Abstract
Tightly regulated activity of the transcription factor MYC is essential for orderly cell proliferation. Upon deregulation, MYC elicits and promotes cancer progression. Proteasomal degradation is an essential element of MYC regulation, initiated by phosphorylation at Serine62 (Ser62) of the MB1 region. Here, we found that Ser62 phosphorylation peaks in mitosis, but that a fraction of nonphosphorylated MYC binds to the microtubules of the mitotic spindle. Consequently, the microtubule-destabilizing drug vincristine decreases wild-type MYC stability, whereas phosphorylation-deficient MYC is more stable, contributing to vincristine resistance and induction of polyploidy. PI3K inhibition attenuates postmitotic MYC formation and augments the cytotoxic effect of vincristine.
The spindle's function as a docking site for MYC during mitosis may constitute a window of specific vulnerability to be exploited for cancer treatment.
Introduction
The nuclear bHLH-Zip transcription factor MYC regulates gene expression programs that govern fundamental cellular processes including proliferation (1–4). The control of MYC activity is tight, with an estimated half-life of MYC protein in the range of 20 to 30 minutes (5–9). MYC protein stability is critically regulated by phosphorylation events in the MB1 box region of the N-terminal transcriptional regulatory domain (10–12). Phosphorylation at Serine62 (Ser62) has been shown to stabilize MYC protein and transcriptional activity (13), whereas subsequent phosphorylation at Threonine58 (Thr58) initiates ubiquitin-mediated proteasomal degradation (11, 14, 15). MYC phosphorylation has been described to occur in mitosis, and several kinases have been involved including the cyclin-dependent kinases Cdk1 and Cdk2 (8, 16, 17). Thus in neuronal precursor cells, Ser54, the paralog MB1 serine phosphorylation site of N-MYC, is phosphorylated by Cdk1 in mitosis-initiating subsequent growth arrest and differentiation of the daughter cells (18). At the same time, MYC activity is a critical determinant of mitotic cell fate, shown to modulate the survival of tumor cells exposed to chemotherapy and to regulate cell fate following asymmetric cell division of antigen-stimulated T cells (19, 20). Given its pivotal role for mitotic progression and proliferation, we here addressed the fate of MYC protein during mitosis in cancer cells and asked for interactions with cytotoxic drugs predominantly acting in mitosis.
Materials and Methods
Cells, tumor model, and small molecules
Lymphoma and cancer cell lines were obtained from public depositories or cooperating groups as listed (Supplementary Materials and Methods). Short-tandem repeat–profiled cells were cultured to obtain frozen master stocks, from which cells were thawed and expanded for experimentation. All cell lines were cultured at 37°C with 5% CO2 in RPMI1640 or DMEM medium containing 10% or 20% FCS (Gibco-BRL) and 1% penicillin/streptomycin (Sigma, Biochrom). Cell cultures were regularly tested for mycoplasma contamination with a PCR-based universal mycoplasma detection kit (VenorGeM, Minerva) according to kit instructions, and cells from mycoplasma-free frozen master stocks negative cultures were used for experimentation. Animal experiments were conducted in accordance with Institutional Animal Care and Use Committee–approved protocols at Beth Israel Deaconess Medical Center. TA lymphoma experiments were performed using a well-established mouse model of cancer, in which tumors arising spontaneously in K14-Cre+ Brca1f/f;Trp53f/f mice are orthotopically transplanted into K14-Cre–negative genetically identical littermates, as previously described (21). Dimethylanthrone (DME) and cycloheximide (CHX) were obtained from Sigma-Aldrich, and pharmaceutical grade vincristine and paclitaxel were purchased from Cell-Pharm and Hexal, respectively. The PI3K inhibitors NVP-BKM120 (BKM120), idelalisib (CAL-101), and copanlisib were obtained or purchased from Selleckchem, Gilead Sciences, and Bayer AG, respectively. Further information on antibodies, small molecules, plasmids, and vectors is provided in the Supplementary Data section. Flow cytometry, cell sorting, confocal microscopy, SDS-PAGE, immunoprecipitation, Western blot, viability tests, RNA preparation, PCR, and colony formation assays in semi-solid methylcellulose were performed according to standard protocols (Supplementary Materials and Methods).
Nuclear magnetic resonance spectroscopy
All nuclear magnetic resonance (NMR) experiments were acquired using a 900 MHz spectrometer equipped with a 5-mm triple resonance cryoprobe. The transfer Nuclear Overhauser enhancement (trNOE) experiments were done using spectral windows of approximately 14 ppm in t1 and t2 dimensions with a mixing time of 350 ms, 64 scans per transient, and 400 t1 increments per experiment. Saturation transfer difference (STD) experiments were done using a standard Bruker pulse sequence, acquiring 16k data points with a saturation time of 2 seconds. The spectra were processed and analyzed using TopSpin (Bruker BioSpin).
