Inhibition of Hsp90 Via 17-DMAG Induces Apoptosis in a P53-Dependent Manner to Prevent Medulloblastoma

Inhibition of Hsp90 via 17-DMAG induces apoptosis in a p53-dependent manner to prevent medulloblastoma

Olivier Ayraulta,1, Michael D. Godenyb,1, Christopher Dillonb, Frederique Zindya, Patrick Fitzgeraldb, Martine F. Roussela, and Helen M. Beereb,2

Departments of aGenetics and Tumor Cell Biology and bImmunology, St. Jude Children’s Research Hospital, 262 Danny Thomas Place, Memphis, TN 38105

Edited by Susan L. Lindquist, Whitehead Institute for Biomedical Research, Cambridge, MA, and approved August 18, 2009 (received for review March 20, 2009) Elevated expression of HSP90 is observed in many tumor types and to be challenged by features of tumor cells that confer resistance is associated with a limited clinical response. Targeting HSP90 to conventional therapies. using inhibitors such as 17-DMAG (17-desmethoxy-17-N,N-dimeth- Elevated expression of HSP90 in human medulloblastoma ylaminoethylaminogeldanamycin) has shown limited therapeutic (16) suggests that it may represent a candidate for therapeutic success. HSP90 regulates the function of several proteins impli- intervention in this disease. Here, we sought to preclinically cated in tumorigenesis although the precise mechanism through evaluate if the HSP90 inhibitor, 17-DMAG, affects the growth which 17-DMAG regulates tumor cell survival remains unclear. We of medulloblastoma, a form of pediatric cancer arising in the observed a requirement for p53 in mediating 17-DMAG-induced cerebellum that develops largely after birth due to the failure of cell death. The sensitivity of primary mouse embryonic ﬁbroblasts granule neuron precursors (GNPs) to exit the cell cycle and and tumor cells to 17-DMAG-induced apoptosis depended on the differentiate (17). This aberrant process has been linked to p53 status. Wild-type MEFs underwent 17-DMAG-induced caspase- human medulloblastomas involving TP53 inactivation (10% of dependent cell death, whilst those lacking p53 failed to do so. human cases), defective Sonic Hedgehog/PATCHED (SHH/ Interestingly p53-dependent cell death occurred independently of PTCH) signaling (27% of human cases), lesions in the WNT Atm or Arf. Primary tumor cells derived from two models of murine signaling pathway (15% of human cases), as well as the persistent ؉ ؊ ؊ ؊ ؉ medulloblastoma (Ptch1 / ;Ink4c / and p53FL/FL;Nestin-Cre ; expression of pro-proliferative genes (18). ؊ ؊ Ink4c / ) that retain and lack p53 function, respectively, displayed Several murine models for medulloblastoma that recapitulate a dependence on functional p53 to engage 17-DMAG-induced causative genetic lesions identified in human medulloblastoma (19) apoptosis. Strikingly, 17-DMAG treatment in an allograft model of are characterized by activation of the Shh/Ptch signaling pathway, ؉ ؊ ؊ ؊ ؉ ؊ ؊ Ptch1 / ;Ink4c / but not p53FL/FL;Nestin-Cre ; Ink4c / tumor two of which were used in our studies (20). The first (denoted cells prevented tumor growth in vivo. Our data suggest that p53 throughout as Ptch1ϩ/Ϫ;Ink4cϪ/Ϫ) was generated through a germ- status is a likely predictor of the sensitivity of tumors to 17-DMAG. line deletion of one copy of the Patched gene (Ptch1), the receptor CELL BIOLOGY for Shh (21), which, when combined with the deletion of Ink4c (22), cquisition of inactivating p53 mutations or aberrant expression induce medulloblastomas with an approximate 60% incidence (23). Aof signaling molecules that engage p53 are extremely common Importantly, all tumors retain functional p53 (20, 23) but lose in tumors and can render them refractive to conventional therapies expression of the wt Ptch1 allele (20). The second model (denoted (1). Because the tumor suppressor activity of p53 is mediated largely throughout as p53FL/FL;Ink4cϪ/Ϫ) is generated by the conditional by its ability to engage apoptosis, its inactivation provides tolerance deletion of floxed p53 in the cerebellum using Cre recombinase to the tumor microenvironment (1) and is analogous to the survival under the control of the Nestin promoter (24) which, when com- promoting effects of heat shock proteins (HSPs) that, in response bined with germ line deletion of Ink4c and irradiation of postnatal to stresses including hypoxia and nutrient deprivation and in day 7 mice displays complete penetrance of medulloblastomas (20). collaboration with co-chaperone proteins, regulate the refolding Importantly, tumors arising in p53-deficient mice are characterized and repair of damaged proteins (2). By doing so, they preserve by the homozygous deletion of Ptch1 (20). Therefore, both the protein function and maintain cellular survival in part by preventing Ptch1ϩ/Ϫ;Ink4cϪ/Ϫ and p53FL/FL;Ink4cϪ/Ϫ models of medulloblas- apoptosis (3). The expression of several HSPs, including HSP90 is toma are characterized by the constitutive activation of Shh/Ptch increased in tumors (4), suggesting that elevated HSP expression signaling, regardless of their founding mutations, with the only may contribute to aberrant tumor survival. New clinical strategies difference being the presence or absence of functional p53. aim to exploit this weakness by targeting components of the stress Using in vitro assays in primary wt and p53-null mouse pathway (5). embryonic fibroblasts (MEFs) and purified GNP-like tumor HSP90 displays the unique ability to selectively associate with cells, we show that 17-DMAG induced apoptosis in a p53- and signaling molecules implicated in the aberrant survival of tumor caspase-dependent manner that required Puma or Bax/Bak, but cells (4). These include mutant (6, 7) and wild-type (wt) p53 (8), was independent of p19Arf and Atm signaling. Transfer of tumor Raf-1 (9), and Akt (10). HSP90 is ubiquitously expressed in both cells derived from each of the murine models into immunocom- normal and malignant tissues, but its altered ‘high-affinity’ conformation in tumor cells confers 100-fold selectivity for HSP90 inhibitors (11). Consequently, several HSP90 inhibitors Author contributions: O.A., M.D.G., M.F.R., and H.M.B. designed research; O.A., M.D.G., derived from the ansamycin antibiotic geldanamycin (GA) are in C.D., and F.Z. performed research; P.F. contributed new reagents/analytic tools; O.A., clinical trials for the treatment of cancer (5, 12, 13). Ansamycin M.D.G., C.D., M.F.R., and H.M.B. analyzed data; and M.F.R. and H.M.B. wrote the paper. compounds bind tightly to the ATP-binding pocket of HSP90 to The authors declare no conﬂict of interest. prevent its stable interaction with substrates and to target them This article is a PNAS Direct Submission. for proteasomal degradation (14, 15). HSP90 inhibitors have 1O.A. and M.D.G. contributed equally to this article. shown promising but limited signs of clinical activity (5, 12, 13). 2To whom correspondence should be addressed. E-mail: [email protected]. It therefore remains important to understand how 17-DMAG This article contains supporting information online at www.pnas.org/cgi/content/full/ acts as an effective anti-tumor agent and if its efficacy is likely 0902880106/DCSupplemental.

