Sirtuin-6–dependent genetic and epigenetic alterations are associated with poor clinical outcome in hepatocellular carcinoma patients


  • Jens U. Marquardt,

    1. First Department of Internal Medicine , University Medical Center, Johannes Gutenberg University Mainz, Mainz, Germany
    2. Institute of Human Genetics, University Medical Center, Johannes Gutenberg University, Mainz, Germany
    Search for more papers by this author
    • These authors contributed equally to this work.

  • Kerstin Fischer,

    1. First Department of Internal Medicine , University Medical Center, Johannes Gutenberg University Mainz, Mainz, Germany
    Search for more papers by this author
    • These authors contributed equally to this work.

  • Katharina Baus,

    1. First Department of Internal Medicine , University Medical Center, Johannes Gutenberg University Mainz, Mainz, Germany
    Search for more papers by this author
  • Anubha Kashyap,

    1. First Department of Internal Medicine , University Medical Center, Johannes Gutenberg University Mainz, Mainz, Germany
    Search for more papers by this author
  • Shengyun Ma,

    1. First Department of Internal Medicine , University Medical Center, Johannes Gutenberg University Mainz, Mainz, Germany
    Search for more papers by this author
  • Markus Krupp,

    1. First Department of Internal Medicine , University Medical Center, Johannes Gutenberg University Mainz, Mainz, Germany
    Search for more papers by this author
  • Matthias Linke,

    1. Laboratory of Experimental Carcinogenesis, Center for Cancer Research, National Cancer Institute, National Institutes of Health, Bethesda, MD, USA.
    Search for more papers by this author
  • Andreas Teufel,

    1. First Department of Internal Medicine , University Medical Center, Johannes Gutenberg University Mainz, Mainz, Germany
    Search for more papers by this author
  • Ulrich Zechner,

    1. Laboratory of Experimental Carcinogenesis, Center for Cancer Research, National Cancer Institute, National Institutes of Health, Bethesda, MD, USA.
    Search for more papers by this author
  • Dennis Strand,

    1. First Department of Internal Medicine , University Medical Center, Johannes Gutenberg University Mainz, Mainz, Germany
    Search for more papers by this author
  • Snorri S. Thorgeirsson,

    1. Institute of Human Genetics, University Medical Center, Johannes Gutenberg University, Mainz, Germany
    Search for more papers by this author
  • Peter R. Galle,

    1. First Department of Internal Medicine , University Medical Center, Johannes Gutenberg University Mainz, Mainz, Germany
    Search for more papers by this author
  • Susanne Strand

    Corresponding author
    • First Department of Internal Medicine , University Medical Center, Johannes Gutenberg University Mainz, Mainz, Germany
    Search for more papers by this author

  • Potential conflict of interest: Nothing to report.

  • This study was supported by grants from the “Inneruniversitäre Forschungsförderung Stufe 1,” Johannes Gutenberg University, Mainz (to S. S.) and in part by the Forschungszentrum Immunologie, Imaging Core Facility, Johannes Gutenberg University, Mainz (to D. S.).

Address reprint requests to: Susanne Strand, Ph.D., First Department of Internal Medicine, Molecular Hepatology, University Medical Center, Johannes Gutenberg University Mainz, Obere Zahlbacherstrasse 63, 55131 Mainz, Germany. E-mail:; fax: +49-6131-179963.


Sirtuin 6 (SIRT6) is a member of the sirtuin family of NAD+–dependent deacetylases. Genetic deletion of Sirt6 in mice results in a severe degenerative phenotype with impaired liver function and premature death. The role of SIRT6 in development and progression of hepatocellular carcinoma is currently unknown. We first investigated SIRT6 expression in 153 primary human liver cancers and in normal and cirrhotic livers using microarray analysis. SIRT6 was significantly down-regulated in both cirrhotic livers and cancer. A Sirt6 knockout (KO) gene expression signature was generated from primary hepatoctyes isolated from 3-week-old Sirt6-deficient animals. Sirt6-deficient hepatocytes showed up-regulation of established hepatocellular carcinoma (HCC) biomarkers alpha-fetoprotein (Afp), insulin-like growth factor 2 (Igf2), H19, and glypican-3. Furthermore, decreased SIRT6 expression was observed in hepatoma cell lines that are known to be apoptosis-insensitive. Re-expression of SIRT6 in HepG2 cells increased apoptosis sensitivity to CD95-stimulation or chemotherapy treatment. Loss of Sirt6 was characterized by oncogenic changes, such as global hypomethylation, as well as metabolic changes, such as hypoglycemia and increased fat deposition. The hepatocyte-specific Sirt6-KO signature had a prognostic impact and was enriched in patients with poorly differentiated tumors with high AFP levels as well as recurrent disease. Finally, we demonstrated that the Sirt6-KO signature possessed a predictive value for tumors other than HCC (e.g., breast and lung cancer). Conclusion: Loss of SIRT6 induces epigenetic changes that may be relevant to chronic liver disease and HCC development. Down-regulation of SIRT6 and genes dysregulated by loss of SIRT6 possess oncogenic effects in hepatocarcinogenesis. Our data demonstrate that deficiency in one epigenetic regulator predisposes a tumorigenic phenotype that ultimately has relevance for outcome of HCC and other cancer patients. (Hepatology 2013;53:1054–1064)




