The BMI1 Polycomb Protein Represses Cyclin G2-Induced Autophagy to Support Proliferation in Chronic Myeloid Leukemia Cells

ORIGINAL ARTICLE The BMI1 polycomb protein represses cyclin G2-induced autophagy to support proliferation in chronic myeloid leukemia cells

L Mourgues1,2,3, V Imbert1,2,3, M Nebout1,2,3, P Colosetti1,3,4, Z Neffati1,2,3, P Lagadec1,2,3, E Verhoeyen1,4,5, C Peng6, E Duprez7, L Legros3,8,9, N Rochet3,9, V Maguer-Satta10, F-E Nicolini11, D Mary1,2,3,12 and J-F Peyron1,2,3,12

The BMI1 polycomb protein regulates self-renewal, proliferation and survival of cancer-initiating cells essentially through epigenetic repression of the CDKN2A tumor suppressor locus. We demonstrate here for the ﬁrst time that BMI1 also prevents autophagy in chronic myeloid leukemia (CML) cell lines, to support their proliferation and clonogenic activity. Using chromatin immunoprecipitation, we identiﬁed CCNG2/cyclin G2 (CCNG2) as a direct BMI1 target. BMI1 downregulation in CD34+ CML cells by PTC-209 pharmacological treatment or shBMI1 transduction triggered CCNG2 expression and decreased clonogenic activity. Also, ectopic expression of CCNG2 in CD34+ CML cells strongly decreased their clonogenicity. CCNG2 was shown to act by disrupting the phosphatase 2A complex, which activates a PKCζ-AMPK-JNK-ERK pathway that engages autophagy. We observed that BMI1 and CCNG2 levels evolved inversely during the progression of CML towards an acute deadly phase, and therefore hypothesized that BMI1 could support acute transformation of CML through the silencing of a CCNG2-mediated tumor-suppressive autophagy response.

Leukemia (2015) 29, 1993–2002; doi:10.1038/leu.2015.112

INTRODUCTION of BMI1 − / − phenotypes, suggesting the implication of other The BMI1 polycomb and transcriptional repressor exerts BMI1-controlled loci. an epigenetic control over cell fate, development and The CDKN2A locus is one of the most frequently lost gene in 1 human cancer (30% of cases), in particular during cancer transformation. BMI1 is preferentially expressed in stem cells 13 such as hematopoietic stem cells (HSCs),2 to support self-renewal. progression. The chronic myeloid leukemia (CML) blast crisis is Transgenic mice expressing BMI1 have enhanced self-renewal of associated with the acquisition of additional genetic/epigenetic HSC,3 whereas inactivation of BMI1 results in long-term hemato- defects and in particular with loss of CDKN2A function as the poiesis failure because of impaired HSC self-renewal.3–5 BMI1 was promoters of p16INK4A and p14ARF genes are frequently silenced first identified as an oncogene, cooperating with c-MYC to induce by methylation during disease progression from the chronic phase (CP) to the accelerated phase (AP).14,15 Meanwhile, expression of mouse lymphomas.6 Overexpression of BMI1 in a lymphocyte BMI1 in CML primitive cells increases gradually during CML specific manner induces T-cell lymphomas.7 BMI1 is also crucial for transformation,16 which strongly suggests that BMI1 controls leukemogenesis as BMI1 − / − HSC, transformed with Hoxa9 and other important, tumor suppressor, loci. Meis1a genes, failed to generate leukemia into secondary hosts 4 We used the K562 CML cellular model that harbors a deletion of because of proliferation arrest, differentiation and apoptosis. the CDKN2A locus17 to search for new modes of action for BMI1. It has recently been demonstrated that pharmacological targeting Here, we demonstrate that BMI1 represses an autophagic of BMI1 interferes with colon cancer tumor formation by affecting 8 response that interferes with proliferation and clonogenicity and self-renewal of cancer-initiating cells. depends on the atypical cyclin G2. These results highlight a new A major target of BMI1 is the CDKN2A locus that encodes the function for BMI1 that could be relevant to the physiopathology of tumor suppressor genes p16INK4A and p14ARF. By repressing leukemic cells. CDKN2A, BMI1 prevents engagement of RB1 and TP53 pathways, allowing progression through the cell cycle, maintenance of stemness and restriction of differentiation.9 If the dual deletion of MATERIALS AND METHODS Ink4A and Arf genes almost completely restores BMI1 − / − HSC Cells 10 fi 11 survival, prevents senescence of BMI1-de cient MEFs and The human CML lines JURL-MK1, LAMA-84 and K562 (ATCC, Molsheim, blocks transformation of hematopoietic progenitors by the France) were grown at 37 °C under 5% CO2 in RPMI 1640 medium E2a-Pbx1 oncogene,12 it did not always lead to a complete rescue supplemented with 5% fetal calf serum, 50 U/ml penicillin, 50 μg/ml

1INSERM, UMR1065 Centre Méditerranéen de Médecine Moléculaire C3M, Nice, France; 2Equipe Inﬂammation, Cancer, Cancer Stem Cells, Nice, France; 3Université de Nice-Sophia Antipolis, UFR Médecine, Nice, France; 4Equipe Cell Death, Differentiation and Cancer, Nice, France; 5Equipe Metabolic control of Cell Death in Cancer, Nice, France; 6Department of Biology, York University, Toronto, ON, Canada; 7CRCM, U1068, INSERM, Marseille, France; 8Service d'Hématologie Clinique, Hôpital Archet 1, Nice, France; 9Institut de Biologie Valrose (iBV), UMR CNRS 7277-UMR INSERM 1091, Nice, France; 10CNRS, UMR5386, INSERM, U1052, Lyon, France and 11Service d’Hématologie, Hospices Civils de Lyon, Centre Hospitalier Lyon Sud, Pierre Bénite, France. Correspondence: Dr D Mary or Dr J-F Peyron, INSERM, UMR1065 Centre Méditerranéen de Médecine Moléculaire C3M, 151 Route de Ginestière BP 2 3194, Nice 06204, France. E-mail: [email protected] or [email protected] 12These authors share senior coauthorship. Received 8 September 2014; revised 27 March 2015; accepted 30 March 2015; accepted article preview online 30 April 2015; advance online publication, 22 May 2015 BMI1 blocks autophagy through cyclin G2 silencing L Mourgues et al 1994

