Abstract
Hepatic steatosis is a strong risk factor for the development of hepatocellular carcinoma (HCC), yet little is known about the molecular pathology associated with this factor. In this study, we performed a forward genetic screen using Sleeping Beauty (SB) transposon insertional mutagenesis in mice treated to induce hepatic steatosis and compared the results to human HCC data. In humans, we determined that steatosis increased the proportion of female HCC patients, a pattern also reflected in mice. Our genetic screen identified 203 candidate steatosis-associated HCC genes, many of which are altered in human HCC and are members of established HCC-driving signaling pathways. The protein kinase A/cyclic AMP signaling pathway was altered frequently in mouse and human steatosis-associated HCC. We found that activated PKA expression drove steatosis-specific liver tumorigenesis in a mouse model. Another candidate HCC driver, the N-acetyltransferase NAT10, which we found to be overexpressed in human steatosis–associated HCC and associated with decreased survival in human HCC, also drove liver tumorigenesis in a steatotic mouse model. This study identifies genes and pathways promoting HCC that may represent novel targets for prevention and treatment in the context of hepatic steatosis, an area of rapidly growing clinical significance. Cancer Res; 77(23); 6576–88. ©2017 AACR.
Introduction
Hepatocellular carcinoma (HCC) is the second leading cause of death from cancer worldwide (1). It occurs two to four times more frequently in men than women (2). Hepatic steatosis is an increasingly common and strong HCC risk factor (3) induced by chronic heavy alcohol intake or metabolic syndrome (4, 5). Oxidative stress due to high intrahepatic fatty acids induces DNA damage, leading to hepatocyte injury, inflammation, fibrosis, and tumorigenesis (6). Alcohol consumption further increases HCC risk and is among the most significant factors associated with progression of steatosis to HCC (3).
There is considerable genetic heterogeneity in HCC (7). The majority of mutations occur infrequently (8) and chromosome aberrations involving large gains and losses are prevalent (8, 9), making the distinction between molecular tumor drivers and passenger alterations challenging. Driver mutations can be identified by comparative analysis of molecular alterations in human tumors to data obtained through insertional mutagenesis using transposons, like Sleeping Beauty (SB) or PiggyBac (PB) (10, 11). Mutagenic transposons have been engineered to induce gain- and loss-of-function mutations upon insertion into the genome, tagging proto-oncogenes and tumor suppressors in resulting tumor cells (10). SB transposon-based mutagenesis in mice has identified HCC drivers in normal livers and in livers with common HCC-promoting mutations (12). Although earlier transposon-based HCC studies have been informative, few recapitulate the chronic liver damage in which HCC usually develops, which changes the selective pressures that nascent HCC cells encounter (2). We expected SB mutagenesis in steatotic livers would reveal genetic drivers relevant for steatosis-associated HCC prevention and treatment.
Using liver-specific SB insertional mutagenesis in mice treated with ethanol and a choline-deficient diet, we induced liver tumors and identified over 200 recurrent insertion sites, many of which affected genes and signaling pathways altered in human steatosis-associated HCC. We validated the oncogenic roles of Nat10 and the PKA/cAMP signaling pathway using an in vivo steatotic mouse model.
Materials and Methods
Transgenic mouse models
Animal studies were conducted using procedures approved and monitored by the Institutional Animal Care and Use Committee at the University of Minnesota. SB-mutagenized transgenic mice with Cre recombinase expressed from the Albumin promoter, Cre-inducible SB transposase, and T2/Onc transposon concatemer on chromosome 15 (Alb-CreTg/WT; Rosa26-lsl-SB11Tg/WT; T2/OncTg/WT) and control mice lacking transposon mobilization (Rosa26-lsl-SB11Tg/WT; T2/OncTg/WT) were generated and genotyped as described (12). After weaning at 21 days, mice were placed on 5% (w/w) ethanol drinking water ad libitum for 2 months, then normal drinking water and a choline-deficient diet (#0296021050, MP Biomedicals) ad libitum for the remainder of the study. Mutagenized mice were sacrificed with age-matched control mice at approximately 60 day intervals to assess tumor formation, when mice became moribund, or by 400 to 436 days. Transgenic Fah-deficient mice expressing SB (Fah−/−; Rosa26-SB11Tg/WT) were generated and maintained as described (13). Mice were assigned to either normal diet (ND) or a hepatic steatosis-inducing diet (eCDD). ND mice were maintained on nitisinone drinking water until plasmid injection at 8 weeks of age, then placed on normal drinking water and standard diet. After weaning, eCDD mice were placed on 5% (w/w) ethanol plus nitisinone drinking water. After plasmid injection at 8 weeks of age, eCDD mice were placed on normal drinking water and a choline-deficient diet ad libitum for the remainder of the study. Mouse Nat10 cDNA (#MC202909; Origene) and Prkaca transcript variant 1 (accession no. NM008854) open reading frame with nucleotides 616 and 617 mutated from TT to CG (generating the L206R mutation) with a C-terminal V5-tag were cloned into the previously generated pT2/GD-IRES-GFP (14). Gateway destination SB transposon plasmid coexpressing Fah and GFP using Gateway LR clonase mix (#11791-020; Thermo Fisher Scientific) to generate pT2/GD-Nat10 and pT2/GD/PrkacaL206R. pKT2/GD-GFP and pT2/shp53 plasmids were generated previously (13). Twenty micrograms of each plasmid was delivered by hydrodynamic tail vein injection as described (15). Mice were sacrificed at 150 to 180 days postinjection, livers examined using GFP goggles (#FHS/EF-2G2; BLS-Ltd.), and GFP-positive tumors counted and collected. All liver tumor analysis was performed as described (12).
Quantitative reverse transcriptase PCR
RNA was extracted from mouse liver tissue using Trizol (#15596026; Invitrogen) or the RNeasy Kit (#74106; Qiagen) and DNase treated with the DNA-Free Kit (#AM1906; Ambion) per manufacturers' instructions. cDNA was synthesized using the Transcriptor First Strand cDNA Synthesis Kit (#04379012001; Roche) or SuperScript IV VILO Master Mix (#11756050; Invitrogen) per manufacturers' instructions. Primer sequences are listed in Supplementary Table S1. qPCR was performed on a C1000 Touch Thermal Cycler with CFX96 Real-Time System (#1855195; Bio Rad) using SYBR Green Master Mix (#4309155; Applied Biosystems) per manufacturer's instructions.
Tissue microarray preparation
A human HCC tissue microarray was procured from the University Of Washington Department of Pathology. HCC and paired tumor-free liver tissue cores from cases with etiologies of Hepatitis B (11 cases), Hepatitis C (58 cases), autoimmune hepatitis (3 cases), alcohol (1 case), a combination thereof (9 cases), or unknown (18 cases) collected from surgeries performed between 1999 and 2007 were formalin-fixed, paraffin-embedded, sectioned at 5 μm on a standard microtome, and heat-fixed onto glass slides.
Histologic staining
Formalin-fixed, paraffin-embedded tissue section slides were either stained with hematoxylin and eosin (H&E) and analyzed by color bright-field microscopy by two board-certified pathologists (KA and ML, American Board of Pathology) or stained by immunohistochemistry as described (13) using the antibodies listed in Supplementary Table S2. Representative fields were displayed from stained tissue sections photographed at ×10 magnification.
Histologic stain quantification
Slides scanned at ×20 magnification with Aperio ScanScope XT (Leica) were analyzed using Aperio ImageScope Positive Pixel Count software (PPC; Leica). Parameters to quantify lipid droplet tissue percentage from H&E-stained slides, FOXA1 immunohistochemistry levels, and percent phospho-T156-FOXA2 (p-FOXA2)-positive nuclei are described in Supplementary Tables S3–S5. Stain levels from images of immunohistochemically stained tissue sections taken at ×10 magnification were measured with ImageJ software (NIH) using the Immunohistochemistry Image Analysis Toolbox (16) default H-DAB settings (imagej.nih.gov/ij/plugins/immunohistochemistry-toolbox/index.html) to detect brown staining, then converting the brown stain image to 32-bit grayscale. Quantification parameters are listed in Supplementary Table S6.
