antagonistic modulation of gliomagenesis by pax6 and olig2 in pdgf-induced oligodendroglioma

Antagonistic modulation of gliomagenesis by Pax6 andOlig2 in PDGF-induced oligodendroglioma

Irene Appolloni1*, Filippo Calzolari2*, Manuela Barilari3, Marta Terrile4, Antonio Daga3 and Paolo Malatesta1,3

1 Department of Experimental Medicine (DIMES), University of Genoa, Genoa 16132, Italy2 Institute for Stem Cell Research, National Research Center for Environment and Health, Neuherberg D-85764, Germany3 IRCCS A.O.U. San Martino—IST, Largo Rosanna Benzi 10, Genoa 16132, Italy4 Department of Cancer Genetics, Royal College of Surgeons in Ireland, Dublin 2, Ireland

Gliomas are aggressive tumors of the central nervous system originating from proliferating neural cells. Regulators of neural

stem or progenitor cells biology may thus influence aspects of brain tumorigenesis, such as the maintenance of

tumor-propagating potential. We investigated the role of Pax6, a neurogenic transcription factor already suggested as a

positive prognostic marker for human gliomas, in a well-characterized in vivo model of PDGF-B-driven oligodendroglioma. In

this system, the expression of Pax6 severely impairs tumor propagation by inducing a reduction of cell proliferation and the

acquisition of differentiation traits in tumor-initiating cells. The overexpression of Pax6 correlates with a downregulation of

Olig2, a bHLH transcription factor that normally antagonizes Pax6 in adult neurogenic niches and that plays a key role in the

maintenance of neural stem and progenitor cells. Furthermore, we found that Olig2 is strictly required to maintain the

malignancy of oligodendroglioma cells, since its silencing by interfering RNA abrogates tumor propagation. We finally show

evidence that this function depends, at least in part, on the silencing of ID4, a dominant negative bHLH protein, whose

upregulation follows Olig2 loss. In our model, the upregulation of ID4 mimics the loss of Olig2 in impairing the

tumor-propagating potential of glioma cells. Our data, therefore, establish the relevance of physiological regulators of neural

stem cell biology in regulating glial tumor malignancy and provide support for their functional interactions in this context.

Gliomas are highly aggressive brain tumors originating fromsome of the different stem and progenitor cell populations ofthe adult mammalian brain.1 Important insights into thebiology of these tumors may thus come from studying therole of regulators of neural stem and progenitor biology. Twosuch molecules are the transcription factors Pax6 and Olig2,which modulate adult neurogenesis exerting divergent andopposing functions, by respectively promoting neurogenesisand oligodendrogliogenesis from stem cells residing in theperinatal and adult subventricular zone.2–4 Pax6 thus poten-tially opposes glial tumorigenesis by promoting the neuronaldifferentiation of the progeny of neural stem cells at theexpense of gliogenesis. Olig2, on the other hand, drives oligo-dendroglial lineage commitment and is required for properoligodendroglial differentiation.5–7 Aberrant execution of anOlig2-driven gliogenic program may thus represent a poten-tial early step in glial tumorigenesis, as suggested by in situobservations of nascent lesions in the neural stem cell niche8

and by the tumorigenic consequences of sustained Olig2expression in oligodendroglial progenitors activated by amyelin lesion.9 Furthermore, Olig2 is important for the pro-liferation of neural progenitors under normal conditions andfollowing EGFR-induced transformation.10,11 Moreover, whileOlig2 is expressed in all human diffuse gliomas,12 low Pax6levels correlate with higher malignancy in patients13 and itsforced expression negatively affects the tumorigenic potentialof xenotransplanted human glioma cell lines.14 Albeit intrigu-ing, these latter data may suffer from the well-appreciated

Key words: glioma, neural stem cells, neurogenesis, gene silencing,

murine model

Abbreviations: bHLH: basic helix loop helix; DsRed: red fluorescent

protein; EdU: 5-ethynyl-2’-deoxyuridine; FACS: fluorescence-

activated cell sorter; GFAP: glial fibrillary acidic protein; GFP: green

fluorescent protein; IRES: internal ribosome entry site; NPCs: neural

precursor cells; OPC: oligodendrocyte precursor cell; SVZ:

subventricular zone; TUNEL: terminal deoxynucleotidyl transferase

dUTP nick end labeling

Additional Supporting Information may be found in the online

version of this article.

*I.A. and F.C. contributed equally to this work.

Grant sponsors: Ministero della Salute; Grant number: GR-2008-

1135643; Grant sponsor: AIRC; Grant number: NUSUG 1180;

Grant sponsor: Fondazione S. Paolo (Molecular and cellular basis of

glioma), Fondazione CARIGE

DOI: 10.1002/ijc.27606

History: Received 7 Dec 2011; Accepted 11 Apr 2012; Online 19

Apr 2012

Correspondence to: Paolo Malatesta, Department of Experimental

Medicine (DIMES), University of Genoa, Via Leon Battista Alberti 2,

16132 Genoa, Italy, Tel.: þ390105737403, Fax: þ390105737405,

E-mail: [email protected]; or Filippo Calzolari, Institute for

Stem Cell Research, National Research Center for Environment and

Health, Ingolst€adter Landstrasse 1, Neuherberg D-85764, Germany,

E-mail: [email protected]

