BAP1 Downregulates NRF2 Target Genes and Exerts Anti-Tumorigenic Effects by Deubiquitinating KEAP1 in Lung Adenocarcinoma

KELCH-ECH-associated protein 1 (KEAP1) is an adaptor protein of Cullin 3 (CUL3) E3 ubiquitin ligase that targets a redox sensitive transcription factor, NF-E2-related factor 2 (NRF2). BRCA1-associated protein 1 (BAP1) is a tumor suppressor and deubiquitinase whose mutations increase the risk of several types of familial cancers. In the present study, we have identified that BAP1 deubiquitinates KEAP1 by binding to the BTB domain. Lentiviral transduction of BAP1 decreased the expression of NRF2 target genes, suppressed the migration and invasion, and sensitized cisplatin-induced apoptosis in human lung adenocarcinoma (LUAD) A549 cells. Examination of the lung tissues in KrasG12D/+ mice demonstrated that the level of Bap1 and Keap1 mRNAs progressively decreases during lung tumor progression, and it is correlated with NRF2 activation and the inhibition of oxidative stress. Supporting this observation, lentiviral transduction of BAP1 decreased the growth of A549 xenografts in athymic nude mice. Transcriptome analysis of human lung tissues showed that the levels of Bap1 mRNA are significantly higher in normal samples than LUAD samples. Moreover, the expression of Bap1 mRNA is associated with a better survival of LUAD patients. Together, our study demonstrates that KEAP1 deubiquitination by BAP1 is novel tumor suppressive mechanism of LUAD.


Introduction
NF-E2-related factor 2 (NRF2) is a transcription factor that plays an important role in the detoxification of reactive oxygen species (ROS) by transcriptional activation of phase II cytoprotective enzymes [1]. Under basal conditions, NRF2 is located in the cytosol and constantly targeted for polyubiquitination by KELCH-like ECH-associated protein 1 (KEAP1), an adaptor protein for Cullin 3 (CUL3) E3 ubiquitin ligase [2]. Exposure of oxidative insults halts polyubiquitination of NRF2, which causes NRF2 to translocate into the nucleus and initiates transcriptional activation of phase II cytoprotective enzymes by binding to the antioxidant response element (ARE), a cis-acting motif sequence existing in the promoter of NRF2 target genes [3]. While NRF2 activation has been regarded as a target for chemoprevention and treatment of proinflammatory diseases [4], the studies DUB cDNAs were amplified from various human cell lines by RT-PCR and subcloned into pcDNA3 expression vectors. Human Nrf2, Keap1, and ubiquitin cDNAs were amplified from various cell lines. Full-length sequence of human cDNAs was confirmed by DNA sequencing. Site-directed mutagenesis and the deletion of individual domains were conducted using overlapping PCR technique. pcDNA3-Myc-Ubiquitin, pcDNA3-HA-NRF2, and pcDNA3-HA-KEAP1 plasmids were cloned using In-Fusion ® HD Cloning Plus kit (Takara Korea Biomedical Inc., Seoul, Korea). DNA oligonucleotides were purchased from Macrogen (Seoul, Korea).

Immunoprecipitation
For immunoprecipitation of epitope-tagged plasmids, confluent 293T cells were cotransfected with FLAG-tagged and HA-tagged pcDNA3 vectors for 48 h using Jet-PEI reagent. Then, 293T cells were lyzed with 1 mL NP-40 buffer (50 mM Tris-HCl at pH 7.5, 150 mM NaCl, 0.5% NP-40, 50 mM NaF, protease inhibitors) for 1 h on ice and cell lysates were incubated with anti-FLAG magnetic beads or anti-HA magnetic beads overnight at 4 • C. Magnetic beads were washed with NP-40 buffer three times and denatured in 2× sample buffer. Eluted samples were resolved by SDS-PAGE and analyzed by Western blot.
For immunoprecipitation of endogenous BAP1, confluent 293T cells were lyzed with 1 mL NP-40 buffer (50 mM Tris-HCl at pH7.5, 150 mM NaCl, 0.5% NsdP-40, 50 mM NaF, protease inhibitors) for 1 h on ice, and cell lysates were incubated with mouse IgG or anti-BAP1 antibody overnight at 4 • C. Protein A/G magnetic beads were added to the tubes and incubated for 4 h at room temperature. Protein A/G magnetic beads were washed with NP-40 buffer three times and denatured in 2× protein sample buffer. Eluted samples were resolved by SDS-PAGE and analyzed by Western blot analysis.

