The opposite prognostic effect of NDUFS1 and NDUFS8 in lung cancer reflects the oncojanus role of mitochondrial complex I

A recent surge of research on complex I mitochondrial DNA indicates that complex I disassembly regulated by mutation threshold plays a critical role in tumor progression. However, nuclear DNA (nDNA)-encoded core subunits are still a neglected area for cancer investigation. In this study, respective prognostic contributions of 7 nDNA-encoded core subunits were analyzed by immunohistochemical staining and RNA expression data extracted from public resources. The results showed that NDUFS1 and NDUFS8 had the most significant prognostic power in NSCLC patients among all 7 nDNA-encoded core subunits. Patients with low NDUFS1 or high NDUFS8 IHC and RNA expression levels had poor overall survival. Because of the significant correlation between expressions of 7 nDNA-encoded core subunits, multivariate analysis was performed and identified NDUFS1 and NDUFS8 IHC and RNA expression levels retained their leading prognostic roles. By combining NDFUS1 and NDUFS8 as a panel, the most unfavorable prognostic group had a 14-fold increased risk of poor prognosis than the most favorable prognostic group. In conclusion, the opposite prognostic effect of nDNA-encoded core subunits suggests the oncojanus role of nuclear genes regulating complex I dysfunction. The panel with NDUFS1 and NDUFS8 reflecting tumor metabolism status is a novel prognostic predictor for lung cancer.

A cancer cell with its rapid proliferative and aggressive behavior is regarded as a highly metabolic and energy demand biosystem compared to normal cell 1 . Since Warburg effect which stated that cancer cells produce energy predominantly by aerobic glycolysis instead of oxidative phosphorylation proposed over several decades ago, tumor metabolism became a rising field in cancer research 2,3 . Mitochondria as the energy factory of the cell is the main place of various energy producing processes and plays an important role in tumor metabolism 4 . From tricarboxylic acid cycle to oxidative phosphorylation, dysfunction of steps in mitochondrial biogenesis cascade alters the balance of intracellular environment during cancer tumorigenesis and progression 5 .
Mitochondrial complex I (NADH dehydrogenase), the first entrance step of oxidative phosphorylation, generates mitochondrial membrane potential for ATP production by converting NADH to NAD + and produces reactive oxygen species (ROS) as byproducts 6 . Complex I as the largest complex in mitochondrial respiratory chain composes of 45 subunits. Among them, 14 subunits form core subunits of complex I, which consists of 7 mitochondrial DNA (mtDNA)-encoded core subunits and 7 nuclear DNA (nDNA)-encoded core subunits 7 . Complex I deficiency caused by complex I gene mutation is known to be the culprit of mitochondrial disorder which leads to neuromuscular symptoms and various clinical presentations 8 . Since cancer is also considered as a mitochondrial disease in a broad definition, there has been a tremendous wave of interest in the relationship between complex I dysfunction and cancer in recent years 9 . Alteration of NAD+/NADH caused by dysfunction of complex I, in turn, activates Akt pathway and inhibits autophagy and results in cancer progression 10,11 . Restoring complex I activity or normalizing the NAD+/NADH balance was shown to inhibit cancer metastasis and tumor progression 11 . However, studies focusing on mitochondrial DNA (mtDNA) mutation demonstrated conflicting results and further expanded the debate on the functional role of complex I in cancer. mtDNA mutations are commonly seen in various kinds of cancer due to its vulnerable to the oxidative damage and carcinogens 12,13 . Accumulated Results NDUFS1 and NDUFS8 have the leading prognostic relevance among 7 nDNA-encoded core subunits in NSCLC patients. To evaluate the importance of nDNA-encoded core subunits in lung cancer, we first analyzed the prognostic values of all 7 nDNA-encoded core subunits through immunohistochemistry (IHC) analysis. Interestingly, our results showed that there were differences in prognostic predictive power between 7 nDNA-encoded core subunits. Among all 7 nDNA-encoded core subunits, only NDUFS1 and NDUFS8 IHC expression reached statistical significant correlation with prognosis ( Fig. 1). Patients with low IHC expression of NDUFS1 (P = 0.000001) and high IHC expression of NDUFS8 (P = 0.000198) were found significantly associated with poor overall survival. We further validate the results of our IHC analysis by using Kaplan-Meier plotter web resource. Consistent results were also observed in the survival analysis results of public microarray database. Most significant correlations were observed between low NDUFS1 (P < 0.000001) and high NDUFS8 (P < 0.000001) RNA expression and poor overall survival (Fig. 2). Patients with high NDUFV1 (P = 0.008), NDUFV2 (P = 0.000001), NDUFS3 (P = 0.000047), and NDUFS7 (P = 0.002) RNA expression levels also had poor prognosis values (Fig. 2).
