The MYC mRNA 3′‐UTR couples RNA polymerase II function to glutamine and ribonucleotide levels

Abstract

Deregulated expression of MYC enhances glutamine utilization and renders cell survival dependent on glutamine, inducing “glutamine addiction”. Surprisingly, colon cancer cells that express high levels of MYC due to WNT pathway mutations are not glutamine‐addicted but undergo a reversible cell cycle arrest upon glutamine deprivation. We show here that glutamine deprivation suppresses translation of endogenous MYC via the 3′‐UTR of the MYC mRNA, enabling escape from apoptosis. This regulation is mediated by glutamine‐dependent changes in adenosine‐nucleotide levels. Glutamine deprivation causes a global reduction in promoter association of RNA polymerase II (RNAPII) and slows transcriptional elongation. While activation of MYC restores binding of MYC and RNAPII function on most promoters, restoration of elongation is imperfect and activation of MYC in the absence of glutamine causes stalling of RNAPII on multiple genes, correlating with R‐loop formation. Stalling of RNAPII and R‐loop formation can cause DNA damage, arguing that the MYC 3′‐UTR is critical for maintaining genome stability when ribonucleotide levels are low.

Introduction

Deregulated and enhanced expression of one of three MYC genes (MYC, MYCN, and MYCL) is observed in multiple tumor entities, and inhibition of MYC function can selectively eradicate tumor cells (Soucek et al, 2008; Dang, 2012; Annibali et al, 2014). MYC proteins are transcription factors that broadly bind to promoters and enhancers with an open chromatin structure (Wolf et al, 2014; Kress et al, 2015). Enhanced expression of MYC proteins promotes cell growth and cell proliferation in tissue culture and in vivo, suggesting that promoting both processes is a critical oncogenic function of MYC (Barna et al, 2008; Dang, 2012).

To promote growth and proliferation, MYC proteins exert profound effects on many aspects of cell metabolism (Carroll et al, 2015). MYC regulates multiple genes involved in glucose utilization and thereby increases the aerobic use of glucose and its conversion into lactate (Shim et al, 1997; Doherty et al, 2014). Similarly, MYC increases uptake of glutamine and its flux into downstream metabolic pathways via regulation of genes encoding glutamine transporters (Wise et al, 2008), enzymes of glutamine metabolism (Eberhardy & Farnham, 2001; Bott et al, 2015), and microRNAs that control expression of these genes (Gao et al, 2009). Collectively, these changes promote the biosynthesis of macromolecules and enable cells to avoid apoptosis (Hsu & Sabatini, 2008).

These observations have led to the hypothesis that targeting alterations in cell metabolism is a potentially universal strategy to selectively eradicate MYC‐driven tumor cells (Dang, 2011; Gouw et al, 2016). This notion is supported by observations that MYC‐driven tumor cells can be “glutamine‐addicted”, a term that describes the dependency of MYC‐driven tumor cells on an external supply of glutamine for survival (Wise et al, 2008). Glutamine addiction of MYC‐driven tumor cells has been observed in multiple experimental systems, both in tissue culture and in vivo (Yuneva et al, 2007, 2012; Gao et al, 2009; Qing et al, 2012; Xiang et al, 2015). The concept has spurred the development of glutaminase inhibitors for tumor therapy. One of these inhibitors is in current clinical trials, using amplification of MYC as a marker for patient stratification (ClinicalTrials.gov: NCT02071862).

Somewhat paradoxically, a parallel series of experiments has demonstrated that expression of MYC is sensitive to stress, including metabolic stress, in many experimental systems. For example, translation of MYC is regulated by internal ribosomal entry sites that are present in the 5′‐UTR (Stoneley et al, 2000; Wiegering et al, 2015). Conversely, the 3′‐UTR of MYC is targeted by a large number of microRNAs, including LET‐7 (Kim et al, 2009), the miR‐34 family (Cannell et al, 2010; Christoffersen et al, 2010; Kress et al, 2011), miR‐24 (Lal et al, 2009), miR‐17‐19b (Mihailovich et al, 2015), and others. Expression of these miRNAs (in particular the miR‐34 family) is regulated by FoxO and p53 transcription factors, coupling endogenous MYC expression to the cellular energy status and stress‐dependent signals (He et al, 2007; Hermeking, 2007; Cannell et al, 2010; Kress et al, 2011; Zheng et al, 2011).

Many of the experimental strategies used to demonstrate glutamine addiction use ectopic expression of MYC as model, and most transgenes used do not contain the 3′‐UTR. However, the 3′‐UTR is not mutated in human tumors, suggesting that metabolic controls that impinge on MYC expression via the 3′‐UTR may have been overlooked. To address this question, we studied the glutamine dependence of colon carcinoma cells, since colon carcinomas universally express high levels of MYC due to mutations in the WNT signaling pathway (He et al, 1998; van de Wetering et al, 2002; Sansom et al, 2007).

Results

Glutamine deprivation induces a reversible S‐phase arrest in colon carcinoma cells

In order to test whether colon cancer cells depend on an external supply of glutamine for survival, we cultured HCT116 carcinoma cells, which express high levels of MYC due to an activating point mutation in the beta‐catenin gene (Polakis, 1999), in complete medium, in medium lacking glutamine, or in medium lacking glucose. Deprivation of either glutamine or glucose suppressed growth of HCT116 to a varying extent (Fig 1A). FACS analyses of annexin V/propidium iodide‐stained cells after 36 h of nutrient deprivation showed that depletion of glucose led to a large increase in both early and late apoptotic cells, whereas deprivation of glutamine led to only a moderate increase in apoptosis (Fig 1B). Parallel FACS analyses of cells labeled with BrdU after 15 h of glutamine deprivation provided little evidence for apoptosis but instead showed a cell cycle arrest predominantly in late S‐phase (Fig 1C). Re‐addition of glutamine restored BrdU incorporation and promoted rapid cell cycle progression (Fig 1C and D). We concluded that glutamine deprivation suppresses growth of HCT116 cells primarily via inducing a reversible cell cycle arrest.

Figure 1. Effects of glutamine starvation on growth and MYC levels in HCT116 cells.

