EHMT2 is crucial for Kras-mediated tumor initiation and are consistent with the requirement of EHMT2 to enable AT2 and TPC regenerative capacity and cell fate.

Discussion

Wnt signals are critical for maintaining or acquiring facultative stem-like properties in lung tissue and distinct disease contexts. Our work reveals a cell-intrinsic mechanism of Wnt pathway activation governed by EHMT2 methyltransferase activity. We find that EHMT2 inhibition directly activates transcriptional activity of chromatin-bound β-catenin within cells, revealing an intracellular mechanism of Wnt signaling control. Interestingly, we detected RUVBL2 and β-catenin interaction in both the cytoplasm and nucleus. We currently do not completely understand the role of RUVBL2-β-catenin interaction in the cytoplasm and in which subcellular compartment this interaction is regulated. One possible explanation is that RUVBL2-β-catenin complex shuttles between the nucleus and cytoplasm compartments. In the context of lung, EHMT2 activity functions to support regenerative properties of KrasG12D tumors and normal AT2 cells – the predominant cell of origin of this cancer. By discovering this alternative cell-intrinsic mechanism of governing Wnt signaling in stem-like cells, we are able to manipulate cell fate decisions to predictably limit the functionality of these cells to disable tumor propagation.

The ability of AT2 stem populations to give rise to differentiated AT2 and AT1 cells is a critical feature that aids in maintaining the integrity of the distal alveolar compartment (Nabhan et al., 2018; Zacharias et al., 2018). How stemness is regulated and maintained is of considerable interest since its dysregulation is an underlying contributor to diseases of the airway, including idiopathic pulmonary fibrosis and cancer. Wnt signaling is a master regulator of AT2 cell fate within the lung, paramount for development, homeostasis and damage repair. Regulation and tuning of Wnt signals are necessary to ensure proper repair in the alveolar compartment, but the mechanisms are context dependent. For example, β-catenin loss in AT2 cells enhances AT1 conversion, indicating Wnt signaling functionally prevents differentiation of alveolar AT2 stem cells (Frank et al., 2016; Nabhan et al., 2018). Our findings are consistent with this paradigm as functional perturbation of EHMT2 activity enhances Wnt signaling, indicating impairment of AT2 plasticity, as well as the regenerative capacity of TPCs. Of note, another report found reduced Wnt activity following EHMT2 perturbation, in three cell tumor cell lines; A549, H1299, and H1975 which invokes an APC2-mediated mechanism, albeit in a distinct cellular context from what we describe (Zhang et al., 2018). Indeed, the authors report that EHMT2 inhibition impairs cell growth by inhibiting Wnt. Importantly in our studies, we do not observe Wnt inhibition upon EHMT2 manipulation in either the TPC or non-TPC populations (Figure 4A). Additionally, the TPC population responsible for long-term regenerative capability of autochthonous tumors is not represented in established cell lines. Therefore, we view this report as a distinct context, represented by effects seen in three lung cell lines that notably all harbor mutations in ARID4A, a protein responsible for HDAC recruitment. More recent work outlines another link between EHMT2 and Wnt pathway suppression. Pal and colleagues described a role for EHMT2-mediated suppression of Wnt signaling in Rhabdomyosarcoma through activation of DKK1 (Pal et al., 2020). Of note, DKK1 is not expressed in AT2 cells (Habermann et al., 2020), thus represents a distinct mechanism of Wnt regulation from what our work describes. Although the Wnt-mediated suppression differs mechanistically from our report, this work reinforces the concept of a role for EHMT2 in suppression of Wnt signaling. Recent reports (Nabhan et al., 2018; Zacharias et al., 2018; Zepp et al., 2017) describe, how Wnt regulation is remarkably nuanced within the distal lung with the discovery and characterization of distinct niches that provide exogenous Wnt ligands to support neighboring stem cells. In the context of tissue damage, these interactions are likely disrupted and thereafter differentiated cells can adopt a facultative state defined by changes in Wnt signaling. Although autocrine Wnt secretion has been proposed as a mechanism to self-sustain the Wnt signaling requirements while outside of the niche, our discovery provides a means to enable cell intrinsic regulation of β-catenin transcription, providing a safe switch to maintain AT2 cell identity when a mesenchymal Wnt-providing niche, is no longer tethered to the epithelial cell.

In lung tumorigenesis, we observe that EHMT2 loss at the time of tumor initiation leads to a significant reduction in tumor formation, as well as, tumor burden resulting in significantly increased overall survival. Of note, genetic cooperativity between Ras and Wnt signaling pathways has been reported (Juan et al., 2014; Pacheco-Pinedo et al., 2011) and linked to cell fate effects. In the context of Kras-mutant lung tumor initiation, genetic activation of Wnt signaling using mutant beta catenin within Scgb1A1 + cells leads to a distal cell fate change consistent with our results. However, in contrast to our AT2-derived tumor-initiation model, mutant beta catenin expression results in enhanced tumor formation within Scgb1A1+expressing cells (Pacheco-Pinedo et al., 2011). Notably, this work differs in the cell of origin, but in that regard, it is not yet clear how distinct, alternate thresholds of Wnt signaling contribute to determining cell fate within different cell lineages in the lung. Moreover, Pacheco-Pinedo et al., used a stable form of beta catenin, whereas loss of EHMT2 activity leads to signaling of chromatin-bound beta catenin, which likely differ in both signal strength and duration. Others have reported the necessity of Wnt signaling to maintain tumors (Juan et al., 2014) and suggested that Wnt gradients might be critical for self-renewal, in line with the necessity for a specific Wnt threshold to support this process (Tammela et al., 2017). Of note, a recent report surprisingly demonstrated enhanced tumor formation when initiating tumors with mutant Kras and EHMT2 knockdown (Rowbotham et al., 2018), This is in direct contradiction to our findings using a genetic model to delete EHMT2 concurrent with KrasG12D in tumor initiation. Importantly, the shRNA seed sequence used in the Rowbotham study targets nine other target transcripts with 100% homology in addition to Ehmt2 (Supplementary file 1), raising the possibility that genes other than Ehmt2 may be implicated in the described phenotype. Notably, one of the genes targeted by this shRNA construct is Babam1, which is known to increase the metastatic capability of a murine Kras-mutant lung cancer-derived cell line (Chen et al., 2015), (Supplementary file 1). While in contrast to our findings, these technical differences preclude definitive conclusions from those experiments within their study.

Our work implicates EHMT2 as an important mediator of Wnt signals in order to maintain tumor propagating capacity. This cell autonomous mechanism for activating Wnt signaling may be beneficial in contexts of tumor seeding and regrowth, allowing independence from niche factors. EHMT2 activity functions similarly in primary AT2, as well as, Kras-transformed tumor cells, suggesting that TPCs may share functional features of their respective cell of origin. The discovery that EHMT2 functions to directly regulate Wnt signaling deepens our mechanistic understanding of EHMT2 activity and presents potential opportunities for targeting in both lung cancer, as well as other AT2-mediated lung pathologies.

Materials and methods

Key resources table
Reagent type (species) or resourceDesignationSource or referenceIdentifiersAdditional information
Cell lines (Homo-sapiens)A549ATCCCRM-CCL-185Short tandem repeat profiling, SNP fingerprinting, and mycoplasma testing were used for strict quality control.
Cell lines (Homo-sapiens)MRC5ATCCCCL-171Short tandem repeat profiling, SNP fingerprinting, and mycoplasma testing were used for strict quality control.
Cell lines (Homo-sapiens)RKOATCCCRL-2577Short tandem repeat profiling, SNP fingerprinting, and mycoplasma testing were used for strict quality control.
Antibody(rat monoclonal); CD45 biotin conjugatedBD Biosciences553078; clone 20-F11(1:200)
Antibody(rat monoclonal);Ter119 biotin conjugatedBD Biosciences553672; clone Ter119(1:200)
Antibody(goat polyclonal); clone MEC13.3; CD31 biotin conjugatedBD Biosciences553371(1:200)
Antibody(rat monoclonal); CD24 PerCP-eFluor 710eBioscience46–0242(1:300)
Antibody(rat monoclonal); EPCAM-FITCBiolegend118208; Clone G8.8(1:20)
Antibody(rat monoclonal); ITGB4-PEBiolegend123602; clone 346–11 A(1:20)
Antibody(armenian hamster monoclonal); Notch1-APCBiolegend130613; clone HMN1-12(1:80)
Antibody(armenian hamster monoclonal); Notch2-APCBiolegend130714; clone HMN2-35(1:80)
Antibody(armenian hamster monoclonal); Notch3-APCeBioescience17-5763-82; clone HMN3-133(1:80)
Antibody(armenian hamster monoclonal); Notch4-APCBiolegend128413; clone HMN4-14(1:80)
Antibody(rat monoclonal); Fc-BlockBD Biosciences553141; clone 2.4 /G2(1:1000)
Antibody(mouse monoclonal); CD74-BUV395BD Biosciences740274; clone ln-1(1:25)
Antibody(rabbit polyclonal); Pro-SPCAbcamab170699(1:200)
Antibody(rabbit monoclonal); G9aAbcamab185050; clone EPR18894(1:1000)
Antibody(mouse monoclonal); PodoplaninThermoFisherMA5-16113; clone 8.1.1(1:200)
Antibody(rabbit polyclonal); Alexa Fluor 488ThermoFisherA-21206(1:500)
Antibody(mouse monoclonal); BrdUNeoMarkersMS-1058-PO; clone BRD.3(1:200)
Antibody(mouse monoclonal); Cleaved Caspase-3Cell signaling Technology9661(1:300)
Antibody(rabbit polyclonal); CC10SantaCruz Biotechnology9772(1:200)
Antibody(mouse monoclonal); FoxJ1eBioscience14-9965-82; clone 2A5(1:25)
Antibody(rat monoclonal); RageR&D Systems175410; clone MAB1179(1:100)
Antibody(mouse monoclonal); ActinBD Biosciences612656; clone C4(1:20,000)
Antibody(rabbit polyclonal); Axin2Abcam109307; clone EPR2005(1:100)
Antibody(mouse monoclonal); H3K9me2AbcamAb1220(1:1000)
Antibody(rabbit polyclonal); RUVBL2Bethyl LaboratoriesA302-536Western
Antibody(mouse monoclonal); Beta-cateninBD Biosciences610153Western
Antibody(rabbit polyclonal); H3K9me3Active Motif39161(1:1000)
Antibody(mouse monoclonal); Histone H3Cell Signaling Technology3638; clone 96C10(1:1000)
Antibody(rabbit polyclonal); RUVBL2Bethyl LaboratoriesA5302-537AImmunoprecipitation
Antibody(rabbit polyclonal); Beta-cateninThermoFisher71–2700Immunoprecipitation
Chemical compound, drugPE/Cy7 StreptavidinBiolegend405206(1:300)
Chemical compound, drugUNC0642Biotechne5132
Chemical compound, drugCHIR99021Biotechne4423
Commercial assay, kitMatrigelCorning356231
Commercial assay, kitDynabeads Co-IP kitThermoFisher14–321
Commercial assay, kitPerfeCTaQuantaBio95146–005
Commercial assay, kitTrue-ChIP KitDiagenodeC01010140
Commercial assay, kitSAGM mediaLonzaCC-3118
Commercial assay, kitCollagenase/DispaseRocheCOLLD-RO2 ug/ml final concentration
Transfected construct (bacteriophage P1)Adeno CMV-CreBaylor College of MedicineAdeno CMV-CreConcentration 1.2 × 107 plaque-forming units
Transfected construct (bacteriophage P1)Adeno-Flp-Ires-CreBaylor College of MedicineAdeno-Flp-Ires-CreConcentration 1.2 × 107 plaque-forming units
Transfected construct (bacteriophage P1)AAV9-CreVirovekAAV9-Cre2E13 vg/ml dilution 1:60 in 60 ul vol.
OtherBioruptor PicoDiagenodeB01080010
OtherInflux cell sorterBD BiosciencesNA
Software, algorithmFlowjoBD BiosciencesFlowjo.com
Software, algorithmImageJImageJhttps://imagej.nih.gov/ij/

  1. Primer sequences included in Supplementary file 2.

  2. SPC: surfactant protein C.

Mice

KrasLSL-G12D (Jackson et al., 2001), Trp53flox/flox (Jonkers et al., 2001), Trp53frt/frt (Lee et al., 2012), Rosa26LSL-tdTomato (Madisen et al., 2010) were licensed by Genentech Inc All animal studies were approved by the Institutional Animal Care and Use Committee at Genentech and adhere to the Guidelines for the Care and Use of Laboratory Animals (protocols 17–1217, 17–0107, and 18–1833 series). Tumors were induced in KP mice at 8–12 weeks of age using intranasal infection of AdCMV-Cre or Adeno-Flp-IRES-Cre (Baylor College of Medicine) at 1.2 × 107 plaque-forming units (PFU). In vivo treatment with EHMT2 inhibitor UNC0642 was performed by treating mice with either vehicle (60% PEG400/40% H20) or with UNC0642 10 mg/kg, IP, daily (60% PEG400/40% H20) for a duration of 6 days. Lungs were harvested and sorted for CD24- as previously described (Barkauskas et al., 2013; McQualter et al., 2010). For in vivo dosing experimentation of models, animals were randomized into treatment cohorts by tumor measurement, with equal numbers of male and female animals. The animals were dosed and monitored according to guidelines from the IACUC at Genentech, Inc Animals were censored for survival in an unblinded manner based on predetermined morbidity criteria.

Ehmt2 conditional knockout design

Request a detailed protocol

Ehmt2 expression vector was constructed by introducing a Frt-PGK-em7-NEO-Frt as a selection marker. LoxP sites were introduced into the 5’ and 3’ homology sequences flanking the targeted exons (genomic location 34908772–34912090 and 34916244–34918676, respectively). The loxP sites, flanked exon 25–27, constituting the SET catalytic domain of Ehmt2. 5’ and 3’ extra genomic regions were used to design PCR primers to validate the targeted deletion which generates a 2.2 kb fragment upon Cre recombinase administration. Conditional gene deletion in the adult was generated in the lung upon Adeno-Cre administration. Licensing and strain availability upon request from Genentech (Scheme 1).

Scheme 1
Download asset Open asset
Ehmt2 CKO-Genomic Structure.

Orthotopic transplantation studies

For tumor formation, 8–12 week-old recipient mice were intratracheally transplanted with 25,000–45,000 inducible shRNA-carrying primary KP tumor cells per mouse. Cells were resuspended in 60 µL MEM alpha (Gibco) before transplantation and monitored using micro CT as previously described (Zheng et al., 2013).

Cell isolation and tumorsphere preparation

Request a detailed protocol

Pooled KP tumors from 3 to 6 mice per experiment, were extracted out of the lung and completely minced with a razor blade. The material was resuspended in DMEM-F12 media containing 10% fetal bovine serum (FBS), 1 × P/S and 2 µg/ml collagenase-dispase in 50 ml conical tubes and incubated for 1–1.5 hr in a 37°C incubator on a shaking platform. Digested material was sequentially filtered through 70 micron and 40 micron strainers, distributed into 15 ml conical tubes and centrifuged for 5 min at 500 g. Pellets were resuspended in hypotonic lysis buffer (15 mM NH4Cl, 10 mM KHCO3, and 0.1 mM EDTA) for 1–2 min, neutralized with DMEM-F12 and spun down on a 1 ml FBS cushion to remove cell debris. Final cell pellets were resuspended in PBS containing 10% FBS. For tumorsphere assays, KP tumor cells were mixed with Matrigel in tumorsphere media as previously described (Zheng et al., 2013). TPCs were sorted and analyzed using the influx machine (BD). Sorting was performed as previously described by Zheng et al., 2013 with the following modifications. Immune cell lineage content was excluded by sorting with the following antibodies: biotin-conjugated CD45 (BD, 553078, 30-F11,1:200), CD31(BD, 553371, MEC13.3, 1:200), Ter119 (BD, 553672, Ter119, 1:200), thereafter biotin-conjugated antibodies were detected with phycoerythrin (PE)/Cy7 streptavidin (Biolegend, 405206: 1:300), and subsequently stained for EpCAM +FITC (Biolegend, 118208, G8.8, 1:20) and TPC markers CD24 PerCP-eFluor 710 (eBioscience, 46–0242, M1/69: 1:300), ITGB4-PE (Biolegend, 123602, 346–11 A, 1:20) and Notch1 (Biolegend, 130613, HMN1-12, 1:80), Notch2 (Biolegend, 130714, HMN-2–35, 1:80), Notch3 (eBioescience, 17-5763-82, HMN3-133, 1:80), Notch4 (Biolegend, 128413, HMN4-14, 1:80). All anti-notch antibodies are allophycocyanin (APC) conjugated and used as a pool.

