Low DMT1 expression associates with increased oxidative phosphorylation and early recurrence in HCC

Background—Despite a high rate of recurrences, long-term survival can be achieved after the resection of hepatocellular carcinoma (HCC) with effective local treatment. Discovery of adverse prognostic variables to identify patients with high risk of recurrence could improve the management of HCC. Accumulating evidence showing a link between carcinogenesis and increased expression of iron import proteins and intracellular iron prompted us to investigate a role Corresponding Authors: Fumito Ito, MD, PhD, Department of Surgical Oncology, Center for Immunotherapy, Roswell Park Comprehensive Cancer Center, Elm & Carlton Streets, Buffalo NY 14263 USA, Tel: +1-716-845-2300, Fax: +1-716-845-1595, fumito.ito@roswellpark.org. Kazuaki Takabe, MD, PhD, Department of Surgical Oncology, Roswell Park Comprehensive Cancer Center, Elm & Carlton Streets, Buffalo NY 14263 USA, Tel: +1-716-845-5540, Fax: +1-716-845-1595, kazuaki.takabe@roswellpark.org. 1These authors equally contributed to this study Presented at the 13th Annual Academic Surgical Congress in Jacksonville, Florida, on January 31, 2018. Author contributions: T.H. designed and conceptualized the study and prepared the article. E.K. conceptualized the study, acquired and analyzed data. L.Y. gave bioinformatical expertise. K.T. interpreted data and revised the article critically for important intellectual content. F.I. interpreted data, revised the article, and provided final approval of the version to be submitted. Disclosure The authors report no proprietary or commercial interest in any product mentioned or concept discussed in this article. Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain. HHS Public Access Author manuscript J Surg Res. Author manuscript; available in PMC 2020 February 01. Published in final edited form as: J Surg Res. 2019 February ; 234: 343–352. doi:10.1016/j.jss.2018.11.008. A uhor M anscript


