Biochemical Journal

Research article

UHRF1 depletion causes a G2/M arrest, activation of DNA damage response and apoptosis

Amy L. Tien, Sucharita Senbanerjee, Atul Kulkarni, Raksha Mudbhary, Bernadette Goudreau, Shridar Ganesan, Kirsten C. Sadler, Chinweike Ukomadu

Abstract

UHRF1 [ubiquitin-like protein, containing PHD (plant homeodomain) and RING finger domains 1] is required for cell cycle progression and epigenetic regulation. In the present study, we show that depleting cancer cells of UHRF1 causes activation of the DNA damage response pathway, cell cycle arrest in G2/M-phase and apoptosis dependent on caspase 8. The DNA damage response in cells depleted of UHRF1 is illustrated by: phosphorylation of histone H2AX on Ser139, phosphorylation of CHK (checkpoint kinase) 2 on Thr68, phosphorylation of CDC25 (cell division control 25) on Ser216 and phosphorylation of CDK1 (cyclin-dependent kinase 1) on Tyr15. Moreover, we find that UHRF1 accumulates at sites of DNA damage suggesting that the cell cycle block in UHRF1-depleted cells is due to an important role in damage repair. The consequence of UHRF1 depletion is apoptosis; cells undergo activation of caspases 8 and 3, and depletion of caspase 8 prevents cell death induced by UHRF1 knockdown. Interestingly, the cell cycle block and apoptosis occurs in p53-containing and -deficient cells. From the present study we conclude that UHRF1 links epigenetic regulation with DNA replication.

  • apoptosis
  • caspase
  • cell cycle
  • methylation
  • phosphorylation
  • ubiquitin-like protein containing PHD and RING finger domains 1 (UHRF1)

INTRODUCTION

Following genotoxic injury, the regulated process of mammalian cell cycle progression is often halted. This arrest occurs through the activation of checkpoints that result in cell cycle blocks, allowing for damage repair or cell death if the damage is irreparable [1]. The consequence of abrogating such checkpoints includes the proliferation of cells with defective DNA content, a frequent cause of carcinogenesis.

UHRF1 [ubiquitin-like protein, containing PHD (plant homeodomain) and RING finger domains 1] was identified as a factor that binds the inverted CCAAT box in the topoisomerase 2a promoter and regulates its expression [2]. However, subsequent studies have not demonstrated that UHRF1 functions as a transcription factor. Instead, UHRF1 functions to regulate gene expression through epigenetic mechanisms including DNA methylation [3,4], histone deacetylation [5], histone methylation [6] and possibly histone ubiquitination [7]. UHRF1 is a multi-domained protein (Figure 1A) that contains: (i) an N-terminal ubiquitylation-like domain; (ii) a PHD, through which it interacts with methylated histones [8,9], retinoblastoma protein [10] and DNMT1 (DNA methyltransferase 1) [3]; (iii) an SRA (SET-and-RING-finger associated domain) that interacts with hemimethylated DNA [1113], and HDAC1 (histone deacetylase 1) [5]; and (iv) a RING finger motif that has an E3-ubiquitin-ligase activity [7,14]. Therefore UHRF1 is thought to regulate gene expression through epigenetic mechanisms.

Figure 1 Depletion of UHRF1 results in a G2/M-phase block

(A) UHRF1 is a multi-domained protein with ubiquitin (UBQ), PHD, SRA and RING finger domains. The location of UHRF1 siRNAs si-A, si-B and si-C targets are shown. (B) Western immunoblotting of HCT116 protein lysates from cells transfected with non-targeting siRNA (NTS) or UHRF1-targeting siRNAs (si-A, si-B and si-C), shows efficient knockdown of UHRF1 with targeting siRNAs. (C) Depletion of UHRF1 leads to a G2/M-phase arrest. The percentage of cells in the G1- (hatched), S- (black) and G2/M-phase (white) in cells transfected with NTS, si-A or si-B. Error bars represent S.D. from the mean of three experiments. (D) Enhanced Ser10 phospho-histone H3 levels in UHRF1-knockdown cells. Depletion of UHRF1 (lanes 2 and 3) leads to increased levels of phosphorylated histone H3 (panel 2). Total histone H3 levels are constant. (E) There is no concurrent G1-phase arrest in UHRF1-depleted cells. The remnant G1-phase population in UHRF1-containing and -deficient cells (top histograms) progress to the nocodazole-induced mitotic block (bottom histograms). (F) The level of G1- and S-phase cyclins D2, E and A, remain stable in UHRF1-depleted cells (top three panels), while the cyclin B1 level is elevated (panel 4). β-Actin is used as a loading control for all experiments.

UHRF1 is required for cell cycle progression of non-cancerous cells. UHRF1 mRNA and protein fluctuate with the cell cycle [15,16], depletion of UHRF1 abrogates S-phase entry [17] and zebrafish with a loss-of-function mutation in uhrf1 have defects in hepatocyte proliferation and increased apoptosis [15]. In cancer cells, UHRF1 levels are high and the protein is equally expressed in all phases of the cell cycle [10,16,18]. However, reports on the effects of UHRF1 depletion in cancer cells have been varied. For example, siRNA (short interfering RNA)-mediated knockdown of UHRF1 in HeLa cells concurrently treated with adriamycin causes a small percentage of cells to arrest in the G1-phase [18]. However, in H1299 cells a modest 2-fold knockdown of UHRF1 by shRNA (small-hairpin RNA) causes cells to arrest in either the G1- or G2/M-phases [14]. Regardless of these differences, it is clear that cell cycle progression requires UHRF1 [10,18]. These data establish the possibility that depleting cancer cells of UHRF1 may lead to cell death.

