The repressor Capicua is a barrier to lung tumor development driven by Kras/Trp53 mutations.

The paper explained
Problem
Lung adenocarcinomas driven by KRAS oncogenes are among the most severe and lethal types of cancer. Yet, the mechanisms linking KRAS mutations and progressive allelic imbalances to tumor formation remain unclear.

Results
Our study shows that mutant Kras copy numbers gradually increase during tumor progression. Genetic disruption of the repressor Capicua (CIC) in Kras/Trp53 mutant mice suppresses these allelic imbalances and promotes the transformation of bronchiolar Club cells, leading to an increased tumor burden and inducing resistance to MAPK pathway inhibition. Restoring CIC repressor activity or silencing of its target genes Etv4 and Etv5 decreases resistance to MAPK pathway inhibition, similar to treatment with drugs (PFK15 and Tx-1123) that selectively affect the viability of Cic-deficient tumor cells.

Impact
We have identified the repressor CIC as a barrier to lung tumor development in Kras/Trp53 mutant mice. The absence or mutational inactivation of CIC in lung cancer patients may have adverse consequences such as increased resistance to MAPK pathway inhibitors, and reveals new vulnerabilities that could be exploited to overcome resistance.

Introduction
Lung cancer is one of the most frequently diagnosed cancers and the leading cause of cancer-related mortality worldwide. Mutations in KRAS contribute to ~25% of lung adenocarcinomas (LUADs), the most frequent lung cancer subtype (Ferrer et al, 2018; Sung et al, 2021). Although initially classified as undruggable, there are now two FDA-approved drugs targeting a specific mutant isoform (KRASG12C). This has been the starting point for the development of a wide variety of other allele-specific, pan-KRAS, or even pan-RAS inhibitors (Holderfield et al, 2024), although the frequent emergence of resistance, at least in the case of KRASG12C inhibitors, has not yet produced major clinical benefits (Drosten and Barbacid, 2022; Zhu et al, 2022; de Langen et al, 2023; Molina-Arcas and Downward, 2024).
Despite these advances, surprisingly little is known about the molecular mechanisms linking KRAS mutations to lung cancer. In mice, only a small fraction of cells expressing a resident Kras oncogene progress to form tumors, suggesting that additional mechanisms must contribute to tumor growth (Guerra et al, 2003; Mainardi et al, 2014). Whole-exome sequencing of lung tumors driven by endogenous Kras oncogenes has revealed a relatively modest number of co-occurring mutations, including a significant fraction of tumors lacking additional mutations in other known cancer genes (Chung et al, 2017; Junttila et al, 2010; McFadden et al, 2016; Westcott et al, 2015). In contrast, most of these tumors carried secondary amplifications of the Kras locus. Importantly, human KRAS-mutant LUADs also show frequent allelic imbalances associated with reduced patient survival (Chiosea et al, 2011; Yu et al, 2017). Moreover, it has been shown that Kras allelic imbalance promotes malignancy, at least in part, through metabolic rewiring and stimulation of tumor cell fitness (Burgess et al, 2017; Kerr et al, 2016). Consistent with the requirement for MAPK signaling in KRAS-driven lung cancer, these findings suggest that Kras allelic imbalance may further promote MAPK pathway activity and tumor growth (Blasco et al, 2011; Burgess et al, 2017; Drosten and Barbacid, 2020; Junttila et al, 2010). While gaining additional copies of mutant Kras alleles is particularly frequent in LUAD (Chung et al, 2017; McFadden et al, 2016; Westcott et al, 2015), concomitant loss of the wild-type allele (loss of heterozygosity) can also enhance cellular transformation and MAPK pathway activity (Ambrogio et al, 2018).
The transcriptional repressor Capicua (CIC) has been identified as a critical substrate of ERK kinases in development and disease (Jiménez et al, 2012; Kim et al, 2021; Lee, 2020; Simón-Carrasco et al, 2018; Wong and Yip, 2020). When ERK kinases are inactive, CIC proteins engage highly conserved octameric DNA binding sites and repress transcription of their target genes by recruiting the SIN3 deacetylation complex (Ajuria et al, 2011; Weissmann et al, 2018; Gupta et al, 2022). In contrast, ERK activation results in CIC phosphorylation and inactivation of its repressor activity, thereby leading to derepression of its targets (Bunda et al, 2019; Dissanayake et al, 2011; Okimoto et al, 2017; Park et al, 2022).
Accumulating evidence supports the view that CIC is a tumor suppressor in human cancer (Kim et al, 2021). CIC mutations have been detected at high frequencies in oligodendrogliomas, and several other tumor types, including stomach adenocarcinoma, melanoma or lung cancer, also show significant mutation rates (Kim et al, 2021; Simón-Carrasco et al, 2018). Moreover, increased proteasomal degradation of CIC has been observed in glioblastomas (Bunda et al, 2019), and homozygous deletions of CIC occur in a variety of cancers (Kim et al, 2021). Of note, CIC deletions as well as inactivating mutations were associated with metastasis of lung cancer cells (Okimoto et al, 2017).
In addition, CIC mutations have been linked to resistance to drugs targeting RTK/RAS/MAPK signaling (Simón-Carrasco et al, 2017; Kim et al, 2022; Wang et al, 2017; Liao et al, 2017). Indeed, inactivating mutations were identified as the principal mechanism of resistance in patients treated with MAPK pathway inhibitors (Hashiba et al, 2020; Da Vià et al, 2020). These observations suggest that CIC is a critical effector of the RTK/RAS/MAPK signaling pathway in cancer development and therapy resistance, although its contribution to KRAS-driven LUAD has remained unknown. Hence, in this study, we have interrogated the role of CIC in LUAD formation and drug resistance using genetically-engineered mouse (GEM) models as well as patient-derived organoids, and reveal that CIC inactivation enhances tumor initiation. We also show that, while promoting resistance to inhibitors of the MAPK pathway, absence of CIC creates exploitable vulnerabilities in these tumors.