Introduction
Incidence of liver cancer has more than tripled since 1980. It has increased by about 3% per year in the U.S. from 2004 to 2013. An estimated 42,220 new cases of liver cancer will be diagnosed in the U.S. during 2018, and approximately three fourths of which will be hepatocellular carcinoma (HCC). 1 HCC is the sixth most frequently diagnosed cancer and responsible for the second most common cause of cancer mortality worldwide with estimated 700,000 deaths per year. 2,3 HCC generally develops in patients with underlying chronic liver disease, which include cirrhosis from any cause such as chronic infection of hepatitis B virus (HBV) or hepatitis C virus (HCV), excessive alcohol intake, and nonalcoholic fatty liver disease (NAFLD). NAFLD and its advanced form, nonalcoholic steatohepatitis (NASH), which occurs with metabolic syndrome, obesity, and diabetes mellitus (DM), are now increasingly frequent underlying liver disease in patients with HCC. [4][5][6][7][8] Furthermore, numbers of clinical observational studies have provided a link between obesity and risk of HCC. [9][10][11][12][13] Several studies have indicated an association between the sequelae of NAFLD and the development of HCC even in the absence of cirrhosis; 8,14,15 however, pathogenesis of HCC development in NAFLD/NASH remains to be elucidated.
For patients who are medically fit with healthy background liver function and have resectable disease, liver resection continues to be a mainstay treatment. 16,17 However, even with complete surgical extirpation, recurrence rates within 5 years have been reported to exceed 70%. 18,19 Since the majority of recurrences are intrahepatic due to local recurrence or a new second primary tumor, 17,20,21 the goal of post-resectional surveillance is early detection of disease that might be amenable to subsequent local therapy such as repeat liver resection, liver transplantation, thermal ablation, and transarterial chemoembolization (TACE) or radioembolization. [22][23][24][25] Several tumor-related biologic as well as histologic factors such as high preoperative alpha fetoprotein (AFP) levels, tumor size, vascular invasion, resectional margin status, spontaneous tumor bleeding, and poor histologic grade of differentiation have been identified as potential predictors of recurrence, 17,20,21,24 yet the management of surveillance remains challenging and additional indicators will improve the detection of recurrence in HCC. Recent advances in RNA-seq transcriptomics allow identification of novel transcripts and molecular mechanism governing carcinogenesis, progression, and prognosis in solid malignancies.
Divalent metal-ion transporter-1 (DMT1) is a transmembrane protein involved in transportation of divalent metals including cadmium, cobalt, copper, nickel, lead, manganese, zinc, and in particular iron. 26,27 DMT1 is ubiquitously expressed, most notably in proximal duodenum, immature erythroid cells of the bone marrow, brain, and kidney. 28 Dietary iron is imported into the enterocyte through DMT1 on its apical surface and enters into the blood stream through ferroportin 1 (Fpn1). 29 Once absorbed, transferrin-bound iron is endocytosed and released into hepatocytes via DMT1. [30][31][32] Non-transferrin-bound iron, in case of iron overload, is also taken up by hepatocytes though DMT1 directly. 33 Some degrees of iron overload have been shown to be present in 10% to 30% of patients with chronic liver disease, and known as one of the causes of hepatocarcinogenesis through generating reactive oxygen species. 34 We previously reported aberrant expression of DMT1 in the duodenum, where iron absorption takes place, led to liver iron accumulation in patients with NASH. 35 Boult et al. investigated the expression of iron transport proteins in the premalignant lesions, Barrett's metaplasia and esophageal adenocarcinoma, and found that progression to adenocarcinoma was associated with increased expression of iron transport proteins including DMT1. 36 Moreover, Brookes et al. showed progression of normal colon and precancerous state such as low/high grade dysplastic adenomas to colorectal carcinoma was associated with increased expression of DMT1. 37 While DMT1 is involved in the carcinogenesis, its contribution to tumor progression and relapse of HCC remains unclear. Interestingly, DMT1 expression was found to be upregulated in iron-loaded, non-cirrhotic, and non-tumorous liver tissues compared with normal liver controls. 38 However, Deugnier et al. described lack of iron accumulation within HCC in hereditary hemochromatosis patients, 39 indicating the role of DMT1 differs between hepatocarcinogenesis and tumor progression. To date, no study has been reported on the significance of DMT1 gene expression in the context of recurrence following liver resection.
In the present study, we employed publicly available database, The Cancer Genome Atlas (TCGA) for genomic analysis to elucidate a role of DMT1 in recurrence after liver resection for HCC and performed gene set enrichment analysis (GSEA) to interpret genome-side expression profile.

Data acquisition and pre-processing
The National Cancer Institute (NCI) and National Human Genome Research Institute (NHGRI) obtained human data from doctors who collected tissue for TCGA after gaining approval with informed consent documents and local Institutional Review Boards (IRBs). The TCGA Biospecimen Core Resource laboratory extracted RNA from all samples and distributed the RNA to their Genome Sequencing Centers, where the uniform sequencing technique was used by TCGA researchers. There were 440 HCC cases of liver resection in TCGA cohort (http://cancergenome.nih.gov). RNA seq v2 z-scores and clinical data were obtained through the cBioportal for Cancer Genomics (http://cbioportal.org) on January 15th, 2018, and processed as previously described. 40,41 Of 440 cases, there were 369 primary tumors available for gene expression data from RNA-sequence. Two recurrent tumor samples from two patients whose primary tumors were also registered were excluded from our study to avoid duplication of patients. We also excluded two patients with neoadjuvant therapy to eliminate the gene expression affected by therapy prior to surgery. Patients were divided into DMT1 high and DMT1 low groups according to their gene expression levels with a cutoff being determined as the lower quartile value.

Statistical Analysis
Statistical analyses were performed using unpaired two-sided Student's t test or one-way ANOVA followed by Tukey's test for continuous variables when appropriate and Fisher's exact test for categorical variables. Disease-free survival (DFS) was estimated by the Kaplan-Meier (KM) method and the log-rank test. Censoring times were designated with vertical tick-marks on the KM curves. Correlations between DMT1 and the other gene mRNA expression levels were assessed with Pearson's correlation coefficients. P values of less than 0.05 were considered statistically significant. Data analyses and creation of graphs were carried out using R version 3.4.3 (http://www.r-project.org) and Bioconductor packages (http://www.bioconductor.org).

