Sperm-specific COX6B2 enhances oxidative phosphorylation, proliferation, and survival in human lung adenocarcinoma

Cancer testis antigens (CTAs) are proteins whose expression is normally restricted to the testis but anomalously activated in human cancer. In sperm, a number of CTAs support energy generation, however, whether they contribute to tumor metabolism is not understood. We describe human COX6B2, a component of cytochrome c oxidase (complex IV). COX6B2 is expressed in human lung adenocarcinoma (LUAD) and expression correlates with reduced survival time. COX6B2, but not its somatic isoform COX6B1, enhances activity of complex IV, increasing oxidative phosphorylation (OXPHOS) and NAD+ generation. Consequently, COX6B2-expressing cancer cells display a proliferative advantage, particularly in low oxygen. Conversely, depletion of COX6B2 attenuates OXPHOS and collapses mitochondrial membrane potential leading to cell death or senescence. COX6B2 is both necessary and sufficient for growth of human tumor xenografts in mice. Our findings reveal a previously unappreciated, tumor-specific metabolic pathway hijacked from one of the most ATP-intensive processes in the animal kingdom: sperm motility.


Introduction
Tumors frequently activate genes whose expression is otherwise restricted to testis; these genes are known collectively as cancer-testis antigens (CT-antigens, CTAs). The expression of these genes outside their native and immune privileged site has been the basis for immunotherapeutic approaches including vaccines and adoptive T-cell therapy (Hunder et al., 2008;Gjerstorff et al., 2015). Historically, functional knowledge of the contribution of these proteins, if any, to neoplastic behaviors has significantly lagged behind immune-targeting. In recent years, the functional characterization of CTAs has gained broader interest and these proteins have been implicated in tumor cell survival, TGF-b signaling, mitotic fidelity, polyadenylation, mRNA export, tumor suppressor silencing, and DNA damage repair (Gibbs and Whitehurst, 2018;Viphakone et al., 2015;Nichols et al., 2018).
Multiple reports now indicate that anomalous expression of testis proteins can be engaged to support the tumorigenic program. The biased expression pattern of these proteins to tumors and testis may offer an extraordinarily broad therapeutic window if they can be targeted directly.
One unique aspect of mammalian sperm physiology is the tremendous energy demand required for motility while preserving the integrity of their precious DNA cargo within the hostile environment of the female reproductive tract. To meet this demand, sperm contain a number of tissue-specific protein isoforms for glycolysis and oxidative phosphorylation (OXPHOS) that mediate the increase in ATP production. For example, lactate dehydrogenase C, LDHC, is a testis-specific isoform of the terminal enzyme in glycolysis that catalyzes the reduction of pyruvate to lactate essential for male fertility (Odet et al., 2008;Krisfalusi et al., 2006;Li et al., 1989). Similarly, COX6B2 is a testis-specific subunit of cytochrome c oxidase (complex IV) (Hüttemann et al., 2003). mRNA encoding of either LDHC or COX6B2 is undetectable in normal tissue, but both are upregulated in a number of different tumor derived cell lines, classifying them as CTAs (Maxfield et al., 2015, (CTpedia (http://www. cta.lncc.br/index.php))). However, it is unclear whether these proteins support metabolic programs in tumor cells.
In a large-scale lossof-function analysis to annotate the contribution of individual CTAs to neoplastic behaviors, we found that COX6B2 is essential for survival of non-small cell lung cancer (NSCLC) cell lines (Maxfield et al., 2015). COX6B2 is a nuclear encoded, sperm-specific component of complex IV (Hüttemann et al., 2003). By transferring electrons from reduced cytochrome c to O 2, complex IV is the rate-limiting step for ATP production by the electron transport chain (ETC). Thirteen subunits make up complex IV: three are mitochondrial encoded and ten are derived from nuclear DNA. Six of the 10 nuclear encoded subunits have tissue-specific isoforms that permit regulation of this complex in response to environmental cues (e.g. pH, hormones, metals, ATP/ADP ratio etc.) (Kadenbach and Hüttemann, 2015). The somatic isoform of COX6B2 is COX6B1. These two proteins share 58% amino acid identity and~80% similarity. Based on structural information, it is apparent that COX6B1/2 are the only complex IV subunits that are not membrane bound. Instead, their localization is confined to the intermembrane space where cytochrome c associates. It is inferred that COX6B1/B2 participate in dimerization of complex IV and cytochrome c association (Sampson and Alleyne, 2001;Tsukihara et al., 1996). While limited, studies on COX6B1 indicate that it is essential for complex IV activity. In particular, human mutations in COX6B1 (R20C or R20H) abrogate complex IV assembly and lead to severe cardiac defects (Abdulhag et al., 2015). In addition, biochemical analysis indicates that removal of COX6B1 from assembled complexes enhances complex IV activity, implying a negative regulatory role for COX6B1 (Weishaupt and Kadenbach, 1992). To-date no reports detail the function of COX6B2 nor indicate the physiological relevance of this sperm-specific subunit to fertility. Furthermore, there are no reports describing mechanism(s) that regulate COX6B1 or COX6B2 expression. Overall, compared to the enormous data on complex IV in general, the COX6B proteins have been understudied.
We have undertaken a detailed investigation into the mechanism of action of COX6B2 when this protein is aberrantly expressed in NSCLC. Here, we report that COX6B2 enhances mitochondrial oxidative phosphorylation (OXPHOS) in tumor cells. This activity accelerates proliferation in vitro and in vivo. In contrast, silencing of COX6B2 attenuates OXPHOS, reduces tumor cell viability and dramatically decreases growth in vivo. Importantly, we find that hypoxia enhances COX6B2 expression, which confers a selective advantage for proliferation under low oxygen. Indeed, elevated COX6B2 mRNA correlates with reduced survival of LUAD patients. Cumulatively, this study demonstrates the remarkable capacity of tumor cells to integrate primordial gene products into their regulatory environment as a means of promoting unrestrained proliferation. In turn, COX6B2 becomes a liability that may be exploited for tumor selective targeting of OXPHOS.

