ZRF1 controls the retinoic acid pathway and regulates leukemogenic potential in acute myeloid...

transcript

ORIGINAL ARTICLE

ZRF1 controls the retinoic acid pathway and regulatesleukemogenic potential in acute myeloid leukemiaS Demajo1,2, I Uribesalgo1,2, A Gutierrez1,2, C Ballare1,2, S Capdevila3, M Roth4, J Zuber4, J Martın-Caballero3 and L Di Croce1,2,5

Acute myeloid leukemia (AML) is frequently linked to epigenetic abnormalities and deregulation of gene transcription, which leadto aberrant cell proliferation and accumulation of undifferentiated precursors. ZRF1, a recently characterized epigenetic factorinvolved in transcriptional regulation, is highly overexpressed in human AML, but it is not known whether it plays a role in leukemiaprogression. Here, we demonstrate that ZRF1 depletion decreases cell proliferation, induces apoptosis and enhances celldifferentiation in human AML cells. Treatment with retinoic acid (RA), a differentiating agent currently used to treat certain AMLs,leads to a functional switch of ZRF1 from a negative regulator to an activator of differentiation. At the molecular level, ZRF1 controlsthe RA-regulated gene network through its interaction with the RA receptor a (RARa) and its binding to RA target genes. Ourgenome-wide expression study reveals that ZRF1 regulates the transcription of nearly half of RA target genes. Consistent with ourin vitro observations that ZRF1 regulates proliferation, apoptosis, and differentiation, ZRF1 depletion strongly inhibits leukemiaprogression in a xenograft mouse model. Finally, ZRF1 knockdown cooperates with RA treatment in leukemia suppression in vivo.Taken together, our data reveal that ZRF1 is a key transcriptional regulator in leukemia progression and suggest that ZRF1 inhibitioncould be a novel strategy to be explored for AML treatment.

Oncogene advance online publication, 2 December 2013; doi:10.1038/onc.2013.501

Keywords: acute myeloid leukemia; ZRF1; retinoic acid; transcription; differentiation

INTRODUCTIONA characteristic abnormality of leukemia cells is an early stageblock in their development, with a subsequent failure todifferentiate into mature functional cells.1 Acute myeloidleukemia (AML) is the most frequent type of leukemia in adultsand has the lowest survival rate.2 AML is characterized by anaberrant proliferation and accumulation of granulocyte ormonocyte precursors in bone marrow, blood and other tissuessuch as the spleen.3,4 Since certain chemicals can induceleukemic cells to differentiate, one strategy for treatingleukemia is to force malignant cells into terminal differentiationrather than eliminating them. For instance, patients with acutepromyelocytic leukemia (APL), a subtype of AML, are treated withthe differentiating agent all-trans retinoic acid (RA).1 However, RAtreatment does not work for other types of AML besides APL, andit can lead to toxic secondary effects known as RA syndrome.5

For these cases, alternative therapies are being pursued.RA, a derivative of vitamin A, has an essential function during

vertebrate development, as it regulates numerous processes oforganogenesis and differentiation in multiple tissues.6 Duringhematopoiesis, RA controls hematopoietic stem cell developmentand helps to regulate granulocytic differentiation.7,8 Thus, RAtreatment of AML drives the differentiation of immature leukemiccells arrested at the promyelocytic stage to mature granulocytes.Importantly, the RA signaling pathway not only controls celldifferentiation but also directly regulates cell proliferation andapoptosis.9

RA transmits it signal predominantly by binding to the RAreceptor (RAR) family of nuclear receptors. RARs bind to specific

regions of DNA called RA responsive elements (RARE) toregulate the transcription of their target genes.9 In the absenceof RA, RARs repress their target genes by recruiting co-repressorproteins, such as histone deacetylase (HDAC), which results inhistone deacetylation, chromatin compaction and silencing. UponRA binding, RARs undergo a conformational change thatpromotes co-activator recruitment, which leads to chromatindecondensation and activation of transcription.10

How transcription factors regulate differentiation of hemato-poietic cells, and how they are disrupted in leukemia, has been thesubject of intense research.1 Growing evidence shows thatleukemia is frequently linked to epigenetic abnormalities.4 Theseinclude defects in diverse chromatin-related features, such as DNAmethylation, histone acetylation and methylation, and thefunction of Polycomb repressive complexes. Alterations in thesemechanisms lead to deregulation of gene transcription, which cancause aberrant silencing of genes necessary for myeloiddevelopment. Therefore, epigenetic defects act in concert withother types of genetic anomalies to trigger leukemictransformation.4

Recently, our group characterized ZRF1 (also known as MPP11)as an epigenetic regulator that displaces the Polycomb repressivecomplex 1 from chromatin during the onset of differentiation.11

We also showed that ZRF1 is able to control oncogene-inducedsenescence by transcriptionally regulating the INK4-ARF locus.12

Interestingly, previous studies showed that ZRF1 is overexpressedin several types of cancer, such as AML,13–15 chronic myeloidleukemia,16 chronic lymphocytic leukemia17 and head and necksquamous cell carcinoma.18 ZRF1 is highly overexpressed in

1Centre de Regulacio Genomica (CRG), Barcelona, Spain; 2UPF, Barcelona, Spain; 3Unidad de Animal de Laboratorio, PRBB, Barcelona, Spain; 4Institute of Molecular Pathology(IMP), Vienna, Austria and 5Institucio Catalana de Recerca i Estudis Avancats (ICREA), Barcelona, Spain. Correspondence: Dr L Di Croce, ICREA and Centre de Regulacio Genomica(CRG), c/ Dr Aiguader 88, 08003 Barcelona, Spain.E-mail: luciano.dicroce@crg.euReceived 15 May 2013; revised 15 October 2013; accepted 18 October 2013

www.nature.com/onc

leukemic blasts from patients with AML and in AML-derived celllines as compared with blasts from healthy donors. However,there are no reports addressing the role of ZRF1 in leukemia. Here,we investigated the function of ZRF1 in human AML cells. Ourresults indicate that ZRF1 regulates the leukemogenic potential bycontrolling cell proliferation, apoptosis and cell differentiation. Atthe molecular level, ZRF1 controls the RA pathway by regulatingthe expression of RA target genes. Depletion of ZRF1, alone or incombination with RA treatment, inhibits leukemia progression ofhuman AML xenografted blasts in vivo. Collectively, our datasuggest that ZRF1 is a potential novel target to be explored inleukemia treatment.