www.pnas.org͞cgi͞doi͞10.1073͞pnas.0902880106 PNAS ͉ October 6, 2009 ͉ vol. 106 ͉ no. 40 ͉ 17037–17042 Downloaded by guest on October 1, 2021 promised recipients demonstrated that 17-DMAG efficiently prevented medulloblastoma tumor formation and growth in vivo but only when p53 was functional. Our studies establish a relationship between Hsp90 and p53 activity in vivo and provide evidence that the Hsp90 inhibitor, 17-DMAG requires an intact p53 response to exert its anti- tumorigenic effect. Although the relevance of these findings in a clinical setting remains to be examined, they predict that HSP90 inhibitors may be an effective treatment option for human medulloblastoma, a tumor type in which a significant percentage of tumors retain functional p53.

Results Inhibition of Hsp90 Engages a p53-Dependent Pathway to Apoptosis in Primary Mouse Embryo Fibroblasts. The relevance of p53 status in response to Hsp90 inhibition in non-transformed cells was tested using primary wt p53 (p53ϩ/ϩ) and p53-deficient (p53Ϫ/Ϫ) MEFs. 17-DMAG-induced cell death in p53ϩ/ϩ cells was significantly reduced in their p53Ϫ/Ϫ counterparts (Fig. 1A). In addition, MEFs expressing a tamoxifen inducible p53-ER fusion protein (p53- ERtam) (25) were sensitive to 17-DMAG-induced cell death but only when tamoxifen was present to engage the activity of p53-ER (Fig. 1B). These data indicate that p53 is an important modulator of 17-DMAG-induced cell death. DNA damage or oncogenes engage the p53 response via activation of Atm or p19Arf, respectively. However, we failed to observe any reduced sensitivity to 17-DMAG in either ArfϪ/Ϫ or AtmϪ/Ϫ MEFs as compared to wt MEFs (Fig. 1 C and D), suggesting that neither pathway is necessary for 17-DMAG-induced cell death. The pathways through which p53 engages apoptosis universally require the pro-apoptotic multidomain proteins Bax and Bak. p53 can activate Bax either directly (26), independently of its transcrip- tional activity or indirectly by inducing expression of Puma (27). We observed that 17-DMAG induced apoptosis in wt MEFs but not in those lacking both Bax and Bak (Fig. 1E) or Puma (Fig. 1F), suggesting that p53-dependent 17-DMAG-induced cell death required Puma or Bax and Bak.

