Myc and ChREBP transcription factors cooperatively regulate normal and neoplastic hepatocyte proliferation in mice

Analogous to the c-Myc (Myc)/Max family of bHLH-ZIP transcription factors, there exists a parallel regulatory network of structurally and functionally related proteins with Myc-like functions. Two related Myc-like paralogs, termed MondoA and MondoB/carbohydrate response element–binding protein (ChREBP), up-regulate gene expression in heterodimeric association with the bHLH-ZIP Max-like factor Mlx. Myc is necessary to support liver cancer growth, but not for normal hepatocyte proliferation. Here, we investigated ChREBP's role in these processes and its relationship to Myc. Unlike Myc loss, ChREBP loss conferred a proliferative disadvantage to normal murine hepatocytes, as did the combined loss of ChREBP and Myc. Moreover, hepatoblastomas (HBs) originating in myc−/−, chrebp−/−, or myc−/−/chrebp−/− backgrounds grew significantly more slowly. Metabolic studies on livers and HBs in all three genetic backgrounds revealed marked differences in oxidative phosphorylation, fatty acid β-oxidation (FAO), and pyruvate dehydrogenase activity. RNA-Seq of livers and HBs suggested seven distinct mechanisms of Myc–ChREBP target gene regulation. Gene ontology analysis indicated that many transcripts deregulated in the chrebp−/− background encode enzymes functioning in glycolysis, the TCA cycle, and β- and ω-FAO, whereas those dysregulated in the myc−/− background encode enzymes functioning in glycolysis, glutaminolysis, and sterol biosynthesis. In the myc−/−/chrebp−/− background, additional deregulated transcripts included those involved in peroxisomal β- and α-FAO. Finally, we observed that Myc and ChREBP cooperatively up-regulated virtually all ribosomal protein genes. Our findings define the individual and cooperative proliferative, metabolic, and transcriptional roles for the “Extended Myc Network” under both normal and neoplastic conditions.

The c-Myc (Myc) oncoprotein is deregulated in many human tumors while also playing roles in normal proliferation, differentiation, and metabolism (1)(2)(3)(4)(5). Myc coordinates the transcription of thousands of protein-coding genes, rRNAs and tRNAs, microRNAs, and long-noncoding RNAs (6 -15). However, its role in maintaining growth and survival of normal or neoplastic cells depends on the timing and level of its expression and its cellular context. For example, whole-body depletion of Myc in the embryo is lethal but is compatible with postnatal survival (16 -18). Indeed, total-body dominant-negative inhibition of endogenous Myc in older mice is associated with mild and reversible intestinal epithelial and hematopoietic hypoplasia and long-term survival (19), yet these mice demonstrate regression of Ras oncogene-induced lung adenocarcinomas, thereby demonstrating the differential Myc dependence of normal and transformed tissues. Myc hypomorphism is associated with smaller organs resulting from a reduction in cell number but not size (18).
The consequences of Myc knockout in the liver have been controversial, particularly with regard to its role in regeneration following partial hepatectomy (PH) 3 (20 -23). We recently reexamined this in a novel mouse model of Type I hereditary tyrosinemia in which the replacement of diseased hepatocytes by transplanted nontyrosinemic hepatocytes can be monitored over several months (24 -26). Using competitive transplants, we demonstrated that WT and myc Ϫ/Ϫ (Myc-KO) hepatocyte populations possessed similar long-term repopulation poten-This work was supported by National Institutes of Health RO1 Grants DK103645 (to A. W. D.) and CA174713 (to E. V. P.). The authors declare that they have no conflicts of interest with the contents of this article. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. This article contains Figs. S1-S11 and Tables S1-S11.

Raw and processed original data have been deposited in the National Center for
Biotechnology Information (NCBI) Gene Expression Omnibus (GEO) under accession number GSE114634. 1  cro ARTICLE tial. In contrast, hepatoblastoma (HB) development driven by mutant forms of ␤-catenin and yes-associated protein (YAP) was markedly delayed in Myc-KO livers (13,27), indicating that Myc plays distinct roles in normal and neoplastic proliferation. Alternatively, Myc-KO hepatocytes could be rescued by factors that functionally substitute for Myc. Indeed, Myc-KO HBs upregulated expression of the weakly transforming Myc paralog L-Myc (13,28,29).
Myc, L-Myc, and another paralog, N-Myc, are members of a group of related bHLH-ZIP transcription factors that bind canonical E-box elements (CA(C/T)GTG) and regulate gene expression. The central member of this group is Max, whose dimerization with Myc is necessary for E-box binding. In contrast, heterodimers formed between Max and the four bHLH-ZIP members of the Mxd family compete with Myc-Max heterodimers for DNA binding and transcriptional modulation (30,31). Hereafter, we refer to these proteins as the "Myc-Max-Mxd" family.
Following the discovery of the Myc-Max-Mxd family, it was appreciated that a related bHLH-ZIP family with Myc-like functions exists. Analogously, this also consists of a central, Max-like bHLH-ZIP protein, Mlx; two closely related Myc-like factors termed MondoA/Mlxip and MondoB/ChREBP/Mlxipl (hereafter ChREBP) that positively regulate gene expression; and a negative regulator termed Mnt (32)(33)(34). We refer to this as the "ChREBP-Mlx-Mnt" family and to the two families collectively as the "Extended Myc Network" (Fig. S1A). Cross-talk occurs by virtue of the fact that Mxd1 and Mxd4 also interact with Mnt. The canonical binding site for the ChREBP-Mlx-Mnt family, termed the carbohydrate response element (ChoRE), consists of two Myc-like E-box elements separated by 5 nucleotides (CAYGYGnnnnnCRCRTG) (35)(36)(37)(38). Whereas the two families regulate common genes (14,39,40), those under ChREBP-Mlx-Mnt family control comprise a subset of those regulated by the Myc-Max-Mxd family, particularly those involved in carbohydrate and lipid metabolism (14,32,34,36,(41)(42)(43)(44)(45). Moreover, unlike nuclearly localized Myc proteins, MondoA and ChREBP shuttle between the nucleus and cytoplasm in response to their binding of glucose 6-phosphate, thereby suggesting that they serve as nutrient sensors and regulate target gene expression in accord with metabolic status (34,40,46,47).
We have now compared both the repopulation and tumorigenic potential of Myc-KO (or MKO) hepatocytes with those lacking ChREBP (ChREBP-KO or CKO) and both Myc and ChREBP (Double-KO or DKO). We chose for study ChREBP rather than the related MondoA because, even though both are widely distributed, ChREBP expression is more prominent in liver and appears to play a more important role (32,48,49). Our findings demonstrate that, in contrast to Myc, ChREBP's absence leads to a profound proliferative disadvantage in competitive hepatocyte repopulation assays and that Double-KO hepatocytes are even more impaired. HB growth rates were also markedly slowed, although tumor initiation remained unaffected. A combination of metabolic, biochemical, and transcriptional profiling better defined the abnormal phenotypes. Our findings indicate that the inhibition of normal and neoplastic proliferation resulting from Myc and/or ChREBP loss originates from multiple and in some cases overlapping and cooperative effects on key metabolic processes, primarily centered around carbohydrate and lipid metabolism and protein synthesis.