COX6B2 is expressed in human lung adenocarcinoma (LUAD) tumors and correlates with poor survival
Using a large-scale functional genomics approaches, we previously found that depletion of COX6B2 activates cleaved caspase 3/7 in cell lines derived from breast, melanoma and NSCLC, with the most potent activation in H1299 NSCLC cells (Figure 1-figure supplement 1A; Maxfield et al., 2015). Based on these findings, we examined the relationship between COX6B2 expression and patient outcome in human NSCLC. Strikingly, elevated expression of COX6B2 is associated with significantly shorter overall survival (OS) time (p=5.3Â10 À6 , log rank test; hazard ratio (HR): 1.46; 95% confidence interval (CI): 1.24-1.73) ( Figure 1A). In addition, COX6B2 expression positively correlates with time to first progression (FP) (p=4.1Â10 À4 , log rank test; HR: 1.63; 95% CI: 1.24-2.14) ( Figure 1A). Separation of the two major histological subtypes of NSCLC revealed a strong correlation for both outcomes in LUAD (OS: p=1.6Â10 À4 , log rank test; HR: 1.59; 95% CI: 1.25-2.03; FP: p=9.4Â10 À5 , log   Figure 1C). Analysis of TCGA data sets also revealed a significant upregulation of COX6B2 mRNA in LUAD tumors as compared to normal tissues ( Figure 1D). We next immunoblotted for COX6B2 protein in a panel of human LUAD tumors and normal lung tissue using an antibody that detects COX6B2 ( Figure 1E; Figure 1-figure supplement 1B). We found that COX6B2 protein expression is low or undetectable in normal lung tissue, but present in all tumor-derived tissues irrespective of stage and grade ( Figure 1E; Figure 1-figure supplement 1C). A similar pattern is also observed in LUAD-derived cell lines ( Figure 1F). Furthermore, immunohistochemical staining of COX6B2 in a panel of human non-malignant (n = 22) and LUAD tissues (n = 32) indicated accumulation of COX6B2 in tumor, but not non-malignant tissues ( Figure 1G). This expression of COX6B2 in tumors does not have an apparent correlation with the presence or absence of COX6B1, the somatic isoform of COX6B2. We did observe a moderate correlation with the complex IV subunit, COXIV, suggesting COX6B2 accumulation coincides with increased mitochondria in cancer cells ( Figure 1-figure supplement 1D). Furthermore, depletion or overexpression of either COX6B1 or COX6B2 does not impact protein accumulation of the corresponding isoform (Figure 1-figure supplement 1B). These findings indicate that while the isoforms are closely related in sequence their regulation may be through independent mechanisms. Immunofluorescence of endogenous COX6B2 in intact LUAD cells revealed specific localization to the mitochondria as judged by overlap with Tom20, a mitochondrial resident protein ( Figure 1H). Thus, COX6B2 is frequently expressed in LUAD, correlates with poor survival and when ectopically expressed can localize to its native mitochondrial location.

