HACE1 Prevents Lung Carcinogenesis via Inhibition of RAC-Family GTPases

Lung cancer is the leading cause of cancer-related death in developed countries (1). The prognosis of patients with non–small cell lung carcinoma remains poor despite therapeutic advances due to late diagnosis, limited efficacy of available treatments, and development of drug resistance (2–4). Lung cancer initiation and progression are the result of genetic alterations and deregulation of several critical signaling pathways that prevent DNA damage and control critical cellular processes, such as proliferation, DNA repair, and apoptosis (2, 5). Unraveling the complex molecular mechanisms of lung cancer development is therefore crucial to enable the identification of novel therapeutic targets.

HACE1 (HECT-domain and ankyrin-repeat containing E3 ubiquitin protein ligase 1) was identified in the context of Wilms' tumor (6). We previously showed that genetic inactivation of Hace1 renders mice more susceptible to spontaneous and induced tumors, highlighting the function of HACE1 as a tumor suppressor (7). Since then, several independent studies have shown that HACE1 is mutated or functionally inactivated in different types of human cancer, including osteosarcoma (8), B-cell lymphoma (9), colon carcinoma (10), gastric carcinoma (11), and breast carcinoma (12). In addition, low HACE1 expression has been linked to a poor overall survival of patients with osteosarcoma (8), hepatocellular carcinoma (13), and neuroblastoma (14). We recently reported that Hace1-deficient mice show increased susceptibility to inflammation-driven colon cancer due to deregulated tumor necrosis factor receptor 1 (TNFR1) signaling and exaggerated induction of RIPK3-dependent necroptosis (15). However, dependent on the type of tumor and based on in vitro experiments, other mechanisms have been suggested to mediate tumorigenesis in the absence of HACE1, including impaired autophagic uptake of oxidatively damaged proteins (16) or deregulation of RAC1 activity (17–19).

RAC1 is a small GTPase that regulates cell proliferation, motility, stress responses, or production of reactive oxygen species (ROS; ref. 20). RAC1 belongs to a RAS superfamily of small GTP-binding proteins that also includes RAC2 and RAC3. These three proteins are 90% identical in their amino acid sequence but have different tissue-specific expression profiles. Although RAC1 is widely expressed, RAC2 expression is mainly confined to hematopoietic cells; RAC3, though more broadly expressed than RAC2, is most abundant in the brain (21, 22). Active GTP-bound RAC1 (GTP-RAC1) initiates cellular ROS production by plasma membrane and endosomal nicotinamide adenine dinucleotide phosphate (NADPH) oxidase complexes (17). Similarly to RAC1, RAC2 has also been implicated in NADPH oxidase–dependent ROS production (23). Although the ability of RAC3 to activate the NADPH oxidase complex was proposed, a role for RAC3 in ROS production has not yet been formally established (24).

Deletion of Rac1 was shown to impede tumorigenesis in animal models of skin cancer (25, 26), pancreatic cancer (27), and KRas^G12D-driven lung cancer (28). Conversely, hyperactivation of RAC1 and RAC2 can promote tumorigenesis in various cancer types (29, 30), such as breast cancer (12) and melanoma (31, 32). Importantly, HACE1 was shown to preferentially ubiquitylate and drive proteasomal degradation of GTP-RAC1 when it is bound to the NADPH oxidase complex (17–19, 33). As a consequence, Hace1-deficient cells show deregulated activation of RAC1 and increased levels of genotoxic cellular ROS, leading to genomic instability, DNA damage, and cyclin D1–mediated hyperproliferation (17). In addition, HACE1 can mediate the activation of NRF2 (34), reinforcing the function of HACE1 in the control of cellular oxidative stress.

We previously reported an increased susceptibility of Hace1^–/– mice to urethane-driven lung tumor (7). However, the molecular mechanisms by which HACE1 controls lung cancer formation remained unknown. Here, we investigated the role of HACE1 in KRas^G12D-driven lung tumors and uncovered a key role of deregulated activation of RAC1 and, surprisingly, also RAC2 in the pathogenesis of lung tumors in the absence of HACE1. We also report that in patients with lung cancer, increased expression of HACE1 is associated with prolonged survival, whereas elevated expression of RAC1 and RAC2 correlates with a poor prognosis, highlighting HACE1 and RAC-family GTPases as key drivers and potential targets for the treatment of lung cancer.

Hace1^−/− mice were generated in our laboratory as described previously (7) and crossed to LSL-KRas^G12D mice (35) to generate KRas^G12D;Hace1^−/− and control KRas^G12D;Hace1^+/+ animals. To analyze the role of RAC1 and RAC2 in KRas^G12D-driven lung cancer in the absence of HACE1, Rac1^fl/flRac2^−/− mice on a C57BL/6J background (generated by David A. Williams, Boston Children's Hospital) were crossed to KRas^G12D;Hace1^−/− mice, generating KRas^G12D;Hace1^+/–Rac1^fl/+Rac2^+/– mice. These triple-heterozygous mice were then interbred to yield KRas^G12D;Hace1^−/−, KRas^G12D;Rac1^fl/fl, KRas^G12D;Hace1^−/−Rac1^fl/fl, KRas^G12D;Rac2^−/−, KRas^G12D;Hace1^−/−Rac2^−/−, KRas^G12D;Rac1^fl/flRac2^−/−, and KRas^G12D;Hace1^−/−Rac1^fl/flRac2^−/− as well as control KRas^G12D;Hace1^+/+Rac1^+/+Rac2^+/+ mice. The generation of Ripk3^–/– animals has been previously described (36). Ripk3^–/– and KRas^G12D;Hace1^−/− mice were bred to generate KRas^G12D;Hace1^−/−Ripk3^–/– mice. Animals were genotyped by PCR analysis of genomic DNA. If not stated otherwise, 8- to 12-week-old, sex- and age-matched mice were used for experiments. All animal experiments were carried out complying with the current Austrian and European legislation and have been approved by the Austrian Federal Ministry of Science, Research and Economy (GZ) BMWFW-66.015/0008-WF/II/3b/2014 and (GZ) BMBWF-66.015/0030-V/3b/2019.