Fig. 1. Inhibition of Hsp90 engages a p53-dependent pathway to apoptosis. Hsp90AA1 Protein and RNA Levels Are Elevated in Primary GNP-Like Primary MEFs from embryos expressing wt p53 (p53ϩ/ϩ) or lacking p53 (p53Ϫ/Ϫ) Cells Isolated from Murine Medulloblastomas. Hsp90AA1 protein were treated with 17-DMAG (0–10 ␮M) overnight and cell death determined in levels were elevated in GNP-like tumor cells isolated from medul- three independent experiments (A). MEFs isolated from mice expressing a ta- ϩ Ϫ Ϫ Ϫ loblastomas in both Ptch1 / ;Ink4c / (tumors 1–4) and p53FL/FL; moxifen inducible p53-ER fusion protein (p53-ERtam) were treated with or with- Ink4cϪ/Ϫ (tumors 1Ј–4Ј) mice as compared to GNPs isolated from out 17-DMAG in the presence or absence of tamoxifen and cell death assessed (B). 7-day-old (P7) mice or post-mitotic neurons in mature cerebella Cell death was also determined in primary MEFs at early passages (Ͻ6) with the following genotypes: Arfϩ/ϩ or ArfϪ/Ϫ (C); Atmϩ/ϩ or AtmϪ/Ϫ (D); Baxϩ/ϩ and from P30 mice (Fig. 2 A and B, respectively, and Fig. S1). qPCR ϩ/ϩ Ϫ/Ϫ Ϫ/Ϫ ϩ/ϩ Ϫ/Ϫ analysis of the same tumor samples showed that Hsp90AA1 gene Bak or Bax and Bak (E); and Puma or Puma (F). Data depict the annexin V-positive and propidium iodide-negative population and represent the expression was at least equal to, or greater than that observed in wt mean values of triplicate experiments. P7 GNPs (Fig. 2 C and D). Interestingly, Hsp90AA1 RNA and protein levels decreased as proliferating GNPs (P7) exited the cell cycle and differentiated into post-mitotic granular neurons (P30), to engage 17-DMAG-induced cell death in primary GNP-like an expression pattern that is observed with other genes implicated medulloblastoma cells. in medulloblastoma genesis (23). We also observed that 17-DMAG induced a rapid accumulation of p53 and p21Cip1 protein in GNP-like tumor cells isolated from 17-DMAG Treatment of Primary Medulloblastoma Cells In Vitro In- Ptch1ϩ/Ϫ;Ink4cϪ/Ϫ mice (Fig. 4 A and C). As expected, no p53 was duces Caspase-Dependent Cell Death but Only in the Presence of detected in tumor cells isolated from p53FL/FL;Ink4cϪ/Ϫ mice in Functional p53. Inhibition of Hsp90 can engage cell death in a response to either 17-DMAG or UV treatment (Fig. 4B). In variety of tumor cell lines (28). We observed an accumulation of addition, Cip1, Puma and Mdm2 gene expression was induced in a cells in the subG1 phase of the cell cycle in 17-DMAG treated dose- and time-dependent manner in tumor cells isolated from GNP-like tumor cells from Ptch1ϩ/Ϫ;Ink4cϪ/Ϫ mice but not in Ptch1ϩ/Ϫ;Ink4cϪ/Ϫ but not p53FL/FL;Ink4cϪ/Ϫ mice (Fig. 4 D–F). similarly treated tumor cells lacking p53 (p53FL/FL;Ink4cϪ/Ϫ) that These results suggest that the inhibition of Hsp90 engages a p53 was inhibited by Q-VD-OPH, a pan caspase inhibitor (Fig. 3 A response in tumor cells that likely accounted for the death observed and B and Fig. S2). Furthermore, reduction of p53 activity by in vitro (Fig. 3 A and B). transduction of Ptch1ϩ/Ϫ;Ink4cϪ/Ϫ GNP-like tumor cells with We also evaluated 17-DMAG-induced cell death in two pairs of Mdm2 or a dominant-negative form of p53 (DN-p53 Q22, S23, human isogenic cell lines U2OS and SAOS-2, osteosarcomas wt or and G281) significantly reduced the sensitivity of tumor cells to null for TP53, respectively, and DAOY and D283 MED, medul- 17-DMAG as compared to those expressing GFP alone (Fig. loblastomas mutant or wt for TP53, respectively. We examined their 3C). Collectively these data indicate that p53 activity is necessary sensitivity to 17-DMAG and ␥ irradiation by annexin V binding and

17038 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0902880106 Ayrault et al. Downloaded by guest on October 1, 2021 Fig. 2. Hsp90AA1 expression is elevated in primary GNP-like tumor cells from medulloblastomas in Ptch1ϩ/Ϫ;Ink4cϪ/Ϫ and p53FL/FL;Ink4cϪ/Ϫ mice. Inducible Hsp90AA1 expression was examined by immunoblotting (A and B) or qPCR (C and D) in tumor cells isolated from four independent medulloblastomas arising in Ptch1ϩ/Ϫ;Ink4cϪ/Ϫ [A and C (labeled 1–4)] or p53FL/FL;Ink4cϪ/Ϫ [B and D (labeled 1Ј–4Ј)] mice. Analysis of GNPs or whole cerebellum of corresponding genetic backgrounds at 7 or 30 days post-partum represent normally proliferating GNPs (P7) and differentiated granular neurons (P30), respectively. qPCR data were Fig. 3. Inhibition of Hsp90 induces caspase-dependent cell death in p53- normalized to the internal standards 18S (shown) and L32 (data not shown). competent medulloblastomas. GNP-like tumor cells isolated from tumors arising in Ptch1ϩ/Ϫ;Ink4cϪ/Ϫ or p53FL/FL;Ink4cϪ/Ϫ mice were treated in triplicate with ␮ ability to induce TP53 and p21CIP1 (Figs. S3 and S4, respectively). 17-DMAG (0–1 M) or DMSO for 24 h with or without addition of Q-VD-OPH (qvD) (40 ␮M). The percentage of subG1 population was determined by pro- Overall, the results demonstrated that a failure to induce p53 in pidium iodide staining and ﬂow cytometric analysis. Death induced in a repre- response to 17-DMAG correlated with insensitivity to 17-DMAG sentative tumor from each of the two genotypes induced by 17-DMAG (0–1 ␮M)