streptomycin and 1 mM sodium pyruvate. K562 cells (Tet-K562) expressing Lentiviral transduction the tetracycline repressor by stable transfection of the pCDNA6-TR plasmid CD34+ cells (106/ml) were ﬁrst maintained 24 h in StemSpan medium 18 (Invitrogen, Life Technologies SAS, Saint Aubin, France) was described. All supplemented with 50 ng/ml stem cell factor, interleukin-3, FMS-like parental cell lines were originated from ATCC; derivatives and parental tyrosine kinase 3-ligand and thrombopoietin. Lentiviral particules cell lines were authenticated and are regularly tested for mycoplasma (Supplementary Experimental Procedures) were added (multiplicity of contamination. Primary samples were obtained from CML patients at infection of 100) with 8 μg/ml of hexadimethrine bromide (Sigma, St Louis, diagnosis and during follow-up after informed consent in accordance with MO, USA). A second transduction step was realised 12 h later. The medium the Declaration of Helsinki and local ethics committee bylaws (from the was replaced the next day and transduction efﬁciency was evaluated at Délégation à la recherche clinique des Hospices Civils de Lyon, France). day 3 by GFP detection.

Primary cell isolation BMI1 silencing by short hairpin RNA interference Bone marrow samples were collected from patients newly diagnosed for Two short hairpin RNA (shRNA)-expressing plasmids were constructed using CML as part of an institutionally approved cellular sample collection the following sequences: sh-A: 5′-CAACCAGAATCAAGATCACTGA-3′;sh-B: protocol (Centre Hospitalier Nice, France). Mononuclear cells were isolated 5′-GGAAGAGGTGAATGATAAA-3′. Oligonucleotides (Eurogentec, Seraing, by density centrifugation (Ficoll-Paque Plus; StemCell, Vancouver, BC, Belgium) were annealed and cloned into the pTER vector using BglII and Canada), washed with phosphate-buffered saline and 5% fetal calf serum. HindIII restriction sites. The resulting Tet-on-inducible shRNA vector was then CML cells were labeled with CD34 microbeads isolated by magnetic- stably transfected into Tet-K562 cells (100 μg/ml zeocin) (A1-K562). positive selection (CD34 MicroBead Kit; Miltenyi, Paris, France). Experi- ments were performed using a StemSpan SFEM medium (StemCell) supplemented with 100 ng/ml stem cell factor and FMS-like tyrosine kinase Chromatin immunoprecipitation assay 3-ligand and 20 ng/ml human recombinant interleukin-3, interleukin-6 and Chromatin immunoprecipitation (ChIP) assay was performed using the granulocyte colony-stimulating factor (Miltenyi). SimpleChIP Enzymatic Chromatin IP Kit (Cell Signaling, Danvers, MA, USA)

35 ) BMI1 100 100 30 HSP60 37%

) 80 80

6 _ _ 25 / +/ DOXY 41% 20 60 60 shBMI1

40 BrdU 40

15 WST-1 10 number (10 20 20 untreated condition Accumulated cell 5 0 0 DOXY DOXY (% of 0 (% of untreated condition) 0 5 10 shCT shBMI1 shCT shBMI1 days

shBMI1 untreated 100 80 NT NT 60 40 20 (% of control)

Number of colonies 0 PMA+Sb DOXY

1 2 3 1 2 3 Replating

K562 shBMI1 shCT

NT PMA+Sb shCT shBMI1 Figure 1. Knockdown of BMI1 inhibits biological parameters in K562 cells. (a) A1-K562 cells expressing shBMI1 were stimulated or not by 1 μg/ml doxy for 20 days. Cells were washed two times and resuspended in fresh medium with or without doxy before counting. (b) Metabolic activity of A1-K562 cells measured 10 days after stimulation or not with doxy. (c) Proliferation analysis of A1-K562 cells (bromodeoxyuridine (BrdU) immunoassay) 4 days after stimulation or not with doxy. (d) Doxy (1 μg/ml) was added or not to A1-K562 cells seeded in methylcellulose medium (3 × 103 cells per ml). Colonies were counted after 7 days. Results are % of control ± s.d. of three different determinations. Every 7 days, cells from each condition were resuspended in fresh medium before replating without/with doxy. Error bars represent s.d. for three biological replicates. *Po0.05; Student's t-test. (e) K562 cells were left untreated (NT) or incubated with a combination of PMA (phorbol-12- myristate-13-acetate; 10 ng/ml) and Sb202190 (Sb, 5 μM) for 24 h to induce vacuolization (arrows), detected after May–Grunwald–Giemsa (MGG) staining. A1-K562 cells with a control shRNA (shGFP) or a shBMI1 were stimulated or not with doxy for 5 days. (f) K562 cells stably expressing RFP-LC3 (red) were left untransfected (NT), stimulated by PMA+Sb (24 h), transfected with a control shRNA (shGFP) or a shBMI1 (72 h). Cells were analyzed using a confocal microscope. DAPI (4',6-diamidino-2-phenylindole) was used to visualize nuclei.

Leukemia (2015) 1993 – 2002 © 2015 Macmillan Publishers Limited BMI1 blocks autophagy through cyclin G2 silencing L Mourgues et al 1995 according to the manufacturer’s protocol. Details are provided in RESULTS Supplementary Experimental Procedures. Inducible knockdown of BMI1 affects proliferation, metabolism, clonogenicity and induces an autophagic response in K562 human CML cells Statistical analysis Cells of the K562 human CML line were engineered to stably Continuous variables and binomial variables, expressed as mean (s.d.), a two-tailed Student's t-test for equivocal variance with Po0.05 deemed as express the Tet repressor and a pTer vector coding an shRNA statistically signiﬁcant. All experiments were repeated with three different against BMI1 (A1-K562 cells) to allow for an inducible down- cell cultures, and WST-1, DNA synthesis and colony formation assays were regulation of BMI1 with doxycycline (doxy). An important decrease performed in quadruplicate. For primary cell experiments, statistical in A1-K562 cell proliferation was observed upon BMI1 knockdown, analyses were performed when n was greater than 5. The nonparametric evaluated by counting viable cells over a 20-day period Mann–Whitney U-test was performed to assess the difference in tumor (Supplementary Figure S1A). After 20 days of culture, A1-K562 volume between control and treatment group in xenograft experiments. cells were washed out and put back into culture (Figure 1a).