Previous studies have shown that UHRF1 functions to conserve epigenetic inheritance [3,4]. UHRF1 interacts with DNMT1, which methylates cytosines on CpG islands of hemimethylated DNA. UHRF1 also interacts with hemimethylated DNA allowing the methyl cytosine of the parent strand to ‘flip out’ of the double helix so that DNMT1 can access the unmethylated cytosine on the daughter strand [1113]. Indeed, depletion of UHRF1 prevents the association of DNMTI with chromatin leading to the hypomethylation of many genes [3]. A role for UHRF1 in maintaining genomic integrity has been suggested in experiments which show that cells lacking UHRF1 are hypersensitive to DNA damage by genotoxic agents [19]. Furthermore DNMT1, which interacts with UHRF1, accumulates at sites of DNA damage [20]. Lastly, the inactivation of DNMT1 in HCT116 cells leads to activation of the DNA damage response pathway and a G2/M-phase block [21]. These studies support the hypothesis that proper UHRF1 function is required for genomic fidelity.

In the present study, we test this hypothesis by depleting UHRF1 from cancer cells and investigating the effects on the cell cycle. We show that UHRF-depleted cells undergo a caspase 8-mediated apoptosis and that cell cycle arrest and cell death in response to UHRF1-knockdown cells does not require p53. Moreover, we find that UHRF1 accumulates rapidly at sites of DNA injury. Taken together, these data support a model in which UHRF1 is required for genomic fidelity and its loss causes activation of the DNA damage responses and cell death.

EXPERIMENTAL

Materials, cell lines, cell culture and siRNA transfections

Caffeine (Sigma) was used at 5 mM, nocodazole (Calbiochem) was used at 40 ng/ml and the ATM (ataxia telangectasia mutated)/ATR (ATM- and Rad3-related) inhibitor CGK 733 (Calbiochem) was used at 10 μM. 5-Azacytidine (MP Biochemicals) was used at 5 μM. Human colorectal cancer cells HCT116 were obtained from the A.T.C.C. (Manassas, VA, U.S.A.), and HCT116 (p53−/−) was a gift from Dr B. Vogelstein (Johns Hopkins University, Baltimore, MD, U.S.A.). Cells were grown in 10% (v/v) FBS (fetal bovine serum) and 5% (w/v) penicillin/streptomycin in McCoy's 5A modified medium. Liver carcinoma cells Huh7 and Hep3B were cultured in DMEM (Dulbecco's modified Eagle's medium) and minimum essential medium respectively, supplemented with 10% (v/v) FBS and 5% (w/v) penicillin-streptomycin antibiotics. Cells were transfected with non-targeting sequence from firefly luciferase (Dhamarcon) or specific siRNA-targeting UHRF1 or caspase 8 (Invitrogen). Two transfections were performed 24 h apart and cells were harvested 48 h after the initial transfection, except for the experiment shown in Supplementary Figure S3 where TUNEL (terminal deoxynucleotidyltransferase-mediated dUTP nick-end labelling) assays were performed 72 h after the initial transfection.

siRNA sequences are shown in Supplementary Table S1 (at http://www.BiochemJ.org/bj/435/bj4350175add.htm).

Immunoblotting

Western immunoblotting was performed as previously described [22]. Cells were lysed in HSLB [high-salt lysis buffer; 25 mM Hepes (pH 7.4), 400 mM KAc, 10% (v/v) glycerol, 5 mM MgCl2, 2 mM EDTA and 0.5% NP40 (Nonidet P40)] on ice for 45 min. An equal volume of LSLB (low-salt lysis buffer; as HSLB except with 0 mM KAc) was added. Approx. 50 μg of protein was loaded for each lane and separated on a 10% polyacrylamide gel.

The antibodies used, the proteins they have been raised against and the dilutions they were used at are listed in Supplementary Table S2 (at http://www.BiochemJ.org/bj/435/bj4350175add.htm).

qPCR (quantitative PCR)

RNA was extracted using an RNeasy spin column kit (Qiagen) according to the manufacturer's protocol. The first strand cDNA was synthesized from 2 μg of total RNA SuperScript III RT (Invitrogen) for RT (reverse transcription)–PCR. qPCRs were performed in triplicate using LightCycler SYBR Green reagent and detector system (Roche). Gene expressions were normalized to β-actin as a reference. A mean±S.E.M. for three individual experiments is shown. Primer sequences for qPCR are shown in Table S3 (at http://www.BiochemJ.org/bj/435/bj4350175add.htm).

Fluorescence microscopy

TUNEL assays were performed using an in situ death detection kit (Roche) according to the manufacturer's protocol. For immunofluorescence, cells were fixed in ice-cold methanol for 20 min and then blocked in blocking buffer [PBS containing 3% (w/v) BSA and 0.1% NP40] for 30 min. Cells were stained with primary antibody dissolved in blocking buffer for 1 h at room temperature (25 °C). After staining, cells were washed in blocking buffer, and then incubated with a secondary fluorescently tagged antibody. The cells were mounted with DAPI (4′,6-diamidino-2-phenylindole), and then visualized by fluorescence microscopy.