Results

Genetic CIC inactivation facilitates lung tumor initiation
LUADs driven by Kras oncogenes exhibit stage-specific amplification of MAPK signaling (Chung et al, 2017; Cicchini et al, 2017; Chen et al, 2019; Feldser et al, 2010). Thus, we hypothesized that amplified MAPK signaling may be relevant for efficient suppression of growth-inhibitory activities that prevent uncontrolled tumor progression, such as those mediated by ERK-regulated transcriptional repressors. Based on this assumption, we embarked on exploring the role of the repressor Capicua (CIC), which is negatively controlled by active MAPK signaling (Simón-Carrasco et al, 2018). To this end, we bred Ciclox/lox mice (Simón-Carrasco et al, 2017) with the Kras+/LSLG12Vgeo;Trp53lox/lox (KP) strain known to develop aggressive lung adenocarcinomas upon Cre-mediated recombination (Fig. 1A) (Drosten et al, 2017). In these mice, CIC can be rendered inactive by eliminating sequences corresponding to exons 2–6, resulting in the expression of non-functional CIC-SΔ2-6 and CIC-LΔ2-6 protein isoforms that lack their HMG-box DNA binding domain required for gene repression (Simón-Carrasco et al, 2017). Intranasal infection of KP mice with Ad-Cre resulted in the death of all animals due to lung tumor development with a median survival of 48 weeks (Fig. 1B). Notably, infection of Kras+/LSLG12Vgeo;Trp53lox/lox;Ciclox/lox (KPCic) mice resulted in significantly reduced survival (median survival 39 weeks) (Fig. 1B). As illustrated in Fig. EV1A,B, KPCic mice displayed a significant increase in grade 3 adenocarcinomas when sacrificed at a humane endpoint (Jackson et al, 2005). However, we did not observe evidence for accelerated tumor progression or a shift towards more aggressive lesions in the absence of functional CIC, as demonstrated by HMGA2 immunostaining as a marker for lung tumor progression and invasion (Fig. EV1C–E) (Winslow et al, 2011).
To ascertain whether the reduced survival of KPCic mice stemmed from more efficient initiation of lung tumors, we infected KPCic as well as KP mice as controls with Ad-Cre and compared their tumor burden 20 weeks post-infection. At this time point, KPCic mice exhibited an approximately 2.5-fold increase in the number of lesions (Fig. 1C). Yet, tumors in KPCic mice were histologically indistinguishable from those in KP mice, both expressing TTF-1 (also known as NKX2-1) and the alveolar marker SPC, but lacking the marker of bronchiolar Club cells CC10 (Fig. 1D). As shown in Fig. EV1F, the overall increase in the number of lesions in KPCic mice resulted from an elevated number of grade 2 adenomas and grade 3 adenocarcinomas. The percentage of HMGA2-positive tumors was similar in both groups, again supporting the idea that inactivation of CIC had no impact on tumor progression (Fig. EV1G,H). Interestingly, tumors from KPCic mice displayed significantly lower levels of pERK+ areas, while the percentage of Ki67+ cells was similar in both groups (Fig. 1E,F).
To better understand the impact of Cic disruption on tumor growth in KP mice, we first determined the genes potentially controlled by CIC by ChIP sequencing (ChIP-seq) in KPCic cells vs. KP cells left untreated or treated with the selective MEK inhibitor trametinib for 24 h to efficiently block the MAPK pathway and maximize CIC DNA binding. While the inactive CICΔ2-6 protein in KPCic cells, as expected (Simón-Carrasco et al, 2017), did not bind to CIC target genes (29% of peaks were located in other promoter regions based on ChIPseeker annotations), CIC proteins in proliferating KP cells showed a significant association with several promoters (69% of all peaks) including known CIC targets (Fig. EV2 and Dataset EV1). However, upon trametinib treatment, CIC strongly relocated to a more limited set of promoters (44% of all peaks), including Etv4 and Etv5, two well-known CIC target genes (Kawamura-Saito et al, 2006), or well-known negative regulators of the MAPK pathway such as Spry2, Dusp4, Dusp5 and Dusp6 (Fig. EV2A,B and Dataset EV1). Yet, expression of most genes, including these CIC targets, was unaltered in tumors from KPCic mice, suggesting that once tumors have formed in the absence of CIC, they are largely indistinguishable from those in KP mice (Fig. 1G, Fig. EV2C and Dataset EV2). Together, these findings indicate that CIC inactivation facilitates lung tumor initiation but does not notably contribute to tumor progression in KP mice.

Genetic CIC inactivation abrogates Kras allelic imbalance
Similar to human tumors, GEM models of KRAS-driven LUAD frequently develop allelic imbalance via acquisition of additional copies of mutant Kras alleles (Chung et al, 2017; McFadden et al, 2016; Westcott et al, 2015). This suggests that such allelic expansions could be, at least to some extent, responsible for amplified MAPK signaling in those tumors. Indeed, we noted that allelic imbalance increased significantly over time in KP mice infected with Ad-Cre as quantified by fluorescence-in situ hybridization (FISH) (Fig. 2A,B). Moreover, MAPK signaling increased concomitantly over time as demonstrated by pERK+ immunostaining (Fig. 2A,C).
Thus, we hypothesized that Kras allelic imbalance and amplified MAPK signaling could represent a mechanism selected to ensure effective inactivation of CIC and thus facilitate tumor initiation. In this view, CIC repressor activity would act as a barrier to LUAD development, while Kras allelic amplification would promote sufficient inactivation of CIC to overcome this barrier. However, as we examined this idea, we unexpectedly found that LUAD tumors harboring excess MAPK activity did not present changes in CIC protein levels or localization (Fig. 2A). Similarly, using our recently described Kras+/G12Vlox;Trp53–/–;Rosa26-CreERT2KI/KI;hUBC-CreERT2+/T (KG12VloxPC2) cell lines, in which the oncogenic KrasG12V allele can be excised by adding 4-hydroxytamoxifen (Salmón et al, 2023), we observed that elimination of KrasG12V had little impact on localization or stability of CIC (Fig. EV3A,B). However, LUAD tumors do show a significant increase in ETV5+ cells (Fig. 2A,D), raising the possibility that amplified MAPK signaling might inhibit CIC repressor activity by a mechanism independent of subcellular localization of protein degradation. Indeed, loss of KrasG12V in KG12VloxPC2 cells resulted in enhanced binding of CIC to its target promoters, suggesting that CIC activity in LUAD is regulated at the level of DNA binding (Fig. EV3C). Notably, we also found that genetic disruption of CIC significantly suppressed the mutant Kras allelic expansion characteristic of KP tumors (Fig. 2E,F). However, this suppression was not as evident in tumors retaining functional p53 (Fig. EV4A,B). Since LUADs also show a frequent gradual increase in ERBB activity (Kruspig et al, 2018), we also analyzed whether the absence of CIC affected expression of ERBB family receptors and ligands. As shown in Fig. EV4C, we did not detect differences in their expression levels in KP and KPCic tumors, suggesting that this mechanism of tumor progression is not affected by the absence of CIC, at least at the time point analyzed. Interestingly, in human LUAD as well as other tumor types carrying KRAS mutations, the presence of CIC mutations correlates with the lack of KRAS oncogene amplification (Fig. EV4D,E). Together, these observations indicate that CIC inactivation, either via mutation or in response to sufficiently strong, amplified MAPK signaling, is key for LUAD development in KP mice.

