key: cord-0738057-apwyl90s
authors: Wang, Huabo; Lu, Jie; Alencastro, Frances; Roberts, Alexander; Fiedor, Julia; Carroll, Patrick; Eisenman, Robert N.; Ranganathan, Sarangarajan; Torbenson, Michael; Duncan, Andrew W.; Prochownik, Edward V.
title: Coordinated Cross-Talk Between the Myc and Mlx Networks in Liver Regeneration and Neoplasia
date: 2022-03-05
journal: Cell Mol Gastroenterol Hepatol
DOI: 10.1016/j.jcmgh.2022.02.018
sha: 281f4d21775f86f7c0c45d981669e737d9736a5e
doc_id: 738057
cord_uid: apwyl90s

BACKGROUND & AIMS: The c-Myc (Myc) Basic helix-loop-helix leucine zipper (bHLH-ZIP) transcription factor is deregulated in most cancers. In association with Max, Myc controls target genes that supervise metabolism, ribosome biogenesis, translation, and proliferation. This Myc network crosstalks with the Mlx network, which consists of the Myc-like proteins MondoA and ChREBP, and Max-like Mlx. Together, this extended Myc network regulates both common and distinct gene targets. Here, we studied the consequence of Myc and/or Mlx ablation in the liver, particularly those pertaining to hepatocyte proliferation, metabolism, and spontaneous tumorigenesis. METHODS: We examined the ability of hepatocytes lacking Mlx (MlxKO) or Myc+Mlx (double KO [DKO]) to repopulate the liver over an extended period of time in a murine model of type I tyrosinemia. We also compared this and other relevant behaviors, phenotypes, and transcriptomes of the livers with those from previously characterized MycKO, ChrebpKO, and MycKO × ChrebpKO mice. RESULTS: Hepatocyte regenerative potential deteriorated as the Extended Myc Network was progressively dismantled. Genes and pathways dysregulated in MlxKO and DKO hepatocytes included those pertaining to translation, mitochondrial function, and hepatic steatosis resembling nonalcoholic fatty liver disease. The Myc and Mlx Networks were shown to crosstalk, with the latter playing a disproportionate role in target gene regulation. All cohorts also developed steatosis and molecular evidence of early steatohepatitis. Finally, MlxKO and DKO mice showed extensive hepatic adenomatosis. CONCLUSIONS: In addition to showing cooperation between the Myc and Mlx Networks, this study showed the latter to be more important in maintaining proliferative, metabolic, and translational homeostasis, while concurrently serving as a suppressor of benign tumorigenesis. GEO accession numbers: GSE181371, GSE130178, and GSE114634.

c -Myc (Myc) is a Basic helix-loop-helix leucine zipper (bHLH-Zip) transcription factor that regulates numerous target genes, which collectively support survival, proliferation, metabolism, ribosome biogenesis and translation. [1] [2] [3] [4] [5] [6] [7] [8] Positive regulation involves Myc's direct sequence-specific DNA binding in heterodimeric association with its obligate bHLH-Zip partner, Max. 6, 7 This occurs at canonical E-box elements that typically reside in the vicinity of proximal promoters. [9] [10] [11] [12] Bound Myc-Max heterodimers recruit an assortment of transcription co-factors and chromatin modifiers such as histone acetylases and methyltransferases, which collectively increase chromatin accessibility, relieve transcriptional pausing, and increase the rate and efficiency of messenger RNA (mRNA) elongation. [13] [14] [15] [16] [17] Down-regulation of these genes, often occurring during cellular quiescence or differentiation, involves a reduction in Myc levels and a shift in E-box occupancy to heterodimers now comprising Max and members of the transcriptionally repressive bHLH-ZIP Mxd family, which includes Mxd1-4 and the less-related Mnt and Mga factors. 3, [18] [19] [20] Together, their binding reverses the chromatin modifications mediated by Myc-Max binding and restores transcriptional repression. Negative regulation by Myc is more indirect and involves interaction with and inhibition of positively acting transcription factors such as Miz1 and Sp1. 21, 22 The loss of transcriptional balance maintained by these different competing interactions is a feature of transformed cells, which often overexpress and/ or otherwise deregulate Myc. 4, 10, 14, 23 The Myc Network crosstalks and shares considerable regulatory overlap with a structurally related but distinct group of bHLH-Zip transcription factors that comprise the so-called Mlx Network. [1] [2] [3] 8, 20, 24, 25 Classically believed to control target gene sets smaller and more functionally restricted than those overseen by Myc, the Myc-like equivalents of the Mlx Network include the transcription factors Carbohydrate response element (ChRE)-binding protein (ChREBP) and MondoA. Upon binding glucose and other nutrients, these cytoplasmic proteins translocate to the nucleus, heterodimerize with the Max-like protein Mlx, and bind to their target genes at carbohydrate response elements (ChoREs) comprising tandem E-boxes separated by 5 nucleotides. 26, 27 Myc Network and Mlx Network members (collectively termed the Extended Myc Network) can bind to one another's DNA target sequences and some genes are dually regulated by both sets of factors; however, the numbers of these genes and the degree to which their regulation is a result of binding to shared vs separate sites has not been delineated clearly. [28] [29] [30] [31] [32] Although the Mlx Network is less widely implicated in tumorigenesis than the Myc Network, recurrent MLX gene deletions nevertheless occur in as many as 10%-20% of several human cancers, with the precise fraction correlating with the size of the deletion (https://portal.gdc.cancer.gov/genes/ ENSG00000108788). [1] [2] [3] 13, 20, 33 We previously explored the roles for these 2 networks in normal hepatocyte proliferation using mice lacking the enzyme fumarylacetoacetate hydrolase (FAH). These animals serve as a model for type I hereditary tyrosinemia in which FAH's absence allows toxic tyrosine catabolites to accumulate, leading to hepatic necrosis and liver failure. [34] [35] [36] Treatment with the drug 2-[2-nitro-4trifluoromethylbenzoyl]-1,3-cyclohexanedione (NTBC) blocks the enzyme 4-hydroxyphenylpyruvic dioxygenase, which catalyzes the second step in tyrosine catabolism, thereby preventing the accumulation of these deleterious intermediates and circumventing the lethal consequences of FAH deficiency. Immunocompromised fumarylacetoacetate hydrolase nonobese diabetic (FRG-NOD) (Fah À/À ) mice thus can be used as a robust and sensitive animal model in which to evaluate the regenerative potential of any other hepatocyte population, so long as it is Fahþ/þ. Cells are delivered intrasplenically followed by the cyclic withdrawal and reinstatement of NTBC over several months. As recipient hepatocytes accumulate toxic tyrosine intermediates and die, they are replaced by the donor cells, which expand as much as 50-to 100-fold before eventually comprising up to 70% of the hepatic mass and allowing the recipients to achieve NTBC independence. 25, 37 The FAH model thus places greater proliferative demands on regenerating hepatocytes than does two-thirds partial hepatectomy (PH), which represents the gold standard for liver regeneration. 38 It also permits the simultaneous delivery of 2 or more competing populations of hepatocytes to the same recipient, thus allowing for a direct comparison of their relative proliferative rates within the identical environment.

Using this approach, we previously showed that wildtype (WT) and Myc-/-(MycKO) hepatocytes possess indistinguishable regenerative potential. 37 This is quite different from most other cases in which Myc's loss in either nontransformed or transformed cells or tissues profoundly suppresses proliferation. 25, 37, [39] [40] [41] [42] [43] In contrast, the proliferation of Chrebp-/-(ChrebpKO) hepatocytes is impaired significantly and MycKO Â ChrebpKO hepatocytes are even more defective. 25 These findings indicated that normal hepatocyte regeneration is more dependent on the Mlx Network than the Myc Network and that the 2 pathways crosstalk and rescue one another's defects to varying degrees. At the same time, they raise questions about the possible functional redundancy of MondoA in the context of ChREBP's loss.