effects. or DMSO vehicle control after 24 h (A) and average percentage of the subG1 CELL BIOLOGY population induced by 1 ␮M 17-DMAG or DMSO vehicle control after 24 h (Ϯ 40 ϩ Ϫ Ϫ Ϫ 17-DMAG Administration Prevents Medulloblastoma Formation in ␮M Q-VD-OPH) in four independent tumors from Ptch1 / ;Ink4c / (black bars) FL/FL Ϫ/Ϫ Vivo. GNP-like tumor cells purified from medulloblastomas arising or p53 ;Ink4c (white bars) mice (B). Primary GNP-like tumor cells isolated ϩ/Ϫ Ϫ/Ϫ in Ptch1ϩ/Ϫ;Ink4cϪ/Ϫ mice were implanted bilaterally into the flanks from medulloblastoma arising in Ptch1 ;Ink4c mice were plated and infected the same day with murine stem cell virus–based retroviruses (MSCV-IRES- of 12 immunocompromised CD-1 nude mice. Twenty-four-hours GFP) expressing GFP alone or co-expressing Mdm2 or DN-p53 (Q22, S23, and post tumor implantation, mice were injected once daily for 3 G281). Thirty-six hours post-infection tumor cells were treated with 17-DMAG (1 consecutive days per week with 17-DMAG (15 mg/kg) or PBS ␮M) (white bars) or PBS (black bars). Annexin V staining (Annexin-APC) and ﬂow control before a 4-day recovery period. We observed an almost cytometric analysis was performed 18 h post-treatment. Data are expressed as complete absence of tumor growth in those mice treated with percent annexin V-positive cells within the GFP-positive population (C). 17-DMAG as compared to those receiving PBS vehicle (Fig. 5A). Similar observations were made using two additional independently ϩ/Ϫ Ϫ/Ϫ retardation of tumor growth in 17-DMAG treated mice trans- arising tumors from Ptch1 ;Ink4c mice (data not shown). ϩ/Ϫ Ϫ/Ϫ Importantly, no overt changes in either body weight (Fig. S5 A and planted with Ptch1 ;Ink4c tumor cells as compared to the B) or standard clinical chemical diagnostic parameters (data not PBS control group (Fig. 5B). Representative 3-D images from the PBS and 17-DMAG treatment groups are shown (Fig. S7 A shown) were detected in 17-DMAG treated mice. Of the 12 mice and B). In stark contrast, growth of p53FL/FL;Ink4cϪ/Ϫ tumors evaluated in the experiment described here, seven had grossly was almost identical in both 17-DMAG and PBS treatment visible tumors and of these, only one had received 17-DMAG groups (Fig. 5C). (representative image shown in Fig. S5C). Tumors harvested from To address whether p53 was induced in situ, a subset of tumors vehicle-treated mice exhibited characteristic tumor morphology in mice receiving 17-DMAG (15 mg/kg) or vehicle were processed (Fig. S6A) and in cases where masses were evident in the 17- for p53 immunostaining. We found a significant and widespread DMAG treatment group, tumors were smaller and comprised accumulation of p53 protein in those tumors harvested from mice largely of proteinaceous material infiltrated by a few inflammatory receiving 17-DMAG as compared to vehicle control (Fig. S8 A–C). cells and only clusters of tumor cells (Fig. S6B). Collectively these These data indicate that inhibition of Hsp90 by 17-DMAG can data indicate that 17-DMAG prevents medulloblastoma engraft- induce a p53 response in Ptch1ϩ/Ϫ;Ink4cϪ/Ϫ tumors that likely ment and growth in vivo. mediates its ability to prevent medulloblastoma tumor growth We also evaluated the impact of 17-DMAG on the growth of in vivo. established tumors. Following injection of primary GNP-like Collectively these data show that 17-DMAG significantly retards ϩ Ϫ Ϫ Ϫ tumor cells from medulloblastomas arising in Ptch1 / ;Ink4c / tumor engraftment and growth in vivo that requires functional p53 or p53FL/FL;Ink4cϪ/Ϫ mice into the flanks of CD1 nude mice, we and furthermore, support the idea that elevated levels of Hsp90 in monitored tumor growth via ultrasound imaging until tumors the context of wt p53, both characteristics of pediatric tumors, may reached approximately 100 mm3 in volume, after which 17- provide a therapeutic opportunity for the treatment of human DMAG or PBS was administered. We observed a significant medulloblastoma.

Ayrault et al. PNAS ͉ October 6, 2009 ͉ vol. 106 ͉ no. 40 ͉ 17039 Downloaded by guest on October 1, 2021 Fig. 4. 17-DMAG induces a p53 response. GNP-like tumor cells isolated from medulloblastomas arising in Ptch1ϩ/Ϫ;Ink4cϪ/Ϫ or p53FL/FL;Ink4cϪ/Ϫ mice were treated with 17-DMAG (0–10 ␮M) or DMSO for 0–24 h before evaluation of p53 (A and B) and p21Cip1 (C) protein expression (two independent tumors). Cells exposed to UV irradiation or recombinant p53 protein served as positive controls. Actin was used as a loading control. Data shown are representative of three independent tumors. Analysis via qPCR for Cip1 (D), Puma (E), and Mdm2 (F) gene expression following treatment with 17-DMAG in tumor cells isolated from Ptch1ϩ/Ϫ;Ink4cϪ/Ϫ or p53FL/FL;Ink4cϪ/Ϫ mice. Data shown were normalized to 18S and L32 (data not shown) and are representative of at least three independent tumors.