Can

cerCellBiology

Int. J. Cancer: 131, E1078–E1087 (2012) VC 2012 UICC

International Journal of Cancer

IJC

limitations of most commonly used human glioma cell lines,which do not faithfully reflect the behavior of glioma cell cul-tures established and maintained according to more recentstandards.15

In our study, we evaluated if Pax6 could affect tumor for-mation in a well characterized syngenic in vivo murine modelof high-grade oligodendroglial tumorigenesis based on theoverexpression of PDGF-B,16,17 a molecular event observed inmany human gliomas.18,19 In utero transduction of embry-onic neural progenitors with PDGF-B-encoding retrovirusesinduces the formation of oligodendroglial tumors,16 which pro-gress to full malignancy.17 These oligodendrogliomas can beserially propagated in vivo by orthotopic transplantations andtheir cells can be genetically manipulated in culture, allowingthe analysis of the role of putative modulators of the tumori-genic process.20 By using this model, we show that Pax6 canimpair tumor malignancy. Pax6 overexpression negativelyaffects the proliferation of tumor cells, induces their differentia-tion, and causes a reduction in the expression of the transcrip-tion factor Olig2. We further demonstrate an essential role forOlig2 by showing that its abrogation reduces tumor cell prolif-eration and the tumorigenic potential of PDGF-induced oligo-dendrogliomas. On Olig2 silencing, we also show that ID4, aninhibitor of oligodendroglial differentiation, is strongly upregu-lated, and demonstrate that ID4 mimics Olig2 silencing impair-ing the malignancy of PDGF-induced oligodendrogliomas.

Material and MethodsAnimal procedures

Mice were handled in agreement with current Italian regula-tions for the protection of animals used for scientific pur-poses (D.lvo 27/01/1992, n. 116). Procedures were approvedby the Ethical Committee for Animal Experimentation of theNational Institute of Cancer Research and by the Italian Min-istry of Health. All experiments were performed on the C57/Bl6 mouse strain as described previously.16,17 Primary tumorcultures from injected brains were established by dissectingDsRed-positive areas under a fluorescence microscope, andtrypsinizing them for 20 min. All tumor cultures were main-tained in DMEM-F12 supplemented with B27 supplementand human recombinant FGF2 and EGF (20 ng/ml) andplated on to Matrigel-coated flasks (1:200; BD Biosciences).Tumor cells were injected in deeply anesthetized animalsusing a stereotaxic stage. Up to 5 ll of cell suspension, con-taining 3,500 to 150,000 cells, were injected using a HamiltonSyringe (Bregma coordinates: AP, 1.0 mm; L, 1.5 mm leftand 2.5 mm below the skull surface).

Retroviral vectors and transduction procedures

Pax6/GFP retroviral vector encoding for the Pax6 isoformwith the canonical PD domain [which lacks the 14 aminoaci-dic insertion in the PD present in the isoform Pax6(5a)21]was provided by Dr. Goetz (Helmholtz Center, Munich,Germany). The coding sequence of mouse ID4 was provided

by Dr. Sabliztky (Univ. of Nottingham, GB) and cloned intothe pCAG:GFP retroviral vector (provided by Dr. M. Goetz)upstream the IRES-GFP region. Control experiments wereperformed with retroviral vectors coding for the GFP reporteralone. The two engineered Olig2-silencing miRNAs (mirOlig2)were designed using the BLOCK.iT RNAi Designer (Invitro-gen, Paisley, UK) and cloned with the EmGFP sequence intothe pCDB-GW retroviral vector as described elsewhere.20 As anegative control we cloned into the pCDB-GW plasmid themiRneg oligo supplied by the BLOCK-iTTM Pol-II miR RNAiExpression Vector kit (Invitrogen). Replication-defectiveretroviral supernatants were prepared by transiently transfect-ing plasmids into Phoenix packaging cells and harvesting thesupernatant after 2 days. The supernatants were concentratedby centrifugation and stored at �80�C before use.

Proliferation and differentiation analysis

For growth curves, two to three independent PDGF-B/Redcultures were transduced with the described retroviral vectorsand seeded onto matrigel-coated wells. Transduced cells wereharvested at different time points and the percentage of GFP-positive cells was counted in a hemocytometer. EdU (Invitro-gen) was added to the cultures 3 days after transduction andcells were fixed and stained, following manufacturer’s instruc-tions, after 32 hr. Cell-cycle analysis was performed by stain-ing transduced cells with Hoechst 33342 and measuring thelevel of fluorescence of each nucleus with an automated ZeissAxioImager M2 equipped with an Axiocam MRM. Thesedata were then utilized to calculate the cell-cycle parametersby using Multicycle AV software. The analysis was performed6 days after transduction with mirOlig2 and ID4, and 3 daysafter transduction with Pax6 to avoid a too dramatic reduc-tion of the percentage of transduced cells. Control analysiswith the control vector expressing only GFP was performedboth at 3 and 6 days after transduction with no significantdifferences. TUNEL staining was performed using the in situcell death detection kit (Roche, Milano, Italy) and followingmanufacturer’s instructions. Cell differentiation experimentswere performed maintaining transduced cells in mediawithout bFGF and EGF for 3–8 days before fixing.

