The BCL-2 family member BOK promotes KRAS-driven lung cancer progression in a p53-dependent manner

A variety of cancer entities are driven by KRAS mutations, which remain difficult to target clinically. Survival pathways, such as resistance to cell death, may represent a promising treatment approach in KRAS mutated cancers. Based on the frequently observed genomic deletions of BCL-2-related ovarian killer (BOK) in cancer patients, we explored the function of BOK in a mutant KrasG12D-driven murine model of lung cancer. Using KrasG12D/+ Bok−/− mice, we observed an overall tumor-promoting function of BOK in vivo. Specifically, loss of BOK reduced proliferation both in cell lines in vitro as well as in KrasG12D-driven tumor lesions in vivo. During tumor development in vivo, loss of BOK resulted in a lower tumor burden, with fewer, smaller, and less advanced tumors. Using KrasG12D/+ Tp53Δ/Δ Bok−/− mice, we identified that this phenotype was entirely dependent on the presence of functional p53. Furthermore, analysis of a human dataset of untreated early-stage lung tumors did not identify any common deletion of the BOK locus, independently of the TP53 status or the histopathological classification. Taken together our data indicate that BOK supports tumor progression in Kras-driven lung cancer.


INTRODUCTION
Despite recent therapeutic advances, including the development of immune checkpoint inhibitors and other targeted therapies, lung cancer remains the leading cause of cancer-related deaths worldwide [1]. Lung adenocarcinoma (LUAD) is the most common subtype of lung cancer and is associated with significant rates of morbidity and mortality. About 33% of LUAD cases are associated with activating mutations in KRAS [2], which are difficult to target in a clinical setting. Furthermore, 46% of LUAD patients have somatic mutations in the tumor suppressor TP53--one of the most common events in cancer [2]. TP53 defends cells from acquired stress by triggering a multitude of anti-proliferative and stress-response signals such as senescence, cell cycle arrest, coordination of DNA damage repair, or apoptosis [3,4].
Notably, defects in apoptosis can promote tumorigenesis per se through extending the life span of a cell, therefore increasing its chances of acquiring additional cancer-promoting mutations [5]. Furthermore, apoptosis inhibition enables cell resistance to stress caused by oncogenic alterations, nutrient deprivation, DNA damage, or the build-up of toxic components [6]. Under normal conditions, apoptosis is controlled by the BCL-2 protein family [7], which comprises both pro-and anti-apoptotic members. Apoptotic signals result in permeabilization of the mitochondrial outer membrane, caspase activation, and cell death [7]. Interestingly, several cancer entities have altered expression of BCL-2 family members [7][8][9].
The BCL-2 family member BOK (BCL-2-related ovarian killer) [10] is a multi-BH-domain family member with sequence and structural homology to pro-apoptotic BAX and BAK [11,12]. BOK localizes primarily to the endoplasmic reticulum (ER) [13], where its activity is negatively regulated by ER-associated degradation [14]. Furthermore, BOK is a critical inducer of apoptosis in response to ER stress [15]. Notably, BOK has also been linked to the pathway of TP53-mediated cellular response to DNA damage, with Zhang et al. showing increased levels of damage in the absence of BOK [16,17]. Recent data point to a role for BOK also in cell proliferation [18] and Srivastava et al. suggested that BOK interacts with the UMP synthetase enzyme (UMPS), aiding its function in nucleotide synthesis and thus promoting cellular proliferation [19].
While its function and purpose have remained in part elusive, Beroukhim et al. found that the genomic locus containing BOK amongst several other genes is frequently deleted across several cancer entities [20]. These data were initially interpreted as pointing to a previously unknown function of BOK as a tumor suppressor and thus underlining the need for further investigation.
Using an inducible model of Kras-driven murine lung cancer, we show that BOK plays a role in promoting tumor development, as Bok-proficient mice developed a larger number of more advanced lesions. This effect is mediated, at least in part, by defects in proliferation, and is entirely dependent on p53. Furthermore, we show that the BOK locus is not genetically deleted in human lung cancer naïve samples. Collectively, our data point towards a function of BOK in promoting the development of lung adenocarcinomas.

