Lung adenocarcinoma (ADC) is the most frequent subtype of lung cancer. Most lung ADCs develop through stepwise progression from atypical adenomatous hyperplasia (AAH) to bronchio-alveolar carcinoma (BAC), and ultimately to multiple types of invasive ADCs (Subramanian and Govindan, 2008; Wistuba and Gazdar, 2006). Human AAH and BAC are considered to be equivalent to mouse lung adenoma (AD). Approximately 25% of human lung ADC cases harbor activating mutations in the K-RAS gene. Although K-RAS signaling has been intensely studied, no drugs have yet been approved to treat K-RAS–mutant cancers; this is primarily because inhibition of K-RAS, which is important for normal cellular function, would be extremely toxic to patients (Drosten et al., 2018). Therefore, identification of effective tumor suppressors of K-RAS–dependent lung cancer would provide new strategies for cancer treatment.

RUNX3 is inactivated by epigenetic alterations in most cases of human lung ADs (AAH and BAC) (Lee et al., 2010; 2013) and K-RAS–activated lung ADCs (Lee et al., 2013). Runx3 ^+/− mice spontaneously develop a variety of tumors in old age, most frequently lung ADs (Lee et al., 2010), and disruption of Runx3 in mouse lung leads to AD formation (Lee et al., 2013). The cellular decision regarding whether to undergo proliferation or death is made at the restriction (R)-point, which is disrupted in nearly all tumors (Weinberg, 2014). RUNX3 plays a key role at the R-point which governs the p14^ARF–p53 pathway (Chi et al., 2017; Lee and Bae, 2020; Lee et al., 2019a; 2019b). Early after mitogenic stimulation, RUNX3 transactivates R-point–associated genes, which include p14^ARF (p19^ARF in mouse, hereafter ARF) by recruiting activator complexes to its target loci. When the RAS-MEK pathway is downregulated, RUNX3 suppresses the target genes by recruiting repressor complexes. The cell then passes through the R-point to S phase. If the RAS-MEK pathway is not downregulated, expression of the R-point–associated genes are maintained and the cell cannot pass through the R-point. Therefore, the RUNX3 → R-point → ARF → p53 pathway constitutes a surveillance mechanism against oncogenic RAS (Lee et al., 2019a). Indeed, inactivation of p53 or Runx3 similarly accelerates malignant progression of oncogenic K-Ras–dependent mouse lung tumors (DuPage et al., 2009; Lee et al., 2013).

However, in mouse lung tumor models, oncogenic K-Ras mutation alone can induce ADCs without p53 mutation, and loss of p53 does not have a significant impact on early K-Ras–induced lung tumorigenesis (Feldser et al., 2010; Junttila et al., 2010; Muzumdar et al., 2016). Therefore, it remains unclear how K-Ras–activated cells evade oncogene surveillance mechanisms. In this study, we analyzed genetically engineered and carcinogen-induced mouse models, and obtained evidence demonstrating that K-Ras–activated cells are selected only if the activating K-Ras mutation arises in AD cells in which Runx3 is inactivated.

Experimental model and subject details

Runx3^flox (Jax 008773), p53^flox (Jax 008462), K-Ras^LSL-G12D (Jax 008179), Rosa26R-Tomato (Jax 007914), Cre^tm/ERT1 (Jax 004682), and Cre^tm/ERT2 (Jax 008463) mice were obtained from Jackson Laboratory (USA). Unless stated otherwise, all mice analyzed had mixed genetic backgrounds and were age-matched (6-8 weeks old). Sample size was determined based on our experience and previous experiments. No data were excluded from the analysis. Animal experiments described were performed with at least three independent replicates, and representative results are shown in the figures. All animal studies were randomized into control or treated groups; in general, all animals housed in the same cage were in the same treatment group. For analysis of tumor samples, identities were blinded from histopathological assessment. All animals were housed in specific pathogen-free facilities and monitored daily. Animal studies were approved by the Institutional Animal Care Committee of Chungbuk National University (CBNUA-1208-18-02).

Method details

Adenovirus carrying Cre recombinase (Ad-Cre) was purchased from Vector Biolabs (USA). Each mouse was treated with 2.5 × 10⁷ Ad-cre viral genome copies diluted in warm 50 µl sterile MEM. After treatment, mice were placed on a warm pad until they woke up .

To determine the order of Runx3 inactivation and K-Ras activation during urethane-induced development of lung ADC, 6-week-old wild-type (WT) mice were injected intraperitoneally with urethane (ethylcarbamate; Sigma, USA) (500 mg/kg body weight).