2 1.5 CCNG2 04567days 1 50 BMI1 0.5 50 CCNG2 0 -0.5 CCNG1 37 -1 BMI1 60 HSP60 mean log 2 difference -1.5 kDa -2 02 4 6 8 10 12 14 16 mean log 2 intensity

3 shCT shBMI1 NT DOXY (shBMI1) 2.5 2 1.5 DAPI 1

μ μ 0.5 0 CD34+ mRNA expression (a.u) BMI1 CCNG2 BMI1 CCNG2 BMI1 CCNG2

BMI1 Patient #1 Patient #2 Patient #3

100

CCNG2 80

48% n = 5 60 42%

38%37% 40 30% n = 6 100 20 80 12% 53% 54% 11% 40% 60 40% 38%

40 CD34+ colony-forming cells (% of control) 0 20 PTC-209 1 μM (% of control) 0 ima 1 μM sh BMI1 ima 0.1 μM μ CD34+ colony-forming cells ima 0.1 M nilo 2 nM nilo 0.2 nM nilo 0.2 nM Figure 2. Knockdown of BMI1 directly regulates CCNG2 expression in K562 cells. (a) Dot plot of gene expression in A1-K562 after shBMI1 induction by doxy (1 μg/ml) for 4 days. (b) A1-K562 cells were stimulated or not with doxy for indicated time. Cell lysates were analyzed by speciﬁc immunoblotting. (c) Immunoﬂuorescence detection of BMI1 (green) and CCNG2 (red) in A1-K562 cells stimulated or not with doxy for 4 days. (d) CD34+ cells isolated from three CML patients were transduced with a lentiviral vector expressing an shRNA against BMI1 or a scrambled shRNA (shCT). Gene expression was studied 4 days later by quantitative real-time-PCR (qRT-PCR) analysis. (e) Clonogenic capacity of primary cells from (d) was examined after 14 days, in the presence or not of suboptimal doses of imatinib (0.1 μM) or nilotinib (0.2 nM). (f) Clonogenic capacity of CD34+ CML cells, treated with PTC-209 alone or in combination with imatinib or nilotinib, was examined after 14 days. Error bars represent s.d. for three biological replicates. *Po0.05; Student's t-test.

© 2015 Macmillan Publishers Limited Leukemia (2015) 1993 – 2002 BMI1 blocks autophagy through cyclin G2 silencing L Mourgues et al 1996 Removal of doxy resulted in a complete recovery of a normal BMI1 directly represses cyclin G2 expression proliferation rate. BMI1 knockdown in A1-K562 cells decreased To visualize BMI1 target genes a gene proﬁling experiment was both cell metabolism (Figure 1b) and proliferation (Figure 1c). performed on the A1-K562 CML cell line that harbors a deletion of ShBMI1-inducible cells were implanted in the ﬂanks of nude mice CDKN2A. cDNA from A1-K562 cells, incubated or not with doxy to treated intraperitoneally every 3 days with doxy, which was also downregulate BMI1, were hybridized to Affymetrix arrays. Among present in drinking water. Doxy strongly affected tumor develop- several upregulated genes (Figure 2a), we selected the tumor ment (Supplementary Figure S1B). Induction of the shBMI1 in cells suppressor CCNG2/cyclin G2. The decrease in BMI1 was associated was associated with a 40% decrease in the number of colonies in a with a higher expression of CCNG2 in vitro (Figure 2b and semisolid methylcellulose assay (Figure 1d). After 7 days, cells from Supplementary Figure S2A) and in vivo in xenografted A1-K562 each condition were seeded at the same density for serial tumors from mice treated with doxy (Supplementary Figure S2B). A replating. Doxy gradually decreased A1-K562 colony numbers confocal analysis (Figure 2c) clearly showed a nuclear localization of after three replating steps. The absence of senescence and BMI1 in untreated A1-K562 cells, whereas CCNG2 was barely apoptotic processes (not shown), as well as the reversibility of detected. The induction of shBMI1 was associated with a strong the biological events that follow BMI1 silencing, prompted us to decrease in BMI1 levels with the appearance of cytoplasmic CCNG2. investigate autophagic events. After shBMI1 induction in In contrast, ectopic expression of BMI1 in K562 cells was associated A1-K562 cells, we detected the presence of cells with vacuoles with markedly decreased expression of CCNG2 (Supplementary by MGG (May–Grunwald–Giemsa) staining (Figure 1e). Figure S2C). A ChIP assay showed that BMI1, within the PRC1 Vacuolated cells were also detected after stimulation with complex, binds to the CCNG2 promoter near the transcription start PMA (phorbol-12-myristate-13-acetate) and SB202190 p38 site (Supplementary Figure S2D). This mechanism was not restricted mitogen-activated protein kinase inhibitor, a combination that to the K562 cell line, as transient knockdown of BMI1 in the JURL- had been shown to trigger autophagy in K562 cells.18 Besides, MK1 and LAMA-84 human CML cell lines by two shRNA species also by using K562 cells expressing an RFP-LC3, we observed that resulted in a higher expression of CCNG2 (not shown). Besides, downregulation of BMI1 resulted in the appearance of LC3-II downregulation of BMI1 in primary CD34+ CML cells, by a lentivirus foci that correspond to the relocalization of RFP-LC3 into vector expressing shRNA, also induced the expression of CCNG2 autophagosomes (Figure 1f). (Figure 2d). PTC-209, a new BMI1 inhibitor, induced a strong

abc CT siCCNG2 * * * * 100 16% 100 100 20% 40% 37% 80 * 80 80 42% 60 * 60 60 67% Brdu