Bok deletion reduces tumor burden
To analyze the role of BOK in lung cancer, we used mice harboring a lox-stop-lox-Kras G12D allele, in which the mutated Kras is expressed upon AdenoCre viral infection of the lungs [21]. Through this synchronized induction of tumor development, we compared the development of lung cancer between littermate mice with two copies (Bok +/+ ), one copy (Bok +/− ) or none (Bok −/− ) of the Bok gene. Intriguingly, Kras G12D/+ Bok −/− and Kras G12D/ + Bok +/− mice showed significantly lower tumor burden than the Bok-proficient animals ( Fig. 1a, b). This was due to a decrease in overall lesion number (Fig. 1c). We then histologically graded the most advanced lesion present in the lungs of experimental mice [21]. Whereas Bok −/− and Bok −/+ mice more frequently presented with histologically well-differentiated lesions, Bok-proficient mice presented with more advanced lesions (Fig. 1d). Notably, most advanced lesions (carcinoma) were only observed in the Bok− proficient mice. These data indicate that loss of BOK leads to a lower tumor burden and lesion progression in Kras-driven lung cancer.
Bok deletion decreases proliferation As the proliferative rate is an important characteristic of malignant lesions and BOK has been shown to affect proliferation [18,19], we analyzed Ki67 in our setting, a well-established immunohistochemical marker of proliferation. Consistent with a prior publication on hepatocellular carcinoma [18], Bok +/+ lesions showed significantly more Ki67 positive cells than Bok −/− lesions (Fig. 2a,  b). However, we did not observe any differences in the number apoptotic cells in Bok +/+ , Bok +/− or Bok −/− lesions (Fig. S1).
To further explore the role of BOK, we used the mutant Krasdriven mouse lung adenocarcinoma cell line LKR10 [22] to generate Bok-deficient cells by CRISPR/Cas9 (LKR10 Bok −/− and, as control, LKR10 LacZ, Fig. S2a). Bok knockout had no effects on the protein expression of its close relatives BAK and BAX, or its well-established pro-survival interaction partner MCL-1 [10,19] (Fig. S2b). As expected, we found that Bok-deficient cells proliferate slower than their counterparts (Fig. 2c). Furthermore, cell cycle analysis revealed that Bok-deficient cells show longer G2 duration (Fig. 2d), suggesting persistent DNA damage. Following this data, we analyzed the levels of phosphorylated histone γ-H2AX, an indicator of DNA double-strand breaks, in our experimental animals. In line with our hypothesis, we detected higher levels of γ-H2AX in Bok-deficient lesions (Fig. 2e). Taken together, these data show that BOK impacts on proliferation as well as on DNA damage repair response, which could explain the decreased tumor burden observed in the absence of BOK. Functional Tp53 mediates the Bok− dependent phenotype Since alterations in the TP53 tumor suppressor gene occur in about 50% of NSCLC cases [23], we investigated the contribution of functional p53 signaling to the effects of BOK in lung adenocarcinoma by crossing the Kras G12D Bok −/− mice in a Tp53 fl/fl background. We generated mice in which the administration of Cre recombinase allowed for the simultaneous deletion of the tumor suppressor p53 as well as Kras G12D activation [23]. Notably, analysis of tumor burden showed no differences across genotypes ( Fig. 3a-c), suggesting that the function of BOK is dependent on the presence of p53 in this model of tumorigenesis. Accordingly, proliferative rates and DNA damage levels were similar in Bok-deficient and -proficient mice ( Fig. 3d-f). Knockout of p53 also rescued the defect in cell cycle progression in the Bokdeficient mouse lung adenocarcinoma cell line LKR10 (Fig. S3), supporting the findings in the Kras G12D/+ Tp53 Δ/Δ mouse model.
Bok-deficiency sensitizes NSCLC cells against DNA damaging agents As our previous data suggested that Bok-deficient lung cancer cells have a defect in DNA repair leading to a p53-dependent arrest in G2 (Figs. 2 and S3), we wanted to determine if Bokdeficient cells are thus more sensitive to DNA damaging agents. First, to measure differences in DNA damage and DNA repair capacity in Bok-deficient versus Bok-proficient lung cancer cells upon induction of DNA damage, we performed comet assays in the LKR10 LacZ and Bok −/− cells treated with Etoposide. As expected, Etoposide-treated Bok-deficient cells had significantly higher levels of DNA damage (% DNA in tail) than the LacZ control cells, indicating that Bok-deficient cells have a decreased capacity to efficiently repair DNA strand brakes (Fig. 4a). Accordingly, LKR10 Bok −/− cells showed significantly increased sensitivity to DNA damaging standard chemotherapeutics such as Etoposide and Irinotecan (Fig. 4b). These results highlight the important role of BOK for DNA repair processes in lung cancer cells.
The BOK locus remains unaffected in human lung cancer patients As stated before, BOK has been shown to be genomically deleted across several cancer entities [20]. To gain additional information specific to lung cancer, we interrogated an extended dataset of human patients from the TRACERx dataset [24]. The TRACERx prospective study was designed to analyze intra-tumor heterogeneity and the evolutionary trajectory of tumor cell genomes specifically in treatment-naïve lung squamous carcinoma (LUSC) and LUAD patients. Using the TRACERx data, we detected that BOK is not frequently deleted in either cohort (Fig. 5a, b). Our data in the mouse models suggest an interplay between BOK and TP53, therefore we re-run the same analysis in the TP53 wild type (WT) subpopulation. Again, both in LUSC and LUAD, we detected no significant BOK deletion (Fig. 5c, d).
Taken together, our data suggest that, while BOK plays a role as tumor-promoting factor in lung cancer, and clearly does so in collaboration with TP53, it is not genomically deleted in lung cancer from patients in the TRACERx study.