RESULTSZRF1 depletion decreases cell proliferation and increasesapoptosisTo explore the functional role of ZRF1 in AML, we stably knockeddown ZRF1 in five different human AML cell lines: HL60, NB4, U937,THP1 and NB4.007/6. We used two independent shRNA constructsthat efficiently downregulated ZRF1 mRNA and protein levels(Figure 1a and Supplementary Figure S1A). Interestingly, ZRF1depletion led to a strong decrease in growth rates in all the AML celllines analyzed, including the RA-resistant NB4.007/619 (Figure 1band Supplementary Figure S1B). We investigated if the observeddecrease in cell growth was caused by a reduction in proliferationand/or an increase in apoptosis. Analysis of BrdU incorporationrevealed that the cell proliferation of the ZRF1-depleted cells wassignificantly decreased as compared to control cells (Figure 1c).ZRF1-depleted cells also had an increased rate of spontaneous celldeath, as assessed by trypan blue staining. Annexin V stainingfollowed by FACS analysis revealed that the increase in dead cellswas caused by an induction of apoptosis (Figure 1d).

We next focused on HL60 cells as a model cell line to study theeffects of ZRF1 depletion in combination with RA treatment. RAreduces cell proliferation, increases apoptosis and induces celldifferentiation.9,10 In line with our previous results, we observed acooperative effect of RA and ZRF1 downregulation in growthinhibition of leukemic cells (Supplementary Figure S1C).Moreover, we found that ZRF1 knockdown strongly enhancedRA-induced apoptosis, thereby decreasing leukemic cell viabilityup to 50% (Figures 1e and f). Taken together, these resultsshow that ZRF1 depletion leads to a reduction in cell growth inAML cells due to both a decrease in cell proliferation and anincrease in apoptosis.

ZRF1 regulates RA-induced differentiationWe next studied the role of ZRF1 in leukemic cell differentiation, asa better understanding of differentiation could potentially help todevelop novel leukemia treatments. FACS analysis of thedifferentiation marker CD11b showed a consistent increase inthe rate of spontaneous differentiation after ZRF1 had beendepleted from leukemic cells (Figures 2a and b). Strikingly,

however, upon RA treatment, ZRF1-depleted HL60 cells had areduced potential to differentiate as compared to control cells,determined by the percentage of CD11b-positive cells. This effectwas observed after one day of RA treatment and was significantby day two (Figure 2c). Similar results were obtained by thenitroblue tetrazolium assay, which labels mature granulocytes(Supplementary Figure S1D). Accordingly, western blot analysisshowed a decreased expression of CD11c, a late differentiationmarker in HL60, in ZRF1 knockdown cells as compared to thecontrol (Figure 2d). In summary, ZRF1 depletion increases thebasal differentiation state of leukemic cells and impedes properdifferentiation following RA treatment.

We next generated an HL60 cell line that stably overexpressedHA-tagged ZRF1 (Figure 2e). In agreement with the resultsobtained from ZRF1 depletion, overexpression of ZRF1 increasedthe cell differentiation potential following RA treatment. Remark-ably, the ZRF1 effect was dose dependent, since the cells thatoverexpressed the highest levels of ZRF1 had the highest rates ofdifferentiation (Figure 2f and Supplementary Figure S1E). It shouldbe noted that ZRF1 protein levels were stable during RA-induceddifferentiation (Supplementary Figure S1F). Taken together, theseresults show that ZRF1 regulates cell differentiation in AML cells.As reported in the literature for RA receptor a (RARa),20 ZRF1seems to have a dual role, as a differentiation repressor in basalconditions but then switching to an activator following RAinduction.

ZRF1 regulates RA target gene expressionIn order to study the molecular mechanisms underlying the effectof ZRF1 in cell proliferation, apoptosis and cell differentiation, weperformed a genome-wide expression analysis in ZRF1-depletedand control HL60 cells in three different conditions: untreated(RA0) or treated with RA for 4 h (RA4h) or 48 h (RA48h). These earlyand late RA treatment time points were selected to be able tostudy both the direct effect of RA in transcription (RA4h) and thetranscriptome in the onset of RA-induced differentiation (RA48h).Comparing ZRF1 depleted and control cells, we found that theexpression of more than 5000 genes was altered in each of thethree RA conditions, with approximately half downregulated andhalf upregulated (Supplementary Figure S2A). We extracted thebona fide RA-activated genes by comparing untreated control cellsto RA4h-treated cells from our gene expression array and found1075 genes to be direct RA target genes (Supplementary FigureS2A). We next applied the Ingenuity Pathway Analysis todetermine the pathways and networks that were significantlyregulated by RA and ZRF1. Interestingly, we found that the cohortof genes regulated by ZRF1 both in basal and RA conditions werein the same pathway categories as the genes directly regulated byRA (Figure 3a and Supplementary Figure S2B). In particular, amongthe five most significantly overrepresented categories both for RA-and ZRF1-regulated genes, we found ‘cell development’, ‘cellgrowth and proliferation’ and ‘cell death and survival’, supportingour previous finding of the cellular functions regulated by ZRF1