FACS analysis

Cell cycle analysis was performed as previously described [22]. Briefly, following fixation and PI (propidium iodide) staining, PI-positive cells were sorted and histograms were analysed using Modfit LT (version 3.0, Verity Software House). For nocodazole treatment, cells were transfected with control or UHRF1-targeting siRNA for 24 h and then incubated with or without nocodazole (40 ng/ml) for an additional 24 h. Cells were then collected for FACS analysis.

UVA-laser-scissor-induced DNA injury

HeLa cells were treated with 10 μM 5-iodo-2-deoxyuridine (Sigma) for 24 h prior to laser irradiation. LabTek chambers were mounted on a Zeiss Axiovert 200 microscope integrated with the P.A.L.M. Microlaser workstation (P.A.L.M. Laser Technologies). Cells were visualized under visible light and laser-targeted nuclei were selected using the supplied software. A pulsed UVA laser (30 Hz, 337 nm) coupled to the bright-field path of the microscope was focused through a LD 40×; NA (numerical aperture) 0.6 Zeiss Achroplan objective to yield a spot size of approx. 1 μm. Nuclei were subsequently irradiated with a pulsed solid-state UVA laser (30 Hz, 337 nm) with following settings: energy, 35; focus, 57; and cut speed between 10 and 15 with laser output set to 50%. An average of 100 cells were micro-irradiated within 2–5 min, and cell nuclei were exposed to the laser beam for less than 500 ms.

RESULTS

UHRF1 depletion arrests cells in the G2/M-phase

UHRF1 is expressed at high levels in many cancer cell lines and primary tumours [14,16]. Studies have suggested that loss of UHRF1 prevents cell cycle progression and thus UHRF1 may be a potential target for decreasing tumour growth in vivo. To characterize the cell cycle behaviour of colorectal cancer cells (HCT116) depleted of UHRF1, we used siRNA to knockdown UHRF1. Three siRNAs (si-A, si-B and si-C) targeting distinct regions of UHRF1 (Figure 1A) effectively depleted UHRF1 as assessed by Western blot analysis (Figure 1B). Similar UHRF1 depletion was achieved in liver tumour cell lines (Huh7 and Hep3B) and breast cancer cell lines (MCF7) transfected with these siRNAs (results not shown). To control for potential off-target effects, the majority of our studies were carried out using two different UHRF1-targeted siRNAs (si-A and si-B).

We evaluated the relationship between UHRF1 depletion and cell cycle progression. Cells harvested 48 h post-transfection were processed for FACS analysis to assess cell cycle stage. As shown in Figure 1(C), the depletion of UHFR1 in HCT116 cells led to a doubling in the percentage of cells in the G2/M-phase. Western immunoblotting (Figure 1D) showed a corresponding increase of Ser10-phosphorylated histone H3, confirming a block in mitosis. In Hep3B cells depleted of UHRF1, we saw a similar pattern (Supplementary Figure S1A at http://www.BiochemJ.org/bj/435/bj4350175add.htm); indicating the G2/M-phase cell cycle arrest is not cell-type specific.

Previous studies in HeLa cells with reduced UHRF1 levels showed that a small percentage of cells remain blocked in G1-phase following DNA damage [18]. We reasoned that if the loss of UHRF1 causes a G1-phase arrest in HCT116 cells, some UHRF1-depleted cells should be retained in the G1-phase when treated with nocodazole, a drug that induces cell arrest in prometaphase. FACS analyses revealed no difference between UHRF1-depleted and control cells: in both cases, the G1-phase population progressed to the G2/M-phase where they arrested (Figure 1E). Consistent with this, Western immunoblotting showed no changes in the levels of G1- or S-phase cyclins, indicating that there is no cell cycle block in the G1- or S-phases. Instead, we saw an increase in the mitotic cyclin B1 (Figure 1F), suggesting that UHRF1 depletion caused a block after cyclin B synthesis, but before its destruction in mid-mitosis [23,24]. These experiments confirm that there is no G1-phase arrest in UHRF1-depleted HCT116 cells. This observation is reminiscent of the phenotype observed in these same cells depleted of DNMT1 [21].

Loss of UHRF1 activates the DNA damage and response pathway

Cyclin-dependent kinases are regulators of cell cycle progression [23]. Because CDK1 (cyclin dependent kinase 1, also called Cdc2) is required for progression of cells from the G2-phase into and through mitosis, we questioned whether the G2/M-phase block was associated with evidence of CDK1 inhibition. Figure 2(A) shows that inhibitory phosphorylation of CDK1 on Tyr15 is enhanced in HCT116 cells depleted of UHRF1, while total CDK1 remains constant (Figure 2A). A similar elevation of Tyr15 CDK1 phosphorylation was also seen in Hep3B cells transfected with UHRF1 siRNA (Supplementary Figure S1B). Taken together, these data indicate that in at least two different cancer cell lines depletion of UHRF1 caused cell cycle arrest in the G2/M-phase.