Fractionation of the Nucleus and the Cytosol
After lentiviral transduction of FLAG-BAP1, A549 cells were lysed with 1mL NP-40 buffer (50 mM Tris-HCl at pH 7.5, 150 mM NaCl, 0.5% NP-40, 50 mM NaF, protease inhibitors) for 1 h on ice. Lysates were centrifuged at 15,000 rpm for 10 min and supernatant was collected as a cytosolic fraction. After washing the remnant pellets three times with NP-40 buffer, the pellets were resuspended in NP-40 buffer and heavily sonicated. Lysates were centrifuged at 15,000 rpm for 15 min and the supernatant was collected as the nuclear fraction. Both nuclear and cytosolic fractions were subjected to Western blot analysis.

Real-time Reverse Transcription-Polymerase Chain Reaction (RT-PCR)
Total RNA of A549 cells and mouse lung tissues was extracted by Hybrid-R RNA extraction kit (GeneAll, Seoul, Korea). A total of 1 µg of total RNA was subject to cDNA synthesis, using PrimeScript RT-PCR kit (TAKARA Korea, Seoul, Korea). The real-time RT-PCR analysis was performed using SYBR mix (ELPIS Biotech. Daejeon, Korea) on CFX384 real-time system as recommended by manufacturer (BioRad, Hercules, CA, USA). The mRNA level of individual genes was normalized by that of GAPDH. The sequence of RT-PCR primers are listed in Table 1.

Wound Healing Assay
A549 cells (2.0 × 10 5 cells/well) and A549-FLAG-BAP1 cells (2.0 × 10 5 cells/well) were seeded in 24-well plates and a straight wound was created by scratching with a yellow pipette tip. Cells were cultured in serum-free medium and allowed to migrate into the wound area and the images of cell movement were obtained using the Eclipse Ti-U inverted microscope (Nikon, Tokyo, Japan). Wounded areas were calculated using NIS Elements F software (Nikon, Tokyo, Japan).

Cell Migration Assay
Migration assays were performed using transwell inserts (Neuro Probe Inc., Gaithersburg, MD, USA). After lower surface of transwell inserts was immersed in fibronectin solution (10 µg/mL) overnight, A549 and A549-FLAG-BAP1 cells (5 × 10 5 cells/mL) suspended in serum-free medium were added to the upper chamber of each inserts and the lower chambers were filled up with medium containing 3% FBS. After incubation for 6 h, cells that failed to migrate were scraped off from the upper surface of the membrane and those that reached the lower surface were stained by the Diff Quik staining kit (Sysmex, Kobe, Japan). The image of cells in the lower chamber was observed and captured using the Eclipse Ti-U inverted microscope (Nikon, Tokyo, Japan), and the number of migrated cells was manually counted using the computer.

TUNEL Assay
TUNEL assay was conducted using DeadEND TM fluorometric TUNEL system kit (Promega, Madison, WI, USA) as recommended by the manufacturer.

Animal Experiments
Six-week-old Balb/c nude mice were purchased from Daehan Biolink Co. (Eumseong, Korea) and used for xenografts. After a week acclimation, 12 mice were divided into two groups of six mice, and they were subcutaneously injected into the dorsal flank with A549 cells (5 × 10 6 cells/0.2 mL DMEM media) or A549-FLAG-BAP1 cells (5 × 10 6 cells/0.2 mL DMEM media). The body weights of mice were measured every three days during the course of study. After 18 days, mice were sacrificed by asphyxiation with CO 2 and the weight of xenografts was measured.
Kras G12D/+ mice that can form lung tumors by spontaneous recombination events [16] were obtained from Dr. Tyler Jacks (Massachusetts Institute of Technology, Cambridge, MA, USA). Kras G12D/+ mice were housed in the sterile filter-capped microisolator cages and provided with water and diet ad libitum. After Kras G12D/+ mice were bred for 1 week, 8 weeks, 24 weeks, and 40 weeks, they were sacrificed, and the lung tissues were obtained. The lung tissues were stored in the deep freezer for Western blot and RT-PCR or in 10% paraformaldehyde for hematoxylin and eosin staining and immunohistochemistry.

