COX5B-Mediated Bioenergetic Alterations Modulate Cell Growth and Anticancer Drug Susceptibility by Orchestrating Claudin-2 Expression in Colorectal Cancers

Oxidative phosphorylation (OXPHOS) consists of four enzyme complexes and ATP synthase, and is crucial for maintaining physiological tissue and cell growth by supporting the main bioenergy pool. Cytochrome c oxidase (COX) has been implicated as a primary regulatory site of OXPHOS. Recently, COX subunit 5B (COX5B) emerged as a potential biomarker associated with unfavorable prognosis by modulating cell behaviors in specific cancer types. However, its molecular mechanism remains unclear, particularly in colorectal cancers (CRCs). To understand the role of COX5B in CRCs, the expression and postoperative outcome associations using independent in-house patient cohorts were evaluated. A higher COX5B tumor/nontumor expression ratio was associated with unfavorable clinical outcomes (p = 0.001 and 0.011 for overall and disease-free survival, respectively. In cell-based experiments, the silencing of COX5B repressed cell growth and enhanced the susceptibility of CRCs cells to anticancer drugs. Finally, downstream effectors identified by RNA sequencing followed by RT-qPCR and functional compensation experiments revealed that the tight junction protein Claudin-2 (CLDN2) acts downstream of COX5B-mediated bioenergetic alterations in controlling cell growth and the sensitivity to anticancer drugs in CRCs cells. In conclusion, it was found that COX5B promoted cell growth and attenuated anticancer drugs susceptibility in CRCs cells by orchestrating CLDN2 expression, which may contribute to unfavorable postoperative outcomes of patients with CRCs.


Introduction
Colorectal cancers (CRCs) are among the most common malignancies worldwide as well as in Taiwan. The majority of CRCs are developed from adenomatous polyps named as the adenoma-carcinoma sequence, which eventually leads to metastasis [1]. In recent decades, bioenergetic alterations have been implicated in numerous types of cancers [2], including in CRCs [3][4][5], through their capability to control tumorigenesis and/or progression, thereby emerging as potent targets for anti-cancer treatments [6][7][8].
Bioenergy originates from anaerobic lactate fermentation and aerobic glycolysis as well as mitochondrial oxidative phosphorylation (OXPHOS) [9]. In normal cells under aerobic conditions, bioenergy is mainly generated through OXPHOS. Nearly a century ago, Otto Warburg discovered that cancer cells secrete more lactate than normal cells under aerobic conditions, indicating their higher glucose usage [10]. Although impaired 1:10000 dilution), and the rabbit polyclonal antibody against GDAP1 (Abclonal, Cat. A6601, 1:2000 dilution), and SerpinB8 (Abclonal, Woburn, MA, USA, Cat. A13039, 1:4000 dilution) were used for Western blot analysis in this study.

Immunohistochemical Staining (IHC)
The IHC was conducted as procedures reported previously [26]. The rabbit monoclonal antibody against COX5B (abcam, Cambridge, UK, Cat. ab180136), and rabbit polyclonal antibody against CLDN2 (Abclonal, Cat. A6560) were used for IHC staining in 1:200 dilution. The intensity of staining signals was acquired by Image J (National Institutes of Health, Bethesda, Maryland, USA, Fiji version) and used for subsequent analysis.

Seahorse Assay
Simultaneous measurement of oxidative consumption rate (OCR), indicator of OX-PHOS activity and extracellular acidification rate (ECAR), indicator of glycolysis activity, in either patient-derived tissues or cultured cells was conducted using the Seahorse XF24 analyzer (Agilent, Santa Clara, CA, USA) as previously described [3]. OCR and ECAR were reported as absolute rates normalized by the concentration of proteins extracted from tissues (mMol/min/mg for OCR while mpH/min/mg for ECAR) or from cultured cells (pMol/min/µg for OCR while mpH/min/µg for ECAR).