Figure 2 Depletion of UHRF1 activates the DNA damage response pathway

(A) Western immunoblotting shows similar levels of total cellular CDK1 (panel 2), but there are enhanced levels of Tyr15-phosphorylated CDK1 in UHRF1-depleted cells (panel 3, compare lanes 2 and 3 with 1). (B) Additional markers of the DNA damage response are present in UHRF1-depleted cells. Total CHK2 is unchanged (panel 2), but CHK2 is phosphorylated on Thr68 (panel 3), CDC25C is phosphorylated on Ser216 (panel 4) and histone H2AX is phosphorylated on Ser139 (panel 5). β-Actin is used as a loading control for the experiments above. (C) Knockdown of CHK2 in UHRF1-depleted cells reduces Tyr15 phosphorylation. The increased Tyr15 phosphorylation of CDK1 seen in UHRF1-depleted cells (lane 2, panel 4) is reduced when CHK2 is concurrently knocked down (lane 4, panel 4). si-CHK2, siRNA against CHK2. (D) Caffeine abrogates the G2/M-phase checkpoint in UHRF1-depleted cells. The marked Tyr15 phosphorylation of CDK1 in UHRF1-depleted cells (panel 3, lane 3) is absent in UHRF1-deficient cells treated with caffeine (CAF). CDK1 levels for NTS-transfected cells are similar (panel 2, lanes 1 and 2) as well as for si-A-transfected cells (panel 2, lanes 3 and 4). (E) Representative FACS analysis of caffeine-treated cells. Percentage of cells at various stages of the cell cycle following caffeine treatment. NTS represents UHRF1-containing cells, while si-A represents UHRF1-depleted cells. + and - represent the presence or absence of caffeine. Error bars represent S.D. from the mean of replicates. Note that caffeine abrogates the increase in G2/M-phase population seen with UHRF1 depletion. NTS, non-targeting siRNA.

The G2/M-phase checkpoint is activated as a response to DNA damage. For example, ultraviolet and ionizing irradiation or genotoxic drug treatment of cells causes cells to arrest in order to repair the damage [25]. Because UHRF1 depletion renders cells more sensitive to DNA damaging agents [18,19], we hypothesized that the G2/M-phase block observed in UHRF1-depleted cells was associated with activation of the DNA damage pathway. To address this, we evaluated several markers of the DNA damage response in HCT116 cells depleted of UHRF1. (i) CHK (checkpoint kinase) 1 and CHK2 are activated through phosphorylation by ATM and/or ATR kinase-mediated pathways in response to DNA damage [1]. We did not see any phosphorylation of CHK1, but did find increased Thr68 phosphorylation of CHK2 in UHRF1-depleted cells (Figure 2B). (ii) CDC25 (cell divison control 25) is the phosphatase responsible for removing the inhibitory Tyr15 phosphorylation of CDK1. CDC25 is inactivated when it is phosphorylated on Ser216 by CHK1 and/or CHK2. We found increased phosphorylation of Ser216 on CDC25 in cells depleted of UHRF1, confirming the downstream effect of CHK2 phosphorylation (Figure 2B). (iii) Activation of the DNA damage pathway causes phosphorylation of the histone variant H2AX on Ser139. Western blotting (Figure 2B) showedenhanced H2AX phosphorylation, and immunofluorescence showed positive phospho-H2AX cells (Supplementary Figure S2 at http://www.BiochemJ.org/bj/435/bj4350175add.htm) in UHRF1-depleted cells. These data demonstrate that the DNA damage response system is activated in response to UHRF1 depletion.

We rationalized that if the G2/M-phase block is due to the activation of the DNA damage repair pathway, then the loss of CHK2 in UHRF1-depleted cells should affect phospho-Tyr15 CDK1 levels. Figure 2(C) shows that increased Tyr15 phosphorylation of CDK1 in UHRF1-depleted cells is reduced when CHK2 is concurrently knocked down. The G2/M-phase checkpoint can be abolished in cells following caffeine treatment, probably through inhibition of ATM and ATR [26]. We treated transfected cells with caffeine and then harvested them for Western blotting. Figure 2(D) shows that there is no change in total CDK1 levels whereas Tyr15 phosphorylation of CDK1 is markedly increased in UHRF1-depleted cells (Figure 2D). However, CDK1 phosphorylation is abolished when the cells are treated with caffeine (Figure 2D). In support for the role of this checkpoint in the observed G2/M-phase block, FACS analysis of UHRF1-containing and -depleted cells, treated with or without caffeine, showed that caffeine treatment abolished the increase in the population of the G2/M-phase cells seen with UHRF1depletion (Figure 2E). These data suggest that loss of UHRF1 activates DNA damage signals initiated through ATM and/or ATR, leading to the inhibitory phosphorylation of CDK1 and the subsequent cell cycle block.