Figure 1. ZRF1 depletion decreases cell proliferation and increases apoptosis (a) Western blot analysis of ZRF1 in control (shCtr) and ZRF1-depleted (shZRF1 #1 and #2) cells, in five AML cell lines. Tubulin was used as a loading control. (b) Growth curves of control andZRF1-depleted cells. Data are the means±s.e.m. of at least three independent experiments. (c) BrdU cell proliferation assay in controland ZRF1-depleted cells. Data are the means±s.e.m. of at least three independent experiments. (d) Cell death in control and ZRF1-depletedcells. Top panel, cell death assay by trypan blue-positive cell count; bottom panel, apoptosis assay determined by FACS after Annexin V/SytoxGreen double staining. Data are the means±s.e.m. of four independent experiments. (e) Cell death assay by trypan blue-positive cell count incontrol and ZRF1-depleted HL60 cells, untreated or treated with RA (1mM) for 3 or 5 days. Data are the means±s.e.m. of four independentexperiments. (f ) Cell death assay determined by FACS after Annexin V/Sytox Green double staining in control and ZRF1-depleted HL60 cells,untreated or treated with RA (1mM) for 3 or 5 days. Viable cells in green (Q3), apoptotic cells in blue (Q4), late apoptotic or necrotic cells in red(Q2). Graphs are representative of three independent experiments; numbers represent percentage of cells. Statistical analysis was performedbetween the two shZRF1 and the shControl cells at each of the three conditions. Statistical significance was assessed by a two-tailed Student’st-test; *Po0.05.

ZRF1 regulates leukemogenesisS Demajo et al

Oncogene (2013) 1 – 10 & 2013 Macmillan Publishers Limited

(Figures 1 and 2). Strikingly, we also found that there was a highlysignificant overlap between the 1075 bona fide RA target genesand those regulated by ZRF1, with almost half of the RA

target genes in HL60 cells co-regulated by ZRF1, both in basaland RA conditions (Figure 3b and Supplementary Figure S2B).In addition, a motif analysis of the genes regulated by ZRF1

& 2013 Macmillan Publishers Limited Oncogene (2013) 1 – 10

indicated that the RARE was significantly overrepresented in thepromoters of these genes (Supplementary Figure S2C).

Focusing on the subset of genes co-regulated by RA and ZRF1,we observed that, overall, ZRF1 carried out opposite transcrip-tional roles in untreated and RA-treated cells. In the absence of RA,70.6% of the 449 co-regulated genes were upregulated in theZRF1 knockdown cells (untreated; Figure 3c and SupplementaryFigure S2D). In contrast, with RA treatment, 62.7% of the 505co-regulated genes were downregulated in the ZRF1 depletedcells (RA treated for 48 h; Figure 3c and Supplementary FigureS2D). Analysis of the overlap between the RA target genesupregulated and downregulated by ZRF1 knockdown in untreatedversus RA treated cells revealed that ZRF1 functions either as arepressor or as an activator in different subsets of genes, likelydepending on the promoter context (Supplementary Figure S2E).We next analyzed the RA target genes that were upregulated by

the ZRF1 knockdown in basal conditions and those down-regulated by ZRF1 inhibition following 48 h of RA treatment.Significantly, the list of genes included important regulators ofdevelopment and differentiation in the myeloid linage(Supplementary Figure S3A), such as ICAM1, HOXA5, RGS2, THBD,CSF1R, ICAM4, ICAM3 and CSF3R, in line with the differentiationresults after ZRF1 knockdown (see Figure 2). Their expression wasmeasured in independent experiments by quantitative PCR withreverse transcription (qRT–PCR) (Figure 3d). Moreover, analysis ofthe RA target genes upregulated by ZRF1 knockdown in bothuntreated and RA-treated cells showed an enrichment of positiveregulators of apoptosis and negative regulators of proliferation(Supplementary Figures S3A and B), which also supports theresults shown above (see Figure 1). Confirming the role of ZRF1 asa gene activator that we had previously reported,11,12 we foundthat nearly half of the RA targets depended on ZRF1 for proper

Figure 2. ZRF1 regulates RA-induced differentiation (a) Differentiation assay by CD11b-positive surface marker measured by FACS, in control(shCtr) and ZRF1-depleted (shZRF1 #1 and #2) HL60 cells. Only viable cells were considered for the analysis. Data are the means±s.e.m. of fourindependent experiments. (b) Differentiation assay in control and ZRF1-depleted cells, in AML cell lines; results are shown as percentage ofCD11b or CD11c (in the case of NB4 cells) positive cells. Only viable cells were considered for the analysis. Data are the means±s.e.m. of threeindependent experiments. (c) Differentiation assay (CD11b-positive surface marker) in control and ZRF1-depleted HL60 cells, untreated (0) ortreated with RA (1 mM) for up to 3 days. Only viable cells were considered for the analysis. Right, FACS profile of a representative experiment.Data are the means±s.e.m. of four independent experiments. (d) Western blot analysis of the differentiation marker CD11c and ZRF1 incontrol and ZRF1-depleted HL60 cells, untreated (no RA) or treated with RA (1 mM) for 3 or 5 days. Tubulin was used as a loading control.(e) Western blot analysis of ZRF1 in HL60 cells infected with GFP-HA-empty (control, Ctr) or GFP-HA-ZRF1 (ZRF1) expression vector. Arrowheadindicates HA-ZRF1 fusion protein. Tubulin was used as a loading control. (f ) Differentiation assay (CD11b-positive surface marker) in controland ZRF1-overexpressing HL60 cells, untreated (0) or treated with RA (100 nM) for up to 3 days. ZRF1 high corresponds to the cellsoverexpressing the highest levels of ZRF1 (see Supplementary Figure S1E). Only viable cells were considered for the analysis. Data are themeans±s.e.m. of four independent experiments. ZRF1 overexpression does not affect gene expression of RA targets in untreated cells (seeSupplementary Figure S3E). Statistical significance was assessed by a two-tailed Student’s t-test; *Po0.05.

gene induction after RA administration (Supplementary FigureS3C). This is the first time, however, that ZRF1 has been reportedto act as a transcription repressor. We therefore extended thisanalysis to NB4 cells, for which we likewise observed anupregulation in the basal state of the RA target genes uponZRF1 knockdown (Supplementary Figure S3D). Taken together,these results strongly suggest that ZRF1 is directly involved in thecontrol of RA-regulated gene network, thus regulating prolifera-tion, apoptosis and differentiation processes.

