Mitochondria are complex organelles of bioenergetic, biosynthetic, and signaling that are correlated to several diseases, including cardiovascular diseases, neurological disorders, and metabolism disorders (Vyas et al., 2016; Genovese et al., 2020). In addition, recently, mitochondria are demonstrated to participate in multiple aspects of tumor formation and progression (Hsu et al., 2016; Bonora et al., 2021). The mechanisms of mitochondria involved in tumorigenesis have been extensively studied (Nunes et al., 2015), and some specific nuclear mitochondrial genes (NMGs) were regarded as the potential targets for the development of the next generation of cancer therapeutics (Su et al., 2016; Zong et al., 2016).

mtDNA is the own genome of mammalian mitochondria, 16.5- kb double-stranded circular DNA, which encodes a total of 13 proteins that are all core components of oxidative phosphorylation (OXPHOS). However, all remaining mitochondrial functions proteins (∼1,300) are encoded in the nuclear DNA (nDNA). Following their translation, they are imported into the organelle through the import machinery instead (Anderson et al., 1981; DiMauro and Schon, 2003; Schon and Przedborski, 2011). Thus, the mitochondrial function requires cooperation between the nuclear and mitochondrial genomes. Mitochondrial genome instability and mitochondrial dysfunction are novel markers for cancer (Wang et al., 2015; Idaghdour and Hodgkinson, 2017). Mitochondrial dysfunction, caused by smoking or other factors, plays a vital role in the carcinogenesis of multiple cancers, especially lung cancer (Cloonan et al., 2020). The dysfunction of mitochondria may shift the energy production from mitochondria-mediated oxidative to glycolysis, which in turn increases the cell metabolism and activity (Chen et al., 2021). Meanwhile, mitochondrial dysfunction causes ROS outbreak, which mediates epithelial-mesenchymal transition and cell invasion in lung cancer via coordinate intracellular signal transduction (Nunes et al., 2015; He et al., 2016; Han et al., 2018). mtDNA variations have been frequently suggested to serve as initiators for a variety of tumors, such as prostate cancer (Hopkins et al., 2017), LUAD (Yuan et al., 2015), breast cancer (Weerts et al., 2018), pancreatic cancer (Lam et al., 2012), gynecological malignancies (Guerra et al., 2014), and acute myeloid leukemia (Reznik et al., 2016; Tyagi et al., 2018). Contrary to the increased study on the association of mtDNA alterations and cancer, fewer studies have shed light on the role of NMGs on cancer. The germline or somatic alteration in nuclear mitochondrial complex II genes encoding SDH subunits (SDHB, C, and D) were found associated with pheochromocytomas and paragangliomas (Gimm et al., 2000; van Nederveen et al., 2007; Bardella et al., 2011). Similarly, one study found that the NMGs NDUFS1 and NDUFS8 (encoding subunits of mitochondrial complex 1) had significant prognostic power in the patients with non-small cell lung cancer (NSCLC) by analyzed immunohistochemical staining and RNA expression data (Su et al., 2016). Alterations in the genes mentioned above have also been detected in patients with mitochondrial disease, such as mitochondrial complex I deficiency and mitochondrial respiratory chain deficiency (Bourgeron et al., 1995; Loeffen et al., 1998; Björkman et al., 2015). Therefore, we examined the hypothesis that the NMGs may have significant predictive value in the identification of high-risk cancer patients with poor overall clinical survival.

Lung cancer is the leading cause of cancer-related mortality worldwide causing over 1.7 million deaths annually (Bray et al., 2018), of which LUAD are the major subtypes. In the present study, we investigate the NMGs associated with LUAD survival by bioinformatics analysis, following univariate Cox regression analyses to develop an NMGs signature to provide new clues for improving the diagnosis and treatment of patients with LUAD.

The pivotal roles of mitochondria in maintaining normal cell function as well as in cancer development are widely acknowledged. Dysfunction in some specific members of mitochondrial genes, especially FH, SDH family, and IDH1/2, have been demonstrated as the milestone event for the development of multiple cancers like gliomas (Hartmann et al., 2009), glioblastomas (Balss et al., 2008), acute myeloid leukemia (Paschka et al., 2010), pheochromocytoma (Astuti et al., 2001), paraganglioma (Niemann and Müller, 2000; Astuti et al., 2001), and papillary renal cell cancer (Tomlinson et al., 2002; Luo et al., 2020). However, the integrated role of NMGs in the prediction for the prognosis of lung cancer patients remains undefined.