Hematoxylin and Eosin (H&E) Staining and Immunohistochemistry (IHC)
Dehydration of the lung tissues in 10% paraformaldehyde was performed by serially immersing the tissues into 75%, 80%, 85%, 90%, 95%, and 100% ethanol, and xylene solution for 1 h at each step. Dehydrated tissues were embedded in the paraffin block. Paraffin-embedded tissues were sectioned at the length of 5 µm, mounted on the slide, and deparaffinized. To conduct the H/E staining, the tissues were stained with Mayer's hematoxylin solution for 5 min at room temperature and rinsed in tap water until the water became clear. In the bluing step, the tissues were stained with repeated cycles of eosin Y ethanol solution for 70 s, 5 dips in 95% ethanol, and 5 dips in 100% ethanol at room temperature. The tissues were rinsed with distilled water and the images were taken using the Eclipse Ti-U inverted microscope (Nikon, Tokyo, Japan).
To conduct the IHC, the paraffin-embedded tissues were sectioned at the length of 5 µm, mounted on the slide, and deparaffinized. Tissues on the slide were heated with citrate buffer (pH 6.0) in the microwave for 10 min and the slides were blocked by blocking solution (ScyTek Inc., Logan, UT, USA) for 30 min. After washing three times with 1× PBS, the slides were incubated with primary antibodies (1:200) overnight at 4 • C. After washing three times with 1× PBST, the slides were incubated with antimouse UltraTEk HRP antibodies (ScyTek Inc., Logan, UT, USA), and development of the slides was performed with DAB kit (GBI Labs, Bothell, WA, USA). The slides were sealed with a mounting medium and the images were captured using the Eclipse Ti-U inverted microscope (Nikon, Tokyo, Japan). The percentage of stained area under the bright field was monitored by the ImageJ software (National Institute of Health, Bethesda, MD, USA).

Statistical Comparison of A549 Cells and A549 Tumor Xenografts
Student's t-test was used to statistically analyze the difference in cell line experiments and A549 tumor xenograft experiments. Asterisks indicate statistical a significance with * p < 0.05, ** p < 0.01, and *** p < 0.001.

Comparison of Bap1 mRNA Expression between LUAD and Normal Lung Tissues of Humans
The levels of Bap1 mRNA expression (Transcripts Per kilobase Million, TPM) from RNA-sequencing (RNA-seq) data of 513 LUAD samples and 288 normal lung samples were obtained from the UCSC Xena [17], where the RNA-seq data of the TCGA and the GTEx were reprocessed by using a uniform RNA-seq analysis pipeline (alignment to hg38 genome and quantification using RSEM). Student's t-test was applied to determine the difference in mean expression levels of Bap1 mRNA between tumors and normal samples.

Survival Analysis of LUAD Cancer Patients
Transcriptome and clinical data of LUAD and non-small cell lung cancer (NSCLC) patients were obtained from the TCGA and the Gene Expression Omnibus (GEO) databases, respectively. In TCGA data, RNA-seq profiles and clinical information were downloaded via the R package "TCGA biolinks", and TPM was taken as the gene expression level. In each cohort, patients were divided into two groups, high and low, based on whether their expression level of BAP1 was higher or lower than the median level. Survival analysis was performed to test the difference in overall survival (OS) rates between the high and low groups via the R package "survival". Hazard ratio (HR) and p-value (P) were calculated using Cox proportional hazards regression analysis and log-rank test, respectively.