Cell Culture and Transfection
HT-29 and WiDr were used and maintained, respectively, in RPMI-1640 and DMEM mediums under standardized culture conditions with 5% CO 2 in a humidified 37 • C incubator in this study. Both of the cell lines were routinely examined for mycoplasma contamination. For knockdown of COX5B, the smart pool siRNAs (Dharmacon, Lafayette, CO, USA, M-013632), siRNA 1: CGACUGGGUUGGAGAGGGA, siRNA 2: GAGCAC-CUGCACUAAAUUA, siRNA 3: GGGACUGGACCCAUACAAU and siRNA 4: GAGAA UAGUAGGCUGCAUC, were used. The non-targeting pool included four independent scramble siRNAs, UGGUUUACAUGUCGACUAA, UGGUUUACAUGUUGUGUGA, UG-GUUUACAUGUUUUCUGA and UGGUUUACAUGUUUUCCUA, and was used as control siRNA (Dharmacon, D0018101020). The Lipofectamine RNAiMAX transfection reagent (Invitrogen, Waltham, MA, USA, 13778) was employed according to the protocol provided by the manufacturer. The plasmids capable of expressing COX5B and CLDN2 with myc-DDK tag were purchased from Origene (Rockville, MD, USA, Cat. RC202511 for COX5B and RC229728 for CLDN2). For transient expression of COX5B and/or CLDN2, the cells were seeded 16 h before transfection. Five µg of plasmid DNA was used for transfection in a 6 cm plate. The MaestroFectin (Omics Bio, New Taipei City, Taiwan, Cat. MF002) transfection reagent was used for transfection, according to the procedures provided by the manufacturer.

Cytochrome c Oxidase (COX) Activity Measurement
To measure the activity of COX under changes to COX5B in CRCs cells, the cytochrome c oxidase assay kit (Sigma-Aldrich, Cat. CYTOCOX1) was employed. The mitochondrial extracts were isolated using a mitochondria isolation kit (Sigma-Aldrich, Cat. MITOISO2). All the experiments were conducted according to the instructions provided by the manufacturer.

Cell Proliferation and Viability Assay
The cell proliferation assay was conducted as reported previously [27]. Briefly, 3 × 10 3 cells were seeded in each 96-well plate. After at least 16 h post-seeding, the Alamar Blue cell viability reagent (Invitrogen, Cat. DAL1100) was directly supplemented to the culture medium 3 h before quantification of the fluorescence of the metabolite for day 1. Then, the quantification was measured every day until day 4. The cell viability in response to anticancer drugs was initiated from seeding 1 × 10 4 cells in each well of a 96-well plate at least 16 h before treatments. The cells with 5-FU or Oxaliplatin supplemented were incubated at 37 • C incubator. After 24 h incubation, the medium of each well was refreshed with Alamar Blue cell viability reagent added. The quantification of the fluorescence of the metabolite was conducted after 3 h incubation at 37 • C.

Transcriptomic Analysis
The transcriptomic profiling was conducted using RNA-sequencing (RNA-seq). A total of 2 µg total RNA was subjected into RNA-seq analysis. The RNA samples passed through the quality assessment by Agilent 2100 Bioanalyzer were then used for library construction according to the manufacturer's instructions (Illumina, San Diego, CA, USA, Cat: 20020596). The sequencing was conducted using a NovaSeq 6000 system (Illumina) with a paired-end 150 bp program. The cleaned reads were aligned to human genome GRCh38 to obtain annotations for subsequent analysis of differentially expressed genes.

Real-Time Quantitative Polymerase Chain Reaction (RT-qPCR)
The total RNA isolation was conducted as previously described [28]. The ToolScript MMLV RT kit (BIOTOOLS, New Taipei City, Taiwan, Cat. TGKRA04) was used for first strand cDNA synthesis for up to 5 µg total RNA. The QuantStudio 5 real-time PCR system was employed. The primers used in this study are listed in Table S1.

Statistical Analysis
Parametric data in normal distribution was presented as mean ± standard deviation and compared by Student's t-test. Dichotomized data were presented as numbers and percentages (%) and compared by utilizing the Chi-square test. Univariate and following multivariate Cox proportional hazard models were performed to estimate survivals for clinical factors and other variables. In this study, significant factors determined from the univariate analysis could be included for multivariate Cox proportional hazards. The Kaplan-Meier method was performed to estimate the survival probabilities between groups, and the log-rank test was performed to compare the survivals. All tests were two-tailed, and p < 0.05 was considered statistically significant. All statistical analyses were performed using the Statistical Package for the Social Sciences (SPSS) statistics (IBM, Armonk, NY, USA, Version 20).