1. Effects of glucose (Glc) or glutamine (Q) deprivation on cell growth. Colonies were stained with crystal violet after starvation for 36 h. Quantification of three independent experiments is shown on the right. Results represent mean + SD. P‐values were calculated using Student's t‐test.
2. FACS analysis showing the percentage of early apoptotic (AnnV+; Ann: annexin) and late apoptotic (AnnV+/PI+; PI: propidium iodide) HCT116 cells after 36 h of starvation. Results represent mean + SD (n = 3). P‐values were calculated for all apoptotic cells using Student's t‐test.
3. FACS analysis of BrdU/PI‐stained cells. Cells were grown in complete medium or medium without glutamine for 15 h. Where indicated, glutamine was re‐added for 8 h. Cells were labeled with BrdU for 1 h.
4. Quantification of BrdU‐positive cells. Results represent mean + SD (n = 4). P‐values were calculated using Student's t‐test.
5. Top: Immunoblot documenting MYC protein levels in HCT116 cells cultured in medium containing the indicated amounts of glutamine for 24 h (n = 3). Bottom: Quantification of three independent experiments. Each point represents mean ± SD. Curve fitting was performed using a sigmoidal dose–response equation.
6. Time course of MYC protein levels in cells starved for 15 h, followed by re‐addition of glutamine at the indicated time points. Each value was normalized to +Q at time point 0. Results represent mean ± SD (n = 4).

HCT116 express approximately 300,000–600,000 molecules of MYC per cell, which is about 10‐ to 20‐fold higher than levels expressed in several non‐transformed proliferating cell lines (Lorenzin et al, 2016). The comparison with U2OS cells expressing doxycycline‐inducible alleles of MYC showed that these levels are sufficient to suppress colony formation and induce apoptosis in U2OS cells (Walz et al, 2014; Lorenzin et al, 2016). To understand why HCT116 cells undergo a reversible cell cycle arrest upon glutamine deprivation, we determined endogenous MYC levels by immunoblotting. These experiments showed that glutamine deprivation suppressed expression of MYC (Fig 1E). Titration of glutamine showed that the EC₅₀ value for MYC was approximately 0.5 mM, which is slightly below glutamine concentrations in human plasma (0.6–0.8 mM; Felig et al, 1970), suggesting that glutamine availability in tissues is a physiologically relevant regulator of MYC expression (Fig 1E). In contrast, deprivation of glucose did not significantly alter MYC expression (Appendix Fig S1A). Re‐addition of glutamine after 15 h of deprivation rapidly restored MYC expression (t _1/2 = 13.5 min; Fig 1F). Similarly, glutamine deprivation suppressed MYC expression and cell growth in four additional colon carcinoma cell lines (Appendix Fig S1B). We concluded that expression of MYC is tightly and rapidly regulated by extracellular glutamine levels in colon carcinoma cells.

Glutamine regulates MYC expression via the 3′‐UTR

The activating mutation in beta‐catenin leads to deregulated and constitutive expression of MYC mRNA in HCT116 cells (He et al, 1998). In response to glutamine deprivation, we observed a moderate further increase in MYC mRNA levels, which is due to a well‐known negative feedback loop, in which MYC proteins autosuppress transcription of MYC (Penn et al, 1990; Appendix Fig S1B and C). MYC proteins are rapidly turned over by the ubiquitin–proteasome system, and one of the critical ubiquitin ligases, FBXW7, recognizes a phosphodegron that is phosphorylated by MAP kinases and GSK3, suggesting that glutamine‐dependent changes in growth factor‐dependent signaling may regulate MYC via FBXW7 (Sears et al, 2000; Welcker et al, 2004). However, determination of MYC levels after addition of cycloheximide in medium containing either 2 mM or 0.5 mM glutamine did not reveal a change in MYC stability (Fig EV1A; note that MYC levels in the complete absence of glutamine were too low to reliably determine stability). Furthermore, the abundance of other short‐lived proteins such as c‐JUN or MCL1 that are turned over by FBXW7 and carry a similar phosphodegron did not alter in response to glutamine deprivation (Fig EV1B). Finally, the low levels of MYC in the absence of glutamine were not significantly elevated by addition of the proteasome inhibitor, MG132 (Fig EV1C). In contrast, pulse‐labeling using ³⁵S‐methionine showed that glutamine deprivation essentially shut down translation of MYC and that re‐addition of glutamine restored translation (Fig 2A). We concluded that translation of MYC is controlled by glutamine levels.

Figure EV1. Cellular responses to glutamine starvation and regulation of MYC .

1. Quantification of immunoblots documenting MYC levels at the indicated times after addition of cycloheximide (CHX) to HCT116 cells cultured in medium containing 2 mM or 0.5 mM glutamine for 24 h. Each point shows mean ± SD (n = 3).
2. Immunoblots documenting levels of c‐JUN, MCL‐1, and cyclin E before and after glutamine withdrawal for 1 h.
3. Immunoblot documenting MYC levels in the presence of the proteasomal inhibitor MG132. Cells were cultured for 15 h in medium containing or lacking glutamine and treated with 20 μM MG132 or EtOH for 6 h.
4. Immunoblots documenting levels of the indicated proteins and phosphoproteins in HCT116 cells pre‐treated with rapamycin (Rap) or DMSO for 2 h and starved for 15 h in the presence of rapamycin. Glutamine was re‐added after starvation for the indicated time points (n = 2).
5. The experiment was performed as described in (C) using the inhibitor OSI‐027.
6. Immunoblots showing levels of the indicated proteins and phosphoproteins in cells cultured for 15 h in the presence or absence of the indicated amino acids (M = methionine; L = leucine). Each amino acid was re‐added for 2 h after starvation (−/+) (n = 3).
7. Immunoblots documenting levels of the indicated proteins and phosphoproteins in wild‐type (p53^+/+) or p53‐deficient (p53^−/−) HCT116 cells cultured in medium containing the indicated amounts of glutamine for 15 h (n = 2).

Source data are available online for this figure.

Figure 2. The 3′‐ UTR of MYC mediates the response to glutamine.