H&E staining was conducted according to a standard protocol. Briefly, slides were rehydrated by ethanol, xylene, and water to remove the paraffin. The nuclei were stained with hematoxylin (#S3309; DAKO, USA) for 3 min, and the cytoplasm was stained with eosin (HT110280; Sigma) for 30 s. Slides were mounted with Permount (SP15-500; Thermo Fisher Scientific, USA) after the dehydration and clearing steps.

For histological analysis, lungs were inflated with 4% paraformaldehyde or formalin (3.7% formaldehyde) and fixed for 36 h. Fixed paraffin sections were rehydrated, subjected to antigen retrieval, and blocked in Tris-buffered saline (0.1% Triton X-100 containing 1% bovine serum albumin) or protein-free blocking solution (DAKO), and sequentially incubated with specific primary antibodies and biotinylated (DAKO) or Alexa Fluor-conjugated secondary antibodies (Invitrogen, USA). Images were produced with a conventional microscope mounted with a DP71 digital camera (Olympus, Japan), an LSM 710 T-PMT confocal microscope (Carl Zeiss, Germany), and an AXIO Zoom.V16 and ApoTome.2 (Carl Zeiss). Images were processed with equivalent parameters using the ZEN Light Edition software (Carl Zeiss).

Standard exome-capture libraries were generated from 1 µg input DNA using the Agilent SureSelect Target Enrichment protocol for Illumina paired-end sequencing library (ver. B.3, June 2015). Probe sets were SureSelect Human All-Exon V6 or SureSelect Mouse All-Exon (Agilent, USA). DNA was quantitated using PicoGreen, and DNA quality was assessed by agarose gel electrophoresis. One microgram of DNA from each cell line was diluted in EB buffer and sheared to a target peak size of 150 to 200 bp using the Covaris LE220 focused ultrasonicator (Covaris, USA). The 8-microTUBE Strips were loaded into the tube holder of the ultrasonicator, and the DNA was sheared using the following settings: mode, frequency sweeping; duty cycle, 10%; intensity, 5; cycles per burst, 200; duration, 60 s × 6 cycles; temperature, 4°C to 7°C. The fragmented DNA was repaired, an ‘A’ was ligated to the 3′ end, and Agilent adapters were ligated to the fragments. Once ligation was assessed, the adapter ligated product was amplified by polymerase chain reaction (PCR). The final purified product was quantified using TapeStation DNA ScreenTape D1000 (Agilent). For exome capture, 250 ng of DNA library was mixed with hybridization buffers, blocking mixes, RNase block, and 5 µl of SureSelect All-Exon capture library according to the standard Agilent SureSelect Target Enrichment protocol. Hybridization to the capture baits was conducted for 24 h at 65°C in a PCR machine, with the thermal cycler lid heated to 105°C. The captured DNA was washed and amplified. The final purified product was quantified by qPCR according to the qPCR Quantification Protocol Guide (KAPA Library Quantification kits for Illumina Sequencing platforms) and quantified again using the TapeStation DNA ScreenTape D1000 (Agilent). Finally, the DNA was sequenced on the HiSeq 2500 platform (Illumina, USA).

Runx3 inactivation, but not p53 inactivation, enables immediate proliferation of K-Ras–activated lung epithelial cells

To determine whether the oncogenic K-RAS → RUNX3 → ARF → p53 pathway suppresses proliferation of K-Ras-activated lung epithelial cells in vivo, we crossed Rosa26R-Tomato (Tomato*) mice with strains harboring Runx3^flox/flox, p53^flox/flox, and K-Ras^{LoxP-STOP-LoxP-G12D/+} (K-Ras*), yielding K-Ras*;p53^flox/flox;Tomato* (KPT) and K-Ras*;Runx3^flox/flox;Tomato* (KRT) mice. The mice were targeted by Cre-expressing adenovirus infection (Ad-Cre, 2.5 × 10⁷ pfu/mouse; DuPage et al., 2009), and the Tomato-positive cells were traced 10 days after infection. The results revealed that most of the p53-inactivated and K-Ras–activated cells remained as single cells, i.e., they had not divided (Fig. 1A). By contrast, most of the Runx3-inactivated and K-Ras–activated cells had formed clusters (Fig. 1B). These results indicate that Runx3 inactivation, but not p53 inactivation, enables immediate proliferation of K-Ras–activated lung epithelial cells.