Decreased DMT1 expression is associated with worse DFS in HCC patients after liver resection
Next, 369 patients were divided into DMT1 high and DMT1 low groups. DFS data were available in 317 HCC patients who underwent surgical resection, and there were 242 DMT1 high and 75 DMT1 low patients. We found the DMT1 low group showed significantly worse DFS compared with the DMT1 high group (P = 0.044) ( Fig. 2A). To explore potential correlation between DFS and DMT1 expression in different stages, subgroup analyses were conducted in early-stage (AJCC stage I) patients and advanced-stage (AJCC stage II, III, and IV) patients. The DFS of DMT1 high and low groups were similar in 148 early-stage patients (P = 0.815), whereas the DMT1 low group showed significant worse DFS compared with the DMT1 high group in 148 advanced-stage patients (P = 0.008) ( Fig. 2B and C). High expression of DMT1 has been reported to contribute to carcinogenesis, while our data illustrated low expression of DMT1 caused worse prognosis, suggesting the role of DMT1 might be different between carcinogenesis and tumor progression.

DMT1 expression does not differ in etiologies, stages, and differentiation status of HCC
To identify the features of the low DMT1-expressing HCC associated with worse DFS, we evaluated the expression of DMT1 in etiologies, stages, and differentiation status of HCC. We found that DMT1 expression levels were independent of virus infectious status (Fig.   3A), and comparable in different stages (Fig. 3B). The comparison of clinicopathological characteristics between DMT1 high (n = 277) and low (n = 92) groups were shown in Table  1. There was no difference in the any features listed between DMT1 high and low groups. There was also no difference in DMT1 expression levels among grade 1 (well differentiated), grade 2 (moderately differentiated), and grade 3/4 (poorly differentiated/ undifferentiated) (Fig. 3C). As NAFLD is a growing cause of HCC and a manifestation of metabolic syndrome, we sought to investigate the relation between DMT1 expression and metabolic status. There was little information available on metabolic syndrome including diabetes in this cohort. We compared 66 patients with BMI > or = 30, which indicates high risk for diabetes, with 265 patients with BMI < 30, and there was no difference of DMT1 expression (Fig. 3D). Taken together, we found DMT1 expression levels were not associated with underlying disease, or any known factors related to prognosis or recurrence of HCC. Of note, K-ras and H-ras are known as oncogenes found in less than 7% of human liver cancers, 46 and the activation of Ras pathway has a role in HCC initiation and progression. 47 Loss of TP53 function, which is a frequently mutated tumor suppressor gene in HCC, has been shown to be associated with hepatocellular carcinogenesis 48 and poor prognosis. 49 Therefore, we looked at P53 pathway, K-ras, and H-ras signaling in the GSEA; however, we did not find any significant enrichments of those gene sets (Fig. 3E).

DMT1 expression is not associated with iron regulatory or heme transport genes in HCC
Since DMT1 is known as an iron importer, we evaluated correlation between DMT1 expression and iron regulatory genes or heme transport gene expressions in HCC. These include ceruloplasmin, cytochrome b reductase (Dcytb), Fpn1, hepcidin, hephaestin, ironresponsive element-binding protein 1 (IRP1), iron-responsive element-binding protein 1 (IRP2), transferrin receptor 1 (TfR1), transferrin receptor 2 (TfR2), transferrin (Trf), and solute carrier family 39 member 14 (Zip14) as iron regulatory genes, and hemoglobin scavenger receptor (CD163) and low density lipoprotein receptor-related protein 1 (LRP1/ CD91) as heme transport genes. Despite its function known to be as an iron importer, DMT1 expression correlated with neither of the iron regulatory genes nor heme transport genes ( Fig. 4A and B). Furthermore, GSEA revealed that none of iron regulatory or heme transport-related gene sets were enriched in DMT1 high group (Fig. 4C).