Genetic CIC inactivation promotes expansion and lineage conversion of Club cells
Next, we aimed at identifying the origin of the elevated tumor burden observed upon genetic disruption of Cic. To this end, we infected KP as well as KPCic mice with Ad-Cre and quantified clusters of X-Gal+ cells as a surrogate marker for KRASG12V expression 4 weeks post-infection (Guerra et al, 2003; Mainardi et al, 2014). As illustrated in Fig. 3A,B, we did not observe differences in the number and size of X-Gal+ clusters between KP and KPCic mice within alveoli. However, KPCic mice exhibited significantly more X-Gal+ clusters with 5 or more cells within or close to bronchioles, suggesting that, despite the absence of the bronchiolar marker CC10 in full-blown tumors, CIC inactivation facilitated expansion of bronchiolar cells. Indeed, accumulating evidence indicates that bronchiolar Club cells can serve as the cellular origin for KRAS-driven lung tumors while transitioning from expressing bronchiolar to alveolar lineage markers (Rosigkeit et al, 2021; Spella et al, 2019; Cicchini et al, 2017; Sutherland et al, 2014; Nieto et al, 2017). To ascertain whether additional tumors arising in KPCic mice originated from Club cells, we infected KP and KPCic mice with Ad-CC10-Cre particles that restrict Cre expression to CC10-positive cells such as Club cells (Sutherland et al, 2011). Immunostaining of CC10 and SPC markers revealed the emergence of double-positive cells within bronchioles 4 weeks post-infection in both KP and KPCic mice, indicating that activation of KRASG12V with concomitant elimination of p53 caused a subset of Club cells to express alveolar type 2 (AT2) cell markers (Fig. 3C). Yet, these cells expanded more rapidly in KPCic mice as shown by the appearance of significantly larger clusters of double-positive cells that ceased to express CC10 when they continued to grow, while mostly retaining expression of SOX2, another Club cell marker (Sutherland et al, 2014) (Figs. 3C,D and  EV4F). Importantly, KPCic mice also displayed significantly more tumors 4 months after infection with Ad-CC10-Cre, indicating that accelerated conversion of Club cells into AT2-marker-positive tumor cells and their subsequent expansion could be a possible explanation for their increased tumor burden (Fig. EV4G).

Latent Kras mutations cooperate with concomitant loss of Trp53 and Cic in tumor initiation
The above results underscore the contribution of CIC inactivation to lung tumor development driven by Kras oncogenes. Although we previously observed that systemic disruption of Cic in mice for up to 9 months did not result in any histopathological alteration in lung tissue (Simón-Carrasco et al, 2017), we now explored whether latent mutations in other oncogenes might cooperate with the disruption of Cic in lung tumor development. Interestingly, two thirds of human CIC-mutant LUADs harbor co-occurring mutations in TP53 or other genes mutated in lung cancer, including KRAS (Fig. 4A). Thus, we infected Ciclox/lox, Trp53lox/lox or Trp53lox/lox;Ciclox/lox (PCic) mice with Ad-Cre and evaluated their tumor burden after 1 year. Although none of these animals became sick during this time, further histopathological analyses revealed the presence of lung adenocarcinomas positive for TTF-1 and SPC in 10 out of 22 PCic mice (Fig. 4B–D). Interestingly, neither Ciclox/lox nor Trp53lox/lox mice infected with Ad-Cre showed histological alterations in their lungs at the same time point. Further characterization of these tumors by whole exome sequencing revealed the presence of latent or spontaneous KrasQ61R mutations in 3/3 tumors analyzed (Fig. 4E). Moreover, these mutations were present at low frequencies, suggesting that none of these tumors showed notable Kras oncogene amplifications, as typically observed in KP mice. These tumors also had significantly lower levels of pERK than those from KP mice when sacrificed at a humane endpoint (Fig. 4F). A fraction of KRAS/TP53 mutant LUAD patient samples in The Cancer Genome Atlas (TCGA) database also harbor loss-of-function mutations in CIC that lack KRAS imbalances (Appendix Fig. S1A). Interestingly, these tumors show no differences in the expression of a genetic signature indicative of MAPK pathway activation that includes multiple CIC target genes (Pratilas et al, 2009), consistent with our findings in mice (Appendix Fig. S1B). Yet, KRAS/TP53/CIC-mutant LUADs display multiple differentially expressed genes when compared with KRAS/TP53-mutant tumors, although this, as well as the lack of allelic imbalances, needs to be confirmed in additional tumor samples (Appendix Fig. S1C). Moreover, since the number of tumors with this mutational profile is relatively low, it remains to be determined whether CIC mutations influence any clinical parameters. Taken together, our data suggest that disruption of Cic facilitates initiation of lung adenocarcinomas in KP mice, possibly by rendering Kras allelic expansion unnecessary at early stages.

Reintroduction of CIC decreases tumor cell proliferation and reverts resistance to MAPK pathway inhibitors
The above results underscore the impact of CIC inactivation in KRAS-driven LUAD. Hence, we aimed at exploring whether reactivation of CIC could represent a potential therapeutic scenario. To this end, we established cell lines from tumors that developed 5 months after infection in KP and KPCic mice. These tumor cell lines did not differ in their growth properties or the activity of classical signaling pathways downstream of KRAS (Figs. 5A and  EV5A). Yet, as expected (Simón-Carrasco et al, 2017; Wang et al, 2017; Liao et al, 2017), KPCic cells exhibited resistance to trametinib (Fig. 5B). To test whether reactivation of CIC affected cell proliferation and/or resistance to trametinib, we infected KP and KPCic tumor cells with adenoviral vectors expressing GFP as a control, the CIC-S cDNA fused to GFP (GFPCIC) (Dissanayake et al, 2011) or phosphorylation-insensitive GFPCICS173A known to more effectively repress its target genes (Park et al, 2023). Moreover, to efficiently reconstitute CIC’s transcriptional repressor activity, we ectopically expressed the corepressor ATXN1L in these experiments (Fig. 5C) (Lee et al, 2011; Wong et al, 2019), since cell lines established from KP and KPCic tumors displayed significantly lower levels of Atxn1 and Atxn1l than the respective tumors, which were insufficient to allow efficient transcriptional repression upon adenoviral expression of CIC proteins (Fig. EV5B,C). As shown in Fig. 5D, ectopic expression of GFPCIC still caused little or no repression of Etv4 and Etv5. However, ectopic expression of GFPCICS173A resulted in efficient repression of these genes in most KP and KPCic cells. Likewise, GFPCICS173A was also more effective than GFPCIC in preventing colony growth of KPCic cells, while both CIC variants were equally effective in KP cells (Fig. 5E,F; Appendix Fig. S2). More importantly, however, expression of GFPCIC and GFPCICS173A re-sensitized resistant KPCic cells to trametinib treatment (Fig. 5F and Appendix Fig. S2).

