Intricate crosstalk between MYC and non‐coding RNAs regulates hallmarks of cancer

Myelocytomatosis viral oncogene homolog (MYC) plays an important role in the regulation of many cellular processes, and its expression is tightly regulated at the level of transcription, translation, protein stability, and activity. Despite this tight regulation, MYC is overexpressed in many cancers and contributes to multiple hallmarks of cancer. In recent years, it has become clear that noncoding RNAs add a crucial additional layer to the regulation of MYC and its downstream effects. So far, twenty‐five microRNAs and eighteen long noncoding RNAs that regulate MYC have been identified. Thirty‐three miRNAs and nineteen lncRNAs are downstream effectors of MYC that contribute to the broad oncogenic role of MYC, including its effects on diverse hallmarks of cancer. In this review, we give an overview of this extensive, multilayered noncoding RNA network that exists around MYC. Current data clearly show explicit roles of crosstalk between MYC and ncRNAs to allow tumorigenesis.


Introduction
The MYC gene family consist of three members, that is, c-MYC, n-MYC, and l-MYC. c-MYC forms a central hub in all cells by regulating many cellular processes, while n-MYC and l-MYC are more tissue-specific regulators. MYC proteins are overexpressed in more than half of all human cancers, including lung, breast, and colon cancers (Albihn et al., 2010). This overexpression is caused by diverse mechanisms including amplifications, translocations, and epigenetic alterations (Kalkat et al., 2017). In this review, we will focus on c-MYC, hereafter referred to as MYC.
MYC belongs to the basic helix-loop-helix superfamily and functions as a transcription factor. Upon dimerization with its binding partner MAX, the MYC-MAX dimer binds to E-box sequences in the promoter region of its targets genes, thereby activating transcription of these genes (Tu et al., 2015). In addition to interacting with MAX, MYC can also interact with other transcription factors, histone-modifying enzymes, and DNA methyltransferases to repress transcription. MYC regulates the transcription of many different genes, which include protein-coding as well as noncoding genes (Dang, 2012;Hart et al., 2014;Winkle et al., 2015). These noncoding genes can include various RNA molecules, for example, miRNAs and lncRNAs. miRNAs are noncoding, regulatory RNA molecules of about 22 nucleotides in length. A miRNA is transcribed as a longer primary transcript, which is processed in two steps into a mature single-stranded miRNA and subsequently incorporated into the RISC. The miRNA guides the RISC complex to its target mRNA by recognition of a complementary sequence, most often in the 3 0 UTR. Usually, conserved Watson-Crick pairing with nucleotides 2-7 of the miRNA, the so-called seed region, is essential for target recognition (Bartel, 2009). Binding to the target mRNA will subsequently result in mRNA cleavage by AGO2 in case the miRNA has high complementarity with the binding site region on the mRNA. In case of a low level of complementarity, binding will lead to translational repression.
LncRNAs are defined as noncoding RNA molecules of more than 200 nucleotides in length. Their expression is often tissue specific or cell type specific, and their transcripts can have subcellular compartment-specific localizations. Together, this restricts their function to specific cell types and locations. LncRNAs can regulate gene expression at the transcriptional and post-transcriptional level, as well as by modulating protein stability, localization, and functionality via diverse mechanisms. In the nucleus, lncRNAs can regulate transcription of nearby genes in cis or of more distant genes in trans, for example, by recruiting transcription factors, chromatinmodifying complexes, or heterogeneous nuclear ribonucleoprotein (hnRNP) complexes. LncRNAs residing in the cytoplasm can modulate mRNA stability, translation efficiency, or protein stability, localization, or activity. Cytoplasmic lncRNAs can act as decoys to sequester RNA binding proteins or miRNAs (sponges or ceRNAs) or interfere with post-translational modifier proteins (Chen, 2016;Schmitt and Chang, 2016).
Over the last decades, it has become clear that MYC is not only regulated by and regulates many proteincoding genes, but this extensive network also includes the family of ncRNAs. The overall aim of this review was to present an overview of the intricate crosstalk between ncRNAs and MYC. We first focus on ncRNAs acting upstream of MYC by regulating its transcription, translation, and activity. In addition, we focus on ncRNAs acting downstream of MYC and pinpoint their contributions to crucial hallmarks of cancer.