Immunostainings

Primary antibodies: mouse monoclonal antibodies againstbIII-tubulin (1:100, Sigma-Aldrich, St. Louis, MO), O4 (1:800,Chemicon, Billerica, MA), GFAP (1:100, Sigma-Aldrich), Nes-tin (1:100, BD-Pharmingen, Franklin Lakes, NJ), NeuN (1:50,Chemicon); rabbit polyclonal antibodies against Olig2 (1:200,Chemicon), Pax6 (1:100, Covance, Rome, Italy), Nucleostemin(1:1,000, Chemicon), FoxG1 (1:1,000, Sigma-Aldrich customservice); rat polyclonal antibody against Ki67 (1:25; Dako, Mi-lano, Italy); chicken polyclonal antiserum against GFP (1:500,Abcam, Cambridge, UK).

Secondary antibodies and streptavidin: Alexa488- andAlexa546-conjugated antibodies (1:500, Immucor, Norcross,GA), anti-rat Cy3-conjugated antibody or biotinylated

Can

cerCellBiology

Appolloni et al. E1079


secondary antibodies (1:50, Dako). Streptavidin-conjugatedAlexa 350 (1:500, Molecular Probes, Paisley, UK). Nuclei werestained via 5 min incubation in DAPI solution (1 lg/ml,Sigma).

Immunostainings were analyzed directly under a fluores-cent microscope or automatically using an ImageJ plugin(available on request). ImageJ (Rasband, W.S., ImageJ, U.S.National Institutes of Health, Bethesda, MD, http://rsb.info.nih.gov/ij/, 1997–2007).

Levels of Pax6 achieved upon its overexpression in gliomacells, and the amount of endogenously expressed Olig2were analyzed by immunocytochemistry; Pax6 and Olig2immunoreactivities in glioma cells were compared to that ofembryonic neural telencephalic progenitors. Immunoreactiv-ity was measured on 16 bit microphotographs obtainedwith an AXIOCAM MRM (Zeiss, Oberkochen, Germany)using ImageJ.

All stainings were performed on cells or tissue sectionsfixed in 4% paraformaldehyde.

Real-time PCR

RNA was extracted from tumor cultures with QIAzol reagent(Qiagen, Hilden, Germany), following the manufacturer guide-lines. cDNA was obtained from 500 ng of RNA using theiScript (Bio-Rad Laboratories) retrotranscription kit. Quantita-tive real-time PCR was performed on 1/100 of the retrotran-scription reaction using iQ SyBr green supermix (Bio-Rad Lab-oratories, Segrate (MI), Italy). mRNA quantifications werenormalized to the housekeeping gene Rpl41 (NM_018860).Sequences of the primers used are available on request.

ResultsPax6 expression impairs the tumorigenic potential

of PDGF-induced oligodendroglioma cells

To assess the potential oncosuppressive role of the neuro-genic factor Pax6, we employed cultures obtained fromindependent murine high-grade oligodendrogliomas inducedby in utero transduction of PDGF-B and DsRed encodingretroviruses at embryonic day 14 (E14), as previouslyreported.16,17,20 These cultures, herein referred to as PDGF/Red glioma cells, when transplanted in adult mouse brains,are able to induce the formation of secondary tumors remi-niscent of the original one, showing features typical of high-grade oligodendrogliomas (Supporting Information Fig. 1).PDGF/Red glioma cells were infected in vitro with Pax6/GFPencoding retroviral vectors producing a mixed population ofuntransduced and transduced cells (Supporting InformationFig. 2a), with the latter stably integrating the provirus andectopically expressing Pax6 and GFP (Supporting Informa-tion Figs. 2b and 2c). The level of Pax6 overexpression drivenby the retroviral vector was found to be about three timeshigher than that exhibited by the neural progenitors of themouse embryonic forebrain which endogenously expressPax6 (Supporting Information Figs. 2b and 2c), consistent

with levels reported in a previous study using a similar vec-tor.22 Following Pax6 overexpression in glioma cells, weobserved a gradual and significant decrease in the relativeabundance of Pax6/GFP-expressing cells in the mixed (trans-duced/untransduced) glioma cell population, while controlcultures transduced with a GFP-only encoding retroviralvector, displayed stable percentages of GFP-positive cells(Fig. 1a). To clarify the reason for this dramatic effect, weevaluated the ability of Pax6 to affect apoptosis, proliferationand differentiation of glioma cells.

Pax6 overexpression only slightly affected cell survival inculture, since the frequency of apoptotic cells (showed byTUNEL) increased from 0.8 6 0.2% to 1.8 6 0.3% of Pax6-expressing cells. The absolute frequency of apoptotic cells inthese cultures was thus always very low even after Pax6transduction, suggesting for apoptosis a marginal role inmediating the effects of Pax6.