DMT1 expression level inversely associates with oxidative phosphorylation and glycolysis in HCC
In order to identify the features of the low DMT1-expressing HCC, which have worse DFS after liver resection, GSEA was conducted based on the 50 hallmark gene sets. 44 Intriguingly, oxidative phosphorylation and glycolysis gene sets were found to be enriched in DMT1 low tumors in addition to DNA repair and reactive oxygen species gene sets ( Table  2). Respiratory metabolism-related gene sets, such as glycolysis (NES = −1.547, P = 0.022), oxidative phosphorylation (NES = −2.166, P < 0.001), which generates adenosine triphosphate (ATP) using pyruvate, a product of glycolysis, and reactive oxygen species pathway (NES = −1.766, P = 0.011), were enriched in the low DMT1-expressing HCC among 50 hallmark gene sets ( Fig. 5A and B). Furthermore, respiratory chain known to be a major source of reactive oxygen species and the site of oxidative phosphorylation was also enriched in DMT1 low group (NES = −2.148, P < 0.001) (Fig. 5B). Our unexpected findings of inverse correlation between DMT1 and oxidative phosphorylation and glycolysis in HCC patients prompted us to further examine mitochondrial metabolism-related gene sets by GSEA. Since oxidative phosphorylation is a process that takes place in mitochondria, as we expected, "Mitochondrion pathway", a broad function of mitochondria, was found to be significantly enriched in DMT1 low tumor (NES = −2.099, P < 0.001) (Fig. 5C). Next, we further evaluated roles of DMT1 in regulation of other mitochondrial functions. Despite increased oxidative phosphorylation and mitochondrion pathway in DMT1 low group, other known mitochondrial functions including citric acid cycle (Kreb pathway; NES = −1.302, P = 0.193), fatty acid beta-oxidation (mitochondrial fatty acid beta oxidation; NES = −1.199, P = 0.332), storage of calcium ions (calcium ion homeostasis; NES = −0.990, P = 0.428), or apoptosis (apoptotic mitochondrial changes; NES = −1.381, P = 0.053) were not enriched (Fig. 5D). These findings suggest that DMT1 low tumors with higher risk of recurrence have increased glycolysis and oxidative phosphorylation facilitating higher ATP production.