Transduction of primary KP tumor cells

Request a detailed protocol

shRNAs containing the following Ehmt2 hairpins shG9.1: 5’ acagcaagtctgaagtcgaa 3’, shG9a.2: 5’ cactgtcaccgtcggcgatga 3’ were synthesized, cloned into a mirE backbone and then subsequently sub-cloned into pInducer-10 (Fellmann et al., 2013; Meerbrey et al., 2011) and transfected with packaging constructs into 293T cells using Lipofectamine 2000 (Thermo Fisher). Viral supernatants were collected 72 hr-post transfection and subsequently concentrated using ultracentrifuge at 25,000 RPM for 2 hr to generate high-titer virus. Freshly-sorted primary KP tumor cells were infected over-night using ultra-low attachment plates (corning) and subsequently transplanted or grown in Matrigel as tumorspheres.

TPC gating strategy

Request a detailed protocol

Cells and doublet cells were excluded from the sort, respectively. EpCAM-Lineage scatter plot was used to gate on epithelial cells. Each TPC marker was gated in an individual scatter plot against a mock gate in the following sub-gating scheme: CD24 + cells gated from EpCAM + cells, ITGB4 + cells gated from CD24 + cells and NotchHI gated from ITGB4 + cells. Non-TPCs were classified as the ‘non-positive’ cells for each individual marker scheme and gated accordingly. All non-TPC were combined in one single tube.

Tumorsphere and TPC assays

Request a detailed protocol

For EHMT2 inhibition assays in KP tumorspheres, 10,000–20,000 KP primary tumor cells from 3 to 6 mice for each biological replicate were seeded in Matrigel for 4–5 days before treatment. Each biological replicate contained n=3–4 technical replicates. Tumorspheres were then treated with either vehicle or 2 µM of the EHMT2 inhibitor Unc0642 (Bio-techne) for an additional 5–7 days. For Wnt activation assays, established tumorspheres were treated with either 0.5 µM or 3 µM CHIR99021 (Tocris) for 5 days. For EHMT2 inhibition assays in TPCs, 5x106–10 × 106 KP primary tumor cells from 6 to 8 mice per experiment were seeded per 5 ml Matrigel plug for 4–5 days before treatment and treated as described above. Cells were then extracted out from Matrigel as single cell suspensions and sorted for the TPC cell subset. TPC isolation directly from treated or untreated cultured tumorspheres was performed using the influx sorter (BD).

Secondary sphere formation assays

Request a detailed protocol

For secondary sphere formation assays following EHMT2 inhibition or depletion, established KP tumorspheres were mechanically and enzymatically dissociated with 2 µg/ml collagenase/Dispase (Roche). To obtain single-cell suspensions, dissociated KP tumorspheres were further treated with Accutase (Corning) for 5–10 min, counted and re-seeded in Matrigel in 3–4 technical replicates per experiment. Matrigel, spheres were then counted using a 10×20 eyepiece containing 0.5×0.5 mm grid.

Alveolar and AT2-derived alveosphere preparations

Request a detailed protocol

AT2 derived alveosphere assays were generated by isolating Epcam+/CD24- cells as previously described (Barkauskas et al., 2013; McQualter et al., 2010). Sorted cells were either intracellularly stained for Axin2 Abcam (109307, EPR2005, 1:100) or alternatively, resuspended in SAGM media (Lonza) and mixed in a 1:1 ratio with growth factor-reduced Matrigel (BD Biosciences). 100 μl of mixed cells/Matrigel suspension was placed in 24-well Transwell inserts (Falcon) and allowed to solidify. To allow alveospheres growth, 5×104 MRC5 human lung fibroblasts (ATCC CCL-171) were seeded in the bottom chamber supplied with 500 μl MTEC media. Media was changed every other day. EHMT2 inhibitor was replaced every 2–3 days.

Tumorsphere immunofluorescence (IF)

KP tumorspheres were either stained in Matrigel or were paraffin-embedded, sectioned and stained as previously described (Huber et al., 2015). For Matrigel preparations, tumorspheres were fixed in 4% PFA for 40 min, washed 3×5 min, permeabilized with 0.5% Triton X-100 for 30 min then blocked in 4% BSA for 30–60 min. Staining was performed in blocking buffer with 0.05% Triton X-100 and washes were performed with 0.1% Triton X-100. Tumorspheres were imaged using a Leica SPE confocal microscope. The following antibodies were used for IF: BrdU (NeoMarkers, MS-1058-PO, BRD.3 1:200), cleaved caspase-3 (Cell Signaling Technology, 9661 1:300) pro-SPC (Abcam, ab170699: 1:200), CC10 (SantaCruz, 9772, 1:200) FoxJ1 (eBioscience, 14-9965-82, 2A5), RAGE (R&D, 175410, MAB1179, 1:100).

BrdU incorporation in tumorspheres

Request a detailed protocol

BrdU labeling reagent (ThermoFisher) was incorporated for 3 hr. Tumorspheres were then processed according to IF procedures described in the previous section.

Quantification of tumorsphere IF images

Request a detailed protocol

Images were taken using SP5 Confocal (Leica). Single projection images of 4–6 z-stack sections from at least 15–20 tumorspheres were constructed and analyzed in the Matlab software package (version R2016b by Mathworks, Natick, MA). Individual cell nuclei were segmented using regional intensity maxima and watershed thresholding on the DAPI channel, and then scored by the presence of relevant IF signal above a global intensity threshold in the area immediately surrounding each nucleus. For BrdU images, signal was calculated in each nucleus around intensity maxima. Quality control images were also created by superimposing the cell scoring mask on the raw image data.

IP and chromatin immunoprecipitation (ChIP)

Request a detailed protocol

For Co-IP experiments β-catenin and RUVBL2 were pulled-down from vehicle or EHMT2 inhibitor-treated A549, using Thermo Fisher (71–2700) or Bethyl (A5302-537A) antibodies, respectively, and utilizing the Dynabeads Co-IP kit (14-321) as indicated by protocol. To detect interacting proteins, membrane was blotted for β-catenin BD (610153) or RUVBL2 Bethyl (A302-536A). ChIP experiments were performed on tumor propagating cells (TPCs) sorted directly from vehicle or EHMT2 inhibitor treated-tumorspheres using the Diagenode True-ChIP protocol (C01010140) with the following modifications. Briefly, 50 K-100K cells were crosslinked with 1% formaldehyde for 8 min, chromatin was sheared to 100–500 bp fragments using the Bioruptor Pico sonicator for 3 cycles of 30 s on/off in 0.5 ml tubes (Diagenode). Quantitation of chromatin was performed using a qubit fluorimeter where at least 10 ng of chromatin were used per experiment. Pulldown was performed with 0.5 ug RUVBL2 antibody Bethyl (A302-537A) overnight then incubated with 30 min preblocked protein A conjugated Dynabeads. Immunoprecipitated material was then washed as indicated by protocol with an additional LiCl buffer wash (0.25 M LiCl, 1% IGEPAL, 1% deoxycholic acid, 1 mM EDTA, 10 mM Tris, pH 8.1). Samples corss-links where then reversed and then samples were preamplified using PerfeCTa, QuantaBio (95146–005). The precipitated DNA and input DNA were quantified by qPCR using specific primers flanking a TCF4 binding site in the promoters of the following genes, Etv5, Slc34a2, Lamp3.

Cell lineage marker analysis

Request a detailed protocol

Mouse cell lineage markers were obtained from Treutlein et al., 2014 using the following selection criteria: Pearson correlation ≥ 0.5, p-value (GBA, BH corrected)<0.05. This resulted in 30 mouse genes characteristic of AT1 cells, 28 genes for AT2 cells, 43 genes for Club cells, and 84 genes for ciliated cells. The biomaRt R package was used to map mouse genes to human orthologs, resulting in 28 human AT1 markers, 26 AT2 markers, 39 club markers, and 65 ciliated markers. For the assessment of lineage markers in human and mouse samples, expression data Z scores were first calculated for each gene across the 546 human LUAD tumors, and across the 10 mouse KP tumorsphere samples. Cell lineage signature scores were then calculated for human adenocarcinoma tumors as the average z-score expression of the human markers for each cell lineage and for mouse samples as the weighted average z-scored expression of the mouse markers with weights set to -log10(p-value). P-values were obtained from Table S4 (Treutlein et al., 2014).

Transmission Electron microscopy (TEM)

Request a detailed protocol

Lung tumorspheres were first fixed in modified Karnovsky’s fixative and then post-fixed in freshly prepared 1% aqueous potassium ferrocyanide-osmium tetroxide (EM Sciences, Hatfield, PA), for 2 hr followed by overnight incubation in 0.5% Uranyl acetate at 40C. The samples were then dehydrated followed by propylene oxide (each step was for 15 min) and embedded in Eponate 12 (Ted Pella, Redding, CA). Ultrathin sections (80 nm) were cut with an Ultracut microtome (Leica), stained with 0.2% lead citrate and examined in a JEOL JEM-1400 TEM at 80kV. Digital imaged were captured with a GATAN Ultrascan 1000 CCD camera. For each biological replicate, 20 randomly selected cells in each treatment were manually scored and quantified for lamellar bodies. Sections were imaged at low and high magnifications, scale bars 2 µm and 0.5 µm, respectively.

qRT-PCR analysis

Request a detailed protocol

RNA isolation for tumorspheres was performed using the RNAeasy Micro plus kit (Qiagen). Tumorspheres were dissociated as described. Samples were then measured using nanodrop and subsequently reverse transcribed using SuperScript III (ThermoFisher). For TPCs an additional step of pre-amplification was performed using TaqMan Preamp Master Mix (ThermoFisher). qRT-PCR was performed using fast advanced PCR Master Mix (ThermoFisher). Hprt and Actin were used as reference genes for all assays. All Taqman Assays used in the study are described in Supplementary file 2.

Tumorsphere subcellular fractionation

Request a detailed protocol

Pooled primary KP tumorspheres (n=6) from EHMT2-treated and control Matrigel cultures were extracted, fractionated and processed to generate cytoplasmic, nuclear and chromatin fractions, using Subcellular Protein Fractionation Kit Thermo Fisher (78840), as defined by the protocol.

Antibody validation assays

Axin2 validation was conducted using both human RKO cells and intestinal sections of APCloxp/loxp vs APCfl/fl; ROSA26CreERT2. Briefly, RKO cells were GSK3-β (CHIR)-treated for 48 hr vs controls, harvested, 4% paraformaldehyde fixed for 10 min and subsequently washed and permeabilized with 0.25% Triton X-100. Cells were then blocked with 2.5% horse serum and stained for Axin2 Abcam (ab109307, EPR2005, 1:100) for flow cytometry. Intestinal epithelial sections were stained with Axin2 (ab109307, EPR2005, 1:25, AR; Citrate pH = 6).

Western blot for TPCs and tumorspheres

Request a detailed protocol

TPCs and non-TPCs were sorted from 60 tumor-bearing mice to generate material for Western. Whole cell extracts were lysed in RIPA buffer (50 mM Tris-HCl, 150 mM NaCl, 1% Deoxycholate 0.1% SDS 1% Triton-X100) supplemented with Halt protease and phosphatase inhibitor cocktails (ThermoFisher). Lysates were quantified using pierce BCA protein assay kit (Pierce). To detect histones in tumorspheres, acid extraction method was used as previously described (Shechter et al., 2007). Antibodies used for Western blotting of TPCs and non-TPCs: G9a (Abcam, ab185050, EPR18894, 1:1000), Actin (BD, 612,656, C4, 1:20,000), was used as loading control. In tumorspheres; H3K9me2 (Abcam, Ab1220, 1:1000), H3K9me3 (Active Motif, 39,161 1:1000) histone H3 (cell signaling technology, 3638, 96C10, 1:1000), was used as loading control.

RNA-sequencing data for tumorspheres

Request a detailed protocol

5 biological repetitions of primary tumorspheres were generated from pooled KP tumors from 6 to 10 mice per experiment. In each experiment matched KP tumorspheres were treated with either vehicle or G9a inhibitor (UNC0642). Total RNA was extracted using Qiagen RNeasy kit as per the manufacturer’s protocol and quality control of RNA samples was performed to determine their quantity and quality. The concentration of RNA samples was determined using NanoDrop 8000 (Thermo Scientific) and the integrity of RNA was determined by fragment analyzer (Advanced Analytical Technologies). 0.5–100 ng of total RNA was used as an input for library preparation using TruSeq RNA Sample Preparation Kit v2 (Illumina). Size of RNA-seq libraries was confirmed using 4200 TapeStation and high sensitivity D1K screen tape (Agilent Technologies). Library concentration was determined by a qPCR-based method using library quantification kit (KAPA). The libraries were multiplexed and then sequenced on Illumina HiSeq4000 (Illumina) to generate 30 M of single end 50 base pair reads.

RNA sequencing data were analyzed with HTSeqGenie (Pau et al., 2012) in BioConductor (Huber et al., 2015) as follows: first, reads with low nucleotide qualities (70% of bases with quality <23) or rRNA and adapter contamination were removed. Reads were then aligned to the reference genome GRCm38 using GSNAP (Wu and Nacu, 2010). Alignments that were reported by GSNAP as ‘uniquely mapping’ were used for subsequent analysis. Gene expression levels were quantified as reads per kilobase of exon model per million mapped reads normalized by size factor (nRPKM), defined as number of reads aligning to a gene in a sample/(total number of uniquely mapped reads for that sample × gene length × size factor).

Assay for transposase-accessible chromatin using sequencing (ATAC-Seq)

Request a detailed protocol

Cells were aliquoted into cryovials, frozen, and shipped to Epinomics, Menlo Park, CA. Cells were processed as previously described (Buenrostro et al., 2013). Paired-end reads were aligned to mouse reference genome GRCm38 using GSNAP (Wu and Nacu, 2010) version ‘2013-10-10’, allowing a maximum of two mismatches per read sequence (parameters: ‘-M 2 n 10 -B 2 -I 1–pairmax-dna=1000–terminal-threshold=1000–gmap-mode=none–clip-overlap’). Duplicate reads and reads aligning to locations in the mouse genome containing substantial sequence homology to the mitochondrial chromosome or to the ENCODE consortium blacklisted regions were omitted from downstream analyses. Remaining aligned reads were used to quantify chromatin accessibility according to the ENCODE pipeline standards with minor modifications as follows. Accessible genomic locations were identified by calling peaks with MACS 2.1.0 (Zhang et al., 2008) using insertion-centered pseudo-fragments (73 bp–community standard) generated on the basis of the start positions of the mapped reads and a width of 250 bp. Accessible peak locations were identified as described; briefly, we called peak significance (cutoff of P=1e-7) on a condition-level pooled sample containing all pseudo-fragments observed in all replicates within each condition. Peaks in the pooled sample, independently identified as significant (cutoff P=1e-5) in two or more of the constituent biological replicates were retained, using the union of all condition-level reproducible peaks to form the atlas. (https://www.encodeproject.org/atac-seq/). The atlas consisted of 184,032 peaks with a median width of 266 bp (ranging from 250 bp to 1531 bp). 20.4% of peaks were located in promoter regions, 32.6% in intergenic regions, and 43.3% in introns.