ZRF1 interacts with RARa and controls histone acetylationOur genome-wide expression analysis showed that ZRF1 controlsa large proportion of RA target genes. The RAR family mediatesvirtually all the physiological effects of RA.10 RARa is the main

isoform expressed in myeloid leukemic cells,21 as we confirmed inour cellular model by determining the expression of the threeRARs (Supplementary Figure S4A). Given the important effect thatZRF1 inhibition had on the RA transcriptome, we hypothesizedthat ZRF1 regulates RA target gene expression through itsinteraction with RARa. Indeed, we observed that ZRF1 interactedwith RARa in pulldown assays of His-ZRF1 with 293T nuclearextracts (Supplementary Figure S4B). Importantly, endogenous co-immunoprecipitation experiments confirmed this ZRF1-RARainteraction in HL60 cells (Figure 4a). To study whether theinteraction between ZRF1 and RARa was direct, we carried outin vitro pulldown experiments using recombinant proteins. Wefound that ZRF1 interacted directly with RARa irrespective of thepresence or absence of RA (Figure 4b and SupplementaryFigure S4C). In addition, we observed that its binding to RARa

Figure 3. ZRF1 regulates RA target gene expression. (a) Ingenuity Pathway Analysis of the gene expression microarray showing the top fivemost significantly overrepresented categories of the cohort of RA direct targets (as defined in Supplementary Figure S2A) and ZRF1-regulatedgenes (both up- and downregulated) corresponding to RA 0 and RA 48 h. In bold, the categories shared in the three analyses. For ZRF1-regulated genes corresponding to RA 4 h, see Supplementary Figure S2B. (b) Proportion of RA direct targets co-regulated by ZRF1, comprisingboth upregulated and downregulated genes in shZRF1 cells as compared with shControl, corresponding to RA 0 and RA 48 h. For RA 4 h, seeSupplementary Figure S2B. (c) Microarray heat-map of RA and ZRF1 co-regulated genes corresponding to untreated cells (RA0, left) and RA-treated cells (RA48h, right). In each case, the first columns corresponds to the expression levels of control (shCtr) and ZRF1-depleted (shZRF1)replicates; the last two columns correspond to the direct effect of RA (RA activation: comparing the expression at RA4h with RA0, in controlcells) and the effect of ZRF1 (shZRF1 effect: comparing shControl with shZRF1 cells). Genes were sorted by how they were affected by shZRF1,from the most upregulated to the most downregulated. The percentages of genes upregulated (UP) and downregulated (DOWN) in each caseare shown. (d) qRT–PCR analysis of representative RA-target genes previously reported to be involved in myeloid differentiation, in shZRF1relative to shControl cells. Top panel: genes upregulated in shZRF1 at RA0. Bottom panel: genes downregulated in shZRF1 at RA48h; thenumbers above the shControl bars correspond to the fold induction as compared with untreated cells. Expression was normalized to thePUM1 housekeeping gene. Data are the means±s.e.m. of four independent experiments.

was mediated by the N-terminal part of ZRF1 (Figure 4c andSupplementary Figure S4D).

Although RARa regulates both gene activation and repressionby binding to RARE sequences in the DNA,10 the aberrant basalrepression of RA target genes by RARa is the key feature of somesubtypes of leukemia. Given the promising anti-leukemic effects ofZRF1 depletion in cell proliferation, apoptosis and differentiation(Figures 1 and 2), we focused on studying ZRF1 regulation of RAtarget genes in undifferentiated, untreated leukemic cells.Chromatin immunoprecipitation experiments showed that ZRF1occupied RAREs on the promoters of RA target genes in untreatedHL60 cells (Figure 4d). Chromatin immunoprecipitation alsorevealed that ZRF1 depletion affected the chromatin state ofthese RA target genes by increasing the level of histoneacetylation (Figure 4e). Specifically, we observed an increasedlevel of global histone H3 acetylation (H3Ac) and of histone H3acetylation at lysine 27 (H3K27Ac), both of which are marksassociated with active chromatin.22 This observation correlatedwith our finding in the genome-wide expression study that a largeproportion of RA target genes were upregulated following ZRF1inhibition in basal conditions (Figures 3c and d). Taken together,

these results suggest that ZRF1 regulates RA target geneexpression by interacting with RARa and thereafter controllinghistone acetylation.

ZRF1 depletion inhibits leukemogenesis in vivoWe have shown that depletion of ZRF1 in AML cells leads toa decrease in cell proliferation and an increase in apoptosis andcell differentiation in proliferating cells in vitro. These results,together with the fact that ZRF1 is highly overexpressed in humanAML,13–15 suggested that targeting ZRF1 could be a potentialnovel strategy to be explored for leukemia treatment.Xenotransplantation of human AML leukemic cells into severecombined immunodeficient (SCID) mice recapitulates most of thefeatures of the human malignancy, such as the presence ofleukemic cells in peripheral blood and their homing to thespleen.23 Therefore, we chose this experimental model to validatethe anti-leukemic potential of ZRF1 inhibition in vivo. Aftergenerating an HL60 cell line that expressed luciferase, we stablyknocked down ZRF1 in this cell line (Supplementary Figure S5A).We injected control and ZRF1-depleted cells in SCID mice and

Figure 4. ZRF1 interacts with RARa and controls histone acetylation. (a) Co-immunoprecipitation (IP) assay of endogenous RARa and ZRF1 inHL60 cells. IgG was used as a control. (b) In vitro GST pulldown assay with recombinant GST or GST-RARa and recombinant His-ZRF1, in theabsence (–) or presence of RA (þ : 100 nM; þ þ : 1 mM), as detected by anti-His immunoblotting. (c) In vitro GST-pulldown assay withrecombinant GST or GST-RARa and recombinant His-ZRF1, full-length or deletion mutants, as indicated in the corresponding diagram;detected by anti-His immunoblotting. The different domains of ZRF1 are indicated. (d) Chromatin immunoprecipitation (ChIP) of ZRF1 inHL60, control (shControl) and ZRF1-depleted (shZRF1) cells, followed by real-time qPCR analysis of the indicated RA target genes at the RAREregions. IgG was used as a control. Results are shown as percentage relative to input. Data are the means±s.e.m. of three independentexperiments. (e) ChIP of global histone H3 acetylation (H3Ac) and histone H3 lysine 27 acetylation (H3K27Ac) in shControl and shZRF1 HL60cells, followed by qRT–PCR analyses of the indicated RA target genes at the RARE regions. Results are shown as enrichment relative to totalhistone H3. MAT2A gene was used as a positive control for acetylation.36 Data are the means±s.e.m. of three independent experiments.