Myc-KO, ChREBP-KO, and Double-KO livers accumulate neutral lipid
Livers from young Myc-KO mice appear relatively normal but gradually accumulate neutral lipids (25). We confirmed this by demonstrating more prominent vacuolization in hematoxylin and eosin-stained Myc-KO livers (Fig. S1B). ChREBP-KO livers from comparably aged mice were even more abnormal, with lipids tending to be particularly prominent in subcapsular and periendothelial regions. Similar changes were seen in Double-KO livers, with certain areas being more noticeably affected. Hepatic triglyceride content largely reflected these findings (Fig. 1C). Thus, ChREBP appears to play a more important role than Myc in maintaining hepatic triglyceride content.
In contrast, HBs were largely devoid of neutral lipids, although occasional areas of vacuolization were observed. However, the adjacent regions of nontransformed parenchyma largely retained the patterns of neutral lipid staining described above (Fig. S1B). Tumor triglyceride content was also uniformly low (Fig. S1B).

ChREBP-KO and Double-KO hepatocytes function poorly in repopulation assays
Liver repopulation by WT and Myc-KO hepatocytes occurs at similar rates (25). To assess this, we employed the hereditary tyrosinemia model described above. Donor mice lack the enzyme fumaryl acetoacetate hydrolase (FAH) (50) and succumb to the progressive accumulation of toxic tyrosine catabolites unless rescued with the drug 2-(2-nitro-4-trifluoromethyl-benzoyl)-1,3-cyclohexanedione (NTBC), which blocks the upstream enzyme p-hydroxyphenylpyruvate dioxygenase. Mice can also be rescued by the intrasplenic injection of fah ϩ/ϩ hepatocytes (25,26,51,52). During iterative periods of NTBC discontinuation, donor cells gradually replace their fah Ϫ/Ϫ counterparts and, upon reaching an equilibrium state, comprise as much as 80% of the recipient's hepatocytes. This assay is superior to PH for assessing donor hepatocyte proliferative potential in that it imposes greater replicative demand that can reveal subtle differences in repopulation potential not appreciable with the PH model. Additionally, whereas 30 -40% of post-PH hepatocytes do not replicate and about an equal fraction undergo hypertrophy rather than actual division (53,54), repopulation following transplant assesses only proliferation.
We initially performed transplants using single populations of hepatocytes and noted the time needed to achieve NTBC independence, as defined by the ability of recipients to discontinue the drug and maintain Ͼ80% of their pretransplant weight (25). Consistent with previous findings, we found this time to be 9.5 Ϯ 0.3 and 10.1 Ϯ 0.4 weeks following transplantation with WT and Myc-KO hepatocytes, respectively (not significant as determined by Student's two-tailed t test) (25). In contrast, transplantation with ChREBP-KO hepatocytes required 11.8 Ϯ 0.4 weeks to achieve NTBC independence (p ϭ Myc and ChREBP cooperation in hepatocyte proliferation 5.18 ϫ 10 Ϫ3 , compared with WT). Double-KO hepatocytes required 18.9 Ϯ 1.0 weeks (p ϭ 1.40 ϫ 10 Ϫ8 , compared with WT). Additionally, the time to NTBC independence was significantly prolonged in Double-KO mice versus ChREBP-KO mice (p ϭ 1.91 ϫ 10 Ϫ5 ). Thus, both ChREBP-KO and Double-KO hepatocytes are replication-impaired.
We next performed competitive repopulation assays with ChREBP-KO ϩ WT or Double-KO ϩ WT hepatocytes ( Fig.  1A) (25,51). Upon attaining NTBC independence, a TaqManbased approach was used to quantify total donor and recipient populations (Fig. 1). In both cases, ϳ75% of hepatocytes were found to be of donor origin (Fig. 1C). Further investigation of the ChREBP-KO ϩ WT transplants showed that ChREBP-KO hepatocytes comprised 61.5 and 27.7% of the pre-and posttransplant donor population, respectively ( Fig. 1 (D and E), p ϭ 5.2 ϫ 10 Ϫ10 ). Even more striking were the results obtained with Double-KO ϩ WT transplants, where 52.8% of the pretransplant input donor population but only 7.5% of the posttransplant hepatocytes were Double-KO cells (p ϭ 5.2 ϫ 10 Ϫ10 ). The difference in repopulation efficiencies between ChREBP-KO and Double-KO was highly significant (p ϭ 5.0 ϫ 10 Ϫ6 ). Thus, both ChREBP-KO and even more so Double-KO hepatocytes demonstrated a profound competitive disadvantage.  The values for C and E were obtained from reconstitution experiments using eight different known ratios of each of the two DNA populations being quantified as described previously (51). Correlation coefficients in each case were Ͼ0.99 (not shown).