Cell proliferation was instead severely affected by Pax6,since the fraction of proliferating cells (growth fraction)dropped from 97 6 0.5% of the control to 34 6 5% of Pax6-expressing cells (p < 0.001), as shown by incorporation ofEdU after a pulse of 32 hr (Fig. 1b). To identify whether Pax6overexpression induced an accumulation of cells in a specificphase of the cell cycle, we performed an analysis of DNA con-tent of GFP-positive cells (Supporting Information Fig. 3). Wenoticed that Pax6 transduction induces a strong increase ofthe proportion G0/G1-phase cells at the expenses of cells inG2-phase which almost disappeared (Supporting InformationFig. 3). These data were paralleled by a decrease of KI67-posi-tive cells from 87 6 4% to 60 6 2% (p < 0.01; Supporting In-formation Figs. 4a and 4b). All these data clearly demonstratedthat Pax6 expression provokes a very strong decrease of cellproliferation probably due to a block of cells in G0 phase.

Along with a reduction of the growth fraction, Pax6expression promoted the acquisition of more pronouncedneuron-like features, while inhibiting oligodendrocyte differ-entiation. This effect was assessed in culture conditions thatallow differentiation (i.e., in absence of EGF and bFGF). Inthese conditions, Pax6 overexpression, compared to controlvector, caused a decrease of the stem cell markers nucleoste-min23 from 71 6 4% to 29 6 8% (p < 0.01) and FoxG124

from 72 6 2% to 53 6 2% (p < 0.01), an increase of the neu-ronal marker bIII-tubulin from 25 6 4% to 36 6 4% (p <

0.05) and a decrease of the oligodendrocyte differentiationmarker O4 from 12 6 1% to 7 6 1% (p < 0.05; Fig. 1c andSupporting Information Figs. 4c and 4d). The expression ofthe astroglial marker GFAP, scarcely expressed by PDGF-induced oligodendroglioma cells,17 did not change (GFAP-pos-itive cells < 0.1%; Supporting Information Figs. 4e and 4f).

We then analyzed whether the effect of Pax6 on cell pro-liferation and differentiation would affect the tumor-propa-gating potential of PDGF-induced gliomas in vivo. PDGF/Red tumor cells transduced with Pax6/GFP retroviruses wereharvested few days after infection and the resulting mixedcell population (constituted by about 50% transduced cells)

Can

cerCellBiology

E1080 Role of Pax6 and Olig2 in oligodendrogliomas


was directly injected intracranially into the brains of adultmice20 (n ¼ 15; Fig. 2a). Remarkably, two thirds of the gen-erated tumors lacked any visible GFP-positive fraction andamong the remaining all but one showed a reduced amountof GFP-positive cells. Interestingly, the immunostaining ofthe few Pax6-transduced cells retrieved in the tumors showedthat they were negative for nestin (Supporting InformationFig. 5a) corroborating the hypothesis of a loss of their imma-ture features. However they did not show immunoreactivityfor differentiated cell markers such as GFAP or NeuN (Sup-porting Information Figs. 5b and 5c). An abundant GFP-pos-itive cell fraction was present in all tumors derived fromtransplantations of a mixed population obtained with the

control retroviral vector (n ¼ 15; Fig. 2b). Altogether, theseresults demonstrate that the expression of Pax6 limits themalignancy of PDGF-induced gliomas, by limiting the prolif-eration of cancer-propagating cells, likely by inducing theirdifferentiation.

Pax6 downregulates Olig2 which is essential for the

tumorigenic potential of PDGF-induced oligodendrogliomas

The expression of Pax6 is known to induce telencephalic pro-genitor and neural stem cells to arrest their proliferation andundergo neuronal differentiation,22,25 likely at the expense ofglial production from SVZ progenitors.2,4 An opposite role inneural stem cells is played by the bHLH transcription factor

Figure 1. Pax6 overexpression in PDGF-B induced glioma cells affects cell proliferation and induces cell differentiation. (a) Fraction of GFP-

positive cells at different time points after transduction with Pax6-expressing or control vector. (b) Percentage of EdU incorporating cells

after a pulse of 32 hr. (c) Percentage of nucleostemin-, bIII-tubulin- or O4-positive cells among GFP-positive cells and corresponding

immunostainings. Scale bar: 20 lm.

Can

cerCellBiology



Olig2, which was suggested to inhibit neuronal differentiationand Pax6 expression4,26 and that has been shown to bedirectly inhibited by Pax6 at the transcriptional level.3 SincePDGF-induced gliomas strongly resemble oligodendrocyteprecursor cells (OPCs)16,17 and they express similar levels ofOlig2 compared to primary OPCs in vitro (Supporting Infor-mation Fig. 6a), we were prompted to analyze whether theeffect of Pax6 on tumorigenesis could be mediated by aneffect on Olig2.

We therefore transduced Pax6 in PDGF/Red cultures andfound that the percentage of Olig2 positive cells droppedfrom 95 6 3% to 40 6 17% (p < 0.01; Fig. 3a). A similareffect was observed when forcing Pax6 expression in vitro inventral telencephalic embryonic progenitors, which normallyexpress Olig2 but not Pax6 (Supporting Information Fig. 6b).We then evaluated the role of Olig2 in regulating the survivaland proliferation of PDGF/Red glioma cells by silencingOlig2 with distinct retroviral vectors expressing two inde-pendent engineered anti-Olig2 microRNAs (mirOlig2) andGFP. Both silencing constructs efficiently abrogated Olig2expression (Supporting Information Fig. 6d) but this did not

result in the activation of Pax6 expression, neither in tumorcells, nor in ventral embryonic neural progenitors (Support-ing Information Fig. 6c), suggesting that, also in these sys-tems, Olig2 lays genetically downstream Pax6.