hepatocellular carcinoma


insulin-like growth factor




mitogen-activated protein kinase


Nicotinamide adenine dinucleotide


nuclear factor kappa B


quantitative real-time polymerase chain reaction


sirtuin 6



Hepatocellular carcinoma (HCC) is the most deadly consequence of the majority of chronic liver diseases.[1] Whereas vaccination programs in several Asian countries have effectively controlled incidence rates in recent decades, incidences in several Western countries and Japan have increased steadily, mainly due to constant elevation of hepatitis C infections.[2] Moreover, predisposing risk factors for HCC development, such as alcohol and metabolic disease, exhibit alarmingly increasing trends in the Western world. Among these, metabolic syndrome and nonalcoholic fatty liver disease are of particular interest due to a predicted raise in prevalence and high numbers of HCC without underlying cirrhosis.[3, 4] Although considerable efforts to unravel genetic determinants of liver cancer have been made in recent decades, the exact pathogenesis remains to be elucidated and significantly varies between the different etiologies. In nonalcoholic steatohepatitis patients, the molecular changes are highly associated with the development of insulin resistance.[4] However, in addition to etiological differences, a common phenotypic hallmark feature of the majority of HCCs is the so-called inflammation-fibrosis-cancer axis, orchestrated by a complex interplay of different cell types and molecular features.[5]

Sirtuin 6 (SIRT6) is a member of the evolutionarily conserved sirtuin family of NAD+-dependent protein deacetylases and is involved in the regulation of glucose metabolism, triglyceride synthesis, and fat metabolism.[6-8] Sirt6-deficient animals present with early lethality due to profound abnormalities, including hypoglycemia and premature aging.[9, 10] Moreover, conditional disruption of Sirt6 in hepatocytes leads to increased glycolysis, triglyceride synthesis, reduced beta oxidation, and, ultimately, fatty liver formation. Furthermore, specimens from steatotic human livers show significantly lower levels of SIRT6 than control tissues, indicating a prominent role of SIRT6 in liver homeostasis.[11] A well-known mechanism in expediting the inflammation-fibrosis-cancer sequence is the activation of nuclear factor kappa B (NF-κB).[12] Although the regulation of NF-κB is complex, epigenetic modulation of NF-κB activation (e.g., by histone deacetylation) is well characterized.[8, 13] Recently, it was demonstrated that SIRT6 is a key component of histone H3 lysine 9 activity and plays a prominent role in the regulation of NF-κB signaling during inflammation, stress response, and aging.[14, 15]

Over the last decade, comparative functional genomics have been repeatedly and successfully employed to reproduce molecular features of human hepatocellular cancers using appropriate mouse models. This approach contributed significantly to a better understanding of the molecular features of HCC and led to the discovery of novel therapeutic targets.[16-18] Given the importance of SIRT6 in hepatocyte function and homeostasis of liver metabolism, we applied comparative and integrative genomics to determine the role of SIRT6 in human hepatocarcinogenesis. As a result, we demonstrated a stepwise reduction of SIRT6 levels from preneoplastic stages of hepatocarcinogenesis to human HCCs as well as an association of SIRT6 signaling with the outcome of liver and other cancers. The Sirt6-deficient microarray gene expression signature we generated from isolated hepatocytes of 3-week-old Sirt6−/− mice showed significant up-regulation of known HCC biomarkers. Using western blot and quantitative real-time polymerase chain reaction (qRT-PCR) analysis, the expression of these biomarkers was validated in Sirt6−/− mouse hepatocytes and human hepatoma cell lines. Further, re-expression of SIRT6 in HepG2 cells restored sensitivity to apoptotic stimuli. Global transcriptomic analyses confirmed the prominent role of Sirt6 signaling in the regulation of key hepatocyte functions such as cell cycle, metabolism, and oxidative stress response. On the molecular level, genetic loss of Sirt6 caused changes in the methylation pattern of affected livers leading to a metabolic and pro-oncogenic phenotype. Together, our results indicate a clinical significance of SIRT6 and disrupted SIRT6 signaling during liver carcinogenesis.