Higher COX5B Expression Levels in CRCs Correlate with Poorer Clinical Outcomes
To understand whether COX5B is involved in the oncogene-like properties of CRCs, its expression was assessed in patient-derived tissues. As shown in Figure 1A, unlike in HCC, the level of COX5B was significantly downregulated in tumorous tissues (p < 0.001). According to the ratio of tumorous/nontumorous (T/N ratio) of COX5B expression, approximately one-third of patients (40 of 126) exhibited COX5B T/N ≥ 1. Comparison of the baseline characteristics between subgroups with COX5B T/N < 1 and ≥ 1 are listed in Table S2. Patients with COX5B T/N ≥ 1 had greater height (p = 0.021), tumors located within the proximal colon (p = 0.007), tumors with good differentiation (p = 0.016), a shorter distance from the tumor to the serosa (p = 0.048), and a higher oxygen consumption rate OCR T/N ratio (p = 0.027). The subgroups of patients were separated according to the COX5B T/N ratio ≥ 1 or < 1 obtained from the staining intensities of IHC. (E,F) The OCR and ECAR in either non-tumorous (N or NT) or CRCs tumorous (T) tissues. The p values in the left panels were acquired by two-tailed paired Student's t-test, while two-tailed unpaired Student's t-test was used for the middle panel. *, p < 0.05; ***, p < 0.001. Pearson correlation analysis was employed for the right panel.
Taken together, one-third of patients, those with higher COX5B expression in the tumorous area (COX5B T/N ≥ 1), exhibited associations with unfavorable clinical outcomes, suggesting that COX5B plays an oncogenic role in CRCs.

Higher COX5B Expression Levels Correlate with Increased OCR in CRCs
As a subunit of COX, a key modulator of OXPHOS, it was hypothesized that changes in COX5B expression disturbed bioenergetic homeostasis in CRCs. To test this hypothesis, the OCR and ECAR in the same cohort as shown in Figure 1A were measured in surgically resected tissues. As shown in Figure 1E left panel, the OCR was markedly decreased in the tumorous tissue (p < 0.001). Separating the patients into two groups according to the COX5B T/N ratio, T > N (n = 40) or T < N (n = 86), revealed an evident increase in the OCR T/N in those with higher COX5B T/N ( Figure 1E middle panel, p < 0.001). Pearson correlation analysis also demonstrated that the COX5B T/N ratio was positively correlated with the OCR T/N in CRCs tissues ( Figure 1E right panel, r = 0.727, p < 0.001).
In contrast, ECAR was significantly increased in the tumorous area ( Figure 1F left panel, p = 0.035), whereas no obvious difference was observed between subgroups of patients with COX5B T > N and T < N ( Figure 1F left panel, p = 0.178). In addition, COX5B T/N and ECAR T/N showed no significant correlation in tissues derived from patients with CRCs ( Figure 1F right panel, r = 0.163, p = 0.069).
These results suggested that altered COX5B expression primarily disturbed OXPHOS activity but had only limited effects on promoting the activation of glycolysis in tissues from patients with CRCs.