We now have explored the relationship between the Myc and Mlx Networks further by generating 2 additional mouse strains. In the first (hereafter referred to as MlxKO), deletion of the Mlx gene functionally inactivates the entire Mlx Network, including any potential rescue by MondoA that might have existed in ChrebpKO mice. 25 The second mouse strain contains a double knockout (DKO) of both Myc and Mlx that further inactivates the Extended Myc Network. We show that hepatocytes from both strains, but particularly the latter, are profoundly compromised in repopulating the livers of Fah-/-recipients. They also show markedly attenuated expression of genes that are direct targets for both the Myc and Mlx Networks and that control mitochondrial structure and function, ribosomal biogenesis, and more general aspects of mRNA processing and translation. Older mice of both groups also develop steatosis akin to that previously described in MycKO, ChrebpKO, or MycKO Â ChrebpKO mice. Finally, and unexpectedly, more than one third of older MlxKO and DKO mice develop multifocal hepatic adenomas occasionally associated with small regions of hepatocellular carcinoma (HCC). These results further support the idea that the Myc and Mlx Networks crosstalk and cooperatively regulate a range of pathways related to energy metabolism, lipid balance, translation, and proliferation. Finally, they show a heretofore unsuspected role for the Mlx Network as a suppressor of benign hepatic adenomatosis. 44 

Donor mouse strains used for competitive hepatocyte repopulation studies carried homozygous floxed alleles of the Mlx and/or Myc genes ( Figure 1A and B and Table 1 ) and expressed an albumin promoter-driven tamoxifen-inducible estrogen receptor and Cre recombinase (CreER). 25, 45 Five daily intraperitoneal injections of tamoxifen were sufficient to allow inactivation of each allele by the time hepatocytes were transplanted 3-4 months later ( Figure 1C ).

Using FRG-NOD mice as recipients, we previously showed that WT donor hepatocytes can outcompete an equal number of ChrebpKO hepatocytes whereas WT and MycKO hepatocytes compete equally. 25, 35, 37 Suspecting that MlxKO hepatocytes would be even more defective, and to emphasize this, we delivered a total of 3 Â 10 5 donor hepatocytes intrasplenically into recipient mice at an approximately 1:6 WT:MlxKO ratio (Figure 2A and B) . After 24-28 weeks of NTBC cycling, a number of recipients had died and no survivors had achieved NTBC independence, possibly as a result of the deliberate under-representation of WT hepatocytes in the initial inoculum. Indeed, quantification of the total donor population in the surviving recipients indicated that it comprised only 2%-46% of all hepatocytes, which is both lower and more variable than typically achieved when mice receive larger numbers of replicationcompetent donor cells ( Figure 2C ). 25, 37 Despite this low-level reconstitution, the surviving donor hepatocytes were nearly all WT despite their initial minority status (P < .001) ( Figure 2D ).

Although Myc deletion alone does not confer a replicative disadvantage to hepatocytes, Chrebp deletion does and is exacerbated further by the concurrent inactivation of Myc. 25 This suggests that the Myc and Mlx Networks are redundant and compensate for one another under certain circumstances. Because MycKO Â ChrebpKO hepatocytes still express MondoA, 25 we asked whether its redundancy might mask more prominent phenotypes. We therefore compared the replicative potential of a mixed population of WT and DKO hepatocytes (1:10 ratio) in which the latter cells have functionally inactivated both ChREBP and Mon-doA as a consequence of Mlx deletion. This experiment achieved a somewhat greater rate of transplant success, with more animals surviving and with recipient livers eventually containing more than 50% donor hepatocytes ( Figure 2E ). As before, however, virtually all of these were of WT origin (P < .001) ( Figure 2F ).

To determine more directly which KO population was more proliferatively challenged, an additional competitive transplant experiment was performed using a 1:1 input ratio of MlxKO and DKO donor hepatocytes. Overall survival again was low, no animals achieved NTBC independence and less than 2% of hepatocytes isolated from recipients were of donor origin ( Figure 2G ). However, despite their own inherent replicative compromise, MlxKO hepatocytes showed an overwhelming survival advantage ( Figure 2H ) and comprised nearly 95% of the recovered donor population (P < .001). Together, these results argue that, in a highly demanding, long-term model of liver regeneration, 25, 37, 46, 47 Mlx loss and the ensuing functional inactivation of ChREBP and MondoA markedly compromise donor hepatocyte proliferation and/or survival, with the additional loss of Myc strongly reinforcing the defect.

Before comparing whole-transcriptome profiles of WT, MlxKO, and DKO livers, we confirmed that the latter 2 evidenced the predicted dysregulation of their direct target genes. For this, gene set enrichment analysis 48 Figure 1A Mycþ loxp site target gene sets also were enriched significantly in MlxKO liver RNA sequencing profiles, indicating as previously shown that some Myc-regulated genes also are responsive to Mlx Network inactivation ( Figure 3A ). 13, 24, 25 The broader and more pronounced enrichment of these transcripts in DKO livers indicated that the Myc Network further contributes to the Mlx Network-mediated regulation of these targets as expected. In the second case, MondoA/ChREBP/ Mlx target genes were enriched significantly in both MlxKO and DKO livers ( Figure 3B ). These results confirmed that

Myc and/or Mlx inactivation was associated with both unique and shared responses of each network's respective target genes.

To assess the effect of progressive dismantling of the extended Myc Network on target gene sets, volcano plots were used to compare individual gene expression profiles in the earlier-described livers and previously described MycKO, ChrebpKO, and MycKO Â ChrebpKO livers (GEO accession number: GSE114634). 25 In the latter 3 cases, fewer than 30 differences were identified relative to normal livers from age- Mlx Network and its E-box-related but more complex ChoRE binding site. 27, 46 Mlx interacts with the nutrient-regulated positive factors ChREBP and MondoA or the negative factors Mxd1, Mxd4, and Mnt. 2, 13, 24 The latter cross-talk with the Myc Network (dotted arrow). (B) Hepatocyte transplantation strategy. Isolated Fahþ/þ WT or KO hepatocytes were mixed at the desired ratio and injected intrasplenically into FRG-NOD Fah À/À mice maintained continuously on NTBC. NTBC cycling was continued until mice achieved NTBC independence or for 24-28 weeks, at which time total hepatocytes were isolated and the fractional representation of the total donor and recipient populations was determined ( Figure 1 ). The contribution of each donor set then was further determined and compared with that of the input inoculum. (C) After intrasplenic injection of 3 Â 10 5 donor hepatocytes comprising a 1:6 ratio of WT and MlxKO cells, NTBC cycling was continued for 24-28 weeks in the 3 animals that survived, with none achieving NTBC independence. Hepatocyte DNA was isolated from the transplanted animals and the percentage of recipient and total donor cells was determined. (D) DNA from panel C was used to determine the ratio of the WT and MlxKO donor populations. DNA from an aliquot of hepatocytes at the time of transplant was used to confirm the input donor cell ratio. (E) Transplants performed in FRG-NOD Fah À/À mice using inocula containing an approximate 1:10 ratio of WT:DKO cells. Hepatocytes isolated after 24-28 weeks of NTBC cycling showed that, on average, 52.5% of hepatocytes were composed of donor cells. (F) Hepatocyte DNA from panel E was used to determine the fraction of WT and MlxKO donor hepatocytes. Lane 1 shows the approximate 1:10 ratio of the initial input inoculum. (G) Transplants performed in FRG-NOD Fah À/À mice using inocula containing an approximate 1:1 ratio of MlxKO and DKO cells. Total hepatocytes isolated after 24-28 weeks of NTBC cycling were evaluated for the fractional representation of total donor and recipient populations showing that, on average, less than 2% of the total hepatocyte mass was of donor origin. (H) Fractional make-up of the donor population from panel G. Lane 1 shows the approximate 1:1 ratio of the input population. matched animals (differential expression, !1.5-fold and q < 0.05), whereas MlxKO and DKO livers showed up to 60-fold more differences ( Figure 3C ). This more pronounced gene dysregulation again suggested that the combined loss of ChREBP and MondoA eliminated all redundant functions from the Mlx Network, thereby allowing a much larger complement of gene expression differences to be shown. The relatively few differences between MlxKO and DKO expression profiles ( Figure 3D and insert) was consistent with the notion that, at least in the normal nonproliferating liver, the Mlx Network contributes more to regulating both the direct and indirect targets of both networks. The differences between the DKO vs MlxKO groups in Figure 3D thus were comparable with those between WT and MycKO groups shown in Figure 3C .