Chromatin accessibility within each peak for each replicate was quantified as the number of pseudo-fragments overlapping a peak and normalized these estimates using the TMM method (Robinson and Oshlack, 2010). Differentially accessible peaks between control and EHMT2 -inhibited samples were identified using the framework of a linear model, accounting for TPC and non-TPC and implemented with the edgeR R package. Significant differences in chromatin accessibility levels within a peak between EHMT2 i-treated and control-treated samples was set to log2-fold change >1.5 and FDR <0.05. The fold-change of ATAC-seq peaks were used as input for gene set enrichment analysis using the Hallmark MsigDB gene set collection, v6.1 (Liberzon et al., 2015).The integrative genomics viewer (IGV) was used for visualization of ATAC-seq peaks near genes of interest (Robinson et al., 2011). All source code and sequencing data are available at https://github.com/anneleendaemen/G9a.CellIdentity.Lung; Pribluda, 2020.

Cancer genome atlas (TCGA) RNA-sequencing data analysis

Request a detailed protocol

RNA-sequencing (RNAseq) data for 546 LUAD tumors from TCGA (Cancer Genome Atlas Research, 2014) were obtained from the National Cancer Institute Genomic Data Commons (https://gdc.cancer.gov). We employed the same approach for RNAseq data processing and quantification, with human reference genome GRCh38.

Motif enrichment near lineage markers

Request a detailed protocol

For each cell lineage, function findMotifs.pl from HOMER 4.7 (Heinz et al., 2010) was used to identify enriched motifs near cell lineage markers from Treutlein et al., 2014. Motif enrichment was performed with promoter regions defined from –2000 bp to 500 bp relative to TSS, and with the background sets defined as the full genome. Locations of the TCF4 motif near cell lineage markers (from –2000 bp to 500 bp relative to TSS) were determined using findMotifs.pl from HOMER 4.7. The TCF4 motif file was obtained from the HomerMotifDB (http://homer.ucsd.edu/homer/motif/HomerMotifDB/homerResults.html).

Data availability

All source code used in this study has been made available in the R computer language, in a fully documented software and data package. This package is freely available under the Creative Commons 3.0 license and can be downloaded from https://github.com/anneleendaemen/G9a.CellIdentity.Lung, (copy archived at swh:1:rev:207a6bd1022c199eaae3ef0d2fcc9e4edcbb3c69).

The following previously published data sets were used
    1. Treutlein
    (2014) NCBI Gene Expression Omnibus
    ID GSE52583. High throughput quantitative whole transcriptome analysis of distal mouse lung epithelial cells from various developmental stages (E14.5, E16.5, E18.5 and adult).
    https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE52583
    1. The cancer genome atlas
    (2014) tcga-data
    ID luad_2014. Comprehensive molecular profiling of lung adenocarcinoma.
    https://tcga-data.nci.nih.gov/docs/publications/luad_2014/

References

    1. Balis JU
    2. Conen PE
    (1964)
    The role of alveolar inclusion bodies in the developing lung
    Laboratory Investigation; a Journal of Technical Methods and Pathology 13:1215–1229.

Decision letter

  1. Martin McMahon
    Reviewing Editor; University of Utah Medical School, United States
  2. Edward E Morrisey
    Senior Editor; University of Pennsylvania, United States
  3. Martin McMahon
    Reviewer; University of Utah Medical School, United States

In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.

Decision letter after peer review:

Thank you for submitting your article "G9a methyltransferase governs cell identity in the lung and is required for KRAS G12D tumor development and propagation" for consideration by eLife. Your article has been reviewed by 3 peer reviewers, including Martin McMahon as the Reviewing Editor and Reviewer #1, and the evaluation has been overseen by Edward Morrisey as the Senior Editor.

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

As the editors have judged that your manuscript is of interest, but as described below that additional experiments are required before it is published, we would like to draw your attention to changes in our revision policy that we have made in response to COVID-19 (https://elifesciences.org/articles/57162). First, because many researchers have temporarily lost access to the labs, we will give authors as much time as they need to submit revised manuscripts. We are also offering, if you choose, to post the manuscript to bioRxiv (if it is not already there) along with this decision letter and a formal designation that the manuscript is "in revision at eLife". Please let us know if you would like to pursue this option. (If your work is more suitable for medRxiv, you will need to post the preprint yourself, as the mechanisms for us to do so are still in development.)

This is generally a well-designed and carefully conducted study that is likely to be of interest to many both inside and outside of the field of lung tumorigenesis and normal lung development and tissue homeostasis. However, there are substantial conceptual issues and issues of data interpretation that the reviewers have raised regarding the research described in this manuscript. Although it may not be feasible for the authors to address every critique listed below experimentally, the authors should attempt to address the reviewers' critiques. Some of the key issues that were raised include:

1. The results on tumor inhibition in vivo and proliferation of tumorspheres in vitro are not entirely surprising, given previously published literature by Zhang et al. (Molecular Cancer 17: 153 (2018)) showing that UNC0642 and/or shEHMT2 inhibit multiple pathways, including WNT signaling in cancer cell lines in vitro and in xenografts. The inhibitory effect on WNT signaling was ascribed to effects on the expression of APC2 resulting in increased destruction of β-catenin, a paper that authors of this manuscript do not reference. Rather, Pribluda et al. propose a model in which inhibition of G9a/EHMT2 activity in AT2 "stem cells" and in "tumor propagating cells" leads to increased WNT activity in a subset of AT2 cells and tumor cells as measured by Axin2 expression. In the case of AT2 cells this WNT pathway activation is proposed to drive the terminal differentiation of cells into a fully mature, non-proliferative AT2 cell. The authors argue that this loss of self-renewal and the potential to differentiate into AT1 cells limits tumorigenesis driven by the combination of mutationally-activated KRAS(G12D) and silencing of TP53.

2. Others have shown that either pharmacologic or genetic inhibition of WNT signaling has striking inhibitory effects on lung tumorigenesis driven by the RAS/BRAF pathway. In that context, the authors should consider the results presented in Juan et al: "Diminished WNT->β-catenin->c-MYC signaling is a barrier for malignant progression of BRAFV600E-induced lung tumors" (Genes Dev. 2014 28: 561-75). In addition, Pacheco-Pineda et al., "Wnt/B-catenin signaling accelerates mouse lung tumorigenesis by imposing an embryonic distal progenitor phenotype on lung epithelium" (J Clin Invest 121: 1935-1945) further supports the hypothesis that elevated WNT signaling promotes lung tumorigenesis. This research clearly indicates that an insufficiency of WNT signaling is inhibitory to lung tumorigenesis, which stands in contrast to the authors contention that sustained WNT signaling promotes differentiation of TPC cells to a more differentiated AT2 phenotype thereby suppressing lung tumorigenesis.

Major comments:

1. Throughout the manuscript, the authors manipulate their genes/pathways of interest largely in one direction, by inhibiting/depleting G9a, or by stimulating WNT signaling, without looking at the effects of over-expressing or rescuing G9a, or inhibiting WNT signaling. Their findings would be considerably strengthened if they could show that overexpressing G9a and inhibiting WNT signaling produced the reverse outcomes.

2. In the authors' tumorsphere assays, they have not ruled out the effects of G9a inhibition on cell proliferation and/or apoptosis, or assessed the proportion of CD24+ ITGB4+ NOTCH(hi) cells by FACs. These experiments are absent for both the in vitro and transplant experiments. These findings contradict previous literature that suggests that G9a is lost in a different TPC population and G9a inhibition increases another TPC marker, SCA-1 (Rowbotham et al. 2018). The authors should explain if these populations/markers are affected in their experiments and address any differences that they have observed.

3. It is surprising that the increase in CD74 (Difference in SP-C is not significant?) is only restricted to the TPCs and is only ~2 fold. In Figure 3, Supplement 5 the data indicates that TPCs make up ~0.12% of the sphere culture. If this was the case the increase in SPC in the whole sphere culture would be insignificant. However, in Figure 3A SPC levels are significantly higher in the G9a inhibited samples and appear widely distributed across the tumorsphere suggesting that many cells and not just the 0.12-0.02% that are TPCs have increased SPC. How do the authors account for this observation?

4. It is perhaps surprising that the authors did find many changes in chromatin accessibility in their inhibited tumorspheres, whence they conclude that G9a inhibition does not alter chromatin in their cultures. In Figure 1E however they clearly show that levels of the chromatin modifications H3K9me2 and H3K9me3 are significantly reduced suggesting that the chromatin has been altered. The authors should account for this and rule out technical problems with their ATAC-Seq data. Indeed, although the data on β-catenin⌠RUVBL2 is interesting, it remains unclear whether this interaction is indeed brokered by changes in RUVBL2 lysine methylation.

5. In the distal lung, WNT responsive Axin2+ AT2 cells have been demonstrated to act as progenitors which can regenerate a substantial fraction of the alveolar epithelium (Frank et al. 2016, Zacharias et al. 2018, Nabhan et al. 2018). Zacharias et al. suggest that SP-C+/Axin2+ cells are an alveolar progenitor distinct from AT2 cells and Nabhan et al. suggest that Axin2+ AT2 cells have stem cell activity. The authors are suggesting that in the context of a KRAS(G12D)-driven lung adenocarcinoma, WNT signaling and Axin2 expression have the opposite effect and oppose the stem cell activity of the TPCs. As this is in direct contrast to a study using the same mouse model which shows that WNT signaling supports the self-renewal and tumor propagation of TPCs, and inhibiting WNT signaling from the niche impairs tumor growth and progression (Tammela et al. 2017), the authors should test their mechanism in vivo or explain the discrepancy.

6. The authors show the RUVBL2 and β-Catenin interaction is altered by G9a inhibition in a human cell line. Can the authors also show this interaction in their tumorsphere cultures or at least with a mouse KRAS(G12D)/TP53(Null) cell line for relevance to the rest of their results?

7. In Figure 6 the authors should use a pharmacodynamic biomarker to show that the G9a inhibitor reaches the epithelial cells of the lungs and is actively inhibiting di- or tri-methylation of H3K9. The results are not consistent with the conclusion that AT2 stemness has been altered by G9a inhibition, as this has not been directed tested. The data from Figure 6B showing no difference in sphere forming ability suggests against this interpretation. The authors should passage their G9a inhibited alveolar spheres and test this directly. As the proportion of AT2 cells between the two cultures does not appear to be altered we might not expect to see a significant difference. The authors should perform statistical tests to show that the differences between cells with AT2 and AT1 markers are significantly different, even though they appear large. The results seem to suggest not that AT2 stemness is lost, but that the differentiation to AT1 cells is impaired, leading to a majority intermediate type cell (SP-C+/PDPN+) instead of a majority of AT1 cells. This would be consistent with existing literature showing that activating WNT signaling with the CHIR GSK3-β inhibitor increases the proportion of AT2 cells in culture whilst inhibiting WNT signaling increases the proportion of AT1 cells (Zacharias et al. 2018).

8. The authors administer Cre adenovirus to a G9a(fl/fl) reporter mouse to further study the effects of G9a loss in vivo. However, the authors do not describe this allele in their Materials and Methods nor do they cite an original publication of such an allele/mouse. (Perhaps this is the relevant citation: Lehnertz et al., 2010 Activating and inhibitory functions for the histone lysine methyltransferase G9a in T helper cell differentiation and function. J Exp Med. 207: 915-922). However, if this mouse is not published the authors should provide additional information on the provenance of this allele/mouse. The authors should demonstrate that G9a expression is successfully silenced in this tumor model, or if the mechanism they propose is active in vivo, whereby G9a deletion in tumors would be presumed to establish continuous WNT signaling to maintain AT2 differentiation and either prevent tumor initiation or retard tumor growth.

9. The statistical analysis of the SPC+/PDPD+ cells following hyperoxic injury should be reassessed. The >10,000 cells the authors enumerate are not independent of each other and should not be analyzed by a t-test in this manner. An n>10,000 is too high for a t-test and will almost always result in p<0.05, rendering the statistical analysis meaningless. A more relevant way to analyze this data is to average the % SPC+/PDPN+ cells per mouse and then perform a statistical test on these values (n=6), assuming each mouse was independent of the others. A t-test would be appropriate if the data are normally distributed, and if not, a non-parametric ranked test such as Mann-Whitney.

10. If the mechanism suggested by the authors is correct, then they should show that G9a is deleted in the alveolar cells that become SPC+/PDPD+ after hyperoxic injury and not in the cells that escape G9a deletion. Whilst the authors do present evidence that alveolar cell fate appears to be affected by G9a, the 'stemness' of the AT2 cells has not been measured either directly (by lineage tracing their progeny and quantifying their contribution to the new epithelia) or indirectly by studying the dynamics (i.e. timing) of regeneration following injury in G9a(fl/fl) vs normal mice.

11. Throughout the manuscript the authors refer to AT2 "cell fate identity" and to the "trans-differentiation" of AT2 to AT1 cells. This is quite confusing since there has been important work on the role of transcription factors such as SOX2 and NKX2.1 on determining the "identity" of epithelial cells in NSCLCs (e.g. Tata et al. 2018 Dev Cell 44: 679-693 and Snyder et al. 2013 Mol. Cell. 50: 185-99). In this previous research a change in identify involves true switching of fate from lung to non-lung (e.g. intestinal or esophageal) cell types. Consequently, it is unclear whether the differentiation of AT2 cells to AT1 cells should be considered "trans-differentiation" or loss of "cell fate identify" but rather the normal behavior of AT2 cells in homeostasis and repair of the adult lung.

12. The authors should consider the confusing fact that two papers that describe active WNT signaling in a subset of AT2 "stem cells" in the normal mouse lung (Nabhan et al. 2018 and Zacharias et al. 2018) report very different proportions of these cells (~3% vs. ~30%). Indeed, the authors assume that there is a fixed subset of AT2 stem-like cells and that these are maintained in a "plastic" self-renewing state by the correct level of WNT signaling. Perhaps AT2 cells can oscillate between states in which they are more or less differentiated, and that it takes longer for the fully differentiated cells packed with lamellar bodies and associated secretory machinery to start to divide and/or give rise to AT1 cells? Indeed, there is little evidence in vivo that KRAS-driven tumors can only arise in Axin2+ AT2 "stem cells", albeit that WNT signaling is essential for both KRAS and BRAF-driven lung tumorigenesis.

13. To support their model the current authors should consider testing whether their tumorspheres contain both AT2 and AT1 cells and whether treatment with UNC0642 specifically reduces differentiation of AT1 cells and does not affect other pathways regulating cell proliferation and survival, as proposed by Zhang et al. The experiments in which Pribluda et al. test the effect of inhibiting EMHT2 on primary mouse AT2 cells in organoids (alveolosphere) culture are confusing. The authors state, "During alveosphere formation and expansion, emerging cells express transcriptional and surface markers consistent with an AT2 to AT1 cell fate transition". A "cell fate transition" was not ascribed to the normal differentiation of AT2 to AT1 cells by the original authors. The fact that the treated spheres are smaller than normal may reflect reduced cell proliferation – however, this was not assayed and the morphology of the spheres and immunohistochemistry is not shown. The authors state "Interestingly, while the proportion of SPC+/PDPN- AT2 cells (SPC+/PDPN-) remained unchanged, we observed a marked increase in double positive cells (SPC+/PDPN+) from 15% to 62%. Together these data indicate that G9a is required for differentiation of the AT2 progenitor pool and for complete and proper AT1 cell fate transition". However, the significance of dual positive cells in vivo or in vitro is not known and it is unclear how the lack of change in the proportion of AT2 cells can be interpreted as reduced differentiation of AT2 cells. Finally, it should be noted that neither LGR5 nor LGR6 are not normally expressed in lung alveolar epithelial cells in vivo – rather expression is confined to mesenchymal cells.