Next to regulating MYC in a direct fashion, miR-24-3p can also influence MYC protein levels indirectly by targeting OGT. OGT can O-GlcNAcylate the MYC protein and thereby increase its stability . A second miRNA that can act indirectly on MYC is miR-375-3p, which targets CIP2A. CIP2A prevents phosphorylation of Ser62 on MYC by PP2A and thereby prevents degradation of MYC (Jung et al., 2013). So, miR-24-3p and miR-375-3p can downregulate MYC protein levels indirectly by targeting OGT and CIP2A, respectively.
Many of the miRNAs that can directly downregulate MYC by binding to the MYC mRNA, show reduced levels in cancer. The decreased expression of these miRNAs can thus contribute to the high levels of MYC as commonly observed in cancer. Examples are the let-7-5p family, miR-148a-5p, miR-331-3p, and miR-363-3p, which are downregulated in Burkitt lymphoma compared to normal lymph nodes (Bueno et al., 2011). A well-known exception is miR-17-5p, which is part of the oncogenic miR-17~92 cluster that is often upregulated in MYC-driven cancers. As too high MYC levels are potentially dangerous for cancer cells, targeting of MYC by miR-17-5p may be a means to maintain optimal MYC levels and sustain continuous tumor growth .

lncRNAs regulating MYC
Expression of MYC is controlled at the level of transcription, translation, and protein stability. Several lncRNAs have been demonstrated to play a role in these regulatory processes. Here, we describe the lncRNAs with a well-characterized role in MYC regulation (Fig. 2).
enhancing the activity of TCF7L2, a transcription factor for MYC (Ling et al., 2013). Thus, both CCAT1-L and CCAT2 positively regulate MYC transcription.
Interaction between an enhancer region downstream the first transcriptional start site of PVT1 and the PVT1 promoter itself has tumor suppressor activity by reducing MYC transcription (Cho et al., 2018). Silencing of the PVT1 promoter increased MYC expression independent of the PVT1 transcript itself. The underlying mechanism has been identified as a competition between the PVT1 promoter and the MYC promoter for interaction with the intragenic enhancer region in the PVT1 locus. Under normal conditions, these enhancers preferentially bind to the PVT1 promoter. Silencing of the PVT1 promoter allowed interaction of enhancers with the MYC promoter, leading to increased MYC transcription. Importantly, this effect is restricted to cells where MYC forms chromatin loops with PVT1, for example, breast cancer, as opposed to colorectal cancer or cervical carcinoma cells where MYC loops to the CCAT1 enhancer. The levels of three partially overlapping lncRNA transcripts antisense to the 3 0 distal region of MYC, NAT6531, NAT6538, and NAT7281, are regulated by histone H3 acetylation in prostate cancer cells. Under normal conditions, NAT6531 is expressed and processed by DICER into several short RNAs, which have a repressive effect on MYC transcription, possibly by binding to the MYC promoter and intron 1 through partial sequence complementarity. Partial inhibition of histone deacetylation shifts transcription from NAT6531 to NAT6538, and this releases the block on MYC transcription. Strong inhibition of histone deacetylation results in transcription of the longer NAT7281, which strongly represses MYC transcription (Napoli et al., 2017).

LncRNAs controlling MYC mRNA stability and translation
IGF2BPs enhance mRNA stability and promote translation by binding to the MYC mRNA . A number of cell type-specific lncRNAs have been identified that modulate this interaction. Interaction of IGF2BP1 with lncRNA GHET1 in gastric cancer and THOR in renal and skin cancer increased MYC mRNA and protein levels Yang et al., 2014;Ye et al., 2018). In contrast, binding of the skeletal muscle-specific lncRNA lncMyoD to IGF2BP2 decreased MYC mRNA levels by preventing binding of IGF2BP2 to MYC mRNA (Gong et al., 2015).