Runx3 inactivation allows initiation of K-Ras–dependent lung cancer

Currently, all available genetically engineered K-Ras–dependent mouse lung cancer models require activation of K-Ras in a large number of cells (10⁶-10⁸ cells/mouse) and a long latency period (Guerra et al., 2003; Tuveson et al., 2004). Under physiological conditions, however, activation of K-Ras does not occur simultaneously in a large number of cells. Therefore, we asked whether activation of K-Ras, alone or in combination with inactivation of p53 or Runx3, in a very small number of cells could induce lung cancer. To target Runx3, p53, and/or K-Ras in a very small number of cells, we crossed Cre^tm/ERT1 mice with strains harboring Runx3^flox/flox, p53^flox/flox, and K-Ras* and obtained various genotype combinations. In these mice, Cre recombinase can be activated with tamoxifen to activate K-Ras and inactivate p53 or Runx3 (Fig. 1C). However, due to weak Cre leakage in this model, in the absence of tamoxifen, genes are targeted in very few cells (Kemp et al., 2004), providing a convenient means of targeting a limited number of cells. Therefore, we examined tumor formation and survival in these mice in the absence of tamoxifen.

All K-Ras*;Runx3^flox/flox (KR), K-Ras*;Cre^tm/ERT1 (K-Cre^ERT1), K-Ras*;p53^flox/flox;Cre^tm/ERT1 (KP-Cre^ERT1), and Runx3^flox/flox;Cre^tm/ERT1 (R-Cre^ERT1) mice survived for more than 1 year (Fig. 1C). Pathological analysis revealed that the mice were completely tumor-free (Fig. 1D), whereas R-Cre^ERT1 mice developed a few small ADs (Fig. 1D). By contrast, all mice with the K-Ras*;Runx3^flox/flox;Cre^tm/ERT1 (KR-Cre^ERT1) genotype developed lung ADCs rapidly, as early as 2 weeks after birth, and died within 12 weeks of birth in the absence of tamoxifen (median survival, 48 days) (Figs. 1C and 1D). These results indicated that while activation of K-Ras alone or in combination with inactivation of p53 in a small number of cells does not result in lung cancer formation, concurrent K-Ras activation and Runx3 inactivation does.

To determine whether the same thing would happen when K-Ras activation and Runx3 inactivation occurred in an even smaller number of cells, we used Cre^tm/ERT2, which is much more tightly regulated by tamoxifen than Cre^tm/ERT1. These experiments were performed in Rosa26R-Tomato (Tomato*) mice to allow visualization of the targeted cells. Cre^tm/ERT2 mice and Tomato* mice were crossed with Runx3^flox/flox, p53^flox/flox, and K-Ras* mice to obtain Tomato*;Cre^tm/ERT2 (T-Cre^ERT2), K-Ras*;Tomato*;Cre^tm/ERT2 (KT-Cre^ERT2), K-Ras*;p53^flox/flox;Tomato*; Cre^tm/ERT2 (KPT-Cre^ERT2), and K-Ras*; Runx3^flox/flox;Tomato*;Cre^tm/ERT2 (KRT-Cre^ERT2) mice. The mice were maintained in the absence of tamoxifen. Six months after birth, a few Tomato-positive cells (average, 20 cells per section) were detected in the lungs of T-Cre^ERT2 mice (Fig. 2A), confirming that Cre^tm/ERT2, like Cre^tm/ERT1, is leaky. However, Tomato-positive cells in the lungs of KT-Cre^ERT2 mice were extremely rare (average, 0.5 cells per section) (Fig. 2A). Simple calculation suggests that only about 2.5% of K-Ras–activated cells survived, and that these cells were in a quiescent state. These results suggest that the majority of K-Ras–alone-activated cells are eliminated through oncogene surveillance mechanisms.

Similar to what we observed with Cretm/ERT1 mice, KPT-Cre^ERT2 mice (n = 5) were completely tumor-free 6 months after birth (Fig. 2B). By contrast, all KRT-Cre^ERT2 mice (n = 5) developed lung ADCs (Fig. 2B). In the lungs of KPT-Cre^ERT2 mice, we detected clusters of Tomato-positive cells, but these cells formed a normal alveolar architecture and did not exhibit any dysplastic features (Fig. 2C). These results suggest that p53 inactivation enables K-Ras–activated cells to survive and weakly proliferate, but not to develop into tumors. By contrast, Tomato-positive cells in KRT-Cre^ERT2 mouse lungs formed tumors (Fig. 2D), indicating that Runx3 inactivation enables K-Ras–activated cells not only to survive and proliferate, but also to develop into tumors. These results suggest that p53 inactivation abrogates the oncogene surveillance mechanism only partly, whereas Runx3 inactivation does so completely.