Transposon insertion site analysis
Transposon–genomic DNA junctions were amplified from genomic DNA isolated from SB-induced liver tumors by linker-mediated PCR as described (17) and sequenced on the Illumina MiSeq with 150 cycle paired-end reads. Nonredundant insertions were calculated using TAPDANCE; to limit analysis to trunk drivers, only reads over 1/10,000th of the total reads for the tumor were included (18). Identification of TAPDANCE common transposon insertion sites (CIS) and gene centric CIS (gCIS) was performed as described (10, 18). To account for T2/Onc local-hopping, the donor chromosome 15 was excluded from analysis. Known CIS artifacts caused by T2/Onc sequence elements (En2, Foxf2) or amplified regions within the mouse genome (Serinc3, Sfi) were excluded (10, 18). Mice with two or more tumors with at least 1,000 reads in each tumor and insertions with at least 10 reads were further selected for clonality analysis. Each mouse was analyzed separately. The insertions and tumors for each mouse were hierarchically clustered using average clustering method and Euclidean distance.
CIS enrichment analysis
For each CIS gene with >3 transposon insertions in SB-induced liver tumors in eCDD-treated mice, the number of SB-induced liver tumors in eCDD-treated mice with and without insertions was compared to the number of SB-induced liver tumors in ND-treated mice from a previous study (12) with and without insertions. Steatosis-enriched CIS were defined as CIS with a significantly higher proportion of tumors with insertions in eCDD-treated than ND-treated mice with a Q value <0.05 by Benjamini–Hochberg-corrected Fisher exact test.
Sex distribution of human HCC cases with steatosis
Clinical information abstracted from records of 460 HCC patients at the Mayo Clinic, Rochester, between January 2007 and December 2009 was reported previously (19). The number of male and female HCC cases in the overall cohort and the subset without hepatic steatosis was compared to the subset with nonalcoholic hepatic steatosis as the primary HCC etiology by χ2 test.
Analysis of genomic data from The Cancer Genome Atlas
RNA-seq normalized counts, copy number variation, mutation, and survival data generated by The Cancer Genome Atlas (TCGA) Research Network (http://cancergenome.nih.gov/) were extracted from TCGA and cBioportal (20, 21). Normalized expression count values under 0.1 were adjusted to 0.1 to minimize effects of low count genes in differential expression analysis. Genes with significant expression changes between tumor and normal liver were identified by Benjamini–Hochberg adjusted Student t test.
High fat diet induced wild-type mouse liver tumor generation and RNA-sequencing
Animal studies were conducted using procedures approved and monitored by the Institutional Animal Care and Use Committee at the Pacific Northwest Research Institute. Starting at 5 weeks of age, male C57BL/6J mice were fed a high-saturated fat, high-sucrose diet for 400 days (#D12331; Research Diets) ad libitum, then sacrificed for tissue collection. Tumors and adjacent tumor-free liver samples were frozen in liquid nitrogen and stored at −80°C. RNA was extracted using the Direct-zol RNA Miniprep Kit (#R2050; Zymo Research). Library preparation was performed with the TruSeq Stranded mRNA kit (#RS-122-2101; Illumina) and samples were sequenced on the Illumina NextSeq 500 platform with a high output 2 × 75 bp Paired-End Kit. Quantification of all RefSeq genes from the GRCm38/mm10 genome assembly was performed using the kallisto algorithm (22).
Ingenuity pathway analysis
Gene lists were subjected to a core analysis including direct and indirect relationships using Ingenuity Pathway Analysis (IPA) Software (Ingenuity Systems; www.ingenuity.com).
Results
A transposon insertional mutagenesis-driven mouse model of steatosis-associated liver cancer
We used a previously described conditional SB transposon insertional mutagenesis system to induce liver tumors, generating a mutagenized transgenic mouse cohort with one copy each of: Albumin-cre, T2/onc mutagenic transposon, and Rosa26-lsl-SB11 transposase and a control cohort carrying two of the three transgenes, thereby lacking transposition (12). Mice were treated with a hepatic steatosis-inducing diet (eCDD) consisting of drinking water with 5% ethanol ad libitum for 2 months after weaning followed by untreated drinking water and a choline-deficient diet for the remainder of the study (Fig. 1A). Choline-deficient diet reliably induces hepatic steatosis and leads to liver tumorigenesis in 22% of mice after 19 months (23). The addition of 5% ethanol increases steatosis severity and sensitization to carcinogen-induced tumorigenesis (24, 25). Here, SB insertional mutagenesis drove hepatocarcinogeneis in eCDD-treated mice.
SB transposition and eCDD treatment promotes steatosis-associated liver tumorigenesis in mice. A, Genotype and treatment model. EtOH, 5% ethanol drinking water. CDD, choline-deficient diet. RLS, Rosa26-Lsl-SB11. Alb-cre, Albumin-cre. d, days. mo, months. B, H&E and immunohistochemistry for SB and ALB-stained mouse liver sections. Liver section scale bars, 100 μm. T, tumor. N, nontumor liver. C and D, Tumor penetrance (C) and box-and-whisker (D) plot of tumor burden in control (n = 32) and SB-mutagenized (n = 49) eCDD-treated mice. E, Gross livers from SB-mutagenized and control eCDD-treated mice. Arrows, tumors. Gross liver scale bars, 0.5 cm. ****, P < 0.0001. M, male. F, female.
SB transposition and eCDD treatment promotes steatosis-associated liver tumorigenesis in mice. A, Genotype and treatment model. EtOH, 5% ethanol drinking water. CDD, choline-deficient diet. RLS, Rosa26-Lsl-SB11. Alb-cre, Albumin-cre. d, days. mo, months. B, H&E and immunohistochemistry for SB and ALB-stained mouse liver sections. Liver section scale bars, 100 μm. T, tumor. N, nontumor liver. C and D, Tumor penetrance (C) and box-and-whisker (D) plot of tumor burden in control (n = 32) and SB-mutagenized (n = 49) eCDD-treated mice. E, Gross livers from SB-mutagenized and control eCDD-treated mice. Arrows, tumors. Gross liver scale bars, 0.5 cm. ****, P < 0.0001. M, male. F, female.
All livers from eCDD-treated mice examined histopathologically (34/34) displayed steatosis (Fig. 1B and Supplementary Fig. S1A and S1B). Lipid droplet tissue percentage, measured by quantifying percent unstained pixels from H&E-stained slides using PPC, was approximately eight-fold higher in eCDD-treated than ND-treated mouse livers (P = 0.012, Student t test; Supplementary Fig. S1A and S1B).
Hepatic SB transposase expression was confirmed by immunohistochemistry (Fig. 1B). To determine whether SB insertional mutagenesis drove liver tumorigenesis, tumor penetrance and multiplicity were compared between eCDD-treated SB-mutagenized mice and age-matched controls. Mutagenized mice had higher tumor penetrance, with 76% (n = 49) versus 22% (n = 32) controls developing tumors by 436 days (P < 0.0001, χ2 test; Fig. 1C). Tumor multiplicity was also higher, averaging 4.8 tumors per mutagenized mouse versus 0.3 tumors per control (P < 0.0001, Wilcoxon rank-sum test; Fig. 1D). Liver tumors from mutagenized mice were classified as well-differentiated hepatocellular neoplasms, analogous to adenomas (81%), or HCCs (19%). Cholangiocarcinoma-like morphology was not observed (Fig. 1B and E). Albumin was detected in tumors by immunohistochemistry, confirming their hepatocyte origin (Fig. 1B; ref. 12). Thus, SB insertional mutagenesis induced steatosis-associated liver cancer in eCDD-treated mice.