IPA profiling of the differentially expressed transcripts in DKO livers (and by extension MlxKO livers) showed that 6 of the top 7 most affected pathways were those with roles in mRNA translation and its control, energy metabolism, and mitochondrial structure and function ( Figure 3E ). The seeming exception (coronavirus pathogenesis pathway) contained numerous ribosomal protein transcripts whose dysregulation accounted for this pathway's inclusion. Collectively, these findings agreed with previous reports in livers, liver cancers, and other cell types showing roles for the extended Myc pathway in the earlier-described processes. 1, 2, 13, 17, 18, 24, 25, 37, 47, [51] [52] [53] [54] Gene expression profiles compiled from the pathways shown in Figure 3E showed the down-regulation of numerous transcripts encoding proteins involved in translation and mitochondrial structure and function in MycKO and MlxKO livers and an even greater degree of down-regulation in DKO livers ( Figure 3F and G). 25 To explore the potential co-regulation of direct target genes by the Myc and Mlx Networks, we obtained data from the most current version of the Encyclopedia of DNA Elements database (https://www.encodeproject.org) 55 and focused on the HepG2 HCC cell line, which was deemed the most relevant to the current work. CRISPR editing had inserted in-frame 3xFLAG epitope tags into the 3' end of the endogenous MYC or MLX coding regions, thus allowing all Chromatin immunoprecipitation (ChIP) sequencing studies to be performed under identical conditions with the same anti-FLAG antibody. The results were trimmed using a default setting that enumerated only those binding sites within ±2.5 kb of the transcriptional start site of each gene, thereby maximizing the likelihood of functional relevance. In this way, we identified 4152 genes that bound only Myc, 748 that bound only Mlx, and 2433 that bound both factors at 6047 sites, 5267 of which overlapped, either entirely or partially ( Figure 3H ). Thirty-seven percent of Myc target genes (2433 of 6558) also bound Mlx, whereas 76% of Mlx target genes (2433 of 3181) also bound Myc, thus indicating that Mlx target genes are twice as likely to also bind Myc. Fifty percent of the Myc and Mlx binding site peaks mapped to within 65 bp of one another and 75% mapped to within 170 bp, thus indicating that the 2 factors either bound to the same E-box or ChoRE, or to more than 1 element in such close proximity that their individual peaks could not be resolved simply by examining the ChIP sequencing footprints ( Figure 3I ).

To confirm the earlier-described results and obtain greater resolution and characterization of Myc and Mlx binding, we analyzed the sequences flanking the sites of maximum factor binding (ie, the ChiPseq binding peaks). The earlier-described 6047 binding sites were merged to 2863 distinct sites for motif analysis. A total of 1220 of these (42.6%) contained consensus E-boxes for Myc-Max and 714 contained consensus ChoREs 56 ( Figure 3J ). A total of 45.2% of the sites contained neither E-boxes nor ChoREs, despite the presence of prominent Myc and/or Mlx footprints, thereby indicating either that the motifs did not conform to the conservative consensus sequences used in our search or that Myc and Mlx binding was indirect as a result of association with other DNA binding factors.

Remarkably, the E-boxes and ChoREs in Figure 3J showed a nonrandom distribution and tended to reside within close proximity of factor-binding peaks. The consensus binding sites located closest to or at the peak centers tended to be those whose adjacent sequences contained either the fewest numbers of additional motifs or tightly clustered ones. This suggested that many ChIP sequencing peaks represent the integrated signal of multiple variably overlapping and unresolvable individual binding sites and thus do not necessarily directly overlie a particular site. Collectively, these findings confirm the presence of multiple E-boxes and/or ChoREs within the majority of common target genes as well as direct evidence of Myc's binding to ChoREs and Mlx's binding to E-boxes in select cases. They further suggest a means by which the apposition of multiple binding sites within some genes could allow for the simultaneous binding of different combinations of factors as well as their direct interaction and crosstalk.

Cataloging the functions of the 2433 common genes shown in Figure 3H using the IPA and MitoProteome databases and a bespoke collection of previously published genes 25, 51, 57, 58 showed that many could be categorized as supporting mitochondrial and ribosomal structure and function ( Figure 3K ). 25, 57 Thus, the common Myc-and Mlxbound target genes in human HepG2 cells faithfully reflect both the current transcript changes and those previously documented in MycKO and/or ChrebKO murine livers and hepatoblastomas. 25, 57 Interestingly, the 4152 genes bound only by Myc and the 748 genes bound only by Mlx were in somewhat different IPA categories than the common genes ( Figure 3K ). For example, Myc-specific genes also were involved in more restricted and/or unique functions such as transforming growth factor-b signaling, cell cycle, and oxidative phosphorylation. Similarly, Mlx-specific genes also tended to belong to distinct functional categories such as those related to retinoic acid signaling, xenobiotic metabolism, fatty acid b-oxidation, and the tricarboxylic acid (TCA) cycle.

Given the extended Myc Network's dynamic nature (Figure 2A) , the potential for different members to bind multiple closely neighboring sites with different affinities ( Figure 3J ), and their ability to either augment or antagonize one another's transcriptional impact, we hypothesized that Myc and Mlx binding alone ( Figure 3I and J) would not necessarily predict target gene expression levels. We further hypothesized that the transcriptional impact on any individual target gene ultimately would reflect the entire extended network's integrated action. 59 We thus examined the expression of the 2433 common Myc and Mlx target genes ( Figure 3H ) in 371 human HCCs using data from The Cancer Genome Atlas. Target gene expression could be categorized into 4 groups (designated A-D) that correlated with 4 patterns of extended Myc Network member expression (groups 1-4) ( Figure 3L ). Two of the tumor groups also showed significant differences in survival ( Figure 3M ). Together with the results of Figure 3J , these findings support the idea that the binding of Myc, Mlx, or any other extended Myc Network factor to a target gene likely reflects only 1 aspect of the complex and integrative interplay among other network members that collectively dictates the gene's expression level and downstream biological consequences. 20 Myc, particularly when it is overexpressed by tumor cells, promotes the Warburg effect by up-regulating genes encoding glucose transporters and glycolytic enzymes. 12, 25, 40, 53, [60] [61] [62] [63] [64] [65] However, none of these showed altered expression in MycKO livers (Table 3 ). In contrast, the progressive inactivation of the extended Myc Network was associated with the down-regulation of 3 transcripts encoding rate-limiting transporters or enzymes, including the major hepatocyte glucose transporter Glut2/Slc2a2 and the glycolytic enzymes liver-type phosphofructokinase (Pfkl) and liver-type pyruvate kinase (Pklr). Glut2/Slc2a2 also is required for the proper regulation of glucosesensitive genes and for glucose-stimulated insulin secretion. 64, 66 These findings were consistent with the previous IPA showing that genes comprising a glycolysis-related set were co-bound by Myc and Mlx ( Figure 3K ).