COX6B2, but not COX6B1, enhances oxidative phosphorylation in LUAD
To investigate whether COX6B2 supports OXPHOS when ectopically expressed in tumors, we generated a LUAD cell line that overexpresses COX6B2 protein~3 fold (HCC515-COX6B2-V5) ( Figure 2A). This overexpression is not associated with an increase in mitochondrial DNA or total mitochondria ( Figure 2B-C). We then measured the oxygen consumption rate (OCR) at baseline and following exposure to electron transport chain perturbagens. This analysis revealed a significant enhancement in basal, ATP-linked, maximal and reserve OCR in COX6B2-V5 expressing cells as compared to control ( Figure 2D). A similar trend is observed in an independent LUAD cell line, H2122, expressing COX6B2-V5 ( Figure 2E-F). The OCR increase is associated with an elevation in total ATP in both cell lines ( Figure 2G). In addition, we observed elevated NAD + /NADH in COX6B2-V5 cells, indicative of an overall increase in electron transport activity ( Figure 2H). We did not observe a significant or reproducible alteration in extracellular acidification rate (ECAR) in either setting ( Figure 2I). We also studied whether COX6B2's somatic isoform, COX6B1, was sufficient to enhance OCR ( Figure 2J). COX6B1-V5 expression appears to have no effect or slightly decreases OCR for each parameter ( Figure 2K), consistent with prior biochemical studies where COX6B1 removal enhances the activity of complex IV (Weishaupt and Kadenbach, 1992).

COX6B2 enhances complex IV activity and is incorporated into mitochondrial supercomplexes
To further characterize the impact of COX6B2 on complex IV, we directly assessed complex IV activity using a technique to pharmacologically isolate complex IV in permeabilized COX6B1-V5 or COX6B2-V5 cells (Salabei et al., 2014;Divakaruni et al., 2014; Figure 3A). Here, we find that COX6B2-V5 overexpression confers a significant increase (~25%) in complex IV activity in both LUAD  Figure 2 continued on next page cell lines tested ( Figure 3B). In contrast, a minimal, statistically insignificant increase in activity was observed following COX6B1-V5 overexpression ( Figure 3C). The increase in complex IV activity did not correlate with any apparent increase in the abundance of complex IV as judged by COXIV accumulation ( Figure 3-figure supplement 1A). This finding suggests that COX6B2 is a rate-limiting subunit of complex IV.
Recently, a number of reports have suggested that tissue-specific complex IV isoforms may promote supercomplex formation and selectively and specifically increase ATP production in tumor cells (Maranzana et al., 2013;Ikeda et al., 2019). Supercomplexes contain multiple electron transport sub-complexes containing complex III and IV and are hypothesized to promote local channeling of substrates, trapping of reactive intermediates and/or subunit stabilization (Acín-Pérez et al., 2008). To test this possibility for COX6B2, we used Blue-native PAGE (BN-PAGE) analysis to separate electron transport chain complexes (monomers (IV) and dimers (IV 2 ) and supercomplexes (III 2 +IV and I +III 2 +IV n )) in control and COX6B1/B2-V5 expressing cells. Here, we observed that >90% of endogenous COX6B2 is found in in complex IV dimers or incorporated into supercomplexes. In contrast, >98% COX6B1 is found in monomeric complex IV ( This trend is maintained when either isoform is overexpressed (>85% of COX6B2 is in complex IV in dimers/supercomplexes versus 40% for COX6B1) ( Figure 3F).
We next tested the polymeric distribution of complex IV, using analysis of COXIV, a core protein subunit of complex IV, following COX6B1-V5 or COX6B2-V5 expression. Here, we found that upon expression of COX6B2-V5, but not COX6B1-V5, the distribution of complex IV shifted from monomers to increased formation of supercomplexes ( Figure 3G). Supercomplexes are hypothesized to enhance electron transport activity without a compensatory increase in ROS (Maranzana et al., 2013). Indeed, COX6B2 overexpressing cells do not display a concomitant induction of superoxide ( Figure 3H). These data indicate that COX6B2 preferentially integrates into dimeric complex IV and supercomplexes, and COX6B2 expression promotes incorporation of complex IV into supercomplexes. These alterations in distribution are proposed to enhance mitochondrial OXPHOS while limiting oxidative stress.