Increased HCC incidence in females with hepatic steatosis
As reported previously (19), the sex distribution of a cohort of 460 adult HCC patients at the Mayo Clinic, Rochester, followed the typical bias, with 72% (332 patients) males and 28% (128 patients) females. We examined the sex distribution among subsets of patients with and without hepatic steatosis as the primary predisposing factor. The 397 HCC patients without steatosis were 75% (296 patients) men and 25% (101 patients) women. The 63 HCC patients with steatosis were 57% (36 patients) men and 43% (27 patients) women, a significantly higher female proportion than the non-steatotic subset and the overall cohort (P = 0.004 and 0.014, respectively, χ2 test; Fig. 2A).
Reduced male liver cancer sex bias with steatosis. A, Sex distribution of human HCC cases by steatosis status (n = 460 for overall cohort, 397 for no steatosis cohort, 63 for steatosis cohort). B, Box-and-whisker plot of tumor burden per ND-treated, SB-mutagenized male (n = 27) and female (n = 11) mouse. C, Box-and-whisker plot of tumor burden per eCDD-treated, SB-mutagenized male (n = 23) and female (n = 26) mouse. #, P > 0.05; *, P < 0.05; **, P < 0.01.
Reduced male liver cancer sex bias with steatosis. A, Sex distribution of human HCC cases by steatosis status (n = 460 for overall cohort, 397 for no steatosis cohort, 63 for steatosis cohort). B, Box-and-whisker plot of tumor burden per ND-treated, SB-mutagenized male (n = 27) and female (n = 11) mouse. C, Box-and-whisker plot of tumor burden per eCDD-treated, SB-mutagenized male (n = 23) and female (n = 26) mouse. #, P > 0.05; *, P < 0.05; **, P < 0.01.
A previous study of ND-treated mice with identical SB transgenes (12, 14) modeled the typical human sex bias, with an average of 8.1 liver tumors per male (n = 27) and 2.3 per female mouse (n = 11; P = 0.044, Wilcoxon rank-sum test; Fig. 2B and Supplementary Table S7). There was no statistically significant sex bias and a trend toward reversal in SB-induced liver tumorigenesis in eCDD-treated mice, with an average tumor burden for males (n = 23) and females (n = 26) of 3.3 and 6.0, respectively (P = 0.054, Wilcoxon rank-sum test; Fig. 2C). Tumor penetrance in male mice decreased nonsignificantly from 85% with ND-treatment to 65% with eCDD-treatment (P = 0.12, χ2 test) and increased nonsignificantly in females from 64% with ND-treatment to 81% with eCDD-treatment (P = 0.27, χ2 test). Thus, the SB system modeled human HCC sex bias, with higher prevalence of non-steatotic liver cancer in males, and a reduction of this bias in steatosis-associated liver cancer.
FoxA1 and FoxA2 repression in steatotic livers
Forkhead box transcription factors FoxA1 and FoxA2 contribute to liver cancer sex bias in mice (26). Their activities are metabolically regulated in human livers; FOXA1 transcript is downregulated with steatosis (27) and FOXA2 function is negatively regulated by phosphorylation at T156 downstream of insulin signaling (28).
We observed similar effects in steatotic mice. FoxA1 transcript was lower in eCDD-treated than ND-treated mouse livers (P = 0.014; Student t test; Supplementary Fig. S1C). Hepatic steatosis increases with age in humans and mice (29, 30), and in ND-treated mice, FoxA1 expression significantly decreased with age (Pearson's correlation coefficient, r = −0.82, P = 0.002; Supplementary Fig. S1D). We examined FOXA1 expression by immunohistochemistry staining quantification in serial sections from the same liver samples used to quantify lipid droplet percentage, finding a nonsignificant inverse correlation between the two measures (r = −0.62, P = 0.26; Supplementary Fig. S1A and S1E). Additional serial sections stained for p-FOXA2 revealed increased staining intensity (r = 0.89, P = 0.04) and nonsignificantly increased nuclear localization (r = 0.77, P = 0.12) with higher liver lipid content (Supplementary Fig. S1A, S1F, and S1G). Expression of ApoM, a transcriptional target of FOXA1 and FOXA2, was reduced in eCDD-treated versus ND-treated control mouse livers (P = 0.02, Student t test; Supplementary Fig. S1H). Collectively, these data suggest FOXA1 and FOXA2 are repressed in steatotic livers. Because loss of FoxA1 and FoxA2 eliminates sex bias in chemically-induced mouse HCC (26), repression of FOXA1 and FOXA2 may contribute to the reduced sex bias with steatosis.
CIS identify candidate steatosis-associated HCC genes
Transposon insertions from 159 tumors collected from 36 eCDD-treated SB-mutagenized mice (15 males and 21 females; Supplementary Table S7) were amplified and sequenced on the Illumina platform, revealing 7010 nonredundant insertions. From these, TAPDANCE analysis (18) identified 67 CIS with 73 associated genes, and gCIS analysis (10) identified 188 CIS genes. Fifty-eight CIS genes were identified by both methods, totaling 203 candidate steatosis-associated liver cancer genes (Table 1 and Supplementary Table S8). No annotated genes were associated with one CIS (Supplementary Table S8), although it is possible this region contains a currently unannotated noncoding RNA. Clonality analysis of tumors originating from the same animals revealed that, whereas most insertions were unique, identical insertions were present in multiple tumors from the same mouse for 14 individuals (Supplementary Figs. S2A–S2G, S3A–S3F, S4A–S4D). Of the 20 CIS identified in tumors from <3 animals, only two shared exact insertion locations in tumors from the same individual, indicating possible clonal derivation (Supplementary Table S9). These did not contribute substantially to the findings described below.