Consistent with findings that the Myc and Mlx Networks both impact pathways involved in carbohydrate and lipid metabolism ( Figure 3K ), 1, 2, 7, 24, 25, 37, 67, 68 young MycKO, ChrebpKO, and MycKO Â ChrebpKO mice develop steatosis. 25, 37, 69 However, these studies did not determine if this was progressive or if the dual compromise of the Myc and Mlx Networks increased its severity. We therefore examined the livers of older (14-16 mo) MycKO, ChrebpKO, MycKO Â ChrebpKO, MlxKO, and DKO mice to evaluate the extent of lipid imbalance. Relative to WT livers, all KO livers showed more intense Oil Red O staining but did not differ significantly from one another ( Figure 4A -F). They also contained more total triglycerides than livers from younger MycKO, ChrebpKO, and MycKO Â ChrebpKO mice 25,37 ( Figure 4G ). These findings suggest that steatosis appears earlier in MlxKO and DKO mice, with MycKO mice eventually achieving a similar degree of severity.

In further support of the earlier-described conclusions, we also found evidence for enrichment of a 163-member gene set 

associated with human nonalcoholic fatty liver disease (NAFLD) in all cohorts except ChrebpKO (https://www. wikipathways.org/instance/WP4396_r98945) ( Figure 4H ). Thus, despite the fact that the causes of steatosis in the earlier-described mice and human beings differ considerably, in most cases there is considerable similarity in the diseaserelated gene expression profiles. 70 Finally, we performed IPA to identify additional diseaserelated pathways that sometimes are dysregulated in NAFLD 70 and in all cases found several relating to lipid synthesis/metabolism and peroxisome proliferator-activated receptor(PPAR) activation (Table 4 ). While noting little histologic evidence for inflammatory cell infiltrates or fibrosis in our KO livers, several of these pathways were in fact associated with nonalcoholic steatohepatitis (NASH) ( Table 4 ). Although these transcriptome-based findings may represent early evidence of actual NASH, the altered expression of inflammatory markers also could be indicative of the immune function changes that accompany Myc dysregulation in nonhepatic tissues. 37, 71 We believe the most conservative interpretation of the earlier-described results is that the steatosis accompanying the loss of most extended Myc Network members is progressive and eventually leads to a mild NASH-like picture as indicated by significantly altered molecular markers of this state, but little documentable histopathologic change.

Unexpectedly, 36% of MxlKO and DKO animals (15 of 42) of both sexes developed multiple small-to mediumsized hepatic neoplasms, which were never observed in WT, MycKO, ChrebpKO, or MycKO Â ChrebpKO mice ( Figure 5A and B) . 25, 37 These were mostly welldifferentiated and/or myxoid-type tumors with numerous balloon cells, nuclear enlargement, and microvesicular steatosis. A minority also showed small foci of welldifferentiated HCC, which sometimes is associated with hepatic adenomas in human beings ( Figure 5C ). 44, 72 Regardless of histology, adenomas showed significantly more staining for Ki-67 than did the adjacent non-neoplastic liver parenchyma ( Figure 5D ) (P ¼ 3.4 Â 10 -5 ). Adenomas did not express Mlx protein, indicating that they did not originate from residual hepatocytes that had escaped Mlx locus excision and maintained their growth advantage ( Figure 5E ). Normal livers and DKO adenomas also did not express detectable Myc protein, and neither did most MlxKO adenomas. This was consistent with their slow growth rates and a likely consequence of their extended Myc Network defects ( Figure 3C-G) . 25, 37 An exception was seen in a single large adenoma containing elements of HCC from an MlxKO mouse with marked hepatomegaly (liver weight, 6.7 g, or approximately 3 times normal) ( Figure 5E ).

We performed RNA sequencing on 5 adenomas from MlxKO mice and compared their transcriptome profiles with those of WT livers and 45 primary murine hepatoblastomas (HBs) generated by overexpressing the Hippo pathway terminal effector yes-associated protein (YAP S127A ) and 1 of 9 different patient-derived oncogenic mutants of b-catenin. 73 The transcriptional profiles of the adenomas were distinct from those of both livers and all HBs ( Figure 5F and G). Adenomas did not overexpress wild-type b-catenin or YAP as is common for HBs, 74 but did dysregulate 15 of 22 transcripts that are expressed aberrantly in all murine HBs regardless of etiology and that correlate with survival in human HB and other cancers ( Figure 5H ). 47, 75 Our results indicate that dismantling the Mlx Network, either alone or concurrently with Myc, leads to the eventual emergence of multiple adenoma-like hepatic neoplasms (adenomatosis), which, similar to their human counterparts, may evolve further and acquire HCC-like features. 72

Most previous investigations into Myc's role in hepatic regeneration have relied on the PH model and yielded conflicting results that likely reflected differences in how and when regeneration was assessed and quantified. 37 The short time frame over which this process occurs and the dependency on separate groups of mice may have further contributed to disparate outcomes. Because post-PH hepatocytes require fewer than 2 divisions to replace the missing mass, the model also poses a comparatively modest regenerative challenge. Indeed, even this low number overestimates the actual contribution made by dividing hepatocytes given that approximately half the response to PH involves hypertrophy of the liver remnant plus replicative contributions by nonhepatocyte populations such as endothelial, Kupfer, and stellate cells. 38, 76, 77 In contrast, the FAH model is associated with a more sustained and robust 50-to 100-fold expansion of pure populations of transplanted hepatocytes. It also provides a well-defined point at which a stable level of regeneration can be assessed and a means by which competing donor populations within the same liver can be simultaneously distinguished and quantified after their delivery at any desired and preselected ratio. 25, 37, 52 Using this approach, we previously showed that the long-term proliferation of otherwise normal hepatocytes requires ChREBP but not Myc, although the loss of both factors was additive. 25, 37 Even more pronounced interdependencies were seen during HB tumorigenesis, with HB growth impaired markedly in both MycKO and ChrebpKO livers, and even more so in MycKO Â ChrebpKO livers. 25, 46, 47 These findings implied a means of communication between the Myc and Mlx Networks, with each one being able to rescue, at least partially, defects in the other. 1 They also showed that the requirement for Myc becomes progressively more critical as proliferative demand increases, thus emphasizing its strong contextual dependency ( Figure 6 ).

Despite our previous transplant studies having been performed with different input ratios of WT and KO hepatocytes, 25, 37 their outcomes were consistent with those reported here and allowed us to extend our conclusions regarding the relative importance of the Myc and Mlx Figure 3C ), the criteria for differential expression were relaxed to include those genes with more than 1.2-fold differences and P values < .05.