Patients and Methods

Isolation of Primary Mouse Hepatocytes

Mice of the strain 129-Sirt6tm1Fwa/J were obtained from The Jackson Laboratory and interbred to obtain mice homozygous for the Sirt6tm1Fwa allele. Hepatocytes from Sirt6−/− and Sirt6+/+ mice were isolated from mouse livers via hepatic portal vein perfusion as described.[19] Mice were kept in the central laboratory animal facility (ZVTE) of the Johannes Gutenberg University. Blood glucose levels were measured in whole blood with an Accu-Chek Sensor (Roche).

Global DNA Methylation Analysis

For genomic DNA preparation, tissues were lysed at 37°C overnight in a buffer containing 75mM NaCl, 30 mM EDTA, 0.5% SDS and 250 μg/mL proteinase K (pH 8.0). After addition of NaCl to a final concentration of 2M, the lysate was centrifuged for 20 minutes at 10,000 rpm. Genomic DNA was precipitated, washed with 70% ethanol, air-dried, and resuspended in TE buffer. The global DNA methylation status of livers from Sirt6−/− and Sirt6+/+ mice was determined using the colorimetric MethylFlash Methylated DNA Quantification Kit (Epigentek Inc.) according to the manufacturer's instructions.[20] Relative quantification of 100 ng genomic DNA was performed on an enzyme-linked immunosorbent assay plate reader at 450 nm. All investigations were performed in triplicate using two independent replicates.

Microarray Analysis

Total RNA from isolated hepatocytes was extracted using Absolutely RNA Miniprep Kit (Agilent Technologies) following the instructions of the manufacturer. RNA quantity was estimated using a NanoDrop ND-1000 Spectrophotometer (NanoDrop Technologies, Wilmington, DE). Gene expression microarrays were performed using Affymetrix GeneChip Mouse Genome 430 2.0 arrays. The arrays were deposited at EMBL-EBI (accession number: E-MTAB-1477). Arrays were normalized based on mean intensity values across the chips. Changes in expression levels were calculated based on log2 ratio. Publically available microarray data (Gene Expression Omnibus accession number: GSE21965)[11] was downloaded from GEO, processed, and analyzed using BRB ArrayTools V3.3.0 software package 3 (Biometric Research Branch, National Cancer Institute). Samples were normalized using Significance Analysis of Microarrays (SAM), and differentially expressed genes were identified at a nominal P ≤ 0.05. Unsupervised cluster analysis was performed using Cluster and TreeView programs.2. Only genes with a fold change ≥2 were included in the analyses. Functional classification and network analysis were performed using Ingenuity Pathway Analysis tool (Ingenuity Systems Inc.) and the GeneGo microarray tool.

Patients, Databases, and Statistical Analysis

Microarray data from 139 HCC samples[21] were used for the survival analysis according to the SIRT6 signatures. SIRT6 expression was investigated in a subcontingent of 53 HCC tumor specimens.[22] The Oncomine Cancer Microarray database ( was used to study gene expression of the SIRT6 signature in human HCC and conduct a meta-analysis for the predictive value of the SIRT6 signature in more than 40 different cancer types. Expression values of tumor samples were log-transformed and median-centered and standard deviation was normalized to one per array before comparison to their normal tissue counterparts as described recently.[23] Statistical analysis was performed using Student t test or analysis of variance as indicated. P ≤ 0.05 was considered statistically significant. Results are presented as the mean ± SD or mean ± SEM as indicated. Univariate and multivariate analysis were performed using a chi-squared test and Cox proportional hazard regression, respectively. For the multivariate analyses, only significant variables with sufficient data points were included.


Down-Regulation of SIRT6 in Human HCC

To investigate the relevance of SIRT6 for primary human HCC, we first used publically available gene expression data of liver cancer patients from the Oncomine Cancer Microarray database.[23] A significant reduction of SIRT6 expression was revealed in cirrhotic livers and HCC specimens (P < 0.001) compared with levels observed in noncirrhotic livers (Fig. 1A). In confirmation of these findings, a down-regulation of SIRT6 in HCC tissues compared with nondiseased normal livers was also observed in around 45% (24/53) using independent gene expression data from our recently published cohort of 53 human HCCs (Fig. 1B, upper panel).[22] Consistently, around 42% (16/38) of the tumor samples showed SIRT6 levels below the median center of the expression data of all samples (normalized expression units < 0) of patient samples analyzed in Fig. 1A (Fig. 1B, lower panel). These data indicate a stepwise reduction of SIRT6 in both premalignant and malignant stages of hepatocarcinogenesis.