Overall, these findings indicate that K-Ras activation in a limited number of cells, either alone or in combination with p53 inactivation, does not induce lung cancer, whereas the combination of K-Ras activation and Runx3 inactivation immediately induces lung cancer. These results suggest that Runx3 functions as a gatekeeper against K-Ras–dependent lung tumorigenesis.

The gatekeeper activity of Runx3 against K-Ras activation also operates in other tissues

Leaky activation of Cre^tm/ERT1 occurs in a wide range of tissues (Kemp et al., 2004). Indeed, tumors in KR-Cre^ERT1 mice were not limited to the lungs. Although lung ADCs developed in all KR-Cre^ERT1 mice, other types of tumors also developed. In particular, six of twenty KR-Cre^ERT1 mice developed skin cancers in the anus (Fig. 3A), six had a grossly enlarged spleen with massive lymphoid hyperplasia (Fig. 3B), and seven had an enlarged thymus with thymic lymphoma (Fig. 3B). None of these tumors were detected in KP-Cre^ERT1 mice.

Notably, thymic lymphoma cells infiltrated into the lung along the broncho-vascular bundle in some KRT-Cre^ERT2 mice (Fig. 3C). The targeting of K-Ras* and Runx3^flox/flox by leaky activation of Cre^tm/ERT1 or Cre^tm/ERT2 in the tumors was confirmed by genotyping (Fig. 3D). These results imply that the proposed gatekeeper activity of Runx3 against K-Ras activation is not limited to the lung, but also operates in other tissues.

Lung ADs develop via inactivation of Runx3 and progress into ADCs via K-Ras activation

Mouse lung tumors induced by urethane, a tobacco carcinogen, recapitulate the natural history of smoking-associated, K-RAS–driven human lung ADCs (Westcott et al., 2015). The mice develop slow-growing ADs about 20 weeks after urethane treatment and fast-growing ADCs about 58 weeks later (Fig. 4A). We previously showed that Runx3 expression is downregulated in nearly all urethane-induced lung ADs and ADCs, mainly due to hypermethylation of the Runx3 CpG island (Lee et al., 2010). In this study, we detected K-Ras pathway activation (Erk phosphorylation) in almost all ADCs (19 out of 20), but not in any ADs (n = 20) in which Runx3 was inactivated (Fig. 4B). Consistent with this, whole-exon sequencing revealed that oncogenic K-Ras mutation was present in all ADCs but in no ADs (Fig. 4C, Supplementary Data 1). No mutations within exons of other known oncogenes and tumor suppressors involved in lung cancer were detected in ADs or ADCs (Supplementary Data 1). Notably, the proportion of oncogenic K-Ras–mutated alleles (representing the proportion of K-Ras–mutated cells within the tumor) increased with cancer growth (Fig. 4C). These results, together with our previous observations that Runx3 deletion in mouse lung results in development of AD (Lee et al., 2013), suggest that lung ADs develop via inactivation of Runx3, and that AD cells that acquire oncogenic K-Ras mutation selectively proliferate and form ADCs (Fig. 4D).

Although K-RAS mutation is the most frequently detected oncogenic mutation in human cancers, to date no drugs have been approved to treat K-RAS–mutant cancers. Recently, oncogenic K-RAS–specific inhibitors have been developed and are currently in clinical trials (Canon et al., 2019). However, more recent reports show that shortly after treatment with these inhibitors, some cancer cells bypass the effect of oncogenic K-RAS inhibition and resume proliferation (Xue et al., 2020). In addition, even when a cancer is effectively regressed by knockdown of oncogenic K-Ras in a mouse lung cancer model, the cancer recurs rapidly, not because the gene knockdown was unsuccessful but because other oncogenes were activated (Shao et al., 2014). These results suggest that existing strategies for inhibiting oncogenic K-RAS have a limited ability to achieve a durable response in cancer treatment. Given that normal mice do not suffer from tumors as frequently as tumor-regressed mice, the rapid recurrence phenomenon observed in cancer-regressed mice (Shao et al., 2014) implies that a defense mechanism is abrogated in K-Ras–activated lung tumors. However, loss of p53 does not have a significant impact on early K-Ras–induced lung tumorigenesis (Feldser et al., 2010). These observations imply the existence of an effective defense mechanism against oncogenic K-Ras signaling, and that this mechanism is disrupted by a hidden molecular event in K-Ras–dependent lung cancers. Therefore, to achieve a complete and durable response, it will be necessary to identify the hidden defense mechanism against K-Ras–dependent lung cancers.