Networks in liver regeneration. For example, our studies comparing WT and ChrebpKO hepatocytes used an input inoculum comprising 62% of the latter population that was reduced by more than half after competitive repopulation. 25 Our current results in which MlxKO hepatocytes comprised approximately 84% of donor cells but only approximately 4% of the final population ( Figure 2D ) provided strong evidence that the concurrent functional inactivation of both ChREBP and MondoA confers an even more profound proliferative disadvantage. This could be a direct effect resulting from the concurrent loss of ChREBP and MondoA binding to their respective target genes, either individually or collaboratively with Myc, thereby eliminating any possibility of rescue of 1 factor by another (Figure 2A) . 24, [26] [27] [28] [29] A nonmutually exclusive indirect effect that allowed Mxd1, Mxd4, and Mnt to suppress Myc target genes more effectively by increasing their association with Max also remains possible (Figure 2A) . The relative importance of these 2 models could vary among different target genes at different times during repopulation or in different liver compartments. We also previously showed that the combined loss of Myc and ChREBP suppressed regeneration more than the KO of either individual gene, thereby corroborating previous evidence for internetwork crosstalk. 2, 3, 13, 24, 25 In 1 such study, performed with nearly equal contributions of WT and MycKO Â ChrebpKO donor hepatocytes, the latter was reduced to 7.5% after repopulation. 25 Although impaired markedly, the residual proliferative activity of these cells could have reflected the redundant function of MondoA (Figure 2A ), which is supported by 2 separate aspects of the current work. The first was the approximately 45-fold repopulation advantage of WT hepatocytes over DKO hepatocytes, whereas the second was the approximately 10fold repopulation advantage of MlxKO hepatocytes over DKO hepatocytes ( Figure 2F and H) . Collectively, our current results indicate that both the Myc and Mlx Networks play distinct as well as redundant roles in normal hepatocyte replication. However, much of the proliferative drive needed to sustain hepatocyte expansion in FAH mice is subsumed by the Mlx Network regardless of the Myc Network's status. 25, 37 This is supported by the progressive deterioration of repopulation potential as the Extended Network is gradually dismantled (Figure 2C-H) . 25 In proliferatively quiescent cells or organs such as the liver, Myc usually is expressed at low levels and regulates relatively few genes in contrast to Mlx (Figures 3C and  5E) . 25, 37, 78 Myc's contribution to genome-wide transcription therefore may be better appreciated in tumors where its overexpression can activate genes that are otherwise nonphysiologic targets owing to their low-affinity binding sites. 10, 11, 25, 68 Another plausible explanation for the seemingly modest transcriptional consequences of Myc loss in some normal tissues is that at least some Myc target gene expression is maintained by the Mlx Network with redundant contributions being made by MondoA and/or ChREBP. 25, 37, 78 This is best appreciated in livers and tumors when the Myc and Mlx Networks are either individually or concurrently inactivated ( Figure 3C ). 25 In tumors, the Myc Network positively regulates most of the genes encoding glycolytic enzymes and strongly contributes to the Warburg effect, 5, 12, 40, 51, 53, 61, 65, 80 as it does in rapidly growing fibroblasts in vitro. 40, 80 In contrast, our transcriptomic studies have not shown such widespread roles for the Myc and Mlx Networks in maintaining glycolysis in vivo (Table 3 and Figure 3K ), which may reflect Myc's low-level expression, the relative proliferative quiescence of the normal liver, and its greater reliance on fatty acid oxidation as an energy source. 25, 46, 51, 73 Nonetheless, among the most down-regulated genes in MlxKO and/or DKO livers were Glut2/Slc2a2, Pfkl, and Pklr, whose encoded proteins are rate-limiting for glucose uptake and glycolysis. In rat insulinoma cells, the Pklr proximal promoter binds both ChREBP and Myc, with the former interacting with a ChoRE element and the latter binding elsewhere. 28, 29 These results suggest that, in normal liver, glucose uptake and oxidation are more reliant on the Mlx Network ( Figure 6 ) whereas, in response to Myc-driven transformation or normal proliferation, more extensive transcriptional regulation of glucose uptake and its oxidation is achievable. 4, 14, 23, 25, 51, 81 This could have the additional benefit of maximizing glycolytic efficiency and sustaining cell division when microenvironmental glucose and oxygen supplies were limiting and nutrient-dependent functions of MondoA and ChREBP were attenuated. 82 Coordinated changes in direct and/or functionally related Myc and Mlx Network target genes were identified in MlxKO and DKO livers but were more pronounced in the latter ( Figure 3A-D) . 25 Figure 6 . A model for gene regulation and control of normal and neoplastic proliferation by the extended Myc Network. During quiescence, Myc levels are low and its target genes, which are relatively few in number (Figure 3 and Hofmann et al 78 ) , tend to be those with high-affinity E-box binding sites. Slow and controlled replication, such as that occurring during the replacement of Fah-/-hepatocytes, is largely regulated via the Mlx Network. 25, 37 As replication increases or as cells acquire transformed features, increased Myc expression 25,51 activates genes with low-affinity binding sites, including those encoding glycolytic enzymes that collectively are responsible for the Warburg effect as well as other metabolic pathways and functions that support increased energy demands and rapid growth. 2, 3, 24, 40, 53, 65, 69, 79 Despite the rapid growth that occurs in response to the overexpression of Myc and mutant forms of b-catenin and YAP S127A . 25, 51, 73 The Mlx Network, which supports this rapid growth, is proposed to contribute to tumor suppression as well ( Figure 5 ).

to support protein translation and its control as well as mitochondrial structure and function ( Figure 3E-G) . 18, 25, 37, 40, 51, 69, 73 As was seen for individual glycolysis-related transcripts (Table 3) , the collective expression of these sets became increasingly compromised as the extended Myc Network was progressively inactivated. The dramatic up-regulation of these pathways that accompanies tumorigenesis in WT livers also had been shown to be attenuated in response to Myc and/or ChREBP inactivation and to correlate with impaired rates of tumor growth. 25 We provided a mechanistic underpinning for the coordinated response of the relevant gene sets associated with these pathways by showing that, in HepG2 cells, 37% of Myc's direct target genes also bind Mlx, while 76% of direct Mlx target genes also bind Myc ( Figure 3H ). Many of the previously mapped binding sites for these 2 factors overlapped and/or contained multiple E-boxes and/or ChoREs ( Figure 3I and J). Although this sometimes made it difficult to attribute Myc or Mlx binding precisely to a specific motif within a factor's ChIP sequencing footprint, the collective binding landscape suggested a model for target gene regulation that accommodates this and all other observations. The model accounts for the fact that many sites containing only E-boxes coincided with Mlx binding peaks whereas many ChoRE-only sites coincided with Myc binding peaks ( Figure 3J ). This indicated that crosstalk between the Myc and Mlx Networks occurs by virtue of shared common binding sites as previously suggested. The nonrandom distribution of E-boxes and ChoREs around Myc and Mlx binding peaks ( Figure 3J ) also suggests that more than one such site could be occupied at any given time, that binding might be cooperative, and that the composition of the bound factors, their interactions, with each other and differential protein-DNA affinities are dynamic and serve to fine-tune the target gene's transcriptional output. Although we examined only Myc and Mlx binding, these motifs also could bind other extended Myc Network members such as those between Max and Mxd proteins, which would not have been detected with our ChIP sequencing analysis. Whether closely spaced Mlx sites contained ChREBP or MondoA heterodimers also potentially could determine if, when, and the degree to which a gene was responsive to metabolic substrate-mediated regulation. Finally, the 4 groups into which the expression patterns of the 2433 common Myc þ Mlx direct target genes in HCCs could be compiled correlated with the patterns of extended Myc Network transcript expression and, in 2 cases, with significant survival differences ( Figure 3L and M) . In future studies, it will be important to determine the degree to which different neighboring heterodimeric combinations of extended Myc Network members either cooperate with or antagonize one another under different conditions in different cell types.