Discussion
Despite a higher recurrence rate, long-term survival can be achieved after resection of HCC because in most cases recurrences are confined to the liver and may be amenable to local therapies. 17,20,21 Thus identification of risk factors, close follow-up evaluation, and early detection are of great importance. DMT1 was first identified as a transmembrane iron-transport protein, and found to play an important role in intestinal iron absorption. 51 DMT1 is also expressed in the liver; however, a role of DMT1 in hepatocytes remains unclear. Although transferrin-bound iron as well as free iron is uptaken by hepatocyte via DMT1, Wang et al. showed that DMT1 in hepatocytes is dispensable for hepatic iron accumulation and non-transferrin-bound iron uptake using mice with the Dmt1 gene selectively inactivated in hepatocytes (Dmt1 liv/liv ). 52 In line with this, expression of DMT1 did not correlate with iron regulatory genes or heme transport genes.
While iron deposition in hepatocyte has been shown to induce hepatocarcinogenesis, how iron and its importer, DMT1, are associated with cancer recurrence remains to be elucidated. In this study, we employed gene expression data from TCGA HCC cohort, and found that higher expression of DMT1 in HCC was associated with significantly longer DFS after resection ( Fig. 2A). This was somewhat an unexpected finding given the previous studies suggesting potential role of DMT1 in iron overload frequently seen in chronic liver disease and one of causes of hepatocarcinogenesis, 33,35 but might suggest potentially unrecognized roles of DMT1 in HCC.
In 1950s, Otto Warburg observed that cancer cells preferentially relied on aerobic glycolysis, which is less efficient than mitochondrial oxidative phosphorylation in terms of ATP generation, even in the presence of abundant oxygen. 53,54 For decades, the respiratory alteration had been regarded as a result of compensation for mitochondrial dysfunction, while recent studies have revealed oncogenes, such as K-ras and c-Myc, and hypoxia environment induce glycolysis. 50,55 Of note, iron is required as a cofactor for PHD1 to hydroxylate proline residues on HIF-α in regulation of hypoxia; 56 however, hypoxia pathway was not associated with DMT1 expression levels in HCC in hallmark gene set (Fig.  5E), and we did not find significant correlation between DMT1 expression and PHD1 (Fig.  5F) in the present study. Low DMT1 group with worse DFS showed enrichments of both glycolysis and oxidative phosphorylation, suggesting ATP production depends on both glycolysis and oxidative phosphorylation. Although it is well known that glycolysis is enhanced in most malignant tumors, contribution of oxidative phosphorylation for ATP production to HCC growth remains elusive. 57 In this study, we found that HCC with low DMT1 correlates not only with increased glycolysis but also with enhanced oxidative phosphorylation. While it is unclear whether oxidative phosphorylation is augmented in the process of an adaptation of hepatocellular carcinoma cells to meet energy demands in the tumor microenvironment, HCC usually develops blood supply predominantly from the arterial system during progression, 58 and can provide more oxygen for oxidative phosphorylation than portal vein. Some tumors are relatively glycolytic, the others have high rates of mitochondrial respiration. 57 Inactivation of the mitochondrial transcription factor has been shown to impaire lung tumor cell proliferation, 59 suggesting mitochondrial function is inevitable for cancer survival. Recently, Tan et al. have shown that tumor microenvironment instructs cancer cells to restore respiratory function, recover mitochondrial function, and reestablish tumor-initiating efficacy. 60 ATP is known as the main energy fuel and also reported to promote cell proliferation and drug resistance in cancer cells. 61 Since we have observed inverse association between DMT1 and oxidative phosphorylation, our results might indicate that DMT1 plays a role in directly or indirectly inhibiting oxidative phosphorylation, or vice versa. Indeed, our findings of strong correlation between increased mitochondrial metabolism and worse prognosis are in line with the profound influence of mitochondrial metabolism on all steps of oncogenesis such as malignant transformation, tumor progression and response to treatment. 62,63 Given cancer cells require higher levels of cytosolic ATP than normal tissue to sustain elevated rates of growth and division, 64 we assume elevated ATP is associated with worse prognosis in HCC. ATP is used for acquiring drug resistance in cancer cells through ATPdependent efflux pump. 65 ATP also increases metastatic efficiency by improving ability of cancer cells to withstand traumatic deformation in the microvasculature of target organs. 66 Cancer cells secrete ATP into their microenvironment at high concentrations than healthy tissues. 67,68 The extracellular ATP is known to be unstable and can be hydrolyzed to ADP, AMP, and eventually adenosine, and the final degradation product suppresses cellular immunity in the tumor environment and stroma. 69 DMT1 has thus far not been reported to be associated with activation of mitochondrial metabolism. Recently, DMT1 was found to be located in outer mitochondrial membrane and play roles in mitochondrial iron import and other metals. 70 The fact that iron is of vital importance for mitochondrial energy metabolism in oxidative phosphorylation as electron carriers contradicts our data where low DMT1 expression associates gene enrichment of mitochondrial function. Since our study investigated oxidative phosphorylation between high and low DMT1 expression and did not compared with DMT1 expression in normal liver, the relatively low DMT expression does not necessarily reflect a shortage of DMT1 and iron. The involvement of other metals may underlie the association between DMT1 low and up-regulated mitochondrial respiration.
The present study provides novel links between low DMT1 and increased mitochondrial and respiratory function in HCC with worse DFS, yet there are some limitations. First, this study was conducted using only one publicly available data without being validated using other cohorts. Second, this study is based on the gene expression of the primary tumor in TCGA cohort, that is, we have not verified through any in vitro or in vivo studies. To determine the role of DMT1 gene in association with oxidative phosphorylation and glycolysis, further studies would be warranted.

Conclusions
Lower DMT1 mRNA expression associates with poor DFS in patients with HCC regardless of underlying disease, stages, and differentiation status, and is associated with increased mitochondrial oxidative phosphorylation and glycolysis. Further studies are warranted to uncover how DMT1 affects the respiratory chain in mitochondria.