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

Thank you for resubmitting your work entitled "G9a methyltransferase governs cell identity in the lung and is required for KRAS G12D tumor development and propagation" for further consideration by eLife. Your revised article has been evaluated by Edward Morrisey (Senior Editor) and a Reviewing Editor.

The manuscript has been improved but there are some remaining issues that need to be addressed, as outlined below. Indeed, as you will see from the reviews below, two of the three reviewers are willing to accept the manuscript, whereas the third reviewer has raised a number of additional concerns, not the least of which is the fact that only one of the shEHMT2 hairpins reduces sphere formation despite the fact both hairpins elicit strong reduction of EHMT2 expression. Hence, all three reviewers are concerned about the possibility of off-target effects. Hence, in order to satisfy all three reviewers, and to allow us to accept your manuscript, we ask that:

1. You address the comments raised by Reviewer #2 on a point-by-point basis.

2. That you more explicitly state in the text a concern regarding the issue of the shEHMT2 hairpins.

3. That you address R3's comments regarding citations that document expression of LGR5 in the epithelial compartment.

At this point the BRE, acting in consultation with the eLife Senior Editor overseeing this submission, will assess your response and make a final decision on your manuscript recognizing that, with your departure from Genentech, it is likely not possible for additional experiments to be conducted on this project.

Reviewer #1 (General assessment and major comments):

The authors present a substantially revised manuscript and an accompanying thorough and conscientious response to the reviewers' critiques that, in my opinion, addresses the main concerns that were expressed in the original review. While there remain some important unresolved questions, this is more the hallmark of good science and does not reflect a deficiency in the revised manuscript. Moreover, the authors have further elaborated on the generation of new alleles of Ehmt2. Consequently, I am happy to recommend that the revised manuscript be accepted for publication in eLife.

Reviewer #2 (Recommendations for the authors):

Pribula et al. have revised their manuscript "EHMT2 methyltransferase governs cell identity in the lung and is required for KRASG12D tumor development and propagation. In a lengthy response to the first round of reviews, the authors make some changes to improve their manuscript, particularly with regards to the interpretation of their data. However, they do not appear to have attempted any of the experiments suggested by the reviewers to improve their manuscript, particularly with regard to testing the effects of G9a inhibition/knockdown on the Sca-1+ population of KP tumors in their systems, which has been published to interact differently with G9a loss than the authors report, and connecting the mechanism established in A549 cells to their phenotypes in tumors and alveospheres.

Furthermore, in response to a minor point, the authors have highlighted a significant problem with their manuscript that was overlooked in the previous round of revisions. In Figure 1I it was noted that only one hairpin was denoted as significantly reducing tumorsphere formation. Initially assumed to be a mistake, the authors clarify that indeed only one hairpin significantly reduced tumorsphere formation. This leaves only one valid hairpin showing the same phenotype as G9a inhibition and as shRNA's are notorious for off-target effects, it cannot be concluded from just one hairpin that effects were due to observed reduction in EHMT2 mRNA.

This concern carries over into figure 2 where inducible hairpins are used to study the effects of EHMT2 knockdown on transplanted tumorigenesis in vivo. In the description of these experiments, it is not clear if the same two hairpins were used, and if both hairpins independently produced the same phenotype. Indeed, from the legend of Figure 2E, it appears that only one hairpin was used (shEHTM2 2.1). All shRNA hairpin experiments must be completed with at least two, and ideally 3+ independent hairpins to ensure that the phenotypes observed are not due to off-target effects. The sequence of the hairpins used does also not appear to be in the manuscript. Because these are serious concerns regarding experiments key to the manuscripts claim, we must conclude that the manuscript is still not suitable for publication in it's current state.

Specific points:

In response to point 1. the authors note that in discussing Zhang et al. 2018, and the results that UNC0642 and shEHMT2 inhibited wnt signaling this study relied on three human cancer cell lines, A549, H1299 and H1975. They argue that the results of this paper are context dependent on these cell lines, and not applicable to their primarily mouse GEMM studies. However, the biochemical experiments linking G9a to Wnt regulation in Pribula et al. are performed in A549 cells and not in primary GEMM tumors, organoids or AT2 cells (For obvious practical reasons). As the authors establish their G9a-wnt mechanism in this same cell line where g9a inhibition was shown to suppress wnt signaling, this should be addressed. If A549 cells are such a different biological context to their GEMM model (ARID4A mutations etc), then can they expect that their mechanism found in A549 cells carries through to GEMM tumors but the inverse finding on wnt signaling activation does not? Do the authors observe the same results with G9ai in A549 cells as Zhang et al.?

2. The authors make a complex argument that G9a regulates a precise level of wnt activity, when G9a is inhibited an increase in wnt activity drives tumors cells to a differentiated AT2 cell state. This idea of a distinct Wnt threshold is intriguing. This concept and the context of pacheo-Pinedo et al. and Juan et al. should be addressed much earlier in the manuscript when wnt signaling is first introduced and not only in the discussion. The Wnt-threshold concept suggests that either activating or inhibiting wnt could both be deleterious or supportive to lung tumors, depending on context (i.e cell of origin) which would make clinical use of wnt inhibitors/activators a challenging prospect.

Comments:

The authors regret that they were unable to experimentally inhibit wnt signaling, arguing that because in their model, because G9a functions downstream of where most wnt inhibitors act, e.g. on ligand interactions and the B-catenin complex, and because they argue that only suprabasal wnt activity leads to the observed phenotype. If this is interpretation is correct, it would actually present an elegant method to test their model, as wnt inhibition (very easy to technically administer and can be carefully titred to the correct level) should therefore not be able to rescue the effects of G9a inhibition in their system. We believe that this experiment should be completed and reported as it is simple and would greatly strengthen their claims.

2. The authors highlight some tumorsphere data showing that both proliferation and apoptosis are reduced. The purpose of asking for these data was to confirm if the effects of G9a inhibition were truly on stemness which leads to more organoids forming, or because G9a inhibition might reduce proliferation or increase apoptosis of the tumor cells. The results highlighted from figure S 3-1 show that whilst apoptosis is down, proliferation is also strongly reduced. Therefore are the results from G9ai on tumor growth, formation etc due to changes in proliferation rather than AT2 cell differentiation? Or do the authors consider that these are indistinguishable?

Whilst the discussion of the different TPC populations described is appreciated and adds important context, noting that they have different abundances. The question of whether the other TPC marker, Sca-1 and the Sca-1+ fraction is affected in their experiments remains unaddressed when a simple QPCR for Sca-1 in any of their tumor systems could help to address this.

3. We appreciate the reviewer's explanation, highlighting the difference between the in vivo and tumorsphere results. To incorporate this explanation into the text, the authors should mention that the broadly increased SPC expression observed in the cultures is likely from progenitors and differentiated tumor cells whilst in vivo only TPCs show an increase in genes associated with AT2 differentiation.

4. The authors helpfully remind the reviewer that chromatin accessibility was measured in TPCs and not tumorspheres, where they demonstrate that H3K9me2 and H3K9me3 are, as expected, reduced by G9a inhibition. Although they find very few changes in chromatin accessibility in TPCs, the method they describe is appropriate and if carried out correctly should find differences if they are there. This explanation seems to suggest that chromatin modifications would not be changed in TPCs and differences are restricted to tumorspheres "a more diverse population including both TPCs and non-TPC progeny". The authors would seem to be suggesting, therefore, that unlike tumorspheres in culture, H3K9me2 and H3K9me3 are not reduced by G9a inhibition in TPCs. This would be a surprising result, and at the very least should be investigated by the authors.

5. In the explanation provided the authors refine their explanation of the effects of G9a and WNT signaling on TPCs, to align their conclusions with the existing literature. As mentioned in earlier rebuttal points, they interpret their data to suggest that G9a restriction of Wnt signaling is supportive of TPC function, and if G9a is lost or inhibited then excess wnt signaling can lead to loss of TPC function. This represents an interesting finding, although the authors perhaps overinterpret their findings when they suggest that G9a regulates Wnt to a distinct threshold (author's emphasis).

With the authors data and the existing literature suggest that either too little, or now too much Wnt signaling can be deleterious to TPCs and AT2 cells. Whether there is a distinct threshold of wnt signaling for TPCs/AT2 cells to maintain proper function, and more importantly, what this threshold might be is interesting to speculate about, but goes beyond what the author's data show. It would take a considerable amount of experimentation to demonstrate that such a distinct threshold exists, more than could be reasonably expected in a revision of this manuscript. The authors should amend the description of their results and the discussion appropriately.

6. It is unfortunate that the reviewers were not able to confirm the RUVBL2 and B-Catenin interaction in tumorspheres or TPCs, and that mouse cell lines do not behave like TPCs in order for them to be used as a model. This raises the question then, if the A549 cells used for the mechanistic experiments contain "the relevant population to progeny relationship found in de novo kp tumors". Do A549 cells behave similarly to TPCs in organoid culture in response to G9a inhibition? This is important for the connection of the proposed mechanism to the phenotypes observed in the mouse. If the mechanistic relationship cannot be tested in the phenotypic model(s), then the phenotype should be confirmed in the mechanistic model, otherwise we cannot connect the mechanism to the phenotype even if they are 'operating in the same direction' (positively corelating). As mentioned earlier in this discussion, Zhang et al. found that Wnt signaling was inhibited by G9a inhibition and EHMT2 shRNA in cultures and xenografts of human cells including A549 cells. Whilst this study should not be considered the last word on this topic, it highlights the importance for the authors to test to see if their phenotype can be reproduced in this system used to elucidate their proposed mechanism.

Whilst the authors have convincingly shown that Wnt signaling is modulated upwards in response to G9a inhibition in tumors and AT2 cells, a stronger connection to the proposed mechanisms is required for publication.

7. It is helpful that the authors provide pharmacokinetic data showing consistent plasma concentration of their inhibitor following. We take it that the G9a inhibitor is highly selective over other histone methyltransferases which has been amply demonstrated in the original literature. Whilst direct evidence of G9a inhibition in vivo would (through histone methylation) would be greatly convincing, we appreciate that confirmation of the main phenotypes of increases of wnt signaling and AT2 gene expression through genetic deletion makes it likely that G9a inhibitor is functioning as expected in vivo. In the absence of confirming whether G9a inhibition reduces the stem cell capacity by serial passaging of inhibited AT2 cell cultures, the text changes made by the authors is appropriate.

8. The author's test if there are changes in the proportion of AT1 markers in their tumorsphere assays with G9a inhibition, as they see with their alveolar spheres. Here they do not see a significant change AT1 cell markers, highlight differences between how tumor cells and AT2 cells respond to G9a inhibition, which should be commented on in the discussion.

The authors have already addressed wording changes to the descriptions of their alveolar sphere cultures to reflect that AT2 to AT1 is the normal differentiation process and not an example of trans-differentiation.

Whilst the authors' results are consistent with the explanation that a change in AT2 to AT1 differentiation, the small size of the alveospheres does suggest that changes in proliferation could also contribute to these differences, and really should be explored by the authors. The immunohistochemistry would also be extremely informative as well, what to the AT1/AT2 cells look like?

The authors point out that it is not possible to perform secondary transplantations with their inducible due to the re-expression of G9a in vivo. As the effects on self-renewal cannot be experimentally verified, the authors should modify their claims in the manuscript.

Reviewer #3 (Recommendations for the authors):

1. The authors have addressed some of my previous comments. However, I find their responses to comments 12 and 13 vague and inadequate

2. Provide one or more citations for the claim that, "…multiple lines of evidence support the expression of LGR5 in the epithelial cell compartment." or, if that is not possible, then omit reference to LGR5 expression.

https://doi.org/10.7554/eLife.57648.sa1

Author response

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

As the editors have judged that your manuscript is of interest, but as described below that additional experiments are required before it is published, we would like to draw your attention to changes in our revision policy that we have made in response to COVID-19 (https://elifesciences.org/articles/57162). First, because many researchers have temporarily lost access to the labs, we will give authors as much time as they need to submit revised manuscripts. We are also offering, if you choose, to post the manuscript to bioRxiv (if it is not already there) along with this decision letter and a formal designation that the manuscript is "in revision at eLife". Please let us know if you would like to pursue this option. (If your work is more suitable for medRxiv, you will need to post the preprint yourself, as the mechanisms for us to do so are still in development.)

This is generally a well-designed and carefully conducted study that is likely to be of interest to many both inside and outside of the field of lung tumorigenesis and normal lung development and tissue homeostasis. However, there are substantial conceptual issues and issues of data interpretation that the reviewers have raised regarding the research described in this manuscript. Although it may not be feasible for the authors to address every critique listed below experimentally, the authors should attempt to address the reviewers' critiques. Some of the key issues that were raised include:

1. The results on tumor inhibition in vivo and proliferation of tumorspheres in vitro are not entirely surprising, given previously published literature by Zhang et al. (Molecular Cancer 17: 153 (2018)) showing that UNC0642 and/or shEHMT2 inhibit multiple pathways, including WNT signaling in cancer cell lines in vitro and in xenografts. The inhibitory effect on WNT signaling was ascribed to effects on the expression of APC2 resulting in increased destruction of β-catenin, a paper that authors of this manuscript do not reference. Rather, Pribluda et al. propose a model in which inhibition of G9a/EHMT2 activity in AT2 "stem cells" and in "tumor propagating cells" leads to increased WNT activity in a subset of AT2 cells and tumor cells as measured by Axin2 expression. In the case of AT2 cells this WNT pathway activation is proposed to drive the terminal differentiation of cells into a fully mature, non-proliferative AT2 cell. The authors argue that this loss of self-renewal and the potential to differentiate into AT1 cells limits tumorigenesis driven by the combination of mutationally-activated KRAS(G12D) and silencing of TP53.

To our knowledge, there are only two reports of G9a-mediated WNT regulation – one was retracted and therefore we will not comment on it (Cha et al., NCB 2016; PMID: 28008183). The Zhang et al. work (Zhang et al., Mol Cancer 2018; PMID: 30348169) which the reviewer refers to in his comments, relied on three cell tumor cell lines, A549, H1299 and H1975, to determine the role of G9a in this disease context. Indeed, the authors report that G9a inhibition impairs cell growth and they attribute this effect to Wnt inhibition. In our studies, we do not observe Wnt inhibition upon G9a manipulation in either the tumor-propagating (TPC) or non-TPC populations (Figure 4A). We view this report as a unique single context, represented by effects seen in three lung cell lines that notably all harbor mutations in ARID4A, a protein responsible for HDAC recruitment.The range of response to G9a perturbation varies widely across lung cell lines (https://depmap.org/portal/gene/EHMT2?tab=overview) and therefore it is unknown how the effects in these three cell lines apply. We do not believe that this report retracts from our work. We want to reiterate that we are studying a role for G9a in a distinct cancer context – one not observed in cell lines. Autochthonous primary tumors, known to harbor a unique and rare tumor propagating population (TPC) responsible for long-term regenerative capability. This biological situation is not represented in established cell lines. Furthermore, we also extend these findings beyond the tumor context and into the normal lung. The fidelity between normal lung and the tumor context strengthens our work and provides a broader applicability for G9a function outside of lung cancer. We agree with the reviewer that this work should be cited and discussed and have added the Zhang et al. reference and the above discussion to our manuscript.