Myc and/or ChREBP inhibition are widely associated with lipid accumulation, which stems from an over-reliance on fatty acid oxidation and a resulting increase in lipid uptake that exceeds the amount necessary to satisfy energy demands ( Figure 4) . 25, 37, 67, 83, 84 Young mice with hepatocyte-specific loss of Chrebp or combined Myc þ Chrebp loss also accumulate more neutral lipids than do those with isolated Myc loss. 25 Although we did not serially follow these animals, our findings suggest that, early in life, the partial or complete inactivation of the Mlx Network promotes a more rapid genesis of steatosis than inactivation of Myc alone. Over time, however, lipid accumulation equalizes, with little differences among the various KO groups being discernible in older individuals ( Figure 4G ). KO livers dysregulated many of the same gene sets and/or IPA pathways that have been described in NAFLD and its progression in human beings ( Figure 4H and Table 4 ), thereby further supporting the mechanistic relatedness of the various factors responsible for this state. Because KO livers also showed molecular evidence of incipient NASH, it will be important in future work to determine whether these features become more pronounced with age and whether histologic findings of inflammation and fibrosis eventually emerge. 85, 86 An unanticipated finding was the development of hepatic adenomatosis in more than one third of MlxKO and DKO mice ( Figure 5A and B). This incidence likely represents an underestimate because animals older than 14-16 months were not investigated and microscopic adenomas may have been overlooked in some instances. That similar neoplasms did not appear in WT, MycKO, ChrebpKO, or MycKO Â ChrebpKO mice makes it likely that complete Mlx Network inactivation is a prerequisite for their development. Although these neoplasms bear the hallmarks of actual human adenomas, 44, 72 several features suggest more aggressive and malignant predilections, despite the lack of Myc expression in most. These include their multifocality, their occasional HCC-like features, and their robust Ki-67 expression ( Figure 5C and D) . In contrast, human adenomas, although well known for their occasional conversion to HCC, typically are few in number and tend to show only modestly higher Ki-67 expression. 72 Molecular features suggestive of more aggressive behavior in our adenomas include the dysregulation of 15 of 22 transcripts that we recently identified as predicting inferior outcomes in human HBs and more than a dozen other human cancer types ( Figure 5H ). 47 Recurrent MLX gene deletions are associated with at least 8 human cancer types and provide further reason to implicate the Mlx Network in the pathogenesis of hepatic adenomatosis (https://portal.gdc.cancer.gov/genes/ ENSG00000108788). 20 Genetic suppressors of hepatic adenomas and other benign tumors such as meningiomas, neurofibromas, and uterine fibroids are well documented but are distinct from more notorious counterparts such as TP53, RB, PTEN, APC, and BRCA1/2 that are associated with malignant tumors. [87] [88] [89] [90] However, the role of Mlx and its members also may be more indirect and nuanced. For example, NAFLD ( Figure 4) is a known predisposing factor for the development of both adenomas and HCC and we are currently unable to determine how it might affect tumorigenesis in MlxKO mice. 72, 91, 92 On the other hand, the failure of MycKO, ChrebpKO, or MycKO Â ChrebpKO mice to develop adenomas or HCCs, despite their equally pronounced steatosis as well as the fact that high-fat diets can actually suppress hepatic tumor growth, 93 argues for a more direct role for the Mlx Network in adenoma suppression. It will be of interest to determine whether MlxKO mice are more susceptible to transformation by other oncogenic stimuli even though the emergence of the ensuing tumors may be delayed and their subsequent growth slowed.

In summary, we have shown that the Mlx Network engages in considerable biological and molecular crosstalk with the Myc Network and plays a more substantive role in long-term liver regeneration. 25, 29, 35, 37, 46 Both networks, but the former in particular, alter the expression of numerous genes responsible for broad aspects of translation and energy generation by both aerobic and anaerobic pathways. 25, 37, 57 The majority of Myc and Mlx targets are coregulated or at least bound by both factors, which appear to share many of the same binding sites, often lying in close proximity to one another. The actual expression of these target genes further correlates with the expression patterns of all 13 members of the extended Myc Network, thereby suggesting complex interactions and interdependent crosstalk at their sites of binding. Mechanistically, the defects that ensue in KO cells as a result of compromising these genes reflect an inability to maintain energy production and translation at levels commensurate with their proliferative demands. This is particularly acute in the neoplastic setting where tumor growth, but not induction, may be severely compromised. 25, 37, 57 The presumptive energy dysequilibrium that arises as a consequence of perturbing either or both of the networks likely is addressed by the increased uptake and storage of fatty acids, leading to eventual steatosis. 25, 73 The hepatic adenomatosis and occasional HCC seen in response to Mlx Network compromise suggests that the tumor suppressor-like activity of the Mlx Network counters the more pro-oncogenic tendencies of Myc overexpression. Our findings emphasize the elaborate orchestration of the Extended Myc Network in balancing energy demands and metabolism with normal and neoplastic proliferation.

All breeding, care, husbandry, and procedures were approved by The University of Pittsburgh Department of Laboratory and Animal Resources and the Institutional Animal Care and Use Committee, with standard animal chow and water provided ad libitum. C57BL6 mice expressing green fluorescent protein (GFP) (C57BL/6-Tg[UBC-GFP] 30Scha, MGI:3057178) have been described previously and were used as a source of WT control hepatocytes because of the ease with which the GFP gene could be identified. 25 C57BL6

c-myc LoxP/LoxP mice (B6.129S6-Myc tm2Fwa , MGI:2178233) have been described previously ( Figure 1A) , 25, 37, 94, 95 and were obtained as a gift from I. Moreno de Alboran. The generation of mice bearing a 1717bp deletion spanning exons 3-6 of the Mlx locus (MlxKO mice) ( Figure 1B ) also has been described recently. 8 Transgenic mice expressing a fusion protein comprising the hormone-binding domain of CreER and under the control of the albumin promoter that allows CreER to be activated in hepatocytes after tamoxifen exposure (B6.129S2-Alb tm1(cre/ ERT2)Mtz , MGI:3052812) were a kind gift from Dr Frank Gonzalez (Laboratory of Metabolism, Center for Cancer Research, National Cancer Institute). The latter mice were bred to homozygosity with Mlx LoxP/LoxP mice or Myc LoxP/LoxP Â Mlx LoxP/LoxP mice. 25, 37 At weaning, mice were subjected to 5 daily intraperitoneal injections of tamoxifen (75 mg/kg each) in corn oil (Sigma-Aldrich, St. Louis, MO). Several weeks later, hepatocytes were harvested as previously described. 25, 37 An aliquot of these was used for DNA isolation and to quantify the extent of Myc and/or Mlx knockout (Figure 1 ). The remainder of the MycKO, ChrebpKO, or MycKO Â ChrebpKO hepatocytes then were combined in the indicated proportions with WT hepatocytes and a total of 3 Â 10 5 cells were injected intrasplenically into Fah-/-FRG-NOD mice (Yecuris, Inc., Tualatin, OR) 25,37 (ChreBP mice: B6.129S6-Mlxipl tm1Ku , MGI:3043871; Fah mice: NOD.Cg-Rag1 tm1Mom Fah em1Mvw Il2rg tm1Wjl MGI:5485380). All animals were maintained on 8 mg/L NTBC (Ark Pharm, Libertyville, IL) in their drinking water. After 4 days, NTBC was discontinued until mice lost approximately 20% of their body weight. NTBC then was reinstated until mice regained their age-appropriate weight. NTBC cycling was continued either until mice had become NTBC-independent (at least 20 weeks after transplantation) or until week 28 in those cases in which NTBC independence was not achieved. Hepatocyte DNAs then were isolated from recipients and the TaqMan-based approaches shown in Table 1 and Figure 1 were used again to determine the donor:recipient ratio and the relative contribution of each donor population. 25, 37 Polymerase chain reactions were performed in a volume of 12 mL with 50 ng of genomic DNA. Conditions for amplification were 95 C for 5 minutes, 40 cycles at 95 C for 15 seconds, and 60 C for 60 seconds.

