A C-terminal HSP90 inhibitor restores glucocorticoid

letters

A C-terminal HSP90 inhibitor restores glucocorticoid sensitivity and relieves a mouse allograft model of Cushing disease © 2015 Nature America, Inc. All rights reserved.

Mathias Riebold1, Christian Kozany2, Lee Freiburger3,4, Michael Sattler3,4, Michael Buchfelder5, Felix Hausch2, Günter K Stalla1 & Marcelo Paez-Pereda1 One function of the glucocorticoid receptor (GR) in corticotroph cells is to suppress the transcription of the gene encoding proopiomelanocortin (POMC), the precursor of the stress hormone adrenocorticotropin (ACTH)1. Cushing disease is a neuroendocrine condition caused by partially glucocorticoidresistant corticotroph adenomas that excessively secrete ACTH, which leads to hypercortisolism2–4. Mutations that impair GR function explain glucocorticoid resistance only in sporadic cases5,6. However, the proper folding of GR depends on direct interactions with the chaperone heat shock protein 90 (HSP90, refs. 7,8). We show here that corticotroph adenomas overexpress HSP90 compared to the normal pituitary. N- and C-terminal HSP90 inhibitors act at different steps of the HSP90 catalytic cycle to regulate corticotroph cell proliferation and GR transcriptional activity. C-terminal inhibitors cause the release of mature GR from HSP90, which promotes its exit from the chaperone cycle and potentiates its transcriptional activity in a corticotroph cell line and in primary cultures of human corticotroph adenomas. In an allograft mouse model, the C-terminal HSP90 inhibitor silibinin showed anti-tumorigenic effects, partially reverted hormonal alterations, and alleviated symptoms of Cushing disease. These results suggest that the pathogenesis of Cushing disease caused by overexpression of heat shock proteins and consequently misregulated GR sensitivity may be overcome pharmacologically with an appropriate HSP90 inhibitor. HSP90 is essential for the proper folding of the ligand-binding domain of GR7,8. After the maturation of GR and its final folding, however, continued binding of HSP90 to GR inhibits the latter’s transcriptional activity by reducing its DNA binding and disrupting its formation of transcriptional complexes9,10, which evokes the partial glucocorticoid resistance of corticotroph adenomas. We therefore investigated the expression of the α and β isoforms of HSP90 as well as HSF1, the transcription factor that controls their expression. We found strong overexpression of these heat shock proteins in corticotroph adenomas

as compared with the normal human pituitary and non-functioning pituitary adenomas (Fig. 1a–d and Supplementary Fig. 1). Both HSP90 isoforms form complexes with GR (Supplementary Fig. 2). To test whether the overexpression of HSP90 has a role in the pathogenesis of corticotroph adenomas, we inhibited the chaper one in the mouse corticotroph cell line AtT-20 using distinct classes of compounds: the N-terminal HSP90 inhibitor 17-AAG, the C-terminal HSP90 inhibitor novobiocin11, and the recently identified HSP90 inhibitor silibinin12. These compounds reduced AtT-20 cell proliferation and altered the distribution of these cells among the phases of the cell cycle in a dose-dependent manner (Fig. 1e,f and Supplementary Fig. 3). The pharmacologic inhibition of HSP90 leads to the degradation of its client proteins13. Therefore, we investigated the stability of two HSP90 clients: cell division cycle protein 2 (Cdc2, ref. 14) and GR, which control, respectively, cell proliferation and ACTH production through independent mechanisms in corticotroph cells. All three inhibitors induced the degradation of Cdc2 in AtT-20 cells but only 17-AAG also triggered the degradation of GR, an effect similar to that previously reported for geldanamycin (refs. 15,16 and Fig. 1g). Of note, silibinin and novobiocin did not affect GR protein levels at concentrations that induced degradation of Cdc2 in the same experiment (Fig. 1g), suggesting that they may act through a novel biochemical mechanism. Novobiocin binds to a pocket in the C-terminal domain (CTD) of HSP90 (refs. 11,17). As silibinin shows pharmacologic similarity to novobiocin, it might target the same domain12. To test this hypothesis and clarify the mechanism of action of C-terminal HSP90 inhibitors, we refined a binding assay in which silibinin was immobilized and recombinant HSP90-α CTD18 was allowed to bind. Bound HSP90-α CTD was then displaced by excess silibinin or novobiocin in solution (Fig. 2a). Both inhibitors induced chemical shift perturbations for a number of amide signals, as detected by nuclear magnetic resonance (NMR) spectroscopy (Fig. 2b,c). A few NMR signals were perturbed by one or the other compound, suggesting that both compounds directly bind HSP90-α CTD in a similar region (Fig. 2b,c).