Growth inhibition of Myc-KO, ChREBP-KO, and Double-KO HBs
We next examined HB tumorigenesis in response to liver expression of oncogenic forms of ␤-catenin and YAP (13,27,51). WT animals succumbed with a mean survival of ϳ13 weeks ( Fig. 2A) (13,27,51), whereas survival in all knockout groups was delayed, and tumor sizes were somewhat smaller (Fig. 2B). Proliferating cell nuclear antigen and phosphohistone H3 levels as markers of proliferation (55,56) were indistinguishable among the tumor groups (Fig. 2C). Thus, ChREBP loss, or combined Myc ϩ ChREBP loss, affects tumor growth rates but not initiation.
We examined expression of other members of the Extended Myc Network as well as other Myc target genes capable of restoring myc Ϫ/Ϫ cell growth, including myct1, hmga1, and shmt (57-60) (Fig. 2D). Aside from confirming that the three knockout tissues lacked expression of their respective fulllength transcripts, we showed variable degrees of change in the livers and tumors of mice with regard to a number of these transcripts, including those for L-Myc, N-Myc, Mxd3, Mnt, Myct1, and Shmt.
Members of the Extended Myc Network possess varying degrees of structural and functional redundancy (14,28,36,(62)(63)(64)(65), suggesting that they might be able to rescue the proliferative defects of KO hepatocytes. As expected, Myc-KO livers and tumors did not express Myc, ChREBP-KO livers and tumors did not express ChREBP, and Double-KO livers and tumors did not express either protein ( Fig. 2E and Fig. S2). Regardless of genotype, all livers and tumors expressed low or undetectable levels of N-Myc. L-Myc levels were variably elevated in all tumors. Similarly, livers expressed low levels of MondoA, which was increased modestly in all tumors. ChREBP was expressed at similar levels in WT and Myc-KO livers and tumors. From these studies and the fact that L-Myc and MondoA are about equally up-regulated in all tumors regardless of genotype, we conclude that neither of these is likely to be expressed at levels sufficient to support non-

Myc and ChREBP cooperation in hepatocyte proliferation
transformed hepatocyte proliferation. The poor correlation between transcript and protein levels for some of these factors ( Fig. 2 (D and E) and Fig. S2) indicates that much of their control was post-transcriptional.

Metabolic alterations in livers and tumors
WT and Myc-KO liver and tumor metabolism differ in several critical ways that impact TCA cycle function, FAO, and levels of critical substrates, such as ATP and acetyl-CoA (AcCoA) (13,25). Because of ChREBP's known influence on carbohydrate and lipid metabolism and its cooperation with Myc in regulating these activities (40, 41, 44, 45, 49, 66 -69), we reevaluated some of these behaviors in our four liver and HB cohorts.
Oxygen consumption rates (OCRs) of WT and Myc-KO livers were similar, as reported previously (Fig. 3A) (13). In contrast, ChREBP-KO liver OCRs were lower than those of WT or Myc-KO livers. This discrepancy was corrected in Double-KO livers, suggesting that, in this largely quiescent population, Myc and ChREBP affected readout oppositely. These differences were largely attributable to alterations in Complex II function, which accounts for the majority of Oxphos in liver (Fig. 3A) (13,25,51). Complex I and Complex II responses were distinguished by measuring OCR before and after the addition of the Complex I inhibitor rotenone. B, reduction of mtDNA content in HBs is independent of tumor growth rates and the presence of Myc and/or ChREBP. Two different sets of TaqMan probes (Set 1 and Set 2) were used to independently amplify different regions of the mitochondrial genome. C, ATP levels in livers and tumors. D, AcCoA levels in livers and tumors. E, PDH activity in fresh livers and tumors. PDH was assessed by measuring the release of 14 CO 2 from 14 C-labeled pyruvate (13,51). F, FAO rates in fresh livers and tumors. FAO was assessed by measuring the release of water-soluble products from 3 H-labeled palmitate-BSA (13,51). G, representative Coomassie Bluestained nondenaturing gels of ETC complexes from livers (L) and tumors (T) isolated from the indicated groups. The panel shows two sets of results. H, typical appearances of developed nondenaturing gels following in situ enzymatic assays for the ETC complexes I, III, IV, and V. I, quantification of ETC complex activities. Four sets of gels similar to those shown in H were scanned and normalized to the activity of Complex IV, which showed little variation among the different tissues (13,25,105). J, complex II activities were assayed in vitro on tissue lysates because the in situ assay is unreliable. All results were normalized to the mean activity of each complex measured in WT livers. Error bars, S.D.

Myc and ChREBP cooperation in hepatocyte proliferation
Also, as shown previously, WT and Myc-KO tumor OCRs were significantly lower than their corresponding normal livers, indicating greater reliance on glycolysis (13,24,51). In contrast, both ChREBP-KO tumors and Double-KO tumors were indistinguishable from their corresponding livers, and this was also reflected in their or in vitro activities The reduction of WT and Myc-KO tumor OCRs was an expected consequence of the Warburg effect, which diverts glycolytic substrates away from mitochondria and into anabolic pathways that abet tumor proliferation (70 -73). We recently reported that, like many human cancers (74), both HBs and Myc-induced hepatocellular carcinomas have a reduced mitochondrial DNA (mtDNA) content relative to livers, thus providing a potential physical basis for their reduced Oxphos and the Warburg effect (13,75). Indeed, this was true for all four sets of HBs (Fig. 3B). Our results indicate that certain ETC functions are under differential control by ChREBP and Myc but that neither is responsible for the maintaining mtDNA.
Despite their reduced mtDNA content, WT and Myc-KO HBs had higher ATP levels than their corresponding livers (13). We confirmed this and further showed that these persisted in ChREBP-KO and Double-KO HBs (Fig. 3C). This correlated with a higher AcCoA content in WT HBs but not in Myc-KO, ChREBP-KO, or Double-KO tumors (Fig. 3D). Thus, the impaired growth of knockout tumors cannot be attributed to ATP deficiencies but could be due to an AcCoA deficit.
Increased pyruvate dehydrogenase (PDH) activity accompanies HB and hepatocellular carcinoma tumorigenesis and may reflect a reduced reliance on FAO and/or an attempt to maximize the conversion of limiting amounts of pyruvate to AcCoA (76). We confirmed this but also showed that Myc-KO, ChREBP-KO, and Double-KO livers and tumors had progressively larger reductions in PDH activity. Most strikingly, PDH activity in Double-KO tumors, although still elevated relative to its corresponding liver, was not significantly higher than the activity in WT livers (Fig. 3E).
The relatively high FAO rate of normal liver is Myc-dependent and is reduced by ϳ50% in its absence. A further ϳ50% decline in FAO rates then occurs in both WT and Myc-KO HBs, with the latter reflecting the already reduced activity in Myc-KO livers (13). FAO rates were equally dependent on ChREBP, and Double-KO livers and HBs had the lowest FAO rates as well as the lowest PDH activities among their respective groups (Fig. 3F). These results reveal that FAO is positively regulated by the combined actions of Myc and ChREBP. Tumors and their corresponding livers therefore progressively suppress PDH and FAO activities that normally function reciprocally and serve as major sources of ATP and AcCoA.
Myc's loss is associated with significant ETC dysfunction in both livers and HBs (Fig. 3A) (13,25). To determine how ChREBP and Myc ϩ ChREBP affected these processes, we resolved ETC complexes using blue native gel electrophoresis (Fig. 3G) and performed functional assays for all complexes. As seen in Fig. 3 (H-J), the only consistent abnormality was a loss of Complex V (ATP synthase) activity in Double-KO livers that was lower than the corresponding activity in Double-KO tumors or other liver or tumor groups.