Inhalation of 8- to 12-week-old mice with Cre-expressing adenovirus was performed as previously reported (35, 37, 38). Briefly, mice were anesthetized with Ketasol/Xylasol and placed on a heated pad. An Adeno-Cre-CaCl₂ precipitate was produced by mixing 60 μL MEM, 2.5 μL Adeno-Cre (10¹⁰ pfu/mL; University of Iowa, Gene Transfer Vector Core, Iowa, USA), and 0.6 μL CaCl₂ (1 mol/L) for each mouse and incubated for 20 minutes at room temperature (21–22°C). Adeno-Cre was delivered by intratracheal administration to induce KRas^G12D expression and Rac1 deletion in pneumocytes in vivo. In vivo treatment with N-acetylcysteine (NAC) was performed starting at day 9 after lung tumor induction by supplementing NAC (Sigma, A9165) in drinking water (1 g/L).

For histopathology analysis of lung tumors, lungs were fixed in 4% paraformaldehyde overnight at 4°C and embedded in paraffin after dehydration in ascending concentrations of ethanol. For each lung, sections were prepared at a thickness of 2 μm at three different levels and stained using hematoxylin and eosin (H&E) with an automated stainer (HMS 740, Microm). Sections were imaged and digitized by the 3D Histech Pannoramic Flash II whole slide scanner. Tumor burden was automatically evaluated by an algorithm programmed and executed using the Definiens Developer software suite. IHC was performed using an automatic staining system (Leica Bond III). After dehydration and heat-induced antigen retrieval using a 10 mmol/L citrate solution (pH 6.0), sections were incubated with rabbit monoclonal anti-γH2AX (Cell Signaling Technology, 9718, 1:200 dilution) or rabbit monoclonal anti-Ki67 (Abcam, ab16667, 1:200 dilution). γH2AX- and Ki67-positive tumor cells were manually counted by a pathologist using a Zeiss Axioskop 2 MOT microscope. Positive tumor cell nuclei were counted in 10 representative 40x objective fields per tumor section. Representative H&E and IHC images were acquired by the Pannoramic Viewer software and a SPOT Insight color camera (SPOT Imaging, Diagnostic Instruments, Inc.). Data were validated by a certified pathologist.

mRNA was enriched using the NEBNext Ultra RNA Library Prep Kit and sequenced using singleend sequencing with 50bp reads on an Illumna HiSeq2500. RNA sequencing (RNA-seq) reads were trimmed using trim_galore v0.3.7, and reads mapping to known mitochondrial and ribosomal sequences were removed using bowtie2 v2.1.0. Thereafter, reads were aligned to the Mus musculus genome mm10 (iGenomes) using star v2.5.0a, and reads in genes were counted with featureCounts from subread v1.4.6. Gene expression measurements in fragments per kilobase per million reads and transcripts per million were calculated using RSEM v1.2.25. The RNA-seq dataset produced in this study is available at NCBI Gene Expression Omnibus database, accession number GSE149137 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE149137).

All cell lines were cultured in DMEM (Dulbecco) supplemented with 10% FCS, nonessential amino acids, and penicillin–streptomycin. Wild-type (+/+) and Hace1 knockout (−/−) mouse embryonic fibroblasts (MEF) were derived from Hace1^+/+ and Hace1^−/− littermates bred on a C57BL/6J background and allowed to spontaneously immortalize (17). HEK293 cells were purchased from the ATCC (CRL-1573) and authenticated by the provider. The University of Arizona Cancer Center (UACC) melanoma cell lines were obtained from TGen and were started using fresh patient material at the University of Arizona and originally banked at the UACC. The HACE1 mutations in the UACC lines were initially discovered using exome sequencing and authenticated with Sanger sequencing. UACC lines used for experiments were of low passage number. All cell lines were tested negative for Mycoplasma before experiments. MSCV-HA and MSCV-HA-Hace1 vectors were cloned as described in ref. 7. Transfections of siRNAs were performed with 25 nmol/L siRNA using Lipofectamine RNAiMax (Invitrogen). The following siRNAs were used in the study: Control siRNA (C): (5′–3′) AUAUCGGCUAGGUCUAACA; HACE1-1 (H1): Hs_HACE1_ 1 (FlexiTube, Qiagen); HACE1-2 (H2): Hs_HACE1_4 (FlexiTube, Qiagen). Hydrogen peroxide (H₂O₂) and 2-acetylphenothiazine (ML171) were purchased from Sigma. Piperlongumine (PL) was purchased from RD Chemicals.