Steatosis-associated CIS list: top 51 CIS genes
CIS-associated gene . | Tumors with tdCIS . | tdCIS P value library number . | Mice with tdCIS . | Tumors with gCIS . | gCIS q value library number . | Mice with gCIS . | Predicted effect on gene function . | Predicted Oncogene or TSG? . |
---|---|---|---|---|---|---|---|---|
Gnai3 | 9 | 7.14E−11 | 8 | 9 | 0.00E+00 | 8 | Disrupts | TSG |
Adk | 14 | 5.69E−09 | 11 | 14 | 6.83E−23 | 11 | Disrupts | TSG |
Sfpq | 5 | 3.02E−06 | 5 | 4 | 1.07E−32 | 4 | Disrupts | TSG |
Hipk3 | 5 | 3.02E−06 | 4 | 5 | 2.25E−16 | 4 | Truncates C terminus or disrupts | TSG |
Egfr | 5 | 3.02E−06 | 4 | 6 | 1.60E−11 | 5 | Truncates C terminus | Oncogene |
Gigyf2 | 6 | 2.24E−05 | 6 | 7 | 9.34E−18 | 6 | Disrupts | TSG |
Arih1 | 8 | 1.72E−04 | 7 | 7 | 1.07E−20 | 6 | Disrupts | TSG |
Dpyd | 10 | 2.27E−04 | 9 | 18 | 1.42E−13 | 13 | Disrupts | TSG |
Gbe1 | 9 | 3.24E−04 | 8 | 9 | 2.16E−13 | 8 | Disrupts | TSG |
Dhx9 | 4 | 4.30E−04 | 4 | 4 | 1.08E−18 | 4 | Disrupts | TSG |
Ubl3 | 4 | 4.30E−04 | 3 | 4 | 4.64E−16 | 3 | Disrupts | TSG |
Palmd | 4 | 4.30E−04 | 3 | 4 | 2.35E−14 | 3 | Disrupts | TSG |
Mkln1 | 4 | 4.30E−04 | 3 | 6 | 9.43E−14 | 4 | Disrupts | TSG |
Pafah1b1 | 5 | 1.15E−03 | 5 | 5 | 4.72E−20 | 5 | Disrupts | TSG |
Ppp2r2a | 5 | 1.15E−03 | 4 | 5 | 1.83E−19 | 4 | Disrupts | TSG |
Ppm1b | 5 | 1.15E−03 | 5 | 5 | 1.57E−17 | 5 | Disrupts | TSG |
Arhgap5 | 5 | 1.15E−03 | 5 | 5 | 2.95E−16 | 5 | Disrupts | TSG |
Rundc3b | 6 | 2.06E−03 | 3 | 3 | 8.96E−06 | 2 | Drives N terminal truncation | Oncogene |
Mtus1 | 7 | 3.09E−03 | 5 | 7 | 1.34E−18 | 5 | Disrupts | TSG |
Kynu | 7 | 3.09E−03 | 6 | 5 | 9.66E−08 | 5 | Disrupts | TSG |
Cfh | 7 | 3.09E−03 | 5 | 4 | 5.12E−06 | 4 | Disrupts | TSG |
Zbtb20 | 7 | 3.09E−03 | 6 | 7 | 4.98E−05 | 6 | Disrupts | TSG |
Atrn | 5 | 2.98E−09 | 5 | Disrupts | TSG | |||
Ddrgk1 | 8 | 2.62E−02 | 6 | — | — | — | Drives N terminal truncation | Oncogene |
Prr16 | 8 | 2.62E−02 | 6 | — | — | — | Disrupts | TSG |
Acbd5 | 6 | 4.85E−02 | 6 | 4 | 2.02E−14 | 4 | Disrupts | TSG |
Ranbp9 | 6 | 4.85E−02 | 5 | 5 | 2.12E−14 | 5 | Disrupts | TSG |
Rnf111 | 6 | 4.85E−02 | 5 | 5 | 3.88E−14 | 5 | Disrupts | TSG |
Mllt10 | 6 | 4.85E−02 | 5 | 6 | 2.41E−10 | 5 | Disrupts | TSG |
Suclg2 | 6 | 4.85E−02 | 6 | 7 | 2.93E−10 | 7 | Disrupts | TSG |
Gfi1 | 6 | 4.85E−02 | 6 | 5 | 1.40E−08 | 5 | Drives | Oncogene |
Chchd3 | 6 | 4.85E−02 | 5 | 6 | 8.71E−06 | 5 | Disrupts | TSG |
Bmpr2 | 6 | 4.85E−02 | 5 | — | — | — | Drives N terminal truncation | Oncogene |
Thrap3 | 4 | 4.93E−20 | 4 | Drives/Drives N terminal truncation | Oncogene | |||
Sh3d21 | 5 | 4.86E−02 | 4 | — | — | — | Drives N terminal truncation | Oncogene |
Atp2a2 | 5 | 4.86E−02 | 5 | 4 | 2.19E−16 | 4 | Disrupts | TSG |
Usp47 | 5 | 4.86E−02 | 4 | 5 | 1.51E−11 | 4 | Disrupts | TSG |
Cacul1 | 5 | 4.86E−02 | 4 | 4 | 1.51E−11 | 3 | Drives N terminal truncation | Oncogene |
Grb14 | 5 | 4.86E−02 | 4 | 5 | 1.05E−10 | 4 | Disrupts | TSG |
Gopc | 5 | 4.86E−02 | 4 | 3 | 2.07E−07 | 3 | Disrupts | TSG |
Rhobtb1 | 5 | 4.86E−02 | 5 | 3 | 8.24E−06 | 3 | Drives | Oncogene |
Trip11 | 3 | 4.53E−05 | 2 | Disrupts | TSG | |||
Cpsf2 | 5 | 4.86E−02 | 3 | — | — | — | Drives | Oncogene |
Yeats2 | 5 | 4.86E−02 | 5 | — | — | — | Disrupts | TSG |
Sucla2 | 4 | 4.89E−02 | 4 | 4 | 1.37E−15 | 4 | Disrupts | TSG |
Scyl2 | 4 | 4.89E−02 | 3 | 4 | 8.94E−15 | 3 | Disrupts | TSG |
Rnf169 | 4 | 4.89E−02 | 3 | 4 | 1.70E−13 | 3 | Drives N terminal truncation | Oncogene |
Notch2 | 4 | 4.89E−02 | 4 | 5 | 6.81E−12 | 4 | Disrupts | TSG |
Tax1bp1 | 4 | 4.89E−02 | 2 | 4 | 1.01E−11 | 2 | Disrupts | TSG |
Nat10 | 4 | 4.89E−02 | 4 | 3 | 9.06E−11 | 3 | Drives | Oncogene |
Caprin1 | 3 | 1.81E−09 | 3 | Disrupts | TSG |
CIS-associated gene . | Tumors with tdCIS . | tdCIS P value library number . | Mice with tdCIS . | Tumors with gCIS . | gCIS q value library number . | Mice with gCIS . | Predicted effect on gene function . | Predicted Oncogene or TSG? . |
---|---|---|---|---|---|---|---|---|
Gnai3 | 9 | 7.14E−11 | 8 | 9 | 0.00E+00 | 8 | Disrupts | TSG |
Adk | 14 | 5.69E−09 | 11 | 14 | 6.83E−23 | 11 | Disrupts | TSG |
Sfpq | 5 | 3.02E−06 | 5 | 4 | 1.07E−32 | 4 | Disrupts | TSG |
Hipk3 | 5 | 3.02E−06 | 4 | 5 | 2.25E−16 | 4 | Truncates C terminus or disrupts | TSG |
Egfr | 5 | 3.02E−06 | 4 | 6 | 1.60E−11 | 5 | Truncates C terminus | Oncogene |
Gigyf2 | 6 | 2.24E−05 | 6 | 7 | 9.34E−18 | 6 | Disrupts | TSG |
Arih1 | 8 | 1.72E−04 | 7 | 7 | 1.07E−20 | 6 | Disrupts | TSG |
Dpyd | 10 | 2.27E−04 | 9 | 18 | 1.42E−13 | 13 | Disrupts | TSG |
Gbe1 | 9 | 3.24E−04 | 8 | 9 | 2.16E−13 | 8 | Disrupts | TSG |
Dhx9 | 4 | 4.30E−04 | 4 | 4 | 1.08E−18 | 4 | Disrupts | TSG |
Ubl3 | 4 | 4.30E−04 | 3 | 4 | 4.64E−16 | 3 | Disrupts | TSG |
Palmd | 4 | 4.30E−04 | 3 | 4 | 2.35E−14 | 3 | Disrupts | TSG |
Mkln1 | 4 | 4.30E−04 | 3 | 6 | 9.43E−14 | 4 | Disrupts | TSG |
Pafah1b1 | 5 | 1.15E−03 | 5 | 5 | 4.72E−20 | 5 | Disrupts | TSG |
Ppp2r2a | 5 | 1.15E−03 | 4 | 5 | 1.83E−19 | 4 | Disrupts | TSG |
Ppm1b | 5 | 1.15E−03 | 5 | 5 | 1.57E−17 | 5 | Disrupts | TSG |
Arhgap5 | 5 | 1.15E−03 | 5 | 5 | 2.95E−16 | 5 | Disrupts | TSG |
Rundc3b | 6 | 2.06E−03 | 3 | 3 | 8.96E−06 | 2 | Drives N terminal truncation | Oncogene |
Mtus1 | 7 | 3.09E−03 | 5 | 7 | 1.34E−18 | 5 | Disrupts | TSG |
Kynu | 7 | 3.09E−03 | 6 | 5 | 9.66E−08 | 5 | Disrupts | TSG |
Cfh | 7 | 3.09E−03 | 5 | 4 | 5.12E−06 | 4 | Disrupts | TSG |
Zbtb20 | 7 | 3.09E−03 | 6 | 7 | 4.98E−05 | 6 | Disrupts | TSG |
Atrn | 5 | 2.98E−09 | 5 | Disrupts | TSG | |||
Ddrgk1 | 8 | 2.62E−02 | 6 | — | — | — | Drives N terminal truncation | Oncogene |
Prr16 | 8 | 2.62E−02 | 6 | — | — | — | Disrupts | TSG |
Acbd5 | 6 | 4.85E−02 | 6 | 4 | 2.02E−14 | 4 | Disrupts | TSG |
Ranbp9 | 6 | 4.85E−02 | 5 | 5 | 2.12E−14 | 5 | Disrupts | TSG |
Rnf111 | 6 | 4.85E−02 | 5 | 5 | 3.88E−14 | 5 | Disrupts | TSG |
Mllt10 | 6 | 4.85E−02 | 5 | 6 | 2.41E−10 | 5 | Disrupts | TSG |
Suclg2 | 6 | 4.85E−02 | 6 | 7 | 2.93E−10 | 7 | Disrupts | TSG |