ETV4 and ETV5 mediate proliferation and resistance to MAPK pathway inhibition
Next, we examined which CIC-controlled genes contribute to KRAS-driven tumor cell proliferation and drug resistance. To this end, we conducted RNA-seq of three independent KP and KPCic tumor cell lines either treated with 20 nM trametinib for 24 h or DMSO as a control. We reasoned that while trametinib treatment would inhibit the MAPK pathway in KP cell lines, resulting in higher CIC repressor activity, the lack of functional CIC in KPCic cells should prevent repression of its targets. As demonstrated in Fig. 6A; Dataset EV3, the genes that best follow this expression pattern are Etv4 and Etv5. Constitutive derepression of these genes, as well as Ccnd1, another known CIC target (Dataset EV1), was also confirmed by quantitative PCR in KPCic cells, while other known MAPK-regulated genes do not follow this pattern (Figs. 6B and  EV5D), despite binding of CIC to their promoters (Fig. EV2A,B). These results also indicate that, although endogenous Atxn1/Atxn1l levels are significantly reduced, they remain sufficient to sustain CIC repressor activity in KP cell lines without the need for ectopic ATXN1L expression. Interestingly, ETV4 and ETV5 protein levels were not entirely MAPK pathway-independent in KPCic cells, suggesting the existence of CIC-independent mechanisms controlling their translation or stability (Fig. 6C).
Based on these observations, we downregulated Etv4 or Etv5 expression using two independent shRNAs in KP as well as KPCic cell lines. As depicted in Appendix Fig. S3, individual downregulation of either gene had no effect on colony formation in the absence of trametinib and only moderately affected colony growth in its presence, albeit with slightly stronger growth inhibition after knockdown of Etv5. Hence, we assessed the consequences of the combined knockdown of Etv4 and Etv5. While their combined absence faintly reduced colony formation in KP cell lines, it strongly inhibited colony growth in KPCic cell lines and re-sensitized CIC-deficient cells to treatment with trametinib (Fig. 6D–F). Likewise, overexpression of ETV4 also promoted resistance to trametinib (Fig. EV5E,F). Interestingly, knockdown of either ETV4 or ETV5 strongly inhibited the growth of human KRAS-mutant lung cancer cell lines, suggesting that the requirement for these transcription factors extends to human lung cancer (Fig. 6G,H). Together, these results indicate that ETV4 and ETV5 are involved in tumor cell proliferation and are the main mediators of drug resistance upon inactivation of CIC.

Genetic inactivation of CIC creates selective vulnerabilities to overcome resistance
Since CIC inactivation promotes resistance to MAPK pathway inhibitors, we reasoned that it may also expose specific vulnerabilities that could be exploited to selectively block the proliferation of tumor cells that have lost CIC activity. To this end, we used a collection of 114 cancer drugs to identify compounds that selectively interfere with the proliferation of KPCic cells (Dataset EV4). We identified three drugs that significantly inhibited the growth of KPCic cells by more than 20% when compared to KP cell lines (Fig. 7A). These drugs (quizartinib, Tx-1123 and PFK15) were then validated individually in dose-response experiments in three independent KP and KPCic cell lines (Fig. 7B). While quizartinib did not exhibit substantial activity in KP or KPCic cell lines, Tx-1123 and PFK15 more strongly affected the survival of KPCic cells (Fig. 7B). Given their selectivity for KPCic cell lines, we next explored whether Tx-1123 and PFK15 were able to revert resistance to trametinib due to genetic CIC inactivation. As depicted in Fig. 7C, treatment with Tx-1123 and PFK15 had little effect on the viability of KP cell lines upon treatment with trametinib, especially at high doses. However, they strongly enhanced the response to trametinib in KPCic cell lines to a level observed in KP cells. Next, we aimed to recapitulate the effect of PFK15 in human lung cancer cells. To this end, we eliminated CIC from the PDX-derived cell line PDX-dc1 (Sanclemente et al, 2018) using CRISPR/Cas9. As illustrated in Fig. 7D,E, elimination of CIC triggered resistance to trametinib, which was reverted by the addition of PFK15. More importantly, PFK15 strongly reduced the survival of trametinib-resistant patient-derived organoids that expressed low CIC and high ETV5 levels, while trametinib-resistant organoids showing high CIC and low ETV5 expression levels were not affected (Fig. 7F,G; Appendix Fig. S4). Together, these results indicate that genetic inactivation or absence of CIC creates vulnerabilities that can be exploited pharmacologically.