COX6B2 enhances cell proliferation and confers a growth advantage in low oxygen
We next investigated whether the expression of COX6B2 is sufficient to enhance the growth of LUAD cells. We observed an increase in DNA synthesis as measured by EdU incorporation of 50.39 ± 6.35% in HCC515 and 38.7 ± 8.4% in H2122 ( Figure 4A; Figure 4-figure supplement 1A). Notably, an increase is not observed for COX6B1-V5 overexpression consistent with no significant change in OCR ( Figure 4A, middle panel; Figure 2K). In addition, we observed an enhanced population double rate of 37.16 ± 10.71% and 29.88 ± 6.74% for the HCC515 and H2122 cells, respectively ( Figure 4B  Mitochondrial OXPHOS has been linked to proliferation via regulation of the NAD + /NADH ratio, which is elevated in COX6B2-V5 overexpressing cells ( Figure 2H; Birsoy et al., 2015;Sullivan et al., 2015). Consistent with this model, supplementation with pyruvate, an electron acceptor which drives NAD + production, largely diminishes the proliferative effects of COX6B2-V5 overexpression ( Figure 4B). A similar proliferative advantage is also observed when COX6B2-V5 cells are grown in the absence of extracellular matrix, suggesting that COX6B2 promotes anchorage-independent growth ( Figure 4C). Bars represent mean ± SD (n = 3). (F) As in (D) (n = 8). (G) ATP content of indicated cell lines. Bars represent mean + SEM (n ! 6). HCC515 p value calculated by Student's t-test and H2122 p value calculated by Mann-Whitney test. (H) NAD + /NADH measurement in indicated cell lines. Bars represent mean ± SEM (n ! 5). p-Value calculated by Student's t-test. (I) Basal extracellular acidification rate (ECAR) are presented as mean ± SEM (n ! 6) from CTRL and COX6B2-V5 cells of HCC515 and H2122 cells. p-Value calculated by Student's t-test. (J) Immunoblots of indicated whole cell lysates with specified antibodies. MW markers are indicated. (K) As in (D) using indicated cell lines. Bars represent mean ± SEM (n ! 9). Basal and ATP-link OCR p value calculated by Student's t-test. Maximal OCR and reserve capacity p value calculated by Mann-Whitney test. Isoforms of complex IV subunits can be selectively expressed to permit adaptation to tissue-specific needs or environmental conditions (Sinkler et al., 2017). In particular, hypoxia is reported to regulate complex IV subunit expression and permit continued respiration in low oxygen (Ikeda et al., 2019;Hayashi et al., 2015;Fukuda et al., 2007). Therefore, we cultured control and COX6B2-V5 overexpressing cells in low (2.5%) O 2 for 4 days, monitoring EdU incorporation every 24 hr. Under these conditions, control cells reduce EdU incorporation by~60%, while COX6B2-V5 expressing cells decrease incorporation by 15% ( Figure 4D; Figure 4-figure supplement 1C). This data indicates that COX6B2 expression can enhance proliferation in both normoxia and low oxygen conditions. Given this phenotype, we examined whether COX6B2 is regulated in response to low oxygen. ChIP data from human breast cancer cells indicates that HIF-1 is bound near the COX6B2 promoter (Zhang et al., 2015). In addition, we also identified 11 putative HREs (A/G-CGTG) which are enriched at the HIF-1 peaks (Figure 4-figure supplement 1D). Indeed, we observed a greater than twofold increase in COX6B2 but not COX6B1 mRNA following 12 hr of 2.5% oxygen ( Figure 4E). This accumulation of mRNA was associated with stabilization of COX6B2 protein ( Figure 4F). Under both normoxic or hypoxic conditions, we did not observe any significant alteration in COX6B2 or COX6B1 accumulation when the alternative isoform was overexpressed, indicating that the changes in expression are not an isoform switch (Figure 1-figure supplement 1B; Figure 4-figure supplement 1E). Collectively, these data suggest that the enhanced electron transport activity in COX6B2 expressing cells confers a selective advantage by promoting key tumorigenic behaviors including unrestrained proliferation, anchorage independent growth and enhanced survival in hypoxia.

Depletion of COX6B2 impairs mitochondrial electron transport chain function in LUAD cells
Given the pro-proliferative phenotype following COX6B2 expression, we next examined the consequences of its depletion on OXPHOS and tumor cell viability. Reduction of COX6B2 protein using shRNA pools (four independent sequences) or siRNA pools (two independent sequences), led to a decrease in basal, ATP-linked, maximal and reserve OCR in HCC515 and H2122 cell lines ( Figure 5A-B; Figure 5-figure supplement 1A). We were unable to establish a stable clone of COX6B2 deleted cells by CRISPR/Cas9. However, COX6B2 directed CRISPR/Cas9 did generate a population with reduced COX6B2 protein expression ( Figure 5-figure supplement 1B). In agreement with the si and shCOX6B2 data, sgCOX6B2 cells also exhibit attenuated OCR ( Figure 5-figure supplement 1C). The reduction in OCR following COX6B2 depletion is accompanied by a reduction in the NAD + /NADH ratio ( Figure 5C). In addition, total cellular ATP is decreased following siRNA-mediated COX6B2 depletion with two independent siRNA sequences across a panel of COX6B2-expressing LUAD cell lines ( Figure 5D). Consistent with a loss of electron transport chain function, depletion of COX6B2 resulted in decreased mitochondrial membrane potential ( Figure 5E) and increased intracellular hydrogen peroxide ( Figure 5F). Collectively, these findings indicate that COX6B2 is essential for respiration and mitochondrial integrity in tumor cells.