1. ³⁵S‐methionine labeling of HCT116 followed by immunoprecipitation with α‐MYC or control antibodies. Glutamine starvation was performed for 2 or 15 h. Where indicated, adenosine (Ado, 150 μM) was supplemented or re‐added to the medium after starvation. The arrow indicates the specific MYC band detected by autoradiography (n = 2).
2. Diagram depicting the doxycycline (Dox)‐inducible MYC constructs used for lentiviral infection of HCT116 cells. All constructs contain a carboxy‐terminal HA‐tag. CDS, coding sequence.
3. Immunoblot documenting levels of ectopically expressed MYC. Glutamine starvation was started after 2 h of Dox‐mediated induction and maintained for 15 h in the presence of Dox. Exogenous MYC levels were detected by α‐HA immunoblot (n = 3).
4. Top: Immunoblot showing levels of exogenous MYC. The experiment was performed as in (C), using Dox‐inducible MYC constructs (n = 3). CDS and 3′‐UTR constructs are described in (A). Four 3′‐UTR deletions are depicted at the bottom. Numbers refer to the annotated MYC sequence NM_002467.4. Vertical dashed lines indicate the region responsible for MYC regulation. Black boxes indicate the position of the AU‐rich elements (AUUUA sequences) present in MYC 3′‐UTR. Gray dashes indicate the predicted binding sites of miRNAs (shown in Appendix Fig S3).

Glutamine activates mTORC1, which is a potent activator of translation of MYC, suggesting that activation of mTORC1 may account for the regulation of MYC levels by glutamine (West et al, 1998; Nicklin et al, 2009; Duran et al, 2012). To test this hypothesis, we depleted and re‐added glutamine to HCT116 as before in the presence of the mTORC1 inhibitor rapamycin, and in the presence of OSI‐027, an inhibitor of both mTORC1 and mTORC2 (Bhagwat et al, 2011). Consistent with the published data, glutamine deprivation decreased mTORC1 activity as witnessed by a reduction in phosphorylation of 4EBP1 at T70 and of p70S6 kinase at T389; re‐addition of glutamine strongly stimulated phosphorylation (Fig EV1D and E). Addition of rapamycin blunted phosphorylation of 4EBP1 and attenuated phosphorylation of S6 kinase at T389, demonstrating that it is effective in inhibiting mTORC1. Importantly, rapamycin did not affect downregulation of MYC levels upon glutamine deprivation and only moderately attenuated induction upon re‐addition of glutamine (Fig EV1D). Identical results were obtained with OSI‐027 (Fig EV1E). Starvation for two other amino acids, methionine and leucine, also reduced MYC levels and TORC activity, consistent with known effects of amino acid availability on mTORC1 (Fig EV1F; Nicklin et al, 2009). Importantly, re‐addition of either amino acid to depleted cells restored phosphorylation of S6K by mTORC1, but, unlike glutamine, had no rapid effect on MYC levels (Fig EV1F). Finally, neither siRNA‐mediated depletion of raptor or rictor abolished regulation of MYC levels in response to glutamine starvation (Appendix Fig S2). We concluded that changes in mTORC activity do not account for regulation of MYC by glutamine. We also speculated that stress due to glutamine deprivation might induce p53 to regulate MYC expression, but experiments using P53‐deficient HCT116 cells established that p53 is not required for regulation of MYC levels by glutamine (Fig EV1G).

Multiple stress‐response pathways impinge on MYC via the 5′‐ and 3′‐UTR (see Introduction), prompting us to map the elements that respond to glutamine deprivation on the MYC mRNA. To do so, we stably expressed doxycycline‐inducible alleles of HA‐tagged MYC from constructs that encompass either the entire MYC mRNA, only the coding sequence or the coding sequence and the 5′‐UTR or the 3′‐UTR, respectively, using lentiviral transduction (Fig 2B). In order not to alter cell physiology or compete out any possible repressive factors, we chose constructs that express low levels of MYC relative to endogenous levels for these experiments. Inclusion of the 5′‐UTR led to the pronounced appearance of a second band with reduced mobility in MYC immunoblots, but not to downregulation of MYC in glutamine‐starved cells (Fig 2C). Previous work has established that a CUG codon upstream of the canonical AUG codon in the 5′‐UTR of MYC can initiate translation (Hann et al, 1988). Mutation of the CUG confirmed that glutamine deprivation stimulated the use of this codon (Fig EV2A). In contrast, inclusion of the 3′‐UTR led to robust downregulation of MYC protein in the absence of glutamine (Figs 2C and EV2B). A deletion series showed that the glutamine‐responsive region is between nucleotides 2,127 and 2,366 of the MYC mRNA (Fig 2D). qRT–PCR assays showed a minor decrease in steady‐state mRNA levels of constructs containing the 3′‐UTR in the absence of glutamine, suggesting that changes in mRNA stability contribute to the effect on MYC translation (Fig EV2C; note that the constructs used in this experiment are not subject to MYC‐dependent autosuppression). We concluded that glutamine deprivation inhibits MYC translation via the 3′‐UTR.

Figure EV2. The 3′‐ UTR of MYC mediates the response to glutamine.

1. Top: Scheme of the constructs used. Bottom: Immunoblot from HCT116 cells expressing the indicated MYC constructs and treated as shown.
2. Quantification of exogenous MYC levels. The graphs show mean expression + SD of each construct relative to expression in the presence of glutamine (n = 3). P‐values were calculated using Student's t‐test.
3. qRT–PCR assays showing levels of HA‐tagged MYC mRNA in Dox‐induced cells. HA‐MYC mRNA in the presence or absence of glutamine is each shown relative to CDS values. Results represent mean + SD (n = 3). P‐values were calculated using Student's t‐test.

The 3′‐UTR of MYC protects tumor cells from glutamine addiction

It is possible that the low level of apoptosis observed in HCT116 cells upon glutamine deprivation is due to the presence of anti‐apoptotic mutations in these tumor cells rather than to downregulation of MYC. To test this possibility, we stably expressed a MYC‐ER protein in HCT116 cells and measured cell growth and apoptosis in response to glutamine deprivation (Fig 3A). Addition of 4‐hydroxytamoxifen to activate MYC‐ER strongly suppressed colony formation (Fig 3B and C) and induced apoptosis (Fig 3D) upon glutamine deprivation. Both observations recapitulate previous work (Yuneva et al, 2007). To test the effect of the 3′‐UTR, we expressed a MYC‐ER construct that contains the 3′‐UTR. As shown above for wild‐type MYC, expression of MYC‐ER‐3′‐UTR was attenuated in the absence of glutamine (Fig 3A). This led to enhanced colony formation of glutamine‐starved cells (Fig 3B and C). In addition, inclusion of the 3′‐UTR reduced apoptosis upon glutamine deprivation, arguing that this contributes to the restoration of colony formation in cells expressing MYC‐ER‐3′‐UTR (Fig 3D). We concluded that the 3′‐UTR of MYC suppresses glutamine addiction of colon cancer cells.