Statistical evaluations
Statistical tests were performed using Excel 2013 or GraphPad Prism version 4.03 for Windows as indicated (Microsoft Excel 2013, GraphPad Software, www.graphpad.com), and P values are given for each experiment. Error bars represent SDs of mean values.
Results
MYC is phosphorylated at Ser62 in mitosis, but a fraction of nonphosphorylated MYC binds to tubulin microtubules of the mitotic spindle apparatus
As MYC degradation requires prior phosphorylation at Ser62 and then Thr58 of the MYC-Box I (MBI) region (15), we measured total and phosphorylated MYC using flow cytometry. We found that total MYC was evenly detectable throughout the cell cycle (Fig. 1A, top plot). Phospho-specific MYC antibodies revealed increased phosphorylation at Ser62 (pS62) and Thr58 (pS62/pT58) distinctly in G2–M cells with high total MYC content, with larger populations of pS62-positive compared with pS62/pT58-dual–positive cells (Fig. 1A, as boxed, Fig 1B; Supplementary Fig. S1A). Ser62 phosphorylation of MYC was associated with phosphorylation of the mitosis marker MPM2 in all cell types examined, including the diffuse large B-cell lymphoma cell line OCI-Ly3 (Fig. 1A), a panel of leukemia, lymphoma, and solid cancer cell lines, as well as in proliferating nontransformed cells such as murine embryonic fibroblasts, CD34-positive human hematopoietic stem cells, and mitogen-stimulated human T cells (Supplementary Fig. S1A). Consistent with these cell line observations, we detected a tight correlation of pSer62-MYC with Ser10 phosphorylation in histone H3, also phosphorylated predominantly during mitosis, in clinical specimen of patients with primary aggressive lymphoma (Supplementary Fig. S1C and S1D). At the subcellular level, we found that pSer62-MYC did not localize to chromosomal DNA in metaphase cells, but rather spread evenly throughout the cytoplasm, from where it could be released by cell membrane perforation with digitonin (Supplementary Fig. S1E).
Upon gel electrophoresis, MYC phosphorylated during mitosis migrated at slightly slower speed, and was separated in a major and a minor band (Fig. 1C), presumably due to conformational changes associated with phosphorylation. Similar observations had been made for the corresponding Ser54 phosphorylation site in the MB1 box of N-MYC (18). As controls for the specificity of the anti-MYC and the anti–pSer62-MYC antibodies, we found no binding of the anti-MYC antibody Y69 nor the anti-pSer62 antibody to c-myc−/− HO15.19 cells (Supplementary Fig. S2A), as well as no binding of the phospho-specific antibody to phosphorylation-deficient c-MYC-S62A even at high expression levels (Supplementary Fig. S2B). Immunoblot detection of pSer62-MYC in mitotic cells was sensitive to alkaline phosphatase (AP) treatment, both from endogenous and ectopic MYC expression in Hek293 cells (Fig. 1D; Supplementary Fig. S2C). We then analyzed the levels of MYC and pSer62-MYC in synchronized cells. Following enrichment in mitosis by thymidine synchronization, 8-hour release, and harvest of nonadherent cells, we noted a decrease of MYC at 2 hours after mitosis, which was replenished by CHX-sensitive de novo protein biosynthesis (Fig. 1E and F). We then applied the mitotic kinesin Eg5/Kif11 inhibitor DME to accumulate cells in mitosis and found postmitosis MYC levels decreased to approximately one third at 2 hours after release from mitotic arrest, as quantified by Western blot in HeLa and HCT116 cells (MoAb Y69, HeLa: 30.6% ± 16.4%, HCT116: 41.0% ± 28.1%, n = 3 each). Importantly, DME arrests cells by inhibiting spindle pole segregation, but leaves the microtubule architecture intact (22). This feature separates DME from most G2–M arresting drugs such as nocodazole, colcemid-derivatives, and vinca-alkaloids, which destroy the spindle. Hence, we used DME as a means to achieve “microtubule-sparing” G2–M arrest.