BAP1 Stabilizes KEAP1 by Binding to the BTB Domain
Because BAP1 bound to the BTB domain ( Figure 2A) and promoted deubiquitination of KEAP1 ( Figure 1E), we assumed that the BTB domain might be responsible for maintaining the protein stability of KEAP1 by BAP1. To address this issue, we generated 293T-HA-KEAP1 and 293T-HA-KEAP1∆BTB cells by stable transfection of pcDNA3-puro-HA-KEAP1 and pcDNA3-puro-HA-KEAP1∆BTB plasmids followed by selection with puromycin. Established 293T-HA-KEAP1 and 293T-HA-KEAP1∆BTB cells were transfected with pcDNA3-FLAG-BAP1 and exposed to cycloheximide (CHX) to inhibit de novo protein synthesis. Our results show that treatment of CHX decreased the level of HA-KEAP1, and this event was significantly attenuated by overexpression of FLAG-BAP1 in 293T-HA-KEAP1 cells ( Figure 2C, left panel). However, FLAG-BAP1 failed to attenuate a decrease in the level of HA-KEAP1∆BTB induced by treatment of CHX in 293T-HA-KEAP1∆BTB cells ( Figure 2C, right panel), suggesting that the BTB domain is responsible for maintaining the protein stability of KEAP1 by BAP1.

Overexpression of BAP1 Inhibits NRF2, Suppresses Migration and Invasion, and Promotes Cisplatin-Induced Apoptosis of A549 Cells
Aberrant NRF2 activity is frequently observed in human LUAD due to loss of function mutations in Keap1 [20]. Thus, we examined whether BAP1 can affect KEAP1 to modulate NRF2 target gene expression in human LUAD A549 cells. To address this issue, we generated stable A549-FLAG-BAP1 cells by lentiviral transduction and fractionated them into the nucleus and the cytosol. Our results show that transduced FLAG-BAP1 is mostly located in the cytosol ( Figure 3A). We also observed that A549-FLAG-BAP1 cells exhibited a lower level of ARE-luciferase activity compared with A549 cells ( Figure 3B). Transfection of pcDNA3-FLAG-BAP1 increased the level of KEAP1 and decreased the level of NRF2 in A549 cells ( Figure 3C). Real-time RT-PCR analysis demonstrates that lentiviral transduction of FLAG-BAP1 significantly decreased the mRNA level of NRF2 target genes such as heme oxygenase-1 (Hmox1), NADPH:quinone oxidoreductase-1 (Nqo1), glutamate cysteine ligase catalytic subunit (Gclc), and aldo-keto reductase family 1 member B10 (Akr1b10) in A549 cells ( Figure 3D). In addition, lentiviral transduction of FLAG-BAP1 suppressed the migration ( Figure 3E) and the invasion ( Figure 3F) of A549 cells, and potentiated cisplatin-induced apoptosis in A549 cells ( Figure 3G,H). Together, these results illustrate that BAP1 lowers the level of NRF2 target genes by deubiquitinating and stabilizing KEAP1, thereby inhibiting the migration and invasion of A549 cells, and sensitizing A549 cells to cisplatin-induced apoptosis.  Direct inhibition of KRAS oncoprotein is a difficult task and, therefore, targeting Kras-mutant lung cancer remains a major challenge [21]. It is known that oncogenic KRAS stimulates transformation [22] and lung cancer cell proliferation, at least in part, by induction of NRF2 [23]. However, the molecular mechanisms underlying how oncogenic KRAS activates NRF2 are still elusive. We hypothesized that oncogenic KRAS might activate NRF2, at least in part, by downregulating BAP1 and KEAP1. To examine this hypothesis, we compared the levels of Bap1, Keap1, Nrf2, and Hmox1 mRNAs in the lung tissues of Kras G12D/+ mice with those of wild-type littermates by real-time RT-PCR analysis. We found that the level of Bap1 and Keap1 mRNAs in the lung tissues of Kras G12D/+ mice progressively decreased from 8 weeks to 40 weeks, compared with those of wild-type littermates ( Figure 4A). Consistent with this, the relative mRNA level of Hmox1, a target gene of NRF2, progressively increased from 24 weeks to 40 weeks in the lung tissues of Kras G12D/+ mice ( Figure 4A). However, the relative level of Nrf2 mRNA in the lung tissues of Kras G12D/+ mice was not different from that of wild-type littermates ( Figure 4A), supporting that the level of NRF2 is regulated mostly by CUL3/KEAP1mediated polyubiquitination [24]. Western blot analysis shows that the levels of BAP1 and KEAP1 were significantly lower in the lung tissues of Kras G12D/+ mice at 40 weeks, when compared with those of wild-type littermates ( Figure 4B). Immunohistochemistry results demonstrate that the levels of BAP1, KEAP1, and 8-hydroxydeoxyguanosine (8-OH-dG), an oxidative damage marker, significantly decreased in the lung tumors of Kras G12D/+ mice, compared with normal tissues in the surrounding area ( Figure 4C). Together, these results demonstrate that KRAS activation downregulates BAP1 and KEAP1, and this event is correlated with NRF2 activation and a decrease in oxidative stress in vivo. Antioxidants 2022, 11, x FOR PEER REVIEW 13 of 18