Dihydroethidium (DHE; Sigma) was used to measure superoxide levels, as previously described (17). In brief, 20-μm-thick sections of frozen tumors were incubated with 10 μmol/L DHE dissolved in methanol in the dark for 30 minutes at room temperature. Then, sections were adhered onto covered slips and mounted with fluorescent mounting medium (Vectorshield mounting medium for fluorescence with DAPI, Cedarlane) and immediately imaged using an epifluorescent microscope (Axio Observer Z1; Carl Zeiss; Excitation 540 nm and emission 605 nm) using a 20× objective lens. All images were captured using identical exposure times, and signal intensities were analyzed using ImageJ software.

Primary lung tumor cells from Adeno-Cre–treated LSL-KRas^G12D–carrying mice were purified as described (38, 39). In brief, lungs were dissected from 8- to 12-week-old mice, infused with IMDM containing 5,000 U/mL dispase (BD) and 200 U/mL DNase (Worthington), followed by an incubation for 1 hour at 37°C. After the isolation, cells were maintained in Ham's F-12 media supplemented with 15 mmol/L HEPES, 0.8 mmol/L CaCl₂, 0.25% BSA, ITS (Sigma), and 2% FCS, at 37°C and 5% CO₂ conditions.

The GST-tagged p21-binding domain (PBD) of PAK1 (GST–PBD) was used to specifically precipitate active GTP-RAC1 from lysates of primary lung tumor cells (40), according to the company's protocol (Thermo Fisher Scientific, 16118).

His-tagged recombinant human RAC1 (Cytoskeleton, RC01) and RAC2 (Cytoskeleton, RC02) were preloaded with GTP or GDP as described (41). Briefly, 300 ng of RAC1 and RAC2 were loaded with GTP (5 mmol/L) or GDP (5 mmol/L) by incubation for 30 minutes at 30°C in a buffer containing 20 mmol/L TRIS-HCl, 0.1 mmol/L DTT, 2.5 mmol/L NaCl, and 4 mmol/L EDTA (pH 7.5). Subsequently, GTP- and GDP-loaded RAC1 or RAC2 (300 ng) were incubated for 3 hours at 37°C with recombinant human HACE1 (wild-type or catalytic dead C876S mutant, 2 μg), Ube1 (E1, 250 ng), Ubch7 (E2, 500 ng), ubiquitin (Sigma-Aldrich U6253, 10 μg), and ATP (Roche, 1051997900, 2 mmol/L) in a reaction buffer consisting of 50 mmol/L HEPES, 150 nmol/L NaCl, and 20 mmol/L MgCl₂ (pH7.5), at 37°C for 3 hours. Reactions were terminated by adding 2xSDS sample buffer and incubating at 96°C for 1 minutes. Samples were then subjected to SDS-PAGE and immunoblotting using the indicated antibodies.

Purification of E1 and E2 was performed as described in ref. 42.The codon-optimized sequence of human wild-type HACE1 or the catalytic dead C876S HACE1 mutant was cloned into pETM33_ccdB (Genewiz). The expression constructs were transformed into Escherichia coli Bl21(DE3) and cultivated overnight at 37°C in Luria-Bertani medium with kanamycin. Note that 3 L of Luria-Bertani medium was inoculated overnight (15 mL/L) and grown at 37°C until an OD600 of 0.5 was reached. Cultures were shifted to a shaker with preset temperature of 20°C. After 15 minutes (OD600 = 1), expression was induced with 1 mmol/L IPTG. After 6 hours, culture was harvested and stored at −80°C overnight. The pellet from the 3 L of expression culture was thawed and resuspended in 75 mL buffer A (50 mmol/L Tris, pH 7, 500 mmol/L NaCl, 0.5 mmol/L TCEP, 4°C). Cells were lysed via sonication. Cleared lysate was loaded onto a 5 mL GSTrap column equilibrated with buffer A. After a wash with 30 column volumes of buffer A, bound protein was eluted with buffer B (50 mmol/L Tris, pH 7, 500 mmol/L NaCl, 0.5 mmol/L TCEP, and 10 mmol/L reduced GSH, 4°C). The elution fraction was concentrated using Vivaspin (MWCO 50 kDa).

Immunoblotting was performed following standard protocols. Blots were blocked for 1 hour with 5% BSA in TBST (1X Tris-buffered saline (TBS) and 0.1% Tween-20) and then incubated overnight at 4°C with the indicated primary antibodies diluted in 5% BSA in TBST. After washing 3 times in TBST for 15 minutes, blots were probed with horseradish peroxidase–conjugated secondary antibodies (anti-mouse: Promega, W402B, 1:2,500 dilution; anti-rabbit: GE Healthcare, NA9340V, 1:2,500 dilution) for 1 hour at room temperature. Proteins were detected by enhanced chemiluminescence (ECL Plus, Pierce, 1896327). The following primary antibodies were used: anti-HACE1 (Abcam, ab133637, 1:500 dilution), anti-RAC1 (Cytoskeleton, ARC03, 1:500 dilution), anti–β-actin (Sigma, A5316, 1:10,000 dilution), anti-His (Thermo Fisher Scientific, MA1-21315, 1:1,000), anti-ubiquitin (P4D1; Santa Cruz Biotechnology, sc-8017, 1:1,000), anti-HA (Covance, MMS-101P, 1:1,000 dilution), anti-p53 (Cell Signaling Technology, 2524, 1:1,000 dilution), anti–phosphoS15-p53 (Cell Signaling Technology, 9284, 1:1,000 dilution), anti–cyclin D1 (Cell Signaling Technology, 2922, 1:1,000 dilution), and anti-GAPDH (Cell Signaling Technology, 2118, 1:2,000 dilution).

