Cell cycle checkpoints are implemented to safeguard genome, avoiding the accumulation of genetic errors1–2. Checkpoint loss results in genomic instability and contributes to the evolution of cancer. Among G1-, S-, G2- and M-phase checkpoints, genetic studies indicate the essence of an intact S phase checkpoint in maintaining genome integrity3–4. Although the basic framework of S phase checkpoint in higher eukaryotes has been outlined, the mechanistic details remain to be elucidated. Human chromosome 11 band q23 translocations disrupting the MLL/HRX/ALL-1 gene lead to poor prognostic leukemias5–9. Here we assign MLL as a novel effector in the mammalian S phase checkpoint network and identify checkpoint dysfunction as an underlying mechanism of MLL leukemias. MLL is phosphorylated at serine 516 by ATR in response to genotoxic stress in S phase, which disrupts its interaction with and thereby degradation by the SCFSkp2 E3 ligase, leading to its accumulation. Stabilized MLL protein accumulates on chromatin, methylates histone H3K4 at late replication origins, and inhibits the loading of CDC45 to delay DNA replication. Cells deficient in MLL exhibited radioresistant DNA synthesis (RDS) and chromatid-type genomic abnormalities, indicative of S phase checkpoint dysfunction. Reconstitution of MLL−/− mouse embryonic fibroblasts (MEFs) with wild-type but not S516A or ΔSET mutant MLL rescues the S phase checkpoint defects. Moreover, murine myeloid progenitor cells (MPCs) carrying an MLL-CBP knock-in allele that mimics human t(11;16) leukemia exhibit a severe RDS phenotype. MLL-fusions function as dominant negative mutants that abrogate the ATR-mediated phosphorylation/stabilization of wild-type MLL upon DNA damage and thus compromise the S phase checkpoint. Altogether, our study identifies MLL as a key constituent of the mammalian DNA damage response pathway and deregulation of the S phase checkpoint incurred by MLL translocations likely contributes to the pathogenesis of human MLL leukemias.
Leukemogenic MLL translocations fuse the common MLL N-terminal 1,400 aa in frame with more than 60 partners8. The MLL gene encodes a 500 kD precursor MLL500 which is processed by Taspase110 to generate mature, heterodimerized MLLN320/C180. MLL participates in embryogenesis, cell fate, cell cycle and stem cell function7,11–14, in part by methylating histone H3 lysine 4 (H3K4) through its C-terminal SET domain15. Although the importance of HOX gene deregulation in the pathogenesis of MLL leukemias has been extensively investigated5–8, physiological MLL-fusion knock-in mouse models indicate that HOX gene aberrations alone are insufficient to initiate MLL leukemias7,16.
MLL participates in the cell cycle control12,17–19 and exhibits a biphasic expression with peaks at G1/S and G2/M transitions12. This unique, two peaks are conferred by proteasome-mediated degradation–SCFSkp2 and APCCdc20 degrade MLL at S and M phases, respectively12. Why MLL needs to be degraded in S and M phases is unclear. The observation that over-expression of MLL impedes S phase progression12 raises a testable thesis that MLL may accumulate in S phase upon DNA damage to delay DNA replication for repair. Indeed, tested DNA perturbation agents, including aphidocolin, hydroxyurea (HU), ultraviolet light (UV), etoposide, and γ-ionizing irradiation (γ-IR), induced the MLL protein expression, (Fig. 1a and Supplementary Fig. 1a, b). The MLL protein was induced upon DNA damage in S but not G1 or M phases through a transcription-independent mechanism (Fig. 1b and Supplementary Fig. 1c).