To confirm the existence of this defense mechanism and identify the critical genes involved, we need to understand why tumor development in animal models requires activation of oncogenes in a large number of cells. For example, K-Ras–dependent mouse models develop lung cancer several months after K-Ras is targeted in a large number of cells (10⁶-10⁸ cells/mouse) (Guerra et al., 2003; Tuveson et al., 2004). Therefore, cancer development must be a very rare event within the large population of K-Ras–activated cells. To explain this, we can consider two possibilities. First, the cells that originate the tumors may be very rare, making the probability of targeting K-Ras in any given origin cell very low (single-step model, assuming the absence of a defense mechanism). Second, the cells of origin of tumors may be relatively abundant, but K-Ras activation alone cannot induce cancer unless the defense mechanism is abrogated (two-step model, assuming the existence of the defense mechanism). In this study, we observed the following: (1) A combination of Runx3 inactivation and K-Ras activation in a very limited number of cells immediately induced lung ADCs, whereas K-Ras activation alone or in combination with p53 inactivation did not. (2) In the carcinogen-induced mouse lung cancer model, Runx3 inactivation occurred earlier than K-Ras activation. Our results strongly support the second possibility described above, which assumes the existence of the defense mechanism: the majority of lung ADs are initiated by Runx3 inactivation, and AD cells, in which R-point-associated oncogene surveillance mechanism is abrogated, cannot defend against K-Ras activation, resulting in a transition from ADs to ADC.

The Arf–p53 pathway is the primary defense against oncogenic Ras in primary cells (Serrano et al., 1997). Why, then, is only limited tumor suppressor activity of p53 observed in mouse lung tumor models (Feldser et al., 2010)? In our interpretation, the results do not imply a limitation of p53 tumor suppressor activity, but instead suggest that p53-independent tumor suppressive mechanisms exist in the lung; hence, if lung tumorigenesis is suppressed by multiple pathways, loss of p53 can be compensated by other pathways. In this regard, it is worth emphasizing that the R-point, which is disrupted in nearly all tumors, governs multiple tumor suppression pathways (Weinberg, 2014). When a death decision is made at the R-point, the cell defends against tumorigenesis by regulating not only intracellular programs (cell cycle, apoptosis, and metabolism) but also extracellular programs (inflammatory and immune responses) (Lee et al., 2019a; Samarakkody et al., 2020; Seo and Taniuchi, 2020). Because Runx3 functions as a decision-maker of the R-point, inactivation of Runx3 abrogates the entire R-point–associated defense programs (Lee et al., 2019a). By contrast, p53 functions as an executor of the R-point decision, and inactivation of the gene abrogates only intracellular programs. The critical roles of Runx3 in R-point regulation may explain how Runx3 functions as a potent gatekeeper against oncogenic K-Ras.

Targeted therapies that inhibit activated oncogenes have yielded clinical responses but eventually lead to cancer recurrence with secondary oncogene activation in almost all malignancies (Janne et al., 2009; Podsypanina et al., 2008). Therefore, it would be of great therapeutic value to understand why cancers recur after the failure of therapies that target activated oncogenes. Our results explain why cancers recur: inhibition of an activated oncogene causes the cancer to regress, but the regressed cells remain cancer-prone because their oncogene surveillance mechanisms are inactivated.

S-C Bae is supported by a Creative Research Grant (2014R1A3A2030690) through the National Research Foundation (NRF) of Korea. Y-S Lee is supported by Basic Science Research Program grant 2017R1D1A3B03034076. J-W Lee is supported by Basic Science Research Program grant 2018R1C1B6001532. Y Ito is supported by the Singapore Ministry of Health’s National Medical Research Council under its Open Fund Large Collaborative Grant (LCG) Programme (MOH-OFLCG18MAY-0003), NRF Singapore, and the Singapore Ministry of Education under its Research Centres of Excellence initiative.

Y.S.L., J.Y.L., and D.M.K. generated mouse cancer models and analyzed the cancers. X.Z.C. analyzed inactivation, activation, and restoration of target genes in mouse cancer models. J.W.L. and S.H.S. analyzed K-Ras mutations in mouse cancers. Y.I. and S.C.B. interpreted the results. S.C.B. planned the experiments and wrote the manuscript. All authors contributed to the editing of the manuscript.

The authors have no potential conflicts of interest to disclose.