followed leukemia progression through bioluminescent imaging.Strikingly, we observed a strong inhibition of leukemia progres-sion in mice injected with ZRF1-depleted cells as compared withthe control (Supplementary Figures S5B and C). Additionally, post-mortem necropsies showed that all the mice injected with controlcells presented several intraperitoneal solid tumors and most ofthem had an enlarged spleen (Supplementary Figure S5D), whileonly two out of eight mice injected with ZRF1-depleted cellsshowed tumor masses. To further analyze the leukemic pheno-type, we collected samples from peripheral blood and spleen andperformed FACS analysis of human CD33, a myeloid cell surfacemarker. In agreement with the bioluminescence results, wedetected a strong reduction in the amount of leukemic cells inmice injected with ZRF1-depleted cells as compared with controlmice (Supplementary Figure S5E).

We next studied the effect of ZRF1 depletion combined with RAadministration. We observed that ZRF1 knockdown inhibitedleukemia progression to a similar extent as RA treatment, and,remarkably, that a combination of both had a cooperative effect inleukemia suppression (Figures 5a and b). Analysis for the presenceof leukemic cells in peripheral blood and spleen confirmed theseresults and showed that mice injected with ZRF1-depleted cellsand treated with RA had almost undetectable levels of leukemiccells (Figure 5c). Moreover, in agreement with our in vitro results,ZRF1 knockdown cells showed a higher level of differentiation ascompared with control cells (Figure 5d). These results, thatdepletion of ZRF1 strongly inhibited leukemia progressionin vivo and that it functioned cooperatively with RA in leukemiasuppression, were confirmed in NB4 cells (Figure 5e andSupplementary Figures S5F and G). These findings suggest thatZRF1 could be a novel target for AML treatment.

DISCUSSIONIntensive research on the process of leukemic cell differentiationin recent years has provided important clues both for under-standing the mechanisms underlying cell fate transitions and fordiscovering novel therapeutic drugs. In this report, we havestudied the function of ZRF1, a recently characterized epigeneticfactor that is overexpressed in AML but with an unknown role inleukemia development. We demonstrate that ZRF1 is animportant regulator of proliferation, apoptosis and differentiationin AML cells, all of which are fundamental processes altered inleukemia. Additionally, ZRF1 depletion strongly inhibited leukemiaprogression in a mouse xenograft model. At the molecular level,we show that ZRF1 has an important interplay with the RApathway through its binding to RARa and its regulation of RAtarget gene transcription.

Our data provide evidence that ZRF1 normally functions as arepressor of differentiation in proliferating cells but turns into anactivator when cells are treated with RA. Strikingly, this dual role isalso found in RARa itself:20 studies using RARa� /� mice and otherapproaches have shown that the RA receptor works both as apositive and a negative regulator of granulocytic differentiation.24

This suggests that there could be a functional cooperation betweenZRF1 and RARa. In fact, our genome-wide expression data showthat ZRF1 regulates nearly half of the RA target genes, pointing toZRF1 as a cofactor of RARa. However, the molecular basis for theZRF1 switch from a repressor to an activator remains unclear. Onepossible explanation is that the protein undergoes a post-translational modification during differentiation. Interestingly, agenome-wide phosphoproteomic study suggested that ZRF1 isphosphorylated during RA-induced differentiation in mouse P19cells.25 In this regard, we recently reported that the oncogene Myccooperates with RARa to repress RARE-containing genes inundifferentiated cells and, upon RA treatment, becomesphosphorylated, which allows it to activate gene expression andenable proper AML cell differentiation.26 Understanding whether a

similar mechanism controls the ZRF1 switch from a repressor to anactivator will be a focus of future research.

We reported previously that ZRF1 is an epigenetic regulator thatdisplaces the Polycomb repressive complex 1 from chromatin,thus facilitating transcriptional activation.11,12 However, themechanism by which ZRF1 is targeted to specific genesremained unclear. Here, we show that ZRF1 binding to RARa isa possible mechanism by which it is recruited to chromatin.Moreover, we show that ZRF1 also works as a transcriptionalrepressor of RA target genes by regulating basal histoneacetylation levels. In fact, RARa-mediated repression alsoinvolves the regulation of histone acetylation, being HDACsfundamental components of the RARa-associated corepressorcomplexes.10 This leads us to hypothesize that ZRF1 may interactwith HDACs and that ZRF1 depletion would displace the HDACs,thereby increasing acetylation levels. If this is the case, differentialbinding to HDACs depending on the cellular and/or chromatincontext may account, at least in part, for the dual role of ZRF1 intranscription. Interestingly, deregulation of histone acetylation byaberrant recruitment of HDACs to RA target genes contributes toleukemogenesis in several types of AML.27

In contrast with the dual role of ZRF1 in differentiation, our datashows a unique function for ZRF1 in proliferation and apoptosis.Specifically, ZRF1 represses both anti-proliferative genes and pro-apoptotic genes, thus promoting proliferation and inhibitingapoptosis. This suggests that ZRF1 overexpression in AML maycontribute significantly to the alteration of these processes that ischaracteristic of leukemia.

In this study, we demonstrate that ZRF1 depletion gives rise tocell growth inhibition by deregulating the proliferation, apoptosisand differentiation processes in five different AML cell lines, whichsuggests that the function of ZRF1 is of broad relevance in AML.This effect of ZRF1 on cell growth is consistent with previous data inhuman and mouse cancer cell lines28,29 and with data we obtainedin other cancer cell lines, such as NT2 (data not shown). However,our previous results showed that this deregulation is not observedwhen ZRF1 is knocked down in normal cells, such as humankeratinocytes12 and mouse embryonic stem cells (data not shown).Taken together, these data demonstrate that growth inhibition byZRF1 depletion depends on the cellular context. Considering alsothat ZRF1 is highly overexpressed in AML blasts as compared withnormal blasts, we hypothesize that a potential inhibitor of ZRF1could specifically affect leukemic cells, which should be tested inthe future with the appropriate experimental models.