Transcriptional profiling of livers
We next conducted RNA-Seq on five tissues from each group and identified distinct transcript expression profiles. The relationships of these to one another and the ways in which Myc and ChREBP may potentially interact to differentially affect target gene expression are depicted in Fig. 4 (A-C).

Myc and ChREBP cooperation in hepatocyte proliferation
genes plus three additional down-regulated Cyp450 genes and Kdm5d (Table S3) red) and down (green) arrows signify general up-or down-regulation of genes belonging to a given pathway. Pathways without arrows contain dysregulated genes but could not be assigned directionality with any degree of confidence. An analysis for livers is not shown, as the number of dysregulated genes in each Venn diagram group were generally too low to reliably predict pathway directionality. cent to MYC, often co-amplifies with it, and stabilizes the Myc protein (Table S4) (85). 29 genes were dysregulated following the loss of ChREBP or Myc, but not both (Shared dependent response; Fig. 4 (B-D), Group 6). Of these, 25 were up-regulated and four were downregulated. Among these were the estrogen sulfotransferase Sult1e1 (up-regulated 11-fold) and three MHC Class II genes (Table S5). Members of this group were too few in number to permit any further functional classification using gene ontology analysis.
494 genes were dysregulated in Double-KO livers compared with normal liver and included genes contained within Groups 1, 2, 3, and 7 (281 up-regulated and 213 down-regulated; Fig. 4, B-D). Similar to Myc-KO livers, and more so ChREBP-KO livers, pathway analysis revealed roles for many of these genes in the metabolism of steroids and other lipids (Fig. S3C and http://prochownik.pitt.edu/chrebp/). Dysregulated genes falling into groups 1-3 have been discussed above. Because 232 of these were only dysregulated in Double-KO livers, we classified them as showing a "cooperative response" indicating that loss of either ChREBP or Myc fully compensated for the other's absence (Fig. 4 (B-D), Group 7). Members included 16 down-regulated ribosomal protein (RP) genes (average down-regulation ϭ 1.43-fold) and the gene encoding acetoacetyl-CoA synthetase (Aacs, down-regulated 2.63-fold) (Table S6). 66 genes were dysregulated in all three liver groups (Dual-dependent response; Fig. 4 (B-D), Group 1 and Table S8). Of these, 38 were up-regulated, and 28 were down-regulated. The latter included genes encoding the leptin receptor (Lepr; downregulated 4-fold) and the rate-limiting enzyme in cholesterol synthesis, HMG-CoA reductase (Hmgcr; down-regulated 1.6fold) (Table S7).
Common to both ChREBP-KO and Double-KO tumors and strictly ChREBP-dependent response (Group 2) were 590 upregulated and 557 down-regulated transcripts (Fig. 4 (A, B, E, and F), Groups 1 and 2). These involved pathways regulating Oxphos and mitochondrial function, including 20 down-regulated NADH:ubiquinone oxidase genes and four ATP synthase genes (Fig. S5A). Genes involved in Wnt/␤-catenin signaling were also among this group (Fig. S5B), as were those encoding the mitochondrially localized members of the sirtuin family, sirt3-5 (Fig. S5C). Sirtuins function as lysine deacylase, ADPribosyltransferase, and desuccinylase proteins, and Sirt1 and Sirt3 also serve as tumor suppressors (86,87). Additionally, NRF2-mediated signaling was predicted to be decreased due to altered expression of 21 genes in this pathway (Fig. S5D). NRF2 is a leucine zipper transcription factor whose own targets include numerous genes involved in cytoprotective and antioxidant responses (88,89). Other notable genes in this group include Igf2 (up-regulated 5.6-fold) and the previously noted Pklr (down-regulated 3.5-fold).
1,657 genes were differentially expressed in Myc-KO tumors compared with WT tumors (816 up-regulated and 841 downregulated), with many of the involved pathways being the same as described in ChREBP-KO tumors (Fig. 4, E and F). These included various growth and cancer signaling pathways, metabolic and lipid metabolism pathways, and intracellular signaling (Fig. S6).
The subset of 248 genes dysregulated in both Myc-KO and Double-KO tumors consisted of 121 up-regulated and 127 down-regulated genes (Myc-dependent response; Fig. 4E, Group 3). Notably down-regulated were 12 genes encoding ribosomal proteins and two encoding eukaryotic initiation factors (Fig. S7A). Elevated expression of seven Cyp450 genes was also observed, as well as variable dysregulation of several genes involved in RXR signaling and lipid metabolism, including three phospholipases, carnitine palmitoyltransferase 1B (Cpt1b; down-regulated 2.8-fold) and others (Fig. S7B). Other notable genes in this group included the previously discussed Pvt1 (up-regulated 1.6-fold) and N-Myc (Mycn; up-regulated 5.7-fold).
Compared with WT tumors, 183 genes (101 up-regulated and 82 down-regulated) were differentially expressed when either ChREBP or Myc, but not both, was knocked out (Shared dependent response; Fig. 4E, Group 6). This group was largely concerned with immune signaling, including pathways involving granulocyte adhesion, cellular movement, phagocytosis, and T-cell function.
Double-KO tumors differentially expressed 3,604 transcripts compared with WT tumors (1,769 up-regulated and 1,835 down-regulated). These most commonly functioned in growth and cancer signaling, lipid metabolism and RXR signaling, the immune response, and intracellular signaling pathways (Fig.  S8). Dysregulated genes belonging to Groups 1, 2, and 3 ( Fig.  4A) have been discussed previously. 1,443 genes were only dys-