Cell cycle arrest is not p53 dependent

In a previous study, adriamycin treatment caused a small population of UHRF1-depleted cells to arrest in the G1-phase, in a p53-dependent manner [18]. p53 is a downstream effector of ATM-mediated DNA damage signalling and co-ordinates DNA repair with cell cycle progression [27]. Activation of ATM/ATR, CHK1 and/or CHK2, leads to p53 phosphorylation and stabilization [28,29]. Interestingly, DNMT1 inactivation also activates and stabilizes p53 [21]. We asked whether the cell cycle arrest and DNA damage response activation in cells depleted of UHRF1 is p53 dependent. We found no difference in total p53 levels in cells depleted of UHRF1 and there was no corresponding increase in the Ser15-phosphorylated moiety (Figure 3A). Next, we assessed whether the cell cycle arrest due to UHRF1 depletion required p53 by knocking down UHRF1 in HCT116 cells devoid of p53 (HCT116 p53−/−). FACS analysis revealed the same increase in the G2/M-phase population in response to UHRF1 depletion (Figure 3B). In support of the FACS analysis data, Western immunoblotting of UHRF1-depleted p53−/− HCT cells showed that Ser10 phosphorylation of histone H3 is increased in UHRF1-knockdown cells (Figure 3C). Western blot analysis of HCT116 p53−/− cells also showed that UHRF1 depletion led to inhibitory phosphorylation of CDK1 and phosphorylation of histone H2AX on Ser139 (Figure 3D). In addition, the abrogation of the G2/M-phase checkpoint by knockdown of CHK2 leads to the alleviation of the cell cycle arrest, as illustrated by the decrease in phospho-histone H3 levels (Figure 3E). These experiments demonstrate that, independent of p53 status, the same DNA damage pathway components are activated in response to UHRF1 depletion.

Figure 3 The cell cycle block in UHRF1-depleted cells is p53 independent

(A) Western immunoblotting of total p53 (panel 2) and Ser15-phosphorylated p53 (panel 3). Total p53 and Ser15-phosphorylated p53 proteins are not increased in HCT116 cells depleted or sufficient for UHRF1. (B). Representative histograms of HCT116 p53−/− cells depleted of UHRF1. Note the increased 4N DNA content in cells transfected with si-A and si-B (panels 2 and 3). (C) Increased phosphoshistone H3 levels in UHRF1-depleted cells. Top panel: p53 is absent in p53−/− cells (compare lanes 2, 3 and 4 with lane 1, arrowhead indicates p53 in WT HCT116 cells). In p53−/− HCT116 cells, knockdown of UHRF1 increases phospho-histone H3 levels (panel 4 compare lanes 2 and 3 to 1). (D) UHRF1 knockdown enhances phosphorylation of CDK1 and enhances phospho-H2AX levels (panels 3 and 4). (E) CHK2 depletion in UHRF1-knockdown cells affects phospho-histone H3 levels. UHRF1 depletion increases phospho-histone H3 levels (panel 3, compare lane 1 and 2). Loss of CHK2 in UHRF1-depleted cells abrogates histone H3 phosphorylation (panel 3, compare lanes 2 and 4). Si-CHK2 represents siRNA against CHK2. β-Actin is used as a loading control. NTS, non-targeting siRNA.

UHRF1-depleted cells undergo a p53-independent apoptosis

We noted that some cells became non-adherent after UHRF1 knockdown. The ATM/ATR-mediated DNA damage signalling pathways can promote apoptosis and thus we examined whether the loss of UHRF1 causes apoptosis in cancer cells, as it does in embryonic cells [15]. Apoptosis can be detected biochemically by assaying for caspase activation through detecting their cleavage products. Caspase 3, an effector caspase downstream of multiple apoptotic stimuli, was detected in its active (cleaved) form in lysates from HCT116 cells depleted of UHRF1 (Figure 4A). Additionally, cleavage of the 116 kDa protein, PARP-1 [poly(ADP-ribose) polymerase-1] to an 89 kDa proteolytic fragment by activated caspase 3 was detected in cells depleted of UHRF1 (Figure 4A). Fluorescence microscopy of UHRF1-depleted cells also showed abundant TUNEL-positive cells (Figures 4B and Supplementary Figure S3 at http://www.BiochemJ.org/bj/435/bj4350175add.htm) as well as numerous cells positive for another marker of apoptosis, annexin V (results not shown). Similar to p53-containing cells, p53-deficient HCT116 cells depleted of UHRF1 also expressed markers of active apoptosis with detection of cleaved PARP-1 (Figure 4C). We conclude that the loss of UHRF1 in human cancer cells results in apoptosis, similar to the finding in non-cancerous embryonic zebrafish liver devoid of uhrf1 [15]. It is also interesting that similar findings were reported in DNMT1-depleted cells. However, unlike the loss of DNMT1 where p53 is stabilized and may be required, apoptosis in response to UHRF1 depletion is at least partly independent of p53.

Figure 4 UHRF1 cells undergo a p53-independent apoptosis

(A) Markers of active apoptosis are present in UHRF1-depleted cells. Caspase 3 is cleaved to generate activated caspase 3 (panel 2, compares lanes 2 and 3 to 1) and PARP is cleaved (panel 4, compare lanes 2 and 3 with 1). (B) TUNEL assays show the presence of TUNEL-positive cells (bottom) in UHRF1-depleted cells. Phase photographs are shown for comparison (top). (C) Apoptosis also occurs in HCT116 p53−/− cells. The second panel shows both the 116 kDa and 89 kDa (arrow) moieties of PARP in UHRF1-depleted cells. Only the 116 kDa PARP is seen in the lysate from NTS-transfected cells (panel 2, lane 1). Blots with an antibody that recognizes the 89 kDa fragment confirms that it is generated from PARP (panel 3). NTS, non-targeting siRNA.