Analogously to Pax6 overexpression, the silencing of Olig2negatively affected in vitro expansion of tumor cells, whoseabundance rapidly and significantly declined (Fig. 3b).Although Olig2 silencing did not significantly modify cell-cycle parameters, only slightly increasing the percentage ofcells in G1/G0 phases at expenses of that in G2/M phases(Supporting Information Fig. 3), it caused a significant reduc-tion of the growth fraction of the PDGF/Red cells, as shownby the decrease of KI67-positive cells from 87 6 4% to 68 6

2% (p < 0.01; Supporting Information Figs. 4a and 4b). Thisproliferative effect was coupled with a decrease in the fractionof nucleostemin-positive cells from 71 6 4% to 38 6 5% (p< 0.01) and of FoxG1-expression from 72 6 2% to 57 6 4%(p < 0.05), observed in absence of bFGF and EGF. Thesechanges were not paralleled by changes in the proportion ofbIII-tubulin and O4 positive cells (Fig. 3c and Supporting In-formation Figs. 4c and 4d). The expression of the astrocytic

Figure 2. Pax6 negatively affects tumorigenesis from PDGF-B expressing glioma cells. (a) Design of the orthotopic transplantation

experiments. (b) Fluorescence images of tumors generated by Pax6- or control-transduced glioma cells and quantification of the percentage

of tumors constituted by DsRed-only or Mixed DsRed-GFP-labeled cells. Scale bar: 1 mm.

Can

cerCellBiology



marker GFAP remained hardly detectable, as in the control(GFAP-positive cells < 0.1%; Supporting Information Figs. 4eand 4f). The percentage of apoptotic cells, analyzed by

TUNEL, was unchanged in Olig2 silenced cells (1.1 6 0.2%)compared to control cells (0.8 6 0.2%). These data indicatethat Olig2 is required for the proper regulation of

Figure 3. Olig2 is essential for tumorigenesis in PDGF-B induced gliomas. (a) Immunostaining for Olig2 performed on Pax6- or control-

transduced glioma cells and corresponding quantification of the percentage of Olig2-positive among GFP-positive cells. (b) Fraction of GFP-

positive cells at different time points after transduction with mirOlig2-expressing or control vector. (c) Percentage of O4-, bIII-tubulin- ornucleostemin-positive cells in mirOlig2- or control-transduced cells and corresponding immunofluorescence stainings. (d) Fluorescence

images of tumors generated by mirOlig2 or control-transduced glioma cells and quantification of the percentage of tumors constituted by

DsRed-only or Mixed DsRed-GFP-labeled cells. Scale bars: 20 lm (a,c); 0.5 mm (d).

Can

cerCellBiology



proliferation of oligodendroglioma cells and suggest that partof the in vitro defects resulting from Pax6 overexpressionmay be due to its negative effects on Olig2 expression levels.

To assess the role of Olig2 in tumor propagation, we usedthe same experimental paradigm employed for Pax6. A popu-lation of PDGF/Red glioma cells showing about 50% of cellstransduced with the GFP/Olig2-silencing vector, or with thecontrol vector, was directly injected intracranially into thebrains of adult mice. While most gliomas (87%) developingfrom the injection of control-transduced cells (n ¼ 15) con-tained GFP-positive cells, only 1 of 16 gliomas generated af-ter the injection of mirOlig2 harbored a GFP-positive fraction(Fig. 3d), showing a strongly impaired tumor-propagating ac-tivity in oligodendroglioma cells lacking Olig2 expression. Todemonstrate that this effect was due to a cell-autonomousloss of tumorigenic potential by Olig2 silenced tumor cells,rather than to proliferative out-competition by wild type tu-mor cells, we FACS-purified the GFP-positive component ofthe PDGF/Red glioma cells transduced with the GFP/Olig2-silencing vector and transplanted them into the brain oftwo adult mice. Only one mouse developed a tumor but itwas exclusively constituted by DsRed-positive cells likelyderived by the small minority of untransduced PDGF/Redcells contaminating the sorted population (about 7%,accounting for <300 cells, Supporting Information Fig. 6f).As expected, control mice transplanted with the complemen-tary GFP-negative FACS-purified cells developed purelyDsRed positive oligodendrogliomas (Supporting InformationFig. 6e).

This observation demonstrates that Olig2 is essential forthe tumor-propagating potential of PDGF-induced gliomas,and suggests that Pax6 effects on tumorigenesis can beexplained by its negative regulation of Olig2.

ID4 is upregulated after Olig2 loss and impairs

proliferation and tumorigenesis

We then evaluated the possibility that ID4, a known antago-nist of Olig2 function that mediates the negative effectsexerted by BMP signaling on oligodendroglial development,27

could be responsible for the consequences of Olig2 silencing.ID4 is transcriptionally upregulated in embryonic neural pro-genitors from Olig2 knockout mice,10 but a direct binding ofOlig2 on ID4 promoter is not yet described. By real-timePCR, we demonstrated that ID4 mRNA levels in PDGF/Redcells were strongly increased following Olig2 silencing (Fig.4a), and also following Pax6 overexpression (data notshown). On the contrary, forced ID4 expression did notinfluence Olig2 levels (Fig. 4b). We therefore investigated therole of ID4 by overexpressing it in PDGF/Red glioma cells byretroviral transduction.