Discussion D283 MED cell line, despite its wt TP53 status, was refractive to the We show that primary MEFs and GNP-like tumor cells isolated pro-apoptotic effects of 17-DMAG but significantly, failed to from murine medulloblastomas require functional p53 for the accumulate TP53 protein despite their sensitivity to irradiation induction of apoptosis following inhibition of Hsp90 with 17- coincident with TP53 accumulation. This further suggests that DMAG. This correlated with an accumulation of p53 protein HSP90 inhibitors engage p53 via a specific mechanism, the dys- independently of either p19Arf or Atm and was coincident with regulation of which can render cells resistant to 17-DMAG. transcription of Mdm2 and Cip1. In addition we demonstrate that Somewhat unexpectedly, AtmϪ/Ϫ MEFs were more sensitive to Puma and/or Bax and Bak act as the downstream effectors of p53 17-DMAG as compared to wt MEFs. AtmϪ/Ϫ cells are notoriously to mediate 17-DMAG-induced apoptosis. Strikingly, we observed fragile and undergo premature entry into replicative crisis unless a profound sensitivity of p53-competent murine medulloblastomas either Arf or p53 are also disabled (30). In addition, Atm deficiency to 17-DMAG treatment in vivo that was reflected by an accumu- is associated with elevated levels of ROS (31) that may render cells lation of p53. We propose that 17-DMAG can engage p53- more sensitive to any added imposed stress including (but not dependent cell death via a novel mechanism that likely mediates its necessarily restricted to) 17-DMAG treatment. Whether the in- anti-tumorigenic activity in vivo in murine medulloblastoma. creased 17-DMAG-induced cell death in AtmϪ/Ϫ cells may be Hsp90AA1 was significantly elevated at RNA and protein levels ϩ/Ϫ Ϫ/Ϫ attributed to an enhanced accumulation of p53 and the mechanism in tumor cells isolated from medulloblastoma in Ptch1 ;Ink4c through which this could occur remain unclear. However, the and p53FL/FL;Ink4cϪ/Ϫ mice. These observations support similar requirement of Atm for Stat3 activation (32) and the ability of Stat3 findings in human medulloblastomas (16) and indicate that HSP90 to repress the p53 promoter (33) suggests that limited Stat3 activity inhibitors could represent an as yet, unexplored treatment option in the absence of Atm could enhance p53 accumulation. Because for this pediatric tumor type. HSP90 inhibitors reduce the survival of tumor cell lines derived Hsp90 can promote the activation of Stat3 (34), predictably, from a variety of pediatric solid tumors including medulloblastomas 17-DMAG could enhance de-repression of p53 accumulation in a (28). We too, assessed the sensitivity of a panel of human cell lines Stat3-dependent manner. Therefore, it will be interesting to deter- to treatment with 17-DMAG and showed that the inability of mine if Stat3 function contributes to a p53-dependent enhancement 17-DMAG to induce p53 correlated with a failure to engage cell of 17-DMAG-induced cell death in Atm deficient cells. Cells death. However, DAOY cells, harboring homozygous C252F p53 lacking Atm also display a reduced ability to induce AMP kinase in mutation were sensitive to 17-DMAG, coincident with a decrease response to several stimuli (35) and the inhibition of AMPK in in mutant TP53 protein levels, likely due to the disruption of its combination with cisplatin-induced DNA damage leads to hyper- interaction with HSP90 (7). How this leads to cell death is unclear induction of p53 (36). It is unclear if this scenario could account for although the destabilization of mutant TP53 may relieve cells of a the increased sensitivity to 17-DMAG in AtmϪ/Ϫ cells but it will be dominant active function of mutant TP53 (29). In addition, the interesting to investigate this possibility in the future.