Figure 1.

SIRT6 expression in human hepatocarcinogenesis. (A) The Oncomine cancer microarray database was used to assess SIRT6 levels during hepatocarcinogenesis. SIRT6 expression was compared in normal tissue (n = 10), cirrhosis (n = 58), and primary human HCC specimens (n = 38). Whisker plots represent the minimum and maximum normalized expression units (P < 0.001). (B) Expression levels of SIRT6 in primary liver cancer from our recently published HCC database in comparison with normal liver (n = 47)21 (upper panel) and from the Oncomine database (n = 38)[40] (lower panel) are shown. Only specimens with detectable expression values for SIRT6 are displayed.

SIRT6 Gene Expression Signature

To investigate the gene expression pattern deregulated by SIRT6 loss, we established a SIRT6 KO gene expression signature. To obtain a hepatocyte-specific transcriptome analysis, we isolated primary mouse hepatocytes from wild-type (WT) and Sirt6-deficient livers at 3 weeks of age. Gene expression levels were then quantified using genome-wide mouse microarrays from Affymetrix for comprehensive covering of changes occurring as a result of Sirt6 loss. Only genes for which expression was significantly altered in Sirt6-null hepatocytes (signal log ratio >1 and filtered for absent calls) were included as part of the Sirt6 signature. The resulting Sirt6 signature contained 1,615 probe IDs representing 1,241 genes (Supporting Table 1). Eighteen of the most deregulated targets were further validated using qRT-PCR (Supporting Fig. 1) overall demonstrating a high concordance (P < 0.001; r = 0.85).

Next, we investigated in more detail the functional enrichment of these genes in different networks and signaling pathways by using ingenuity pathway analysis and the GeneGo microarray analysis tools. The two most significant pathway map folders were related to cell cycle and its regulation and cholesterol/bile acid homeostasis (Table 1). Dysregulated pathways also included tissue remodeling and wound repair, lipid biosynthesis, and immune system response as well as nuclear receptor signaling. Additional map folders with a significant number of genes affected by the loss of Sirt6 were involved in mitogenic signaling, cell differentiation, DNA damage response, and apoptosis. Furthermore, canonical pathways and signaling resembling NF-κB and insulin-like growth factor (IGF) signaling were consistently activated in Sirt6-deficient hepatocytes (Supporting Fig. 2).

Table 1. Top Signaling Pathways From the Sirt6 KO Signature as Determined by GeneGo Analysis
No.Map FoldersPRatio
  1. The expression signatures were generated using isolated primary mouse hepatocytes from WT and Sirt6-deficient livers at 3 weeks of age. The resulting Sirt6 KO signature contained 1,241 genes. Shown are the top functional networks identified by GeneGO

1Cell cycle and its regulation4.57E-1569/444
2Cholesterol and bile acid homeostasis5.20E-1066/528
3Tissue remodeling and wound repair1.47E-0556/554
4Lipid biosynthesis and regulation2.46E-0543/393
5Immune system response3.72E-0581/925
6Nuclear receptor signaling7.51E-0564/698
7Mitogenic signaling1.51E-0453/560
8Cell differentiation1.73E-0478/922
9DNA damage response2.40E-0437/354
12Vascular development (angiogenesis)9.37E-0448/532
13Xenobiotic metabolism and its regulation1.30E-0331/306
14Oxidative stress regulation2.42E-0351/600
15Estrogen signaling3.79E-0328/287
16Protein synthesis4.48E-0329/304
17Inflammatory response1.08E-0249/617
18Androgen signaling1.74E-0221/224