Gfi1 | 6 | 4.85E−02 | 6 | 5 | 1.40E−08 | 5 | Drives | Oncogene |
Chchd3 | 6 | 4.85E−02 | 5 | 6 | 8.71E−06 | 5 | Disrupts | TSG |
Bmpr2 | 6 | 4.85E−02 | 5 | — | — | — | Drives N terminal truncation | Oncogene |
Thrap3 | 4 | 4.93E−20 | 4 | Drives/Drives N terminal truncation | Oncogene | |||
Sh3d21 | 5 | 4.86E−02 | 4 | — | — | — | Drives N terminal truncation | Oncogene |
Atp2a2 | 5 | 4.86E−02 | 5 | 4 | 2.19E−16 | 4 | Disrupts | TSG |
Usp47 | 5 | 4.86E−02 | 4 | 5 | 1.51E−11 | 4 | Disrupts | TSG |
Cacul1 | 5 | 4.86E−02 | 4 | 4 | 1.51E−11 | 3 | Drives N terminal truncation | Oncogene |
Grb14 | 5 | 4.86E−02 | 4 | 5 | 1.05E−10 | 4 | Disrupts | TSG |
Gopc | 5 | 4.86E−02 | 4 | 3 | 2.07E−07 | 3 | Disrupts | TSG |
Rhobtb1 | 5 | 4.86E−02 | 5 | 3 | 8.24E−06 | 3 | Drives | Oncogene |
Trip11 | 3 | 4.53E−05 | 2 | Disrupts | TSG | |||
Cpsf2 | 5 | 4.86E−02 | 3 | — | — | — | Drives | Oncogene |
Yeats2 | 5 | 4.86E−02 | 5 | — | — | — | Disrupts | TSG |
Sucla2 | 4 | 4.89E−02 | 4 | 4 | 1.37E−15 | 4 | Disrupts | TSG |
Scyl2 | 4 | 4.89E−02 | 3 | 4 | 8.94E−15 | 3 | Disrupts | TSG |
Rnf169 | 4 | 4.89E−02 | 3 | 4 | 1.70E−13 | 3 | Drives N terminal truncation | Oncogene |
Notch2 | 4 | 4.89E−02 | 4 | 5 | 6.81E−12 | 4 | Disrupts | TSG |
Tax1bp1 | 4 | 4.89E−02 | 2 | 4 | 1.01E−11 | 2 | Disrupts | TSG |
Nat10 | 4 | 4.89E−02 | 4 | 3 | 9.06E−11 | 3 | Drives | Oncogene |
Caprin1 | 3 | 1.81E−09 | 3 | Disrupts | TSG |
Abbreviation: TSG, tumor suppressor gene.
Many steatosis-associated CIS genes were altered in human HCCs profiled by TCGA (http://cancergenome.nih.gov/). Overall, 68% (139/203) of CIS genes were significantly misexpressed (P < 0.05, Benjamini–Hochberg–adjusted Student t test) in tumors compared to normal livers, 14% (29/203) were misexpressed in HCCs with steatosis as the sole predisposing condition, and 62% (125/203) were misexpressed in alcohol-associated HCCs (Fig. 3A and Supplementary Table S10). Examining mutation and copy number changes for all TCGA HCC cases, 10% of CIS genes (20/203) were altered in >5% of tumors, 15% (30/203) were altered in >5% of alcohol-associated HCCs, and 29% (59/203) were altered in >5% of steatosis-associated HCCs (Fig. 3B and Supplementary Figs. S5–S7 and Supplementary Table S10). This overlap shows the relevance of steatosis-associated CIS genes to human HCC.
Steatosis-associated liver cancer CIS genes altered in human HCC. A and B, Representation of the 203 CIS genes and subsets with significant expression changes (A) and subsets altered in more than 5% of tumors by amplification, mutation, or deletion in all TCGA HCC cases, steatosis- associated HCC, and alcohol-associated HCC (B).
Steatosis-associated liver cancer CIS genes altered in human HCC. A and B, Representation of the 203 CIS genes and subsets with significant expression changes (A) and subsets altered in more than 5% of tumors by amplification, mutation, or deletion in all TCGA HCC cases, steatosis- associated HCC, and alcohol-associated HCC (B).
Most CIS gene expression changes in human steatosis or alcohol-associated TCGA HCC cases were consistent with the predicted role as an oncogene or tumor suppressor in mouse tumors based on transposon location and orientation. 69% (20/29) of significant CIS gene expression changes in steatosis-associated HCC and 66% (83/125) in alcohol-associated HCC [74% (14/19) of significant >two-fold] were consistent with the predicted effect of transposon insertions (Supplementary Fig. S8A and S8B and Supplementary Table S10). These genes represent the most promising human steatosis-associated HCC driver candidates.
Steatosis-enriched HCC CIS
To identify CIS specifically relevant to steatosis-associated HCC, we performed an enrichment analysis comparing our results to a previously published SB HCC screen in ND-treated mice (12). We compared the prevalence of steatosis-associated CIS insertions in tumors generated in the eCDD screen to the ND HCC screen (259 tumors examined; ref. 12) using Fisher exact test with Benjamini–Hochberg correction and a significance threshold of P < 0.05. We found 49 steatosis-enriched TAPDANCE CIS genes and 79 steatosis-enriched gCIS with 30 overlapping, equating to 98 steatosis-enriched CIS genes (Supplementary Table S11).
NAT10 overexpression in HCC
In eCDD-treated SB-mutagenized mice, Nat10 was identified as a candidate oncogene by both TAPDANCE and gCIS (P = 0.049 and P < 0.0001, respectively; Table 1 and Supplementary Fig. S9). Nat10 insertions were enriched in tumors from eCDD-treated versus ND-treated SB-mutagenized mice (Q = 0.038, Benjamini–Hochberg-corrected Fisher exact test; Supplementary Table S11).
Of human TCGA HCC cases, 6.7% (24/357) overexpressed NAT10 more than two-fold, which was associated with decreased survival (P = 0.012, log-rank Mantel–Cox test; Fig. 4A). Ten of the NAT10-overexpressing cases had corresponding copy number gains (P < 0.0001, one-way ANOVA with post hoc test; Fig. 4B). NAT10 overexpression was higher in steatosis-associated HCC (P < 0.0001; Student t test; Fig. 4C).
NAT10 alterations in human HCC. A, Overall survival for TCGA HCC cases with and without NAT10 overexpression. B, NAT10 expression by gene copy number for TCGA HCC cases. C, NAT10 expression in TCGA no risk HCC and steatosis-associated HCC cases compared with average normal liver. D, Representative immunohistochemistry for NAT10-stained sections of human HCC (top) or matched normal liver (bottom) from tissue microarray (TMA) cases with hepatic steatosis indicated in pathology reports. Scale bars, 100 μm. T, tumor. N, nontumor liver tissue. E, NAT10 immunohistochemistry stain intensity of human HCC and matched normal liver tissue microarray cases stratified by mention of steatosis in pathology report. Yes, steatosis reported. No, no mention of steatosis or steatosis reported absent. Unknown, pathology report not available. *, P < 0.05; **, P < 0.01; ****, P < 0.0001. Error bars, SEM.
