Phenobarbital sodium salt, 1,4-bis[2-(3,5-dichloropyridyloxy)]benzene (TCPOBOP), anisomycin, and DEN were purchased from Sigma-Aldrich (St. Louis, MO, USA). Fugene 6 from Promega (Ann Arbor, MI) and Lipofectamine 3000 from Life Technologies (Grand Island, NY, USA) were used for cell transfection of plasmid DNA; TaqMan Gene Expression Assays (Thermo Fisher, Waltham, MA) for gene expression analyses used in this study were as follows: Akr1b7, Mm00477605_m1; Casp1, Mm00438023_m1; Cdh1, Mm01247357_m1; Cyp2b9, Mm00657910_m1; Cyp2b10, Mm00456591_m1; Cyp2b13, Mm00771172_g1; Gadd45β, Mm00435123_m1; Fndc5, Mm01181543_m1; Krt19, Mm00491980_m1; Nedd9, Mm01324843_m1; Sult1e1, Mm00499178_m1; Tgfbr2, Mm03024091_m1; and Mouse GAPDH Endogenous Control. Irisin peptide was a product of Raybiotech (Peachtree Corners, GA, USA). Expression plasmid for FNDC5/Irisin was constructed by amplifying cDNA fragment using mouse liver cDNA and cloned into pcDNA3.1 V5-His TOPO (Thermo Fisher). Human TGFBR2 expression vector (pCMV5B-TGFbeta receptor II) was a gift from Joan Massague and Jeff Wrana (Addgene plasmid #11766) (29),

C57BL/6J (Catalog# JAX 000664) and Gadd45β KO (B6;129S6-Gadd45βtm1Flv/J) (Catalog# JAX 013101) were obtained from Jackson Laboratories (Bar Harbor, ME, USA) and were maintained at the National Institute of Environmental Health Sciences (NIEHS, USA). Mice received a single i.p. treatment with DEN (90mg/kg body weight) at 5 weeks of age as in previous reports (7, 30, 31). Two weeks after the DEN treatment, mice were placed on PB (0.5 g/l) via drinking water or normal water as a control for 50 weeks. The water bottle with or without PB was changed every week. Based on 5 ml water consumption per mouse for one day, we expected 0.875 g of PB/mouse for the study. At the end of the 50-week period, mice were euthanized by CO₂ and livers were excised for mRNA analyses. Histopathological evaluation of livers for tumor induction in Gadd45β KO mice and wild-type (WT) C57BL/6J mice was performed to see changes in hepatocytes including hypertrophy, foci and tumors. All animal procedures were approved by the Animal Care and Use Committee at NIEHS, NIH, and performed humanely in accordance with Public Health Service Policy.

HepG2 (human hepatocellular carcinoma cell line) and AML12 (mouse hepatocyte cell line) were obtained from the American Type Culture Collection (ATCC). HepG2 and AML12 cells were cultured in Eagle’s minimum essential media and DMEM/F12 media, respectively, supplemented with 10% fetal bovine serum (FBS) and penicillin/streptomycin at 37°C with 5% CO₂. HepG2 cells were transfected by Fndc5/Irisin and Tgfbr2 expression plasmids using Fugene 6 reagent according to the manufacturer’s instructions. FNDC5/Irisin and Tgfbr2 protein expressions were detected by western blotting analyses using 10% SDS page gels employing anti V5 tag antibody (Thermo Fisher) and anti Tgfbr2 monoclonal antibody (Santa Cruz Biotechnology), respectively. AML12 cells in 96 well plates (seeded 5000 cells per well) were treated with irisin (1 μg/ml or 2 μg/ml) at 24 hr. Seventy-two hr after transfection (HepG2) or irisin treatment (AML12), cell viability was determined using Cell Counting Kit-8 (Dojindo).

All tissue samples for gene expression analysis were obtained from non-tumorous regions of liver tissues. RNAs were purified from mouse livers using Trizol reagent (Thermo Fisher Scientific) and the RNeasy mini kit (QIAGEN, Valencia, CA, USA). Gene expression analysis was conducted using Agilent Whole Mouse Genome 4 × 44 multiplex format oligo arrays (014868) (Agilent Technologies, Palo Alto, CA, USA) following the Agilent 1-color microarray-based gene expression analysis protocol. Cy3 labeled cRNA was produced following the manufacturer’s protocol. Cy3 labeled cRNAs were fragmented and hybridized for 17 h in a rotating hybridization oven. Slides were scanned with an Agilent Scanner. Data were obtained using the Agilent Feature Extraction software (v12). The Agilent Feature Extraction Software performed error modeling, adjusting for additive and multiplicative noise. Genes were considered differentially expressed if they showed a fold-change of at least 1.5 with p-value < 0.05 tested by an ANOVA and Benjamini-Hochberg multiple test correction performed using OmicSoft Array Studio (Version 10) software. Ingenuity Pathway Analysis software (IPA, Qiagen) was utilized to search the canonical pathway, toxicity function, or upstream regulator. The pathways with Z-score >2 and <-2 were considered significantly activated and repressed pathways, respectively. Gene set enrichment analysis (GSEA, version 4.0.3) was performed with the oncogenic gene sets (H) from Molecular Signature Database (MSigDB, v7.1). Gene set with NOM p-value < 0.05 was considered as significantly enriched.