The analyses suggested that loss of Sirt6 predisposes hepatocytes for oncogenic transformation. To validate the results, we performed qRT-PCR and western blot analyses of selected HCC marker genes in serum samples and isolated hepatocytes from WT and Sirt6-deficient animals (Fig. 2). For these studies, we examined Afp, Igf2, H19, and glypican-3 as well-established HCC biomarkers that we found to be up-regulated in our microarray analysis. Consistently, these genes were more abundantly expressed in Sirt6 KO hepatocytes compared with WT littermates. Afp and Igf2 were readily detectable on western blots of serum, and in the case of Afp, in hepatocytes from Sirt6 KO mice (Fig. 2B). Also, the recently reported H19-derived miRNA-675 was elevated in hepatocytes of KO animals (Fig. 2B, right panel). These results confirm that key oncogenic molecules associated with hepatocarcinogenesis are affected by the loss of Sirt6 signaling, thus strengthening the validity of the results from the microarrays. We next characterized a series of human hepatoma cell lines for SIRT6 expression in comparison with that of the series of HCC biomarkers (Fig. 3). SIRT6 was consistently down-regulated in comparison to primary human hepatocytes in all hepatoma cell lines examined. AFP was up-regulated in all cell lines compared with primary hepatocytes. IGF2 was up-regulated in all cell lines except PLC/PRF/5 cells. H19 was increased in Hep3B only. Taken together, these results suggest that the deregulation of SIRT6 and genes in the SIRT6 signature can at least in part be recapitulated in established hepatoma lines.

Figure 2.

Expression of potential Sirt6 target genes. (A) Gene expression of Afp, Igf2, H19, and glypican-3 (Gpc3) was analyzed in Sirt6−/− in comparison with Sirt6+/+ hepatocytes via qRT-PCR. Experiments were performed in triplicate. (B) Western blot analysis of Afp and Igf2 in serum and isolated hepatocytes of four different mice confirmed preferential activation of both genes in Sirt6−/−. β-Tubulin was used as a loading control. H19-derived miR-675 in hepatocytes of Sirt6−/− mice is elevated (right panel). Sirt6+/+ mice were labeled as WT1-WT3; Sirt6−/− mice were labeled as KO1-KO3.

Figure 3.

SIRT6 expression in human hepatoma cell lines. Gene expression of SIRT6, AFP, IGF2, and H19 was analyzed in four hepatoma cell lines via qRT-PCR. Primary human hepatocytes (ph_Hep) were used as a reference. Data were obtained from triplicate PCR analyses relative to expression level of primary human hepatocytes.

We next compared our microarray results with publicly available microarray data from whole liver tissue with conditional deficiency of Sirt6 in hepatocytes at 2 and 8 months (GEO accession number: GSE21965).[11] A total of 3,909 genes were differentially expressed between WT and Sirt6-deficient livers. From these, 329 genes overlapped with our identified Sirt6 KO signature (26.5%), indicating a high grade of concordance within Sirt6 signaling. In accordance with the previous studies, the overlapping 329 genes were functionally involved in lipid metabolism and cholesterol synthesis, hepatic cholestasis, oxidative stress response, and hepatocellular cancer development, thus independently confirming the probable involvement of SIRT6 in the affected pathways. Consistently, the major associated signaling pathways centered around NF-κB signaling, metabolism, and differentiation. Interestingly, the previously reported association with proliferation, cell death, and hepatocyte function as well as inflammatory signaling and tissue remodeling was less pronounced, potentially due to the confounding signaling of other cell types in whole liver tissues in contrast to isolated hepatocytes, overall warranting our approach. Taken together, these data reveal that genetic loss of Sirt6 causes massive changes in essential hepatocyte functions such as cellular metabolism, stress response, differentiation, and proliferation and are predisposing Sirt6-deficient animals to chronic liver diseases.

Expression of SIRT6 in Human Hepatoma Cell Lines Increases Apoptosis

Resistance or insensitivity to chemotherapy is one of the hallmarks of HCC. To analyze the effect of SIRT6 on apoptosis, we expressed SIRT6 in HepG2 hepatoma cells and studied the functional consequences. Transfection resulted in high expression of SIRT6 (Fig. 4A). Furthermore, while SIRT6 expression did not lead to a change in cell proliferation, a significant increase in apoptosis sensitivity mediated by CD95 stimulation (Fig. 4B) and in response to chemotherapeutic drugs was observed (Fig. 4C,D). These results suggest that loss of SIRT6 contributes to the resistance against cell death in tumor cells and supports a role for SIRT6 in suppressing the development of tumors in the liver.

Figure 4.

Re-expressing SIRT6 increases apoptosis sensitivity. (A) HepG2 cells were transfected with SIRT6 or control vectors. Effective transfection demonstrated by western blotting. (B-D) Cells were treated with 100 ng/mL Anti-Apo-1 (B) or cultured in the presence of 1 μg/mL or 10 μg/mL doxorubicin (Doxo) (C) or daunorubicin (Dauno) (D) for 24 hours. Viability was measured via Celltiter glow. All experiments were perfomed in quadruplicate. Graphs present the percentage of control cells. Statistical analyses were performed using a Student t test (Anti-Apo-1) or analysis of variance (Doxo and Dauno). *P < 0.05; **P < 0.01; ***P < 0.001.