Binding of AUF1 to an ARE site in the 3 0 UTR of the MYC transcript can both positively and negatively affect MYC levels, depending on the cell-type. In normal kidney cells, FILNC1 acts as a decoy for AUF1 preventing binding of AUF1 to the MYC mRNA, thereby resulting in low MYC protein levels. In renal cancer, FILNC1 is downregulated, resulting in an AUF1-dependent increase in MYC protein levels . In breast and colon cancers, binding of linc-RoR to AUF1 inhibits binding of AUF1 to MYC mRNA and thereby increases MYC levels (Huang et al., 2015). It is currently unclear why sequestering of AUF1 has opposite effects on MYC levels in these different cell types. In addition, linc-RoR facilitates binding of RNA binding protein hnRNP-I to MYC mRNA and this also enhances MYC protein levels.
MYC can be translated using an IRES in case the regular cap-dependent translation is compromised. This requires binding of the IRES trans-acting factors PSF and p54nrb (Cobbold et al., 2008). These factors are sequestered by lncRNA NEAT1 to the paraspeckles. In HeLa cells, depletion of NEAT1 during nucleolar stress released PSF and p54nrb from paraspeckles and allowed IRES-dependent translation of MYC (Shen et al., 2017).
LncRNAs can also stimulate MYC mRNA translation by competing with MYC-regulating miRNAs. This has been shown for PCAT-1, which competes with miR-34a-5p for interaction with its binding site in the 3 0 UTR of the MYC mRNA (Prensner et al., 2011(Prensner et al., , 2014. The effect of PCAT-1 can be antagonized by miR-3667-3p, which targets PCAT-1.

LncRNAs affecting MYC protein stability and activity
The stability of MYC protein can be increased by two lncRNAs that both prevent its degradation, but via distinct mechanisms. In contrast to the tumor-suppressive role of the PVT1 promoter, the PVT1 transcript can act as an oncogene. PVT1 stabilizes the MYC protein by preventing phosphorylation of threonine 58, which is a signal for its degradation (Tseng et al., 2014). LINC01638 prevents MYC protein degradation by preventing binding of E3 ubiquitin ligase adapter SPOP to MYC (Luo et al., 2018).
Three lncRNAs modulate interaction of MYC with (subsets of) its target genes by directly binding to MYC. PCGEM1 is a prostate-specific lncRNA, which together with MYC co-occupies the promoter regions of several metabolic genes documented to be MYC targets. Knockdown of PCGEM1 reduced recruitment of MYC to the promoters of these PCGEM1-dependent metabolic genes without affecting MYC protein levels (Hung et al., 2014). Thus, PCGEM1 affects the metabolic state of cancer cells by enhancing MYC occupancy at the promoters of several metabolic genes. LncRNA PDIA3P regulates the metabolic state of multiple myeloma cells via induction of G6PD, an enzyme crucial for promoting the PPP flux . This effect is achieved by interaction of PDIA3P with MYC and promoting MYC binding to the G6PD promoter. Together with MYC, lncRNA EPIC1 co-occupies the promoters of > 97% of EPIC1regulated genes involved in cell cycle progression, and thereby regulates transcriptional activity of these genes in breast cancer cells .
From the studies presented here, lncRNAs emerge as important regulators of MYC expression and activity, either directly or indirectly by interacting with proteins. Often, these lncRNAs are deregulated in cancer and promote high MYC levels and activity. Since expression of lncRNAs is highly cell type specific, many of the lncRNA-MYC interactions are restricted to certain tissues. Future studies will likely broaden the repertoire of lncRNAs regulating MYC and improve the understanding of the underlying mechanisms in normal and cancer cells.

Feedback loops on MYC
Next to the more straightforward regulation of MYC by ncRNAs as described above, more complex feedback loops between MYC and MYC-regulating ncRNAs have been identified. These include feedback loops that involve MYC-regulated miRNAs, as well as MYC-regulated lncRNAs that act as sponges for MYC-regulating miRNAs.

Feedback loops involving MYC-regulated miRNAs
Several miRNAs that regulate MYC can be induced or repressed by MYC as well, resulting in the formation of feedback loops. Examples of this are the feedback loops between MYC and MYC-induced miR-7-5p (Capizzi et al., 2017;Chou et al., 2010), miR-17-5p , and miR-185-3p (Liao and Liu, 2011). For miR-7-5p, a positive feedback loop is formed via the miR-7-5p target AMBRA1, which promotes dephosphorylation of Ser62 on MYC upon binding to PP2A. This leads to stimulation of proteosomal degradation of MYC (Capizzi et al., 2017;Cianfanelli et al., 2015). In this way, miR-7-5p indirectly enhances MYC protein stability and promotes its own MYC-mediated transcription. MiR-17-5p and miR-185-3p were shown to directly target MYC mRNA resulting in a negative feedback loop (Liao and Liu, 2011;Liu et al., 2016).