Interestingly, overexpression of ID4 mimicked Olig2silencing and caused a significant decrease of proliferation inglioma cells, resulting in decreased cell abundance over time(Fig. 4c), reduced EdU incorporation after a 32 hr pulse from97 6 1% to 85 6 4% (p < 0.01) and reduced immunoposi-

tivity to KI67 (Supporting Information Fig. 4a and 4b). Simi-larly to what found after Olig2 silencing, the distribution ofthe cells within the cell-cycle phases (Supporting InformationFig. 3) and the percentage of apoptotic cells was unchanged(data not shown). Moreover, in absence of bFGF and EGF,ID4 promoted the acquisition of more pronounced neuron-like features, while inhibiting oligodendrocyte differentiationas shown by the decreased fraction of nucleostemin-positivecells from 71 6 4% to 52 6 4% (p < 0.01) and FoxG1-posi-tive cells from 72 6 2% to 60 6 4% (p < 0.05), the increaseof bIII-tubulin-positive cells from 25 6 4% to 37 6 3% (p <

0.05) and the decrease of O4-positive cells from 12 6 1% to1 6 0.2% (p < 0.05; Fig. 4d and Supporting InformationFigs. 4c and 4d). The expression of GFAP did not change(Supporting Information Figs. 4e and 4f).

Most importantly ID4-overexpression in PDGF/Red cells,analogously to the effect of Olig2 silencing, caused a dramaticreduction of the tumorigenic potential in vivo. In the sameexperimental paradigm described above for Pax6 and mirO-lig2, only 3 of 11 tumors contained a fraction of GFP/ID4positive cells, while a GFP-positive fraction was present in91% of control tumors (n ¼ 11, Fig. 4e), suggesting that ID4upregulation may mediate at least part of the effect of Olig2loss.

DiscussionElucidation of basic aspects of the biology of glial tumors,such as the molecular mechanisms sustaining their malig-nancy, is necessary for improving our ability to fight thesecancers. Important insights can come from exploring the par-allels between physiological and pathological processes. Usinga well characterized in vivo model of oligodendrogliomagene-sis,16,17 we show that transcription factors involved in nerv-ous system development and homeostasis modulate tumori-genesis. Our observations indicate that Pax6 overexpressioninduces differentiation and impairs the proliferation and tu-mor-propagation potential of oligodendroglioma cells. DNAcontent analysis, showed that Pax6 transduced cells tended toaccumulate in G0/G1 phase, compatible with the observeddifferentiation towards the neuronal lineage. The overexpres-sion experiments were performed with the coding sequenceof the Pax6 isoform bearing the canonical PD domain [whichlacks the 14 aminoacid insertion found in the isoform namedPax6(5a)21]. However, since Pax6 positively self-regulates theexpression of both its isoforms,28 we cannot exclude that partof the effects we observed following the overexpression ofPax6 depend on the induced Pax6(5a) isoform.

Moreover, analogously to what we and other authorsobserved in normal progenitor cells,3 Pax6 expressionabrogates the expression of Olig2, a master regulator of oligo-dendroglial biology. Olig2 is already known to be transcrip-tionally regulated by Pax6 in related systems3 and to be fun-damental for tumorigenic features of EGFRvIII inducedgliomagenesis.10 Noticeably, we show that Olig2 silencing,similarly to Pax6 overexpression, reduces proliferation

Can

cerCellBiology



in vitro in high-grade oligodendrogliomas and causes a dra-matic reduction of tumorigenesis in vivo. We also found thatloss of Olig2 is accompanied by increased levels of a known

antagonist of oligodendroglial differentiation, ID4, and weshow that forced ID4 expression mimics Olig2 loss, suppress-ing growth and tumorigenesis.

Figure 4. ID4 overexpression negatively affects tumorigenesis from PDGF-B expressing glioma cells. (a) ID4 mRNA level in mirOlig2- and

control-transduced glioma cells, normalized to the rpl41 housekeeping gene. (b) Percentage of Olig2-positive cells in PDGF/Red cells

transduced with ID4- or control-expressing vectors. (c) Fraction of GFP-positive cells at different time points after transduction with

ID4-expressing or control vector. (d) Percentage of O4-, bIII-tubulin- or nucleostemin-positive cells in ID4- or control-transduced cells and

corresponding immunofluorescence stainings. (e) Fluorescence images of tumors generated by ID4 or control-transduced glioma cells and

quantification of the percentage of tumors constituted by DsRed-only or Mixed DsRed-GFP-labeled cells. Scale bars: 20 lm (d); 2mm (e).

Can

cerCellBiology



Interestingly, Olig2 silencing caused milder effects on cellcycle and neuronal differentiation when compared to overex-pression of Pax6, suggesting that Olig2 downregulation andthe induction of a more differentiated phenotype may be in-dependently triggered by Pax6. Moreover, Olig2 silencing hada strong impact on in vivo tumorigenesis, suggesting that thePax6-induced reduction in malignancy may mostly beaccounted for by loss of Olig2 expression. Interestingly, ID4-induced effects on the in vivo and in vitro behavior of gliomacells were somehow intermediate between those elicited byPax6 overexpression and Olig2 silencing. However, it mustbe noted that while loss of Olig2 results in ID4 upregulation,the forced expression of ID4 does not obviously affect Olig2levels or localization, suggesting that the presence of Olig2may decrease the impact of ID4 on glioma cells.