BAP1 Suppresses the Growth of A549 Xenografts, and the Expression of Bap1 Is Higher in Normal Lung Tissues and Correlated with a Better Survival of LUAD Patients
We have demonstrated that BAP1 inhibits the expression of NRF2 target genes ( Figure 3D) in A549 cells by promoting KEAP1 deubiquitination ( Figure 1E). To examine whether BAP1 can affect the growth of A549 cells in vivo, A549 and A549-FLAG-BAP1 cells were injected into athymic nude mice, and the growth of xenografts was monitored. While athymic nude mice bearing A549 cells and A549-FLAG-BAP1 cells exhibited no difference in the body weight ( Figure 5A), we observed that the mice bearing A549-FLAG-BAP1 produced fewer and smaller xenografts at sacrifice ( Figure 5B). To examine whether BAP1 might behave as a tumor suppressor of LUAD, we retrieved the RNA-seq data of LUAD and normal lung tissues in humans from The Cancer Genome Atlas (TCGA) and the Genotype-Tissue Expression (GTEx) projects, respectively, and revealed that the levels of Bap1 mRNA expression were significantly higher in normal samples than in LUAD patients ( Figure 5C). In addition, the mRNA level of Bap1 in LUAD patients was positively correlated with an increase in the overall survival ( Figure 5D, left panel). It is known that cisplatin is a chemotherapeutic agent which is mostly administered to LUAD patients at stages II-V, but not at stage I [25]. Because BAP1 sensitized A549 cells to cisplatin-induced apoptosis ( Figure 3G,H), we subdivided LUAD patients into stage I and stages II-IV and investigated the association between Bap1 mRNA expression and the overall survival rate. Our results showed that the level of Bap1 mRNA was positively correlated with a better survival of LUAD patients at stages II-IV, but not at stage I ( Figure 5D). This provides a feasible possibility that BAP1 might provide a clinical benefit to LUAD patients at higher tumor stages.

Discussion
Inactivating mutations or truncation of Bap1 frequently occur in familial cancers such as uveal melanoma, mesothelioma, cutaneous melanoma, and renal cell carcinoma [26,27]. Mice lacking Bap1 also exhibit myeloid transformation [28] and spontaneous malignant mesothelioma [29], demonstrating that BAP1 is a bona fide tumor suppressor. BAP1 was initially identified as a protein that binds to the RING finger domain of BRCA1 [30]. Follow-up studies have demonstrated that BAP1 interacts with chromatin-associated proteins such as host cell factor 1 (HCF1), O-linked N-acetylglucosamine transferase (OGT),