Figure 3 continued
Indicated lysates from HCC515 cells were run on a BN-PAGE gel and stained with Colloidal Blue (left two panels) or immunoblotted with anti-COX6B2, anti-COX6B1 and anti-COXIV (right panels). Representative image of n ! 3. MW markers are indicated to identify different complexes (Mourier et al., 2014). (E) Left: Distribution of COX6B1 and COX6B2 incorporated in complex IV monomers, dimers or supercomplexes as detected by BN-PAGE in (D). Right: Bars represent mean ± SEM (n ! 3) based on quantification of bands in (D). In monomeric complex IV, p value calculated by Mann-Whitney test whereas others calculated by Student's t-test. (F) Left: Distribution of COX6B1/COX6B2 incorporated in monomeric complex IV, dimeric complex IV and supercomplexes based on BN-PAGE in COX6B2-V5 or COX6B1-V5 overexpressing cell lines (D). Right: Bars represent mean ± SEM (n ! 3) based on quantification of bands in (D). In monomeric complex IV, p value calculated by Mann-Whitney test whereas others calculated by Student's t-test.

Depletion of COX6B2 decreases cell survival and induces cell apoptosis and senescence
Based on the observed decrease in OCR, we asked whether tumor cell viability was diminished following COX6B2 depletion in LUAD cells. Depletion of COX6B2 for nine days, results in a decrease in EdU incorporation ( Figure 6A). Furthermore, these cells are severely compromised in their ability to grow in the absence of an extracellular matrix ( Figure 6B). Both of these phenotypes were recapitulated in sgCOX6B2 cells ( Figure 6-figure supplement 1A-B). We next examined the underlying mechanism(s) leading to the observed loss of viability. In our original screen, we found that siRNAmediated depletion of COX6B2 enhances cleavage of caspase 3/7, a phenotype that is rescued in COX6B2-V5 overexpressing cells ( Figure 6C, Figure 6-figure supplement 1C; Maxfield et al., 2015). In cells where COX6B2 is depleted using a pool of shRNAs, with sequences independent of the those in the siRNA pool, we observed cleavage of caspase-3 and PARP ( Figure 6C). These cell death markers are also induced in sgCOX6B2 cells ( Figure 6-figure supplement 1D). The induction of cell death proteins is accompanied by an increase in Annexin-V staining, demonstrating activation of caspase-dependent programmed cell death ( Figure 6D). Visual inspection of COX6B2-depleted cultures also revealed a population of cells that are flattened, accumulate vacuoles, and in some cases, are multi-nucleated, feature reminiscent of senescence ( Figure 6E). Staining of these cells revealed an increase in b-galactosidase, suggesting induction of a cellular senescence program ( Figure 6F). To determine the underlying trigger for the reduction in survival, we attempted to rescue the death through multiple approaches. Supplementing depleted cells with pyruvate and uridine as metabolic and nucleotide fuel was insufficient to rescue the viability defects ( Figure 6-figure supplement 1E-F). However, exposing cells to the ROS scavenger N-acetyl-l-cysteine (NAC) reduced the extent of the cleaved caspase-3 and PARP and b-galactosidase staining ( Figure 6G-H). Thus, COX6B2 is essential for survival of LUAD tumor cells as its suppression reduces ATP, compromises mitochondrial integrity thereby generating excess ROS that leads to apoptosis or senescence.

COX6B2 is necessary and sufficient for tumor growth in vivo
We next sought to determine whether COX6B2 was necessary and/or sufficient for tumor growth in vivo. We implanted control and COX6B2-V5 HCC515 cells into the flank of FOXn1 nu (nude) mice. Importantly, tumors engrafted subcutaneously encounter a highly hypoxic environment (~0.08-0.8% O 2 ) (Helmlinger et al., 1997). In agreement with the cell line data, we observed enhanced growth of COX6B2-V5 expressing cells in vivo ( Figure 7A). After 38 days, COX6B2-V5 tumors exhibit a 4X greater mass than control ( Figure 7B). In the converse experiment, we xenografted control and shCOX6B2 cells in the flank of immunocompromised mice. Complementing the gain-of-function phenotype observed, we found that depletion of COX6B2 severely reduces tumor growth and final tumor mass ( Figure 7C-D). Taken together, our findings indicate that the anomalous expression of COX6B2 is both necessary and sufficient for growth of LUAD tumors.