Both targets are involved in the induction of MYC transcription, creating another feedback loop. MiR-200b-3p participates in a feedback loop that involves MYC protein stability by targeting Akt2 mRNA . Akt2 represses the activity of GSK3b, an enzyme that destabilizes the MYC protein by phosphorylation of threonine residue 58. Thus, by repressing miR-200b-3p, MYC ensures inhibition of GSK3b, thereby stimulating its own stability. In contrast, MYC-repressed miR-30a-5p is involved in a negative feedback loop by targeting UBE3C mRNA, a protein that can ubiquitinate MYC for proteosomal degradation (Chang et al., 2008;Xiong et al., 2016).
One of the first identified MYC-regulated lncRNAs is CCAT1. While the CCAT1-L transcript variant is specifically overexpressed in colorectal cancer, the CCAT1-S variant is upregulated in many other cancers, including gastric carcinoma and colon cancer (He et al., 2014;Yang et al., 2013a). By binding to the E-box element in the promoter region of CCAT1, MYC induces expression of CCAT1-S. As the short transcript variant is most likely formed by 3 0 processing of the long variant, MYC probably induces expression of CCAT1-L, but this has not been proven. Besides CCAT1 and CCAT2, six other colorectal cancer-associated MYC-regulated lncRNAs (MYCLos/ CCAT3-8) have been identified (Kim et al., 2015a,b). Three of them are MYC-induced, and the other three are MYC-repressed. In the last five years, many more MYC-regulated lncRNAs have been identified although for many their function has not yet been identified (Hart et al., 2014;Winkle et al., 2015).
Below, we describe in more detail the MYC-regulated miRNAs (Table 1 and Fig. 3) and lncRNAs (Table 2 and Fig. 4) with a clear role in five main hallmarks of cancer, that is, cell cycle progression, apoptosis, metabolism, angiogenesis, and metastasis.

Cell cycle progression
Nineteen MYC-induced ncRNAs have a role in cell cycle progression. LncRNA-assisted stabilization of transcripts (LAST) stimulates CCND1 expression by stabilizing CCND1 mRNA together with CNBP . MiR-378a-3p ensures CCND1 expression by targeting mRNA encoding TOB2, which is a repressor of CCND1 expression (Feng et al., 2011). CASC11 (CARLo-7) promotes CCND1 transcription by stabilizing the hnRNP-K mRNA, which leads to an hnRNP-K-dependent enhanced nuclear accumulation of b-catenin (Zhang et al., 2016b). This leads to activation of WNT/b-catenin signaling, and the subsequent induction of CCND1 transcription. The MYC-induced lncRNA MY (VSP9D1-AS1) associates with hnRNP-K and stimulates CDK6 mRNA translation by competing with miR-16-5p for binding to CDK6 mRNA (Kawasaki et al., 2016). CDKN2B transcription is repressed by lncRNA CCAT-6 upon binding of this lncRNA to hnRNP-K (Kim et al., 2015b). All three lncRNAs interacting with hnRNP-K (CASC11, MYU, and CCAT-6) have been shown to stimulate cell cycle progression in colon cancer. The four lncRNAs HOTAIR, MYCLo-1, CCAT1-S, and DANCR all repress CDKN1A transcription (Kim et al., 2014(Kim et al., , 2015bLiu et al., 2013;Lu et al., 2018;Ma et al., 2014). HOTAIR represses CDKN1A transcription by recruiting EZH2 and inducing epigenetic changes, while MYCLo-1 is assisted by HuR to repress the transcription of CDKN1A. The mechanisms by which CCAT1-S and DANCR repress CDKN1A transcription are not yet known. Members of the miR-17-5p seed family have been strongly implicated in stimulation of cell cycle progression by targeting CDKN1A (Ivanovska et al., 2008;Kim et al., 2009;Trompeter et al., 2011). Conversely, the same seed family represses cell cycle progression by targeting CCND1/2 transcripts (Trompeter et al., 2011;Yu et al., 2008) and E2F1-3 transcripts (He et al., 2005;Luan et al., 2018;O'Donnell et al., 2005;Trompeter et al., 2011). This is consistent with the cell type-specific roles as oncomiR as well as tumor suppressor miR that have been observed for individual members of the miR-17-5p seed family (He et al., 2005;O'Donnell et al., 2005). The MYC-induced lncRNA CONCR plays a role during S-phase and is required for cell division by regulating the activity of helicase DDX11, which is involved in DNA replication and sister chromatid cohesion (Marchese et al., 2016). The MYCinduced lncRNA MINCR promotes MYC-mediated transcription of a selected set of cell cycle genes (Doose et al., 2015), although there is some debate about whether this lncRNA is a direct MYC-induced lncRNA or not (Doose et al., 2015(Doose et al., , 2016Hart et al., 2016). Besides, MINCR functions as a sponge for miR-26a-5p to stimulate cell cycle progression .