The transcriptome data as fragments per kilobase of transcript per million mapped reads upper quartile (FMPK-UQ) and clinical information of hepatocarcinoma patients in TCGA-LIHC was downloaded from the GDC data portal (https://portal.gdc.cancer.gov/). The hepatocarcinoma samples of 377 patients were downloaded. The patients were grouped as low and high GADD45B by FMPK score of GADD45B as lower than 40 (n=121) and greater than 85 (n=133), respectively. Subsequently, the patients in high GADD45B were grouped as low and high FNDC5 by FMPK score of FNDC5 as lower than 4 (n=55) and greater than 10 (n=57), respectively. The hepatocarcinoma patients were grouped as low and high FNDC5 by FMPK score of FNDC5 as lower than 1 (n=133) and greater than 6 (n=137), respectively. Kaplan-Meier survival curve with log-rank test was performed to compare differences in survival distributions by using Python lifelines (https://lifelines.readthedocs.io/en/latest/, version 0.27.1).

Statistical analyses were performed using Graphpad Prism (version 8.3.0, Graphpad Software, La Jolla, CA). A Grubbs test was performed for the statistical outlier evaluation for the assessment of liver weight/body weight.

In previous studies using the DEN/PB mouse HCC experimental system, our group found Gadd45β is induced in mouse liver through CAR activation by PB. Since Gadd45β has been considered to play an anti-tumor role in some cancers (18), our logical hypothesis was that this molecule is playing a key role in HCC tumorigenesis. Here we compared HCC patient survival rates between groups with low and high GADD45B expression. For this analysis, human HCC patients from the public domain (TCGA-LIHC) were grouped based on liver cancer stages. The data shows that GADD45B expression levels decreased according to liver cancer stage progression ( Figure 1A ). Strikingly, the Kaplan-Meier survival curve with log-rank test showed that the group with lower GADD45B expression has a poorer prognosis than the group with higher GADD45B expression, suggesting that GADD45B has antitumor functions in human HCC.

To evaluate the effect of Gadd45β null mutation for PB-promoted liver tumors in mice, approximately 40 C57BL/6 WT (Gadd45β WT) and 40 Gadd45β null mutant mice (Gadd45β KO) were injected with DEN, and half of the mice in each group were treated with PB in drinking water. As shown in Figure 2A , fifty weeks after administration of PB, necropsies were performed, livers were removed, and mRNA expression levels were analyzed (N=6). Real-time PCR confirmed that Gadd45β was not expressed in Gadd45β KO mice ( Figure 2B ). Consistent with our previous results, PB treatment significantly increased the expression of Gadd45β. The numbers of basophilic foci, eosinophilic foci, adenoma, and carcinoma were analyzed in all livers excised from each group. The numbers of foci and tumors in each group are shown in Table 1 . Five cases of adenoma were observed in Gadd45β WT and only one in Gadd45β KO in the groups with PB treatment ( Table 1 ). In addition, carcinoma was observed in only one case in Gadd45β WT, which was not observed in Gadd45β KO ( Table 1 , Figure 2C ). These results strongly suggested that Gadd45β plays a promoting role in PB-mediated tumorigenesis.

Subsequently, liver weight was analyzed for the assessment of hepatocyte proliferation. The Grubbs test detected the values of liver weight per body weight of two mice, one with carcinoma and the other with adenoma, of PB/DEN treated Gadd45β WT group as outliers. Therefore, these two samples were excluded in the following analysis (data including those two are shown in Supplemental Figure 1 ). Although liver weight per body weight was increased by PB in both Gadd45β WT and Gadd45β KO mice, that of livers from PB-treated Gadd45β KO mice was significantly smaller compared to that of livers from Gadd45β WT mice ( Figure 2D ). These results support that Gadd45β promotes hepatocyte proliferation in the presence of PB. In contrast, without PB, Gadd45β WT mice developed no adenoma, whereas Gadd45β KO had two adenomas ( Table 1 ). This result, which may indicate that Gadd45β functions as a tumor suppressor in the absence of PB, is consistent with previously known Gadd45β function in tumorigenesis.

Eosinophilic foci and adenoma development in rodents are characteristic of liver tumor promotion by PB (32). Eosinophilic foci typically consist of hepatocytes which are stained more eosinophilic and are often larger than surrounding hepatocytes (33.). This is consistent with our previous tumorigenesis study with Car KO mice, which were changed to a liver tumor-susceptible mouse strain by repeated backcrossing to C3H/HeNCrlBR mice. In the present study we also observed PB administration increased eosinophilic foci. However, even though Gadd45β KO mice exhibited lower incidences of adenoma and carcinoma than Gadd45β WT mice in the presence of PB, the same mice showed higher incidences of eosinophilic foci. This histological result indicates that hepatic conditions of eosinophilic foci may not be enough to promote tumorigenesis.