Discussion
Activation of the MAPK pathway is critical for KRAS-driven LUAD development (Drosten and Barbacid, 2020). However, it has remained unknown how the MAPK pathway contributes to tumor growth, especially downstream of ERK kinases. Our findings unveil a link between KRAS copy number gains, amplification of MAPK signaling and CIC inactivation during the initiation of KRAS-driven lung tumors, which also has significant therapeutic implications.
It was previously shown that KRAS-driven LUADs show stage-specific amplification of MAPK signaling, but the molecular basis for amplified signaling or its consequences were unclear (Chung et al, 2017; Cicchini et al, 2017; Chen et al, 2019; Feldser et al, 2010). Here, we show that progressive allelic imbalance of mutant Kras correlates with amplified MAPK signaling in Kras/Trp53 mutant mice. Allelic imbalance, which can develop either through copy number gains of mutant Kras or LOH of the wild-type allele, is known to enhance tumor initiation in various contexts (Ambrogio et al, 2018; Najumudeen et al, 2024; Zhang et al, 2001). Our results also offer a plausible explanation for Kras copy number gains and amplified MAPK signaling. Since concomitant genetic inactivation of Cic abrogates Kras allelic imbalance and amplified MAPK signaling, we propose that Kras copy number gains may be selected for in tumors to promote stronger functional inactivation of CIC. In this model, Kras allelic imbalance leads to MAPK signal amplification and correspondingly stronger downregulation of CIC, i.e., a graded signal-response mechanism that appears to be conserved from Drosophila to mammals (Jiménez et al, 2012; Rodríguez-Muñoz et al, 2022). Conversely, when oncogenic KRAS signaling is accompanied by genetic inactivation of CIC, the complete lack of CIC-mediated repression would allow tumor initiation without KRAS allelic imbalances (Appendix Fig. S5). Yet, suppression of Kras amplification in the absence of Cic is less evident in tumors retaining p53. The reason for this observation is currently unknown, but could be related with overall slower tumor progression rates in this model and the advanced age of the animals analyzed. Interestingly, when we inactivated Cic and Trp53 alone or in combination in mice, tumors only formed when cells concomitantly lacking Cic and Trp53 acquired latent Kras mutations, suggesting that the absence of p53 is required to expose the effect of Cic inactivation on tumor initiation.
In contrast to an earlier report (Okimoto et al, 2017), we found no evidence for an elevated metastatic potential of Cic-deficient tumor cells. The reasons for this discrepancy are also unclear, but our results are consistent with a more recent study demonstrating a role for CIC specifically during initiation of KRAS-driven lung tumors (Cai et al, 2021). Nevertheless, as opposed to the deletion of Nf1 (Wang et al, 2019), for instance, the disruption of Cic only moderately enhances tumor initiation in our model. While loss of Nf1 accelerates tumorigenesis via multiple mechanisms, including amplification of KRAS signaling and activation of FAK1, disruption of Cic will only affect a small subset of genes controlled by KRAS. Moreover, since the transcriptional profile of tumors lacking Cic obtained 5 months after infection does not substantially differ from those retaining Cic, notable derepression of CIC target genes might be confined to a narrow window during the early steps of tumor initiation in which Kras amplifications do not yet occur.
Interestingly, functional CIC inactivation may not be the only critical consequence of amplified MAPK signaling, since concomitant elimination of the tumor suppressor Rb1 also abrogates the requirement for amplified MAPK signaling in KRAS-driven LUAD (Walter et al, 2019). Interestingly, both CIC and RB1 proteins function as tumor suppressors that can be inactivated through phosphorylation, suggesting that amplified MAPK signaling might ensure maximal inactivation of CIC and RB1, together with other potential molecular events that contribute to the transformed phenotype. In this context, loss of CIC or RB1 alone may be sufficient to induce gene expression changes that facilitate tumor growth even without MAPK signaling amplification, especially at early stages of tumorigenesis.
Several studies have demonstrated that amplified MAPK signaling can promote transformation of Club cells, suggesting that these cells require a higher threshold of MAPK pathway activity for transformation (Cicchini et al, 2017; Nieto et al, 2017). Our data show that genetic inactivation of CIC phenocopies the effect of amplified MAPK signaling in Club cells, providing further evidence for CIC as a major target of Kras allelic imbalance and amplified MAPK signaling. Furthermore, our results align with a recently observed transient Club/AT2-marker double-positive cellular state that originates during lineage conversion upon oncogenic insult (Chen et al, 2022). A similar cellular state emerges during lung regeneration upon tissue damage, thus raising the possibility that CIC inactivation also contributes to lineage conversion of Club cells in this context (Zheng et al, 2013). Alternatively, this transient double-positive condition may reflect a common response to various, unrelated stress stimuli such as tissue damage or oncogenic signaling, including CIC inactivation.
CIC mutations have been identified in experimental models and patients that developed resistance to drugs targeting the MAPK pathway (Simón-Carrasco et al, 2017; Wang et al, 2017; Liao et al, 2017; Da Vià et al, 2020; Hashiba et al, 2020). In line with these observations, our results show that KP tumor cells expressing inactive CICΔ2-6 proteins are resistant to the MEK inhibitor trametinib. Interestingly, reconstitution of CIC repressor activity via ectopic expression of CICS173A, a mutant CIC protein that is less responsive to ERK-mediated inactivation, and ATXN1L, a corepressor required for CIC stability and DNA binding (Lee et al, 2011; Wong et al, 2019), not only reduced tumor cell proliferation, but also restored sensitivity to trametinib. Concomitant ATXN1L overexpression was only required upon adenoviral expression of CIC variants, since endogenous Atxn1/Atxn1l levels, albeit reduced, were sufficient to sustain CIC’s repressor activity in KP cells. Overall, these observations demonstrate that repression of CIC target genes is required for the anti-tumor effects of MAPK pathway inhibitors. Indeed, the well-known CIC targets ETV4 and ETV5 are key mediators of tumor cell proliferation and drug resistance. Both ETV4 and ETV5 are frequently overexpressed in human NSCLC and are associated with poor patient prognosis (Wang et al, 2020; Cheng et al, 2019; Li et al, 2023). While both genes encode transcriptional activators, how exactly they contribute to these activities remains to be determined.
In addition, we reasoned that genetic inactivation or mutation of Cic may also expose specific vulnerabilities that can be exploited therapeutically. Indeed, we identified two drugs, Tx-1123 and PFK15, with preferential activity in Cic-deficient tumor cells, suggesting that these drugs may be an alternative for patients who develop resistance via mutation of CIC. While the pleiotropic effects of Tx-1123 make it difficult to predict how this drug could preferentially affect Cic-deficient cells (Hori et al, 2002), PFK15 specifically targets the rate-limiting glycolytic enzyme PFKFB3 (Shi et al, 2017; Clem et al, 2013). This suggests that CIC-deficient tumor cells are particularly dependent on glucose metabolism, likely because of constitutive derepression of specific target genes. Indeed, our data (Datasets EV1, EV3) as well as previous observations indicate that CIC controls multiple genes implicated in metabolic processes, suggesting that CIC-deficient cells could be particularly vulnerable to interference with metabolic requirements (Park et al, 2023; Wong et al, 2019).

Methods

Mouse strains and tumor induction
KrasLSLG12Vgeo (Guerra et al, 2003), Trp53lox (Jonkers et al, 2001), and Ciclox (Simón-Carrasco et al, 2017) alleles have been published. To induce lung tumor formation, 8–12 weeks old mice were intranasally infected as described (Esteban-Burgos et al, 2020) with 5 × 105 or 5 × 107 pfu Ad-Cre or 5 × 107 pfu Ad-CC10-Cre (Sutherland et al, 2011). Adenoviral vectors were purchased from the Viral Vector Core at the University of Iowa. Animals were maintained in a mixed 129/Sv-C57BL/6 background. Female and male mice were used for the experiments, and no blinding was done. No sample size estimations were performed. All animal experiments were approved by the Ethical Committees of the Spanish National Cancer Research Centre (CNIO), the Carlos III Health Institute and the Autonomous Community of Madrid (PROEX 161/14), or the Bioethics Committee of the University of Salamanca as well as the Castilla y Leon Autonomous Government (716 CSIC-USAL) and were performed in accordance with the guidelines stated in the International Guiding Principles for Biomedical Research Involving Animals (CIOMS). Mice were housed in specific-pathogen-free conditions at the Animal Facilities of the CNIO (Association for Assessment and Accreditation of Laboratory Animal Care, JRS: dpR 001659) or the Cancer Research Center (CIC) in Salamanca.