We then stained cells under conditions that preserved microtubule integrity. We discovered that MYC associated with the mitotic spindle apparatus, evident from metaphase through telophase (spontaneous mitotic HeLa cell in Figs. 1G, 2E, and 3A and B). The stains were most clearly defined when using the monoclonal rabbit antibody Y69, a high avidity antibody directed against an epitope within the first 100 amino acids of MYC. Similar staining was seen using a polyclonal rabbit antibody directed against N-terminal (ab N262) and a murine monoclonal antibody against C-terminal (ab 9E10) epitopes (Supplementary Fig. S3A). In addition, we confirmed the finding of MYC association with the spindle by electron microscopy (Supplementary Fig. S3C). MYC had previously been shown to interact with tubulin outside mitosis (23, 24). Applying the method of Alexandrova and colleagues (23) to demonstrate MYC microtubule interactions in vivo in mitosis, we fractionated lysates of mitotic HeLa cells by differential centrifugation which separates MYC bound to polymerized tubulin (P2) from MYC bound to nuclear debris (P1) and soluble MYC in the supernatant (SN2; Fig. 1H). The majority of intact MYC precipitated at 25,000 × g in the P2 fraction, containing the tubulin polymers of the spindle, while in contrast, pSer62-MYC was found predominantly in the supernatant (Fig. 1H; ref. 25). These results clearly confirm the binding of MYC to tubulin microtubules in vivo (23, 24), but importantly, they extend this interaction also to the situation in mitosis. Consistently, immunoprecipitation of tubulin alpha (TUBA) coimmunoprecipitated MYC in lysates of asynchronous and DME-arrested cells, and MYC pull-down coimmunoprecipitated TUBA. However, an immunoprecipitation of TUBA did not precipitate pSer62-MYC from asynchronous cells, and only at low levels from mitotic cells (Fig. 1I).
In summary, we found that in mitosis a proportion of MYC, detached from DNA, is phosphorylated at Ser62 and subsequently at Thr58. A fraction of nonphosphorylated MYC protein, however, binds to the tubulin of the mitotic microtubules, i.e., the spindle.
MYC phosphorylation at Ser62 and Thr58 modulates binding to tubulin
To analyze protein interactions of tubulin and the MYC MB1 region directly, we designed oligopeptides corresponding to amino acid 56 through 68 of MYC [MYC-MBI (56–68)], with and without Ser62 and Thr58 phosphorylation (Fig. 2A). NMR spectroscopy, specifically trNOE experiments, revealed that the unmodified peptide bound to tubulin alpha–beta heterodimers. Binding decreased when modifications were introduced that mimic phosphorylation at Ser62 or Thr58, respectively, by at least a factor of 10 (Fig. 2B and C). We then titrated the peptide competitively against epothilone-A, a ligand to tubulin dimers with a known dissociation constant of 150 μmol/L (26). In these experiments, abolishment upon binding can be used to determine binding affinities (Kd), given a ligand of known affinity is available for the same system. In this case, we have studied epothilone-A bound to tubulin allowing us to determine the Kd of MYC-MBI (56–68) to microtubules to be between 77 ± 7 μmol/L and 565 μmol/L (see Supplementary Information for details). These estimates are in the range of previously reported values for the binding of Tau peptides to microtubules (27) or the binding of MYC-MBI–related peptides to other proteins such as bridging integrator protein 1 (BIN1; ref. 28) or peptides of N-MYC to the kinase Aurora-A (29).
We then mapped the MYC-MBI (56–68) peptide sequence using the Resource Parisienne en Bioinformatique Structurale (RPBS) data bank to obtain binding predictions to the tubulin alpha-/beta tubulin dimer (RPBS, Alland, 2005; refs. 30, 31). Eight of the ten best fitting predictions clustered at the plus-end site of the tubulin alpha–beta heterodimer (Fig. 2D), consistent with binding of the MYC-MB1 region to tubulin close to the dimer interface.