Metabolism
Three MYC-induced and eight MYC-repressed ncRNAs are involved in the regulation of aerobic glycolysis, a feature of cancer cells. By targeting PTEN and PP2K transcripts, miR-19a/b-3p enhances PI3K activity (Mavrakis et al., 2010;Mu et al., 2009;Olive et al., 2009). This results in phosphorylation of Akt by PDK1, which stimulates glycolysis through multiple mechanisms, such as increased expression of several glucose transporters, activation of PFK1/2 (important regulatory enzymes of glycolysis), and mTOR. To further ensure high mTOR activity, miR-19a/b-3p also targets AMPK, an inhibitor of mTOR activity (Bolster et al., 2002;Mavrakis et al., 2010). MiR-106a-5p targets the E2F3 transcript, which results in repression of the glucose metabolism (Luan et al., 2018). This is antagonized by H19, which has been proposed to promote glucose metabolism by acting as a sponge for miR-106a-5p. MIF influences the glycolytic activity by sequestering miR-586, thereby preventing expression of MYC target genes involved in glycolysis, that is, GLUT1, LDHA, PKM2, and HK2 (Zhang et al., 2016a). miRNAs repressed by MYC typically inhibit high metabolic activity. The initial uptake of glucose is regulated by miR-195-5p, which targets GLUT3 (Fei et al., 2012). MiR-23a/b-3p targets the mRNA Cell cycle progression, metastasis Cell cycle progression, metabolism, angiogenesis, metastasis HOTAIR ↓ CDKN1A, WIF1, miR-34a-5p Cell cycle progression, metastasis, LAST ↑ CCND1 Cell cycle progression Linc00176 ↓ miR-9-5p, miR-185-5p Cell cycle progression LncRNA-MIF ↓ miR-586-5p Metabolism MINCR ↓ miR-26a-5p Cell cycle progression, apoptosis, metastasis Cell cycle progression MYCLo-5 Unknown Cell cycle progression MYCLo-6 ↑ GADD45A Cell cycle progression ↑ indicates induced/stabilized/activated by the lncRNA, and ↓ indicates being repressed by the lncRNA. a Not all proven target genes mentioned in column two are involved in the cellular processes mentioned here. encoding GLS, which converts glutamine to glutamate and thereby contributes to production of ATP ). In addition, miR-23a-3p targets LDH subunits A and B (LDHA/LDHB), which convert the glycolytic end product pyruvate to lactate (Poyyakkara et al., 2018). Moreover, LDHA is also targeted by miR-30a-5p (Chang et al., 2008;Li et al., 2017a). MiR-26a-5p inhibits PDH activity by targeting PDHX and therefore inhibits the conversion of pyruvate to coenzyme A, an important component of the TCA cycle (Chen et al., 2014a). Instead, pyruvate is converted to lactate, showing an oncogenic role for miR-26a-5p in metabolism. In contrast, miR-129 targets PDK4 mRNA, thereby stimulating PDH activity . MYC-repressed lncRNA IDH1-AS1 stimulates homodimerization of IDH1 by forming a ternary structure with the enzyme, thereby enhancing its activity . IDH1 converts isocitrate to a-ketoglutarate, which is an intermediate in the TCA cycle and can inhibit glycolysis via degradation of HIF1a under normoxic condition (MacKenzie et al., 2007). By repressing IDH1-AS1, MYC downregulates IDH1 activity and ensures glycolysis.