17040 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0902880106 Ayrault et al. Downloaded by guest on October 1, 2021 In addition, we have not formally eliminated the possibility that 17-DMAG induces DNA damage. If this were the case, the damaged AtmϪ/Ϫ cells would fail to engage G1 arrest and progress through S-phase, leading instead to Atr-dependent stalled replica- tion forks and an increase in p53-dependent apoptosis rather than DNA repair. Regardless of the underlying mechanism through which the loss of Atm renders cells more sensitive to 17-DMAG, importantly for our studies, 17-DMAG-induced cell death clearly proceeds in the absence of Atm. 17-DMAG induced an accumulation of p53 protein but not RNA, implicating regulation of p53 turnover. This is supported by similar observations using chronic lymphocytic leukemia (CLL) tumor cells treated with GA (37) but contrasts to observations that HSP90 inhibition can lead to the degradation of mutant TP53 (7). Hsp90 client proteins that influence p53 stability include Mdm2, the E3 ligase that directly ubiquitylates and promotes the degradation of p53, Chk1, a downstream kinase of Atm that phosphorylates p53 to disrupt its interaction with Mdm2 and Akt that phosphorylates Mdm2 to enhance p53 accumulation. The mechanism through which disruption of Hsp90 leads to p53 accumulation in our model is unclear but preliminary studies show that 17-DMAG induces a rapid loss of Mdm2 protein in GNP-like tumor cells isolated from medulloblastoma arising in Ptch1ϩ/Ϫ;Ink4cϪ/Ϫ mice (despite an induction of Mdm2 RNA) (Godeny and Beere, unpublished observations). Acute loss of Mdm2 protein is compatible with a model in which 17-DMAG disrupts a tertiary complex comprised of Hsp90, Mdm2 and p53 leading to an accumulation of p53 protein. Alternatively, disruption of Akt/Hsp90 interactions would lead to the destabilization of Akt protein and prevent it from phosphory- lating and maintaining Mdm2 levels, leading to the accumulation of p53. Our ongoing studies will define the mechanism through which the disruption of Hsp90 engages a p53 response. The absence of p53 in medulloblastoma cells from p53FL/FL; Ink4cϪ/Ϫ mice or its inactivation via Mdm2 or DN-p53 expression in GNP-like tumor cells from Ptch1ϩ/Ϫ;Ink4cϪ/Ϫ mice significantly repressed the pro-apoptotic activity of 17-DMAG in vitro. CELL BIOLOGY Tumor cells isolated from medulloblastomas in Ptch1ϩ/Ϫ; Ink4cϪ/Ϫ mice and implanted into nude recipients, failed to grow when mice were treated with 17-DMAG. Furthermore, 17- DMAG treatment of mice harboring established tumors from Ptch1ϩ/Ϫ;Ink4cϪ/Ϫ mice retarded tumor expansion as compared to the control group. In contrast, GNP-like tumor cells lacking p53 function (p53FL/FL;Ink4cϪ/Ϫ) displayed identical growth characteristics in vivo in both vehicle and 17-DMAG treatment groups. These findings substantiated our in vitro observations that p53 mediates the pro-apoptotic effects of 17-DMAG and suggest that an intact p53 response may be a predictor of clinical outcome. Preclinical testing of alvespimycin, a water-soluble analog of 17-DMAG, revealed no significant impact on medulloblastoma tumor growth in vivo (38). However, the cell line tested (BT-45) harborsaCtoTtransition at position 993 (http://pptp.stjude.org/ documents.php) that generates a mutant TP53 protein (R248W) that is impaired in both its DNA binding ability and its ubiquiti- Fig. 5. Inhibition of Hsp90 by 17-DMAG prevents medulloblastoma tumor nation rendering it susceptible to 17-DMAG-induced degradation growth in vivo. Primary tumor cells (2 ϫ 106) isolated from Ptch1ϩ/Ϫ;Ink4cϪ/Ϫ (39). It remains unclear whether these studies reconcile the failure (A and B)orp53FL/FL;Ink4cϪ/Ϫ (C) mice were suspended in Matrigel and s.c. of a medulloblastoma harboring mutant TP53 (R248W) to respond injected bilaterally into the ﬂanks of 12 CD1 nude mice. Twenty-four hours to 17-DMAG in vivo (38) with our proposed model through which post-implantation mice were treated weekly with either PBS vehicle or 15 the anti-tumorigenic effect of 17-DMAG is mediated by an intact mg/kg 17-DMAG for 3 consecutive days per week followed by a 4-day recovery wt TP53 response. period. Perpendicular tumor diameters were measured using calipers and The administration of 17-DMAG both retards tumor growth and ␲ ϫ 3 average tumor volume calculated from /3 D (A). Data shown are repre- engages a p53 response in vivo and is consistent with the ability of sentative of replicate experiments conducted using three independent 17-DMAG to induce apoptotic cell death in vitro but only in Ptch1ϩ/Ϫ;Ink4cϪ/Ϫ medulloblastomas. Allograft tumors (Ptch1ϩ/Ϫ;Ink4cϪ/Ϫ)(B) or (p53FL/FL;Ink4cϪ/Ϫ)(C) were imaged via ultrasound every 3–4 days until medulloblastoma cells retaining functional p53. Furthermore we tumor volume reached approximately 100 mm3 (day 14) after which 17-DMAG have revealed a pathway through which the p53 response can be (15 mg/kg) or PBS control injections were initiated. Tumor volumes in both directly activated independently of the upstream mediators of p53 groups were monitored by ultrasound and volumetric tumor measurements activation, p19Arf and Atm. This may be of significance in those derived from 3-D reconstructions of the ultrasound datasets. tumor types that harbor defects or mutations in these key activators