Figure 3. Role of the 3′‐UTR in cellular responses to glutamine deprivation.

1. Immunoblot of HCT116 cells stably expressing the indicated MYC‐ER constructs or empty vector (EV) as control. MYC‐ER activity was induced overnight with 20 nM 4‐hydroxytamoxifen (OHT), followed by 24 h of glutamine starvation in the presence of OHT (n = 4).
2. Crystal‐violet‐stained tissue culture plates documenting growth of HCT116 cells cultured in glutamine‐depleted medium for 48 h in the presence of OHT.
3. Quantification of relative cell number at the indicated time points after glutamine starvation. Each sample was normalized to the value obtained after 48 h of starvation. Each point represents the mean ± SD (n = 3). P‐values were calculated using Student's t‐test, comparing EV or MYC‐ER‐3′‐UTR to MYC‐ER.
4. PI‐FACS documenting number of HCT116 cells with a sub‐G1 DNA content after 72 h of glutamine starvation. Results show mean + SD (n = 3). P‐values were calculated using Student's t‐test.

The 3′‐UTR of MYC responds to cellular adenosine levels

A major function of glutamine is to supply intermediates for the TCA cycle after its conversion to glutamate and α‐ketoglutarate (Fig EV3A). Consistently, glutamine deprivation led to a rapid decrease in steady‐state levels of glutamate and the TCA cycle intermediates α‐ketoglutarate, fumarate, and malate (Fig 4A). In addition, glutamine deprivation reduced steady‐state levels of ornithine and putrescine, indicating an overall reduction in nitrogen metabolism (Fig EV3B). In contrast, glutamine starvation had little effect on steady‐state levels of several intermediates of glycolysis (Fig 4A). Glutamine also contributes to redox homeostasis, since glutamate is incorporated into glutathione and since glutamate dehydrogenase contributes to the pool of reducing equivalents (Fig EV3A), and we observed an increase in levels of reactive oxygen species upon glutamine deprivation (Fig EV3C). Finally, glutamine donates amino groups for the synthesis of pyrimidine and purine nucleotides and levels of multiple nucleotides decreased after glutamine deprivation (Fig 4B).

Figure EV3. Characterization of metabolic effects of glutamine starvation.

1. Diagram summarizing main glutamine‐dependent metabolic pathways. Filled dots indicate the metabolites used in this study.
2. Relative concentrations of putrescine and ornithine after 15 h of glutamine starvation in HCT116 cells, normalized to +Q. Data represent mean ratios + SD (n = 6, each measurement carried out in duplicates). P‐values were calculated using Student's t‐test (***P < 0.001; **P < 0.01).
3. FACS assays of 2′,7′‐dichlorofluorescin diacetate (DCFDA)‐stained HCT116 cells documenting levels of reactive oxygen species in control cells and after 15 h of glutamine starvation. The right panel shows the quantification of a representative experiment (n = 2).
4. Immunoblot of glutamine‐starved cells (15 h) after addition of the indicated substrates, either alone or in combination, for 2 h: Nucls: nucleosides; GSH‐MEE: glutathione‐reduced ethyl ester; DMKG: dimethyl‐α‐ketoglutarate; DMS: dimethyl succinate; Pyr: sodium pyruvate; Glu: glutamate; Q: glutamine
5. (Left) α‐HA immunoblot of HCT116 cells expressing the indicated Dox‐inducible constructs and cultured with or without Ado in glutamine‐depleted medium for 2 h (n = 3). (Right) α‐HA immunoblot of cells that were glutamine‐starved for 15 h, followed by re‐addition of the indicated substrates for 2 h (n = 2).

Source data are available online for this figure.

Figure 4. Glutamine‐dependent metabolites affecting MYC expression.

1. Concentrations of selected metabolites after 2 and 15 h of glutamine starvation in HCT116 cells relative to non‐starved cells. Glu: glutamate; α‐KG: α‐ketoglutarate; Fum: fumarate; Mal: malate; DHAP: dihydroxyacetone phosphate; Pyr: pyruvate; Lac: lactate. Data show mean ratios + SD (n = 6, each measurement performed in technical duplicates). P‐values were calculated using Student's t‐test (***P < 0.001; **P < 0.01; *P < 0.05).
2. Concentrations of nucleotides in HCT116 cells that were glutamine‐starved for 15 h, followed by re‐addition of the indicated substrates for 2 h. DMKG: dimethyl‐α‐ketoglutarate; Ado: adenosine. Values were normalized to +Q. Data represent mean ratios + SD (n = 6, each measurement carried out in triplicates). P‐values were calculated using Student's t‐test (***P < 0.001; **P < 0.01; *P < 0.05).
3. Immunoblot of glutamine‐starved HCT116 cells after addition of the indicated substrates. DMS: dimethyl succinate; NAC: N‐acetyl‐cysteine; Nucls: mix of five ribo‐ and deoxy‐ribonucleosides (n = 2).
4. Immunoblot of glutamine‐starved cells after addition of the indicated substrates. Ado: adenosine; Guo: guanosine; Urd: uridine; Cyt: cytidine; Pur: adenosine and guanosine; Pyri: cytidine and uridine; dAdo: deoxy‐adenosine; dGuo: deoxy‐guanosine; Thd: thymidine; dCyd: deoxy‐cytidine; dPur: deoxy‐adenosine and deoxy‐guanosine; dPyri: thymidine and deoxy‐cytidine (n = 3).
5. Concentrations of adenosine‐derived nucleotides after 2 h of glutamine starvation in HCT116 cells, relative to non‐starved cells. Data represent mean ratios + SD (n = 6, each measurement carried out in triplicates). P‐values were calculated using Student's t‐test (**P < 0.01; *P < 0.05).
6. Numbers of HCT116 cells grown for 48 h in media supplemented with the indicated substrates. Cell numbers were determined at 0 and 48 h and shown relative to 0 h. Results represent mean + SD (n = 5, each measurement carried out in duplicates). P‐values were calculated using Student's t‐test (***P < 0.001; **P < 0.01).