Discussion
COX6B2 function has remained obscure in both tumor cells and sperm since its discovery (Hüttemann et al., 2003). Our data indicate that COX6B2 enhances complex IV activity leading to an increase in mitochondrial oxidative phosphorylation. The function of COX6B2 appears to differ from COX6B1, which did not enhance OXPHOS when overexpressed, a finding consistent with previous biochemical analyses that indicate removal of COX6B1 enhances complex IV activity (Weishaupt and Kadenbach, 1992). Genetic studies demonstrate that human mutations in COX6B1 reduce assembly of complex IV and other ETC complexes. Based on this cumulative data, we hypothesize that COX6B1 (and perhaps B2) are essential for assembly and regulation of complex IV.
In normal tissues, COX6B1 may regulate complex IV to prevent the generation of excessive and unused energy and also damaging ROS. In sperm, where ATP demand is high, COX6B2 expression could promote complex IV dimerization and incorporation into supercomplexes to promote efficient OXPHOS and limit ROS production that would otherwise damage DNA (Agarwal et al., 2014). This function may be particularly important during early tumorigenesis, where a highly hypoxic environment can lead to proliferative arrest and excessive ROS generation. We propose that tumor cells   Figure 6 continued on next page that express and engage the COX6B2-based, sperm-specific mechanism can overcome this innate barrier to transformation that would otherwise restrict cell survival.
One of the most surprising aspects of our study was the loss of tumor cell viability following the depletion of COX6B2. As COX6B2 enhances OXPHOS, its loss would be expected to reduce ATP levels and the NAD + /NADH ratio and slow proliferation, but not necessarily kill tumor cells. Tumor cell glycolytic pathways should theoretically compensate for reduced energy generation. Moreover, the presence of COX6B1 should be able to support complex IV function. Instead, we observe ROS production and an exit from the cell cycle either to death or senescence. This finding implies that these cells are highly dependent upon OXPHOS for survival. In addition, the loss of COX6B2 may promote apoptosis due to a collapse of mitochondrial membrane potential and cytochrome c egress (Smaili et al., 2001;Tai et al., 2017).
The majority of tumor metabolism studies have focused on the contribution of the Warburg effect and have led to the notion that OXPHOS is dispensable for many transformed cells. However, a growing body of evidence indicates that OXPHOS is essential for tumor cell division and survival. First, OXPHOS contributes to >80% of total ATP production in tumor cells, which is essential for satisfying the demands of proliferation and tumor progression (Rao et al., 2019;Lissanu Deribe et al., 2018;Zu and Guppy, 2004). Second, OXPHOS provides electron acceptors to generate metabolic intermediates (e.g. aspartate) that are required for continued cell division (Sullivan et al., 2015). Third, tumor cells exhibit enhanced sensitivity to inhibitors of OXPHOS (Nayak et al., 2018;Scotland et al., 2013). Fourth, isoform switching of complex IV subunits is observed in tumors and is associated with aggressive malignant behaviors (Oliva et al., 2015;Zhang et al., 2016;Minton et al., 2016). Finally, OXPHOS is upregulated and essential for survival during ER-stress, hypoxia and glucose limiting conditions, and exposure to targeted therapies (Balsa et al., 2019;Ikeda et al., 2019;Fukuda et al., 2007;Guièze et al., 2019). Based on these and a vast array of additional data, it is accepted that inhibition of OXPHOS is a tractable therapeutic entry point for treating cancer (Weinberg and Chandel, 2015). Indeed, extensive studies with metformin have demonstrated anti-proliferative activities and reduction of tumor incidence through the inhibition of Complex I (Weinberg and Chandel, 2015). However, this and other ETC inhibitors cannot discriminate between tumor or normal cells, limiting their therapeutic window. We propose that enhanced accumulation of COX6B2 is an acquired trait that is essential for OXPHOS in LUAD and perhaps additional cancers. At minimum, elevated COX6B2 protein expression is a functional enrollment biomarker for OXPHOS-inhibitor-based treatment. Moreover, COX6B2 structural elements that are distinct from COX6B1 could represent targets for small molecule inhibitors. Finally, due to expression in an immune privileged site, CTAs have long been considered ideal targets for immunotherapy. However, no objective criterion exists to prioritize them for immunotherapy. We propose that COX6B2 represents an ideal candidate for immunotherapy because it has a bona fide tumorigenic function. Specifically, sequence differences between COX6B2 and COX6B1 could permit antigen recognition by patient immune cells and thus act as targets of adoptive T-cell transfer. Continued on next page

Kaplan-Meier analysis
Overall survival and time to first progression graphs were generated using kmplotter (

TCGA data analysis
We used the UCSC Xena program (https://xenabrowser.net) to download RNA-Seq data from the cohort of TCGA Lung Cancer (LUNG). Values were analyzed using Prism Graphpad software (v 8.4.3).
Immunoblotting SDS-PAGE and immunoblotting were performed as previously described (Maxfield et al., 2015).