Ayrault et al. PNAS ͉ October 6, 2009 ͉ vol. 106 ͉ no. 40 ͉ 17041 Downloaded by guest on October 1, 2021 of a p53 response since HSP90 inhibitors may not, under these RNA Isolation and qPCR. Total RNA was isolated using TRIzol (Invitrogen) and circumstances, represent a viable therapeutic approach. after reverse transcription, real-time PCR was performed using a 7900HT Fast Real time PCR system (Applied Biosystems). Details of primers used are described in SI Materials and Methods Text. Our methods are essentially as described previously (20, 23, 40) with additional details in SI Text. Whole Cell Lysate Preparation and Immunoblotting. Cells (primary GNP-like tumor cells, MEFs or cell lines) were lysed in appropriate lysis buffers (see SI Text) and approximately 30 ␮g protein separated via standard SDS/PAGE (Criterion, Mouse Husbandry. p53FL/FL;Nestin-Creϩ;Ink4cϪ/Ϫ (referred to throughout as Bio-Rad). After transfer to Hybond-C membrane (Amersham) immunoblotting p53FL/FL;Ink4cϪ/Ϫ) and Ptch1ϩ/Ϫ;Ink4cϪ/Ϫ mice were derived and maintained as was performed and specific signals visualized using appropriate HRP-conjugated described previously (20, 23, 40). secondary antibodies and Supersignal chemoluminescence technology (Pierce Biotechnology). SI Text provides details of primary antibodies used. Purification of Primary Cerebellar Granule Neuron Progenitors. Purification of granule neuron progenitors (GNPs) and GNP-like tumor cells from mouse cere- Histopathology and Immunohistochemistry. Recovered tumors were fixed in bella or medulloblastoma, respectively (41), with minor modifications (23). 10% neutral buffered formalin and embedded in paraffin wax, 4-␮m sections prepared, stained with H&E, and examined microscopically. Medulloblastoma Tumor Implantation. GNP-like tumor cells (2–3 ϫ 106) from medulloblastomas arising in Ptch1ϩ/Ϫ;Ink4cϪ/Ϫ or p53FL/FL;Ink4cϪ/Ϫ mice were Viral Transduction of Primary Tumor Cells. GNP-like tumor cells isolated from resuspended in Matrigel and implanted s.c. into each of the flanks of 12 CD-1 medulloblastomas arising in Ptch1ϩ/Ϫ;Ink4cϪ/Ϫ mice were plated and infected the nude mice before administration of vehicle or 17-DMAG and monitoring of same day with murine stem cell virus–based retroviruses (MSCV-IRES-GFP) either tumor volume as described in figure legends and SI Text. as an empty vector or co-expressing Mdm2 or a dominant-negative form of p53 (DN-p53) (Q22, S23, and G281). Preparation and Maintenance of Mouse Embryo Fibroblasts. Timed breedings between appropriate genotypes were used to generate embryos positive for tam ACKNOWLEDGMENTS. We acknowledge Drs Stephen Tait, Jerry Chipuk, and p53-ER (25) and wt or null for each of p53, Arf, Atm, Puma,orBax/Bak. Kirsteen MacLean for providing valuable reagents. Thanks go to Robert Jenson, Suqing Xie, and Shelly Wilkerson for genotyping and to Jennifer Craig for Maintenance of Cell Lines. U2OS, SAOS-2 and DAOY cell lines were maintained technical assistance. Also special thanks to Kelli Boyd (K.L.B.) for histological in DMEM, and D283 MED in EMEM media supplemented with 10% FBS, 4 mM analysis, to the animal husbandry staff of the Animal Resource Center and to glutamine and 100 U each of penicillin and streptomycin (GIBCO). Christopher Calabrese and Melissa Johnson in the Imaging Center. We also thank Douglas R. Green and Charles J. Sherr for valuable input during the preparation of this manuscript. This work was funded in part by National Cancer Institute ϫ 6 Cell Death and Cell Cycle Analyses. GNPs (0.5 10 /well) were treated with grant CA-096832–10 (to M.F.R.), core funding and Cancer Center Supporting 17-DMAG or DMSO before analyses using annexin V and PI staining and flow Grant developmental funds (5P30CA021765–29) (to M.F.R. and H.M.B.), the cytometric analysis (BD FACScan). All data were analyzed using FlowJo software Gephardt Endowed Fellowship of Signal Transduction (to O.A.), and the American (Tree Star, Inc.). Lebanese-Syrian Associated Charities of St. Jude Children’s Research Hospital.