Results obtained in Figure 2 suggest that the Gadd45β null mutation provoked fundamental changes in mouse liver responses against DEN/PB treatment. Therefore, we analyzed PB-induced transcriptome responses of wild type and Gadd45β null mutant mice livers expecting to find the key gene expression changes derived from the null mutation resulted in altered tumor development. For this purpose, RNAs isolated from the livers in Figure 2 were subjected to cDNA microarray analyses as described in the Methods section. We found differentially expressed genes by PB treatment in Gadd45β WT (134 genes) and Gadd45β KO (262 genes) and applied Ingenuity Pathway Analysis (IPA) for the results ( Figures 3A, B ). Canonical pathway analysis results shown in Figure 3A suggested Gadd45β gene null mutation provoked widespread effects on PB-induced gene expression in numerous pathways including those involved in endo- and exogenous metabolism. For example, genes in nicotine degradation, estrogen biosynthesis, and melatonin degradation were strongly affected by the null mutation. Furthermore, provocative differences were found by IPA Toxicity functional analysis ( Figure 3B ) in which clear differences were observed in cancer-related genes of multiple tissues including those of the liver. Subsequently, the Venn diagram visualized the unique differentially expressed genes in Gadd45β WT and KO mice ( Figure 3C , Supplementary Table 1 ). Among those genes, we focused on two genes, Fndc5 and Tgfbr2. Increased FNDC5 expression in human HCC was observed in a previous report (28) and this gene was known to be a CAR target gene (34). On the other hand, Tgfbr2 was reported to have putative CAR response elements and to be CAR dependently induced in DEN/PB-treated mice (11). Our qPCR confirmed these two genes were induced by DEN/PB compared with DEN alone in WT mice and the inductions were much less prominent in Gadd45β KO mice ( Figures 3D, E ).

Next, we assessed the effect of single-dose PB treatment on gene expression levels of Fndc5 and Tgfbr2 in CAR WT and KO mice ( Figure 4 ). Fndc5 mRNA was induced by PB treatment in CAR WT mice whereas no change was observed in CAR KO mice, indicating Fndc5 is a CAR target gene. In contrast, Tgfbr2 was induced by PB treatment in neither CAR WT mice nor CAR KO mice, suggesting the expression change of Tgfbr2 in DEN/PB-treated Gadd45β WT mice was caused by in-direct functions of CAR activated by long-term treatment with PB.

Irisin has been reported to induce cell proliferations for many cell types including HepG2 cells, one of human hepatocellular carcinoma derived cell lines. Here we analyzed the effects of irisin treatment on the proliferation of mouse hepatocyte AML12. The commercial Irisin we used here has 100% identical peptide sequences between humans and mice. Cell numbers of AML12 cells were increased by irisin treatment after 72 hr, indicating proliferation was promoted by irisin ( Figure 5A ). Irisin induced gene expression changes in AML12 cells were analyzed by cDNA microarray followed by GSEA with C6 gene sets and IPA with upstream analysis ( Figure 5B , Table 2 ). In this GSEA analysis, we compared Fndc5/Irisin-regulated genes found in our microarray with previously found genes regulated by TGF-β. The comparison revealed genes down-regulated by transiently expression of TGF-β, were enriched in up-regulated genes by irisin treatment ( Figure 5B ). Moreover, the IPA analysis found 10 upstream regulator candidates that are affected by irisin treatment. Among them, irisin has a potential function for suppression of the Tgfbr1 signaling pathway ( Table 2 ), suggesting that these factors, Fndc5/Irisin, TGF-β and Tgfbr1, have functional interactions.

Next, we assessed the effect of irisin treatment on human hepatocarcinoma HepG2 cell growth. Irisin has been reported to bind with Tgfbr2 resulting in interrupted TGF-β1 signaling pathway (35). We assessed the effect of the combination of transiently expressing irisin and TGFBR2 on the proliferation of HepG2 cells. FNDC5/Irisin and Tgfbr2 protein expressions were confirmed by western blotting analyses in Figure 6A . The expression of either irisin or TGFBR2 promoted cell proliferation and simultaneous expression of both factors yielded a greater promotion ( Figure 6B ). These results suggested that the two factors have a functional interaction for HepG2 cell proliferation. Next, we analyzed the relation of FNDC5 expression levels with HCC patients’ survival rates ( Figures 6C, D ). Whereas a higher FNDC5 expression group does not show a poorer prognosis ( Figure 6C ), higher FNDC5 expression in higher GADD45B expression groups tends to have a poorer prognosis ( Figure 6D ). Thus, FNDC5 has cooperative effects with GADD45B on prognosis of cancer patients.