A meta-analysis performed on a public dataset(GSE16011), in line with our data, shows that Pax6 expres-sion is often downregulated in human gliomas while Olig2results highly upregulated in all human glioma histopatholog-ical subtypes (Supporting Information Fig. 7a), thus suggest-ing the suitability of our murine model to represent thehuman pathology. With regard to ID4, we noticed, analo-gously to other studies using different models, that it is usu-ally overexpressed in glial tumors.29,30 Moreover, someauthors suggest that ID4 may promote, rather than restrain,tumorigenesis,29 possibly reflecting different roles of ID4 indifferent cell types. However, our bioinformatics study ontwo different public data sets (GSE16011 and NCI REM-BRANDT, 2005; accessed 2011 April), clearly showed that,although ID4 is overexpressed in all subtype of gliomas,patients showing higher levels of ID4 are characterized bylonger survival times, also when limiting the analysis only tothose diagnosed for glioblastoma (Supporting InformationFig. 7b). These data are therefore in line with our observationin which strong ID4 induction, due to retroviral transduction,

caused a reduction of the tumorigenic potential of gliomacells, thus strengthening the relevance of our animal model.Furthermore, ID4 is a known mediator of BMP signaling,which negatively affects oligodendrogenesis while promotingadult neurogenesis.31,32 BMP ligands are also well knowninhibitors of gliomagenesis,33 and our data suggest that sucheffects may depend on interferences with the activity ofOlig2. Interestingly, Olig2 is also required to maintain thetumorigenic potential of murine gliomas induced by overex-pressing the glioma-associated EGFRvIII oncogene inp16ink4a/p19arf null neural precursors,10,11 and its loss leadsto ID4 upregulation in that system, in agreement with ourobservations. All these data may point to a general role forOlig2 in the maintenance of the tumorigenic potential inmultiple glioma subtypes, suggesting that targeting Olig2, ormediators of its effects on tumorigenesis, could have broadtherapeutic applications.

In summary, our data point to the relevance of importantphysiological regulators of neural stem cell activity and theirpotential interaction in regulating the malignancy of gliomas.We demonstrate an oncosuppressive role for the neurogenictranscription factor Pax6 and highlight the central role playedin this pathway by Olig2 in the maintenance of the tumori-genic potential of PDGF-induced oligodendrogliomas. Fur-ther, we provide evidence that Olig2 sustains malignancy byrestraining the expression of ID4, a known antagonist of oli-godendroglial differentiation.

AcknowledgementsThe authors remember the late Prof. Giorgio Corte for the many helpful dis-cussions they had together. They thank Dr. Daniele Reverberi and Dr. Fabri-zio Loiacono (IRCCS A.O.U. San Martino – IST) for FacSorting. They alsothank Dr. M. Goetz (Institute of Stem Cell Research, Helmholtz Center, Mu-nich, Germany) and Dr. F. Sabliztky, (Univ. of Nottingham, GB) for sharingplasmids.

References

1. Canoll P, Goldman JE. The interface betweenglial progenitors and gliomas. Acta Neuropathol2008;116:465–77.

2. Hack MA, Saghatelyan A, de Chevigny A, et al.Neuronal fate determinants of adult olfactorybulb neurogenesis. Nat Neurosci 2005;8:865–72.

3. Jang ES, Goldman JE. Pax6 expression issufficient to induce a neurogenic fate in glialprogenitors of the neonatal subventricular zone.PLoS One 2011;6:e20894.

4. Marshall CA, Novitch BG, Goldman JE. Olig2directs astrocyte and oligodendrocyte formationin postnatal subventricular zone cells. J Neurosci2005;25:7289–98.

5. Copray S, Balasubramaniyan V, Levenga J, et al.Olig2 overexpression induces the in vitrodifferentiation of neural stem cells into matureoligodendrocytes. Stem Cells 2006;24:1001–10.

6. Takebayashi H, Nabeshima Y, Yoshida S, et al.The basic helix-loop-helix factor olig2 is essentialfor the development of motoneuron andoligodendrocyte lineages. Curr Biol 2002;12:1157–63.

7. Zhou Q, Choi G, Anderson DJ. The bHLHtranscription factor Olig2 promotesoligodendrocyte differentiation in collaborationwith Nkx2.2. Neuron 2001;31:791–807.

8. Wang Y, Yang J, Zheng H, Tomasek GJ, ZhangP, McKeever PE, Lee EY, Zhu Y. Expression ofmutant p53 proteins implicates a lineagerelationship between neural stem cells andmalignant astrocytic glioma in a murine model.Cancer Cell 2009;15:514–26.

9. Kim HM, Hwang DH, Choi JY, Park CH, Suh-Kim H, Kim SU, Kim BG. Differential andcooperative actions of Olig1 and Olig2transcription factors on immature proliferatingcells after contusive spinal cord injury. Glia 2011;59:1094–106.