For gene expression profiling, the earlier-described Myc LoxP/LoxP , Mlx LoxP/LoxP , and Myc LoxP/LoxP Â Mlx LoxP/LoxP mice were bred to B6.129-Gt(ROSA)26Sor tm1(cre/ERT2)Tyj /J mice (MGI:3699244), which express CreER under the control of the ubiquitously expressed ROSA26 promoter (Jackson Labs, Bar Harbor, ME). 96 Excisional inactivation of each locus was initiated at the time of weaning and confirmed as described earlier. Liver RNAs then were obtained from mice that were the same age as those used for hepatocyte transplants (w5 mo).

Total lipid was extracted from approximately 50 mg liver using the Folch et al 97 method. Total triglyceride content then was determined as described previously using the Free Triglyceride Reagent (Sigma-Aldrich, Inc). 25, 98 Histology, Immunohistochemistry, and Immunoblotting Fresh tissues sections were immediately fixed in formalin, embedded in paraffin, and stained with H&E as previously described. 25, 37, 51 Oil Red O staining and immunohistochemistry on snap-frozen sections also were performed as previously described. 25, 37 Tissue samples for immunoblotting were disrupted in sodium dodecyl sulfate-polyacrylamide gel electrophoresis lysis buffer containing protease and phosphatase inhibitors but lacking b-mercaptoethanol or bromophenol blue as previously described. 25, 46, 51, 73 Protein quantification was performed using the bicinchoninic acid (BCA) reagent according to the supplier's directions (Thermo Fisher Scientific, Rockford, Illinois). After b-mercaptoethanol (1%) and bromophenol blue (10%) addition, samples were boiled for 5 minutes, dispensed into small aliquots, and stored at -80ºC until analysis. Sodium dodecyl sulfate-polyacrylamide gel electrophoresis and semidry transfer to PVDF membranes (Sigma-Aldrich) was performed as previously described. 25, 46, 51, 73 Antibodies used for immunoblotting included rabbit monoclonals directed against Mlx and Myc (85570 and 13987; Cell Signaling Technologies, Inc, Danvers, MA) and a mouse monoclonal antibody against glyceraldehyde-3-phosphate dehydrogenase (G8795; Sigma-Aldrich). A mouse monoclonal anti-Ki-67 antibody used for immunohistochemistry also was from Cell Signaling Technologies (#12202). Horseradish-peroxidase secondary antibodies were from Santa Cruz Biotechnology (Santa Cruz, CA). Ki-67 immunostain quantification was performed using the ImageJ Immunohistochemistry Image Analysis Toolbox (https://imagej.nih.gov/ij/plugins/ihc-toolbox/index.html; National Institutes of Health, Bethesda, MD). All antibodies were used at the dilutions recommended by the suppliers. Immunoblots were developed using an enhanced chemiluminescent assay kit as directed by the supplier (SuperSignal West Pico Plus; Thermo-Fisher, Inc, Waltham, MA).

RNAs were purified from 5 replica tissues from each group of mice using the Qiagen RNeasy Mini Kit (Qiagen, Inc, Germantown, MD), followed by DNase digestion. 25,37 RNA concentration and integrity was confirmed on an Agilent 2100 Bioanalyzer (Agilent Technologies, Foster City, CA) and only those samples with RIN values greater than 8.5 were used for sequencing. All subsequent analyses were performed as previously described. 25, 73 Sample preparation and sequencing was performed on a NovaSeq 600 instrument (Illumina, Inc, San Diego, CA) by Novagene, Inc (Sacramento, CA) and raw data were deposited in the National Center for Biotechnology Information GEO database (accession number: GSE181371). Data sets from previous RNA sequencing studies of MycKO, ChrebpKO, and MycKO Â ChrebpKO livers and mutant forms of b-cateninþYAP S127A HBs are available from the GEO database sets GSE114634 and GSE130178. 25, 73 Differential gene expression was assessed by 3 different approaches, namely EdgeR, CLC Genomics Workbench version 21 (Qiagen), and DeSeq2, as previously described. 25, 47 When low-abundance reads (counts per million <1) were encountered for both comparisons, they were eliminated. Reads from FASTQ files were mapped to the GRCm38.p6 mouse reference genome using STAR (https://github.com/alexdobin/STAR/releases) version 2.7.5. BAM-formatted output was analyzed and transcript abundance was determined by featureCounts (http://bioinf.wehi.edu.au/featureCounts). Where necessary, IPA (Qiagen) was used to classify transcripts into pathways whose significance was adjusted for false discovery using the Bonferroni-Hochberg correction. We further used gene set enrichment analysis 48 to identify alterations of functionally related groups of transcripts from the MSigDB C2 collection (v.7.4) (http://www.gsea-msigdb. org/gsea/msigdb/index.jsp) or from the MitoProteome database (http://www.mitoproteome.org). Volcano plots were generated using the R software package ggplot2 (https://ggplot2.tidyverse.org), with significant differences between samples being defined as having fold differences greater than 1.5 and false discovery rates less than 0.05. Heat maps were generated using the ComplexHeatmap package (version 2.6.2; https://bioconductor.org/ packages/release/bioc/html/Complex Heatmap.html). Statistical analyses were performed with R software v4.0.3 (R Foundation for Statistical Computing, Vienna, Austria) and GraphPad Prism v9.00 (GraphPad Software, Inc, San Diego, CA).

To analyze ChIP sequencing data, we explored binding Myc and Mlx to their target gene sequences in 2 different HepG2 cell lines that had been modified using CRISPR so as to introduce 3 Â FLAG tags at the C-termini of each protein. This allowed ChIP sequencing to be performed under identical conditions with a single anti-FLAG antibody. The results were downloaded from the Encyclopedia of DNA Elements website (https://www.encodeproject.org) and analyzed using ChIPpeakAnno version 3.13 and the annotation database TxDb.Hsapiens. UCSC.hg38.knownGene (R package version 3.13.0). 99 Only binding sites residing within ±2.5 kb of the transcriptional start of each target gene were considered for the current analysis. 55, 100 Overlap between Myc and Mlx binding regions was obtained using the findOverlapsOfPeaks function (set maxgap ¼ 0; minoverlap ¼ 0). Venn diagrams were used to show unique and overlapping binding sites. FIMO (version 5.4.1) from the MEME software suite was used to identify E-boxes and ChoREs most closely associated with ChIP sequencing peaks. Categorization of genes associated with bound peaks was performed using previously described collections of functionally related genes or those from the IPA and MitoProteome databases (Qiagen, Inc, and http://www. mitoproteome.org).