Only NDUFS1 and NDUFS8 remain their independent prognostic significances in multivariate analysis. Considering that NDUFS1, NDUFS2, NDUFS3, NDUFS7, NDUFS8, NDUFV1, and NDUFV2 being core subunits of mitochondrial complex I, we further analyzed the correlation between their IHC expression levels in clinical NSCLC specimens. Through Spearman's rank correlation analysis (Supplementary Table 2 (Table 1). After false discovery rate correction for multiple testing, NDUFS1 (P = 0.01) and NDUFS8 (P = 0.04) also retained their prognostic significance (Table 1).

NDUFS1 IHC expression has significant correlation with T stage, distant metastasis, and pathological stage. Clinicopathological analysis of NDUFS1 and NDUFS8 indicated that low NDUFS1
IHC expression were significantly correlated with higher T stage (P = 0.001), distant metastasis (P = 0.040) and higher pathological stage (P = 0.016) ( Fig. 3A and Supplementary Table 4). However, no significant association between NDUFS8 IHC expression and clinicopathological parameters was seen (Supplementary Table 4). IHC images in Fig. 3B representatively shown loss of NDUFS1 IHC expression in adenocarcinoma and squamous cell carcinoma patients with higher stages.
Combining NDUFS1 and NDUFS8 IHC expression as a panel reflects the oncojanus role of complex I for prognosis prediction. The opposite prognostic effect of NDUFS1 of NADH dehydrogenase module and NDUFS8 of hydrogenase module revealed in our survival analysis is the clinical evidence of the oncojanus role of complex I. Figure 3C showed representative NDUFS1 and NDUFS8 IHC staining images of two NSCLC cases. Patient 1 had strong NDUFS1 IHC staining and weak NDUFS8 IHC staining and patient 2 with weak NDUFS1 expression and strong NDUFS8 expression. Therefore, we combined NDUFS1 and NDUFS8 IHC expression as a prognostic panel which could further separate patients into 3 groups (Fig. 3D). Patients with low NDUFS1 and high NDUFS8 expression had worst prognosis at both IHC (P < 0.000001) and RNA level , and NDUFV2 in NADH dehydrogenase module and NDUFS2, NDUFS3, NDUFS7, and NDUFS8 in hydrogenase module, low NDUFS1 RNA expression level and high NDUFS8 RNA expression level are most significantly associated with poorer overall survival (P < 0.000001). High NDUFV1 RNA expression level (P = 0.008), high NDUFV2 expression level (P = 0.000001), high NDUFS3 expression level (P = 0.000047), and high NDUFS7 expression level (P = 0.002) are also correlated with poor overall survival.
(P < 0.000001). In multivariate analysis, compared to the most favorable prognostic group (high NDUFS1 and low NDUFS8 IHC expression levels), the most unfavorable prognostic group (low NDUFS1 and high NDUFS8 IHC expression levels) showed a 14-fold increased risk of a poor prognosis after adjustment for the pathological stage (Table 3).