NAT10 alterations in human HCC. A, Overall survival for TCGA HCC cases with and without NAT10 overexpression. B, NAT10 expression by gene copy number for TCGA HCC cases. C, NAT10 expression in TCGA no risk HCC and steatosis-associated HCC cases compared with average normal liver. D, Representative immunohistochemistry for NAT10-stained sections of human HCC (top) or matched normal liver (bottom) from tissue microarray (TMA) cases with hepatic steatosis indicated in pathology reports. Scale bars, 100 μm. T, tumor. N, nontumor liver tissue. E, NAT10 immunohistochemistry stain intensity of human HCC and matched normal liver tissue microarray cases stratified by mention of steatosis in pathology report. Yes, steatosis reported. No, no mention of steatosis or steatosis reported absent. Unknown, pathology report not available. *, P < 0.05; **, P < 0.01; ****, P < 0.0001. Error bars, SEM.
Immunohistochemical analysis of a human HCC tissue microarray showed NAT10 overexpression in most HCCs, with 82/99 displaying stronger NAT10 staining than matched tumor-free liver. Five cases had steatosis indicated in the pathology reports, three displaying stronger NAT10 staining than matched tumor-free liver (P < 0.0001 and P = 0.003, respectively; Wilcoxon signed-rank test; Fig. 4D and E).
Nat10 overexpression promotes hepatocarcinogenesis
We used SB transposon-based gene delivery to stably overexpress Nat10 in the selective Fah-deficient mouse model with or without hepatic steatosis as described previously (12–14). Fah-deficient mice expressing SB transposase (12) were maintained on the protective drug nitisinone until delivery of transposon plasmids by hydrodynamic tail vein injection (15) at 8 weeks of age and selection of hepatocytes with stable Fah-rescue transgene expression by nitisinone withdrawal. Mice were treated with either eCDD or ND. Negative control mice of both sexes were injected with an SB transposon-based GFP expression vector coexpressing Fah (pKT2/GD-GFP; ref. 13) and short hairpin against p53 (pT2/shp53; ref. 13; GFP/shp53), and experimental mice of both sexes were injected with an SB transposon-based expression vector for mouse Nat10 cDNA co-expressing Fah with a GFP reporter (pT2/GD-Nat10) plus pT2/shp53 (Nat10/shp53; Fig. 5A and B and Supplementary Table S7).
Nat10 overexpression drives tumorigenesis in mouse livers. A, Transposons for tumor induction. Red triangles, SB inverted repeat/direct repeat sequences. Caggs, Caggs promoter. Gene-of-interest, mouse cDNA sequence for either Nat10 or PrkacaL206R with V5 tag. IRES, internal ribosomal entry site. F. luc, firefly luciferase gene sequence. pA, polyadenylation signal. PGK, PGK promoter. Fah, mouse Fah cDNA. GFP, GFP cDNA sequence. GOI, gene of interest refers to Nat10 or PrkacaL206R. B, Treatment plan. EtOH, 5% ethanol CDD, choline-deficient diet. d, days. mo, months. C, Immunohistochemistry for NAT10, FAH, and ALB-stained liver sections from eCDD-treated mice injected with GFP/shp53 (top) or Nat10/shp53 (middle and bottom). Scale bars, 100 μm. T, tumor. N, nontumor liver. D, Nat10 expression measured by qRT-PCR from nontumor liver tissue (L) or liver tumor (T) from Nat10/shp53 or GFP/shp53-injected eCDD-treated mice, normalized to Actb and to wild-type mouse liver Nat10 (n = 5 each). E, NAT10 immunohistochemical stain intensity of nontumor liver tissue (L) or liver tumors (T) from Nat10/shp53 (n = 11 L; n = 17 T) or GFP/shp53-injected mice (n = 6 L) from both diets combined. F and G, Box-and-whisker plot of tumor burden (F) or tumor penetrance for Nat10/shp53-injected mice treated with eCDD (n = 24) or ND (n = 25) or GFP/shp53-injected mice treated with eCDD (n = 36) or ND (n = 42; G). *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001. Error bars, SEM.
Nat10 overexpression drives tumorigenesis in mouse livers. A, Transposons for tumor induction. Red triangles, SB inverted repeat/direct repeat sequences. Caggs, Caggs promoter. Gene-of-interest, mouse cDNA sequence for either Nat10 or PrkacaL206R with V5 tag. IRES, internal ribosomal entry site. F. luc, firefly luciferase gene sequence. pA, polyadenylation signal. PGK, PGK promoter. Fah, mouse Fah cDNA. GFP, GFP cDNA sequence. GOI, gene of interest refers to Nat10 or PrkacaL206R. B, Treatment plan. EtOH, 5% ethanol CDD, choline-deficient diet. d, days. mo, months. C, Immunohistochemistry for NAT10, FAH, and ALB-stained liver sections from eCDD-treated mice injected with GFP/shp53 (top) or Nat10/shp53 (middle and bottom). Scale bars, 100 μm. T, tumor. N, nontumor liver. D, Nat10 expression measured by qRT-PCR from nontumor liver tissue (L) or liver tumor (T) from Nat10/shp53 or GFP/shp53-injected eCDD-treated mice, normalized to Actb and to wild-type mouse liver Nat10 (n = 5 each). E, NAT10 immunohistochemical stain intensity of nontumor liver tissue (L) or liver tumors (T) from Nat10/shp53 (n = 11 L; n = 17 T) or GFP/shp53-injected mice (n = 6 L) from both diets combined. F and G, Box-and-whisker plot of tumor burden (F) or tumor penetrance for Nat10/shp53-injected mice treated with eCDD (n = 24) or ND (n = 25) or GFP/shp53-injected mice treated with eCDD (n = 36) or ND (n = 42; G). *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001. Error bars, SEM.
Nat10/shp53 and GFP/shp53 mouse livers expressed comparable levels of Fah transcript (P = 0.37, Student t-test; Supplementary Fig. S10) and FAH was detected by immunohistochemistry in livers and tumors from both cohorts (Fig. 5C), indicating efficient liver repopulation with transfected cells. Nat10/shp53-injected mouse livers and tumors overexpressed Nat10 transcript (P < 0.0001 and P = 0.03, respectively, Student t test; Fig. 5D) and NAT10 protein (P = 0.004 and 0.046, respectively; Wilcoxon rank-sum test; Fig. 5E) compared to GFP/shp53 mouse livers. More eCDD-treated mouse livers (13/13) examined histopathologically showed fatty change than ND-treated (4/14) (P = 0.0001, χ2 test).
eCDD-treated Nat10/shp53 mice (n = 24) developed more tumors than eCDD-treated GFP/shp53 mice (n = 36), with a mean of 4.4 versus 0.7 tumors per mouse (P = 0.0005, Wilcoxon rank-sum test), and more mice developed tumors, 71% vs. 22% (P = 0.0002, χ2 test; Fig. 5F and G). Similar results were observed in ND-treated mice, with a mean of 2.8 versus 0.1 tumors per Nat10/shp53 (n = 25) and GFP/shp53 mouse (n = 42; P < 0.0001, Wilcoxon rank-sum test) and tumor penetrance of 68% vs. 7% (P < 0.0001, χ2 test; Fig. 5F and G). Tumor burden and penetrance in Nat10/shp53 mice were not significantly higher with eCDD treatment than ND (P = 0.65 and 0.83; Wilcoxon rank-sum test and χ2 test, respectively).