SIRT6 Signature Is Associated With Clinical Outcomes in Liver and Other Cancers

To test the clinical significance of the SIRT6 KO signature for human hepatocellular cancers, we used a comparative genomic approach[17] and integrated the generated SIRT6 signature with our previously published gene expression dataset from 139 human HCC[21] (Fig. 5A) based on the expression of 958 orthologous genes. Hierarchical clustering analysis successfully identified two distinct subtypes concordant with published prognostic subtypes of HCC.[21] Further, Kaplan-Meier plots and log-rank statistics revealed a significant (P < 0.001) association with shortened mean survival time (306.7 days versus 1,611.2 days) among these two identified subclasses (Fig. 5B). As an independent prognostic factor, we also compared the recurrence between the subgroups of HCC. In accordance with poor prognosis demonstrated by survival analysis, a significant (P < 0.015) association to a shorter time to recurrence (703 days versus 1,520 days) could be demonstrated for tumors with disrupted SIRT6 signaling (Fig. 5C and Supporting Table 2). Notably, when we compared the expression of the SIRT6 signature in the human HCCs around 182 genes (around 15%) significantly differed between both subclasses. These genes again were significantly associated with the prognosis of patients overall, indicating that the these tumors retain a core SIRT6 signature (P < 0.001; data not shown).

Figure 5.

The SIRT6 KO signature is associated with outcome in liver cancer. The prognostic implication of the SIRT6 KO signature was determined using gene expression microarray data from 139 HCC patients.[21] (A) Hierarchical cluster analysis based on the SIRT6 KO (red) and WT (green) signature. Clustering was performed using Euclidean distance and average linkage analyses. Bars under the cluster tree represent integrated cell fractions and overlap with previously generated HCC subclasses (subtype A, B, HB, and HC). (B,C) Kaplan-Meier plots of (B) overall survival (n = 110) and (C) recurrence (n = 63) of HCCs with SIRT6 KO (red) and WT (green) signature (log-rank Mantel-Cox test).

Furthermore, to evaluate the clinical significance of the SIRT6 KO signature in molecular classification of HCC, we then compared the distribution of several clinical and pathological variables of the two subclasses using an univariate analysis (Table 2). The two subtypes of HCC were comparable with respect to sex, presence of cirrhosis in surrounding tissues, tumor size and stage, and vascular invasion. In contrast, a significant association with patient age, overall survival and recurrence, and Edmondson grade could be found. Furthermore, the two subclasses differed with respect to plasma AFP levels, which confirms the results of our microarray analyses. Interestingly, while distribution of HCV-positive and HCV-negative patients was similar among the two subgroups, a significantly higher proportion of HBV-positive patients was found in the poor prognosis cluster. Notably, the significant association with overall survival remained present using multivariate analyses (P = 0.0157; hazard ratio, 1.9273; 95% confidence interval, 1.1351-3.2724).

Table 2. Univariate Analysis of the Sirt6 Signature to Clinico-pathological Features in 139 HCCs
FeatureCluster 1Cluster 2P
  1. Values are presented as no. (% positive for the corresponding feature).

Subtype A49 (98)12 (14)5.85E-21
Age >60 years15 (30)42 (51)0.020043
AFP level (>300 ng/mL)24 (63)24 (40)0.025452
HBV-positive31 (78)31 (47)0.001987
Grade >240 (80)38 (46)0.000104
Survival7 (15)30 (47)0.000527
Recurrence20 (91)24 (54)0.003135
Alcoholic liver disease2 (5)15 (23)0.015914
Size (>5 cm)26 (63)39 (64)0.957307
HCV-positive5 (13)16 (27)0.141492
Cirrhosis27 (54)39 (48)0.47293
Invasion13 (69)14 (50)0.210026
Stage >212 (75)38 (73)0.878814
Male sex37 (74)60 (72)0.829696

Moreover, both gene set enrichment analysis and Oncomine meta-analysis suggested that the SIRT6 signature was significantly associated with cancer development, progression, and clinico-pathological features in several different tumor entities other than liver cancer, suggesting prognostic relevance of the SIRT6 signature for cancers other than HCC (Supporting Table 3 and Supporting Fig. 3). Thus, the SIRT6 signature is characterized by an unfavorable patient outcome with reduced survival and aggressive tumor phenotype in liver and other cancers.