The S phase checkpoint senses DNA damage, activates ATM/ATR, inhibits the firing of late replication origins, and enlists repair machineries. “Chromatid-type” genomic errors, accrued during S phase, include quadriradials, triradials, and chromatid gaps and breaks20. Metaphase spread analysis demonstrated a higher incidence of chromatid-type errors in mitomycin C treated MLL−/− than wild-type cells (Fig. 1c). Cells with compromised S phase checkpoints exhibit radioresistant DNA synthesis (RDS)3. Knockdown of MLL in 293T cells or genetic deletion of MLL in MEFs resulted in RDS (Fig. 1d), confirming a critical role of wild-type MLL in the mammalian S phase checkpoint. To explore whether MLL-fusions incur S phase checkpoint defects, we generated myeloid precursor cells (MPCs) from MLL+/ex7(stop)CBP mice that carry a knock-in inducible MLLex7-CBP allele (Supplementary Fig. 2)21. MLL+/ex7(stop)CBP MPCs retained only one copy of wild-type MLL and consequently exhibited a partial RDS phenotype (Fig. 1e). Remarkably, a severe RDS phenotype was observed in MLL+/ex7-CBP MPCs (Fig. 1e). These data suggest that MLL-CBP functions as a dominant negative mutant that actively compromises the S phase checkpoint, contributing to the acquisition of additional chromosomal translocations observed in MLL+/ex7-CBP leukemias21. Furthermore, expression of MLL-AF4 or MLL-AF9 in Jurkat T cells resulted in an RDS phenotype despite the presence of two wild-type MLL alleles (Supplementary Fig. 3). Consistently, expression of MLL-ENL in progenitor cells increased chromosomal abnormalities upon etoposide treatment22.
MLL is normally degraded in S phase by SCFSkp2 which directly binds to the N-terminal 1,400 aa of MLL12. It is conceivable that signal transduction triggered by DNA damage disrupts the MLL-Skp2 interaction and thereby induces MLL, which was indeed observed (Fig. 2a). As the DNA damage response network relays signals mainly through phosphorylation, we examined whether inhibition of proximal kinases including ATM, ATR and DNA-PKcs prohibited the DNA damage-induced MLL accumulation. LY294002 and Wortmannin abolished the MLL accumulation upon DNA damage (Fig. 2b). To specify key kinase(s) required for such signaling, we employed MEFs with deletion of ATM, ATR or DNA-PKcs23–24. Deficiency in ATR greatly reduced the accumulation of MLL upon DNA damage (Fig. 2c and Supplementary Fig.4), identifying ATR as the principal kinase for the MLL induction.
As DNA damage signals disrupt the MLL-Skp2 interaction, ATR might directly or indirectly phosphorylate MLL and/or Skp2, leading to their dissociation. Bioinformatics analysis (scansite.mit.edu) identified a candidate ATM/ATR site at conserved serine 516 (LPISQSP) of MLL (Fig. 3a). In contrast to the disrupted interaction between wild-type MLL and Skp2 upon HU treatment, a comparable interaction was detected between MLL(S516A) and Skp2 irrespective of DNA insults (Fig. 3b), suggesting that S516 phosphorylation dissociates MLL from Skp2. The S516 of MLL became phosphorylated after HU treatment (Fig. 3b), correlating with diminished MLL-Skp2 interaction. To assess if ATR can directly phosphorylate MLL, we performed in vitro kinase assays using affinity-purified ATR and recombinant MLL proteins. As ATR needs to be fully activated by TopBP1 or Claspin25, purified ATR only weakly phosphorylated wild-type but not S516A MLL protein (Fig. 3c). Once activated by TopBP1, ATR effectively phosphorylated MLL, detected by both anti-phospho-ATM/ATR substrate and anti-phospho-MLL(S516) antibodies (Fig. 3c). To assess the functional significance of S516 phosphorylation, MLL−/− MEFs reconstituted with wild-type or S516A human MLL (Supplementary Fig. 5) were subjected to RDS and metasphase spread assays (Fig. 3d, e). Unlike wild-type MLL, S516A MLL failed to fully rescue the RDS defects and chromatid-type errors of MLL−/− MEFs (Fig. 3d, e). Although S516A MLL is defective in the S phase checkpoint, its interaction with Menin and WDR5 and targeting to the promoters of HoxA9 and Meis1 remain intact as wild-type MLL (Supplementary Fig. 6).
To investigate the mechanism(s) by which MLL engages S phase checkpoint, we determined if MLL deficiency affects the upstream signal transduction upon DNA damage. Like wild-type MEFs, MLL−/− cells were competent in the formation of γH2AX foci (Supplementary Fig. 7a). The autophosphorylation of ATM, the S139 phosphorylation of H2AX, the ATM-mediated activating phosphorylation of Chk2, and the ATR-mediated activating phosphorylation of Chk1 were not affected by the MLL deficiency (Supplementary Fig. 7b, c). Furthermore, the phosphorylation of SMC1 by ATM/ATR, the degradation of CDC25A signaled by Chk kinases, and the Y15 phosphorylation of CDK2 were also not altered in MLL deficient cells (Supplementary Fig. 7d).