Discussion
Our study evaluated the expression status of 7 nDNA-encoded core subunits in clinical lung cancer patients and the results revealed that different subunits have different or even opposite prognostic effect on lung cancer prognosis. The significant predictive powers of low NDUFS1 expression and high NDUFS8 expression on poor prognosis are the clinical evidence of the oncojanus role of complex I. After adjustment for pathological stage and the correlation between all 7 nDNA-encoded core subunits in multivariate analysis, NDUFS1 and NDUFS8 still are independent prognostic markers for overall survival. With the leading prognostic impact, low NDUFS1 immunoexpression level correlated with higher T stage, distant metastasis, and higher pathological stage in the clinicopathological analysis. Taking together, the panel combining NDUFS1 and NDUFS8 expression could be used to predict the risk of poor prognosis more precisely by reflecting the metabolism status of lung cancer.
Mitochondria has long been considered to have a critical role in cancer progression through regulating the metabolism of rapidly growing tumors 4 . Mitochondrial dysfunctions result from mutations in mitochondrial DNA or nuclear-encoded mitochondrial genes exist in various cancer types 16 . Although not as extensively investigated as mitochondrial DNA mutations, nuclear-encoded mitochondrial gene mutations have gradually been shown to possess cancer-promoting properties 17 . For instance, mutations in succinate dehydrogenase, complex II of oxidative phosphorylation subunits, have association with tumorigenesis of paraganglioma, pheochromocytoma, kidney cancer, gastrointestinal stromal tumor, and breast cancer [18][19][20] . Besides, mutations in isocitrate dehydrogenase (IDH), an enzyme of tricarboxylic acid cycle, are the most frequent mutations in glioma and can also be found in many kinds of cancer 21 . However, there have been few attempts to clarify the link between complex I nDNA-encoded core subunits and cancer. According to our knowledge, current research is the first study focusing on the role of complex I nDNA-encoded core subunits in lung cancer and unraveled the prognostic power of complex I nDNA-encoded core subunits in cancer patients. Our findings revealed that nDNA-encoded subunits are a critical issue, though it has a relative little discussion in the considerable amount of research in cancer metabolism. A few recent studies on breast cancer shed some light on the relation between complex I nDNA-encoded core subunits and cancer progression. The expression level of NDUFS3 was shown to be positively correlated with nuclear grade in invasive ductal carcinoma and significantly higher in cancer than in  11 . This contrary function among the members of nDNA-encoded core subunits also fell into the long debate of oncogenic or tumor suppressive role of mitochondrial dysfunction in cancer 23,24 . Different prognostic powers of nDNA-encoded core subunits in our results provided further evidence for the concept of the oncojanus role of complex I dysfunction 25 . The plausible explanation of this two-sided effect may lie on the structure assembly process of complex I. Research on mtDNA induced complex I disassembly have shown that complex I is essential for cancer cells to enter the process of Warburg effect 26 . Dissembling destruction of complex I could not maintain the basic energy demand for alive in cancer cells and results in an anti-tumorigenic effect. The assembly of complex I composed of 45 subunits is a step by step process in which several subcomplexes enter into the processing by different stage 27,28 . Seven nDNA-encoded core subunits could be cluster into 2 functional modules. NADH dehydrogenase module (N module) including NDUFS1, NDUFV1, and NDUFV2 mainly functions on oxidizing NADH and hydrogenase module (Q module) including NDUFS2, NDUFS3, NDUFS7, and NDUFS8 mainly functions on reducing ubiquinone 6 . N module was proposed to enter the assembly process in late stage and Q module enters the process in early stage 29 . Therefore, defects in N module may cause complex I dysfunction without disassembly and lead to cell survival and proliferation. In contrast, defects in Q module may result in complex I disassembly and cell death.