The mutational data of HACE1 in human cancers were obtained from published and publically available datasets on cBioPortal (www.cbioportal.org), including The Cancer Genome Atlas (TCGA) datasets and TGen datasets. For mRNA expression analyses, z-scores (RNA Seq V2 RSEM) for the lung adenocarcinoma (LUAD) TCGA cohort were obtained from cBioPortal. The cohort was stratified based on HACE1 (high z-score>0, low z-score<0), RAC1, and RAC2 (high z-score>1.5, low z-score<1.5) mRNA expression. Only groups with 10 or more patients were included in the analysis. Survival analysis was performed with R package survminer (43); P values were computed with log-rank test and adjusted by the Benjamini and Hochberg procedure (44). For construction of the correlation heatmap and correlation matrix, only samples with an absolute z-score higher than 2 in any of the 4 mRNA expression profiles were included. Correlation matrices were plotted with R package corrplot (45). GISTIC copy-number alterations and mutation data were retrieved from a 507 patient Lung Adenocarcinoma (TCGA, PanCancer) cBioPortal cohort.

All values are given as mean ± SEM). The Student t test was used to compare two groups. Multiple comparisons were analyzed by one-way ANOVA and two-way ANOVA followed by Tukey post hoc tests for multiple comparisons. Details of the statistical tests used are indicated in the respective figure legends. For the Kaplan–Meier survival analysis, a log-rank test was performed. In all tests, P < 0.05 was considered statistically significant. Statistical analyses were performed using GraphPad Prism (GraphPad Software).

To explore the tumor suppressor function of HACE1 in patients with cancer, we performed a meta-analysis of somatic mutations of HACE1 in several human tumors using published and publically available datasets from cBioPortal and TGen (46, 47). HACE1 consists of a catalytic HECT domain (amino acids 572–909), the linker domain (amino acids 258–571), and ankyrin repeats (ANK; amino acids 1–257). Although the HECT domain is responsible for the ubiquitin ligase activity of HACE1, the highly conserved ANK domain, especially ANK 4–7, mediates the interaction of HACE1 with its targets. Mutations in the HECT as well as in the ANK domains have been implicated to play a critical role in the tumor suppressor function of HACE1 (17–19, 33). We identified numerous mutations in the catalytic HECT domain as well as in the noncatalytic ANK and linker domains of HACE1 in patients with cancer (Fig. 1A; Supplementary Table S1). Furthermore, these somatic mutations occurred in multiple cancer types. Forty-one of the total 292 HACE1 mutations we identified in the databanks occurred in patients with lung cancer (lung adenocarcinomas, lung squamous cell carcinomas, and small cell carcinomas; Fig. 1B; Supplementary Table S1).

Previous studies reported that deletion/downregulation of HACE1 results in increased levels of ROS (17). Therefore, we used DHE to evaluate ROS production in patient-derived melanoma cells harboring HACE1 mutations (Fig. 1C). Tumor cells from patients carrying the mutations P811L and E739K in the HECT domain showed significantly higher ROS levels compared with tumor cells from patients with either wild-type HACE1 (wt) or a HACE1 mutation R332X in the linker domain, though the sample size is too small to also exclude a role of the linker region in ROS production. Thus, HACE1 is frequently mutated in different cancer types including lung cancer, and somatic mutations in the HECT domain of HACE1 in cancer cells derived from patients with melanoma can lead to elevated levels of cellular ROS.

As we have previously reported, aged Hace1^–/– mice have a higher incidence of spontaneous lung tumors, which is significantly increased upon urethane treatment compared with Hace1^+/+ littermate controls (7). Therefore, we speculated that HACE1 might also control tumorigenesis in KRas^G12D-driven lung tumors (35). In this model, mice bearing the Lox-Stop-Lox (LSL)-KRas^G12D allele can be induced to express oncogenic mutated KRas^G12D in lung epithelial cells after the removal of the Stop cassette following intratracheal administration of Cre recombinase–expressing adenovirus (Adeno-Cre), which leads to the development of epithelial hyperplasia and progression to benign adenomas and malignant adenocarcinomas (Fig. 2A; ref. 35). Therefore, we crossed Hace1-deficient (Hace1^–/–) animals to LSL-KRas^G12D conditional mice, generating KRas^G12D;Hace1^+/+ and KRas^G12D;Hace1^–/– animals. Indeed, KRas^G12D;Hace1^–/– mice showed significantly reduced survival in the KRas^G12D-driven lung tumor model compared with littermate controls following lung tumor induction (Fig. 2B). These data show that loss of Hace1 markedly enhances the development of KRas^G12D-driven lung tumors.

We have recently shown that HACE1 is involved in TNFR1-RIP3 kinase signaling and plays a role in inflammation-driven colon cancer (15). We therefore first determined if deregulation of TNFR1-mediated RIP3 kinase signaling contributes to the increased tumorigenesis in KRas^G12D;Hace1^–/– mice crossing KRas^G12D;Hace1 mutant mice to Rip3k knockout animals and induced lung tumors following Adeno-Cre administration (28). However, in KRas^G12D;Hace1^–/–Ripk3^–/– mice, we failed to observe a rescue of the enhanced lung tumor phenotype in KRas^G12D;Hace1^–/– mice (Supplementary Fig. S1A). Thus, loss of Hace1 increases the susceptibility to KRas^G12D-driven lung tumors independently of pronecroptotic RIP3 kinase.