40 40 WST-1 40

(% of control) 20 20 20

Number of colonies 0 0 0 (% of untreated condition) (% of untreated condition) - + - + shBMI1 -+ - + shBMI1 12 12 CT siCCNG2 CT siCCNG2 Replating 99% Rescue 46% Rescue

deCT siCCNG2 CT siCCNG2 - + - + shBMI1

NT 44 BMI1 42

50 CCNG2

62 P-AMPKα shBMI1 (T172)

62 AMPKα

18 LC3-I 16 LC3-II PMA+Sb 90 HSP90 kDa

Figure 3. Expression of CCNG2 is required to inhibit proliferation, metabolism and clonogenicity and to promote autophagic process after BMI1 silencing. (a–c) Biological analysis of A1-K562 cells stably expressing an empty vector (CT) or an siRNA against CCNG2 (siCCNG2). Doxy was added when indicated for 4 days. Quantiﬁcation of clonogenicity (a), proliferation (b) and metabolic activity (c). (d) A1-K562 cells with empty vector (CT) or siCCNG2 were stimulated or not with doxy for 5 days. Lower panel: K562 cells were stimulated by a combination of PMA (phorbol-12-myristate-13-acetate; 10 ng/ml) and Sb202190 (Sb, 5 μM) for 24 h to induce vacuolization (arrows). Cells were stained by MGG (May–Grunwald–Giemsa). Error bars represent s.d. for three biological replicates. *Po0.05; Student's t-test. (e) Cells lysates of A1-K562 cells with empty vector (CT) or siCCNG2 stimulated 4 days or not with doxy were examined by immunoblotting. Error bars represent s.e.m. for three biological replicates. *Po0.05; Student's t-test.

Leukemia (2015) 1993 – 2002 © 2015 Macmillan Publishers Limited BMI1 blocks autophagy through cyclin G2 silencing L Mourgues et al 1997 increase of CCNG2 expression in K562 cells (Supplementary Figure sharp contrast, the metabolic activity was rescued by only 46% S2E) associated with a dose-dependent decrease in clonogenic (Figure 3c). We then examined the implication of CCNG2 in the capacity (Supplementary Figure S3A). PTC-209 also induced autophagic process observed after BMI1 silencing. The extinction autophagic events in K562 cells such as vacuolization of CCNG2 markedly decreased the number of vacuolated cells (Supplementary Figures S3B–D). PTC-209 alone or in combination observed in doxy-stimulated A1-K562 cells (Figure 3d). Besides, with imatinib did not apppear cytotoxic in primary cells from an downregulation of BMI1 was associated with an increase in healthy donor (Supplementary Figure S3E). PTC-209 also induced an phosphorylation of AMPKα and a conversion of LC3-I into LC3-II, a increase of CCNG2 expression in CML primary cells (Supplementary hallmark of autophagy (Figure 3e). Remarkably, silencing of Figure S3F). Both shBMI1 lentiviral vector and PTC-209 reduced the CCNG2 totally abrogated these two events. To further investigate clonogenic capacity of primary CD34+ CML cells (Figures 2e and f). the direct implication of CCNG2 on cellular functions, we Suboptimal doses of imatinib and nilotinib combined with BMI1 developed a CCNG2-inducible model in K562 cells (CCNG2-TO) downregulation by either shRNA or PTC-209 did not show additivity (Supplementary Figure S5A). Induction of ectopic CCNG2 reduced (respectively, Figures 2e and f). On the other hand, imatinib and DNA synthesis (Figure 4a), cell metabolism (Figure 4b) and nilotinib inhibitors induced similar effects on BMI1 and CCNG2 clonogenic capacity of the cells (Figure 4c). This last effect was expression in CML primary cells (Supplementary Figure S3F), which reversible as washing the cells out from doxy before replating could explain the lack of additivity. restored a normal colony-forming capacity (Figure 4c). Lentiviral expression of CCNG2 gene also reduced clonogenic capacity of + CCNG2 mediates the antiproliferative effects of BMI1 knockdown primary CD34 CML cells (Supplementary Figure S5B). CCNG2 To verify that CCNG2 is implicated in the biological functions overexpression did not affect the action of suboptimal doses of regulated by BMI1, A1-K562 cells were stably transfected by a GFP- imatinib or nilotinib tyrosine kinase inhibitor. siCCNG2 vector (Supplementary Figures S4A and B). Silencing of CCNG2 strongly rescued the clonogenic ability of BMI1- CCNG2 induces autophagy in K562 cells knockdown cells (Figure 3a) and completely abolished the Stimulation of the CCNG2-TO cells with doxy resulted in an antiproliferative effect of BMI1 downregulation (Figure 3b). By important cell vacuolization (Figure 4d). Ectopic CCNG2 expression

119% 120 100 100 100% 80 80 100 80 60 60 51% 60 49% BrdU 40 WST1 40 40

20 20 (% of control) 20

0 0 Number of colonies 0 CCNG2 CCNG2 +/ +/ CCNG2 (% of untreated condition) (% of untreated condition) Replating

CCNG2 AICAR 50 CCNG2 (myc-Tag)

62 P-AMPKα (T172)

62 AMPKα

1 1.25 0.4 2.25 Ratio P-AMPKα /AMPKα NT CCNG2 59 BECN1

55 ATG5

70 ATG7

60 ATG12-ATG5

90 HSP90 pcDNA4 CCNG2 kDa Figure 4. Ectopic CCNG2 inhibits proliferation, metabolism and clonogenicity and promotes autophagy in K562 cells. (a and b) CCNG2-TO cells stimulated for 4 days were analyzed for proliferation (a) and metabolic activity (b). (c) Clonogenic capacity of CCNG2-TO cells was examined after 7 days, under CCNG2 induction (doxy 1 μg/ml) or not. Cells from the doxy-stimulated condition (+) were replated for 7 additional days with (+/+) or without (+/ − ) doxy. Error bars represent s.d. for three biological replicates. *Po0.05; Student's t-test. (d) Visualization of vacuolization (MGG (May–Grunwald–Giemsa) staining) in CCNG2-TO cells treated 3 days without or with doxy. (e) CCNG2-TO cells cultured for 3 days without or with doxy were stimulated with AICAR (0.5 mM) for the last 18 h when mentioned. Expression of indicated proteins was visualized by immunoblotting. (f) K562 cells stably expressing RFP-LC3 (red) were transfected with an empty vector (pcDNA4) or a vector expressing CCNG2 (72 h). Cells were analyzed using a confocal microscope. DAPI (4',6-diamidino-2-phenylindole) was used to visualize nuclei.