As mentioned above, previous studies have shown that ZRF1(termed MPP11) is overexpressed in several types of cancer,including leukemia16 and, specifically, in AML.13–15 Moreover,these reports showed that ZRF1 is among the leukemia-associatedantigens, which are considered to be promising potential targetsfor immunotherapy in AML.30 We demonstrate here that ZRF1inhibition in AML cells leads to a decrease in cell proliferation andan increase in apoptosis and basal differentiation. This triple effectof ZRF1 depletion is translated into a strong reduction of theleukemogenic potential in vivo. Furthermore, ZRF1 depletioncooperates with RA treatment in leukemia suppression, openingthe possibility to investigate combination therapies. Remarkably,the effect observed in our xenograft experiments seems to bestronger than the one observed in vitro, which suggests that thelack of ZRF1 hinders cell engraftment and/or dissemination in anin vivo environment. Overall, our data suggest that ZRF1, alone orin combination with RA, is a potential novel target to be exploredfor AML treatment.

MATERIALS AND METHODSPlasmids, cloning, primers and antibodiesTo overexpress ZRF1, ZRF1 was subcloned from pet28A11 to a lentiviralpEV833 plasmid. Plasmids for the recombinant proteins GST-RARa and

His-ZRF1 were described previously.11,26 pLPNIG (MSCV-Luc2-PGK-Neo-IRES-GFP) was generated by replacing the miR30 cassette in pLMN31 with amammalian codon-optimized luc2 transgene (taken from pGL4.13,Promega, Madison, WI, USA). Supplementary Table S1 shows theantibodies, primer sequences and shRNA sequences used in the study.

Cell culture, transfection and lentiviral infectionHL60, NB4, U937, THP1 and NB4.007/6 cells were cultured at 37 1C and 5%CO2 in RPMI medium supplemented with 10% fetal bovine serum.HEK293T cells were cultured at 37 1C and 5% CO2 in DMEM supplementedwith 10% fetal bovine serum.

To produce lentivirus, HEK293T cells (2� 106) were transfected using thecalcium phosphate co-precipitation method with 5 mg of pCMV-VSV-G,6 mg of pCMVDR-8.91 and 7mg of either pLKO-shRNA (Sigma, St Louis, MO,USA) for knockdown or pEV833-GFP for overexpression. After 48 h,infection was performed as described before.32 Cells were eitherselected with 2 mg/ml of puromycin (Sigma) or sorted for GFP-positivecells 24–48 h after infection.

Western blot analysisCell extracts for Western blot analysis were prepared in lysis buffer(25 mM Tris–HCl, pH7.6, 1% SDS, 1 mM EGTA, 1 mM EDTA), incubated10 min at 100 1C, sonicated for 30 s in a Bioruptor (Diagenode, Liege,Belgium) and centrifuged for 30 min at maximum speed at 4 1C. Proteinsupernatant was quantified by Bradford assay, diluted in Laemmli bufferand analyzed by SDS–PAGE.

Cell proliferationCells were treated with 10 mM of BrdU solution for 30 min and thenanalyzed for BrdU incorporation using the APC BrdU Flow Kit (BDPharmingen, Franklin Lakes, NJ, USA) according to the manufacturer’sprotocol. The percentage of BrdU-positive cells was analyzed by flowcytometry on a Becton Dickinson FACSCanto.

Cell viability and apoptosisCells were stained with trypan blue and counted under a light microscope.Apoptosis analysis was performed using a Violet Annexin V/deadcell apoptosis kit (Invitrogen, Carlsbad, CA, USA) according to themanufacturer’s protocol. After staining, cells were analyzed by flowcytometry on Becton Dickinson LSRII.

Cell differentiationDifferentiation experiments by CD11b/CD11c expression and nitrobluetetrazolium assay were performed as described previously.32

qRT–PCR analysisRNA was extracted with the RNeasy Mini Kit (Qiagen, Hilden, Germany).cDNA was generated from 1 mg of RNA with the First Strand cDNASynthesis Kit (Fermentas GmbH, St Leon-Rot, Germany). qRT–PCR reactionwas performed using SYBR green (Roche, Basel, Switzerland).

Agilent gene expression microarrayRNA from four independent experiments was isolated, amplified, labeledand subsequently hybridized to the human Agilent SurePrint G3 geneexpression 8� 60K microarray (Agilent Technologies, Santa Clara, CA, USA).Two samples (one corresponding to shControl RA0 and the other toshZRF1 RA48h) were excluded from the analysis due to low quality data.ZRF1- and RA-target genes were selected by considering all probe setswith an adjusted P-value lower than 0.05 and a fold-change cutoff of atleast 1.2. The overlap analyses were performed using the GenomatixSoftware (http://www.genomatix.de). Gene lists were analyzed withIngenuity Pathways Analysis (IPA; Ingenuity Systems; http://www.ingenui-ty.com) and DAVID Gene Functional Classification.33 Analysis of RARE motifoverrepresentation was performed using sequences located within 5kilobases upstream of the transcription start site. Sequences were scanned

with the Clover software34 using a matrix describing the RARE motif.35 Foreach group of genes analyzed, the E-values were estimated by comparingthe obtained scores with a 1000 random sample obtained from the non-regulated genes.