10. Ligon KL, Huillard E, Mehta S, et al. Olig2-regulated lineage-restricted pathway controlsreplication competence in neural stem cells andmalignant glioma. Neuron 2007;53:503–17.

11. Sun Y, Meijer DH, Alberta JA, et al.Phosphorylation state of olig2 regulates

proliferation of neural progenitors. Neuron 2011;69:906–17.

12. Ligon KL, Alberta JA, Kho AT, et al. Theoligodendroglial lineage marker OLIG2 isuniversally expressed in diffuse gliomas. JNeuropathol Exp Neurol 2004;63:499–509.

13. Zhou YH, Tan F, Hess KR, et al. The expressionof PAX6, PTEN, vascular endothelial growthfactor, and epidermal growth factor receptor ingliomas: relationship to tumor grade and survival.Clin Cancer Res 2003;9:3369–75.

14. Zhou YH, Wu X, Tan F, et al. PAX6 suppressesgrowth of human glioblastoma cells. J Neurooncol2005;71:223–9.

15. Li A, Walling J, Kotliarov Y, et al. Genomicchanges and gene expression profiles reveal thatestablished glioma cell lines are poorlyrepresentative of primary human gliomas. MolCancer Res 2008;6:21–30.

16. Appolloni I, Calzolari F, Tutucci E, et al. PDGF-B induces a homogeneous class ofoligodendrogliomas from embryonic neuralprogenitors. Int J Cancer 2009;124:2251–9.

Can

cerCellBiology



17. Calzolari F, Appolloni I, Tutucci E, et al.Tumor progression and oncogene addictionin a PDGF-B-induced model of gliomagenesis.Neoplasia 2008;10:1373–82, following82.

18. Brennan C, Momota H, Hambardzumyan D,et al. Glioblastoma subclasses can be defined byactivity among signal transduction pathways andassociated genomic alterations. PLoS One 2009;4:e7752.

19. Verhaak RG, Hoadley KA, Purdom E, et al.Integrated genomic analysis identifiesclinically relevant subtypes of glioblastomacharacterized by abnormalities in PDGFRA,IDH1, EGFR, and NF1. Cancer Cell 2010;17:98–110.

20. Terrile M, Appolloni I, Calzolari F, et al. PDGF-B-driven gliomagenesis can occur in the absenceof the proteoglycan NG2. BMC Cancer 2010;10:550.

21. Haubst N, Berger J, Radjendirane V, et al.Molecular dissection of Pax6 function: thespecific roles of the paired domain andhomeodomain in brain development.Development (Cambridge, England) 2004;131:6131–40.

22. Heins N, Malatesta P, Cecconi F, et al. Glial cellsgenerate neurons: the role of the transcriptionfactor Pax6. Nat Neurosci 2002;5:308–15.

23. Tsai RY, McKay RD. A nucleolar mechanismcontrolling cell proliferation in stem cells andcancer cells. Genes Dev 2002;16:2991–3003.

24. Fasano CA, Phoenix TN, Kokovay E, et al. Bmi-1cooperates with Foxg1 to maintain neural stemcell self-renewal in the forebrain. Genes Dev 2009;23:561–74.

25. Hack MA, Sugimori M, Lundberg C, et al.Regionalization and fate specification inneurospheres: the role of Olig2 and Pax6. MolCell Neurosci 2004;25:664–78.

26. Buffo A, Vosko MR, Erturk D, et al. Expressionpattern of the transcription factor Olig2 inresponse to brain injuries: implications forneuronal repair. Proc Natl Acad Sci U S A 2005;102:18183–8.

27. Samanta J, Kessler JA. Interactions between IDand OLIG proteins mediate the inhibitory effectsof BMP4 on oligodendroglial differentiation.Development (Cambridge, England) 2004;131:4131–42.

28. Pinson J, Simpson TI, Mason JO, et al. Positiveautoregulation of the transcription factor Pax6 in

response to increased levels of either of its majorisoforms, Pax6 or Pax6(5a), in cultured cells.BMC Dev Biol 2006;6:25.

29. Kuzontkoski PM, Mulligan-Kehoe MJ, Harris BT,et al. Inhibitor of DNA binding-4 promotesangiogenesis and growth of glioblastomamultiforme by elevating matrix GLA levels.Oncogene 2010;29:3793–802.

30. Liang Y, Bollen AW, Nicholas MK, et al. Id4 andFABP7 are preferentially expressed in cells withastrocytic features in oligodendrogliomas andoligoastrocytomas. BMC Clin Pathol2005;5:6.

31. Colak D, Mori T, Brill MS, et al. Adultneurogenesis requires Smad4-mediated bonemorphogenic protein signaling in stem cells. JNeurosci 2008;28:434–46.

32. Jablonska B, Aguirre A, Raymond M, et al.Chordin-induced lineage plasticity of adult SVZneuroblasts after demyelination. Nat Neurosci2010;13:541–50.

33. Piccirillo SG, Reynolds BA, Zanetti N, et al.Bone morphogenetic proteins inhibit thetumorigenic potential of human braintumour-initiating cells. Nature2006;444:761–5.

Can

cerCellBiology



antagonistic modulation of gliomagenesis by pax6 and olig2 in pdgf-induced oligodendroglioma

Documents