© 2015 Macmillan Publishers Limited Leukemia (2015) 1993 – 2002 BMI1 blocks autophagy through cyclin G2 silencing L Mourgues et al 1998 was also associated with (i) stimulation of AMPK phosphorylation, CCNG2-TO cells, the Aa and Ac subunits of the PP2A complex were (ii) increased protein levels for beclin, ATG5, ATG7 and ATG12-ATG5 associated, whereas PKCζ interacts only with the PP2Aa subunit complexes (Figure 4e) and (iii) induction of LC3-II foci (Figure 4f). (Supplementary Figure S6A). Addition of doxy was followed by a These results demonstrate that CCNG2 can induce several crucial strong association of ectopic CCNG2 with the PP2Ac subunit, with molecular steps involved in the initiation of autophagosomes a concomitant separation of the two PP2A subunits and a release formation at the onset of autophagy. Furthermore, the p62/SQSTM1 of PKCζ from Aa subunit. At the same time, we observed that adaptor that mediates recruitment of LC3 into autophagosomes ectopic CCNG2 expression was associated with an increased was shown to accumulate in the presence of bafilomycin A1 phosphorylation of PKCζ, ERK1/2 and JNK2/3 (Figure 5a). Small- (Supplementary Figure S5C) arguing that CCNG2 stimulated an molecule inhibitors were then used to assess the implication of autophagic flux. Knockdown of AMPK (Supplementary Figure S5D) each of the three kinase pathways. The blockade of the JNK counteracted the effects of CCNG2 induction on proliferation pathway strongly rescued the effects of CCNG2 induction on the decrease in DNA synthesis and in clonogenicity, whereas (Supplementary Figure S5F), metabolism (Supplementary Figure fi S5G) and clonogenicity (Supplementary Figure S5H). Silencing inhibition of AMPK and of ERK1/2 were slightly less ef cient BECN1 (Supplementary Figure S5E) appeared more efficient to (Supplementary Figures S6B and C). An inhibitory cell-permeant PKCζ pseudosubstrate, PKCζPS,21 prevented phosphorylation of rescue these parameters (Supplementary Figures S5F–H). These AMPK, ERK1/2 and JNK2/3 (Figure 5a) and strongly rescued the results show that by acting on AMPK and BECN1, CCNG2 can inhibitory effects of ectopic CCNG2 expression on DNA synthesis engage the cells in an autophagic response. (Figure 5b) and clonogenicity (Figure 5c). Furthermore, cell vacuolization observed following CCNG2 induction did not occurr Association of CCNG2 with the phosphatase 2A complex induces in the presence of PKCζPS (not shown). Altogether, these results activation of PKCζ that stimulates autophagic pathways support a key mediator role of PKCζ in the autophagic flux CCNG2 was described to bind phosphatase 2A (PP2A)19 that regulated by AMPK, JNK and ERK1/2 kinases downstream was demonstrated to control activation of PKCζ.20 In untreated expression of CCNG2.

PKCζPS none * -+-+ CCNG2 100 80 50 CCNG2 (myc-Tag) 60

90 BrdU 40 ζ 62 P-PKC (T410) 20 0

ζ (% of untreated condition ) - + - + CCNG2 62 PKC none PKCζPS 1 1.5 1 1.5 Ratio P-PKCζ/PKCζ 90% Rescue

62 P-AMPKα (T172)

62 AMPKα

1 1.2 1 0.5 Ratio P-AMPKα /AMPKα

54 P-JNK 2/3 (T183/Y185) * P-JNK 1 (T183/Y185) 100% 95% 46 100 80 * 1 1.8 1 1 Ratio P-JNK 2/3 /HSP90 59% 60

44 40 42 P-ERK1/2 (T202/Y204)

(% of control) 20

44 Number of colonies 0 42 ERK1/2 -+ -+CCNG2 none PKCζPS 1 1.3 1 0.7 Ratio P-ERK1/2 /ERK1/2 87% Rescue

90 HSP90

kDa Figure 5. CCNG2 binds PP2Ac to release PKCζ that stimulates AMPK, JNK and ERK pathways required for autophagy. (a) CCNG2-TO cells stimulated without/with doxy, and were incubated in the presence of PKCζPS (5 μM) for the last 6 h. Phosphorylation and expression levels of indicated proteins were measured by western blotting. (b and c) Biological analyses of CCNG2-TO cells incubated without/with doxy in the presence of PKCζPS (5 μM) for the last 18 h. Proliferation measured after 3 days (b) and clonogenicity after 7 days (c). The extent of the rescue effect induced by PKCζPS is indicated. Error bars represent s.d. for three biological replicates. *Po0.05; Student's t-test.

BMI1 CCNG2 BMI1 CCNG2 BMI1 CCNG2

Chronic Phase Blast crisis Remission

ATG5 ATG7 BECN1

Chronic Blast Remision Chronic Blast Remision Chronic Blast Remision Phase crisis Phase crisis Phase crisis

282 300 chronic phase

250 accelerated phase

200 168 165

150 104 100

33 40 33 50 31 23 8

mRNA expression (a.u) 0 BMI1 CCNG2 BMI1 CCNG2 BMI1 CCNG2 BMI1 CCNG2 BMI1 CCNG2

#1 #2 #3 #4 #5

350 318

300 ATG5 100 ATG7

250 80 200 60 49 150 97 40 100 62 56 15 20 50 24 8

mRNA expression (a.u) 4 1 0 mRNA expression (a.u) 0 cp ap cp ap cp ap cp ap cp ap cp ap cp ap cp ap cp ap cp ap

#1 #2 #3 #4 #5 #1 #2 #3 #4 #5 Figure 6. BMI1 level inversely correlates with CCNG2, ATG5 and ATG7 expression in CML. (a) BMI1 and CCNG2 expression profiling data from CML patients at different leukemic stages (GSE4170) were extracted using the R software. (b) ATG5, ATG7 and BECN1 expression profiling data from CML patients at different leukemic stages (GSE4170) were extracted using the R software. (c–e) RNA from patients were prepared from BM or PB mononuclear cells from five CML patients at the CP (light gray) or AP (dark gray). BMI1, CCNG2 (c), ATG5 (d) and ATG7 (e) RNA levels were assessed by real-time PCR. Columns: mean of quadruplicate samples. Similar results were obtained in two independent experiments.