To determine which—if any—of these changes mediates regulation of MYC, we performed re‐addition experiments of glutamine‐derived metabolites. Addition of N‐acetyl‐cysteine, glutamate, and membrane‐permeable forms of α‐ketoglutarate, succinate, and pyruvate, either alone or in combination, had little effect on MYC protein levels in glutamine‐deprived cells (Figs 4C and EV3D). In contrast, addition of a mixture of four ribonucleosides and thymidine largely restored MYC protein expression (Fig 4C). Furthermore, both a mixture of purine nucleosides and adenosine, but neither pyridimine nucleosides nor any desoxy‐nucleoside, restored expression of MYC in glutamine‐deprived cells (Fig 4D). Measurements of nucleotide pools showed that addition of glutamine or adenosine, but not of methylated α‐ketoglutarate, reverted the decrease in cellular adenosine‐nucleotide levels caused by glutamine deprivation (Fig 4B). Adenosine addition had no significant effect on cellular levels of other ribonucleotides, but led to increased levels of several desoxy‐ribonucleotides. Consistent with these observations, a MYC construct that comprises only the coding sequence did not respond to re‐addition of adenosine to glutamine‐deprived cells, whereas expression of MYC from a construct comprising the 3′‐UTR was stimulated by addition of adenosine (Fig EV3E). In line with this, addition of adenosine partially restored translation of MYC in glutamine‐deprived cells (Fig 2A). The role of glutamine as donor of amino groups in nucleotide biosynthesis does not require glutaminase; consistently, inhibition of glutaminase by three different inhibitors had no effect on MYC levels (Fig EV4A). We confirmed that two of the inhibitors, BPTES (bis‐2‐(5‐phenylacetamido‐1,3,4‐thiadiazol‐2‐yl)ethyl sulfide) and CB839, decreased label incorporation from ¹³C‐glutamine into glutamate and the TCA intermediates α‐ketoglutarate and malate (Fig EV4B) and that inhibition of glutaminase had no significant effects on levels of adenosine nucleotides (Fig EV4C). In contrast, steady‐state levels of adenosine nucleotides declined within 2 h after glutamine deprivation, paralleling the decrease in MYC levels (Fig 4E). Finally, we found that addition of adenosine to glutamine‐deprived cells restored MYC expression, but not cell proliferation, arguing that the increase in MYC is not an indirect consequence of changes in cell proliferation (Fig 4F). We concluded that the MYC 3′‐UTR senses cellular adenosine levels to regulate MYC expression.

Figure EV4. Effect of glutaminase inhibitors on MYC levels.

1. Immunoblot showing MYC levels in glutamine‐starved HCT116 cells or treated with 10 μM of the indicated inhibitors for 15 h (n = 2).
2. Relative ¹³C incorporation into the indicated metabolites 45 min after addition of ¹³C‐glutamine to starved HCT116 cells in the presence of BPTES (10 μM), CB839 (10 μM), or DMSO as solvent control.
3. Concentrations of adenosine‐derived nucleotides in cells that were glutamine‐starved for 15 h, followed by re‐addition of the indicated inhibitors for 2 h, relative to non‐starved cells.

Data information: Data in (B, C) represent mean ratios + SD (n = 6, each measurement carried out in triplicates). P‐values were calculated using Student's t‐test (***P < 0.001; **P < 0.01; *P < 0.05).

To understand how glutamine and adenosine levels signal to the 3′‐UTR, we profiled expression of microRNAs by RNA sequencing. HCT116 expressed several microRNAs that target the MYC 3′‐UTR at significant levels, but glutamine deprivation did not alter expression of any of them (Appendix Table S1). Furthermore, anti‐miRNAs targeting 11 of the most strongly expressed miRNAs enhanced MYC protein levels but did not affect regulation of MYC in response to glutamine deprivation (Appendix Fig S3), suggesting that regulation is independent of microRNAs.

Glutamine deprivation has global and MYC‐dependent effects on RNAPII function

To understand the physiological relevance of glutamine‐dependent regulation of MYC, we initially examined whether activation of MYC in the absence of glutamine depletes critical metabolite pools. However, measurements of multiple metabolite concentrations and reactive oxygen species after 15 h of glutamine deprivation provided no experimental support for this model (Fig EV5A and B). Activation of MYC can have global effects on both promoter binding and elongation by RNA polymerase II (RNAPII) and, in some situations, globally enhances RNAPII function (Rahl et al, 2010; Lin et al, 2012; Nie et al, 2012; Walz et al, 2014). To test how regulation of MYC levels by glutamine affects RNAPII function, we measured association of RNAPII with the promoters and of the elongating form of RNAPII, (pSer2 RNAPII) with the transcription end sites (TES) of two highly transcribed genes, ACTB (encoding actin B) and NCL (encoding nucleolin; Fig 5A; control immunoblots for MYC and RNAPII are shown in Fig EV5C). Both genes showed decreased RNAPII binding to the promoter and association of elongating RNAPII with the TES upon glutamine deprivation. Both effects were largely reverted by activation of MYC‐ER. To obtain a global view of RNAPII function, we performed ChIP sequencing under these experimental conditions (Fig EV5C). Inspection of multiple individual traces (Fig 5B) and global analyses (Figs 5C and EV5D) showed that glutamine deprivation globally reduced association of RNAPII with promoters and of elongating RNAPII with the TES. Both decreases were partially reverted upon activation of MYC‐ER. Quantitative analyses showed that the decrease in RNAPII binding upon glutamine deprivation and the restoration by MYC were uniform for all promoters where RNAPII binding could be detected (Fig 5D). Similarly, the effect of MYC on Ser2‐RNAPII occupancy at the TES was uniform across all active genes (Fig 5D). Addition of adenosine to glutamine‐deprived cells restored binding of RNAPII to three promoters that we tested (Fig EV5E), consistent with the effects on MYC levels. In contrast, effects on elongation by RNAPII were variable, most likely due to decreased and varying levels of other ribonucleotides (F. R. Dejure, unpublished observation).