More recently, an additional report published in eLife outlined a link between G9a and Wnt pathway suppression, similar to our own work (Ananya et al., EHMT2 epigenetically suppresses Wnt signaling and is a potential target in embryonal rhabdomyosarcoma. eLife 2021). Ananya et al. described a role for G9a-mediated suppression of Wnt signaling in Rhabdomyosarcoma through activation of DKK1, representing a distinct mechanism of G9a/Wnt regulation from what our work describes. DKK1 is not expressed in AT2 cells (Arun et al. Sci. Advanced Medicine 2020), however, the work still reinforces the concept of a role for G9a in suppressing Wnt signaling and we therefore also reference and discuss this work.

Lastly, regarding the role of p53 in our work, while indeed the tumor biology work is performed in the absence of p53 (KP GEMM), the Wnt modulation we observe in primary AT2 cells harbors intact p53. Therefore, the Wnt activation observed upon G9a loss appears to be independent of p53 status.

2. Others have shown that either pharmacologic or genetic inhibition of WNT signaling has striking inhibitory effects on lung tumorigenesis driven by the RAS/BRAF pathway. In that context, the authors should consider the results presented in Juan et al: "Diminished WNT->β-catenin->c-MYC signaling is a barrier for malignant progression of BRAFV600E-induced lung tumors" (Genes Dev. 2014 28: 561-75). In addition, Pacheco-Pineda et al., "Wnt/B-catenin signaling accelerates mouse lung tumorigenesis by imposing an embryonic distal progenitor phenotype on lung epithelium" (J Clin Invest 121: 1935-1945) further supports the hypothesis that elevated WNT signaling promotes lung tumorigenesis. This research clearly indicates that an insufficiency of WNT signaling is inhibitory to lung tumorigenesis, which stands in contrast to the authors contention that sustained WNT signaling promotes differentiation of TPC cells to a more differentiated AT2 phenotype thereby suppressing lung tumorigenesis.

We agree with the reviewer's comments and we would like to explain what seems to the reviewer as an inconsistency of our data with the literature. We are not disputing a requirement for Wnt signaling in maintaining stemness in the lung or in tumor propagation. While the reviewer is correct that Wnt has already been implicated as being required in these processes, in all those cases WNT signaling is not “modulated” per se, but either completely inhibited or constitutively activated.

In our study, we report an increase in basal Wnt signaling/Axin2 expression in response to G9a inhibition from the baseline steady-state level. We can directly link the outcome of this Wnt signaling increase to a cell differentiation effect through mechanistic studies showing a role for G9a in repressing B-catenin on TCF4 promoters of those same differentiation genes. Importantly, the impact of G9a functional impairment in our studies – either through knockdown, enzymatic inhibition or genetic deletion is consistent – increased Wnt signaling in BOTH TPC and in alveospheres leading to a cell state phenotype, indicative of a role for Wnt signaling in regulating AT2 biology.

These findings are not contrary to our reported data supporting a requirement for Wnt signaling – it is distinct, in that we are describing a requirement for a specific level of Wnt activity to maintain stem-like properties within those cells. In our setting, G9a is functioning to selectively regulate Wnt signaling to maintain a distinct threshold – that when lost, Wnt signaling is elevated above the set point leading to differentiation and a change in cell fate and capabilities.

We refer the reviewer back to our discussion where we addressed the differences between our findings and the previously reported ones, concerning the effects of pan inhibition/activation of Wnt signaling on tumor growth. We now include Juan et al. in with the previously reported findings and hopefully clarify the distinction between our work and that of others.

Please see below for the amended discussion:

“Of note, genetic cooperativity between Ras and Wnt signaling pathways has been reported (Juan et al., 2014; Pacheco-Pinedo et al., 2011) and linked to cell fate effects. In the context of Kras-mutant lung tumor initiation, genetic activation of Wnt signaling using mutant β catenin within Scgb1A1+ cells leads to a distal cell fate change consistent with our results. However, in contrast to our AT2-derived tumor-initiation model, results in enhanced tumor formation within Scgb1A1+ expressing cells (Pacheco-Pinedo et al., 2011). Notably, this work differs in the cell of origin, and in that regard, it is not yet clear how distinct, alternate thresholds of Wnt signaling contribute to determining cell fate within different cell lineages in the lung. Moreover, Pacheco-Pinedo et al., used a stable form of β catenin, whereas loss of G9a activity leads to signaling of chromatin-bound β catenin, which likely differ in both signal strength and duration. Others have reported the necessity of Wnt signaling to maintain tumors(Juan et al., 2014) and suggested that Wnt gradients might be critical for self-renewal, in line with the necessity for a specific Wnt threshold to support this process (Tammela et al., 2017).”

In conclusion, our data is not inconsistent with published work, nor does it dispute those findings. Our study is demonstrating that G9a is functioning to maintain a distinct basal threshold of Wnt.

Major comments:

1. Throughout the manuscript, the authors manipulate their genes/pathways of interest largely in one direction, by inhibiting/depleting G9a, or by stimulating WNT signaling, without looking at the effects of over-expressing or rescuing G9a, or inhibiting WNT signaling. Their findings would be considerably strengthened if they could show that overexpressing G9a and inhibiting WNT signaling produced the reverse outcomes.

We show consistent effects of G9a on AT2 biology by three different approaches;(1) two shRNA constructs, (2) a pharmacological approach and (3) genetic deletion of G9a in vivo. By relying on three orthogonal approaches, the results support our conclusion that the effects on G9a biology are acting through modulation of G9a.

We did consider how to experimentally inhibit Wnt inhibition, as suggested by the reviewer, however, we show that G9a control on Wnt signaling occur directly on chromatin-bound β-catenin and therefore lies downstream of both Wnt-receptor ligand interaction and the β-catenin destruction complex, so inhibition of Wnt signaling upstream of these signaling events would not rescue G9a-mediated effects on chromatin. In addition, our work implicates an induction of Wnt above basal Wnt signaling. Experimentally, we are unable to impair only the G9a-induced suprabasal Wnt activity whilst maintaining the required basal Wnt signaling level.

As outlined above, the requirement of Wnt signaling has been previously explored for lung homeostasis and in the context of KRAS tumorigenesis, however these experiments were designed to explore complete inhibition of Wnt signaling. Indisputably, complete lack of Wnt signaling impacts all cell types, yet these studies do not address the requirements or modulation of thresholds of Wnt signaling between discrete populations. Therefore, due to the requirements of a basal Wnt signaling level for both TPCs and non-TPCs, as well as in AT2 stem cells and differentiated AT2 cells, Wnt inhibition experiments were not feasible.

As an alternative orthogonal approach to G9a inhibition, we relied on titrated activation of Wnt signaling (Figure 4 C-F) to solidify the link between G9a-mediated WNT signaling and cell fate outcomes. In these experiments, pharmacological activation of Wnt signaling using two doses of the GSK3β inhibitor CHIR99021, demonstrated impaired TPC stemness and reduced tumorsphere self-renewal, consistent with our data using G9a inhibition. Moreover, corresponding Wnt pathway activation and transcriptional activation of AT2 marker genes was observed at secondary passage (Figure 4E), with a concordant dose-dependent increase in SPC surface expression (Figure 4F). The consistency between direct WNT activation and cell fate outcomes, together with Wnt activation through the myriad of G9a perturbation approaches (ie. pharmacological, short hairpin and genetic deletion) further solidifies our hypothesis.

2. In the authors' tumorsphere assays, they have not ruled out the effects of G9a inhibition on cell proliferation and/or apoptosis, or assessed the proportion of CD24+ ITGB4+ NOTCH(hi) cells by FACs. These experiments are absent for both the in vitro and transplant experiments. These findings contradict previous literature that suggests that G9a is lost in a different TPC population and G9a inhibition increases another TPC marker, SCA-1 (Rowbotham et al. 2018). The authors should explain if these populations/markers are affected in their experiments and address any differences that they have observed.

We thank the reviewer for this question and would like to draw the reviewer’s attention to several figures to help address this query. We report decreased proliferation and reduced apoptosis in tumorspheres following G9a inhibition (Supplement figure 3-1) and demonstrate the proportion of CD24+ ITGB4+ NOTCH(hi) cells by FACS (Supplement figure 3-5). We further show the molecular (Figure 3A) and morphometric features (scanning electron microscopy in Figure 3D-E) that support AT2 differentiation as the final outcome. Additionally, we show an increase in cell fate differentiation markers in tRFP-sorted tumor cells derived from primary transplants. (Supplement figure 3-4). Figures are included below the next paragraph for the ease of the reviewer.

The reviewer also raises the important distinction between the “TPC” populations identified by Zheng and colleagues (original TPC finding, of which our work builds upon) versus Rowbotham and colleagues. In Rowbotham et al. the CD24+ Sca-1 population consists of >15% of all epithelial cells in KP tumors. However, the population discovered by Zheng et al., CD24+/ITGB4+/NotchHI constitutes ~1-2% of the epithelial population. The difference between these populations and their relative abundance makes it difficult to compare them. However, we show in our work that G9a is enriched in CD24+/ITGB4+/NotchHI TPCs by two methodologies, RNA expression and protein expression. Notably, Rowbotham indeed showed that G9a transcript is 2-fold lower in Sca1+ cells, but no confirmation of correlative protein expression was demonstrated. Regardless of the definition of the sub-population of TPC being studied, G9a protein expression levels are more indicative to extrapolate a putative function of G9a biology in the subpopulation.

3. It is surprising that the increase in CD74 (Difference in SP-C is not significant?) is only restricted to the TPCs and is only ~2 fold. In Figure 3, Supplement 5 the data indicates that TPCs make up ~0.12% of the sphere culture. If this was the case the increase in SPC in the whole sphere culture would be insignificant. However, in Figure 3A SPC levels are significantly higher in the G9a inhibited samples and appear widely distributed across the tumorsphere suggesting that many cells and not just the 0.12-0.02% that are TPCs have increased SPC. How do the authors account for this observation?

We agree with the reviewer’s comments, however we don’t see it as inconsistent. G9a inhibitor is maintained in culture for 5-6 days, in which AT2-like TPCs give rise to progenitors that form these tumorspheres. The progenitors would continue to show sustained AT2 marker expression. Ex vivo the number of TPC is higher (0.5.-2%), which would further explain the elevated SPC expression.

4. It is perhaps surprising that the authors did find many changes in chromatin accessibility in their inhibited tumorspheres, whence they conclude that G9a inhibition does not alter chromatin in their cultures. In Figure 1E however they clearly show that levels of the chromatin modifications H3K9me2 and H3K9me3 are significantly reduced suggesting that the chromatin has been altered. The authors should account for this and rule out technical problems with their ATAC-Seq data. Indeed, although the data on β-catenin⌠RUVBL2 is interesting, it remains unclear whether this interaction is indeed brokered by changes in RUVBL2 lysine methylation.

Please note that chromatin accessibility measurements were assessed in TPCs and not in tumorspheres. Our aim was to assess G9a-specific changes within the TPC population itself to understand the mechanism by which G9a inhibition resulted in reduced regenerative capacity for this population – it is within this population that we did not observe significant chromatin accessibility changes upon G9a inhibition. We agree with the reviewer that one may expect more differences in chromatin if comparing a stem-like population to a progeny population, which is why we chose to focus on the “stem-like” population for this analysis.

However, to further clarify our results, even within the TPC population, we do find changes in chromatin accessibility in TPCs, as described in the manuscript: “accessibility increases at 331 sites and decreases at 57 sites upon G9a inhibition”. Although genes with impacted promoter accessibility could not be linked to the phenotype induced by G9a inhibition. We agree that it can be clarified better in the figure itself. As to Fig1E, we indeed show reduction of H3K9me2 and H3K9me3 marks, but this reduction is in tumorspheres – a more diverse population, including both TPCs and non-TPC progeny.

To assuage any further doubt of the ATAC-seq data analysis, we have included more detailed information methods regarding how the QC was performed:

“Assay for Transposase-Accessible Chromatin using Sequencing (ATAC-Seq) Cells were sequenced by Epinomics, and processed as previously described (Buenrostro et al., 2013). Paired-end reads were aligned to mouse reference genome GRCm38 using GSNAP (Wu and Nacu, 2010) Duplicate reads and reads containing substantial sequence homology to the mitochondrial chromosome or to the ENCODE consortium blacklisted regions were omitted from downstream analyses. Remaining reads were quantified according to the ENCODE pipeline with minor modifications: Accessible genomic locations were identified using MACS 2.1.0(Zhang et al., 2008) as insertion-centered pseudo-fragments (73bp with a width of 250bp). Peak significance (cutoff of p=1e-7) was determined on a condition-level pooled sample containing all pseudo-fragments observed in all replicates within each condition. Peaks independently identified as significant (cutoff p=1e-5) in two or more of the biological replicates were retained, using the union of all condition-level reproducible peaks to form the Encode atlas. The atlas consisted of 184,032 peaks with a median width of 266 bp (ranging from 250 bp to 1531 bp). 20.4% of peaks were located in promoter regions, 32.6% in intergenic regions, and 43.3% in introns. Chromatin accessibility within each peak per replicate was quantified as the number of pseudo-fragments overlapping a peak and normalized using the TMM method (Robinson and Oshlack, 2010). Differentially accessible peaks between control and G9a-inhibited samples were identified using a linear model, accounting for TPC and non-TPC and implemented with edgeR. Significance within a peak between G9ai-treated and control-treated samples was set to log2-fold change > 1.5 and FDR < 0.05. The fold-change was used as input for gene set enrichment analysis using the Hallmark MSigDB gene set collection, v6.1 (Liberzon et al., 2015). Integrative Genomics Viewer (IGV) was used for visualization of ATAC-seq peaks near genes of interest (Robinson et al., 2011).”

5. In the distal lung, WNT responsive Axin2+ AT2 cells have been demonstrated to act as progenitors which can regenerate a substantial fraction of the alveolar epithelium (Frank et al. 2016, Zacharias et al. 2018, Nabhan et al. 2018). Zacharias et al. suggest that SP-C+/Axin2+ cells are an alveolar progenitor distinct from AT2 cells and Nabhan et al. suggest that Axin2+ AT2 cells have stem cell activity. The authors are suggesting that in the context of a KRAS(G12D)-driven lung adenocarcinoma, WNT signaling and Axin2 expression have the opposite effect and oppose the stem cell activity of the TPCs. As this is in direct contrast to a study using the same mouse model which shows that WNT signaling supports the self-renewal and tumor propagation of TPCs, and inhibiting WNT signaling from the niche impairs tumor growth and progression (Tammela et al. 2017), the authors should test their mechanism in vivo or explain the discrepancy.

We thank the reviewer for this comment. As a reminder and as discussed similarly in comment 2 above, our work is not challenging the existence of these cell types, nor their requirement for Wnt signaling. These elegant studies all point to an absolute requirement for constitutive basal Wnt signaling. Our work is focused on describing how increases from this basal level can impact cell biology and fate. In fact, these references support the precise concept that differential Wnt signaling levels (ie. AXIN2 levels) distinguish cell types and their capabilities in the lung.

We observe that a specific level of Wnt activity is required to maintain stem-like properties within TPC and AT2 cells – performed by multiple means of G9a manipulation both in vivo (tumor context and normal lung tissue) as well as ex vivo (with primary KP tumor organoids and isolated AT2 cells). In our settings, G9a is functioning to selectively regulate Wnt signaling to maintain a distinct threshold – when that is lost, Wnt signaling is elevated above a set point leading to differentiation and a change in cell capabilities.