Higher COX5B Expression Levels Correlate with Increased OCR in the Pre-Cancerous Tissues
To test whether elevated COX5B T/N is also correlated with a higher OCR T/N in the tissues under the pre-cancerous state of adenomatous polyps, the assays shown in Figure 1 were conducted. The expression status of COX5B was analyzed using Western blotting ( Figure 2A); the quantitative results are shown in Figure 2B. As observed in CRCs, COX5B expression levels were significantly decreased in the tissues of adenomatous polyps (p < 0.001), suggesting that downregulation of COX5B occurred in tissues during the pre-cancerous state.
To investigate the correlation between COX5B expression and OCR or ECAR in precancerous tissues, freshly acquired tissues were obtained and subjected into Seahorse assays. As shown in Figure 2C left panel, there was no obvious change in the OCR between polyp and non-polyp tissues (p = 0.569). Among all patients, 18 of 64 (around one-third of all patients) showed a COX5B polyp/non-polyp (P/N) ratio ≥ 1, whereas the remaining 46 patients exhibited COX5B P/N < 1. Interestingly, a higher OCR P/N was associated with increased COX5B expression in adenomatous polyps (P/N ≥ 1) ( Figure 2C middle panel, p = 0.006). Pearson correlation analysis also revealed a positive correlation between COX5B expression P/N and OCR P/N, indicating that higher COX5B expression is associated with elevated OCR in the polyps.
In contrast, ECAR was significantly increased in the polyps ( Figure 2D left panel, p = 0.031). However, the COX5B P/N was not associated with the ECAR P/N ( Figure 2D middle panel, p = 0.920). In addition, Pearson correlation analysis showed that COX5B P/N did not correlate with the ECAR P/N in tissues derived from patients with adenomatous polyps ( Figure 2D right panel, r = 0.151, p = 0.093).
These findings suggest that alterations in COX5B expression and disturbance of OX-PHOS activity observed in CRCs originated from the pre-cancerous state. The OCR and ECAR in either normal mucosa (N) and polyp (P) tissue. The p values in the left panels were acquired by two-tailed paired Student's t-test, while two-tailed unpaired Student's t-test was used for the middle panel. *, p < 0.05; **, p < 0.01; ***, p < 0.001. The Pearson correlation analysis was employed for the right panel.

Change in COX5B Expression Influences Bioenergetic Alterations, Cell Growth, and Susceptibility to 5-Fluorouracil and Oxaliplatin in CRCs Cells
To investigate the impact of altering COX5B expression in CRCs, COX5B was knocked down or overexpressed in CRCs cells, including HT29 and WiDr cells ( Figure 3A). As a crucial factor for COX, silencing of COX5B suppressed COX activity, whereas the gainof-function of COX5B ameliorated COX activity in CRCs cells ( Figure 3B). Furthermore, downregulation of COX5B significantly repressed OXPHOS activity but had only limited effects on modulating the glycolysis capacity ( Figure 3C). In contrast, an increase in COX5B expression markedly elevated the OXPHOS capacity but did not change glycolysis activity ( Figure 3D).
To determine the effects of alterations of COX5B expression in CRCs on cell growth, a cell proliferation assay was conducted. As was shown in Figure 3E, silencing of COX5B significantly attenuated the cell proliferation rate, whereas elevation of COX5B increased the cell renewal of CRCs cells. In addition, to examine whether alterations in COX5B expression affect the susceptibility of CRCs cells to the frequently employed anticancer drugs 5-fluorouracil (5-FU) and oxaliplatin, a drug sensitivity assay was conducted following silencing or upregulation of COX5B. As shown in Figure 3F, attenuating COX5B expression significantly sensitized CRCs cells to both 5-FU and oxaliplatin, whereas the opposite effects were observed in cells with elevated COX5B expression. Staining of the terminal deoxynucleotidyl transferase dUTP nick-end labeling (TUNEL)-positive cells supported these results ( Figure S1).
Taken together, these findings partially reflect the findings in clinical patients showing that elevated COX5B expression levels led to bioenergetic alterations in CRCs and may modulate cell renewal and susceptibility to anticancer drugs, which may finally lead to recurrent disease and a poor prognosis.
As repressing COX5B expression may lead to either reduced OXPHOS activity or induced reactive oxygen species (ROS) accumulation ( Figure S2), chemical drugs, including H 2 O 2 , oligomycin, antimycin A, and NaN 3 , were added into the growth medium to mimic the attenuation OXPHOS and accumulation of ROS to further understand the main source required for regulating the expression of these genes. As shown in Figure 4C, after treatment with H 2 O 2 , the mRNA levels of TRIM16L, SEMA7A, and ALDH1A3 were upregulated, indicating that these genes were modulated by elevated ROS. However, none of the candidates predicted to be downregulated were reduced after treatment with H 2 O 2 .
Oligomycin, antimycin A, and NaN 3 directly target components in OXPHOS and thus lead to bioenergetic alterations [12]. As shown in Figure 4D, the mRNA levels of only GDAP1, CLDN2, and SERPINB8 consistently decreased after treatment with these OXPHOS-targeted chemical drugs, suggesting that these genes function downstream of COX5B-mediated bioenergetic alterations.
To determine whether these changes also occurred at the protein level after silencing of COX5B and after treatment with these chemical drugs, the Western blot analysis of samples collected after the indicated treatments was conducted. Except for Claudin-2 (CLDN2), the protein levels of GDAP1 and SERPINB8 were not attenuated as predicted when COX5B was silenced ( Figure 4E) or following treatment with OXPHOS-targeted chemical drugs ( Figure 4F). Therefore, only the CLDN2 showed promise as a downstream protein of COX5Bmediated bioenergetic alterations. Notably, cells with NaN 3 -treated, which directly targets COX complex activity, showed significantly alleviated COX5B expression ( Figure 4F).
Taken together, these findings indicate that expression of the tight junction localized protein CLDN2 [25] is a promising downstream effector orchestrated by COX5B-mediated bioenergetic alterations in CRCs cells.