We did not find evidence that other p53 family members (p63 and p73), which share structural homology and overlapping functions with p53 [30], were activated in response to UHRF1 depletion. Neither p63 nor p73 RNA (Supplementary Figure S4A at http://www.BiochemJ.org/bj/435/bj4350175add.htm) or protein (Supplementary Figure S4B) levels were enhanced with UHRF1 depletion. Consistent with this, cells depleted of UHRF1 did not up-regulate the PUMA (p53-up-regulated modulator of apoptosis) gene or protein (Supplementary Figure S4), a downstream target of p53, p63 and p73, and a factor upstream of caspase 9 [31]. Taken together, these results suggest that neither p53, p63 nor p73 are absolute requirements for apoptosis in UHRF1-depleted cells.

UHRF1 depletion causes apoptosis in a caspase 8-dependent manner

Caspase 8, an initiator caspase, can directly activate caspase 3 bypassing the mitochondrial-mediated death pathway [31]. A p53-independent pathway that regulates caspase 8 activity has been described to account for apoptosis seen following irradiation of p53-deficient glioma cells [32]. The presence of activated and cleaved caspase 8 was detected in lysates from UHRF1-depleted cells, but not controls (Figure 5A). Immunofluorescence with antibodies against activated caspase 8 showed a greater than a 10-fold increase in the number of caspase 8-positive cells, when UHRF1-depleted cells were compared with UHRF1-containing cells (Figure 5B). We depleted caspase 8 from UHRF1-knockdown cells and evaluated whether apoptosis still occurred. As expected, the p89 fragment of PARP-1 (Figure 5C) and activated caspase 3 (Figure 5C) increased with knockdown of UHRF1 in caspase 8-containing cells. However in UHRF1-depleted cells, the loss of caspase 8 blocked PARP-1 cleavage and caspase 3 activation (Figure 5C). These data suggest that loss of UHRF1 causes a cell cycle arrest that is irreparable, causing activation of caspase 8 that leads to activation of caspase 3 and cell death.

Figure 5 Apoptosis in UHRF1-depleted cells occurs through the activation of caspase 8

(A) Proteolytic fragments of caspase 8 representing its activated form (p43/p41 and p18) are detected in UHRF1-depleted cells, but not in UHRF1-containing cells (panel 2, compare lanes 2 and 3 with 1). (B) Immunofluorescence photographs (right-hand panel) of UHRF1-containing and -depleted cells show that activated caspase 8 is present in cells lacking UHRF1. Phase photographs are also shown for comparison (left-hand panel). (C) Depletion of total caspase 8 prevents PARP and caspase 3 cleavage. p89 PARP and activated caspase 3 levels increase in UHRF1-depleted, but caspase 8-containing, cells (panels 3 and 4, lane 2) while low in caspase 8-depleted (C8A), but UHRF1-containing, cells (panels 3 and 4, lane 3). When caspase 8 is depleted in cells lacking UHRF1, PARP and caspase 3 cleavage are suppressed (panels 3 and 4, lane 4). Note that ‘Caspase 8’ represents total caspase 8 levels whereas ‘Caspase 8 (activated)’ represents proteolysed forms of caspase 8. NTS, non-targeting siRNA.

5-Azacytidine treatment reproduces some of the effects of UHRF1 depletion

We next conducted studies designed to begin unravelling the mechanisms governing how UHRF1 interacts with the cell cycle. UHRF1 has been reported to bind the inverted CCAAT box in the TOP2A (topoisomerase 2A) promoter and activate its expression [2]. Given that TOP2A plays important roles in DNA replication and chromosomal segregation, events whose disruption can elicit cell cycle blocks, we investigated whether the effects of UHRF1 depletion could be explained through decreased TOP2A levels. Neither the mRNA nor the protein levels of TOP2A were decreased in UHRF1-depleted HCT116 cells (Figures 6A and 6B), nor in HuH7 cells, MCF7 cells or zebrafish embryos lacking uhrf1 (results not shown). In fact, cells transfected with si-A have a modest increase in TOP2A mRNA and protein. These data suggest that TOP2A is unlikely to play a direct role in the cell cycle arrest seen in UHRF1-depleted cells and support the notion that UHRF1 does not directly regulate TOP2A expression.

Figure 6 TOP2A levels are unchanged with UHRF1 depletion

(A) Quantitative PCR assay shows that TOP2A mRNA levels are not decreased in UHRF1-depleted cells. UHRF1 mRNA levels are decreased by 80% for this experiment. Error bars represent S.D. from the mean of three replicates. (B) TOP2A protein levels are not decreased in UHRF1-knockdown cells. Tubulin is used as a loading control. (C) UHRF1 levels remain unchanged in 5-azacytidine (5-AZA)-treated cells (top panel). Total CDK1 and Tyr15-phosphorylated CDK1 also increased (panel 2 and 3), but p89 levels are not increased in 5-azacytidine-treated cells (panel 4). NTS, non-targeting siRNA.

Loss of UHRF1 may lead to increased concentration of hemimethylated DNA in UHRF1-depleted cells. We suspected that such improperly modified DNA may be viewed as defective by the cell cycle machinery, leading to a cell cycle block. If true, treatment of cells with 5-azacytidine, a drug that inhibits DNA methylation, would recapitulate the phenotype of UHRF1 loss. Our findings partially support defective methylation as a variable. In support of a role for methylation, there is an increase in the G2/M-phase population of cells (Supplementary Figure S5 at http://www.BiochemJ.org/bj/435/bj4350175add.htm), as well as an increase in Tyr15 CDK1 content of cells treated with 5-azacytidine (Figure 6C). However, unlike UHRF1 depletion, there is a parallel increase in total CDK1 levels that is not seen in UHRF1-depleted cells. Secondly, there is no increase in the p89 fragment of PARP-1 in 5-azacytidine-treated cells (Figure 6C). Thus failed DNA methylation is one possible mechanism by which UHRF1 induces the DNA damage response and cell cycle arrest; however, it is not likely that this alone accounts for the phenotype seen with UHRF1 knockdown.