It is, indeed, an excellent question as to whether G9a manipulation would, for example, alter this rare Axin2-GFP+ AT2 cell population as described in Nabhan et al. In fact, Nabhan and colleagues show that increased Wnt signaling within their Axin2-GFP+ AT2 population prevents emergence of AT1 from AT2, which is consistent with our data in Figure 6 with Wnt signaling induction mediated through G9a loss.

6. The authors show the RUVBL2 and β-Catenin interaction is altered by G9a inhibition in a human cell line. Can the authors also show this interaction in their tumorsphere cultures or at least with a mouse KRAS(G12D)/TP53(Null) cell line for relevance to the rest of their results?

Further, we considered performing these mechanistic experiments in cell lines derived from KP GEMM tumors; however, pharmacological inhibition of G9a in these cell lines did not impact sphere formation, suggesting that these lines do not harbor the relevant TPC population to progeny relationship found in de novo KP tumors, precluding us from using cell lines for this purpose. See Author response image 1, showing no significant effect with G9a inhibitor in 3 KP derived cell lines.

Author response image 1
Download asset Open asset

We attempted immunoprecipitation from tumorspheres as the reviewer suggests; however, due to the fact that we are isolating TPCs from primary KP tumors, we were unsuccessful in obtaining sufficient material to perform this experiment fully. Therefore, we needed to rely on proximity ligation assay to allow us to explore the interaction in the TPC population.

7. In Figure 6 the authors should use a pharmacodynamic biomarker to show that the G9a inhibitor reaches the epithelial cells of the lungs and is actively inhibiting di- or tri-methylation of H3K9. The results are not consistent with the conclusion that AT2 stemness has been altered by G9a inhibition, as this has not been directed tested. The data from Figure 6B showing no difference in sphere forming ability suggests against this interpretation. The authors should passage their G9a inhibited alveolar spheres and test this directly. As the proportion of AT2 cells between the two cultures does not appear to be altered we might not expect to see a significant difference. The authors should perform statistical tests to show that the differences between cells with AT2 and AT1 markers are significantly different, even though they appear large. The results seem to suggest not that AT2 stemness is lost, but that the differentiation to AT1 cells is impaired, leading to a majority intermediate type cell (SP-C+/PDPN+) instead of a majority of AT1 cells. This would be consistent with existing literature showing that activating WNT signaling with the CHIR GSK3-β inhibitor increases the proportion of AT2 cells in culture whilst inhibiting WNT signaling increases the proportion of AT1 cells (Zacharias et al. 2018).

UNC0642 is more than 20,000-fold selective for G9a and GLP over 13 other methyltransferases (IC50 > 50,000 nM), more than 2,000-fold selective over PRC2-EZH2 (IC50 > 5,000 nM), and displays low cell toxicity (EC50 > 3,000 nM) (Liu et al., J Med Chem 2013). To address the reviewer’s question regarding in vivo exposure of compound, we have now included the pharmacokinetic data corresponding to the G9a inhibitor used throughout our studies and confirmed the in vivo exposures and metabolic stability of UNC0642 (Liu et al., J Med Chem 2013; Author response image 2). Following a single IP dose of 5mg/kg of the G9a inhibitor, the serum concentration remains stable out to 24 hours, consistent with the reported PK properties of this molecule. Unfortunately, we do not have in vivo pharmacodynamic (PD) data beyond the consistent AT2 markers described. presently.

Author response image 2
Download asset Open asset
Pharmacokinetic behavior of UNC0642: (A) concentration from plasma measured at 8 time points with 3 animals per time point after a single 5 mg/kg intraperitoneal dose of UNC0642 using a PEG400/H2O (60/40) formulation.

(B) Summary table of observed parameters from Liu et al., 2013 and Pribluda et al. Note no formulation information is available from Liu et al., 2013.

In addition, we refer the reviewer to supplement 3-4, where we show an increase in AT2 markers in vivo as a result of G9a inhibition. This observed increase in AT2 markers is consistent with our results reported in the alveosphere and tumorsphere cultures, where in the latter, we assessed H3K9 di and tri methylation as a PD readout for G9a inhibition. Importantly, we consistently observe increases in Axin2 and AT2 markers, using genetic deletion and shRNA approaches as a means to inactivate G9a both ex vivo and in vivo (shRNA, pharmacological inhibition and genetic deletion). Therefore, we are strongly convinced that G9a is inhibited in the lung with treatment.

As the reviewer mentions, we do not see differences in sphere forming ability with the AT2 alveosphere cultures, but we disagree that this means our interpretation is flawed. The plasticity of AT2 cells is measured in part by their ability to differentiate to AT1 cells, and indeed despite the lack of change in sphere forming ability, we do observe a significant increase in both Wnt target genes and AT2 cell-lineage markers, consistent with our early observations in the tumor context. Furthermore, we confirm the increase in SPC+/PDPN+ double-positive cells within alveospheres, using a genetic deletion of G9a in vivo. Lastly, we want to draw the reviewer’s attention to Zacharias et al. Nature 2018, where no change in sphere number is detected after Wnt signaling activation (see AT2 alveospheres treated with CHIR vs. DMSO, extended data Figure 9 in that paper). Therefore, given these consistencies, we do not think our interpretation contradicts our results. Moreover, the assay, as performed, is the standard practice used by others in the field to determine stemness of AT2 cells (Nabhan et al. Science 2018; Zacharias et al. Nature 2018).

We agree with the reviewer’s comment that reduced stemness and impaired differentiation could both generate the outcome observed. Because AT2 stem progenitors are believed to be the population responsible for forming alveospheres, we interpreted impaired differentiation as indicative of reduced stemness. We now include language to cover both possibilities by referring to reduced “plasticity”.

Text changes are as follows:

“Together these data indicate that G9a inhibition impairs the plasticity within the AT2 progenitor pool by disrupting proper AT1 cell fate transition.”

And “Together these data indicate that G9a inhibition impairs the plasticity and complete differentiation within the AT2 progenitor pool by disrupting proper AT1 cell fate transition.”

8. The authors administer Cre adenovirus to a G9a(fl/fl) reporter mouse to further study the effects of G9a loss in vivo. However, the authors do not describe this allele in their Materials and Methods nor do they cite an original publication of such an allele/mouse. (Perhaps this is the relevant citation: Lehnertz et al., 2010 Activating and inhibitory functions for the histone lysine methyltransferase G9a in T helper cell differentiation and function. J Exp Med. 207: 915-922). However, if this mouse is not published the authors should provide additional information on the provenance of this allele/mouse. The authors should demonstrate that G9a expression is successfully silenced in this tumor model, or if the mechanism they propose is active in vivo, whereby G9a deletion in tumors would be presumed to establish continuous WNT signaling to maintain AT2 differentiation and either prevent tumor initiation or retard tumor growth.

We apologize for not providing the targeting details of these new alleles. This is now included in the supplemental information together with a description.

Author response image 3
Download asset Open asset

Ehmt2 expression vector was constructed by introducing a Frt-PGK-em7-NEO-Frt as a selection marker. loxP sites were introduced into the 5’ and 3’ homology sequences flanking the targeted exons (genomic location 34908772-34912090 and 34916244-34918676 respectively). The loxP sites, flanked exon 25-27, constituting the SET catalytic domain of Ehmt2. 5’ and 3’ extra genomic regions were used to design PCR primers to validate the targeted deletion which generates a 2.2kb fragment upon Cre recombinase administration. Conditional gene deletion in the adult was generated in the lung upon Adeno-Cre administration.For supportive data that G9a is lost following Cre exposure, please see supplemental figure 6-3.

9. The statistical analysis of the SPC+/PDPD+ cells following hyperoxic injury should be reassessed. The >10,000 cells the authors enumerate are not independent of each other and should not be analyzed by a t-test in this manner. An n>10,000 is too high for a t-test and will almost always result in p<0.05, rendering the statistical analysis meaningless. A more relevant way to analyze this data is to average the % SPC+/PDPN+ cells per mouse and then perform a statistical test on these values (n=6), assuming each mouse was independent of the others. A t-test would be appropriate if the data are normally distributed, and if not, a non-parametric ranked test such as Mann-Whitney.

We repeated the analysis, averaging the data per mouse as suggested by the reviewer. The change remains significant using t-test for statistical analysis. We now use this figure and analysis in the manuscript.

10. If the mechanism suggested by the authors is correct, then they should show that G9a is deleted in the alveolar cells that become SPC+/PDPD+ after hyperoxic injury and not in the cells that escape G9a deletion.

We are not sure what the reviewer is referring to by “escape” in a genetic deletion in the setting of the hypoxic normal lung experiments. G9a deleted cells are labeled with tdTomato reporter (Supplemental figure 6-4), where we have verified that indeed G9a is deleted in these cells (Supplemental figure 6-3). The analysis was conducted only in tdTomato-positive cells and the legend and graph labels now reflect this for further clarity.

Whilst the authors do present evidence that alveolar cell fate appears to be affected by G9a, the 'stemness' of the AT2 cells has not been measured either directly (by lineage tracing their progeny and quantifying their contribution to the new epithelia) or indirectly by studying the dynamics (i.e. timing) of regeneration following injury in G9a(fl/fl) vs normal mice.

The reviewer raises a relevant point regarding the use of “stemness” to describe the observed cell fate changes in response to G9 perturbation. Furthermore, we do not have lineage-tracing evidence in our in vivo AT2 hypoxia experiments to definitively support the AT2-to-AT1 relationship. Moreover, we agree that both the lineage tracing and the dynamic experiments are important follow-up experiments. To ensure that our conclusions are balanced, we removed mention of “stemness” and instead use “plasticity” to more accurately reflect the data. Additionally, we report only on the increased presence of the double positive cells as supportive of our ex vivo data and we incorporate the caveats described above.

“Studies evaluating the impact of G9a loss at later stages of the tissue repair process will further contribute to our understanding of its role in epithelial cell fate decisions in the lung.”

11. Throughout the manuscript the authors refer to AT2 "cell fate identity" and to the "trans-differentiation" of AT2 to AT1 cells. This is quite confusing since there has been important work on the role of transcription factors such as SOX2 and NKX2.1 on determining the "identity" of epithelial cells in NSCLCs (e.g. Tata et al. 2018 Dev Cell 44: 679-693 and Snyder et al. 2013 Mol. Cell. 50: 185-99). In this previous research a change in identify involves true switching of fate from lung to non-lung (e.g. intestinal or esophageal) cell types. Consequently, it is unclear whether the differentiation of AT2 cells to AT1 cells should be considered "trans-differentiation" or loss of "cell fate identify" but rather the normal behavior of AT2 cells in homeostasis and repair of the adult lung.

We agree with the reviewer that this terminology is important and should be used consistently throughout. We have modified the text to refer to AT2-AT1 fate as differentiation. We have removed all transdifferentiation language throughout the manuscript.

See below for example:

“Deletion of b-catenin in AT2 cells leads to AT1 cell fate transdifferentiation”.

Changed to

“Deletion of b-catenin in AT2 cells leads to AT1 cell fate differentiation”.

12. The authors should consider the confusing fact that two papers that describe active WNT signaling in a subset of AT2 "stem cells" in the normal mouse lung (Nabhan et al. 2018 and Zacharias et al. 2018) report very different proportions of these cells (~3% vs. ~30%). Indeed, the authors assume that there is a fixed subset of AT2 stem-like cells and that these are maintained in a "plastic" self-renewing state by the correct level of WNT signaling. Perhaps AT2 cells can oscillate between states in which they are more or less differentiated, and that it takes longer for the fully differentiated cells packed with lamellar bodies and associated secretory machinery to start to divide and/or give rise to AT1 cells? Indeed, there is little evidence in vivo that KRAS-driven tumors can only arise in Axin2+ AT2 "stem cells", albeit that WNT signaling is essential for both KRAS and BRAF-driven lung tumorigenesis.

We agree with the reviewer’s comments. It has been clearly shown that Wnt is essential for tumorigenesis. We do not dispute that nor are we claiming the “AT2 stem cells” are the cell of origin of KRAS/BRAF driven adenocarcinomas.

13. To support their model the current authors should consider testing whether their tumorspheres contain both AT2 and AT1 cells and whether treatment with UNC0642 specifically reduces differentiation of AT1 cells and does not affect other pathways regulating cell proliferation and survival, as proposed by Zhang et al.

The tumorspheres are initiated through isolation of tumor cells, excluding AT2 and AT1 cells; however, these spheres stochastically express AT2 and AT1 markers, nevertheless we assessed the changes using an AT1 transcriptional signature and immunofluorescence and we found no significant changes in AT1 markers after G9a inhibition (See Figure 3). The effect of G9a on proliferation and survival had already been answered in previous comments (Please refer to our answer from comment no.2).

The experiments in which Pribluda et al. test the effect of inhibiting EMHT2 on primary mouse AT2 cells in organoids (alveolosphere) culture are confusing. The authors state, "During alveosphere formation and expansion, emerging cells express transcriptional and surface markers consistent with an AT2 to AT1 cell fate transition". A "cell fate transition" was not ascribed to the normal differentiation of AT2 to AT1 cells by the original authors.

Agree. The statement has been changed to be consistent with the biology reflected in the assay and now reads: "During alveosphere formation and expansion, emerging cells express transcriptional and surface markers consistent with an AT2 to AT1 cell differentiation".

The fact that the treated spheres are smaller than normal may reflect reduced cell proliferation – however, this was not assayed and the morphology of the spheres and immunohistochemistry is not shown.

Indeed, reduction in overall proliferation could explain these results if AT1 cells were exclusively dying. However, we show that the fraction of AT2 cells remains similar in both G9a-treated cultures and controls, whereas it was previously shown that once AT1 cells are generated they exit the cell cycle (Barkauskas et al. JCI 2013, figure S6). Therefore, we interpreted that the increase in intermediate double positive cells comes at the expense of proper AT1 differentiation.

The authors state "Interestingly, while the proportion of SPC+/PDPN- AT2 cells (SPC+/PDPN-) remained unchanged, we observed a marked increase in double positive cells (SPC+/PDPN+) from 15% to 62%. Together these data indicate that G9a is required for differentiation of the AT2 progenitor pool and for complete and proper AT1 cell fate transition". However, the significance of dual positive cells in vivo or in vitro is not known and it is unclear how the lack of change in the proportion of AT2 cells can be interpreted as reduced differentiation of AT2 cells.

The reviewer is correct that we do not fully understand the significance of the dual positive population. However, the sentence proceeding the referenced excerpt stated “By analyzing expression of surface markers indicative of these fates (Desai et al., 2014), we observed that G9a inhibition significantly reduced the proportion of AT1 cells (SPC-PDPN+) from 67.2% to 24.4% (Figures 6C—figure supplement 2)”. We understand that in this assay the reduced AT1 production is indicative of impaired differentiation, consistent with other reports using this assay system (Nabhan et al. Science 2018; Zacharias et al. Nature 2018). We conclude that the differentiation from AT2 to AT1 is impaired and provide further details regarding the double positive cells. While the double population clearly exists in the alveolar assays, it is typically not reported in these types of assays – rather the focus is primarily on AT1 and AT2 cell fates.

Finally, it should be noted that neither LGR5 nor LGR6 are not normally expressed in lung alveolar epithelial cells in vivo – rather expression is confined to mesenchymal cells.

The reviewer is referring to Figure 6D, where expression of Axin2, Lgr4 and Lgr5 is used to confirm Wnt pathway activation in primary AT2-derived murine spheres following G9a inhibition. We presume that the reviewer is basing his statements of lack of Lgr5 on the data published by Lee et al. Cell 2017, where Lgr5 was not detected in the epithelial population by single cell RNA-seq. Because of dropouts, genes undetected in scRNA-seq assays cannot be considered as a confirmation of lack of expression, especially for lowly expressed genes. Indeed, expression of Lgr5 in our own bulk RNA-seq dataset was very low (2-17 reads out of 30-40M reads per sample), which likely means that Lgr5 cannot be detected in standard scRNA-seq assays. Although Lgr5 is a low-expressed gene, it is readily observable by qPCR, most likely due to the amplification steps included in this procedure. While Axin2 and Lgr4 increases suffice to demonstrate wnt pathway activation, we would like to retain Lgr5 as multiple lines of evidence support its expression in the epithelial cell compartment.