The most powerful independent prognostic effect of NDUFS1 shown in our study reveals the fact that not all nDNA-encoded core subunits have equal prognostic significance. The complicated structure of complex I could also be the explanation for this finding. NDUFS1 as the largest subunit of complex I, its mutations can cause about 80% decrease of complex I activity which implies that NDUFS1 is critical for the stability and function of complex I 30 . Accordingly, loss of NDUFS1 may disrupt the native NADH homeostasis function of complex I and lead to cancer progression. On the other hand, an interesting point worth mention is that, in addition to its native function in tricarboxylic acid cycle, mitochondrial enzyme fumarate hydratase could promote tumor formation through succination of other proteins because of its unique chemical structure 31,32 . Although there is no current evidence to suggest that nDNA-encoded core subunits could modulate tumor progression through NADH independent pathway, this possibility cannot be excluded before further investigation.
In conclusion, direct evaluation of the expression levels of nDNA-encoded core subunits in cancer tissue in our study provided the clinical evidence of the oncojanus role of complex I dysfunction, and also revealed that the oncojanus role of complex I dysfunction not only depends on mtDNA but also relies on the regulation of nDNA. NDUFS1 and NDUFS8 as the leading prognostic factors in 7 nDNA-encoded core subunits can be a useful panel for prognosis prediction and therapeutic decision in lung cancer patients.

Materials and Methods
Patient and ethics statement. One hundred and one NSCLC formalin-fixed paraffin embedded tissue specimens including 62 cases of adenocarcinoma, 32 cases of squamous cell carcinoma, and 7 cases of large cell carcinoma were retrieved from pathology department archive of Kaohsiung medical university Hospital (Kaohsiung, Taiwan) between 1991 and 2007. All cases were staged according to American Joint Committee on Cancer (AJCC) staging manual and histological cancer types were classified by World Health Organization (WHO) classification. According to the hospital treatment guideline, patients with resectable stage I-III NSCLC received lobectomy or pneumonectomy with mediastinal lymphadenectomy. Patients with resectable stage II and III NSCLC received postoperative adjuvant platinum-based chemotherapy. No adjuvant chemotherapy was administered in patients with completely resected stage I NSCLC. Inoperable locally advanced or metastatic cases were treated with chemotherapy with or without radiotherapy. Overall survival (OS) was defined as the interval from the initial treatment time to death. Clinical information was obtained by retrospectively reviewing medical record. The demographics details of patients are shown in supplementary Table 1. To construct the tissue microarray (TMA), representative 1-mm-diameter core from each case was taken and selected by morphology typical of the diagnosis.
This study was approved by the ethics committees of Institutional Review Board of Kaohsiung Medical University Chung-Ho Memorial Hospital (KMUH-IRB-20110286) and was carried out in accordance with the approved guidelines. No informed consent was required because the data were analyzed anonymously and no identifying information relating to participants were included.
Public database. Public microarray database was used to validate the results from our cohort and to do further analysis. Survival analysis of large cohort microarray database with a total of 1145 NSCLC patients was performed using Kaplan-Meier plotter web resource (http://kmplot.com/analysis/) which includes The Cancer Genome Atlas (TCGA) cohort and multiple GSE datasets from Gene Expression Omnibus (GEO). Raw data including RNA expression levels, survival time, and survival status were extracted from the website for correlation analysis and cox univariate and multivariate analysis.
Statistical analysis. Statistical analysis was performed with SPSS 20 software (SPSS, Chicago, Illinois, USA).
The prognostic impacts of 7 nDNA-encoded core subunits were evaluated by the Kaplan-Meier method and compared by the log-rank test. Follow-up time was censored if the patient was lost during follow-up. Cox univariate and multivariate proportional hazards regression models were also performed with and without the adjustment of the expression levels of 7 nDNA-encoded core subunits and pathological stage. The association between clinicopathological parameters and NDUFS1 and NDUFS8 IHC expression were analyzed by Pearson's chi-square test. The correlation between the expression levels of 7 nDNA-encoded core subunits was evaluated by Spearman's rank correlation analysis. For all analyses, a P value of <0.05 was considered statistically significant.