Association of COX5B and CLDN2 Expression in Tissues from Patients with CRCs
To assess whether the identified regulatory event involving CLDN2 downstream of COX5B-mediated bioenergetic alterations also occurred in samples derived from clinical patients with CRCs, IHC staining of both COX5B and CLDN2 on slide sections from these patients was conducted. As shown in Figure 5A, samples with a COX5B T/N ratio of staining intensities ≥ 1 were correlated with elevated CLDN2 T/N intensities, whereas the opposite results were observed in samples with COX5B T/N < 1. Comparisons of the COX5B and CLDN2 T/N ratios between subgroups with COX5B T/N < 1 and ≥ 1 are shown in Figure 5B. Further correlation of COX5B and CLDN2 staining intensities using Pearson correlation analysis revealed no correlation in the non-tumor sections ( Figure 5C, r = 0.045, p = 0.692), but marked correlations in the tumor sections ( Figure 5D, r = 0.557, p < 0.001). In addition, the COX5B and CLDN2 T/N ratios showed strong positive correlations ( Figure 5E, r = 0.695, p < 0.001). Taken together, these findings support that CLDN2 is a potent downstream effector of COX5B-mediated bioenergetic alterations in CRCs, because they showed significantly correlated expression patterns.

CLDN2 Functions Downstream of COX5B to Modulate Ell Growth and Susceptibility to 5-FU and Oxaliplatin
To test whether CLDN2 act as a downstream effector of COX5B in controlling cellbased phenotypes ( Figure 3E,F), CRCs cells with simultaneous silencing of COX5B and overexpression of CLDN2 ( Figure 6A) were utilized to assess cell growth activity and the sensitivity to anticancer drugs. There was a positive feedback loop between COX5B and CLDN2, as overexpression of CLDN2 slightly enhanced COX5B expression, whereas silencing of COX5B repressed CLDN2 expression ( Figure 6B). As shown in Figure 6C, overexpression of CLDN2 relieved the growth-suppressive effect caused by the silencing of COX5B in CRCs cells. Additionally, the susceptibility of CRCs cells to 5-FU and oxaliplatin was compensated when CLDN2 expression was elevated after silencing of COX5B ( Figure 6D,E). The results of TUNEL staining also agreed with this notion ( Figure S3). These findings demonstrate that the tight junction protein CLDN2 serves as a downstream effector of COX5B-mediated bioenergetic alterations in orchestrating cell growth and susceptibility to anticancer drugs in CRCs cells.