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

The manuscript has been improved but there are some remaining issues that need to be addressed, as outlined below. Indeed, as you will see from the reviews below, two of the three reviewers are willing to accept the manuscript, whereas the third reviewer has raised a number of additional concerns, not the least of which is the fact that only one of the shEHMT2 hairpins reduces sphere formation despite the fact both hairpins elicit strong reduction of EHMT2 expression. Hence, all three reviewers are concerned about the possibility of off-target effects. Hence, in order to satisfy all three reviewers, and to allow us to accept your manuscript, we ask that:

1. You address the comments raised by Reviewer #2 on a point-by-point basis.

We have provided a point-by-point response to R2.

We incorporated three text changes to satisfy concerns of R2. Changes are outlined in the point-by-point response below in Point 2 and in Point 14.

2. That you more explicitly state in the text a concern regarding the issue of the shEHMT2 hairpins.

We have added text to more clearly point out that while both hairpins reduce sphere formation, only one reaches statistical significance.

Modified language:

“Both UNC0642 treatment or short hairpin RNA (shRNA)-mediated depletion of Ehmt2 similarly impaired secondary sphere formation (Figure 1F-I), albeit only shEHMT2.2 reached statistical significance, establishing a requirement for EHMT2 activity in TPC self-renewal.”

3. That you address R3’s comments regarding citations that document expression of LGR5 in the epithelial compartment.

The reviewer misunderstood our response to his query regarding LGR5 expression that we detect in our studies. The “multiple lines of evidence” is in reference to our own data contained in this paper, RNAseq and q-PCR. Given the paucity of data on this topic and the fact that LGR5 is a well-established WNT target gene, we are compelled to retain this information for the field, if the editor agrees.

Reviewer #2 (Recommendations for the authors):

Pribula et al. have revised their manuscript “EHMT2 methyltransferase governs cell identity in the lung and is required for KRASG12D tumor development and propagation. In a lengthy response to the first round of reviews, the authors make some changes to improve their manuscript, particularly with regards to the interpretation of their data. However, they do not appear to have attempted any of the experiments suggested by the reviewers to improve their manuscript, particularly with regard to testing the effects of G9a inhibition/knockdown on the Sca-1+ population of KP tumors in their systems, which has been published to interact differently with G9a loss than the authors report, and connecting the mechanism established in A549 cells to their phenotypes in tumors and alveospheres.

Furthermore, in response to a minor point, the authors have highlighted a significant problem with their manuscript that was overlooked in the previous round of revisions. In Figure 1I it was noted that only one hairpin was denoted as significantly reducing tumorsphere formation. Initially assumed to be a mistake, the authors clarify that indeed only one hairpin significantly reduced tumorsphere formation. This leaves only one valid hairpin showing the same phenotype as G9a inhibition and as shRNA’s are notorious for off-target effects, it cannot be concluded from just one hairpin that effects were due to observed reduction in EHMT2 mRNA.

This concern carries over into figure 2 where inducible hairpins are used to study the effects of EHMT2 knockdown on transplanted tumorigenesis in vivo. In the description of these experiments, it is not clear if the same two hairpins were used, and if both hairpins independently produced the same phenotype. Indeed, from the legend of Figure 2E, it appears that only one hairpin was used (shEHTM2 2.1). All shRNA hairpin experiments must be completed with at least two, and ideally 3+ independent hairpins to ensure that the phenotypes observed are not due to off-target effects. The sequence of the hairpins used does also not appear to be in the manuscript. Because these are serious concerns regarding experiments key to the manuscripts claim, we must conclude that the manuscript is still not suitable for publication in it’s current state.

In order to support our conclusions presented in our manuscript we utilized multiple experimental methodologies to target EHMT2. We show consistent effects of EHMT2 on AT2 biology by three different approaches; (1) two shRNA constructs, (2) a pharmacological approach and (3) genetic deletion of EHMT2 in vivo using a new genetic mouse model. By relying on three orthogonal approaches, the results support our conclusion that the effects on EHMT2 biology are acting through modulation of EHMT2.

We acknowledge that only one hairpin results in statistically significant reduction in sphere formation in Figure 1I (shEHMT2.2); however, the lack of statistical significance does not equate to lack of biological relevance. Both hairpins significantly reduce EHMT2 transcript and both result in a clear reduction of sphere formation. To highlight this perceived discrepancy and lack of statistical power, we now include text in the manuscript regarding the lack of statistical significance for one hairpin in this experiment and include the actual p-value (p=0.09) in the legend.

Specific points:

In response to point 1. the authors note that in discussing Zhang et al. 2018, and the results that UNC0642 and shEHMT2 inhibited wnt signaling this study relied on three human cancer cell lines, A549, H1299 and H1975. They argue that the results of this paper are context dependent on these cell lines, and not applicable to their primarily mouse GEMM studies. However, the biochemical experiments linking G9a to Wnt regulation in Pribula et al. are performed in A549 cells and not in primary GEMM tumors, organoids or AT2 cells (For obvious practical reasons). As the authors establish their G9a-wnt mechanism in this same cell line where g9a inhibition was shown to suppress wnt signaling, this should be addressed. If A549 cells are such a different biological context to their GEMM model (ARID4A mutations etc), then can they expect that their mechanism found in A549 cells carries through to GEMM tumors but the inverse finding on wnt signaling activation does not? Do the authors observe the same results with G9ai in A549 cells as Zhang et al.?

The reviewer raises a valid point regarding the differences seen in our work vs. Zhang et al. 2018, whilst both using the A549 cell line. In our work using A549, we observed that EHMT2 perturbation results in increased Axin2 expression, consistent with our hypothesis, please see below our results in A549, therefore we proceeded to work with A549 to help address mechanistic questions not feasible to address in other systems, as the reviewer points out.

Author response image 4
Download asset Open asset
4 day knockdown with siRNA hairpin targeting G9a.

Notwithstanding, the A549 cell line was shown to harbor strains that are genetically and transcriptionally distinct from each other, which in turn could result in different drug responses (Ben-David et al. Nature 2018), thus making it difficult to compare between our conclusions with the A549 cells line vs. Zhang et al. Our claims related to distinct context were referring to this possibility.

2. The authors make a complex argument that G9a regulates a precise level of wnt activity, when G9a is inhibited an increase in wnt activity drives tumors cells to a differentiated AT2 cell state. This idea of a distinct Wnt threshold is intriguing. This concept and the context of pacheo-Pinedo et al. and Juan et al. should be addressed much earlier in the manuscript when wnt signaling is first introduced and not only in the discussion. The Wnt-threshold concept suggests that either activating or inhibiting wnt could both be deleterious or supportive to lung tumors, depending on context (i.e cell of origin) which would make clinical use of wnt inhibitors/activators a challenging prospect.

We agree with the reviewer’s comment.

We added this text in Results section:

“Additionally, genetic cooperativity between Ras and Wnt signaling pathways has been reported (Juan et al., 2014; Pacheco-Pinedo et al., 2011) and linked to cell fate effects. In this case, mutant β catenin within Scgb1A1+ cells caused a distal cell fate change and enhanced tumor formation within Scgb1A1+ expressing cells (Pacheco-Pinedo et al., 2011).”

1. The authors regret that they were unable to experimentally inhibit wnt signaling, arguing that because in their model, because G9a functions downstream of where most wnt inhibitors act, e.g. on ligand interactions and the B-catenin complex, and because they argue that only suprabasal wnt activity leads to the observed phenotype. If this is interpretation is correct, it would actually present an elegant method to test their model, as wnt inhibition (very easy to technically administer and can be carefully titred to the correct level) should therefore not be able to rescue the effects of G9a inhibition in their system. We believe that this experiment should be completed and reported as it is simple and would greatly strengthen their claims.

We thank the reviewer for raising this point again and we will try to further clarify our previous response. The mechanism we are exploring postulates that EHMT2 activity regulates basal steady-state chromatin-bound β catenin-mediated transcription by modulating β catenin-RUVBL2 interactions. Inhibition of upstream signaling with a Wnt inhibitor as the reviewer suggests would therefore not address the EHMT2 effect on the chromatin-bound β-catenin signaling pool. The reviewer is suggesting that we perform an experiment where we inhibit Wnt signaling by a means that would not include the pool of EHMT2/b-catenin of interest and demonstrate that has no impact our phenotype. Due to the requirements of a basal Wnt signaling we do not believe that this experiment is feasible and would not contribute further to our understanding of β catenin-mediated transcriptional activity upon EHMT2 inhibition.

2. The authors highlight some tumorsphere data showing that both proliferation and apoptosis are reduced. The purpose of asking for these data was to confirm if the effects of G9a inhibition were truly on stemness which leads to more organoids forming, or because G9a inhibition might reduce proliferation or increase apoptosis of the tumor cells. The results highlighted from figure S 3-1 show that whilst apoptosis is down, proliferation is also strongly reduced. Therefore are the results from G9ai on tumor growth, formation etc due to changes in proliferation rather than AT2 cell differentiation? Or do the authors consider that these are indistinguishable?

The reduced proliferation and cell death in tumorspheres led us to explore a differentiation as a possible outcome upon G9a inhibition. In tumorspheres, we show that differentiation leads to loss of self-renewal by two different sets of experiments: (1) using G9a inhibitors or shRNA and (2) by stabilizing β catenin using a GSK3b inhibitor. In both sets of experiments, we detect evidence of cell differentiation prior to loss of self-renewal. Therefore, we conclude that effects resulting in sphere size changes are a consequence of cell differentiation.

3. Whilst the discussion of the different TPC populations described is appreciated and adds important context, noting that they have different abundances. The question of whether the other TPC marker, Sca-1 and the Sca-1+ fraction is affected in their experiments remains unaddressed when a simple QPCR for Sca-1 in any of their tumor systems could help to address this.

To adequately address the abundance of the alternative TPC population of interest to the reviewer, would require running additional experiments using a newly optimized flow cytometry panel that included those surface markers. This alternative TPC population was not the focus of the work, moreover the outcome of those experiments would not impact the results or conclusions of the manuscript.

However, our RNAseq dataset will be publicly available for query for any and all genes of interest.

For convenience to R2, we include in Author response image 5 the results of Sca-1 expression from EHMT2 inhibitor treated ex vivo tumor spheres. EHMT2 inhibition did not alter Sca-1 expression.

Author response image 5
Download asset Open asset

4. We appreciate the reviewer's explanation, highlighting the difference between the in vivo and tumorsphere results. To incorporate this explanation into the text, the authors should mention that the broadly increased SPC expression observed in the cultures is likely from progenitors and differentiated tumor cells whilst in vivo only TPCs show an increase in genes associated with AT2 differentiation.

We thank the reviewer for the comment and it is now incorporated into the main text.

5. The authors helpfully remind the reviewer that chromatin accessibility was measured in TPCs and not tumorspheres, where they demonstrate that H3K9me2 and H3K9me3 are, as expected, reduced by G9a inhibition. Although they find very few changes in chromatin accessibility in TPCs, the method they describe is appropriate and if carried out correctly should find differences if they are there. This explanation seems to suggest that chromatin modifications would not be changed in TPCs and differences are restricted to tumorspheres "a more diverse population including both TPCs and non-TPC progeny". The authors would seem to be suggesting, therefore, that unlike tumorspheres in culture, H3K9me2 and H3K9me3 are not reduced by G9a inhibition in TPCs. This would be a surprising result, and at the very least should be investigated by the authors.

Indeed, the reviewer captured the results correctly. Our data indicated that the EHMT2 inhibition-induced phenotype is driven by a mechanism independent of changes in chromatin accessibility. Indeed, H3K9me2 and H3K9me3 marks are reduced by EHMT2 inhibition in TPCs, but few changes in chromatin accessibility were observed in the TPC subset. Moreover, pathway enrichment analysis on changes mediated by EHMT2 inhibition did not explain the observed phenotype. These results led us to propose an alternate hypothesis that, as the reviewer points out, is very interesting and was rigorously investigated and presented in our work where we show that Ruvbl2 acts as a non-histone substrate, regulated by EHMT2 activity and affects β catenin-mediated AT2 gene transcription.

6. In the explanation provided the authors refine their explanation of the effects of G9a and WNT signaling on TPCs, to align their conclusions with the existing literature. As mentioned in earlier rebuttal points, they interpret their data to suggest that G9a restriction of Wnt signaling is supportive of TPC function, and if G9a is lost or inhibited then excess wnt signaling can lead to loss of TPC function. This represents an interesting finding, although the authors perhaps overinterpret their findings when they suggest that G9a regulates Wnt to a distinct threshold (author's emphasis).

With the authors data and the existing literature suggest that either too little, or now too much Wnt signaling can be deleterious to TPCs and AT2 cells. Whether there is a distinct threshold of wnt signaling for TPCs/AT2 cells to maintain proper function, and more importantly, what this threshold might be is interesting to speculate about, but goes beyond what the author's data show. It would take a considerable amount of experimentation to demonstrate that such a distinct threshold exists, more than could be reasonably expected in a revision of this manuscript. The authors should amend the description of their results and the discussion appropriately.

To clarify for R2, we do not state anywhere in our manuscript that EHMT2 regulates a distinct threshold of Wnt, but given the questions by the reviewer this concept was discussed heavily in the rebuttal. We only state, as the reviewer summarized above very well, that EHMT2 restricts WNT signaling and if EHMT2 is lost or inhibited then excess Wnt signaling can lead to loss of TPC function. Any speculation regarding WNT signaling thresholds occurs in the discussion and we have reviewed that section carefully to ensure that it is presented as such. Specifically, WNT signaling thresholds are only mentioned twice in the second to last paragraph of the discussion and only in a manner postulating their existence as a means to explain discrepancies between published work.

7. It is unfortunate that the reviewers were not able to confirm the RUVBL2 and B-Catenin interaction in tumorspheres or TPCs, and that mouse cell lines do not behave like TPCs in order for them to be used as a model. This raises the question then, if the A549 cells used for the mechanistic experiments contain "the relevant population to progeny relationship found in de novo kp tumors". Do A549 cells behave similarly to TPCs in organoid culture in response to G9a inhibition? This is important for the connection of the proposed mechanism to the phenotypes observed in the mouse. If the mechanistic relationship cannot be tested in the phenotypic model(s), then the phenotype should be confirmed in the mechanistic model, otherwise we cannot connect the mechanism to the phenotype even if they are 'operating in the same direction' (positively corelating). As mentioned earlier in this discussion, Zhang et al. found that Wnt signaling was inhibited by G9a inhibition and EHMT2 shRNA in cultures and xenografts of human cells including A549 cells. Whilst this study should not be considered the last word on this topic, it highlights the importance for the authors to test to see if their phenotype can be reproduced in this system used to elucidate their proposed mechanism.

Whilst the authors have convincingly shown that Wnt signaling is modulated upwards in response to G9a inhibition in tumors and AT2 cells, a stronger connection to the proposed mechanisms is required for publication.

We refer the reviewer to several mechanistic assays conducted in TPCs: including proximity ligation assay between RUVBL2 and β catenin, confirming interaction and ChIP-qPCR of RUVBL2 on TCF4 motifs in AT2 genes, proving that EHMT2 directly controls Tcf4-containing AT2 gene expression through RUVBL2-mediated repression of β-catenin transcription. Furthermore, we also want to remind the reviewer of the observed changes in RUVBL2 levels upon EHMT2 inhibition in tumorspheres, indicating a direct relationship between EHMT2 and RUVBL2 (Figure 5c-g). Therefore, we disagree with the general statement that “the mechanistic relationship cannot be confirmed in the phenotypic model”. We used A549 as a complementary approach for experiments that are technically not feasible in TPC. Please also refer to our response for the first point raised by the reviewer (R2 Specific Point 1).