Immunofluorescence
Cells plated on glass coverslips were washed with PBS, fixed with 4% formaldehyde for 15 min at room temperature and blocked for 3 hr at room temperature in blocking buffer (5% BSA in PBS with 0.1% Saponin). Cells were incubated with primary antibodies diluted with blocking buffer overnight at 4˚C followed by washes and incubation with Alexa Fluor-conjugated secondary antibodies (Thermo Fisher Scientific) for 1 hr at room temperature. ProLong Gold Antifade reagent with DAPI (Thermo Fisher Scientific) was used to mount slips on glass slides and images were acquired by Zeiss LSM510 confocal microscope or Leica DM5500 B upright microscope. Antibodies used: COX6B2 (1:50, SAB1401983, MilliporeSigma), Tom20 (1:200, sc-17764, Santa Cruz Biotechnology), COXIV (1:200,4850,Cell Signaling Technology). Staining was quantitated using ImageJ software.

Human lung tissue immunohistochemistry (IHC)
Non-malignant testis, lung and LUAD tissues were obtained with informed consent from the UTSW Tissue Management Shared Resource (TMSR) in compliance with the UTSW Internal Review Board committee. COX6B2 IHC was optimized and performed by the TMSR according to internal protocols using the Dako Autostainer Link 48 system. Briefly, the slides were baked for 20 min at 60˚C, then deparaffinized and hydrated before the antigen retrieval step. Heat-induced antigen retrieval was performed at low pH for 20 min in a Dako PT Link. The tissue was incubated with a peroxidase block and then an antibody incubated (COX6B2; 1:150, SAB1401983, MilliporeSigma) for 35 min. The staining was visualized using the EnVision FLEX visualization system. Staining scores of COX6B2 ranging from 0 to 3 were set based on the positive scores quantitated by using IHC Profiler of ImageJ (Varghese et al., 2014): score 0 (<30), score 1 (30-45), score 2 (45-60), and score 3 (>60). Figure 1G demonstrates a 0 for negative staining, whereas score 3 shows the highest staining intensity.

Lentiviral transduction
Lentivirus was produced through co-transfection of HEK293T cells with 5 mg of viral expression vector, 3 mg of psPAX2 packing vector, and 2 mg of pMD2.G envelope vector. Virus-conditioned medium was harvested, passed through 0.45-mm-pore-size filters, and then used to infect target cells in the presence of 10 mg/ml Sequa-brene (S2667, MilliporeSigma).

Mitochondrial content and mass analysis
Mitochondrial DNA quantification was performed as described (de Almeida et al., 2017). To detect mitochondrial mass, cells were incubated with 50 nM MitoTracker Green (Fisher Scientific) at 37˚C for 25 min and processed according to manufacturer's protocol. A minimum of 20,000 cells were analyzed by flow cytometry using a BD LSRFortessa instrument and BD FACSDiva 6.2 software. Data were analyzed by FlowJo (v10). The median fluorescent intensities in each group were used to calculate the fold change compared to the CTRL group.

Metabolic assays
The oxygen consumption rate (OCR) and extracellular acidification rate (ECAR) were performed by using the Seahorse Xfe96 Extracellular Flux analyzer (Agilent). Cells were plated in the Xfe96 cell culture plates at a density of 10,000 cells (HCC515) or 5000 cells (H2122) in 80 ml of growth medium overnight to generate a confluent monolayer of cells. The assay medium consisted of 10 mM glucose, 2 mM glutamine, 1 mM pyruvate in DMEM (D5030, MilliporeSigma). The reaction was monitored upon serial injections of oligomycin (5 mM), CCCP (1 mM), and antimycin A (2 mM). Results were normalized to total protein amount as determined by Pierce BCA Protein Assay Kit. Non-mitochondrial respiration is subtracted from the presented data. COX6B2-depleted cells were assessed following eight days of shRNA or three days of siRNA. Complex IV activity of isolated mitochondria was directly assessed in permeabilized cells as described (Salabei et al., 2014;Divakaruni et al., 2014), except that cells were permeabilized by XF PMP (1 nM). Complex III was inhibited with antimycin A (2 mM) and then treated with complex IV substrates (Ascorbate 10 mM, TMPD 100 mM) ( Figure 3A).

Measurement of ATP
Cell-Titer Glo (Promega) (CTG) was performed by manufacture's protocol but modified to use 15 ml of CTG for 100 ml of cells in media. Data were normalized to total protein amount as determined by Pierce BCA Protein Assay Kit. Luminescence was read using a Pherastar Plus plate reader (BMG Labtech).

NAD + /NADH assay
Cells were plated into 96-well tissue culture plates (Corning), allowed to adhere overnight and reach 80-90% confluency. Cells were processed according to the manufacturer's protocols for NAD + / NADH-Glo Assay (Promega) with modification as previously described (Gui et al., 2016). NAD + / NADH measured in COX6B2 depleted cells were transduced with shRNA for 8 days.