DISCUSSION
Despite the fact that the BCL-2 family member BOK was discovered more than 20 years ago, its function still remains ambiguous (reviewed in [25]). Using a constitutional Bok −/− knockout mouse model, we found that Bok-deficiency reduces tumor burden, by decreasing lesion number and histological grading. This phenotype was mainly driven by proliferation deficits both in vivo and in vitro models. Furthermore, these effects were entirely lost in the absence of functional p53, together suggesting a tumor promoting role for BOK.
Due to the structural homology of BOK with the apoptosis effectors BAX and BAK [11,12] and the aggravated phenotype seen in BAX/BAK/BOK full body triple knockout mice [11], BOK is considered to be a pro-apoptotic member of the BCL-2 family. As such, its deletion should be an advantage in tumor survival and BOK should act as a tumor suppressor. However, recent publications also point towards cell death-independent functions of BOK, including a pro-proliferative role of BOK (first proposed by [26]). Bok −/− cell lines, including lung cancer cell lines, divided slower [19,27] and Bok −/− hepatocellular carcinomas and cell lines showed lower proliferative indexes [18]. In line with these data, we showed that Bok-deficient lung cancer lesions have lower Ki67 levels and Bok-deficient lung cancer cell line proliferate slower.
It can be argued that apoptosis induction and proliferation inhibition are two sides of the same coin. It is known that apoptosis and cellular proliferation are linked by cell-cycle regulators and signaling pathways, affecting both processes [28]. Under physiologic conditions, apoptosis elicits a signaling cascade inducing cell proliferation in the neighboring cells [29]. Therefore, in our settings, the deletion of Bok may decrease proliferation in surrounding cancer cells, by repressing apoptosis in tumor cells.
Furthermore, the proliferative defect seen after downregulation of BOK in human HCT-116 colorectal carcinoma cells is dependent on increased cell cycle arrest, due to the higher levels of the cyclin kinase inhibitors p19 INK4d and p21 cip1 [18]. Along the same line, p53 was found to be upregulated in BOK-deficient cells [19]. These results are in line with our finding, showing that the absence of p53 eliminates any Bok-driven differences in vivo and in vitro. It could be argued that, in our mouse models, the differences are driven by a ceiling effect due to the different aggressiveness of the tumor models. However, while this could explain the tumor burden, it could not justify the effects on proliferation and DNA damage.
While Beroukhim et al. found the genomic region containing the BOK locus to be deleted in a wide-ranging screen of primary cancer tissue and established cell lines [20], we identified no recurrent deletion of the BOK locus in LUAD and LUSC treatmentnaïve samples coming from the TRACERx dataset. These results are in line with our in vivo data, in which deletion of Bok impaired   tumor proliferation. However, our experimental settings do not allow us to exclude a different BOK involvement at later time points. Moravcikova et al. showed detectable BOK mRNA and protein levels in biopsies from early and late-stage lung cancer patients as well as in various NSCLC cell lines, supporting our finding that the BOK gene is not frequently deleted in NSCLC [27]. Interestingly, the same study indicated that BOK may exert a tumor-suppressor-like function at later stages of disease. While our data suggest that BOK deletion does not represent a common event in early lung cancer as it interferes with tumor proliferation, it remains to be tested if and at what stage the proliferation defects seen upon loss of BOK are overcome by the numerous other transformative events in the course of tumor development.
Taken together, our data show that BOK promotes proliferation of tumor cells in mutant Kras-driven lung cancer and that, in contrast to common opinion, BOK genomic deletion does not represent a recurrent event in lung cancer.
For tumor burden analysis, three slides per animal 100 µm apart were evaluated and the lobe and individual lesion areas were defined manually and measured using the Aperio ImageScope software (version 12.4, Leica Biosystems). The total area of the lesions was normalized over the total area of the lung for each slide. The average of three analyzed H&E slides was used for the reported data. For tumor grading, the slides were evaluated by a pathologist blind to the genotype and were classified as hyperplasia, atypical adenomatous hyperplasia (AAH), adenoma, and adenocarcinomas, as previously described [21,32]. For the analysis of Ki67 and γ-H2AX immunostainings, the six most advanced lesions on each slide by software-based analysis using the Positive Pixel Count v9 algorithm available in the Aperio ImageScope software (version 12.4, Leica Biosystems) and the reported data were normalized to lesion area.