HACE1 was previously shown to ubiquitylate and mediate degradation of the small GTPase RAC1, predominantly in its active, GTP-bound form (17, 33). Deregulation of RAC1 has previously been associated with tumor development (29). To examine if increased tumorigenesis in the absence of HACE1 could be explained by deregulation of RAC1, we crossed KRas^G12D;Hace1^–/– mice to Rac1^fl/fl mice (48) to generate KRas^G12D;Rac1^fl/fl and KRas^G12D;Hace1^–/–Rac1^fl/fl mice, in which deletion of Rac1 and expression of oncogenic KRas^G12D can be simultaneously induced by Adeno-Cre administration (28). Rac1 inactivation in lung tumor cells but not in adjacent tissue and lung macrophages was confirmed by IHC, Western blot, and transcript analysis (Supplementary Fig. S2A–S2D). In line with the survival data in Fig. 2B, deletion of Hace1 resulted in a significantly higher tumor burden at weeks 8 and 16 after lung tumor induction compared with KRas^G12D;Hace1^+/+Rac1^+/+ control mice. Genetic inactivation of Rac1 alone led to significantly reduced tumor burden compared with KRas^G12D;Hace1^+/+Rac1^+/+ control animals (Fig. 2C and D), confirming previous data (28). Eight weeks after lung tumor induction, KRas^G12D;Hace1^–/–Rac1^fl/fl mice, similar to KRas^G12D;Rac1^fl/fl mice, had strongly reduced tumor growth as compared with both KRas^G12D;Hace1^–/– and KRas^G12D;Hace1^+/+Rac1^+/+ control mice. After 16 weeks, KRas^G12D;Hace1^–/–Rac1^fl/fl mice still had a significant reduction in tumor burden compared with KRas^G12D;Hace1^–/– animals, but this reduction was not as pronounced as in the KRas^G12D;Rac1^fl/fl mice (Fig. 2C and D).

We further assessed tumor initiation and progression using established pathology criteria (49, 50). At week 8 after lung tumor induction, we observed foci of hyperplasia and adenomas (benign lesions), irrespective of the genotype (Fig. 2E). Foci of atypical adenomatous hyperplasia (preneoplastic/preinvasive lesions) were more prominent in KRas^G12D;Hace1^–/– mice. Borderline minimally invasive adenocarcinomas (malignant lesions) were also more evident in KRas^G12D;Hace1^–/– and KRas^G12D;Hace1^+/+Rac1^+/+ control mice. Mice lacking RAC1 either in the presence or absence of HACE1 showed notably reduced numbers of preinvasive and malignant lesions. Genetic inactivation of Hace1 resulted in a higher number of tumor foci (Fig. 2E), supporting the notion that Hace1 loss leads to more malignant and aggressive type of lung tumors. Together, these data support a role for RAC1 in the progression of Hace1-deficient KRas^G12D-driven lung tumors.

To further characterize tumor progression, we assessed the proliferation index of the lung tumors at weeks 8 and 16 after lung tumor induction by Ki67 immunostaining. In line with tumor growth, proliferation of lung tumor cells in KRas^G12D;Hace1^–/– mice was elevated compared with the other groups, whereas, consistently, Rac1 deletion led to reduced tumor cell proliferation (Fig. 3A, top plots, and B). In KRas^G12D;Hace1^–/–Rac1^fl/fl mice, the number of Ki67⁺ lung tumor cells was lower than in KRas^G12D;Hace1^–/– mice and comparable with those in KRas^G12D;Hace1^+/+Rac1^+/+ control animals. Furthermore, we assessed DNA damage in lung tumor cells by γH2AX immunostaining. Consistent with accelerated tumorigenesis, we observed increased DNA damage in tumor foci of KRas^G12D;Hace1^–/– mice at weeks 8 and 16 after lung tumor induction, whereas this was reduced in KRas^G12D;Hace1^–/–Rac1^fl/fl tumors to levels detected in KRas^G12D;Rac1^fl/fl and KRas^G12D;Hace1^+/+Rac1^+/+ control mice (Fig. 3A, bottom plots, and C). These data show that the enhanced proliferation as well as increased DNA damage observed in Hace1^–/– lung tumors is reduced upon additional inactivation of Rac1.

Previous in vitro findings have shown that deregulated activation of RAC1 in the absence of HACE1 is responsible for increased production of ROS by the NADPH oxidase pathway (17). Given the important impact of ROS in driving cellular damage and cancer development (51), we performed DHE staining to evaluate ROS generation in lung tumor cells. We indeed observed elevated ROS levels in KRas^G12D;Hace1^–/– mice compared with the other groups (Fig. 4A and B). These findings were supported by staining with additional oxidative stress markers, NRF2 and 4hydroxynonenal (4HNE; Supplementary Fig. S2E and S2F; ref. 52). To test whether absence of HACE1 would lead to the accumulation of active GTP-RAC1, we isolated primary lung tumor cells from KRas^G12D;Hace1^–/– and KRas^G12D;Hace1^+/+ control mice at week 12 after lung tumor induction and measured activation of RAC1 by Western blot after affinity precipitation using the GST-tagged GTP-RAC1 binding motif of PAK (40). The fraction of active GTP-bound RAC1 was markedly increased in KRas^G12D;Hace1^–/– tumor cells compared with tumor cells from KRas^G12D;Hace1^+/+ control mice (Fig. 4C). These results indicate that lung tumors of KRas^G12D;Hace1^–/– have increased levels of active RAC1 and exhibit enhanced ROS production.