© 2015 Macmillan Publishers Limited Leukemia (2015) 1993 – 2002 BMI1 blocks autophagy through cyclin G2 silencing L Mourgues et al 2000 CML progression is associated with increased levels of BMI1 and of this locus. BMI1 can protect cells from a DNA damage response decreased levels of CCNG2 and ATG5/ATG7 that would lead to premature aging and early death, by regulating 23 BMI1 levels were described to gradually increase during CML ROS homeostasis at the mitochondrial level. BMI1 can repress progression from CP to AP and BC.16 apoptosis by downregulating the expression of the proapoptotic We first examined the expression levels for BMI1 and CCNG2 in genes BIM24 or NOXA.25 In glioblastoma, transformation by BMI1 data extracted from a gene expression profiling study on CML does not require CDKN2A, although the resulting tumors have patients22 (Figure 6a). We found that as BMI1 mean levels a later time of onset and a lower grading,26 and in MCF10A human increased in BC samples CCNG2 levels comparatively decreased. mammary epithelial cancer cells, BMI1 cooperates with the H-Ras Interestingly, five patients in remission displayed lower BMI1 levels oncogene to induce transformation independently of Ink4a/ARF.27 compared with BC patients, whereas levels of CCNG2 increased In the human K562 CML cell line that harbors a deletion of the 17 towards those of CP. Levels of ATG7 were decreased during BC CDKN2A locus, we show that decreasing BMI1 levels resulted in and returned to normal levels after remission. No significant the reversible downregulation of cell proliferation and clonogenic variations could be observed for ATG5 nor BECN1 (Figure 6b). activity, without induction of apoptosis. A transcriptomic approach We next checked the relative expression of BMI1 and CCNG2 in and a ChIP assay revealed that BMI1 directly represses the CCNG2 primary CP and AP samples from the same patients. We found the gene, which appeared of interest because (i) it is induced by DNA same inverse evolution of the two genes in four out of five damage,28 (ii) it is upregulated under conditions of cell cycle arrest patients (Figure 6c). We also observed that the progression of the or apoptosis 29 and (iii) it mediates the antiproliferative actions of disease was associated with a decreased level of ATG5 in four out the tumor growth factor-β-like nodal ligand in an ovarian cell of five patients (Figure 6d) and of ATG7 in all five patients line.30 Furthermore, CCNG2 expression was shown to be decreased (Figure 6e). No significant differences could be observed for in several solid tumors from thyroid,31 oral,32 prostate33 and BECN1 (not shown). The results suggest that expression of BMI1 colon34 carcinomas, suggesting that CCNG2 could exert a global and CCNG2 could be functionally linked during CML aggravation, tumor suppressor role. Using ChIP, we demonstrated that BMI1 whereas ATG5 and ATG7 that are important for autophagy directly binds the CCNG2 promoter, close to the transcription induction are frequently downregulated. initiation site. Importantly, we could demonstrate, using RNA interference or inducible CCNG2 expression, that CCNG2 mediates the effects of BMI1 knockdown on K562 cell proliferation and DISCUSSION clonogenicity. These observations were confirmed on primary We show here for the first time that the BMI1 polycomb protein is CD34+ CML cells as downmodulation of BMI1 induced CCNG2 a repressor of autophagy by inhibiting the expression of the expression. Both events strongly interfered with clonogenicity. atypical cyclin G2. BMI1 is a known crucial regulator of stem cell The effects of BMI1 knockdown and of tyrosine kinase inhibitors fate by supporting their self-renewal, proliferation and survival.9 were neutral when combined, demonstrating that they lie on the Although BMI1 acts principally through repression of the CDKN2A same pathway, with BMI1 being downstream of Bcr-Abl. We locus to block engagement of the RB1 and TP53 tumor observed that knockdown of BMI1 as well as CCNG2 ectopic suppressors,9 it also mediates important functions independently expression both triggered an autophagic response that was

Figure 7. Schematic view of the molecular and cellular events induced by CCNG2 expression. Induction of CCNG2 after silencing of BMI1 disrupts the PP2A phosphatase complex liberating PKCζ, which becomes activated to stimulate AMPK, JNK and ERK that are required to initiate autophagosome isolation. Recruitment of LC3-II and ATG5–ATG12 complex can in turn contribute to autophagosome elongation and biogenesis.