Comet assay
Cells were seeded in 10 cm dishes at a density of 1.5 × 10 6 the day before the experiment and then treated with Etoposide (6 ug/ml) at 37°C for the indicated times. The cells were harvested, resuspended in ice-cold PBS at 1.4 × 10 5 , and then prepared using the Trevigen CometAssay Kit and its protocol. The cells were combined with low-melting agarose at a ratio of 1:10 and spread onto the CometSlides. Gelling was achieved at 4°C in the dark for 10 min and the slides were then placed in 4°C Lysis Solution overnight. After incubating the slides in alkaline unwinding solution for 20 min at room temperature (200 mM NaOH, 1 mM EDTA), electrophoresis took place at 4°C and 1 V/cm in alkaline electrophoresis solution (200 mM NaOH, 1 mM EDTA) for 25 min. The slides were dried at 37°C and DNA was stained in the dark for 15 min using propidium iodide (1 mg/ml stock solution) at a 1:1500 dilution. Slides were analyzed using the OpenComet software tool (https://cometbio.org/index.html).

TracerX analysis
The TRACERx study was conducted under the favorable opinion from the NRES Committee London--Camden & Islington Research Ethics Committee [24]. Informed consent was obtained from all subjects. BOK copy number alteration was evaluated using a similar approach to the one described in [32].

Statistical analysis
The details of experiments, including sample number, statistical tests used, number of experiments, and dispersion and precision measures, are stated in the figures or figure legends. The data analysis was performed with GraphPad Prism software and in the R statistical environment version ≥3.3.1. The appropriate statistical test was selected depending on data type and distribution. For all statistical tests statistical significance was determined if the p value was less than 0.05.