Although KRas^G12D;Hace1^–/–Rac1^fl/fl mice displayed a reduced lung tumor burden as compared with KRas^G12D;Hace1^–/– mice, this reduction was not as pronounced as in KRas^G12D;Rac1^fl/fl mice (Fig. 2C and D), indicating that, in addition to RAC1, another factor must be involved in the increased tumorigenesis of Hace1-deficient animals. The RAC family of GTPases encompasses three very closely related structural and functional homologs: RAC1, RAC2, and RAC3 (Fig. 5A). Rac2 mRNA was also expressed in lung tumor cells, albeit at lower levels compared with Rac1, whereas Rac3 was expressed at low levels (Fig. 5B, and Supplementary Fig. S2C and S2D). Given the expression pattern and homology, we hypothesized that RAC2 might compensate for the loss of RAC1 and thereby drive tumor formation in KRas^G12D;Hace1^–/–Rac1^fl/fl mice. Importantly, we confirmed that HACE1 ubiquitylates GTP-bound but not GDP-bound RAC2 (Fig. 5C), indicating that HACE1 might directly contribute to the regulation of the activation state of RAC2, similar to RAC1. Subsequently, we investigated the involvement of RAC2 in KRas^G12D-driven lung tumorigenesis in vivo.

To examine the role of RAC2, we crossed Rac2^–/– as well as Rac1^fl/fl mice onto the KRas^G12D;Hace1^–/– background to generate Hace1, Rac1, and Rac2 triple mutants, plus all relevant genetic cohort controls. The survival of lung tumor–bearing KRas^G12D;Hace1^–/–Rac1^fl/fl mice with a mean survival time of 176 days was enhanced compared with KRas^G12D;Hace1^–/– mice with a mean survival time of 127 days (P < 0.0001, log-rank test) and similar to KRas^G12D;Hace1^+/+Rac1^+/+Rac2^+/+ control animals with a mean survival time of 154 days (P = 0.7122, log-rank test), but still less compared with KRas^G12D;Rac1^fl/fl animals with a mean survival time of 267 days (P < 0.0001, log-rank test; Fig. 5D), corroborating the tumor burden analysis shown in Fig. 2C and D. KRas^G12D;Rac2^–/– mice (mean survival time 136 days) showed no significant survival advantage compared with KRas^G12D;Hace1^+/+Rac1^+/+Rac2^+/+ control animals (P = 0.2422, log-rank test) after induction of lung tumors (Fig. 5D). KRas^G12D;Hace1^–/–Rac2^–/–mice (mean survival time 115 days) also did not show any difference in survival compared with KRas^G12D;Hace1^–/– mice (P = 0.2698, log-rank test). The survival of KRas^G12D;Rac1^fl/flRac2^–/– mice (mean survival time 245 days) was comparable with KRas^G12D;Rac1^fl/fl mice (P = 0.3037, log-rank test). These data indicated that RAC2 has no important role in the progression of KRas^G12D-driven lung tumor and that, in these genetic scenarios, it cannot compensate for the absence of RAC1. However, the genetic inactivation of Rac2 intriguingly did lead to markedly prolonged survival when both Rac1 and Hace1 were knocked-out: KRas^G12D;Hace1^–/–Rac1^fl/flRac2^–/– triple knock-out mice (mean survival time 225 days) survived much longer compared with KRas^G12D;Hace1^–/–Rac1^fl/fl mice (P = 0.0017, log-rank test) and almost reached the survival of KRas^G12D;Rac1^fl/fl mice (P = 0.0019, log-rank test; Fig. 5D). In line with the enhanced survival, KRas^G12D;Rac1^fl/fl, KRas^G12D;Rac1^fl/flRac2^–/– and KRas^G12D;Hace1^–/–Rac1^fl/flRac2^–/– mice exhibited a significant reduction in tumor burden compared with KRas^G12D;Hace1^+/+Rac1^+/+Rac2^+/+ control animals (Fig. 5E and F). We also observed decreased tumor burden in KRas^G12D;Hace1^–/–Rac1^fl/fl mice, albeit this was statistically not significant (Fig. 5E and F), in line with the survival curves of this genetic cohort (Fig. 5D). Expression of Ki67 by tumor cells did not significantly differ among the different groups (Supplementary Fig. S3A). Similarly, infiltration of tumors by different immune cells was largely unaffected across the various genotypes (Supplementary Fig. S3B–S3E). Of note, KRas^G12D;Hace1^–/– mice were not included in this tumor burden analysis because at the time point of analysis half of these mice had already died. Taken together, these data show that concurrent loss of Rac2 reduces the tumor growth of KRas^G12D;Hace1^–/–Rac1^fl/fl mice. Thus, RAC2 appears—at least in part—to compensate for the loss of RAC1 in driving lung cancer development in KRas^G12D;Hace1^–/– mice.

To further investigate the role of HACE1, RAC1, and RAC2 in human lung cancer, we stratified patients with lung adenocarcinoma according to the mRNA expression of HACE1, RAC1, and RAC2 using TCGA RNA-seq datasets (53). These data revealed that patients with lung adenocarcinoma showing high expression of HACE1 and low expression of RAC1 had a significantly better disease-free as well as a better overall survival prognosis than patients with high expression of RAC1 and low expression of HACE1 (Fig. 6A and B). High expression of both RAC1 and RAC2 significantly correlated with even poorer survival as compared with high expression of RAC1 alone (Fig. 6C and D). Thus, deregulated expression of the HACE1/RAC pathway is associated with the prognosis of patients with lung adenocarcinoma.