The protection of spindle-bound MYC protein from degradation raises the possibility that it is carried over into the newly formed cells. To examine this possibility, we released cells from mitotic arrest, and measured MYC in sorted postmitotic cells. At 2 hours after mitosis, we found MYC clearly detectable in the nuclei by confocal microscopy, and significant staining remained after suppression de novo protein synthesis in the newly formed cells (Fig. 2E and F), suggesting that a fraction of MYC had been derived from the parental cells. CHX, which we have employed to inhibit protein biosynthesis, may interfere with the stability of proteins. Although we have not seen MYC stabilization in the corresponding immunoblots (Figs. 1F and 2G), we cannot exclude that CHX may have contributed to MYC stability. We then followed MYC levels from ectopic c-myc-wt or the phosphorylation-deficient variants in asynchronous and mitotic Hek293 cells. In the presence of CHX, MYC levels progressively declined in asynchronous cells and in cells released from mitotic arrest, but at a slower pace in the S62A and T58A variants (Fig. 2G). The analysis of the P2 pellet after cell fragmentation then showed that enforced mitotic arrest led to a significant loss of ectopic wt MYC in complex with tubulin polymers, but not for the Ser62- and Thr58-MYC variants (Fig. 2H). As a prerequisite for functional studies without effects of endogenous MYC, we then created cell lines expressing phosphorylation-deficient MYC-T58A and -S62A after extinguishing endogenous MYC expression. Effective silencing of endogenous MYC was achieved with a 3-UTR-MYC-sh-vector (Supplementary Fig. S4C). MYC depletion killed most cells within 7 days and abrogated cell growth completely in HeLa, Hek293, and P493-6 cells (Supplementary Fig. S4D). However, we rescued HeLa cells by expression of MYC-S62A or MYC-T58A, and then we accumulated HeLa-sh-3UTR-MYC/+pMYC-S62A (HeLa-S62A; Supplementary Fig. S4E), HeLa-sh-3UTR-MYC/+pMYC-S58A (HeLa-S58A), and HeLa control cells in mitosis, and followed MYC protein levels during release in the presence of CHX. MYC levels were higher in the variant cell lines and dropped in all cell types, albeit with persistence of higher MYC protein levels in the phosphorylation-deficient MYC-S62A or MYC-T58A variants (Fig. 2I). These data are compatible with slowed MYC decay by phosphorylation-deficient Ser62 and Thr58 mutations both in asynchronous and in mitotic cells.
Taken together, we propose that MYC's binding to tubulin in the mitotic spindle apparatus is modulated by phosphorylation-sensitive interactions at the MB1 (56–68) of MYC.
Tubulin-interacting vincristine destabilizes mitotic MYC
The observation of MYC transfer through mitosis via the spindle raised our interest in the functional consequences of this mechanism, in particular for cell division in cancer. Substances affecting spindle microtubule dynamics are among the most efficient cytotoxic drugs known (32, 33). They induce death in mitosis, but differ in their mechanism of action: vincristine destabilizes microtubules, and paclitaxel stabilizes microtubules (34). We hypothesized that drugs that interfere with microtubule stability should also affect the stability of MYC, because they would modulate the integrity of the tubulin scaffold. Vincristine would be expected to destabilize MYC in mitosis toward increased decay, whereas paclitaxel would not. This effect should differ between cell populations with low mitotic indices from asynchronous cell cultures, and cell populations with high mitotic index, e.g., after DME arrest. Noteworthy, vincristine and paclitaxel alone already arrest cells in mitosis (35), but we expected that preaccumulation of the cells in G2–M by DME pretreatment should greatly enhance the drug effects in mitotic cells. We found that treatment with paclitaxel in fact did not change MYC protein levels in asynchronous and DME-treated HeLa cells, whereas MYC levels in Burkitt lymphoma cells were slightly reduced in DME-treated cells (Fig. 3A). In contrast, vincristine clearly diminished MYC in cell populations arrested in mitosis after DME treatment (“mitotic”) versus untreated populations (“asynchronous”; Fig. 3A; Supplementary Fig. S5A). Confocal microscopy imaging confirmed that cells treated with paclitaxel after DME-mediated G2–M arrest display typical monaster-shaped spindles with positive staining for MYC, whereas vincristine destabilized both spindle microtubules and MYC (Fig. 3B). For further proof, in situ microtubule precipitation experiments showed that vincristine abrogated the coprecipitation of MYC with tubulin, which was not altered by paclitaxel (Fig. 3C and D). Noteworthy, vincristine did not prevent tubulin pellet formation completely (Fig 3C), suggesting residual microtubule formation or interference from other proteins of the cellular extract. We then addressed the impact of vincristine exposure during mitosis on gene transcription. As early as 2 hours after mitosis, there was no immediate effect of vincristine: the expression levels of 89 MYC target genes as well as of MYC itself were not altered in sorted viable postmitotic cells, except reduced levels of branched chain amino acid transaminase 1 transcripts in HCT116 cells (Supplementary Fig. S4B and S4C).