Nat10/shp53-induced liver tumors were histopathologically classified as well-differentiated hepatocellular neoplasms (57%) or HCCs (43%) and GFP/shp53 control liver tumors classified as well-differentiated hepatocellular neoplasms (100%). Albumin was detected in Nat10/shp53 mouse liver tumors by immunohistochemistry, confirming hepatocyte origin (Fig. 5C).
Signaling pathways associated with CIS genes
To identify signaling pathways promoting steatosis-associated liver cancer, we performed IPA on the steatosis-associated and steatosis-enriched CIS gene lists, finding 90 and 72 significantly associated pathways, respectively (Supplementary Tables S12 and S13). To assess the generalizability of our findings, we analyzed genes with significant expression changes between tumor and adjacent liver from an independent steatosis-associated HCC model using wild-type C57BL/6J mice treated with a high-saturated fat, high-sucrose diet, finding 119 significantly associated pathways (Supplementary Table S14). We performed IPA on lists of genes with significant expression changes in steatosis-associated (all differentially expressed genes) and alcohol-associated (>two-fold change) human HCC from TCGA, finding 137 and 113 significantly associated pathways, respectively (Supplementary Tables S15 and S16). Below we describe our findings on Wnt/β-catenin signaling, a well-characterized HCC-driving pathway, to highlight the relevance of this model to HCC, and on PKA/cAMP signaling, the top steatosis-enriched CIS gene associated pathway (Supplementary Table S13).
Wnt/β-catenin signaling pathway alterations in steatosis-associated HCC
IPA indicated the Wnt/β-catenin signaling pathway was significantly associated with the steatosis-associated (P = 0.001) and the steatosis-enriched (P = 0.007) CIS gene lists. Several Wnt/β-catenin signaling genes showed significant expression changes in wild-type, high-fat diet-fed mouse liver tumors. Like the mouse tumors, we found Wnt/β-catenin signaling was significantly associated with genes altered in human alcohol-associated HCC (P = 0.015) and several pathway genes had expression changes in human steatosis-associated HCC (Supplementary Fig. S11 and Supplementary Table S17).
PKA/cAMP signaling pathway alterations in steatosis-associated HCC
IPA indicated the PKA/cAMP signaling pathway was significantly associated with the steatosis-associated (P = 0.0001) and steatosis-enriched (P < 0.0001) CIS gene lists, and several pathway genes had altered expression in wild-type, high-fat diet-fed mouse liver tumors. PKA/cAMP signaling was also associated with genes with significant expression changes in human steatosis- or alcohol-associated HCC from TCGA (P = 0.016 and P < 0.0001, respectively; Supplementary Fig. S12 and Supplementary Table S18).
Activated PKA catalytic subunit phosphorylated at T197 (p-PKA; ref. 31) was examined by immunohistochemical analysis of a human HCC tissue microarray. p-PKA was significantly higher in HCC, with 75/99 total cases and 3/5 steatosis-associated cases displaying stronger p-PKA staining than matched normal liver (P < 0.0001 and P = 0.005, respectively, Wilcoxon signed-rank test), further supporting a role for PKA/cAMP signaling activation in steatosis-associated HCC (Fig. 6A and B).
Activated PKA drives steatosis-associated HCC. A, p-PKA immunohistochemical stain intensity of human HCC and matched nontumor liver samples on a tissue microarray (TMA), stratified by mention of steatosis in pathology report as in Fig. 4E. B, Representative immunohistochemistry for p-PKA-stained sections of human HCC (top) or matched normal liver (bottom) from tissue microarray cases with hepatic steatosis indicated in pathology reports. C, Immunohistochemistry for p-PKA, FAH, and ALB-stained liver sections from eCDD-treated mice injected with GFP/shp53 (top) or PKA/shp53 (middle and bottom). D, Prkaca expression measured by qRT-PCR from nontumor liver tissue (L) or liver tumor (T) from PKA/shp53 or GFP/shp53-injected eCDD-treated mice, normalized to Actb and to wild-type mouse liver Prkaca (n = 5 each). E, p-PKA immunohistochemistry stain intensity of nontumor liver tissue (L) or liver tumors (T) from PKA/shp53 (n = 17 each) or GFP/shp53-injected mice (n = 13) from both diets combined. F and G, Box-and-whisker plot of tumor burden (F) or tumor penetrance for PKA/shp53-injected mice treated with eCDD (n = 33) or ND (n = 26) or GFP/shp53-injected mice treated with eCDD (n = 36) or ND (n = 42; G). Scale bars, 100 μm. T, tumor. N, nontumor liver tissue. #, P > 0.05; *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001. Error bars, SEM.
Activated PKA drives steatosis-associated HCC. A, p-PKA immunohistochemical stain intensity of human HCC and matched nontumor liver samples on a tissue microarray (TMA), stratified by mention of steatosis in pathology report as in Fig. 4E. B, Representative immunohistochemistry for p-PKA-stained sections of human HCC (top) or matched normal liver (bottom) from tissue microarray cases with hepatic steatosis indicated in pathology reports. C, Immunohistochemistry for p-PKA, FAH, and ALB-stained liver sections from eCDD-treated mice injected with GFP/shp53 (top) or PKA/shp53 (middle and bottom). D, Prkaca expression measured by qRT-PCR from nontumor liver tissue (L) or liver tumor (T) from PKA/shp53 or GFP/shp53-injected eCDD-treated mice, normalized to Actb and to wild-type mouse liver Prkaca (n = 5 each). E, p-PKA immunohistochemistry stain intensity of nontumor liver tissue (L) or liver tumors (T) from PKA/shp53 (n = 17 each) or GFP/shp53-injected mice (n = 13) from both diets combined. F and G, Box-and-whisker plot of tumor burden (F) or tumor penetrance for PKA/shp53-injected mice treated with eCDD (n = 33) or ND (n = 26) or GFP/shp53-injected mice treated with eCDD (n = 36) or ND (n = 42; G). Scale bars, 100 μm. T, tumor. N, nontumor liver tissue. #, P > 0.05; *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001. Error bars, SEM.
Activated PKA promotes steatosis-associated hepatocarcinogenesis in vivo
To test the effects of PKA/cAMP signaling on tumorigenesis in steatotic livers, we used SB transposon-based gene delivery for stable hepatic PrkacaL206R (a constitutively active form; ref. 32) expression in Fah-deficient mice treated with eCDD and ND as described above. We used an SB transposon-based expression vector for V5-tagged mouse PrkacaL206R coexpressing Fah with a GFP reporter (pT2/GD-PrkacaL206R) plus pT2/shp53 (PKA/shp53; Fig. 5A and B and Supplementary Table S7).
PKA/shp53 and GFP/shp53 mouse livers expressed comparable levels of Fah transcript (P = 0.80, Student t test; Supplementary Fig. S13A), and FAH was detected by immunohistochemistry (Fig. 6C), indicating efficient liver repopulation with transfected cells. PKA/shp53-injected mouse livers and tumors overexpressed Prkaca transcript (P = 0.0004 and 0.001, respectively, Student t test; Fig. 6D) and had higher p-PKA immunohistochemical staining levels (P < 0.0001 in both cases, Wilcoxon rank-sum test; Fig. 6C and E) compared to GFP/shp53 mouse livers. V5 was detected by immunohistochemistry in PKA/shp53 mouse livers and tumors but not GFP/shp53 mouse livers (Supplementary Fig. S13B). More eCDD-treated mouse livers (19/24) examined histopathologically showed fatty change than ND-treated (0/9; P < 0.0001, χ2 test).
eCDD-treated PKA/shp53 (n = 33) mice developed more tumors than GFP/shp53 mice (n = 36), with a mean of 2.4 versus 0.7 tumors per mouse (P = 0.017, Wilcoxon rank-sum test) and had higher tumor penetrance, with 52% versus 31% developing tumors (P = 0.011, χ2 test; Fig. 6F and G). Low tumor burdens were observed in ND-treated mice, with a mean of 0.5 versus 0.1 tumors per PKA/shp53 (n = 26) and GFP/shp53 mouse (n = 42; P = 0.10, Wilcoxon rank-sum test; Fig. 6F). Tumor penetrance was higher in ND-treated PKA/shp53 mice than GFP/shp53, at 22% vs. 7% (P = 0.010, χ2 test; Fig. 6G). PKA/shp53 mouse tumor burdens were significantly higher with eCDD treatment than ND (P = 0.036, Wilcoxon rank-sum test), but tumor penetrance was not significantly different (P = 0.11, χ2 test).