Figure EV5. Effects of MYC‐ER on metabolism and RNAPII binding.

1. Relative concentrations of the indicated metabolites in glutamine‐starved HCT116 cells. MYC‐ER activity was induced for 8 h with 100 nM 4‐hydroxytamoxifen (OHT), followed by 15 h of glutamine starvation in the presence of OHT. Values were normalized to EV. Data represent mean ratios ± SD (n = 6 each, nucleotide measurements carried out in triplicates, central carbon metabolites in duplicates).
2. FACS assays of 2′,7′‐dichlorofluorescin diacetate (DCFDA)‐stained cells. HCT116 cells were treated with 100 nM OHT and then glutamine‐starved for 24 h, in the presence of OHT (n = 2).
3. Immunoblot showing levels of MYC, RNAPII and pSer2 RNAPII under the conditions used to perform ChIP sequencing (n = 3).
4. Heat map documenting binding of RNAPII at the TSS and of pSer2 RNAPII at the TES for all genes with detectable RNAPII binding under the indicated conditions.
5. Left: ChIP experiment documenting RNAPII binding to the TSS of the indicated genes and an intergenic control region. Bars represent mean + SD of technical triplicates. Right: Immunoblot documenting MYC levels under the experimental conditions used to perform ChIP experiment. Cells were cultured in presence or absence of glutamine for 3.5 h. Where indicated, 300 μM of Ado was supplemented to the medium.

Figure 5. Effects of glutamine starvation on RNAPII function.

1. ChIP experiment documenting RNAPII binding to the transcription start site (TSS) and transcription end site (TES) of the ACTB and NCL genes and an intergenic control region (same for both targets). MYC‐ER activity was induced overnight with 100 nM OHT. Cells were starved for 5 h in the presence of OHT. Bars represent mean + SD of technical triplicates.
2. ChIP‐sequencing tracks for PTMA and LDHA from HCT116 cells treated as in panel (A).
3. Metagene plots of RNAPII (left) or pSer2 RNAPII (right) for all genes with detectable RNAPII binding under the indicated conditions.
4. Density plots comparing occupancy by RNAPII at the TSS and by pSer2 RNAPII at the TES for all genes with detectable RNAPII binding under the indicated conditions.
5. Heat map showing promoter occupancy by RNAPII and by MYC for all promoters (n = 9,796) at which RNAPII binding is detectable.

Additional ChIP‐sequencing experiments showed that MYC binding was detectable at virtually all promoters transcribed by RNAPII and that—as observed in other cell types—binding of MYC closely paralleled that of RNAPII, arguing that MYC generally occupies active promoters (Fig 5E). Furthermore, the magnitude of the decrease in RNAPII occupancy upon glutamine deprivation and of the restoration by activation of MYC paralleled MYC occupancy (Appendix Fig S4A and B). Previous works had shown that MYC can globally affect both RNAPII binding to promoters and elongation (Rahl et al, 2010; Lin et al, 2012; Nie et al, 2012; Walz et al, 2014). The easiest interpretation of these findings is, therefore, that changes in MYC levels directly and globally couple RNAPII function to ribonucleotide and adenosine availability.

Glutamine deprivation alters transcriptional elongation

To ascertain whether RNAPII functions normally in the absence of glutamine, we determined the change in RNAPII association with promoters versus the change in RNAPII association in the body of the gene for all transcribed genes. Notably, glutamine withdrawal caused a large decrease in RNAPII association with promoters, but a smaller decrease in RNAPII association with gene bodies (Fig 6A). Consistently, glutamine starvation led to a decrease in the traveling ratio of RNAPII, which is the ratio of RNAPII bound at promoters relative to RNAPII bound in gene bodies, suggesting that glutamine withdrawal decreases pause‐release or slows down transcriptional elongation (Fig 6B).

Figure 6. Analysis of transcriptional elongation by RNAPII and R‐loop formation.

1. Kernel density plot documenting the effect of glutamine starvation on RNAPII occupancy at the promoter and in the gene body for all genes with detectable RNAPII binding (Δ: difference between the indicated conditions). Outliers are shown as black dots.
2. Traveling ratio of RNAPII under the indicated experimental conditions.
3. ChIP‐sequencing track showing MYC‐ER‐mediated RNAPII accumulation in the gene body (GB) of TKT. See panel (D) for position of primers.
4. ChIP experiment documenting RNAPII occupancy at the TSS, an intragenic site, and the TES of the TKT gene upon MYC‐ER activation. Description as in Fig 5A. Bars represent mean + SD of technical triplicates (n = 2).
5. DNA–RNA immunoprecipitation (DIP) showing the presence of R‐loops in the TKT gene. HCT116 cells were starved for 5 h. DNA was extracted, treated or not with RNase, and immunoprecipitated with S9.6 antibody or IgG as control. Bars represent mean ± SD of technical triplicates.
6. DIP assays showing R‐loops under the indicated experimental conditions. MYC‐ER‐ or EV‐expressing cells were induced overnight with 100 nM OHT and then starved for 24 h in the presence of OHT. Bars represent mean±SD of technical triplicates (n = 2)

Activation of MYC‐ER in the absence of glutamine partially reverted the decrease in the traveling ratio, consistent with multiple previous observations that MYC promotes pause‐release and elongation and transfers elongation factors onto RNAPII (Fig 6B; Rahl et al, 2010; Jaenicke et al, 2016). However, the effect of MYC was not uniform for all genes: RNAPII association with the gene body increased on 176 genes upon MYC activation in the absence of glutamine to levels higher than in control cells grown in the presence of glutamine although levels of RNAPII at the promoter remained lower than in glutamine‐replete cells, suggesting that activation of MYC in the absence of glutamine promotes stalling of RNAPII on these genes. The strongest accumulation of RNAPII in the gene body was observed in the TKT gene, which encodes transketolase, an enzyme of the pentose phosphate pathway (Fig 6C and D).