ImmunoprecipitationFor co-immunoprecipitation assays, cells were washed in PBS, lysed in lysisbuffer (25 mM Tris–HCl, pH7.4, 150 mM NaCl, 1 mM EDTA, 1% NP40 and 5%glycerol) and sonicated for 1 min (6 cycles of 10 sec) in a Bioruptor(Diagenode). After centrifugation for 30 min at maximum speed, solublematerial was quantified by Bradford. The antibodies were crosslinked toprotein A sepharose beads (GE Healthcare, Chalfont St Giles, UK) using BS3(Thermo Scientific, Hudson, NH, USA), following the supplier’s guidelines,and saturated with BSA. Lysates were incubated overnight with theantibodies conjugated to the beads. Immunoprecipitated material waswashed four times with lysis buffer and eluted either with Laemmli bufferor 0.1 M glycine-HCl pH 2.8 (to avoid co-elution of antibody heavy chainwith the target antigen). Eluates were loaded into SDS–PAGE gels.

Chromatin immunoprecipitationChromatin immunoprecipitation was performed as described before,32

with the following modifications: crosslinking was performed with 1%formaldehyde at 37 1C for 10 min and samples were sonicated for 18 min.

Pulldown assayRecombinant protein expression and purification and precipitations usingeither GST or His were performed as previously described.11,26

Xenotransplantation modelAnimal studies were carried out in the AAALAC international accreditedAnimal Facility of the Biomedical Research Park of Barcelona (PRBB) inaccordance with approved protocols from the Institutional Animal Careand Use Committee. HL60 and NB4 cells were infected with the LPINGplasmid containing luciferase and subsequently with pLKO-shControl orpLKO-shZRF1. Immediately after selection with puromycin, 4� 106 cellswere intraperitoneally inoculated into 8-week-old female CB17 SCID/beigemice. Eight mice per group were used in the experiment without RA, andfive mice per group in the two experiments with RA treatment. Both thevehicle and RA were injected intraperitoneally twice a week at 40 mg/kg,starting at day 7. RA (Sigma) was dissolved in Cremophor EL (Sigma) andthen diluted 1:15 in PBS. For whole-body bioluminescent imaging, micewere injected intraperitoneally with 50 mg/kg D-luciferin (Gold BioTechnology, St Louis, MO, USA) and analyzed after 6 min using anIVIS Imaging system (Caliper Life Sciences, Hopkinton, MA, USA). Imageswere quantified with Living Image software (Caliper Life Sciences). Thestudy was performed under SPF conditions. Mice were euthanized on day22 or 23, and necropsies were performed. Spleens and peripheral bloodwere collected (the latter in the presence of EDTA). Samples wereseparately prepared using 1� RBC lysis buffer (eBioscience, San Diego, CA,USA), following the supplier’s guidelines, and were incubated 30 min withPE-Cy5-CD33 and, if indicated, with PE-CD11b. The presence of CD33-positive cells and the cell differentiation status were analyzed by flowcytometry with a Beckton Dickinson FACScanto.

Figure 5. ZRF1 depletion inhibits leukemogenesis in vivo. (a) Bioluminescent imaging of xenografted SCID mice injected with control(shControl) or ZRF1-depleted (shZRF1) HL60 cells, treated with RA or vehicle, during leukemia progression. One representative mouse fromeach group is shown. (b) Bioluminescent quantification (in photons/sec/cm2/steradian) of xenografted SCID mice. Data are represented on alogarithmic scale as box-and-whisker plots at the corresponding days after injection; boxes represent the quartiles and whiskers mark theminimum and maximum values. Statistical significance was determined with a two-way ANOVA. Right panel: bioluminescent data on a linearscale corresponding to day 21. (c) FACS analysis of HL60 leukemic cells abundance in peripheral blood and spleen corresponding to the fourexperimental groups. Results are shown as the percentage of human CD33-positive cells within the gated population. Data are themeans±s.e.m. Statistical significance was assessed by a two-tailed Student’s t-test; *Po0.05, **Po0.01. (d) FACS analysis of the differentiationstate of HL60 cells in peripheral blood and spleen. Results are shown as expression of human CD11b (mean fluorescent intensity)corresponding to the CD33-positive population. In mice from shZRF1þ RA group, CD33-positive cells were not detected or the levels wereclose to the background. (e) Bioluminescent quantification (in photons/sec/cm2/steradian) of xenografted SCID mice injected with control(shControl) or ZRF1-depleted (shZRF1) NB4 cells, treated with RA or vehicle. Statistical significance was determined with a two-way ANOVA.Right panel: bioluminescent data on a linear scale corresponding to day 22.

CONFLICT OF INTERESTThe authors declare no conflict of interest.

ACKNOWLEDGEMENTSWe thank the members of the LDC laboratory for discussions, the CRG Genomic Unit,and VA Raker for help in preparing the manuscript. This work was supported bygrants from the Spanish ‘Ministerio de Educacion y Ciencia’ (BFU2010-18692), fromAGAUR, from ‘Fundacio La Marato’ and from by European Commission’s 7thFramework Program 4DCellFate grant number 277899 to LDC; SD was supported by aPFIS fellowship of the ‘Instituto de Salud Carlos III’.

REFERENCES1 Nowak D, Stewart D, Koeffler HP. Differentiation therapy of leukemia: 3 decades

of development. Blood 2009; 113: 3655–3665.2 Deschler B, Lubbert M. Acute myeloid leukemia: epidemiology and etiology.

Cancer 2006; 107: 2099–2107.3 Tenen DG. Disruption of differentiation in human cancer: AML shows the way.

Nat Rev Cancer 2003; 3: 89–101.4 Uribesalgo I, Di Croce L. Dynamics of epigenetic modifications in leukemia.

Brief Funct Genomics 2011; 10: 18–29.5 Wang ZY, Chen Z. Acute promyelocytic leukemia: from highly fatal to highly

curable. Blood 2008; 111: 2505–2515.6 Duester G. Retinoic acid synthesis and signaling during early organogenesis.

Cell 2008; 134: 921–931.7 Chanda B, Ditadi A, Iscove NN, Keller G. Retinoic Acid signaling is essential for

embryonic hematopoietic stem cell development. Cell 2013; 155: 215–227.8 Friedman AD. Transcriptional control of granulocyte and monocyte development.