Leukemia (2015) 1993 – 2002 © 2015 Macmillan Publishers Limited BMI1 blocks autophagy through cyclin G2 silencing L Mourgues et al 2001 evidenced by increased elements involved in autophagosome ACCESSION NUMBER elongation such as formation of LC3-II foci, beclin 1, ATG5 and Gene profiling was performed on inducible shBmi1-K562 cells ATG7 expression. Autophagy is a protective response when cells incubated or not with doxycycline (1µg/ml) for 96 h using HG- face starvation conditions, but it can also participate in U133 Plus2 Affymetrix Arrays. differentiation, metabolism reprogramming and cell death, depending on the cellular context.35 It appears to be at the crossroad of different behaviors in HSC as some constitutive REFERENCES autophagy appears important for maintaining HSCs genomic 1 Sparmann A, van Lohuizen M. Polycomb silencers control cell fate, development integrity,36 whereas on the other side, an excessive autophagy can and cancer. Nat Rev Cancer 2006; 6: 846–856. lead to cell death. Also, autophagy appears to exert tumor- 2 Hosen N, Yamane T, Muijtjens M, Pham K, Clarke MF, Weissman IL. Bmi-1-green fluorescent protein-knock-in mice reveal the dynamic regulation of bmi-1 suppressive functions as the lack of autophagy in the hemato- 25 fi expression in normal and leukemic hematopoietic cells. Stem Cells 2007; : poietic system after speci c deletion of Atg7 was shown to 1635–1644. interfere with stem cell functions and to cause an abnormal 3 Iwama A, Oguro H, Negjshi M, Kato Y, Morita Y, Tsukui H, et al. Enhanced self- myeloproliferation.37 Moreover, several genes that are central to renewal of hematopoietic stem cells mediated by the polycomb gene product autophagy have been reported to be deleted in various cancer Bmi-1. Immunity 2004; 21:843–851. cells. For instance, ATG5 is decreased in patients with melanoma,38 4 Lessard J, Sauvageau G. Bmi-1 determines the proliferative capacity of normal and 423 – whereas BECN1 (beclin 1) is monoallelically deleted in breast and leukaemic stem cells. Nature 2003; : 255 260. 39 5 Park I, Qian D, Kiel M, Becker M, Pihalja M, Weissman I, et al. Bmi-1 is required for ovarian cancers. As BMI1 is crucial to stemness, it could be maintainance of adult sef-renewing haematopoietic stem cells. Nature 2003; 423: envisioned that in normal HSCs the repressive action of BMI1 on 302–305. CCNG2 and autophagy is important to limit unwanted differentia- 6 van Lohuizen M, Verbeek S, Scheijen B, Wientjens E, van der Gulden H, Berns A. tion or cell death that would exhaust the stem cell compartment. identification of cooperating oncogenes in E mu-myc transgenic mice by provirus During malignant transformation, an overexpression of BMI1 and tagging. Cell 1991; 31:737–752. 7 Raaphorst F, Otte A, Meijer J. Polycomb-group genes as regulators of mammalian the following silencing of CCNG2 and the tumor-suppressive 22 – function of autophagy would promote tumoral progression. CML lymphopoiesis. Trends Immunol 2001; : 682 690. 8 Kreso A, van Galen P, Pedley NM, Lima-Fernandes E, Frelin C, Davis T, et al. Self- evolution from the CP to BC could be a representative example of renewal as a therapeutic target in human colorectal cancer. Nat Med 2014; 20:29–36. such processes. CML acceleration is associated with an increase in 9 Valk-Lingbeek ME, Bruggeman SW, van Lohuizen M. Stem cells and cancer; the the BMI1/CCNG2 ratio, and therefore an impaired autophagy, polycomb connection. Cell 2004; 118:409–418. could be important for disease progression. In addition, we also 10 Oguro H, Iwama A, Morita Y, Kamijo T, van Lohuisen M, Nakauchi H. Differential observed a decreased ATG5 and ATG7 expression in CML AP that impact of Ink4a and Arf on hematopoietic stem cells and their bone fi 203 could reflect a lower capacity of these cells to mount an marrow microenvironment in bmi-1-de cient mice. J Exp Med 2006; : 2247–2253. autophagic response. At a functional level, we observed that 11 Jacobs J, Kieboom K, Marino S, DePinho RA, van Lohuizen M. The oncogene and ectopic CCNG2 could associate with the PP2A phosphatase C Polycomb-group gene Bmi-1 regulates cell proliferation and senescence through 19 subunit as it was already reported. This binding of CCNG2 to the ink4a locus. Nature 1999; 397: 164–168. PP2A phosphatase C subunit comes with the release of PKCζ that 12 Smith K, Chanda S, Lingbeek M, Ross D, Botstein D, van Lohuizen M, et al. Bmi-1 can then activate downstream AMPK, JNK and ERK pathways. regulation of INK4A-ARF is a downstream requirement for transformation of Moreover, PKCζ inhibition could efficiently rescue CCNG2-induced hematopoietic progenitors by E2a-Pbx1. Mol Cell 2003; 12:393–400. proliferation and clonogenicity defects as well as prevent 13 Kim W, Sharpless N. The regulation of INK4/ARF in cancer and aging. Cell 2006; 127:265–275. vacuolization. Our observations are summarized in Figure 7, where 14 Kusy S, Larsen CJ, Roche J. p14ARF, p15INK4b and p16INK4a methylation status in BMI1, within the PRC1 complex, represses autophagy by blocking chronic myelogenous leukemia. Leuk Lymphoma 2004; 45:1989–1994. the expression of CCNG2. This new function adds CCNG2 to the 15 Nagy E, Beck Z, Kiss A, Csoma E, Telek B, Konya J, et al. Frequent methylation of panel of tumor-suppressive genes that are repressed by BMI1 such p16INK4A and p14ARF genes implicated in the evolution of chronic myeloid as CDKN2A and PTEN.40 Understanding the regulatory roles of leukaemia from its chronic to accelerated phase. Eur J Cancer 2003; 39: – BMI1 during the functional interplay between self-renewal, 2298 2305. 16BhattacharyyaJ,MiharaK,YasunagaS,TanakaH,HoschiM,KimuraA.Bmi-1 proliferation, senescence and now autophagy will be an important expresion is enhanced through transcriptional and posttranscriptional regulation step towards manipulation of these functions to support normal during the progression of chronic myeloid leukemia. Ann Hematol 2009; 88:333–340. stem cells and eliminate cancer stem cells. 17 Rui H, Su J. Co-transfection of p16INK4a and p53 genes into the K562 cell line inhibits cell proliferation. Haematologica 2002; 87:136–142. 18 Colosetti P, Puissant A, Robert G, Luciano F, Jacquel A, Gounon P, et al. Autophagy CONFLICT OF INTEREST is an important event for megakaryocytic differentiation of the chronic myelo- The authors declare no conflict of interest. genous leukemia K562 cell line. Autophagy 2009; 5: 1092–1098. 19 Arachchige Don A, Dallapiazza R, Bennin D, Brake T, Cowan C, Horne M. Cyclin G2 is a centrosome-associated nucleo-cytoplasmic shuttling protein that influences ACKNOWLEDGEMENTS microtubule stability and induces a p53-dependent cell cycle arrest. Exp Cell Res 2006; 312: 4181–4204. We are grateful to Malek Djabali (UMR5088 CNRS, Toulouse, France) for helpful 20 Wu SC, Solaro RJ. Protein kinase C zeta. A novel regulator of both phosphorylation discussion during early phases of the project; Hans Clevers (Hubrecht Laboratory, and de-phosphorylation of cardiac sarcomeric proteins. J Biol Chem 2007; 282: Utrecht, The Netherlands) for the kind gift of the pTER vector; Agnès Loubat for help 30691–30698. in cell cycle analysis; Jan Jacob Schuringa (University Medical Center Groningen, 21 Xiao H, Bai XH, Wang Y, Kim H, Mak AS, Liu M. MEK/ERK pathway mediates PKC Hematology, Groningen, The Netherlands) for the kind gift of the shBMI1 lentiviral activation-induced recruitment of PKCzeta and MMP-9 to podosomes. JCell constructs; Patrick Auberger, Arnaud Jacquel and Sandrine Obba for helpful Physiol 2013; 228:416–427. discussion concerning autophagy; Catherine Frelin for critical review of the 22 Radich JP, Dai H, Mao M, Oehler V, Schelter J, Druker B, et al. Gene expression manuscript. We thank Ms Chimène Morillon and Kevin Lebrigand from the changes associated with progression and response in chronic myeloid leukemia. Plateforme Génomique Fonctionnelle (IPMC, Sophia Antipolis, France) for Affymetrix Proc Natl Acad Sci USA 2006; 103: 2794–2799. analysis assistance and Dr Bernard Mari for help with transcriptomic analyses. We 23 Liu J, Cao L, Chen J, Song S, Lee I, Quijano C, et al. Bmi-1 regulates mitochondrial acknowledge the C3M imaging core facility (Microscopy and Imaging Platform Côte function and the DNA damage response pathway. Nature 2009; 459: 387–392. d'Azur) and the C3M animal room facility. The C3M is supported institutional grants 24 Jagani Z, Wiederschain D, Loo A, He D, Mosher R, Fordjour P, et al. The Polycomb from INSERM. LM is supported by a grant from the French Ministry of Research and a group protein Bmi-1 is essential for the growth of multiple myeloma cells. Cancer 4th year PhD grant from the Société Française d’Hématologie (SFH). Res 2010; 70: 5528–5538.