In this study, we first identified 26 NMGs whose expression were correlated with LUAD patients’ survival. Among those genes, only seven genes (ATP8A2, CABC1, TK2, ANT1, RRM2B, COX14, and ISCU) were associated with a favorable outcome in LUAD patients. Although these genes were previously reported in some cancer types, their specific role in lung cancer is unclarified except for ATP8A2 and RRM2B. ATP8A2, encoded ATPase Phospholipid Transporting 8A2, belongs to the P4-ATPase family that actively flips phosphatidylserine and phosphatidylethanolamine from the exoplasmic to the cytoplasmic leaflet of cell membranes to generate and maintain phospholipid asymmetry (Choi et al., 2019). As this gene involves neurite elongation and neuron survival, various genomic variants of ATP8A2 are identified in patients with neurological disorders (Guissart et al., 2020). Though the relationship between ATP8A2 and lung cancer is poorly investigated, a recent study found the alteration in its expression was associated with the prognosis of LUAD (Wang et al., 2020). ATP8A2 was hypermethylated in tumor tissues and, therefore, poorly expressed in the tumor tissues compared with the normal tissues (Wang et al., 2020). The same as the previously published result by Wang et al. (2020), we found ATP8A2 was associated with a better prognosis in LUAD. RRM2B, which has been widely recognized for its important role in maintaining genomic stability, was widely overexpressed in many types of tumors. However, a previous study suggested the distinct roles of RRM2B in tumor progression: on one hand, its amplification was related to worse outcomes in breast cancers (Chae et al., 2016; Iqbal et al., 2021); on the other hand, in vitro analysis found its overexpression could inhibit the proliferation of lung cancer cells by the regulation work of FOXO3 (Cho et al., 2014). The latter result was also supported by our results as RRM2B was significantly associated with better survival in LUAD patients. Meanwhile, 18 NMGs significantly correlated with unfavorable prognosis in LUAD patients were identified. Among them, SCO1, HSPD1, IARS2, MRPL44, TFAM, NDUFS2, MRPL13, OPA1, and YARS have been previously identified as oncogenes in lung cancer or other types of cancers, serving as unfavorable prognosis biomarkers individually (Telang et al., 2012; Lee et al., 2017; Sotgia and Lisanti, 2017; Di et al., 2019; Dunham-Snary et al., 2019; Liu et al., 2019; Witherspoon et al., 2019; Li et al., 2020; Zhang et al., 2020; Cai et al., 2021). However, it is noteworthy that nearly all these findings were established and assessed in vitro, and all these NMGs were analyzed separately instead of integrating. Our novel identified unfavorable prognosis predictors, including GFM1, KARS, DARS2, GARS, DLP1, MRPL12, COA6, SCO1, and C10ORF have not been functionally analyzed previously. We are the first study to confirm their association with the progression of LUAD patients. Meanwhile, previous research found COA6, an OXPHOS complexes assembly factor, negatively correlated with vimentin levels, which was suggested as a favorable predictor for cancer prognosis (Cruz-Bermúdez et al., 2019). On the contrary, our findings suggested its association with unfavorable survival in LUAD patients. Furthermore, MRPL44 and MRPL12 were both members of mitochondrial ribosomal proteins, whose abnormal expression was found associated with the tumorigenesis and development of lung cancer (Huang et al., 2020). Glycolysis is a dominant metabolism for cancer cells to produce energy, even in the presence of oxygen. Research has demonstrated that a higher glycolytic rate in tumor cells was shown to promote resistance to chemotherapeutics (Ganapathy-Kanniappan and Geschwind, 2013). On the contrary, it has been revealed that mitochondrial OXPHOS is also utilized by cancer cells. Meanwhile, some evidence suggests that OXPHOS contributes to cancer metastasis (Viale et al., 2015; Yu et al., 2017; Rao et al., 2019). In the present study, glycolysis and oxidative phosphorylation signaling pathways were enriched in the high-risk subgroup in HALLMARK and KEGG enrichment analyses. These signaling pathways may play important roles in tumorigenesis and the progression of LUAD. Nowadays, the treatment landscape of NSCLC includes target therapy, immunotherapy, and combination therapy, which is mainly decided by the patients’ driver genes status, especially EGFR and ALK (Grant et al., 2021). Interestingly, our findings showed that LUAD patients with high or low NMGs signature may have distinct treatment choices as the differences in the tumor microenvironment and driver alterations. LUAD patients with low NMGs signature have not only improved outcomes but also significantly higher prevalence of EGFR alterations, which may contribute to the response to EGFR tyrosine-kinase inhibitors. On the contrary, patients with high NMGs have more inflamed immune features, including higher mutation counts, high presence of immune-prone TILs (such as CD8 positive T cells, activated CD4 memory T cell and NK cell), and lower presence of immunosuppressive TILs (such as Treg cells). Previous studies have found the tumor-infiltrated CD8 positive T cells prior to immune checkpoint inhibitors could serve as an indicator for the response of anti-PD-1/L1 therapy (Tumeh et al., 2014; McDermott et al., 2018). Based on these features, it is indicated that LUAD patients with high NMGs signature may benefit more from immune checkpoint inhibitors (Lucibello et al., 2021). In concordance with a more inflamed tumor microenvironment, tumor mutation counts (also known as tumor mutation burden, TMB) level was significantly higher in the high NMGs signature than in the low group, which has already been granted as an effective biomarker for selecting patients who may respond to pembrolizumab by FDA (Marcus et al., 2021).

There were some limitations in our study. Although with ample validation, the signature was established and validated based on the public datasets. Further validation in a local cohort using prospective clinical samples and data is merited. Though we found the association between NMGs signature and tumor microenvironment or driver gene alterations (especially EGFR), which may affect the decision-making for clinical management for LUAD patients, the validation of the indication that the difference in the efficacy of immune checkpoint inhibitor and targeted therapy on patients with high- or low-risk signature is worthy to explore in further study, as well as the underlying regulatory mechanism in vivo or in vitro. Based on the above limitations, we have designed a local LUAD cohort from our hospital and performed testing of the involved NMGs based on real-time PCR to further verify the results of this study.

In conclusion, we conducted an integrative analysis of the role of NMGs in the prognosis of LUAD and successfully established a robust and stable prognosis predicting signature with sufficient validation. Except for prognostic function, this established signature could also distinct LUAD patients who may have more clinical benefits from radiotherapy, targeted therapy, and immune checkpoint inhibitors, which may give clinicians more decision-making help to manage patients.

The datasets presented in this study can be found in online repositories. The names of the repository/repositories is cBioPortal and accession number(s) can be found in the article/Supplementary Material.

WM, XZ, and WD: conception and design, data analysis and interpretation (statistical analysis), and manuscript writing and revision. JZ, WL, DS, and JY: data acquisition and manuscript writing. All authors read and approved the final manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.