Vincristine and paclitaxel differ in postmitotic cytotoxicity
We next compared the features of cell killing by paclitaxel and vincristine during and after mitosis in a panel of B-cell lymphoma cell lines. We found that both vincristine and paclitaxel induced death of cells in G2–M in a dose-dependent manner, with higher efficacy of vincristine, particularly at low concentrations (Fig. 4A and B). Still, we observed a fraction of viable cells to persist after 72-hour exposure to both drugs (boxed in gray, Fig. 4A, bottom plot). To analyze the fate of these persistent cells, we employed colony formation assays as a measure of single-cell regeneration capacity. In these experiments, the effects of vincristine and paclitaxel diverged dramatically. After exposure to vincristine for 72 hours, colony numbers were significantly lower than after exposure to paclitaxel (Fig. 4C and D). When we calculated the effect ratios of vincristine versus paclitaxel at equimolar concentrations, we found a significantly stronger effect of vincristine on colony forming unit (CFU) inhibition (Fig. 4E). The strong suppressive effects of vincristine on lymphoma colony formation were evident in cells with both high and low MYC expression. But, as shown in the P493-6 system, colony suppression by vincristine was much more pronounced when cells expressed high levels of MYC (Fig. 4F). These experiments showed that vincristine and paclitaxel both induced direct death in G2–M with comparable efficacy. In addition, and clearly distinct, vincristine exerted a strong postdrug exposure inhibitory effect on clonal regeneration in the CFU assays. Noteworthy, this effect was not be detected by standard short-term viability tests. However, this effect should be pertinent to the patient situation, where relapse or resistant disease originates from regenerating tumor cells.
We next addressed the contribution of MYC phosphorylation at Ser62 and Thr58 to the specific cytotoxic effect exerted by vincristine. We exposed HeLa-S62A and HeLa-T58A cells, as described above, to vincristine during mitosis. Although in the parental cells vincristine induced significant MYC decline as expected, the variants with phosphorylation-deficient MYC mutations retained MYC protein levels in the presence of vincristine following accumulation in mitosis (Fig. 5A). In addition, persistence of phosphorylation-deficient MYC was associated with improved viability after 72-hour exposure to vincristine (Fig. 5B), albeit with significant morphologic alterations in the drug-exposed cells. Although the parenteral HeLa cells died with increasing vincristine concentrations, a fraction of the cells with phosphorylation-deficient MYC survived and formed clusters of enlarged cells (Fig. 5C and D). Analyzing the cytogenetic status of these cells, we then found that in the surviving cells with phosphorylation-deficient MYC, the number of chromosomes had doubled after 72 hours of exposure to ≥20 nmol/L vincristine (Fig. 5E and F). The occurrence of tetraploid metaphases clearly documented that the MYC-S62A and MYC-T58A cells had overcome mitotic arrest without actually dividing, had undergone one round of cell cycling, and had re-entered mitosis during the 72-hour exposure to vincristine. We conclude that stabilization of MYC through introduction of the S62A and T58A mutations allowed cells to survive microtubule disruption caused by vincristine during mitosis, and to enter into a subsequent cell cycle with a double set of chromosomes, leading to polyploidy of these cancer cells.
Postmitotic PI3K inhibition potentiates cytotoxic efficacy of vincristine
Having shown that MYC protein levels decrease during the M to G0–G1 transition, we searched for a mechanism to further deepen MYC loss at this point in the cell cycle. For unbiased evidence, we first mined the Genomics of Drug Sensitivity (GSDC) data for cell lines with MYC mutations. In this panel of 1,074 cell lines, we identified five cell lines with MYC missense mutations, involving aa138 in two cell lines (acute lymphoblastic leukemia MN-60, Burkitt lymphoma EB-3), and involving aa58 in three cell lines [B-cell lymphoma JM1, Burkitt lymphoma EB2, diffuse large cell B-cell lymphoma (DLBCL) Nu-DUL-1, http://www.cancerrxgene.org]. Consistent with our findings, the analysis revealed that the five cell lines with MYC mutations were resistant to drugs causing cell death in mitosis, but that they were particularly sensitive to the PI3K inhibitor idelalisib (Fig. 6A). The stability of MYC depends on GSK3β-mediated phosphorylation of MYC at Thr58 (11), which is negatively regulated by PI3K/AKT signaling during the G0–G1 phase of the cell cycle (36). Consistently, we found that the pan-PI3K inhibitor BKM120 depleted MYC in postmitotic HeLa cells within 2 to 4 hours at nanomolar concentrations, associated with low phosphorylation of AKT (Fig. 6B; Supplementary Fig. S6A), whereas at the same time interval, BKM120 did not deplete MYC levels in asynchronous cells (Fig. 6B).
