The Myc and Mlx Networks. The year 2021 marked the 30th anniversary of the discovery of Max, the obligate heterodimerization partner of the c-Myc (Myc) bHLH-ZIP transcription factor, which had itself been identified nearly a decade earlier as the evolutionarily conserved cellular homolog of the avian v-myc retroviral oncogene [1,2,3,4,5]. Shortly thereafter, it became clear that Max was necessary for Myc-mediated target gene activation and cellular transformation but that, at higher Max:Myc ratios, it also repressed these functions in a dose-dependent manner [6,7,8,9,10,11]. This initially suggested a simple model, whereby Myc–Max heterodimers, which contain a transactivation domain (TAD) contributed by Myc, and Max homodimers, which lack a TAD, alternatively activate or suppress transcription in accord with their relative abundance [12,13]. This balance was believed to be maintained by the competitive binding of these dimers to so-called “E-box” elements that are typically located in proximity to the transcriptional start sites of target genes (Figure 1). The model immediately implied that Max could serve dual functions—on the one hand, it could activate target genes and drive transformation by virtue of its obligate association with Myc; as a homodimer, on the other hand, it could outcompete Myc–Max binding, repress target gene expression and serve as a tumor suppressor (TS).

Max’s discovery soon paved the way for the rapid identification of six additional bHLH-ZIP Max partners that currently comprise the so-called “Myc Network”, of which Max remains the central member (Figure 1). These proteins, Mxd1–4, Mnt and Mga (hereafter referred to collectively as “Mxd proteins”, despite their structural and functional differences) (Figure 1), necessitated a revision of the above model as did the finding that, in mammalian cells, Max’s high level of phosphorylation, maintained by casein kinase II, prevents its binding to DNA as a homodimer but not as a heterodimer [19,20,43]. Transcriptional repression by Max was thus deemed to be mediated not simply by passive exclusion of Myc–Max heterodimers from binding E-boxes but as an active process that required Max’s heterodimerization with Mxd proteins, the association with obligate transcriptional co-repressors such as mSin3 proteins and the ensuing transcriptionally suppressive modification of chromatin via altered patterns of histone acetylation and methylation [14,28,44,45,46,47]. Collectively, these six Mxd proteins, which show distinct tissue-, developmental- and age-dependent patterns of expression, antagonize Myc’s broad impact on transcription and its highly pleiotropic effects on both normal and neoplastic growth [6,7,8,22,24,46,48,49,50,51,52,53,54,55].

By the turn of the 21st century, the existence of a second network, parallel to that of the Myc Network, and possessing significant structural and functional similarities, had emerged. This “Mlx Network”, also comprised of bHLH-ZIP factors, contained the Myc-like proteins ChREBP (“carbohydrate response element binding protein”) and MondoA and their own dedicated heterodimerization partner, the Max-related Mlx, which together formed the Network’s positively-acting arm [31,56,57,58] (Figure 1). Unlike Myc and Max, which are nuclear proteins, ChREBP, MondoA and Mlx are “conditionally nuclear” in that they translocate to the nucleus only upon binding metabolites such as glucose, glucose-6-phosphate, fructose 2,6-bisphosphate, lactate and adenosine [31,32,33,34,35,36,37,38,40,42,56,59,60,61,62,63,64,65,66,67,68]. The cytosolic location of MondoA and ChREBP is also not random; rather, amphipathic helical domains in the C-termini of these factors allow them to interact with intracytoplasmic lipid droplets and presumably serve as sensors of intracellular lipid content [69]. Lipid droplet depletion, together with the above metabolites, allows MondoA and ChREBP to translocate to the nucleus and activate genes involved in glucose metabolism. The negative arm of the Mlx Network shares three factors from the Myc Network, namely Mxd1, Mxd4 and Mnt, along with mSin3–histone deacetylase complexes [31,57,70]. These three negative regulators link the two Networks both structurally and functionally and provide a means of cross-talk for more refined and coordinated regulation of both common and unique targets.

Mlx Network heterodimers recognize E-box-related binding sites termed “carbohydrate response elements” (ChoREs) (Figure 1). Both E-boxes and ChoREs, but particularly the latter, can be quite variable and cross-binding of one Network’s members to the other Network’s sites is likely more common than previously appreciated [57,71,72,73,74,75,76]. Binding by ChREBP may also facilitate promiscuous, non-E-box-dependent binding by Myc at nearby sites [77]. Mlx Network target genes, while overlapping with those regulated by the Myc Network, are both fewer in number and have been reported to be more functionally restricted [39,41,78]. They tend to encode enzymes involved in carbohydrate and lipid metabolism; mitochondrial and ribosomal proteins and factors that regulate translational initiation, elongation and termination [31,39,41,42,46,70,71,72,73,76,78,79,80,81,82,83,84,85,86,87,88,89]. Together, these findings indicate that, via selective, cytoplasmic-nuclear partitioning, sharing of Mxd1,4 and Mnt and binding to both common and unique target genes, this “Extended Myc Network” cross-talks, while simultaneously communicating with and responding to the metabolic cues and reservoirs needed to support energy-intensive processes such as translation and proliferation [36,42,64,65,69,88,89,90,91,92,93,94,95,96].