We next analyzed the mutation frequency of HACE1 and RAC-family members in 507 cases of human lung adenocarcinoma (Fig. 6E). We observed in 9.1% of the adenocarcinoma cases genetic alterations in the RAC and HACE1 genes (Fig. 6E). The most abundant HACE1 alterations detected in patients with lung adenocarcinoma were deletions, missense mutations, and truncations. Conversely, the predominant genetic alterations for RAC1, RAC2, or RAC3 were amplifications (Fig. 6E). In addition, the mRNA expression level of HACE1 showed a significant negative correlation to that of RAC1 or RAC2 in patients with lung cancer, whereas it did not correlate with RAC3 expression, i.e., high HACE1 mRNA expression correlates with reduced expression of RAC1/2 and vice versa (Fig. 6F and G). Thus, HACE1 is mutated in a subset of patients with lung adenocarcinoma. High mRNA expression of HACE1 has positive prognostic value, whereas RAC1 and RAC2 high mRNA expressions are associated with a poor disease-free and overall survival in patients with lung cancer.

Here, we report that the HECT-family E3 ligase HACE1 is mutated in multiple types of cancers, including lung tumors. Importantly, we provide direct genetic evidence that HACE1 regulates KRas^G12D-dependent lung tumor formation via RAC-family GTPases.

Our in vivo and in vitro data demonstrate that lung tumorigenesis in Hace1-deficient mice is associated with elevated RAC1 activity and enhanced ROS levels. We did not observe any significant effect of RAC2 on lung tumorigenesis while RAC1 was still expressed. However, we found that RAC2 promotes lung tumorigenesis in a scenario where both Rac1 and Hace1 were ablated. These data indicate that RAC2 normally only plays a marginal role in driving pneumocyte transformation and tumorigenesis, probably due to much lower expression levels compared with RAC1. In a context in which the more abundant homolog Rac1 is ablated, RAC2 can drive lung tumor development. However, this compensation only occurs if HACE1 is also absent, indicating that HACE1 can effectively also dampen the potential oncogenic function of RAC2. Indeed, we found that GTP-bound RAC2, similar to RAC1, is ubiquitylated by HACE1, implying an important function of HACE1 as a general “molecular brake” of RAC GTPase activity. Interestingly, the presence of RAC2 did not appear to affect the early stages of tumor growth in KRas^G12D;Hace1^–/–Rac1^fl/fl mice, and these animals only showed increased tumor burdens compared with KRas^G12D;Rac1^fl/fl mice at later time points. This might be indicative of a role of RAC2 in the promotion of tumor growth rather than tumorigenesis in these animals. Although RAC1 and RAC2 seem to be the main drivers of lung tumorigenesis in Hace1-deficient mice, KRas^G12D;Hace1^–/–Rac1^fl/flRac2^–/– triple knock-out mice displayed a slightly reduced survival compared with KRas^G12D;Rac1^fl/fl mice. Therefore, we hypothesize that additional factors (possibly the third homolog RAC3) could also participate in lung tumorigenesis in the absence of HACE1, RAC1, and RAC2. Whether HACE1 also regulates RAC3 needs to be determined in future experiments.

Loss of Hace1 results in lung tumor formation, which is associated with enhanced oxidase-dependent ROS generation. Human melanoma cells harboring mutations of HACE1 also displayed increased ROS levels. Maintenance of redox balance is crucial for the survival and functionality of cells, and intracellular ROS are crucial for maintaining a variety of homeostatic processes, such as proliferation, survival, metabolism, and differentiation (54). The accumulation of ROS also results in oxidative stress, causing genomic instability, mutagenesis, and cell transformation. Active GTP-RAC1 as well as active RAC2 can directly promote ROS generation through NADPH oxidase (24). HACE1 has been previously shown in in vitro experiments to dampen NADPH oxidase–dependent ROS generation by inducing proteasomal degradation of GTP-RAC1 (17). In the present study, we show that, besides RAC1, HACE1 also interacts with and inhibits the oncogenic function of RAC2. Therefore, we postulate that a key function of HACE1 in oncogenesis is to control RAC-family GTPases and consequently NADPH oxidase–dependent ROS generation (Fig. 7). However, our data do not exclude the possibility that generation of ROS might be controlled by RAC1 via NADPH oxidase–independent mechanisms. Importantly, in line with enhanced ROS, we also observed increased DNA damage as determined by γH2AX immunostaining.

To assess whether increased oxidative stress might sustain tumor development in the absence of HACE1, we induced lung cancer in KRas^G12D;Hace1^–/– and control KRas^G12D;Hace1^+/+ mice and treated them with ROS scavenger NAC. Strikingly, NAC-treated KRas^G12D;Hace1^–/– mice showed significantly reduced tumor burdens compared with untreated KRas^G12D;Hace1^–/– animals (Supplementary Fig. S4A and S4B). Interestingly, treatment of NAC was previously suggested to accelerate the development of lung tumors (55), whereas our control KRas^G12D;Hace1^+/+ mice were unaffected by NAC treatment. Although we are unsure about the reason for this discrepancy, differences in the genetic background of the animals (the previous study used mixed Sv129/C57BL/6, whereas our mice were on C57BL/6 background) might affect the absorption and metabolization of NAC, leading to different amounts of available NAC and accordingly a different impact on the redox balance of tumor cells. Our data indicate that elevated ROS levels contribute to increased tumorigenesis in Hace1-deficient animals, although we cannot exclude that ROS-independent functions of deregulated RAC1 might also promote tumor development in KRas^G12D;Hace1^–/– mice.