PKA/shp53-induced liver tumors were histopathologically classified as well-differentiated hepatocellular neoplasms (86%) or HCCs (14%). Albumin was detected in PKA/shp53 mouse liver tumors by immunohistochemistry, confirming hepatocyte origin (Fig. 6C).
Discussion
Hepatic steatosis is a major and increasingly widespread HCC risk factor (3). Here we report the results of a genetic screen for HCC drivers in mice with diet and alcohol-induced steatosis. Because HCC occurs most frequently in men, many mouse models of HCC include only males (33). We found, however, that the proportion of women developing HCC is higher in the context of hepatic steatosis. Our steatosis-associated HCC mouse model recapitulated this reduction in sex bias. The inclusion of females in studying HCC pathogenesis is, therefore, particularly important in this context, and drivers discovered in this screen may have broader relevance to human steatosis-associated HCC than those identified in studies conducted exclusively in males.
The considerable genetic heterogeneity in human HCC necessitates the use of methods complementary to patient sample profiling for driver alteration identification (7). Analysis of recurrently mutated genes in tumors from steatotic mouse livers undergoing transposition identified 203 candidate steatosis-associated HCC genes, the majority of which are altered in human HCC. Most expression changes in human steatosis- or alcohol-associated HCC were consistent with predicted roles of CIS genes as oncogenes or tumor suppressors, providing additional evidence that our screen identified a subset of promising candidate driver alterations in human steatosis- and alcohol-associated HCC.
We identified NAT10 as a hepatic oncogene. NAT10 is a lysine and RNA acetyltransferase with a variety of reported acetylation targets and functions, including regulating gene expression, cell division, and precursor rRNA processing (34–36). Normal expression is required for the formation of N4-Acetylcytidine-modified rRNA, essential for 18S rRNA formation during ribosome biogenesis (36). The rate of ribosome biogenesis regulates cell proliferation (37) and could be rate-limiting in HCC. Ribosome biogenesis inhibition using an RNA polymerase I specific inhibitor, CX-5461, reduces proliferation and induces senescence in many cancer cell types in vitro and in vivo (37, 38). In steatotic livers, NAT10 overexpression may allow cells to respond to abundant acetyl-CoA, produced because of steatosis, by increasing ribosome biosynthesis, potentially bypassing a rate-limiting step in hepatocarcinogenesis.
Our study and other published work (39) showed NAT10 protein overexpression in over 60% of HCC cases regardless of steatosis, a higher proportion than is indicated by transcript overexpression on TCGA (6.7%). High NAT10 expression has been found to correlate with high TNM classification and poor survival outcomes (39) and to promote invasion and migration of HCC cells (40). NAT10 knockdown decreases proliferation in several cancer cell types (36, 41). Our data demonstrates a direct role for Nat10 overexpression in liver tumorigenesis in a mouse model regardless of steatosis. One study found no significant correlation between high NAT10 expression and patient gender, age, or hepatic cirrhosis (39). A NAT10 acetyltransferase chemical inhibitor, Remodelin, has been characterized (42). Our Nat10 overexpression mouse models could be used to test the efficacy of Remodelin and CX-5461 for preventing or treating NAT10-driven HCC within and without the context of hepatic steatosis.
PKA/cAMP signaling pathway genes were commonly altered in SB-induced steatosis-associated mouse liver tumors. This pathway was similarly disrupted in an independent mouse model of high-fat diet-induced steatosis-associated HCC and in human steatosis-associated HCC, demonstrating the broader relevance of our model to hepatocarcinogenesis in steatotic livers. We validated an oncogenic role for activated PKA in steatotic livers in vivo. The PKA/cAMP signaling pathway regulates glucose and lipid metabolism, proliferation, and hepatocyte survival (43–45). The PKA catalytic subunit, PRKACA, is commonly activated by mutations in adrenocortical tumors (32) and recurrent translocations in fibrolamellar HCC (32, 46). β-Adrenergic receptors activate PKA/cAMP signaling in many cell types including hepatocytes and regulate many cancer-relevant cellular processes including growth, metabolism, DNA damage repair, cell motility, survival, and inflammation (47). Several retrospective studies have found inhibition of β-adrenergic receptor signaling by nonselective β-blockers in patients with cirrhosis may decrease HCC incidence, an effect that might involve PKA/cAMP signaling (48–50). Our identification of frequent alterations in human and mouse tumors support a role for PKA/cAMP signaling in steatosis-associated HCC.
The candidate HCC drivers we identified in the highly relevant context of hepatic steatosis may provide insight into which of the many genes altered in human HCC are drivers, the pathways promoting HCC in steatotic livers, and potential strategies for HCC prevention or treatment.
Disclosure of Potential Conflicts of Interest
D. Largaespada is a Chairman at B-MoGen Biotechnologies, Inc., is a Chief Science Officer at Recombinetics/Surrogen, Inc., is a scientific advisor at Immunosoft, Inc. and NeoClone Inc.; reports receiving a commercial research grant from Genentech, Inc.; and has ownership interest (including patents) in B-MoGen Biotechnology, Inc., Immusoft Inc., NeoClone Inc., and Recombinetics Inc. No potential conflicts of interest were disclosed by the other authors.
Authors' Contributions
Conception and design: B.R. Tschida, L.R. Roberts, V.W. Keng, D.A. Largaespada
Development of methodology: A.L. Sarver, L.R. Roberts, V.W. Keng
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): B.R. Tschida, T.P. Kuka, L.A. Lee, J.D. Riordan, C.A. Tierrablanca, R. Hullsiek, S. Wagner, W.A. Hudson, P.J. Beckmann, R.A. Heuer, J.D. Yang, L.R. Roberts, J.H. Nadeau, V.W. Keng
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): B.R. Tschida, N.A. Temiz, J.D. Riordan, W.A. Hudson, M.A. Linden, A.L. Sarver, J.D. Yang, L.R. Roberts, A.J. Dupuy, V.W. Keng, D.A. Largaespada
Writing, review, and/or revision of the manuscript: B.R. Tschida, N.A. Temiz, J.D. Riordan, R. Hullsiek, M.A. Linden, P.J. Beckmann, J.D. Yang, L.R. Roberts, J.H. Nadeau, V.W. Keng, D.A. Largaespada
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): L.R. Roberts, V.W. Keng
Study supervision: L.R. Roberts, V.W. Keng, D.A. Largaespada
Other (revision of manuscript for content): L.A. Lee
Other (histologic analysis of liver tissue): K. Amin
Disclaimer
The content is solely the responsibility of the authors and does not necessarily represent the official views of the funding institutions.
Acknowledgments
These results are in part based upon public data generated by TCGA Research Network: http://cancergenome.nih.gov/. This work utilized University of Minnesota Supercomputing Institute computing resources. The University of Minnesota Genomics Center provided Illumina sequencing, Sanger sequencing, and primer synthesis services. Carlyn Iverson provided illustration assistance.
Grant Support
This work was supported by grants from the NIH to D.A. Largaespada (NCIR01CACA132962), B.R. Tschida (T32 AI083196-04), and J.D. Riordan (F32 DK109651); from the American Cancer Society to D.A. Largaespada (Research Professor Award #123939); from The Hong Kong Polytechnic University to V.W. Keng (G-YBAY); and from The Shenzhen Science and Technology Innovation Commission to V.W. Keng (JCYJ20170413154748190).
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.