These data suggest a specific efficacy of PI3K inhibition against tumors, in which mutations protect mutant MYC from degradation. In addition, as PI3K inhibition and spindle disruption lower MYC levels through different mechanisms and at different time points during the cell cycle, we saw a rationale to test the effect of PI3K inhibition on MYC levels in cells exposed to vincristine during mitosis. We treated aggressive B-cell lymphoma cells first with vincristine followed by idelalisib. As expected, vincristine at 20 nmol/L for 72 hours abrogated mitotic MYC in murine A20 lymphoma cells and in a panel of human lymphoma cell lines (Fig. 6C, top plot; Supplementary Fig. S6B). Subsequent application of idelalisib then prevented postmitotic resurgence of MYC at 8 hours (Supplementary Fig. S6B) and pronounced after prolonged exposure at 48 hours, including the vincristine-resistant cell lines DG75 and CA46 (Fig. 6D; Supplementary Fig. S6C). Analyzing AKT phosphorylation as an event directly downstream of PI3K, idelalisib inhibited Ser473 phosphorylation in A20 cells when applied after vincristine (Fig. 6C). Vice versa we expected increased AKT activity to counteract the cytotoxic effect of vincristine attributable to MYC depletion. Thus, we expressed constitutively active, myristoylated AKT1 in HeLa and Hek293 cells (Fig. 6E, right plot), and we found that the expression of constitutive active AKT significantly increased the number of viable cells after exposure to vincristine (Fig. 6D, left plot).
Next, we addressed the impact of MYC expression on vincristine and idelalisib susceptibility in the P493-6–aggressive B-cell lymphoma model with doxycycline-sensitive ectopic MYC expression. Vincristine alone was highly cytotoxic, but subsequent exposure to idelalisib still significantly deepened the antitumor effects as measured in cell viability, death in mitosis and colony formation assays, and as expected, more efficiently in cells with high rather than low MYC status (Fig. 6F; Supplementary Fig. S7A and S7B). Extending the strategy to a panel of 16 Burkitt lymphoma and DLBCL cell lines confirmed the additional cytostatic effect of PI3K inhibition after prior vincristine (Fig. 6G–I). Although in these experiments idelalisib was moderately effective as a single agent (mean cell death, 4.7%; range, 1%–11.7%), the efficacy of idelalisib increased substantially when applied after prior exposure to vincristine (mean cell death, 9.5%; range, 1.7%–19.7%). Overall, we observed an inverse correlation of single-agent vincristine cytotoxicity and combined vincristine/idelalisib cytotoxicity (Fig. 6H). Importantly, idelalisib also reduced the clonogenic growth from cells that survived vincristine exposure (Fig. 6I), supporting the notion that PI3K inhibition exerts its cytostatic effects predominantly in the postmitotic phase. The sequence of drug exposure was critical for the observed lethality of combined idelalisib and vincristine: when P493-6 cells were incubated with idelalisib before exposure to vincristine, the cells were arrested in G0–G1, resulting in reduced vincristine-mediated cytotoxicity and significantly increased CFU activity (Supplementary Fig. S7C–S7E). Finally, we applied the concept of sequential vincristine and PI3K inhibition in vivo in two murine-aggressive B-cell lymphoma models. We used the A20 B-lymphoblastic lymphoma that truly mimicked the behavior of human aggressive B-cell lymphoma in the in vitro studies, as well as the novel TA model, which had spontaneously developed in a cancer-susceptible mouse model and was passaged as a syngeneic tumor model in vivo only (Supplementary Fig. S8). Syngeneically implanted tumors were allowed to grow to 5 mm in diameter. Vincristine was dosed weekly, followed by PI3K inhibitor daily, starting 8 hours after vincristine. The PI3K inhibitor was paused for 48 hours prior to the next weekly dose of vincristine to allow for wash-out (Fig. 7A). As the pharmacokinetics of idelalisib prevent its use in mice, we employed the pan-PI3K inhibitor BKM120, and the alpha- and delta-preferential PI3K inhibitor copanlisib, both under clinical evaluation in lymphoma (37, 38). Consistent with the findings in vitro, sequential vincristine and BKM120 reduced MYC protein levels in the tumors and induced tumor necrosis (Fig. 7B), whereas neither vincristine nor the PI3K inhibitor, when given as single agents, was effective against the fast growth of these lymphomas. Sequential treatment significantly delayed tumor growth in both tumor models (Fig. 7C and E). In histologic analysis, tumor regressions induced by sequential vincristine and the PI3K inhibitors were associated with a marked reduction of cellularity, leaving only few interspersed tumor cells in the stroma (Fig. 7D and F).
In summary, PI3K inhibition further depletes MYC protein immediately following mitosis contributing to the eradication of persistent cells that can overcome the G2–M arrest induced by vincristine. Hence, PI3K inhibition may be particularly active against aggressive lymphoma when administered after vincristine.