Cell lines and treatments
Tumor cell lines were established from tumors in KP or KPCic mice 5 months after infection with Ad-Cre and grown in DMEM supplemented with 5% FBS. The human PDX-derived cell line PDX-dc1 was previously described (Sanclemente et al, 2018) and maintained in DMEM supplemented with 10% FBS. NCI-H358 cells were obtained from the ATCC, grown in RPMI-1640 medium supplemented with 10% FBS, and authenticated using the CLA IdentiFiler Plus kit (Thermo Fisher Scientific). All cell lines were routinely tested for mycoplasma contamination. Trametinib was purchased from MedChemExpress and used at the indicated concentrations. PFK15 and Tx-1123 were purchased from Sigma-Aldrich.

Histopathology and immunohistochemistry
Tissues were fixed in 10%-buffered formalin (Sigma-Aldrich) and embedded in paraffin. For histopathological visualization, 2.5 µm tissue sections were stained with Hematoxylin & Eosin (H&E) and tumors were classified according to standard histopathological grade criteria (Jackson et al, 2005). Antibodies used for immunostaining included those raised against: Ki67 (Cell Signaling Technology, 12202, 1:50), phospho-ERK1/2 (Cell Signaling Technology, 9101, 1:300), SPC (Abcam, ab212326, 1:4000), CIC (Invitrogen, PA5-83721, 1:100), ETV5 (Proteintech, 13011-1-AP, 1:300), CC10 (Santa Cruz Biotechnology, sc-9772, 1:1000) and TTF-1 (Epitomics, 2044-1, 1:100). For imaging analysis, slides were scanned on a ZEISS Axio Scan.Z1 scanner and processed using QuPath version 0.4.2 software.

Western blot analysis
Proteins were extracted in protein lysis buffer (50 mM Tris-HCl pH 7.5, 150 mM NaCl, 1% NP-40) supplemented with cOmplete Mini protease inhibitors (Roche). A total of 30 μg protein extracts were separated by SDS-PAGE, transferred to a nitrocellulose blotting membrane (GE Healthcare) and blotted with antibodies raised against: CIC (Invitrogen, PA5-83721, 1:1000), ETV4 (Proteintech, 10684-1-AP, 1:1000), ETV5 (Abcam, ab102010, 1:1000), ERK1/2 (Cell Signaling Technology, 4695, 1:1000), phospho-ERK1/2 (Cell Signaling Technology, 9101, 1:500), AKT (Cell Signaling Technology, 9272, 1:1000), phospho-AKT (Cell Signaling Technology, 9271, 1:1000), GFP-tag (Proteintech, 50430-AP, 1:1000), HA-tag (GenScript, A01244, 1:1000), GAPDH (Sigma-Aldrich, G8795, 1:5000) and Vinculin (Sigma-Aldrich, V9131, 1:5000).

Subcellular fractionation
Nuclear and cytoplasmic fractions were extracted with the NE-PER Nuclear and Cytoplasmic Extraction Reagents (Thermo Fisher Scientific) following the manufacturer’s guidelines. Efficient nuclear and cytoplasmic fractionation was confirmed by Western blot analysis using Lamin A/C antibodies (Santa Cruz Biotechnology, sc-376248, 1:1000) for the nuclear fraction and MEK1 antibodies (Santa Cruz Biotechnology, sc-6250, 1:500) for the cytoplasmic fraction.

Knockdown and CRISPR/Cas9 assays
Tumor cells were infected with lentiviral supernatants expressing shRNA against mouse Etv4 (TRCN0000295466 [sh-Etv4_1890] and TRCN0000306806 [sh-Etv4_1137], Sigma-Aldrich) and mouse Etv5 (TRCN0000312867 [sh-Etv5_1762] and TRCN0000312868 [sh-Etv5_1646], Sigma-Aldrich). For combined knockdowns, sh-Etv4_1137 was cloned into a pLKO.1 plasmid that carries a blasticidin resistance cassette and used in combination with sh-Etv5_1762. For knockdown experiments in human cell lines, cells were infected with lentiviral supernatants expressing shRNA against human ETV4 (TRCN000013937 [sh-ETV4_A] and TRCN000013934 [sh-ETV4_B], Sigma-Aldrich) or human ETV5 (TRCN000013938 [sh-ETV5_A] and TRCN000013930 [sh-ETV5_B], Sigma-Aldrich). A non-targeting shRNA vector (SHC002) was used as a negative control. A lentiviral vector expressing an sgRNA targeting human CIC (lentiCRISPRv2-sgCIC#1) has been described previously (Simón-Carrasco et al, 2017).

Construction of viral vectors expressing GFP-CIC variants, HA-ATXN1L, and ETV4-HA
Adenoviral vectors expressing GFPCIC or GFPCICS173A were generated using the AdEasy system (He et al, 1998). In brief, the GFPCIC cDNA was excised from pcDNA5/FRT/TO-GFP-CIC (Dissanayake et al, 2011) by digesting the DNA with AflII and NotI. An AflII restriction site also was introduced 5’ of the NotI restriction site in pShuttle-CMV (He et al, 1998) by site-directed mutagenesis, and the GFPCIC cDNA was subsequently cloned into this modified pShuttle-CMV vector after digestion with AflII and NotI. The S173A mutation was then introduced by site-directed mutagenesis. Adenoviral vectors were finally retrieved by homologous recombination of pAdEasy1 with the respective pShuttle constructs in Escherichia coli BJ5183 and introduction of the PmeI-linearized pAdEasy1-GFPCIC and pAdEasy1-GFPCICS173A into 293A cells. Adenoviral infections were carried out at multiplicities of infection (moi) of 50. Lentiviral vectors expressing HA-ATXN1L were generated by amplifying the human ATXN1L cDNA by PCR with forward primers including an EcoRI restriction site and an HA-tag following the ATG start codon and a reverse primer including an EcoRI site after the STOP codon. PCR products were cloned into pCR2.1 using the TOPO TA-Cloning Kit (Invitrogen) and sequence-verified. Finally, the HA-ATXN1L cDNA was excised with EcoRI and cloned into pLVXpuro (Clontech) after digestion with EcoRI. The correct orientation was confirmed by DNA sequencing. Retroviral vectors expressing ETV4-HA were generated by PCR-amplification of the human ETV4 cDNA from pCMVSport6 ETV4 (BC016623) with forward primers including an EcoRI restriction site and reverse primers adding an HA-tag preceding the STOP codon followed by an EcoRI restriction site. The resulting PCR product was digested with EcoRI and cloned into pLPC after digestion with EcoRI. The correct orientation was confirmed by DNA sequencing.

qRT-PCR
Total RNA was extracted with the RNeasy Mini Kit (Qiagen) and reverse-transcribed using the NZY First-Strand cDNA Synthesis kit (NZYTech) following the manufacturer’s instructions. Quantitative real-time PCR reactions were performed on a QuantStudio 6 Flex Real-Time PCR System (Applied Biosystems) using the PowerTrack SYBR Green PCR Master Mix (Applied Biosystems). Values were quantified according to the ΔΔCt method, and β-actin was used for normalization.