Although increased amounts of ROS might sustain tumorigenesis by promoting genomic instability, excessive amounts of intracellular ROS are highly cytotoxic and induce apoptosis (56). Elevated ROS levels in the absence of HACE1 coupled with the impaired capacity of Hace1-deficent cells to activate crucial antioxidant mediator NRF2 (34) open the possibility for therapeutic opportunities. We therefore hypothesized that treatment with ROS-inducing agents might specifically induce apoptosis in Hace1-deficient cells by further increasing intracellular ROS to cytotoxic levels. In Hace1^–/–MEFs, treatment with the ROS-inducer PL (57) indeed increased cell death, p53 phosphorylation, and cyclin D1 expression compared with control cells and untreated cultures (Supplementary Fig. S4C and S4D). This effect was reversed upon re-expression of HACE1 in Hace1^–/– MEFs. Treatment with 2-acetylphenothiazine (ML171), which specifically inhibits NOX1-containing NADPH oxidase complexes (58), prevented cell death in PL-treated Hace1^–/– MEFs. Furthermore, H₂O₂ challenge reduced the clonogenic survival of Hace1^–/– MEFs in a concentration-dependent manner compared with untreated and control MEFs (Supplementary Fig. S4E). We also observed enhanced cell death in human embryonic kidney (HEK) 293 cells upon knockdown of HACE1 expression when treated with either PL or H₂O₂ (Supplementary Fig. S4F). Thus, further induction of oxidative stress upon treatment with ROS inducers destabilized Hace1-deficient cells to trigger cell-cycle arrest and apoptosis, suggesting that ROS-inducing agents could be employed to therapeutically enhance cellular ROS and preferentially eliminate Hace1-mutated cancer cells while sparing healthy cells.

Our findings complement and highlight the role of HACE1 as a tumor suppressor in lung cancer. Mechanistically, it appears that HACE1 ubiquitylates active RAC proteins, leading to their proteasomal degradation and the subsequent impaired activation of the NADPH oxidase complex. Hence, by inhibiting ROS generation, HACE1 prevents DNA damage and tumor development. Importantly, our data also position the HACE1–RAC axes as a key driver of human lung cancer pathogenesis associated with patient survival.

D.A. Williams disclosed commercial research funding from Bluebird Bio and Novartis. No potential conflicts of interest were disclosed by the other authors.

Conception and design: M. Kogler, L. Tortola, J.M. Penninger

Development of methodology: M. Kogler, L. Tortola, A.M. El-Naggar, S. Mereiter, C. Gomez-Diaz, S. Rao, J.M. Trent, F. Ikeda, M. Daugaard, P.H. Sorensen

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M. Kogler, L. Tortola, A.M. El-Naggar, C. Gomez-Diaz, B.V. Gapp, D. Hoffmann, J.M. Trent, F. Ikeda, M. Daugaard

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M. Kogler, L. Tortola, G.L. Negri, C. Gomez-Diaz, D. Tortora, A.M. Kavirayani, M. Novatchkova, F. Ikeda

Writing, review, and/or revision of the manuscript: M. Kogler, L. Tortola, G.L. Negri, D. Tortora, M. Daugaard, P.H. Sorensen

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): R. Nitsch, D. Tortora

Study supervision: L. Tortola, M. Daugaard, J.M. Penninger

Others (performed analyses and interpretation of lung pathology data; contributed figures and text relevant to histopathology data; coordinated certain aspects of digital whole slide image analysis): A.M. Kavirayani

Others (performed experiments): D. Cikes, A. Hagelkruys

We thank all members of our laboratories for constructive discussions and expert advice. We also thank A. Kavirayani, M. Zeba, T. Engelmaier, J. Klughofer, A. Mancebo Gimenez, and A. Piszczek for histology services, and the members of the IMP-IMBA Biooptics facility, especially G. Schmauss and T. Lendl, for assistance in image quantification. We thank Lilian M. Fennell for the help with the ubiquitylation assay. L. Tortola is supported by the Swiss National Science Foundation (Grant#: PBEZP3_145993). S. Mereiter is supported by funding from the European Union's Horizon 2020 research and innovation program under the Marie Skłodowska-Curie grant agreement (Grant# 841319). I. Uribesalgo is supported by an EMBO Longterm Fellowship and a Marie Curie Fellowship from the European Commission. J.M. Penninger is supported by grants from IMBA, the Austrian Ministry of Sciences, the Austrian Science Fund (Grant# Z 271-B19), the Austrian Academy of Sciences, a European Research Council (ERC) Advanced Grant, an Era of Hope Innovator award, and a Canada 150 Grant. M. Daugaard is supported by the Canadian Institutes of Health Research (CIHR, Grant# 377771). P.H. Sorensen is supported by CIHR Foundation (Grant# FDN-143280) and by the British Columbia Cancer Foundation through generous donations from Team Finn and other riders in the Ride to Conquer Cancer.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.