Defects in elongation prolong both the proximity of nascent RNA with RNAPII and transcription‐dependent supercoiling of DNA, and both factors promote the formation of R‐loops (Costantino & Koshland, 2015). To test whether stalling of RNAPII correlates with R‐loop formation, we immunoprecipitated chromatin using a monoclonal antibody (S9.6) that recognizes RNA/DNA hybrids and—as control—treated an aliquot of the immunoprecipitates with RNaseH1 (Skourti‐Stathaki et al, 2011). We selected multiple primers that cover the TKT gene on which stalling was observed upon MYC activation (Fig 6D). This analysis showed that the region in which stalling was observed had a high propensity for R‐loop formation (Fig 6E). To determine the effect of glutamine starvation and MYC activation, we measured R‐loop accumulation in glutamine‐containing medium and in cells depleted of glutamine, both with and without activation of MYC‐ER. In these experiments, activation of MYC‐ER caused a significant accumulation of R‐loops in glutamine‐starved cells (Fig 6F). Taken together, our data show that activation of MYC‐ER in glutamine‐deprived cells leads to stalling of RNAPII and R‐loop formation.

Discussion

A large body of work has shown that expression of the MYC mRNA is responsive to external growth factors and to growth inhibitory factors such as TGF‐beta (Meyer & Penn, 2008). Similarly, recent work has documented that several stress‐responsive pathways and individual miRNAs have the potential to suppress MYC expression (see Introduction). The physiological significance of the latter regulation has not yet been firmly established. We show here, using the paradigm of glutamine starvation, that the 3′‐UTR couples MYC translation to the cellular metabolic status and specifically to the availability of adenosine nucleotides, which are the ribonucleotides that most strongly depend on glutamine. The MYC 3′‐UTR is not mutated in human tumors, suggesting that this regulatory mechanism is intact in tumor cells. This conclusion is supported by the analysis of several colon tumor cell lines. A plethora of miRNAs and RNA‐binding proteins bind to and regulate MYC via the 3′‐UTR. We did not obtain evidence for involvement of microRNAs in glutamine‐dependent MYC translation; hence, the detailed mechanism by which glutamine levels signal to the MYC 3′‐UTR remains to be determined. However, many nucleotide‐binding enzymes are known to moonlight as RNA‐interacting proteins (Castello et al, 2012, 2015). It is therefore tempting to speculate that a decrease in adenosine‐nucleotide levels changes the MYC mRNA interactome to regulate MYC translation.

MYC binds to essentially all open promoters in tumor cells (Sabo et al, 2014; Walz et al, 2014) and changes in MYC levels can globally enhance association of RNAPII with promoters (Walz et al, 2014) as well as pause‐release and elongation (Rahl et al, 2010; Walz et al, 2014; Jaenicke et al, 2016). These findings raise the possibility that glutamine regulation of MYC levels serves to globally couple RNAPII function to the availability of nucleotides. Consistent with this notion, glutamine deprivation elicited a global decrease in RNAPII promoter association and elongation at virtually all genes that we analyzed. Under these conditions, nucleotide levels decrease rapidly; hence, changes in RNAPII function might be secondary to altered ribonucleotide levels. Concentrations of ribonucleotide triphosphates in tumor cells are estimated to be between 400 μM (CTP) and 3,000 μM (ATP) (Traut, 1994) and thus significantly exceed the K _M values of RNAPII for transcriptional initiation (around 10 μM for ATP; Jiang & Gralla, 1995). We therefore consider it unlikely that the magnitude of the decrease in ribonucleotide triphosphate levels observed after glutamine deprivation has a significant direct impact on promoter association by RNAPII. Instead, we show that activation of MYC‐ER restores RNAPII binding to promoters to a large degree. The simplest explanation of our observations is therefore that the low MYC levels in glutamine‐starved cells limit RNAPII function, consistent with the notion that MYC levels can be rate‐limiting for transcription (Lin et al, 2012; Nie et al, 2012).

In glutamine‐deprived cells, RNAPII binding to promoter decreases strongly, but the decrease in RNAPII association with the gene body is more moderate, resulting in a decreased traveling ratio of RNAPII. During elongation, RNAPII oscillates between forward and backward movements in response to nucleotide misincorporation and overall elongation speed hence depends on absolute and relative nucleotide concentrations (Toulme et al, 1999). The decrease in traveling ratio indicates a reduced pause‐release or reduced elongation rate and possibly stalling of RNAPII. Activation of MYC in glutamine‐deprived cells partially reverts the global decrease in traveling ratio, but has no effect on ribonucleotide levels. Importantly, restoration of elongation by MYC is not perfect: Rather there is a significant increase in RNAPII stalling on approximately 200 genes. The strongest increase is observed on the TKT gene in a sequence that can form an R‐loop. R‐loops contain single‐stranded DNA, which is prone to base modifications and mutagenesis (Costantino & Koshland, 2015). Notably, glutamine deprivation also causes accumulation of reactive oxygen species, which are a significant cause of DNA damage at single‐stranded DNA. Finally, R‐loops, if unresolved, can cause replication/transcription conflicts arguing that even individual R‐loops can be a major source of DNA damage and genomic instability (Helmrich et al, 2013). Both replication stress and overt DNA damage contribute to induction of apoptosis by deregulated MYC (Reimann et al, 2007; Murga et al, 2011). We propose, therefore, that regulation of RNAPII function by glutamine via MYC protects cells from the deleterious consequences of global transcriptional elongation when nucleotide concentrations are low. Intriguingly, the identification of a consistent group of downstream target genes and pathways, by which MYC induces apoptosis, has remained difficult. Our data suggest an alternative model, in which the global stimulation of transcription elongation by deregulated MYC induces apoptosis since it generates aberrant transcription structures, such as R‐loops, in metabolically stressed cells (Fig 7). It remains to be determined whether this model can generally account for MYC‐dependent apoptosis during tumorigenesis.

Figure 7.