8. It is helpful that the authors provide pharmacokinetic data showing consistent plasma concentration of their inhibitor following. We take it that the G9a inhibitor is highly selective over other histone methyltransferases which has been amply demonstrated in the original literature. Whilst direct evidence of G9a inhibition in vivo would (through histone methylation) would be greatly convincing, we appreciate that confirmation of the main phenotypes of increases of wnt signaling and AT2 gene expression through genetic deletion makes it likely that G9a inhibitor is functioning as expected in vivo. In the absence of confirming whether G9a inhibition reduces the stem cell capacity by serial passaging of inhibited AT2 cell cultures, the text changes made by the authors is appropriate.

We are thankful for the input and approval of our edits.

9. The author's test if there are changes in the proportion of AT1 markers in their tumorsphere assays with G9a inhibition, as they see with their alveolar spheres. Here they do not see a significant change AT1 cell markers, highlight differences between how tumor cells and AT2 cells respond to G9a inhibition, which should be commented on in the discussion.

We appreciate the reviewer’s comment and have now commented about these differences in the text.

“In contrast to the TPC subset, we observed significant changes in the proportion of AT1 cells, consistent with the reduced plasticity observed in AT2 cells.”

The authors have already addressed wording changes to the descriptions of their alveolar sphere cultures to reflect that AT2 to AT1 is the normal differentiation process and not an example of trans-differentiation.

We are thankful for the input and approval of our edits.

Whilst the authors' results are consistent with the explanation that a change in AT2 to AT1 differentiation, the small size of the alveospheres does suggest that changes in proliferation could also contribute to these differences, and really should be explored by the authors. The immunohistochemistry would also be extremely informative as well, what to the AT1/AT2 cells look like?

We are relieved to read that the reviewer agrees that our results are consistent with the conclusion of the differentiation phenotype. With respect to the proliferation effects, we openly include data on sphere size and the effects on proliferation and apoptosis, in addition to the differentiation phenotypes (Figure 3, including the IF images, plus quantitation and Figure 6 with supplements with quantitation of the IF images). To further support our reasoning, we include the following statement with references regarding the relationship between proliferation and differentiation:

“Reduced proliferation and cell death were previously associated with cell differentiation (Domen and Weissman, 1999; Ruijtenberg and van den Heuvel, 2016).”

The authors point out that it is not possible to perform secondary transplantations with their inducible due to the re-expression of G9a in vivo. As the effects on self-renewal cannot be experimentally verified, the authors should modify their claims in the manuscript.

To clarify, the in vivo tumor formation studies using shEHMT2 are performed following secondary transplantation. To alleviate any concerns of interpretation that reviewer raises, we modified the section heading as below:

Original:

EHMT2 is required for in vivo tumor self-renewal

New:

EHMT2 is required for in vivo tumor growth

Reviewer #3 (Recommendations for the authors):

1. The authors have addressed some of my previous comments. However, I find their responses to comments 12 and 13 vague and inadequate

We are disheartened to read this response from R3.

Comment 12 seemed to us to reflect thoughts of the reviewer regarding aspects of the field and did not contain a clear question. We agreed with the reviewer’s summary of the field and the speculation they proposed.

Comment 13 inquired about AT2 cells in our tumorsphere cultures, which we viewed as a misunderstanding of the technical aspects of the assay, which we clarified for the reviewer.

2. Provide one or more citations for the claim that, "…multiple lines of evidence support the expression of LGR5 in the epithelial cell compartment." or, if that is not possible, then omit reference to LGR5 expression.

The reviewer misunderstood our response to his query regarding LGR5 expression that we detect in our studies. The “multiple lines of evidence” is in reference to our own data contained in this paper. The original response is copied below for reference. Given the paucity of data on this topic and the fact that LGR5 is a well-established WNT target gene, we are compelled to retain this information.

Response from first round of reviews:

The reviewer is referring to Figure 6D, where expression of Axin2, Lgr4 and Lgr5 is used to confirm Wnt pathway activation in primary AT2-derived murine spheres following G9a inhibition. We presume that the reviewer is basing his statements of lack of Lgr5 on the data published by Lee et al. Cell 2017, where Lgr5 was not detected in the epithelial population by single cell RNA-seq. Because of dropouts, genes undetected in scRNA-seq assays cannot be considered as a confirmation of lack of expression, especially for lowly expressed genes. Indeed, expression of Lgr5 in our own bulk RNA-seq dataset was very low (2-17 reads out of 30-40M reads per sample), which likely means that Lgr5 cannot be detected in standard scRNA-seq assays. Although Lgr5 is a low-expressed gene, it is readily observable by qPCR, most likely due to the amplification steps included in this procedure. While Axin2 and Lgr4 increases suffice to demonstrate wnt pathway activation, we would like to retain Lgr5 as multiple lines of evidence support its expression in the epithelial cell compartment.

https://doi.org/10.7554/eLife.57648.sa2

Article and author information

Author details

  1. Ariel Pribluda

    Department of Translational Oncology, Genentech, Inc, South San Francisco, United States
    Present address
    Surrozen, South San Francisco, United States
    Contribution
    Conceptualization, Formal analysis, Investigation, Methodology, Writing – original draft, Writing – review and editing
    For correspondence
    ariel@surrozen.com
    Competing interests
    was an employee of Genentech when the work was performed and may hold stock
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-2817-2827

  2. Anneleen Daemen

    Department of Bioinformatics & Computational Biology, Genentech, Inc, South San Francisco, United States
    Present address
    ORIC Pharmaceuticals, South San Francisco, United States
    Contribution
    Formal analysis, Investigation, Methodology
    Competing interests
    was an employee of Genentech when the work was performed and may hold stock
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-6287-7105

  3. Anthony Nelson Lima

    Department of Translational Oncology, Genentech, Inc, South San Francisco, United States
    Contribution
    Investigation
    Competing interests
    was an employee of Genentech when the work was performed and may hold stock

  4. Xi Wang

    Department of Translational Oncology, Genentech, Inc, South San Francisco, United States
    Contribution
    Investigation, Methodology
    Competing interests
    holds shares in the company

  5. Marc Hafner

    Department of Bioinformatics & Computational Biology, Genentech, Inc, South San Francisco, United States
    Contribution
    Formal analysis, Investigation, Methodology
    Competing interests
    holds shares in the company
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-1337-7598

  6. Chungkee Poon

    Department of Immunology, Genentech, Inc, South San Francisco, United States
    Contribution
    Methodology
    Competing interests
    holds shares in the company

  7. Zora Modrusan

    Department of Molecular Biology, Genentech, Inc, South San Francisco, United States
    Contribution
    Resources, Methodology
    Competing interests
    holds shares in the company

  8. Anand Kumar Katakam

    Department of Pathology, Genentech, Inc, South San Francisco, United States
    Contribution
    Formal analysis, Methodology
    Competing interests
    holds shares in the company

  9. Oded Foreman

    Department of Pathology, Genentech, Inc, South San Francisco, United States
    Contribution
    Investigation
    Competing interests
    holds shares in the company

  10. Jefferey Eastham

    Department of Pathology, Genentech, Inc, South San Francisco, United States
    Contribution
    Formal analysis
    Competing interests
    holds shares in the company

  11. Jefferey Hung

    Department of Pathology, Genentech, Inc, South San Francisco, United States
    Contribution
    Formal analysis
    Competing interests
    holds shares in the company

  12. Benjamin Haley

    Department of Molecular Biology, Genentech, Inc, South San Francisco, United States
    Contribution
    Resources, Methodology
    Competing interests
    holds shares in the company

  13. Julia T Garcia

    Department of Genetics, Stanford University, Stanford, United States
    Contribution
    Investigation
    Competing interests
    holds shares in the company

  14. Erica L Jackson

    Department of Discovery Oncology, Genentech, Inc, South San Francisco, United States
    Present address
    Scorpion Therapeutics, South San Francisco, United States
    Contribution
    Supervision, Writing – review and editing
    Competing interests
    was an employee of Genentech when the work was performed and may hold stock
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-7100-8021

  15. Melissa R Junttila

    Department of Translational Oncology, Genentech, Inc, South San Francisco, United States
    Present address
    ORIC Pharmaceuticals, South San Francisco, United States
    Contribution
    Conceptualization, Supervision, Writing – original draft, Writing – review and editing
    For correspondence
    melissa.junttila@oricpharma.com
    Competing interests
    was an employee of Genentech when the work was performed and may hold stock
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-3538-1192

Funding

No external funding was received for this work.

Acknowledgements

We appreciate the support of all the Junttila lab members. We would like to thank Elaine Storm and the De Sauvage lab for providing critical reagents. We are grateful to the sequencing lab for their superb work. We also received technical support from the in-house genotyping and murine reproductive technology core groups, especially Tiffany Yuan. We would like to acknowledge the laboratory animal research core for facilitating the support needed to enable in vivo experimentation, as well as the center for advanced light microscopy for providing help with immunofluorescence quantitation and all FACS lab members, especially Jonathan Paw and George Tweet who were instrumental with devising critical sorting strategies.

Ethics

All animal studies were approved by the Institutional Animal Care and Use Committee at Genentech and adhere to the Guidelines for the Care and Use of Laboratory Animals (protocols 17-1217, 17-0107 and 18-1833 series).

Senior Editor

  1. Edward E Morrisey, University of Pennsylvania, United States

Reviewing Editor

  1. Martin McMahon, University of Utah Medical School, United States

Reviewer

  1. Martin McMahon, University of Utah Medical School, United States

Version history

  1. Received: April 7, 2020
  2. Preprint posted: April 20, 2020 (view preprint)
  3. Accepted: August 16, 2022
  4. Accepted Manuscript published: August 19, 2022 (version 1)
  5. Version of Record published: September 2, 2022 (version 2)

Copyright

© 2022, Pribluda et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 863
    Page views
  • 221
    Downloads
  • 1
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Open citations (links to open the citations from this article in various online reference manager services)

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)

  1. Ariel Pribluda
  2. Anneleen Daemen
  3. Anthony Nelson Lima
  4. Xi Wang
  5. Marc Hafner
  6. Chungkee Poon
  7. Zora Modrusan
  8. Anand Kumar Katakam
  9. Oded Foreman
  10. Jefferey Eastham
  11. Jefferey Hung
  12. Benjamin Haley
  13. Julia T Garcia
  14. Erica L Jackson
  15. Melissa R Junttila
(2022)
EHMT2 methyltransferase governs cell identity in the lung and is required for KRAS G12D tumor development and propagation
eLife 11:e57648.
https://doi.org/10.7554/eLife.57648

Categories and tags

Research organisms

Further reading

    1. Cancer Biology
    2. Computational and Systems Biology

    Pancreatic cancer symptom trajectories from Danish registry data and free text in electronic health records

    Jessica Xin Hjaltelin, Sif Ingibergsdóttir Novitski ... Søren Brunak
    Research Article

    Pancreatic cancer is one of the deadliest cancer types with poor treatment options. Better detection of early symptoms and relevant disease correlations could improve pancreatic cancer prognosis. In this retrospective study, we used symptom and disease codes (ICD-10) from the Danish National Patient Registry (NPR) encompassing 6.9 million patients from 1994 to 2018,, of whom 23,592 were diagnosed with pancreatic cancer. The Danish cancer registry included 18,523 of these patients. To complement and compare the registry diagnosis codes with deeper clinical data, we used a text mining approach to extract symptoms from free text clinical notes in electronic health records (3078 pancreatic cancer patients and 30,780 controls). We used both data sources to generate and compare symptom disease trajectories to uncover temporal patterns of symptoms prior to pancreatic cancer diagnosis for the same patients. We show that the text mining of the clinical notes was able to complement the registry-based symptoms by capturing more symptoms prior to pancreatic cancer diagnosis. For example, ‘Blood pressure reading without diagnosis’, ‘Abnormalities of heartbeat’, and ‘Intestinal obstruction’ were not found for the registry-based analysis. Chaining symptoms together in trajectories identified two groups of patients with lower median survival (<90 days) following the trajectories ‘Cough→Jaundice→Intestinal obstruction’ and ‘Pain→Jaundice→Abnormal results of function studies’. These results provide a comprehensive comparison of the two types of pancreatic cancer symptom trajectories, which in combination can leverage the full potential of the health data and ultimately provide a fuller picture for detection of early risk factors for pancreatic cancer.

    1. Cancer Biology

    Disseminating cells in human oral tumours possess an EMT cancer stem cell marker profile that is predictive of metastasis in image-based machine learning

    Gehad Youssef, Luke Gammon ... Adrian Biddle
    Research Article

    Cancer stem cells (CSCs) undergo epithelial-mesenchymal transition (EMT) to drive metastatic dissemination in experimental cancer models. However, tumour cells undergoing EMT have not been observed disseminating into the tissue surrounding human tumour specimens, leaving the relevance to human cancer uncertain. We have previously identified both EpCAM and CD24 as CSC markers that, alongside the mesenchymal marker Vimentin, identify EMT CSCs in human oral cancer cell lines. This afforded the opportunity to investigate whether the combination of these three markers can identify disseminating EMT CSCs in actual human tumours. Examining disseminating tumour cells in over 12,000 imaging fields from 74 human oral tumours, we see a significant enrichment of EpCAM, CD24 and Vimentin co-stained cells disseminating beyond the tumour body in metastatic specimens. Through training an artificial neural network, these predict metastasis with high accuracy (cross-validated accuracy of 87-89%). In this study, we have observed single disseminating EMT CSCs in human oral cancer specimens, and these are highly predictive of metastatic disease.

    1. Cancer Biology
    2. Medicine

    Tissue-derived exosome proteomics identifies promising diagnostic biomarkers for esophageal cancer

    Dingyu Rao, Hua Lu ... Defa Huang
    Research Article

    Esophageal cancer (EC) is a fatal digestive disease with a poor prognosis and frequent lymphatic metastases. Nevertheless, reliable biomarkers for EC diagnosis are currently unavailable. Accordingly, we have performed a comparative proteomics analysis on cancer and paracancer tissue-derived exosomes from eight pairs of EC patients using label-free quantification proteomics profiling and have analyzed the differentially expressed proteins through bioinformatics. Furthermore, nano-flow cytometry (NanoFCM) was used to validate the candidate proteins from plasma-derived exosomes in 122 EC patients. Of the 803 differentially expressed proteins discovered in cancer and paracancer tissue-derived exosomes, 686 were up-regulated and 117 were down-regulated. Intercellular adhesion molecule-1 (CD54) was identified as an up-regulated candidate for further investigation, and its high expression in cancer tissues of EC patients was validated using immunohistochemistry, real-time quantitative PCR (RT-qPCR), and western blot analyses. In addition, plasma-derived exosome NanoFCM data from 122 EC patients concurred with our proteomic analysis. The receiver operating characteristic (ROC) analysis demonstrated that the AUC, sensitivity, and specificity values for CD54 were 0.702, 66.13%, and 71.31%, respectively, for EC diagnosis. Small interference (si)RNA was employed to silence the CD54 gene in EC cells. A series of assays, including cell counting kit-8, adhesion, wound healing, and Matrigel invasion, were performed to investigate EC viability, adhesive, migratory, and invasive abilities, respectively. The results showed that CD54 promoted EC proliferation, migration, and invasion. Collectively, tissue-derived exosomal proteomics strongly demonstrates that CD54 is a promising biomarker for EC diagnosis and a key molecule for EC development.