Discussion
Binding of MYC to the mitotic spindle
Our key finding is that MYC, nonphosphorylated at Ser62, binds to the polymerized tubulin of the mitotic spindle, which may allow transfer of functional MYC to the daughter cells. Although MYC degradation during mitosis may provide a mechanism for transcriptional reset for the nascent cells on the one hand (18), tubule-mediated transmission of MYC would provide a scaffold for regulating the transfer of MYC to the new cells. Although we have no exact data on the amount of MYC at the spindle reservoir, the principle that cytoskeletal structures serve as docking sites from which an essential protein can be mobilized in its active form had been shown for transcription factors such as the release of NF-kB through microtubule depolymerization (39), or metabolic enzymes, such as aldolase that reversibly attaches to the actin cytoskeleton (40). Noteworthy, although we gained evidence for binding of MYC to microtubules, others have found phosphorylation-sensitive binding of MYC family members to other proteins, such as to the tumor suppressor BIN1(28), the binding of N-MYC to the kinase Aurora-A in neuroblastoma cells (41) or the binding of Ser62-phosphorylated MYC to nuclear lamins (42) with comparable affinities of the binding regions. The complex formation with Aurora-A protects N-MYC from phosphorylation and degradation in the G2–M phase, and it serves a function similar to the role described for tubulin here, but it may differ in the subcellular localization: Aurora-A primarily localizes with the centrosomes and extends to the spindle proximal to the poles only (41). Several serine kinases were described to phosphorylate MYC at the priming Ser62, including MAPK family members, polo-like kinase 1, calcium/calmodulin-dependent kinase IIg, and dual specificity tyrosine phosphorylation–regulated kinase 2 (11, 13, 43–45). In mitosis, turnover of N-MYC was found increased in cerebellar neural precursor cells, and phosphorylation of N-MYC at Ser54 required the CDK1 complex for Ser54 phosphorylation (18). Our data are consistent with Ser62 phosphorylation in mitosis acting as priming event for MYC degradation. Given the ambivalent role of Ser62 phosphorylation for both stabilization and degradation of MYC (8, 10–12), a scenario arises, in which several proteins beyond tubulin may compete for binding to the MB1 box and may thus govern the fate decision toward Thr58 phosphorylation–mediated degradation versus MYC protein stabilization. This mechanism may gain particular significance during symmetric and asymmetric cell divisions in stem and immune cells, where mitotic MYC regulation has already been identified as an important determinant of cell fate (19, 20). In mammalian development, mitotic MYC partition might operate in tandem with posttranslational regulation of GSK3ß, which phosphorylates MYC at Thr58 and which is inhibited as key downstream event in Wnt signaling. Importantly, inhibition of GSK3ß activity by Wnt-mediated posttranslational stabilization of proteins has been described also to culminate in mitosis (46).
Our data suggest that nonphosphorylated MYC binds to tubulin polymers, in line with previous results (23, 24, 47), whereas phosphorylated MYC loses its affinity to the tubulin polymer. Our data also show that tubulin polymer destabilization increases MYC destruction (Fig. 5), overall suggesting equilibrium of mitotic microtubule dynamics and the dynamics of MYC phosphorylation and decay in mitosis. This equilibrium is modulated by substances interfering with microtubule stability, e.g., by vincristine. Vinca-alkaloids bind to tubulin-beta within a central domain close to the GTP-binding site of the protein (48). As demonstrated for vinblastine and colchicine, the alkaloids create a wedge at the interdimer interface within the microtubule lattice and bend the protofilament growth axis (49), shifting the microtubule assembly/disassembly cycle toward depolymerization. The putative binding site of MYC at the plus end interface of the alpha/beta heterodimer might infer that the polymer formation affects—as we speculate—the binding of MYC to tubulin. However, at this point, we cannot exclude that tubulin dimers also bind to MYC at low affinity. Further characterizing the MYC/tubulin binding sites might open an avenue to design drugs either to stabilize MYC for use in regenerative medicine or to deplete MYC to fight cancer.
The MYC–microtubule interaction explains resistance to vinca-alkaloids
Perhaps the most immediate implications of our findings concern the mitotic MYC repartition in cancer cells: Mitotic depletion of MYC by vinca-alkaloids adds to explain the eminent efficacy of these drugs against MYC-driven and/or MYC-addicted tumors such as aggressive lymphatic neoplasms (Burkitt and Hodgkin's lymphoma, non-Hodgkin lymphoma) for MYC and neuroblastoma for N-MYC (50), and lends credence to the efficacy of approaches that use supplemental or intensified applications of vinca-alkaloids (51, 52).
Burkitt-type aggressive lymphoma is characterized by MYC activation based on chromosome 8;14 translocation and production of a highly active MYC–IgH fusion protein (53). In addition, MYC mutations in the MB1 region are a recurrent event in Burkitt lymphoma tumors (7, 54, 55), which, according to our data, would cause persistence of phosphorylation-deficient MYC through mitosis resulting in reduced susceptibility to vincristine.