Oncogene 2007; 26: 6816–6828.9 Tang XH, Gudas LJ. Retinoids retinoic acid receptors, and cancer. Annu Rev Pathol

2011; 6: 345–364.10 Altucci L, Gronemeyer H. The promise of retinoids to fight against cancer. Nat Rev

Cancer 2001; 1: 181–193.11 Richly H, Rocha-Viegas L, Ribeiro JD, Demajo S, Gundem G, Lopez-Bigas N et al.

Transcriptional activation of polycomb-repressed genes by ZRF1. Nature 2010;468: 1124–1128.

12 Ribeiro JD, Morey L, Mas A, Gutierrez A, Luis NM, Mejetta S et al. ZRF1 controlsoncogene-induced senescence through the INK4-ARF locus. Oncogene 2013; 32:2161–2168.

13 Greiner J, Ringhoffer M, Taniguchi M, Hauser T, Schmitt A, Dohner H et al.Characterization of several leukemia-associated antigens inducing humoralimmune responses in acute and chronic myeloid leukemia. Int J Cancer 2003; 106:224–231.

14 Greiner J, Ringhoffer M, Taniguchi M, Li L, Schmitt A, Shiku H et al. mRNAexpression of leukemia-associated antigens in patients with acute myeloidleukemia for the development of specific immunotherapies. Int J Cancer 2004;108: 704–711.

15 Siegel S, Wirth S, Schweizer M, Schmitz N, Zeis M. M-phase phosphoprotein 11 is ahighly immunogenic tumor antigen in patients with acute myeloid leukemia. ActaHaematol 2012; 127: 193–197.

16 Schmitt M, Li L, Giannopoulos K, Chen J, Brunner C, Barth T et al. Chronicmyeloid leukemia cells express tumor-associated antigens eliciting specific CD8þT-cell responses and are lacking costimulatory molecules. Exp Hematol 2006; 34:1709–1719.

17 Giannopoulos K, Li L, Bojarska-Junak A, Rolinski J, Dmoszynska A, Hus I et al.Expression of RHAMM/CD168 and other tumor-associated antigens in patientswith B-cell chronic lymphocytic leukemia. Int J Oncol 2006; 29: 95–103.

18 Resto VA, Caballero OL, Buta MR, Westra WH, Wu L, Westendorf JM et al.A putative oncogenic role for MPP11 in head and neck squamous cell cancer.Cancer Res 2000; 60: 5529–5535.

19 Dermime S, Grignani F, Rogaia D, Liberatore C, Marchesi E, Gambacorti-Passerini C.Acute promyelocytic leukaemia cells resistant to retinoic acid show further per-turbation of the RAR alpha signal transduction system. Leuk Lymphoma 1995; 16:289–295.

20 Oren T, Sher JA, Evans T. Hematopoiesis and retinoids: development and disease.Leuk Lymphoma 2003; 44: 1881–1891.

21 Lehmann S, Paul C, Torma H. Retinoid receptor expression and its correlation toretinoid sensitivity in non-M3 acute myeloid leukemia blast cells. Clin Cancer Res2001; 7: 367–373.

22 Campos EI, Reinberg D. Histones: annotating chromatin. Annu Rev Genet 2009; 43:559–599.

23 Uckun FM. Severe combined immunodeficient mouse models of human leuke-mia. Blood 1996; 88: 1135–1146.

24 Kastner P, Chan S. Function of RARalpha during the maturation of neutrophils.Oncogene 2001; 20: 7178–7185.

25 Smith JC, Duchesne MA, Tozzi P, Ethier M, Figeys D. A differential phosphopro-teomic analysis of retinoic acid-treated P19 cells. J Proteome Res 2007; 6:3174–3186.

26 Uribesalgo I, Buschbeck M, Gutierrez A, Teichmann S, Demajo S, Kuebler B et al.E-box-independent regulation of transcription and differentiation by MYC. NatCell Biol 2011; 13: 1443–1449.

27 Minucci S, Nervi C, Lo Coco F, Pelicci PG. Histone deacetylases: a commonmolecular target for differentiation treatment of acute myeloid leukemias?Oncogene 2001; 20: 3110–3115.

28 Jaiswal H, Conz C, Otto H, Wolfle T, Fitzke E, Mayer MP et al. The chaperonenetwork connected to human ribosome-associated complex. Mol Cell Biol 2011;31: 1160–1173.

29 Shoji W, Inoue T, Yamamoto T, Obinata M. MIDA1a, protein associated with Id,regulates cell growth. J Biol Chem 1995; 270: 24818–24825.

30 Anguille S, Van Tendeloo VF, Berneman ZN. Leukemia-associated antigens andtheir relevance to the immunotherapy of acute myeloid leukemia. Leukemia 2012;26: 2186–2196.

31 Zuber J, Rappaport AR, Luo W, Wang E, Chen C, Vaseva AV et al. An integratedapproach to dissecting oncogene addiction implicates a Myb-coordinatedself-renewal program as essential for leukemia maintenance. Genes Dev 2011; 25:1628–1640.

32 Villa R, Pasini D, Gutierrez A, Morey L, Occhionorelli M, Vire E et al. Role of thepolycomb repressive complex 2 in acute promyelocytic leukemia. Cancer cell2007; 11: 513–525.

33 Huang da W, Sherman BT, Lempicki RA. Bioinformatics enrichment tools: pathstoward the comprehensive functional analysis of large gene lists. Nucleic Acids Res2009; 37: 1–13.

34 Frith MC, Fu Y, Yu L, Chen JF, Hansen U, Weng Z. Detection of functionalDNA motifs via statistical over-representation. Nucleic Acids Res 2004; 32:1372–1381.

35 Jolma A, Yan J, Whitington T, Toivonen J, Nitta KR, Rastas P et al. DNA-bindingspecificities of human transcription factors. Cell 2013; 152: 327–339.

36 Berge T, Matre V, Brendeford EM, Saether T, Luscher B, Gabrielsen OS. Revisiting aselection of target genes for the hematopoietic transcription factor c-Myb usingchromatin immunoprecipitation and c-Myb knockdown. Blood Cells Mol Dis 2007;39: 278–286.

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

ZRF1 controls the retinoic acid pathway and regulates leukemogenic potential in acute myeloid...

Documents