Measurement of reactive oxygen species (ROS)
The Fluorescent probe MitoSOX Red(ThermoFisher) was used to measure superoxide. MitoSOX Red is a mitochondrial membrane-dependent dye and therefore could not be used in COX6B2-depleted cells as they lose mitochondrial membrane potential (DÉ m ). In this case, Chloromethyldichlorodihydrofluorescein diacetate (CM-H2DCFDA, Fisher Scientific), detecting H 2 O 2 , was used for total ROS analysis in cells 11 days of shRNA silencing. For MitoSOX Red assay, 3 Â 10 5 cells were stained by MitoSOX Red (5 mM; in HBSS) at 37˚C for 10 min. As a positive control, cells were treated with 50 mM antimycin A during staining. For H 2 O 2 measurements, cells reaching approximately 80% confluency in six-well dishes were stained by CM-H2DCFDA (5 mM; in HBSS) at 37˚C for 20 min, washed with HBSS, and then recovered in culture media at 37˚C for 15 min. Cells were immediately analyzed by flow cytometry using a BD LSRFortessa instrument and BD FACSDiva 6.2 software. A minimum of 20,000 cells were analyzed per condition. Data were analyzed by FlowJo (v10). Forward scatter (FSC) and side scatter (SSC) parameters were used to exclude cellular debris, dead cells, and doublets to retain viable single cell events. The median fluorescent intensities in each group were used to calculate the fold change compared to the CTRL group. CM-H2DCFDA had to be used in COX6B2depleted cells because MitoSOX Red is dependent upon mitochondrial integrity, which is lost under these conditions (Zielonka and Kalyanaraman, 2010).

EdU incorporation assays
Cells were exposed to ethynyl deoxyuridine (EdU) for 1 hr before fixing the cells in 4% formaldehyde. Cells were stained using the protocol for Click-iT EdU Alexa Fluor 488 Imaging Kit (Thermo Fisher) and co-stained with Hoechst 3342. Cells were quantified using fluorescence microscopy. EdU detected in COX6B2-depleted cells were transduced with shRNA for 9 days.

Measurement of mitochondrial membrane potential
Following 6 days of shRNA (3 Â 10 5 cells), cells were collected and resuspended in HBSS containing 0.2 mM JC-1 and incubated at 37˚C in a CO 2 incubator for 25 min. The stained cells were collected by centrifuge and washed twice with HBSS containing 1% FBS. Then cells were resuspended in 500 ml of 1% FBS HBSS buffer. Cells were immediately analyzed by flow cytometry using a BD LSRFortessa instrument and BD FACSDiva 6.2 software. A minimum of 20,000 cells were analyzed per condition. FlowJo (v10) was used to generate scatter plots.

Annexin-V staining
After 7 days of shRNA-mediated knockdown 1 Â 10 5 cells were wash once in PBS and resuspend in Annexin-V Binding Buffer (422201, BioLegend) with 3 ml of Annexin-V-APC (640941, BioLegend) and DAPI (200 ng/ml), and incubated for 15 min at room temperature in the dark. Cells were immediately analyzed by flow cytometry using a BD LSRFortessa instrument and BD FACSDiva 6.2 software. A minimum of 20,000 cells were analyzed per condition. FlowJo (v10) was used to generate scatter plots. Early-stage apoptotic cells are presented in the lower right quadrants (Annexin-V positive and DAPI negative) and late-stage apoptotic cells are presented in the upper right quadrants (Annexin-V positive and DAPI positive).

Senescence-associated b-galactosidase activity
Cells were stained by using Senescence Beta-Galactosidase Staining Kit following 11 days of shRNA (9860, Cell Signaling Technology) according to the manufacture's protocol. Cultures were examined under phase-contrast microscopy to determine the percentage of positive cells with blue precipitate.

Xenograft experiments
All animal experiments were conducted under a UT-Southwestern IACUC approved protocol. Hsd: Athymic Nude-Foxn1 nu female mice were purchased from Envigo. At six to eight-weeks of age (23-25 grams) mice were randomly assigned to the indicated groups. Sample size was determined by the formula: n = 1+2C(s/d) 2 , in which C = 10.51, based on a power of 90% and a confidence interval of 95%, S = standard deviation of tumor volume (20%), D = tumor burden difference to be considered significant (50%). Therefore, five mice per condition were tested. 2 Â 10 6 cells in 100 ml PBS were subcutaneously injected in the flank. Once tumors were visible, tumor volume was measured by caliper in two dimensions, and volumes were estimated using the equation V = length x width 2 /2. Caliper measurements were made twice a week.

Statistical analysis
Graphpad Prism (Graphpad Software) was used to perform all statistical analyses. Data were assessed for normality by Shapiro-Wilk tests. Unpaired Student's t-test was used for data with a normal distribution. Mann-Whitney test was used for data that did not conform to a normal distribution. p-Values less than or equal to 0.050 were considered significant. Biological replicates represent experiments performed on samples from separate biological preparations; technical replicates represent samples from the same biological preparation run in parallel.