PCR analysis to verify KrasG12Vlox excision
PCRs were performed on genomic DNA from KG12VloxPC2 cells (Salmón et al, 2023) using specific forward (5’-GGAACTTCGCATGCATAACTTCGTATAATGT-3’) and reverse primers (5’-CAAAGCACGGATGGCATCTTGGACC-3’) for the KrasG12Vlox allele, yielding a 594 bp band corresponding to the unexcised allele and a 170 bp band for the excised allele. PCRs were performed using the DreamTaq Green PCR Master Mix (Thermo Scientific) with the following PCR conditions: 30 s 94 °C, 30 s 60 °C, 30 s 72 °C (35 cycles).

PCR analysis of Kras allele abundance
The relative abundance of KrasWT and KrasG12V alleles was estimated by semiquantitative PCR with the following PCR conditions: 30 s 94 °C, 30 s 60 °C, 30 s 72 °C (20 cycles); forward primer (5’-GATACATTCCTTTGAGAGCCATT-3’), reverse primer (5’-CTCAGTCATTTTCAGCAGGC-3’). This PCR yielded DNA fragments of 309 bp for the KrasWT allele and 398 bp for the recombined KrasG12V allele (Guerra et al, 2003). PCR band intensities were quantified with ImageJ.

Fluorescence-in situ hybridization (FISH) assays
Two sets of bacterial artificial chromosome (BAC) clones (RP24-359O2 and RP24-175B6 for Kras at 6qG3; and RP24-345H19 and RP24-275J18 for control probe at 6qA3.3) were obtained from the BACPAC Resources Centre (https://bacpacresources.org/) to generate a FISH probe to detect Kras amplification or chromosome 6 aneuploidies. The Kras probe was labeled with Spectrum-Orange and the control probe with Spectrum-Green, using the NICK translation assay (Abbot Molecular). FISH analyses were performed as previously described (Martinez-Lage et al, 2020) on 2.5 µm tissue sections mounted on positively charged slides (SuperFrost, Thermo Scientific). Briefly, the slides were first deparaffined in xylene and gradually rehydrated through a series of ethanol washes. The Histology FISH Accessory Kit (DAKO) was used following the manufacturer’s guidelines. The process included pre-treatment in 2-[N-morpholino]ethanesulphonic acid (MES), followed by protein digestion in a pepsin solution. After dehydration, the samples were denatured with the specific probe at 66 °C for 10 min and allowed to hybridize overnight at 45 °C in a DAKO hybridizer. The slides were then washed with a 20×SSC (saline-sodium citrate) buffer containing Tween-20 detergent at 63 °C and mounted in a fluorescence mounting medium (DAPI). FISH signals were manually counted within the nuclei throughout the tissue, and images were captured using a CCD camera (Photometrics SenSys camera) connected to a PC running the Zytovision image analysis system (Applied Imaging Ltd.) with focus motor and Z-stack software. The z-stack images were manually scored by two independent investigators by counting the number of co-localized signals.

Compound screening
A collection of 114 drugs covering multiple oncogenic pathways (see Dataset EV4) was used at a concentration of 5 μM in KP and KPCic cell lines. 3000 cells were seeded in 96-well plates, and the compounds were added the following day. Cell viability was assayed after 72 h using the Cell Titer Glo Luminescent Cell Viability Kit (Promega, G7571). Luminescence counts were read in a Victor plate reader (Perkin Elmer).

Generation of PDX-derived organoids
The lung PDX tumors TP40 and TP91 have been described previously (Garmendia et al, 2019; Quintanal-Villalonga et al, 2020). To generate organoids, PDX tumors were enzymatically digested with 1.2 mg/ml collagenase (Sigma-Aldrich, C9891), 10 mg/ml DNase (Sigma-Aldrich, D5025), and 0.125 mg/ml dispase (Life Technologies, 17105041) in Basic medium (Advanced DMEM/F12 [Gibco, 12634010], 1xHEPES [Life Technologies, 1560106], 1xGlutamine [Sigma-Aldrich, G8541]) at 37 °C in agitation for 1 h. After incubation, the digested tumor pieces were filtered using 70 mm filters, and the disaggregated cells were centrifuged at 1500 rpm for 5 min. After two washes with Basic medium, live cells were counted.

Organoid treatments
Organoids were treated with a drug matrix using 0, 10, 25, and 50 nM trametinib and 0, 5, and 15 μM PFK15 in triplicate for 7 days. 5000 cells/well were resuspended in 36 μl Matrigel (Corning, CLS356231) and seeded in a 96-well plate. Cells were allowed to solidify at 37 °C in a CO2 (5%) incubator for 10 min and 150 μl/well Complete medium (Basic Medium supplemented with 2% FBS, 3 ng/ml epidermal growth factor [Life Technologies, PMG8041], 5 mg/ml human insulin [Sigma-Aldrich, I3536], 1 mg/ml hydrocortisone [Sigma-Aldrich, H0888], 1x B27 (Life Technologies, 17504-044), 10.5 μM ROCK inhibitor Y-27632 [Sigma-Aldrich, SCM075]) was added to each well. PFK15 and trametinib were added 24 h after seeding and refreshed at day 4. Cell viability was assayed after 7 days using the Cell Titer Glo Luminescent Cell Viability Kit (Promega, G7571). Luminescence counts were read in a Victor plate reader (Perkin Elmer).

RNA sequencing and data analysis
Total RNA samples from cell lines were converted into sequencing libraries using the QuantSeq 3’ mRNA-Seq Library Prep Kit (FWD) for Illumina (Lexogen, Cat. No. 015) and sequenced on an Illumina NextSeq 550 instrument. Sequencing read analysis was performed by an automated analysis service provided by Lexogen. Briefly, reads were successively processed with bbduk from BBTools (BBMap, https://sourceforge.net/projects/bbmap/), aligned to the GRCm38/mm10 genome assembly using STAR v2.5 (Dobin et al, 2013) and counted with HTSeq (Anders et al, 2015). Differential gene expression analysis was performed using DESeq2 (Love et al, 2014). For RNA sequencing of mouse tumors, RNA was extracted using the PureLink RNA Mini Kit (Invitrogen). RNA sequencing was performed by HalpoX Gene Tech (Hong Kong) on a Novaseq Xplus. Reads were aligned to the Mus musculus reference genome (CRCm38/mm10 assembly) using the Burrows-Wheeler aligner BWA-MEM 0.7.15 (github.com/lh3/bwa.git). The aligned reads were converted to BAM files using Picard tools 2.9.0 (broadinstitute.github.io/picard). BAM files were then processed with StringTie (ccb.jhu.edu/software/stringtie) and edgeR (bioconductor.org/packages/edgeR). Differentially expressed genes (DGEs) were defined on an absolute log2 (foldchange) ≥0.6 and FDR <0.05, unless stated otherwise.