Myc and ChREBP cooperation in hepatocyte proliferation
regulated in Double-KO tumors and comprised the largest of the seven groups (Cooperative response, Group 7). 734 were up-regulated, and 709 were down-regulated (Fig. 4, E and F). Analysis of this subset was notable for the predicted downregulation of the RhoA signaling pathway (Fig. S9A) and revealed genes involved in nucleotide synthesis and salvage pathways. Additionally, up-regulation of Kdm5d (3.5-fold) and eight cytochromes P450 (average up-regulation ϭ 2.4-fold) was observed, as well as the down-regulation of two additional CYP450 genes (Fig. S9B).
The 766 genes deregulated in all experimental groups comprised the "dual-dependent response" (Fig. 4, A, D, E, and F,  Group 1). This contained 324 up-regulated and 442 down-regulated genes and was most notable for 64 strongly down-regulated RP genes and six eukaryotic initiation factors, which were most prominent in the Double-KO group (Fig. S9C). This included the down-regulation of 10 genes involved in glycolysis and gluconeogenesis (Fig. S9D) and numerous others involved in granulocyte adhesion and diapedesis (Fig. S9E).
To obtain an approximation of the fraction of transcripts that were products of direct Myc target genes, we queried the ChiP Atlas database with the above 1,657 transcripts whose expression differed between WT and Myc-KO tumors (q Ͻ 0.05). 71.2% of these were contained in this database, indicating that a substantial proportion of our differentially expressed transcript population originates from direct Myc target genes.

Myc and ChREBP cooperatively regulate ribosomal protein transcript (RPT) expression
As a group, genes encoding RPTs were up-regulated 2.2-fold in WT tumors relative to WT livers, but only 1.3-fold in Myc-KO and ChREBP-KO tumors and not at all in Double-KO tumors (Fig. 5 (A and B), p Ͻ 0.0001). Myc-KO tumors downregulated 73 of 80 RPTs relative to WT tumors, with the average expression of these being 52% of WT tumors. Similarly, 62 RPTs were down-regulated in ChREBP-KO tumors (average expression ϭ 62% of WT). Double-KO tumors not only downregulated a larger number of RPTs (79 of 80) but did so to a greater degree (average expression ϭ 36% of WT) (Fig. 5B). This indicated that Myc and ChREBP, either individually or together, regulate virtually all RPTs in tumors and, to a lesser extent, in livers.
RPT expression patterns can distinguish tumor types and normal tissues in ways that are independent of absolute RPT levels (90,91). Indeed, principal component analysis (PCA) showed that liver and tumor RPT expression patterns were distinct (Fig. 5C). Additionally, RPT patterns of knockout livers were distinct from those of WT livers, indicating that any knockout leads to a signature pattern of RPT expression (Fig.  5D). Whereas RPT expression patterns of all three KO liver groups displayed significant overlap, their cognate tumor groups clustered separately. Thus, RPT expression patterns from normal and transformed tissues are distinct and readily discernible from one another. Myc and ChREBP, therefore, in addition to dictating the absolute levels of tumor RPTs (and to a lesser extent liver RPTs), also regulate their expression patterns.
Finally, RP gene proximal promoters were analyzed for the presence of putative canonical E-boxes and ChoREs. Although 60 RP gene promoters contained at least one E-box within the regions analyzed, putative ChoRE sequences were found only in the promoters of Rps12 and Rpl13a (Fig. S10D). The lack of conserved E-boxes in 25% of RP genes plus the virtual absence of ChoREs argues that a significant amount of Myc regulation and virtually all of the ChREBP regulation of the RP gene family are likely to be indirect. However, it remains possible that less conserved E-boxes and ChoREs located in these promoters or outside the regions examined are involved in Myc-and/or ChREBP-dependent regulation.

Metabolic pathway dysregulation in livers and cancers mediated by the losses of Myc and ChREBP
The loss of ChREBP in livers resulted in an overall downregulation of transcripts encoding fatty acid elongation genes and HMG-CoA reductase, which encodes the rate-limiting enzyme in cholesterol synthesis (Fig. 6A). Myc-KO livers similarly down-regulated HMGCR transcripts but did not demonstrate strong down-regulation of those encoding enzymes involved in fatty acid elongation as did ChREBP-KO livers (Fig.  6B). Double-KO livers demonstrated down-regulation of transcripts involved in fatty acid elongation and cholesterol synthesis and a small number of genes involved in various other metabolic processes (Fig. 6C).
ChREBP-KO tumors strongly down-regulated the genes involved in glycolysis, the pentose phosphate shunt, the TCA cycle, Oxphos, and ␤and -oxidation. -Oxidation is an alternative nonenergy-generating cytosolic and microsomal FAO pathway that becomes activated in response to inefficient ␤-FAO and is mediated largely by Cyp450 members and alcohol and aldehyde dehydrogenases (Fig. 7A) (82,92). Myc-KO tumors also down-regulated transcripts for glycolysis and HMGCR and up-regulated those for glutamine metabolism and fatty acid elongation (Fig. 7B). Double-KO tumors also downregulated many glycolytic genes and mitochondrial ␤-oxidation and Oxphos genes and up-regulated numerous genes involved in fatty acid elongation and peroxisomal ␣and ␤-FAO (Fig. 7C).
Additional evidence supporting cooperation between Myc and ChREBP in the regulation of lipid metabolism was the finding that 26 of the 102 known murine Cyp450 genes in livers (25%) and 51 of 102 in tumors (50%) were deregulated in response to knockout of ChREBP and/or Myc (Tables S9 and  S10). In the latter case, 34 (56%) were members of the Cyp2, Cyp3, or Cyp4 families, which play particularly relevant roles in steroid and eicosanoid metabolism (93)(94)(95). In keeping with previous findings that Cyp450 genes are Myc targets both in untransformed hepatocytes and hepatoblastomas (13, 25), we found 35 Cyp450 genes to be exclusive targets of Myc, 15 to be exclusive targets of ChREBP, and six to be deregulated

Myc and ChREBP cooperation in hepatocyte proliferation
only when both Myc and ChREBP were eliminated. These results indicate that, as a gene family with important roles in fatty acid, sterol metabolism, and eicosanoid metabolism, Cyp450 genes are subject to extensive regulation by Myc and/or ChREBP.
Finally, to validate RNA-Seq results, we selected a small number of representative transcripts from the various families discussed above and performed real-time qRT-PCR analyses (Table S11). We found these results to be in excellent overall agreement with those obtained by RNA-Seq (Fig. S11).   The order of transcripts is based on the most to the least abundant in WT livers (WL). C, PCA of RPT relative expression in livers and tumors. Livers and tumors cluster distinctly from one another, indicating that each group possesses distinct patterns of RPT expression. D, PCA of RPT expression performed separately in livers and tumors. WT livers cluster distinctly from knockout livers, but none of the knockout liver groups can be distinguished from one another. All tumor groups cluster distinctly from one another, however, indicating that knockout of Myc, ChREBP, or both leads to distinct expression patterns of RPTs.