UHRF1 is recruited to the sites of DNA injury

Previous studies have shown that the UHRF1-interacting protein DNMT1 is recruited to sites of DNA injury [20]. Because DNMTI is dependent on UHRF1 for interaction with chromatin [3], we questioned whether UHRF1 is recruited to these sites as a possible participant in the repair process. It is also known that the DNA damage response pathway is frequently activated when constituents of the pathway are inactivated. For example, inactivation of Hsu1, Claspin 1 and Chk1 all result in DNA damage repair [3335]. Thus it is possible that UHRF1-depleted cells activate the DNA damage response because UHRF1 participates in this pathway. To address this, we used laser scissors to introduce DNA damage at discrete sites that can then be marked by the presence of phospho-H2AX (Figure 7A). We show that UHRF1 is recruited to these sites and is seen within 5 min, with the peak intensity 30 min after injury. The dynamic nature of UHRF1 in the DNA-damage stripes, and the difference with that of phospho-H2AX, suggest that this is a specific interaction of UHRF1 and the damage site (Figure 7B). To confirm the specificity of UHRF1 for the DNA damage sites, we transfected cells with cDNA encoding FLAG-tagged UHRF1 or empty vector and repeated the experiment with anti-FLAG antibody. Once again we found that UHRF1 (as detected with the anti-FLAG antibody) was seen at the sites of injury in FLAG–UHRF1-expressing cells (Figure 7C, top panel), but not in vector-transfected cells (Figure 7C, bottom panel). These data suggest that UHRF1 senses DNA damage and we speculate that it is recruited to these sites for essential DNA modification linked to repair.

Figure 7 UHRF1 localizes to UV-laser-scissor-induced DNA damage

(A) Phospho-H2AX (pH2AX) accumulates at sites of DNA damage after targeted injury (top row, green). UHRF1 is recruited to these sites within 5 min (second row, red), peaks in intensity at 30 min and then begins to fade. (B) Quantification of UHRF1 localization with pH2AX. The ratio of intensity of UHRF1 at the region of laser-scissors-induced damage (marked by pH2AX) compared with the undamaged region is plotted. Note that the change in intensity of UHRF1 staining is dynamic and differs from that of phospho-H2AX. (C) FLAG–UHRF1 localizes to DNA damage. Cells were transfected with FLAG–UHRF1 (top panel) or empty vector (bottom panel). Cells were subjected to UV laser and processed for immunofluorescence 30 min post-DNA damage with anti-pH2AX and anti-FLAG antibodies. UHRF1 is present in the stripe in transfected cells (top row, column 2), but not in control cells (bottom row, column 2).

DISCUSSION

In the present study, we identify a link between UHRF1, cell cycle progression and DNA damage response. Cancer cells depleted of UHRF1 show an increase in the population of cells with 4N DNA, activation of the DNA damage response and caspase 8-dependent apoptosis. The cell cycle arrest and apoptosis induced by loss of UHRF1 does not appear to require any p53 family members. Lastly, we show that UHRF1 accumulates at sites of DNA injury. These studies suggest a direct link between genomic integrity and UHRF1 function.

A previous report found that cancer cells depleted of UHRF1 arrest in the G1-phase only when treated with adriamycin [18], whereas another showed that reduction of UHRF1 causes arrest in both the G1- and G2-phase [14]. In the present study, we find only a G2/M-phase arrest. If UHRF1 is required for G1/S-phase transition in these cells, then the G1-phase-arrested cells would be revealed when nocodazole is used to trap cells in mitosis. Instead, we find that cells in the G1-phase progressed to the mitotic block in UHRF1-sufficient and -deficient cells. Thus no G1-phase block is present in HCT116 cells depleted of UHRF1. A number of reasons may account for the differences in the cell cycle arrest in our studies when compared with previous studies. (i) The requirement for UHRF1 in cell cycle progression may vary in different cancer cell lines due to differences in mechanisms to bypass cell cycle arrest. (ii) The G1-phase arrest seen in cells with low levels of UHRF1, occurred when cells were concurrently treated with adriamycin, raising the possibility that the arrest may be an indirect effect [18]. (iii) Variations in the degree of UHRF1 knockdown may result in elaboration of different cell cycle phenotypes. Such differences have been seen with DNMT1 depletion, where different levels of DNMT1 loss result in various cell cycle effects [36].

It is noteworthy that we do not see a G2/M-phase block in a higher fraction of our cells. This is partly owing to the fact that only adherent cells are harvested, and thus cells that have died and floating are not analysed. A second reason maybe as a result of the limitations of siRNA-mediated knockdown and possible variation in cell-to-cell levels of UHRF1. Thirdly, it may be that only a small population of cells that have gone through the S-phase and contain inadequately repaired DNA are candidates for cell cycle block and death.