Model summarizing our findings

Exploiting glutamine addiction is a paradigm example of therapeutic strategies that rely on the pro‐apoptotic properties of deregulated MYC. Our data have two implications for these strategies: First, current approaches rely on inhibition of glutaminase and our data show that several glutaminase inhibitors have no effect on adenosine and MYC levels in colon cancers cells, arguing that targeting nucleotide synthesis may target MYC‐driven tumors more directly. Second, tumorigenesis in MYC‐driven transgenic mouse models is limited by MYC‐dependent apoptosis and greatly accelerated by inhibition of apoptosis provided by ectopic expression of Bcl2 family members of loss of the Arf and p53 tumor suppressor (Eischen et al, 1999; Finch et al, 2006; Wang et al, 2011), but whether deregulated MYC expression establishes a dependence on specific anti‐apoptotic mechanisms in human tumors is less clear. We show here that the 3′‐UTR enables colon tumor cells with deregulated MYC expression to escape from apoptosis upon metabolic stress, since it ensures that MYC protein expression remains stress‐responsive even though the promoter is constitutively active. Many currently used MYC transgenes, including those used to characterize pro‐apoptotic properties of deregulated MYC in vivo (Felsher & Bishop, 1999; Murphy et al, 2008; Wang et al, 2011), lack the 3′‐UTR. Since tumors frequently grow in hypoxic and nutrient‐deprived environments (Stine et al, 2015), such transgenes may overestimate MYC's pro‐apoptotic potential. Conversely, translation of endogenous MYC is driven from an internal ribosome entry site (IRES) in the 5′‐UTR during staurosporine‐ and TRAIL‐induced apoptosis (Stoneley et al, 2000), arguing that IRES‐driven translation of MYC may not be subject to repression by the 3′‐UTR. We propose, therefore, that therapeutic strategies that aim to exploit MYC's pro‐apoptotic potential for tumor therapy need to inhibit or bypass the pathways that suppress MYC expression in vivo to be successful.

Finally, we note that transketolase, the enzyme encoded by the TKT gene, in which RNAPII stalling is particularly strong and correlates with R‐loop formation, itself has a critical role in nucleotide biosynthesis, since TKT levels influence whether glucose that enters the pentose phosphate pathway is used for synthesis of nucleotides or for synthesis of reducing equivalents to counteract oxidative stress (Xu et al, 2016). R‐loops also have signaling functions independent from causing DNA damage (Tresini et al, 2015). It is possible, therefore, that RNAPII stalling and R‐loop formation also control metabolic enzyme levels or serve as a sensor during the metabolic adaptation to glutamine starvation (Fig 7).

Materials and Methods

Cell culture and reagents

All reagents were purchased from Sigma, unless otherwise indicated. All cell lines were obtained from ATCC. Cell lines were maintained in DMEM supplemented with 10% FBS (Biochrom) and 1% penicillin/streptomycin and validated using STR analysis. Cell lines were routinely tested for mycoplasma contamination. Experiments were performed using DMEM reconstituted with 2 mM glutamine, 2.5 g/l glucose, 0.015 g/l phenol red, 10% dialyzed FBS, and 1% penicillin/streptomycin, without sodium pyruvate, unless specific concentrations are indicated. For starvation experiments, cells were washed twice with PBS and cultured in reconstituted DMEM without glucose or glutamine. For single amino acids starvation (Q; M; L) medium lacking each amino acid was prepared in Earle's balanced salt solution supplemented with MEM vitamin solution (Life Technologies), 2.5 g/l glucose, 10% dialyzed FBS, 0.0001 g/l Fe(NO₃)₃ · 9H₂O, and the remaining amino acids according to the DMEM composition. Where specified, the following reagents were added at the indicated final concentrations: ¹³C‐glutamine (Cambridge Isotopes Laboratories Inc., 2 mM), dimethyl‐α‐KG (7 mM), dimethyl succinate (20 mM), sodium pyruvate (1 mM), N‐acetyl cysteine (5 mM), glutathione‐reduced ethyl ester (5 mM), nucleoside mix (Merck Millipore, 5×), glutamate (1.35 mM), nucleosides and deoxynucleosides (150 μM, except for thymidine, used at 50 μM), cycloheximide (100 μg/ml, dissolved in EtOH), MG132 (Calbiochem/Millipore, 20 μM, dissolved in DMSO), 4‐hydroxytamoxifen (100 nM, dissolved in EtOH), doxycycline (1 μg/ml, dissolved in EtOH). Rapamycin (LC Laboratories, 20 nM), OSI‐027 (Active Biochemicals, 20 μM), CB‐839 (Selleckchem, 10 μM), BPTES, and Compound 968 (Calbiochem/Millipore, 10 μM) were all dissolved in DMSO.

ChIP, ChIP sequencing, and DIP

ChIP experiments were performed as described previously (Walz et al, 2014). The DIP experiment was adapted from a published protocol (Weber et al, 2005). Cells were harvested, digested overnight in a buffer containing SDS and proteinase K, and sonified to random fragments with a size between 100 and 200 bp. Genomic DNA was purified by phenol–chloroform extraction and either treated with 20 U of RNaseH (NEB) overnight or not. Immunoprecipitation was performed using monoclonal antibody S9.6 and protein G‐ and protein A‐coupled Dynabeads without final de‐cross‐linking. Sequencing was performed using NextSeq 500 Illumina platform.

GC‐MS and direct infusion MS analysis

Metabolite extraction was carried out as described in Kempa et al (2009). Nucleotides were extracted as described in Lorkiewicz et al (2012). See Appendix Supplementary Methods for detailed protocols.

Statistical analyses

Statistical analysis was performed using Prism software (GraphPad), and R. Statistical significance was tested with heteroscedastic two‐tailed Student's t‐tests. Data are presented as mean, and error bars show standard deviation of the indicated number of biological replicates.

Data availability

ChIP‐sequencing datasets are available at the Gene Expression Omnibus under the accession number GEO: GSE86556.

Author contributions

FRD, NR, CPA, GM, JK, EW, and JTV performed the experiments, SW analyzed high‐throughput data, SH, AS, SK, and ME devised and supervised experiments, and SK and ME wrote the manuscript.

Conflict of interest

The authors declare that they have no conflict of interest.

Supporting information

Appendix

Expanded View Figures PDF

Source Data for Expanded View and Appendix

Review Process File

The EMBO Journal (2017) 36: 1854–1868

See also: CV Dang (July 2017)