Discussion
An increasing body of evidence from recent decades has supported that bioenergetic homeostasis is crucial for tumorigenesis and/or the progression of many cancer types, including HCC and CRCs [2][3][4][5][6][7][8]24]. COX5B has been reported as a growth-promoting gene that modulates downstream pathways in response to the induced bioenergetic alterations in HCC, breast cancer, and glioma, but its role in CRCs remains unknown [24,[29][30][31]. As such, clinical associations and experimental assays were conducted to investigate the role and potential mechanism of action of COX5B in CRCs.
COX5B has also been implicated as a predictor of clinical outcomes in the various cancer types, including HCC, breast cancer, glioma, gastric cancer, head and neck squamous cell carcinoma (HNSCC), and clear cell renal cell carcinoma (ccRCC) [24,[29][30][31][32]. In HNSCC, COX5B was reported as a potent tumor-suppressive gene, whereas in almost all other cancer types, this protein exhibited a growth-promoting property, and its higher levels were correlated with unfavorable clinical outcomes. Similar to these previous findings, our clinical association studies revealed that albeit COX5B was downregulated in around twothirds of patients with CRCs, a higher COX5B T/N was associated with unfavorable clinical outcomes and an increased OCR T/N ( Figure 1 and Table 1). Additionally, the correlation between COX5B expression and bioenergetic alterations could be traced to as early as the origin of the adenomatous-carcinoma sequence in the adenomatous polyps ( Figure 2). Furthermore, increased COX5B expression in cells, a condition mimicking the situation found in one-third of patients with CRCs who had elevated COX5B expression in the tumorous tissue, promoted cell growth and reduced the susceptibility to anticancer drugs through orchestrating the COX-mediated bioenergetic alterations; experimentally reduced levels of COX5B, which simulated the situation observed in two-thirds of patients with CRCs who showed downregulated COX5B expression in tumorous tissue, repressed cell growth and sensitized CRCs cells to anticancer drugs ( Figure 3). This evidence suggested that COX5B exerts a cancer-promoting role in CRCs, albeit its molecular mechanisms in modulating CRCs growth and susceptibility to anticancer drugs requires further investigation.
Theoretically, in nontumorous tissues, aerobic OXPHOS, which requires sufficient expression of COX5B, is primarily used to generate cellular bioenergy, but is not involved in glycolysis under normal conditions [9]. As a well-established phenotype, it is considered that bioenergetic production sources shift from OXPHOS to aerobic glycolysis in many cancer types [10]. The downregulation of COX5B in tumorous tissues observed in approximately two-thirds of patients with CRCs may support this fact. A possible mechanism is that the status of hypoxic conditions modulates OXPHOS activity by regulating COX subunits expression, including COX5B, in mammalian cells [33,34]. HIF1A expression is increased and directly represses the expression of COX subunits by binding to the promoter region [33], indicating that COX5B expression is repressed under hypoxic conditions. This may explain why approximately two-thirds of patients with CRCs had reduced COX5B expression in the tumorous section, as the hypoxic status in tumors is well known. However, why one-third of patients with CRCs had unchanged or even elevated COX5B remains unclear. This result may be related to the level of oxygen transported into the tumors. A higher level of oxygen may help to maintain or even enhance COX5B expression in tumors. As described previously, angiogenesis occurring in tumors can promote tumor growth by providing nutrients, including oxygen [35]. The results of previous studies and our study suggest that tumors with low COX5B expression are possibly in an early stage or have not developed angiogenesis; in contrast, tumors with high expression of COX5B may be in a late stage or be undergoing angiogenesis, which may impact the unfavorable prognoses of patients with CRCs.
Analysis of the underlying growth-modulating and anticancer drugs sensitivitycontrolling mechanisms of COX5B in CRCs revealed CLDN2 as a potential effector downstream of COX5B-mediated bioenergetic alterations (Figure 4). Subsequent experiments further confirmed this result ( Figure 5), supporting that COX5B is positively correlated with CLDN2 expression in CRCs. CLDN2 is a promising tight junction localized protein with well-characterized functions, including promoting the tumorigenicity, metastasis, and susceptibility of cells to 5-FU, perhaps through ameliorating cancer stem cell property in CRCs [36][37][38]. Our findings partially reflect that CLDN2 acts as an effector downstream of COX5B-induced control of cell proliferation and sensitivity to anticancer drugs ( Figure 6). Additionally, CLDN2 not only conferred resistance to 5-FU, but also contributed to reduction of the susceptibility to another frequently used anticancer drug, oxaliplatin, in CRCs.
Interestingly, elevation of CLDN2 upregulated COX5B expression ( Figure 6B), suggesting the presence of a positive feedback loop linking COX5B expression or bioenergetic homeostasis with CLDN2 levels in CRCs. This makes it worthy to be investigated in detail in the future.

Conclusions
In conclusion, we demonstrated that COX5B exerts oncogenic properties to ameliorate cell growth and repress susceptibility to anticancer drugs by modulating the expression of the oncogenic tight junction protein CLDN2 in CRCs, thereby contributing to the unfavorable prognostic outcomes of patients with elevated COX5B.