Discussion
It is evident that an extensive, multilayered ncRNA network exists around MYC with critical roles for multiple lncRNAs and miRNAs in crucial cellular processes and in tumorigenesis. The picture that we present here is most likely still far from complete, as functions of most of the MYC-regulated ncRNAs are not known yet (Hart et al., 2014;Robertus et al., 2010;Winkle et al., 2015). It is clear that many miR-NAs and lncRNAs regulate MYC and that they can do this via diverse mechanisms at the level of transcription, translation, protein stability, and functionality. This suggests that redundancy is important to ensure optimal MYC levels and thereby cell viability under various conditions, as well as in different cell types. As MYC is involved in many cellular processes in redundant ways, it is remarkable that repression or reintroduction of a single MYC-regulated ncRNA can already show strong effects on MYC-associated phenotypes, as has been shown for many ncRNAs described in this review.
Expression of lncRNAs was shown to be more cell type specific than that of protein-coding genes (Derrien et al., 2012). Also compared to miRNAs, lncRNAs appear to be more cell type-specific. However, this might be biased as there are many more lncRNAs than miRNAs, which increases the chance to find cell type-specific lncRNAs. Based on current knowledge, it seems that the cell type-specific expression of certain lncRNAs can influence the output of MYC in two ways. First, cell type-specific lncRNAs can influence important cellular processes downstream of MYC (Fig. 4). Second, other cell type-specific lncRNAs, like PCGEM1 and PDIA3P, can modulate binding efficiency of MYC to promoters of a specific set of genes. So, these lncRNAs may direct the cell type-specific target gene repertoire of MYC, rather than MYC acting as a general amplifier of expression. Altogether, a picture is emerging that lncRNAs guide cell type-specific effects of MYC.
Although MYC has a central role in tumorigenesis, no effective MYC-specific drugs are being employed in the clinic to date. Given the crucial functions of multiple lncRNAs and miRNAs in the oncogenic MYC network, it is tempting to speculate that targeting of ncRNAs within the MYC network might be an alternative to explore novel anticancer therapies. These ncRNAs can have profound impacts on MYC levels and activity and can also act downstream of MYC enabling cancer cells to gain the crucial hallmarks of cancer. To allow selection of the most optimal ncRNA targets, a more systematic analysis of their functional networks in normal cells as well as in cancer cells needs to be performed to oversee the consequences of targeting them.
Currently, more and more institutes and companies investigate how to specifically target miRNAs and lncRNAs, using both antisense and small moleculebased strategies (Chakraborty et al., 2017;Warner et al., 2018). Inhibitors for miR-92 and miR-122, as well as mimics of miR-16, miR-29 and miR-34, have been developed and tested or are currently tested in clinical trials (NIH U.S. National Library of Medicine, https://clinicaltrails.gov/ (accessed 06.08.2018)). As miR-34a-5p has tumor suppressor activity by both targeting MYC and stimulating apoptosis, while repressing cell cycle progression and metastasis, it is an attractive target for novel anticancer therapies. MiR-16-5p and miR-29-3p too have tumor-suppressive roles in four of the five hallmarks discussed and form attractive targets as well. The cell type-specific expression of lncRNAs adds to their attractiveness as targets for therapy (Derrien et al., 2012). The choice for an attractive target will therefore depend on the type of cancer. For example, CCAT1-L and CCAT2 form attractive targets to specifically inhibit MYC transcription in colorectal cancer. A drug against CCAT1-L, which will also target CCAT1-S, would be very interesting as it will inhibit cell cycle progression and metastasis, while promoting apoptosis. However, a main problem for testing effectivity of lncRNA-based drugs is the limited conservation for many of the lncRNAs, which prevents pre-clinical experiments in relevant mouse models. Patient-derived xenotransplantation models or organoid cultures might represent an alternative approach to test effectiveness of targeting human-specific lncRNAs.
Thus, although MYC is described as one of the most important oncogenes, it is important to realize that there is an extensive, multilayered ncRNA network around MYC, in which intricate crosstalk contributes to hallmarks of cancer.