Deregulated Myc requires MondoA/Mlx for metabolic reprogramming and tumorigenesis

A novel role for the extended MYC network in cancer cell survival

The MYC transcription factor network: balancing metabolism, proliferation and oncogenesis

Reflecting on 25 years with MYC

MYC, metabolism, and cancer

MYC as a regulator of ribosome biogenesis and protein synthesis

Impact of MYC in regulation of tumor cell metabolism

The glucose-sensing transcription factor MLX balances metabolism and stress to suppress apoptosis and maintain spermatogenesis

Statistics of protein-DNA binding and the total number of binding sites for a transcription factor in the mammalian genome

Cellular MYCro economics: balancing MYC function with MYC expression

Genome recognition by MYC

The c-Myc target gene network

Functional interactions among members of the MAX and MLX transcriptional network during oncogenesis

MYC deregulation in primary human cancers

Making myc

MYC interacts with the G9a histone methyltransferase to drive transcriptional repression and tumorigenesis

Interactions between Myc and MondoA transcription factors in metabolism and tumourigenesis

The Myc/Max/Mad network and the transcriptional control of cell behavior

The Mad protein family links transcriptional repression to cell differentiation

Normal and neoplastic growth suppression by the extended Myc network

Mechanisms of c-myc-mediated transcriptional repression of growth arrest genes

Transcriptional repression: the dark side of myc

MYC oncogenes and human neoplastic disease

The Mlx network: evidence for a parallel Max-like transcriptional network that regulates energy metabolism

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

Integrated expression profiling and genome-wide analysis of ChREBP targets reveals the dual role for ChREBP in glucose-regulated gene expression

Genome-wide analysis of ChREBP binding sites on male mouse liver and white adipose chromatin

c-Myc and ChREBP regulate glucose-mediated expression of the L-type pyruvate kinase gene in INS-1-derived 832/13 cells

c-Myc is required for the CHREBP-dependent activation of glucose-responsive genes

Mlx, a novel Maxlike BHLHZip protein that interacts with the Max network of transcription factors

MondoA, a novel basic helix-loop-helix-leucine zipper transcriptional activator that constitutes a positive branch of a max-like network

Mnt: a novel Maxinteracting protein and Myc antagonist

Cancer Genome Atlas Network. Pan-cancer alterations of the MYC oncogene and its proximal network across the Cancer Genome Atlas

Diagnosis and treatment of tyrosinemia type I: a US and Canadian consensus group review and recommendations

Pharmacological correction of neonatal lethal hepatic dysfunction in a murine model of hereditary tyrosinaemia type I

Tyrosinemia: a review

Abnormal lipid processing but normal long-term repopulation potential of myc-/-hepatocytes

Liver regeneration after partial hepatectomy: critical analysis of mechanistic dilemmas

Placental rescue reveals a sole requirement for c-Myc in embryonic erythroblast survival and hematopoietic stem cell function

Mitochondrial structure, function and dynamics are temporally controlled by c-Myc

Promotion of growth and apoptosis in c-myc nullizygous fibroblasts by other members of the myc oncoprotein family

c-Myc regulates mammalian body size by controlling cell number but not cell size

c-Myc depletion inhibits proliferation of human tumor cells at various stages of the cell cycle

Hepatic adenomatosis: a rare but important liver disease with severe clinical implications

Role of Myc in hepatocellular proliferation and hepatocarcinogenesis

Genetic dissociation of glycolysis and the TCA cycle affects neither normal nor neoplastic proliferation

Patient-derived mutant forms of NFE2L2/NRF2 drive aggressive murine hepatoblastomas

Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles

New functionalities in the TCGAbiolinks package for the study and integration of cancer data from GDC and GTEx

Complex heatmaps reveal patterns and correlations in multidimensional genomic data

Sequential adaptive changes in a c-Myc-driven model of hepatocellular carcinoma

Sugar sensing by ChREBP/ Mondo-Mlx-new insight into downstream regulatory networks and integration of nutrient-derived signals

MYC and metabolism on the path to cancer

Myc stimulates nuclearly encoded mitochondrial genes and mitochondrial biogenesis

Expanded encyclopaedias of DNA elements in the human and mouse genomes

HOCOMOCO: towards a complete collection of transcription factor binding models for human and mouse via large-scale ChIP-Seq analysis

Coordinated activities of multiple Mycdependent and Myc-independent biosynthetic pathways in hepatoblastoma

MitoProteome: mitochondrial protein sequence database and annotation system

Expression patterns of small numbers of transcripts from functionallyrelated pathways predict survival in multiple cancers

Hepatic expression and cellular distribution of the glucose transporter family

Cancer's molecular sweet tooth and the Warburg effect

c-Myc and cancer metabolism

Deregulation of glucose transporter 1 and glycolytic gene expression by c-Myc

GLUT2, glucose sensing and glucose homeostasis

Metabolic reprogramming: a cancer hallmark even Warburg did not anticipate

Glucose transporters in adipose tissue, liver, and skeletal muscle in metabolic health and disease

Targeting of the MYCN protein with small molecule c-MYC inhibitors

The ups and downs of Myc biology

The role for Myc in coordinating glycolysis, oxidative phosphorylation, glutaminolysis, and fatty acid metabolism in normal and neoplastic tissues

Gene expression in human NAFLD

Role of c-MYC in alternative activation of human macrophages and tumor-associated macrophage biology

Hepatic adenomas: classification, controversies, and consensus

beta-Catenin mutations as determinants of hepatoblastoma phenotypes in mice

Hippo/YAP, beta-catenin, and the cancer cell: a "menage a trois" in hepatoblastoma

Reconciling the biological and transcriptional variability of hepatoblastoma with its mutational uniformity

Liver regeneration by stem/progenitor cells

Hypertrophy and unconventional cell division of hepatocytes underlie liver regeneration

Reduced expression of MYC increases longevity and enhances healthspan

In vivo Reprogramming of cancer metabolism by MYC

Metabolic and oncogenic adaptations to pyruvate dehydrogenase inactivation in fibroblasts

Max deletion destabilizes MYC protein and abrogates Emicro-Myc lymphomagenesis

The metabolic fates of pyruvate in normal and neoplastic cells

MYC inhibition induces metabolic changes leading to accumulation of lipid droplets in tumor cells

Structurally diverse c-Myc inhibitors share a common mechanism of action involving ATP depletion

Epidemiology and risk-stratification of NAFLD-associated HCC

Bi-allelic inactivation of TCF1 in hepatic adenomas

The role of Merlin/NF2 loss in meningioma biology

Genome-wide DNA methylation indicates silencing of tumor suppressor genes in uterine leiomyoma

Frequent inactivating mutations of the PBAF complex gene PBRM1 in meningioma with papillary features

Liver cancer: connections with obesity, fatty liver, and cirrhosis

Nonalcoholic fatty liver disease

Inhibition of hepatocellular carcinoma by metabolic normalization

de Alboran IM. c-Myc regulates cell size and ploidy but is not essential for postnatal proliferation in liver

Analysis of C-MYC function in normal cells via conditional gene-targeted mutation

Sloane Stanley GH. A simple method for the isolation and purification of total lipides from animal tissues

Measurement of liver triglyceride content

ChIPpeakAnno: a Bioconductor package to annotate ChIP-seq and ChIP-chip data

The ENCODE Project Consortium. An integrated encyclopedia of DNA elements in the human genome

Correspondence Address correspondence to: Edward V. Prochownik, MD, PhD, Division of Hematology/Oncology, Children's Hospital of Pittsburgh

PhD (Conceptualization: Lead; Data curation: Lead; Formal analysis: Lead; Investigation: Lead; Writing -original draft: Equal

Investigation: Equal)

Investigation: Supporting)

Investigation: Supporting)

Resources: Supporting)

Resources: Supporting)

Formal analysis: Supporting)

Formal analysis: Equal)

Conceptualization: Equal; Funding acquisition: Equal; Investigation: Equal

Conceptualization: Lead; Data curation: Equal

Formal analysis: Lead; Funding acquisition: Lead; Investigation: Lead

Supervision: Lead; Writing -original draft: Lead; Writing -review & editing: Lead) by National Institutes of Health RO1 grants DK103645

Children's Hospital of Pittsburgh Summer Research Undergraduate Fellowship Program