Discussion
Cross-talk among select members of the Extended Myc Network presumably underlies the responsiveness of its members to growth and differentiation signals and helps to coordinate the timing and robustness of these responses with nutrient availability, proliferative needs, and metabolic preparedness (14,32,34,36,46,47,63,96).
The hepatocyte-specific loss of Myc is associated with mild and age-dependent neutral lipid accumulation (25). Even more neutral lipids accumulate in the livers of ChREBP-KO and Double-KO mice (Fig. S1, B and C). Similar observations have been made in other cell types following Myc or N-Myc inhibition (97)(98)(99). This likely represents a generalized loss of coordination between exogenous lipid uptake, FAO, and de novo lipid biosynthesis with the rate of the former exceeding the rate of the latter two. Our finding that all KO livers were defective in FAO, coupled with previous observations that Myc inhibition enhances exogenous fatty acid uptake (13,(97)(98)(99), is consistent with this idea. Mitochondrial dysfunction in KO livers and tumors (Fig. 3, A, E, and F), together with the inability to properly up-regulate Oxphos-related genes (Fig. S5A), is consistent with other abnormalities in Myc-KO livers and HBs (13) and in cell lines with MondoA knockdown (63).
We previously found there to be no discernible differences in the rates at which WT and Myc-KO hepatocytes repopulate the fah Ϫ/Ϫ liver despite the latter's lipid processing defects (25,26,50). In contrast, ChREBP-KO and Double-KO hepatocytes, with more severe lipid abnormalities, were even more impaired (Fig. 1, D and E) (37). This suggests that ChREBP-regulated genes are just as important for sustained hepatocyte proliferation as Myc-regulated genes, if not more so. We cannot currently state with certainty whether this represents a direct effect on proliferation or an indirect one resulting, for example, from an inimical intracellular environment generated by accumulated lipids. In addition, the collaboration between the losses of Myc and ChREBP to produce a more severely impaired proliferative defect where none existed for the knockout of Myc alone suggests that this represents a form of synthetic lethality (100).
Despite differences in hepatocyte proliferative rates (Fig. 1E) (25), tumor growth rates were equally impaired (Fig. 2, A and B) (13). Tumors may rely more on the products of certain Myc and/or ChREBP target genes because of their more rapid growth. The pronounced AcCoA deficit in all KO tumors and their failure to increase RPT expression might be two such examples (Figs. 3D and 5 (A and B)). Additionally, Myc and ChREBP target genes are distinct in hepatocytes and tumors and could perhaps account for differential effects on tumor growth (Fig. 4) (13). Finally, ␤-catenin target gene products other than Myc could alter tumor cell sensitivity to Myc and/or ChREBP in ways that are distinct from those of hepatocytes. For example, Myc and ChREBP regulate smaller subsets of RPTs in livers than in tumors and do so less robustly (Fig. 5, A and B) (90). The defective RPT response of hepatocytes might not be fully appreciated until they are subject to the more demanding proliferative signals associated with transformation. Normal and transformed hepatocytes also have distinct metabolic properties, which may be direct or indirect consequences of Myc, ChREBP, and ␤-catenin expression (Fig. 3, A, E, and F).
WT and all KO tumor groups up-regulated L-Myc equally, which could potentially rescue the Myc-KO phenotypes if not the ChREBP and Double-KO phenotypes (25). L-Myc serves this function in myc Ϫ/Ϫ fibroblasts, although its transcriptional activation potential and transforming activity are markedly weaker than Myc's (28,29,101). Similarly, elevated MondoA in tumors might partially rescue ChREBP-KO tumor growth. MondoA enhances N-Myc transcriptional activity by altering chromatin conformation on some metabolic target gene promoters and by preventing Mlx's heterodimerization with Mxd and Mnt (60,90,91). Whether this occurs with L-Myc as well, however, is unknown.
MondoA and ChREBP target genes have been identified in a variety of cell lines, and ChREBP-bound genes have been cataloged in liver and white adipose tissue (37,40,43,63,67,69). However, it remains unclear how many of the nearly 6,000 ChREBP-binding sites actually alter gene expression (37). The gene expression profiles of our three KO liver and tumor groups suggest that at least seven distinct mechanisms account for these transcriptional responses (Fig. 4) and underscore the interplay between Myc and ChREBP.
In addition to the gene targets mentioned above, we unexpectedly found that ChREBP also controls RP genes, particularly in collaboration with Myc ( Fig. 5 and Figs. S9C and S10). Whereas numerous canonical E-boxes were identified in the proximal promoters of these genes, ChoRE elements were detected in only two (Fig. S10D). This suggests either that ChREBP regulation of most RP genes is indirect, that some ChoREs are more degenerate than those we defined as bona fide, that ChoREs reside outside the proximal promoter regions we analyzed, and/or that ChREBP binds to E-boxes as well as ChoREs. Regardless of the nature of the regulation, ChREBP's cooperation with Myc can be quite strong, as seen by the complete unresponsiveness of RP genes in tumors. This suggests that Myc and ChREBP are the major determinants of RPT expression in normal and neoplastic stimuli and cannot be replaced by other factors. It is difficult to escape the conclusion that a major proliferative defect in hepatocytes and livers lacking Myc and/or ChREBP is their inability to match translational rates with proliferative demands (102)(103)(104).
Given the variability in E-box binding affinities for Myc and a likely similar variability for ChREBP (105)(106)(107)(108), it is not surprising that the patterns of RPT expression in different tumor types    Fig. 6. B, transcriptional changes in metabolic pathways in Myc-KO tumors compared with WT tumors. C, transcriptional changes in metabolic pathways in Double-KO tumors compared with WT tumors. vary (Fig. 5, C and D). These likely reflect both the relative intrinsic affinities of binding sites within their gene promoters and the levels of Myc, ChREBP, and other members of the Extended Myc Network. Conceivably, the presumed resulting changes in ribosomal subunit composition can impact the efficiency with which different mRNAs are selected and translated, thereby achieving a distinct level of translational control independent of the transcriptional control imparted by Myc and ChREBP themselves (109).