1. Vousden KH, Lu X (2002) Live or let die: The cell’s response to p53. Nat Rev 2:594–604. 24. Knoepfler PS, Cheng PF, Eisenman RN (2002) N-myc is essential during neurogenesis for the 2. Parsell DA, Lindquist S (1993) The function of heat-shock proteins in stress tolerance: rapid expansion of progenitor cell populations and the inhibition of neuronal differen- Degradation and reactivation of damaged proteins. Annu Rev Genet 27:437–496. tiation. Genes Dev 16:2699–2712. 3. Beere HM (2005) Death versus survival: Functional interaction between the apoptotic and 25. Christophorou MA, et al. (2005) Temporal dissection of p53 function in vitro and in vivo. stress-inducible heat shock protein pathways. J Clin Invest 115:2633–2639. Nat Genet 37:718–726. 4. Whitesell L, Lindquist SL (2005) HSP90 and the chaperoning of cancer. Nat Rev 5:761–772. 26. Chipuk JE, et al. (2004) Direct activation of Bax by p53 mediates mitochondrial membrane 5. Sharp S, Workman P (2006) Inhibitors of the HSP90 molecular chaperone: Current status. permeabilization and apoptosis. Science 303:1010–1014. Adv Cancer Res 95:323–348. 27. Nakano K, Vousden KH (2001) PUMA, a novel proapoptotic gene, is induced by p53. Mol 6. Muller P, Hrstka R, Coomber D, Lane DP, Vojtesek B (2008) Chaperone-dependent stabi- Cell 7:683–694. lization and degradation of p53 mutants. Oncogene 27:3371–3383. 28. Calabrese C, Frank A, Maclean K, Gilbertson R (2003) Medulloblastoma sensitivity to 7. Peng Y, Chen L, Li C, Lu W, Chen J (2001) Inhibition of MDM2 by hsp90 contributes to mutant p53 stabilization. J Biol Chem 276:40583–40590. 17-allylamino-17-demethoxygeldanamycin requires MEK/ERKM. J Biol Chem 278:24951– 8. Walerych D, et al. (2004) Hsp90 chaperones wild-type p53 tumor suppressor protein. J Biol 24959. Chem 279:48836–48845. 29. Koonin EV, Rogozin IB, Glazko GV (2005) p53 gain-of-function: Tumor biology and 9. Grbovic OM, et al. (2006) V600E B-Raf requires the Hsp90 chaperone for stability and is bioinformatics come together. Cell Cycle 4:686–688. degraded in response to Hsp90 inhibitors. Proc Natl Acad Sci USA 103:57–62. 30. Kamijo T, et al. (1999) Loss of the ARF tumor suppressor reverses premature replicative 10. Sato S, Fujita N, Tsuruo T (2000) Modulation of Akt kinase activity by binding to Hsp90. Proc arrest but not radiation hypersensitivity arising from disabled atm function. Cancer Res Natl Acad Sci USA 97:10832–10837. 59:2464–2469. 11. Kamal A, et al. (2003) A high-affinity conformation of Hsp90 confers tumour selectivity on 31. Ito K, et al. (2004) Regulation of oxidative stress by ATM is required for self-renewal of Hsp90 inhibitors. Nature 425:407–410. haematopoietic stem cells. Nature 431:997–1002. 12. Banerji U, et al. (2005) Phase I pharmacokinetic and pharmacodynamic study of 17- 32. Zhang Y, et al. (2003) Ataxia telangiectasia mutated proteins, MAPKs, and RSK2 are allylamino, 17-demethoxygeldanamycin in patients with advanced malignancies. J Clin involved in the phosphorylation of STAT3. J Biol Chem 278:12650–12659. Oncol 23:4152–4161. 33. Niu G, et al. (2005) Role of Stat3 in regulating p53 expression and function. Mol Cell Biol 13. Solit DB, et al. (2007) Phase I trial of 17-allylamino-17-demethoxygeldanamycin in patients 25:7432–7440. with advanced cancer. Clin Cancer Res 13:1775–1782. 34. Sato N, et al. (2003) Involvement of heat-shock protein 90 in the interleukin-6- 14. Roe SM, et al. (1999) Structural basis for inhibition of the Hsp90 molecular chaperone by mediated signaling pathway through STAT3. Biochem Biophys Res Commun 300:847– the antitumor antibiotics radicicol and geldanamycin. J Med Chem 42:260–266. 852. 15. Blagg BS, Kerr TD (2006) Hsp90 inhibitors: Small molecules that transform the Hsp90 protein folding machinery into a catalyst for protein degradation. Med Res Rev26:310– 35. Sun Y, Connors KE, Yang DQ (2007) AICAR induces phosphorylation of AMPK in an 338. ATM-dependent, LKB1-independent manner. Mol Cell Biochem 306:239–245. 16. Hauser P, et al. (2006) Expression and prognostic examination of heat shock proteins (HSP 36. Fu X, Wan S, Lyu YL, Liu LF, Qi H (2008) Etoposide induces ATM-dependent mitochondrial 27, HSP 70, and HSP 90) in medulloblastoma. J Pediatr Hematol Oncol 28:461–466. biogenesis through AMPK activation. PLoS One 3:e2009. 17. Fogarty MP, Emmenegger BA, Grasfeder LL, Oliver TG, Wechsler-Reya RJ (2007) Fibroblast 37. Lin K, Rockliffe N, Johnson GG, Sherrington PD, Pettitt AR (2008) Hsp90 inhibition has growth factor blocks Sonic hedgehog signaling in neuronal precursors and tumor cells. opposing effects on wild-type and mutant p53 and induces p21 expression and cytotoxicity Proc Natl Acad Sci USA 104:2973–2978. irrespective of p53/ATM status in chronic lymphocytic leukaemia cells. Oncogene 27:2445– 18. Lee Y, et al. (2003) A molecular fingerprint for medulloblastoma. Cancer Res 63:5428– 2455. 5437. 38. Smith MA, et al. (2008) Stage 1 testing and pharmacodynamic evaluation of the HSP90 19. Thompson MC, et al. (2006) Genomics identifies medulloblastoma subgroups that are inhibitor alvespimycin (17-DMAG, KOS-1022) by the pediatric preclinical testing program. enriched for specific genetic alterations. J Clin Oncol 24:1924–1931. Pediatr Blood Cancer 51:34–41. 20. Zindy F, et al. (2007) Genetic alterations in mouse medulloblastomas and generation of 39. Nagata Y, et al. (1999) The stabilization mechanism of mutant-type p53 by impaired tumors de novo from primary cerebellar granule neuron precursors. Cancer Res 67:2676– ubiquitination: The loss of wild-type p53 function and the hsp90 association. Oncogene 2684. 18:6037–6049. 21. Goodrich LV, Milenkovic L, Higgins KM, Scott MP (1997) Altered neural cell fates and 40. Zindy F, et al. (2003) Hemangiosarcomas, medulloblastomas, and other tumors in Ink4c/ medulloblastoma in mouse patched mutants. Science 277:1109–1113. p53-null mice. Cancer Res 63:5420–5427. 22. Franklin DS, et al. (1998) CDK inhibitors p18(INK4c) and p27(Kip1) mediate two separate 41. Hatten ME, Gao WQ, Morrison ME, Mason CA (1998) The cerebellum: Purification and pathways to collaboratively suppress pituitary tumorigenesis. Genes Dev 12:2899–2911. 23. Uziel T, et al. (2005) The tumor suppressors Ink4c and p53 collaborate independently with coculture of identified cell populations. In Culturing Nerve Cells, eds Banker G, Goslin K, Patched to suppress medulloblastoma formation. Genes Dev 19:2656–2667. (Cambridge, MA, The MIT Press), pp 419–459.

17042 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0902880106 Ayrault et al. Downloaded by guest on October 1, 2021