Consistent with this reasoning, recent next-generation sequencing (NGS) analyses in another MYC-driven lymphoma, DLBCL, documented poor treatment outcomes specifically for those patients whose tumor carried a MYC-Thr58 mutation and who received a vincristine-containing regimen (56, 57). At this point, it remains open whether the poor outcome associated with MYC-Thr58 mutations is due to a direct growth advantage for tumor cells with high levels of postmitotic MYC, or indirectly linked to perturbation in chromosome segregation we discovered when cells with this mutation underwent DNA replication without cell division (Fig. 5). In line with these findings, elevated MYC levels have previously been recognized to cause chromosomal instability (58), which represents a pivotal mechanism in the progression and clonal evolution of tumors (59). Notably, the hotspot of MBI mutations found in clinical series extends from amino acid 56 through 63, which includes—but is not exclusive to—the Ser62 and Thr58 phosphorylation sites (55). It remains to analyze, however, whether amino acid exchanges other than Ser62 and Thr58 are relevant for the interaction of MYC with tubulin.
The dependence of daughter cells on inherited MYC is a unique opportunity for cancer treatment
Our observations identify the early postmitotic phase as a window for efficient tumor cell targeting. Suppressing of MYC activity in G1 phase through PI3K inhibition was linked with increased tumor cell death and strong impairment of clonogenic growth from surviving cells, consistent with previous reports that link PI3K signaling and MYC transcriptional activity to cell-cycle entry in G1 (60). The cooperation of PI3K and MYC is dependent on MYC stabilization via GSK3ß inhibition (61) and AKT-mediated phosphorylation of FoxO proteins (62). This cooperation is already active in Burkitt lymphoma tumorigenesis in mice, in which combining constitutive c-Myc expression and PI3K activity in germinal center B cells of the mouse led to Burkitt lymphoma–like tumors (63). Consistently in human Burkitt lymphoma disease, TCF3 mutations activating the PI3K pathway were found to occur at high frequency (54). Although the in vivo interactions of PI3K inhibition and vincristine certainly deserve further dissection, the data in this article support a role of PI3K activity for MYC stability through AKT activation as a potential treatment target. We show here that the oncogenic cooperation can be interrupted by applying G2–M arresting, MYC-depleting drugs first, followed by the PI3K inhibitor shortly thereafter. Repetitive rounds of sequential vincristine/PI3K inhibitor were highly effective in vivo, suggesting a strong potential for this treatment strategy in clinical trials for MYC-driven malignancies.
Disclosure of Potential Conflicts of Interest
G.M. Wulf reports receiving other commercial research support from Merck & Co. No potential conflicts of interest were disclosed by the other authors.
Authors' Contributions
Conception and design: C. Kiecke, E. Schäfer, R. Koch, M. de Oliveira Taveira, G.G. Wulf
Development of methodology: S. Becker, C. Kiecke, E. Schäfer, P. Trigo-Mourino, Z. Rydzynska, S. Dierks, H. Bastians, M. de Oliveira Taveira, G.G. Wulf
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S. Becker, C. Kiecke, E. Schäfer, U. Sinzig, P. Trigo-Mourino, C. Griesinger, R. Koch, F. von Bonin, H. Bohnenberger, J. Flach, S. Dierks, B. Maruschak, K. Bojarczuk, M. de Oliveira Taveira, L. Trümper, G.M. Wulf, G.G. Wulf
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S. Becker, C. Kiecke, E. Schäfer, L. Deuper, P. Trigo-Mourino, C. Griesinger, R. Koch, B. Chapuy, V. Venkataramani, A. Leha, S. Dierks, B. Maruschak, M. de Oliveira Taveira, G.M. Wulf, G.G. Wulf
Writing, review, and/or revision of the manuscript: C. Kiecke, E. Schäfer, P. Trigo-Mourino, R. Koch, B. Chapuy, D. Kube, K. Bojarczuk, M. de Oliveira Taveira, L. Trümper, G.M. Wulf, G.G. Wulf
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): S. Becker, E. Schäfer, L. Deuper, B. Chapuy, F. von Bonin, L. Trümper, G.G. Wulf
Study supervision: C. Griesinger, G.G. Wulf
Others (technician): F. von Bonin
Others (selected experiments and data interpretation): D. Kube
Acknowledgments
This work was supported by the University Medicine Goettingen (Forschungsfoerderung to E. Schäfer) and the Goettinger Gesellschaft zur Unterstuetzung der Krebsforschung e.V (to G.G. Wulf). P. Trigo-Mourino acknowledges the Humboldt Foundation for a postdoctoral research fellowship.
FACS and cell sorting were performed in the Flow Cytometry Core Facility at the University of Goettingen. We thank A. Sands for linguistic editing.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.