Animals, hepatocyte transplants, and tumorigenesis studies
All animal work was performed in compliance with the Public Health Service Policy on Humane Care and Use of Laboratory Animal Research (DLAR) Guide for Care and Use of Laboratory Animals. All procedures were reviewed and approved by the University of Pittsburgh's institutional animal care and use committee. Mice were maintained in a pathogen-free facility under standard conditions with free access to food and water. UBI-GFP/B6-EGFP mice were purchased from Jackson Laboratories (Bar Harbor, ME). These mice, which express enhanced GFP in all tissues under the control of the ubiquitin C promoter, were used as a source of WT hepatocytes for all transplant studies. Mice bearing a conditional, hepatocyte-specific deletion of myc (Myc-KO mice) were generated by crossing c-Myc flox/flox (WT) mice (2,25) with Alb-Cre recombinasepositive mice (25) and were genotyped as described previously (25). chrebp Ϫ/Ϫ knockout (ChREBP-KO) mice (B6.129S6-Mlxipl tm1Kuy /J mice, strain 010537) were purchased from Jackson Laboratories and were genotyped as recommended by the vendor.
Immunocompromised FRG-NOD fah Ϫ/Ϫ mice (Yecuris, Inc., Tualatin, OR) were used as hepatocyte transplant recipients. They were maintained on 8 mg/liter NTBC (Yecuris) in their drinking water. Hepatocyte isolation was performed using a two-step collagenase perfusion method (24,25,51). For competitive transplant studies, WT and KO hepatocytes (Ͼ80% viability) were combined in approximately equal numbers, and 3.0 ϫ 10 5 were injected intrasplenically into 6 -8-week-old recipient FRG-NOD mice. Four days after transplantation, NTBC was discontinued and then reinstated when mice lost 20% of their body weight. NTBC cycling was continued until body weights stabilized. Mice were then maintained off NTBC for 4 -5 additional weeks to ensure maximal hepatic replacement and body weight normalization. Hepatocytes were then isolated as described above (24,25,51), washed twice in PBS, and snap-frozen for subsequent PCR-based analysis.
In some cases, mice were reconstituted noncompetitively using only Myc-KO, ChREBP-KO, or Double-KO donor hepatocytes. In these instances, the time to NTBC independence was noted, and entire livers were removed for histological assessment.
Determining the extent to which recipient fah Ϫ/Ϫ mice were reconstituted with the total donor population used a two-step TaqMan-based assay. PCRs were performed with a Kapa Probe Fast qPCR kit (Kapa Biosystems, Wilmington, MA) in a volume of 12 l with 10 ng of genomic DNA as described previously (51). Conditions for amplification were as follows: 95°C for 10 s, 40 cycles at 95°C for 15 s, 60°C for 60 s. Repopulation was calculated with primers that amplified the Pol2neo cassette inserted in exon 5 of fah Ϫ/Ϫ mice (forward, 5Ј-GGGAGGAT-TGGGAAGACAATAG-3Ј; reverse, 5Ј-ATTCTCCTTGCC-TCTGAACATAA-3Ј). The product was detected and quantified in real time with the probe, 5Ј/56-FAM/CTTCTG-AGG/ZEN/CGGAAAGAA/3IABkFQ/-3Ј. A standard curve with known amounts of fah Ϫ/Ϫ and fah ϩ/ϩ hepatocyte DNA was generated to allow the fractional contribution of each transplanted sample to be determined. To determine the contributions of Myc-KO and ChREBP-KO hepatocytes relative to WT hepatocytes or to one another, we used the approach depicted in Fig. 1A.

Induction of HBs
Tumorigenesis was induced in 6 -8-week-old mice using the hydrodynamic tail vein injection-mediated delivery of Sleeping Beauty plasmids (13,27,51). Mice received 10 g each of plasmids encoding a mutant, patient-derived form of human ␤-catenin (⌬90) and YAP (S27A) along with 2 g of a vector encoding Sleeping Beauty transposase (13,51). Plasmids were purified using Qiagen tip-500 columns (Qiagen, Inc., Valencia, CA) and were delivered in a total volume of 2 ml in 0.9% NaCl solution over ϳ5-9 s. Mice were evaluated three times weekly and euthanized when tumors reached ϳ2 cm in maximal diameter. Tumor weights and survival were recorded in all instances.

Metabolic studies
Livers or HBs were removed, immediately placed in a sterile tissue culture plate, and maintained on ice while pieces were apportioned for each of the assays described below or immediately snap-frozen in liquid nitrogen. Additional tissues were used for the isolation of mitochondria (25), which were snapfrozen in liquid nitrogen and stored at Ϫ80 ºC. Remaining tissues were apportioned in small aliquots, snap-frozen, and stored at Ϫ80º C for studies not requiring fresh tissue.
For FAO assays, ϳ50 mg of fresh tissue was minced in 350 l of SET buffer at 4°C (25,51,75) and disrupted by five passes in a Potter-Elvehjem homogenizer; 5 l of the homogenate was added to 195 l of FAO reaction buffer containing 100 mM sucrose, 10 mM Tris-HCl, pH 7.4, 0.2 mM EDTA, 80 mM KCl, 5 mM KH 2 PO 4 , 1 mM MgCl 2 , 0.1 mM malate, 0.05 mM CoA, 2 mM L-carnitine, 2 mM ATP, 1 mM DTT, 0.3% fatty acid-free BSA, pH 8.0, and 0.5 Ci/ml 3 H-labeled palmitate conjugated to BSA (PerkinElmer Life Sciences). Reactions were incubated at 37°C for 2 h and stopped with the addition of 40 l of 1 M KOH. They were further incubated at 60°C for 1 h to hydrolyze newly synthesized acyl-carnitine esters. 40 l of 4 M perchloric acid was added, and incubations were continued on ice for an additional 60 min. After organic extraction, water-soluble 3 H-labeled products were measured by scintillation counting.
PDH complex activity was determined as described previously (25,51,75). Briefly, 50 -100 mg of fresh tissue was minced in ice-cold glutamine-, glucose-, and pyruvate-free Dulbecco'smodified minimal essential medium and gently homogenized by several passages through an 18-gauge needle. 0.5-ml aliquots were placed into 0.5 ml of 2ϫ assay buffer containing 0.075 Ci