Discussion
Inactivating mutations or truncation of Bap1 frequently occur in familial cancers such as uveal melanoma, mesothelioma, cutaneous melanoma, and renal cell carcinoma [26,27]. Mice lacking Bap1 also exhibit myeloid transformation [28] and spontaneous malignant mesothelioma [29], demonstrating that BAP1 is a bona fide tumor suppressor. BAP1 was initially identified as a protein that binds to the RING finger domain of BRCA1 [30]. Follow-up studies have demonstrated that BAP1 interacts with chromatin-associated proteins such as host cell factor 1 (HCF1), O-linked N-acetylglucosamine transferase (OGT), and additional sex comb-like proteins (ASXL1 and ASXL2) to regulate the gene expression, double-strand DNA repair, and DNA duplication by deubiquitinating histone H2A at Lysine 119 [31]. On the other hand, the localization of BAP1 in the cytosol is tightly controlled by the E3 ligase UBE2O [32], where BAP1 interacts with type 3 inositol-1,4,5-triphosphoate receptor (IP3R3) at the endoplasmic reticulum (ER) and promotes Ca 2+ signaling and apoptosis in response to oxidative stress [33]. While the level of BAP1 was very minimal in A549 cells (data not shown), we observed that transduced BAP1 was mostly located in the cytosol ( Figure 3A), stabilized KEAP1 ( Figure 3C), and suppressed the expression of NRF2 target genes ( Figure 3D). At present, we are unaware whether UBE2O played a role in the cytoplasmic localization and/or the stability of BAP1 in A549 cells.
It was previously demonstrated that USP15 is a DUB of KEAP1 [34]. However, USP15 failed to induce KEAP1 ( Figure 1A) nor bind to KEAP1 in our hands (data not shown). In the present study, we have demonstrated that BAP1 deubiquitinates KEAP1 ( Figure 1E) by binding to the BTB domain ( Figure 2A) and that KEAP1 lacking the BTB domain fails to respond to the stabilization by BAP1 ( Figure 2C). The molecular mechanisms underlying how the BTB domain contributes to the stability of KEAP1 are largely unknown, but it is possible to assume that the BTB domain might possess potential lysine residue(s) susceptible for polyubiquitination of KEAP1 that can be removed by BAP1. Alternatively, it can be speculated that BAP1 might competitively bind to KEAP1 at the BTB domain with unknown protein(s) affecting the stability of KEAP1. However, it is unlikely that TRIM25, an E3 ubiquitin ligase of KEAP1, competes with BAP1 because TRIM25 is reported to bind to KEAP1 at the KELCH domain, not the BTB domain [11]. On the other hand, we observed that KEAP1 binds to the CL domain in BAP1 ( Figure 2B). While the crystal structure of full-length BAP1 is unavailable, De et al. have provided the crystal structure of a BAP1 orthologue from Drosophila melanogaster, Calypso, bound to its chromatin activator ASX, additional sex combs [19]. However, the structure of the CL domain in Calypso was removed from this study because the CL domain exhibited a large conformational flexibility and, therefore, interfered with the crystallization of Calypso.
It is known that RAS activation accounts for 30% human LUAD, in which Keap1 is frequently mutated or deleted [35]. This suggests a possibility that the inhibition of KEAP1 might be a prerequisite event necessary for promoting LUAD caused by RAS activation. Supporting this speculation, we observed that the levels of Bap1 and Keap1 mRNA were progressively decreased in the lung tissues of Kras G12D/+ mice, compared with those of wild-type littermates during tumor progression ( Figure 4A). At present, the mechanisms by which KRAS activation downregulates BAP1 and KEAP1 in vivo are unclear. Because RAS activation induces the activation of various intracellular signaling kinases to promote the growth of tumors [36], it would be interesting to find out whether and, if so, how the activation of these kinases caused by RAS contributes to the inhibition of Bap1 and Keap1 transcription. Examination of the TCGA database also showed that the level of Bap1 mRNA was lower in the lung tissues of human LUAD compared with those of normal people ( Figure 5C) and it was associated with an increase in the survival of LUAD patients at tumor stages II-IV ( Figure 5D). These results illustrate that genetic or pharmacological BAP1 activation could be an effective strategy to treat LUAD patients at advanced tumor stages. Collectively, our study demonstrates that BAP1 is a DUB of KEAP1. While BAP1 has been regarded as a tumor suppressor possessing diverse molecular mechanisms [37], we reveal a possibility that BAP1 might serve as a redox regulator to inhibit NRF2 target genes and suppress the growth of lung cancer cells by deubiquitinating KEAP1. This finding provides a feasible possibility that targeting NRF2 might be helpful in treating familial cancers arising from Bap1 deficiency or mutations.

Conclusions
We have demonstrated that BAP1 directly binds to, deubiquitinates, and stabilizes KEAP1, exerting tumor suppressive effects in vitro. KRAS activation leads to a decrease in the levels of BAP1 and KEAP1 in the lung, resulting in NRF2 activation. Bioinformatics analyses indicate that the level of BAP1 is higher in normal human lung tissues and is positively associated with a better survival LUAD patients. Our results demonstrate that BAP1 is a tumor suppressor, which suppresses NRF2 in LUAD by deubiquitinating KEAP1.