Whole-exome sequencing, sequence alignment, processing, and quality control
DNA extraction from formalin-fixed paraffin-embedded (FFPE) tissue and WES of mouse tumors was performed by CeGaT (Germany). Libraries were prepared using the Twist Human Core Exome kit with RefSeq and Mitochondrial panel (Twist Bioscience) and sequenced on the Illumina platform to depths of approximately 12 Gb/sample. Sequence reads were assessed for quality control using FastQC. FASTQ files were aligned to the Mus musculus reference genome (GRCm38/mm10 assembly) using the Burrows-Wheeler aligner BWA-MEM 0.7.15 with standard settings, and converted to BAM files using Picard tools 2.9.0. Following the alignment, we used Picard to mark duplicates. Collection of alignment and coverage metrics was performed with samtools and Picard. Targeted bases were sequenced to a mean depth of 100, and >75% of targeted bases were sequenced to 30x coverage or higher. The aligned sequence reads were visualized with the Integrative Genomics Viewer (igv.org). For the detection of single nucleotide variants (SNVs) and small insertions and deletions (indels) a Galaxy pipeline (bcftools mpileup/call; usegalaxy.eu) was used. Identified variants were annotated using Ensembl’s Assembly Converter and Variant Predictor tools (ensembl.org).

Chromatin immunoprecipitation (ChIP) assay and ChIP sequencing
ChIP assays were carried out as previously described (Simón-Carrasco et al, 2017). For precipitation of endogenous CIC proteins, 2.4 μg anti-CIC polyclonal antibodies (Invitrogen, PA5-83721) were added to 200 μg chromatin, followed by precipitation with 50 μl Protein A/G PLUS agarose (Santa Cruz Biotechnology, sc-2003). Immunoprecipitated chromatin was analyzed by qRT-PCR and quantified using the ΔΔCt method, with normalization to input DNA. For ChIP sequencing, cells were crosslinked by adding 1% formaldehyde at room temperature for 10–15 min. Crosslinking was quenched by adding glycine to a final concentration of 0.125 M. Fixed cells were collected by centrifugation at 1200 rpm for 5 min at 4 °C. The resulting pellet was washed three times with 1.5 mL of ice-cold PBS supplemented with cOmplete Protease Inhibitor (Roche). The crosslinked pellet was then lysed in 2 mL of LB1 buffer (Glycerol 1%, 50 mM HEPES-KOH pH 7.5, 140 mM NaCl, 0.5% NP-40, 0.25% Triton X-100, 1 mM EDTA, pH 8, cOmplete Mini). Samples were centrifuged at 2000×g for 5 min at 4 °C, and the pellet was resuspended in 2 mL of LB2 buffer (10 mM Tris-HCl, pH 8, 200 mM NaCl, 0.5 mM EGTA, 1 mM EDTA pH 8, cOmplete Mini). The mixture was incubated on a rotating platform for 5 min at 4 °C, followed by centrifugation at 2000×g for 5 min at 4 °C. Finally, the pellet containing isolated nuclei was resuspended in 200 µL of LB3 buffer (0.5% N-lauroyl sarcosine, 10 mM Tris-HCl pH 8, 100 mM NaCl, 0.1% sodium deoxycholate, 0.5 mM EGTA, 1 mM EDTA pH 8, cOmplete Mini). Samples were disrupted by sonication using a Bioruptor Plus (Diagenode) for 20 cycles of 30 s ON and 30 s OFF. After sonication, samples were centrifuged at 14,000×g for 20 min at 4 °C. The supernatant was collected and incubated with 75 µL of Dynabeads Protein A (Thermo Fisher Scientific), previously conjugated with 10 µg of anti-CIC polyclonal antibody (Invitrogen, PA5-83721), for 16 h on a rotator. Beads were collected using a magnetic rack and washed six times with RIPA buffer (50 mM HEPES-KOH, pH 7.6, 0.5 M LiCl, 1% NP-40, 0.7% sodium deoxycholate, 1 mM EDTA, cOmplete Mini) at 4 °C on a rotator for 5 min per wash. Subsequently, 200 µL of elution buffer (50 mM Tris-HCl, pH 8, 1% SDS, 1 mM EDTA), 2 µL of RNase A, and 4 µL of Proteinase K were added. To reverse crosslinking, the eluted chromatin was incubated at 65 °C for 6 h. DNA was purified by phenol-chloroform extraction and quantified using a Qubit 4 Fluorometer (Invitrogen). ChIP sequencing was carried out by HalpoX Gene Tech (Hong Kong) on a DNBSEQ. Raw fastq reads were aligned to the mouse reference genome (GRCm38/mm10 assembly, http://genome.ucsc.edu) using Burrows-Wheeler aligner BWA-MEM 0.7.15 (github.com/lh3/bwa.git) with standard settings. The aligned reads were then converted to BAM files using Picard tools 2.9.0 (http://broadinstitute.github.io/picard). Duplicate reads were removed using Picard tools. BAM files were processed with MACS (github.com/taoliu/MACS/) (Zhang et al, 2008) version 3.0.0b3 for enrichment scoring and peak calling. Peaks were called using the callpeak function in MACS, and differential binding between experimental conditions was analyzed using the bdgdiff function. BED files were imported into RStudio and annotated with the R/Bioconductor package ChIPseeker (github.com/YuLab-SMU/ChIPseeker) (Yu et al, 2015). The promoter region was defined as −1 kb to +200 bp from the transcription start site (TSS).

Genomic datasets and analyses
RNA-Seq TPM data from TCGA-LUAD samples (n = 57) along with corresponding mutational and copy number alteration data were downloaded using UCSC Xena (https://xenabrowser.net/datapages/). Expression of each gene was compared between tumors with KRAS/TP53 and KRAS/TP53/CIC mutations using an unpaired Student’s t-test. Resulting p values were adjusted for multiple comparisons using a Bonferroni correction as an indication of significance. Normalized expression values (TPM) for each gene were also plotted using MORPHEUS software (https://software.broadinstitute.org/morpheus, Broad Institute) as a heatmap.

Statistical analyses
Data were represented as mean ± SD. P values were calculated with the unpaired Student’s t-test, one-way or two-way ANOVA tests, where indicated, using GraphPad Prism (v8.4.0) software. Survival differences were calculated using the log-rank test. Differences in Kras allele frequencies determined by FISH assays were calculated using Χ2 tests of a previously generated contingency table of allele distributions. P values <0.05 were considered statistically significant. Animals were allocated to different groups based on their genotypes, so no randomization could be performed.

Supplementary information