Sirt6 Deficiency Causes a Metabolic Phenotype and Leads to Hypomethylation of Liver Tissue

To support the idea that Sirt6 loss is creating a procancer environment in the liver, we investigated whether other changes played a role in tumorigenesis in Sirt6-deficient livers. SIRT6 plays a major role in the epigenetic regulation by modulating chromatin function.[24] Genetic loss of Sirt6 leads to genomic instability, metabolic defects, and degenerative pathologies with aging-associated degenerative phenotypes.[9, 25] Animals with Sirt6 deficiency die within 3 to 4 weeks of age. The observed phenotypic changes are predominantly caused by profound changes in the regulation of cellular metabolism. Consistent with this phenotype, Sirt6−/− animals show a significantly reduced level of blood glucose (P < 0.001) in comparison with control animals already at 3 weeks of age (Fig. 6A). Additionally, increased hepatic fat deposition was already observed in young homozygous animals with Sirt6 deficiency (Fig. 6B). Similar observations have been demonstrated recently in a hepatocyte-specific model of conditional Sirt6 deficiency when the animals were 3 to 4 months old.[11]

Figure 6.

Loss of Sirt6 is associated with hypomethylation and metabolic changes. (A) Sirt6−/− animals show reduced blood glucose levels at 21 days of age (P < 0.001). (B) Hematoxylin and eosin (H&E) staining of SIRT6+/+ and SIRT6−/− mice (upper panel) and Oil Red O staining showing increased deposition of fat in heptocytes of Sirt6−/− mice (lower panel). (C) Global DNA methylation in Sirt6−/− in comparison with Sirt6+/+ was determined using a colorimetric quantification by specifically measuring levels of 5-methylcytosine (5-mC) (P < 0.01). Experiments were performed in four independent matched sample pairs. The graph shows the mean ± SD from three technical replicates.

Aberrant methylation has been reported in HCCs where global methylation was decreased while local CpG promoter methylation increased.[26] Given the importance of DNA methylation for liver-specific gene transcription, differentiation and essential hepatocyte functions as well as the known interplay between histone modifications and DNA methylation, we next assessed the level of global DNA methylation in Sirt6−/− and control livers. In agreement with the dominant role of Sirt6 in epigenetic regulation, significantly reduced levels of DNA methylation were present in Sirt6-deficient animals (Fig. 5C). Taken together, these results indicate that genetic loss of Sirt6 induces epigenetic changes and disruption of the liver homeostasis leading to a metabolic phenotype; both are associated with progression to cancer.


Aging is an established risk factor for cancer; however, the mechanisms and genes involved are just beginning to be revealed. The dramatic phenotype of Sirt6-deficient mice indicates a critical role for SIRT6 in the physiological processes involved in aging. We used an integrative genomic approach to investigate the importance of the longevity gene Sirt6 in chronic liver disease and progression to HCC. We provide evidence that loss of SIRT6 in hepatocytes results in a procancer milieu by deregulating a suite of genes, including known HCC biomarkers, which contribute to this phenotype. This is supported by our finding that disruption of Sirt6 leads to global hypomethylation and causes metabolic changes consistent with a pro-oncogenic phenotype. Comparison of 139 HCC gene expression profiles with the SIRT6-deficient signature revealed significant association with disease progression and recurrence. Furthermore, comparisons with publicly available data sets of other tumor types revealed that SIRT6 may be involved in other tumor types, since the signature is also linked to the clinical outcome of these cancer patients. Consistent with these findings, a recent study confirmed the crucial role of SIRT6 in cancer metabolism leading to poor prognosis of colorectal and pancreatic cancer patients.[27] Furthermore, the global gene expression changes observed in hepatocytes devoid of Sirt6 support the essential role of SIRT6 for liver homeostasis by maintaining the hepatic epigenome. Our results shed light on SIRT6 as a potential tumor suppressor, since its loss results in an oncogenic phenotype that is associated with poor clinical outcome of human liver cancer patients. Mechanistically, genetic loss of SIRT6 causes resistance against cell death (Fig. 4), a key mechanism in cancer development and progression.[28] Conversely, re-expression of SIRT6 in HepG2 cells partly rescued its oncosuppressive function by restoring apoptosis sensitivity mediated by CD95 stimulation as well as chemotherapy treatment. These results extend a recent finding that suggest a critical involvement of SIRT6 in the early phase of hepatocarcinogenesis.[29]

Already at 3 weeks of age, the genetic loss of Sirt6 causes profound alterations in the liver, including hepatic metabolism. These changes involve the progressive accumulation of fat in Sirt6-deficient hepatocytes as well as dramatic disruption of insulin homeostasis resulting in significantly increased glycolysis.[4, 10, 11] Our analysis revealed up-regulation of HCC biomarker genes in livers of 3-week-old mice with Sirt6 deficiency, even though these mice show no overt tumors. Upon comparing the Sirt6 levels to the biomarker expression levels in primary human hepatocytes and several established human hepatoma cell lines, we found a surprising congruency between the Sirt6 knockout (KO) livers and the human hepatoma cell lines. These results are in line with the dominant role of SIRT6 as a regulator of essential hepatocyte functions and support a role of modulating SIRT6 for the prevention of liver disease.