The results of the present study have a number of similarities with the consequences of DNMT1 inactivation in HCT116 cells [21]. In both cases, cells arrest with a 4N DNA content and the DNA damage response system is activated. In addition, both UHRF1-depleted cells and DNMT1-inactivated cells undergo apoptosis. The similarities between DNMT1 inactivation and UHRF1 depletion, however, differ in one aspect. In cells devoid of DNMT1, p53 levels are increased and the phospho-Ser15 p53 level is elevated [21]. These data suggest that p53 may be integral to the DNA damage activation and the subsequent cell cycle block. However, we do not find any evidence that the response to low levels of UHRF1 involves the stabilization of p53, nor do we find any qualitative evidence that p53 is required for the cell cycle effects of UHRF1 loss. This difference may be attributed to the fact that in our experiments, UHRF1 levels are substantially reduced but not abolished, as is the case for the experiments on DNMT1 in which cells are completely devoid of DNMT1. Alternatively, since UHRF1 has multiple activities, that include but are not limited to driving DNMT1-dependent DNA methylation, loss of UHRF1 may abrogate both DNA methylation as well as other processes. Indeed, insights from the crystal structure of the SRA domain in complex with DNA suggest that in addition to methylation, UHRF1 may also allow other enzymes access to the double helix [12]. Lastly, it remains to be tested whether the apoptosis observed owing to DNMT1 depletion depends on p53.

The apoptosis in UHRF1-depleted cells supports our data that p53 is dispensable given a known role for caspase 8 in a pathway that bypasses p53 [32]. As noted, cell death also occurs in DNMT1-deficient cells [21]. Given the functional parallels between two proteins, it is tempting to speculate that cell death in DNMT1-deficient cells may also occur through a caspase 8-dependent process. In addition, the results of the present study suggest a relationship between UHRF1 and caspase 8 activity and the mechanism of this regulatory effect needs to be examined.

How then does UHRF1 depletion lead to activation of the DNA damage response and the subsequent arrest and apoptosis? We do not believe that this effect is related to the putative effects of UHRF1 on TOP2A expression, since we find no evidence thatUHRF1 directly regulates TOP2A expression. We reasoned that high concentrations of hemimethylated DNA, which probably accumulate in UHRF1-depleted cells, may be viewed by the cell as a loss of genomic fidelity. We show that treatment of cells with 5-azacytidine recapitulates some, but not all, of the phenotypes of UHRF1 loss and suggests that DNA methylation defects may play at least a partial role. We find that UHRF1 is recruited to sites of DNA injury, enhancing our argument that it plays a role in DNA damage repair. This observation is supported by reports linking UHRF1 to DNA damage. In one study, repair factors including PARP-1, XRCC1 (X-ray repair cross-complementing factor 1), TopBP1 [topoisomerase (DNA) II-binding protein 1] and PCNA (proliferating-cell nuclear antigen) were found in complex with UHRF1 [4]. Additionally, Eme1, a component of an endonuclease required for damage repair, is reported to associate with UHRF1 [37]. Therefore loss of UHRF1 may lead to a defective replication product by any of these mechanisms, activating the DNA damage response pathway and arresting the cell cycle in the G2/M-phase (Figure 8). While this is an intriguing finding, we are cognizant of the fact that it begs the question of what UHRF1 does at the damage site. This will be the focus of further studies.

Figure 8 Proposed model of the effects of UHRF1 depletion in HCT116 cells

In summary, in the present study we provide molecular evidence linking UHRF1 and DNA damage repair. Our results suggest that UHRF1 loss leads to genomic defects that are sensed by the cell cycle apparatus and result in cell-cycle block and cell death. Our data, combined with existing evidence that cells lacking UHRF1 are more sensitive to DNA damaging agents [19], provides a basis for investigating UHRF1 as a possible target in cancer.

AUTHOR CONTRIBUTION

Amy Tien, Sucharita SenBanerjee, Atul Kulkarni, Raksha Mudbhary, Bernadette Goudreau and Chinweike Ukomadu performed experiments. Amy Tien and Chinweike Ukomadu prepared manuscript. Chinweike Ukomadu, Kirsten Sadler and Shridar Ganesan edited the manuscript prior to submission.

FUNDING

A.T. was supported by a National Institute of Diabetes and Digestive and Kidney Disease Brigham and Women's Hospital T32 grant. B. G. was supported by the Four Directions Summer Research Program. CU and KS were supported by the National Institute of Diabetes and Digestive and Kidney Diseases [grant number 1R01DK080789]. C.U. was supported by the Dana-Farber Cancer Institute/Harvard Cancer Center, National Cancer Institute Cancer Center Support Grant and Sidney A. Swensrud Foundation grants.

Abbreviations: ATM, ataxia telangectasia mutated; ATR, ataxia telangectasia mutated- and Rad3-related; CDC25, cell division control 25; CDK1, cyclin-dependent kinase 1; CHK, checkpoint kinase; DNMT1, DNA methyltransferase 1; FBS, fetal bovine serum; HSLB, high-salt lysis buffer; NP40, Nonidet P40; PARP-1, poly(ADP-ribose) polymerase-1; PHD, plant homeodomain; PI, propidium iodide; qPCR, quantitative PCR; siRNA, short interfering RNA; SRA, SET-and-RING-finger-associated domain; TOP2A, topoisomerase 2A; TUNEL, terminal deoxynucleotidyltransferase-mediated dUTP nick-end labelling; UHRF1, ubiquitin-like protein containing PHD and RING finger domains 1

References

View Abstract