The key effector step at the initiation of DNA replication is the loading of CDC45 onto the pre-replication complex (pre-RC) which consists of ORC and the MCM2-7 complex26. The chromatin association of CDC45 correlates well with DNA synthesis, and thus marks the firing of replication origins26. Supporting the role of MLL in S phase checkpoint, an aberrant chromatin association of CDC45 was observed in MLL deficient cells upon DNA insults, whereas the chromatin association of MCM2 was not altered (Fig. 4a and Supplementary Fig. 8). The S phase checkpoint commenced at ATM/ATR ultimately inhibits CDC45 loading, in part through inactivating CDK2 and DDK3,3,27–28, which was not affected by the MLL deficiency (Supplementary Figs. 7d and 9). Furthermore, the MLL-mediated inhibition of chromatin association of CDC45 was transcription-independent (Supplementary Fig. 10). ChIP (chromatin immunoprecipitation) assays on the β-globin origin, a well characterized late replication origin in 293T cells29, demonstrated that MLL accumulated and methylated H3K4 at the β-globin origin upon DNA damage, resulting in a decreased CDC45 occupancy (Fig. 4b). These data suggest that the histone methyl transferase (HMT) activity of MLL may be required for the execution of S phase checkpoint, which is corroborated by the inability of ΔSET MLL mutant to fully correct the RDS defects and chromatid-type errors of MLL−/− MEFs (Fig. 3d, e). In fact, histone H3 directly interacted with CDC45 and this interaction was greatly compromised when H3K4 was trimethylated (Fig. 4c, d). Data presented thus far support a model in which stabilized MLL accumulates on chromatin to methylate H3K4 at late replication origins upon S phase checkpoint activation, which inhibits CDC45 loading and thereby delays DNA replication (Fig. 4h). Although MLL likely methylates H3K4 at all late replication origins upon DNA damage, genome-wide studies are required to conclude such a mechanism.
We next investigated how MLL-fusions function as dominant negative mutants in the S phase checkpoint. Of note, all MLL-fusions have lost their C-terminal SET domain (Fig. 3a) and tested MLL-fusions were resistant to the Skp2-mediated degradation due to impaired interaction12. Accordingly, MLL-AF4 and MLL-AF9 were not further stabilized upon DNA damage despite the increased S516 phosphorylation (Supplementary Fig. 11). Chromatin association assays revealed an aberrant loading of CDC45 upon DNA damage in Jurkat T cells that stably express MLL-AF4 or MLL-AF9 (Fig. 4e). While MLL-AF4 and MLL-AF9 stably bound to chromatin, wild-type MLL (MLLC180) failed to accumulate on chromatin in the presence of MLL-Fusions (Fig. 4e). Consequently, ChIP assays demonstrated a stable association of FLAG-MLL-fusions, an ablated accumulation of wild-type MLL (MLLC180), a failed induction of H3K4me3, and an aberrant loading of CDC45 on the late replication origin upon genotoxic stress (Fig. 4f). Altogether, these data suggest that MLL-fusions function as dominant negative mutants by preventing the stabilization of wild-type MLL upon DNA insults. As the ATR-mediated phosphorylation and dissociation of wild-type MLL from Skp2 constitutes the initiating step of MLL-mediated S phase checkpoint response, MLL-fusions might prevent the stabilization of wild-type MLL upon DNA damage by abrogating the S516 phosphorylation of wild-type MLL and thus preserving the interaction between wild-type MLL and Skp2. Since ATR associates with chromatin and becomes active upon DNA insults25, the pre-occupancy of MLL-fusions on chromatin would prohibit the access of wild-type MLL to the activated ATR. Co-expression of MLL-AF9 with wild-type MLL abrogated the S516 phosphorylation of wild-type MLL but not MLL-AF9 upon DNA insults, leading to a constitutive interaction/degradation of wild-type MLL by Skp2 (Fig. 4g). In summary, MLL-fusions stably associate with chromatin and prevent the stabilization/targeting of wild-type MLL to the late replication origin upon DNA damage, which abolish the trimethylation of H3K4 and result in an aberrant loading of CDC45 and dysfunction of S phase checkpoints (Fig. 4e–h and Supplementary Fig. 12). The discovery of MLL in executing S phase checkpoint provides new mechanistic insights concerning not only normal cell biology but also the pathology underlying MLL leukemias.