Our global transcriptome analyses confirmed that the disruption of SIRT6 in hepatocytes leads to activation of multiple key signaling pathways with known association to liver diseases, including hepatocarcinogenesis.[30] This includes activation of genes important for proliferation (cyclins A, A2, B1-2, D1-2, CDC20, CDC34, CDK1, CDK4, casein kinase I) and several members of the mitogen-activated protein kinase (MAPK) members (MAP3K1, MAP3K8, MAP4K4) known to play a role for HCC proliferation, survival, and differentiation.[31, 32] Additionally, other key molecules affected by the loss of SIRT6 were involved in malignancy-associated metabolic abnormalities of cholesterol and bile acid homeostasis (CYP2B6, CYP2C18, CYP2C44, CYP2F1, CYP2J2, CYP2J5, CYP2J9, CYP3A4, CYP4A22, CYP4F12, CYP51A1), as well as lipid biosynthesis and regulation. In addition, SIRT6 loss influences chemoresistance drug transporters (ABCB11, ABCB1B, ABCG8)[33, 34] and oxidative stress regulation (GSTM1, GSTM2, GSTM4, GSTM5, GSTM6, GSTM3), further underlining the essential role of SIRT6 for maintaining hepatocyte stress defense. Importantly, we also demonstrated that SIRT6 deficiency causes aberrant growth receptor signaling (epidermal growth factor receptor, platelet-derived growth factor receptor) and IGF2 expression. The role of IGF2 in many human cancers, as well as HCC, is well recognized. Activation of IGF2 is observed in around 30% of human HCCs.[35] Recently, activation of IGF signaling was demonstrated in a subclass of HCC with poor clinical outcome (referred to as “proliferation class”).[36] This study further showed that modulation of IGF signaling provides a promising target for therapeutic strategies in HCC. Together, these results indicate that loss of the aging-related gene Sirt6 might establish a link between key molecules involved in metabolism, inflammation, and steatosis and lead to the development of chronic liver diseases and HCC formation.

Another key finding of the study is the disruption of the hepatic epigenome caused by the loss of SIRT6 signaling. Compelling evidence indicates a causal role of aberrant epigenetic regulation for the development of a variety of cancers including HCC.[37] Epigenetic changes of the inflamed and chronically diseased liver microenvironment are supposed to be early promoters of oncogenic transformation in HCC. Therefore, epigenetic mechanisms might tie genomic alterations with environmental influences in the liver.[38] It is well known that different epigenetic alterations cause activation of signals from the microenvironment leading to cellular proliferation, disruption of the hepatic metabolism, and ultimately cancer initiation and progression. A multistep disruption of the hepatic epigenome leading to allelic imbalances has recently been confirmed in HBV-mediated HCC.[39] Importantly, global hypomethylation could be associated with poor clinical outcome in HCC patients.[26] Consistent with this, we observed a stepwise reduction of SIRT6 from preneoplastic stages of hepatocarcinogenesis to fully malignant HCC. Furthermore, disruption of Sirt6 was associated with significantly reduced global DNA methylation in mouse livers. Thus, our results highlight the importance of Sirt6 in maintaining the hepatic epigenome and demonstrate that disruption of its function is frequently observed during hepatocarcinogenesis. Furthermore, our results point toward the potential of modulating this pathway in a clinical setting to complement existing treatment strategies; due to the promise of epigenetic therapies in HCC, this may be an important addition.[22] Finally, to further support the role of SIRT6 for hepatocarcinogenesis, we performed integrative transcriptomic analyses of SIRT6 signaling in authentic primary HCC. Similar to previously generated prognostic signatures[30] (such as MET and transforming growth factor β), our integrative strategy uncovered two distinct subclasses of HCC patients based on the molecular features of SIRT6 signaling. These distinct subclasses showed significant differences in biological properties as well clinical outcome underlining the clinical relevance of SIRT6.