© 2015 Macmillan Publishers Limited Leukemia (2015) 1993 – 2002 BMI1 blocks autophagy through cyclin G2 silencing L Mourgues et al 2002 25 Yamashita M, Kuwahara M, Suzuki A, Hirahara K, Shinnaksu R, Hosokawa H, et al. 32 Kim Y, Shintani S, Kohno Y, Zhang R, Wong D. Cyclin G2 dysregulation in human Bmi1 regulates memory CD4 T cell survival via repression of the Noxa gene. J Exp oral cancer. Cancer Res 2004; 64: 8980–8986. Med 2008; 205: 1109–1120. 33 Cui DW, Cheng YJ, Jing SW, Sun GG. Effect of cyclin G2 on proliferative ability of 26 Bruggeman SW, Hulsman D, Tanger E, Buckle T, Blom M, Zevenhoven J, et al. Bmi1 prostate cancer PC-3 cell. Tumour Biol 2013; 35: 3017–3024. controls tumor development in an Ink4a/Arf-independent manner in a mouse 34 Sun GG, Zhang J, Hu WN. CCNG2 expression is downregulated in colorectal model for glioma. Cancer Cell 2007; 12:328–341. carcinoma and its clinical signiﬁcance. Tumour Biol 2013; 35: 3339–3346. 27 Hoenerhoff M, Chu I, Barkan D, Liu Z-y, Datta S, Dimri G, et al. BMI1 cooperates 35 Boya P, Reggiori F, Codogno P. Emerging regulation and functions of autophagy. with H-RAS to induce an aggressive breast cancer phenotype with brain metas- Nat Cell Biol 2013; 15:713–720. tases. Oncogene 2009; 28: 3022–3032. 36 Warr MR, Kohli L, Passegue E. Born to survive: autophagy in hematopoietic stem 28 Bates S, Rowan S, Vousden K. Characterization of human cyclin G1 and G2: DNA cell maintenance. Cell Cycle 2013; 12: 1979–1980. damage inducible genes. Oncogene 1996; 13: 1103–1109. 37 Mortensen M, Watson AS, Simon AK. Lack of autophagy in the hematopoietic 29 Martinez-Gac L, Marques M, Garcia Z, Campanero M, Carrera A. Control of cyclin system leads to loss of hematopoietic stem cell function and dysregulated G2 mRNA expression by forkhead transcription factors: novel mechanism for cell myeloid proliferation. Autophagy 2011; 7: 1069–1070. cycle control by phosphoinositide 3-kinase and forkhead. Mol Cell Biol 2004; 24: 38 Liu H, He Z, Simon HU. Autophagy suppresses melanoma tumorigenesis by 2181–2189. inducing senescence. Autophagy 2014; 10:372–373. 30 Xu G, Bernaudo S, GFu G, Lee D, Yang B, Peng C. Cyclin G2 is degraded through 39 Aita VM, Liang XH, Murty VV, Pincus DL, Yu W, Cayanis E, et al. Cloning and the ubiquitin-proteasome pathway and mediates the antiproliferative effect of genomic organization of beclin 1, a candidate tumor suppressor gene on activin receptor-like kinase 7. Mol Biol Cell 2008; 19: 4968–4979. chromosome 17q21. Genomics 1999; 59:59–65. 31 Ito Y, Yoshida H, Uruno T, Nakano K, Takamura Y, Miya A, et al. 40 Song LB, Li J, Liao WT, Feng Y, Yu CP, Hu LJ, et al. The polycomb group protein Decreased expression of cyclin G2 is signiﬁcantly linked to the malignant Bmi-1 represses the tumor suppressor PTEN and induces epithelial–mesenchymal transformation of papillary carcinoma of the thyroid. Anticancer Res 2003; 23: transition in human nasopharyngeal epithelial cells. J Clin Invest 2009; 119: 2335–2338. 3626–3636.

Supplementary Information accompanies this paper on the Leukemia website (http://www.nature.com/leu)