The Extended Myc Network and tumor suppression. Given the number of Mxd proteins and their crucial roles, particularly with regard to limiting Myc signaling, proliferation and metabolism (Figure 1), it is reasonable to surmise that, at some level, they, and perhaps Max as well, behave as TSs. We discuss below the evidence to support this, the findings that the inactivation of these factors is often confined to certain tumor types and the implications of Mxd member haplo-insufficiency, which is a common theme in many tumors. It is also important to consider these factors in their proper biological context. In other words, do they behave as “classical” TSs such TP53, RB and BRCA1/2, whose germ line or somatic inactivation predispose to spontaneous neoplasms such as leukemias, lymphomas, retinoblastomas, osteosarcomas, and breast, ovarian and prostate cancer [97,98,99,100,101,102]? Or, does the inactivation of these factors simply accelerate the growth of pre-existing tumors without otherwise contributing directly to their initiation?

All Mxd members, as well as the central players Max and Mlx, have been implicated as TSs based on bioinformatics-based surveys of large populations of human tumors [103,104]. However, to our knowledge, neither comprehensive summaries nor intragroup comparisons of the consequences of copy number variations (CNVs) and mutations described in these reports have been published. By way of introduction to the more detailed discussions presented in the following sections, we surveyed all tumors from The Cancer Genome Atlas (TCGA) to identify those associated with CNVs of Extended Myc Network genes (Figure 2). This survey revealed several interesting findings. First, and in keeping with the known role of MYC as a bona fide oncogene and the presumptive roles of CHREBP and MONDOA in supporting transformation and rapid proliferation, these genes showed the highest frequency of amplification [36,41,45,103,104]. Second, recurrent deletions of all other Extended Myc Network members were quite frequent across many tumor types, in keeping with their presumptive role as TS genes (TSGs). Third, none of the genes was exclusively amplified or deleted across all tumor types or, in many cases, even within the same tumor type [104]. Fourth, among the “oncogenes”, high-level gene amplification, defined as >4.4 copies/cell [105], was seen only with MYC and tended to be associated with subgroups of certain cancers such as breast, ovary, liver and lung as previously reported [106,107,108,109]. Fifth, more often than not, amplifications or deletions of more than one Extended Myc Network gene were associated with particular tumor types and were most obvious in breast, ovarian and squamous cell lung cancers. This suggests that oncogenic activation of the Extended Myc Network can be achieved via multiple pathways that are cooperative rather than mutually exclusive. Sixth, despite its relatively high incidence of genomic instability, thyroid cancer was the only tumor in which CNV of any Extended Myc Network member was never observed (Figure 2). This may partly reflect the tumor’s high incidence of oncogenic BRAF, PIK3CA and PTEN mutations, which likely circumvent Myc Network activation [104,110]. In keeping with the theme that high-level MYC amplification tends to be confined to certain tumor types (Figure 2) [104], deletion of presumptive TSGs tend to cluster as well, particularly in those instances associated with two copy deletion as seen with MLX and MNT in ovarian cancer, MXD2 in prostate cancer and MXD3 in squamous cell lung cancer (Figure 2) [104]. Finally, the inactivation of TSG-like members of the Extended Myc Network by frame-shift and truncating mutations is rare, typically accounting for <1% of inactivated alleles, except in the case of MGA, where this approaches 4% [104]. Thus, as discussed more fully in the following sections, rather than the traditional bi-allelic inactivation/deletion of classical TSGs seen in both hereditary and sporadic cancers, hemizygous inactivation of MXD genes is not only common but a recurrent theme [98,99,100,101,102].

As was generally true for oncogenic members of the Extended Myc Network, the correlation between MXD TSG expression and survival is variable. Figure 3 summarizes these relationships among all tumors from TCGA and Figure 4 shows several specific examples of survival differences. Interestingly, survival correlated with the expression of as many as 8–9 of the individual Extended Myc Network members and in the cases of low-grade gliomas (LGGs) and clear cell renal cancers (KIRC), with all 11 members, although not necessarily in the same way. This supports the points mentioned above that perturbing the delicate balance among Extended Myc Network members may have oncogenic consequences that are additive or complementary rather than mutually exclusive. The connections among tumor aggressiveness, survival and the expression of individual members of the Extended Myc Network are thus complex, interdependent and presumably tumor type specific.

The relationships among Extended Myc Network gene CNVs, mutation, expression and survival are in many cases only correlative. Indeed, this cautionary note applies even to an oncogene as unassailable as MYC whose role in tumorigenesis mandated a direct demonstration of its actual causality [113,114,115]. Yet, recent work has indicated that, at least for some tumor types, Myc may function more as a facilitator of tumor growth than as an actual initiator. For example, hepatoblastoma (HB) in mice, induced via the ectopic expression of mutant forms of β-catenin and the Hippo pathway effector protein YAP, is associated with Myc over-expression [78,116,117]. Yet, the rate of tumor induction in livers with hepatocyte-specific knockout of Myc is identical to that seen in wild-type livers [41,78] although tumor growth is slowed. Similarly, tumors induced with different β-catenin mutants express widely different levels of Myc that roughly correlate with growth rates even though they too are initiated at identical frequencies [117]. These results, strongly indicate that the role of Myc in actual tumor induction versus facilitation is quite nuanced and may reflect both the levels at which it is expressed and the tissue environment [118].

A somewhat analogous situation, discussed in greater detail below, occurs with Max. While there is no evidence that MAX is an actual oncogene, it could be viewed as being another tumor facilitator that collaborates with Myc (Figure 1). Yet, the recurrent loss of MAX in pheochromocytomas, parganglioneuromas and several other tumors, as shown in Figure 2 and elsewhere, supports its role as a potent TSG [119,120,121,122,123,124,125]. Similarly, MLX loss dramatically impairs normal hepatocyte proliferation while at the same time serving as a potent suppressor of benign hepatic adenomatosis [76].

2. Myc