Myc and ChREBP cooperation in hepatocyte proliferation
of [1-14 C]pyruvate (PerkinElmer Life Sciences). Reactions were incubated at 37°C for 40 min with continuous shaking and terminated by injecting 0.5 ml of 4.0 M perchloric acid through the rubber stopper, which contained a central hanging basket (Kimble-Chase Life Science and Research Products, Rockwood, TN) into which had been placed a 25-mm glass microfiber filter (Whatman/GE Healthcare) saturated with 0.5 M KOH. 14 CO 2 was collected for 40 min at 37°C onto the filters and quantified by scintillation counting.
Respirometry was performed as described previously (25,51,75). Briefly, ϳ50 mg of fresh tissue was disrupted in ice-cold MiR05 buffer. ϳ300 g of the lysate was used to determine OCRs using an Oroboros Oxygraph 2k instrument (Oroboros Instruments, Inc., Innsbruck, Austria). Each reaction contained cytochrome c (10 M) (25), malate (2.0 mM), pyruvate (5 mM), ADP (5 mM), and glutamate (10 mM) to initiate electron flow via Complex I, followed by the addition of succinate to a final concentration of (10 mM) to measure the combined activities of ETC Complexes I ϩ II. Finally, rotenone was added (0.5 M final concentration) to inhibit Complex I and to allow for Complex II quantification. Activities were normalized to total protein content.
To quantify mtDNA content, two probe sets were used to amplify unique regions of mtDNA from livers and tumors using a TaqMan-based approach as described previously (25,51,75). The amplified segments included 101 bp of the cytochrome c oxidase I gene (probe set 1) and 90 bp of the D-loop region (probe set 2). Results were normalized to the product of a third TaqMan reaction that amplified a 73-bp segment of the nuclear apolipoprotein B gene. All reactions contained 10 ng of total DNA and were performed simultaneously on a CFX96 Touch TM real-time PCR detection system (Bio-Rad) using the following conditions: 95°C for 10 s, 40 cycles at 95°C for 15 s, and 60°C for 60 s.
ATP assays were performed on previously snap-frozen tissues using the ATP Lite detection system (PerkinElmer Life Sciences). AcCoA assays were also performed on frozen tissues using an assay kit (MAK039, Sigma-Aldrich). Both assays were performed according to the suppliers' directions. Triglyceride content was measured as described previously (110).
Purified mitochondria were used in nondenaturing blue native gel electrophoresis to separate each of the multisubunit complexes of the electron transport chain (ETC), followed by Coomassie Blue staining (25,51,75). In situ assays for ETC Complexes I, III, IV, and V were performed as described previously, whereas Complex II (succinate dehydrogenase) assays were performed on separately prepared lysates (13,25,111). Enzymatic activity for each sample was quantified by scanning the bands corresponding to enzymatic activity in gels using a Fluorchem M system (Protein Simple, San Jose, CA), and relative activity was compared with that Complex IV, which showed the least sample-to-sample or group-to-group variability.

Western blotting
This was performed as described (13,51,75). Frozen tissues were disrupted in lysis buffer containing protease and phosphatase inhibitors, sonicated, boiled, and stored in small aliquots at Ϫ80°C. The antibodies used, their dilutions, and the vendors from which they were purchased are provided in Table S1.

Transcriptional profiling of livers and HBs
All procedures were performed as described previously (13,75). Briefly, RNA purification from five representative tissues of each group under study was performed using Qiagen RNeasy columns (Qiagen, Inc., Valencia, CA) followed by DNase digestion. Final sample integrity was assessed on an Agilent 2100 Bioanalyzer (Agilent Technologies, Forster City, CA), and only samples with RNA Integrity Number (RIN) values Ͼ8 were processed further. RNA-Seq was performed at the Children's Hospital of Pittsburgh of UPMC Core Genomics Facility. Single-end single-indexed sequencing (ϳ75 bp) was performed with an Illumina NextSeq 500 sequencer (Illumina, Inc., San Diego, CA). Gene expression was calculated using Cufflinks and expressed as fragments per kilobase million (FPKM). Gene expression counts were normalized across samples and statistically compared between experimental groups using Cuffdiff, with a false discovery rate-adjusted p value (or q value) cutoff of 0.05. Raw and processed original data have been deposited in the National Center for Biotechnology Information (NCBI) Gene Expression Omnibus (61) and are accessible through GEO (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?accϭ GSE114634) at accession number GSE114634. Ingenuity Pathway Analysis (IPA, Qiagen) was used to categorize differentially expressed genes into pathways, with pathway significance was adjusted for false discovery using Bonferonni-Hochberg correction and a p value cutoff of 0.05. For each pathway, the following were recorded for display: overall significance (p value), ratios of dysregulated genes compared with all possible genes associated with that pathway (ratio), and predicted pathway activation or inhibition (Z-score, where applicable). Not all IPA pathways included the ability to estimate a Z-score. Significant pathways were then grouped into "superpathways" according to function using IPA's internal category scheme. These and their component transcripts can be accessed in an interactive format at https://prochownik.pitt.edu/chrebp/.
For select transcripts (Fig. S11), the same RNAs used for the above-described RNA-Seq experiments were used in qRT-PCRs to validate RNA-Seq results. Each qRT-PCR (20 l) was performed with a Power SYBR Green RNA-to-C T TM 1-Step kit (Applied Biosystems, Inc., Waltham, MA) according to the manufacturer's instructions. Real-time reactions were monitored on a StepOne Plus Real PCR system (Applied Biosystem). Each reaction was performed in duplicate with Ͻ10% variation between samples. A list of PCR primers is shown in Table S11.