1Clinical

Neuroendocrinology, Max Planck Institute of Psychiatry, Munich, Germany. 2Department of Translational Research in Psychiatry, Max Planck Institute of Psychiatry, Munich, Germany. 3Center for Integrated Protein Science, Technical University Munich, Munich, Germany. 4Institute of Structural Biology, Helmholtz Zentrum, Munich, Germany. 5Department of Neurosurgery, University of Erlangen–Nürnberg, Erlangen, Germany. Correspondence should be addressed to M.P.-P. ([email protected]). Received 12 August 2014; accepted 7 November 2014; published online 9 February 2015; doi:10.1038/nm.3776

nature medicine advance online publication

letters CPA

NFPA

MW (kDa) 100 100

Hsp90-β

f 17-AAG Silibinin Novobiocin

80 40 0 0.01

0.1

50

70

(Cells × 1,000) ml–1

120

75

60

**

50

**

*

0

NP

17-AAG (µM)

Control 10 µM silibinin 100 nM 17-AAG 100 µM novobiocin

30 20

1.0 10 100 1,000 Compound (µM)

4

8

**

4

0

CPA NFPA

**

NP

CPA NFPA

g

**

*

40

d

**

8

Silibinin (µM)

Novobiocin (µM)

Cdc2 GR Hsp90-α

0

1 2 3 Time (d)

β-actin

4

MW (kDa) 37 25 100 75 100 75 50 37

Figure 1 Human corticotroph adenomas overexpress heat shock proteins; HSP90 inhibitors reduce AtT-20 cell proliferation. (a) Immunohistochemistry of HSP90-α in a cryosection of normal human pituitary (NP; representative of 6 samples), a corticotroph pituitary adenoma (CPA) from a subject with Cushing disease (representative of 12 positive samples out of 14 samples tested), and a non-functioning pituitary adenoma (NFPA; representative of 3 positive samples out of 15 samples tested). Arrows and insets denote cells with typical expression of HSP90-α for each tissue. Non-functioning pituitary adenomas sporadically show clusters of HSP90-α–positive cells. The corticotroph adenoma is flanked at the right by normal pituitary tissue. Scale bars, 40 µm; 4 µm in insets. (b) Immunoblot of HSP90-α, HSP90-β, HSF1, and β-actin (as loading control) in different pituitary tissues. (c,d) Quantification of the expression of HSP90-α (c) and HSP90-β (d) relative to β-actin in the same tissues shown in b. (e,f) Proliferation of AtT-20 cells after treatment with 17-AAG, novobiocin, or silibinin as measured by (e) WST-1 assay (dose response) and (f) cell count (time course). Data in e,f are of one experiment representative of three. Student’s t test; *P ≤ 0.05, **P ≤ 0.01, compound versus control; error bars, mean ± sd. (g) Immunoblot of Cdc2, GR, Hsp90-α, and β-actin in AtT-20 cells after treatment with HSP90 inhibitors. MW, molecular weight.

After binding, novobiocin forces HSP90 into a conformation that releases client proteins 11,19,20. Therefore, using coimmunoprecipitation, we investigated whether silibinin and novobiocin dissociate the GR–Hsp90-α complex in AtT-20 cell lysates. These compounds disrupted this complex in a concentrationdependent manner when either GR or Hsp90-α was immunoprecipitated (Fig. 2d). GR requires dynamic interactions with HSP90 to maintain its highaffinity binding conformation21. In the initial steps of the chaperone cycle, HSP70 partially unfolds GR, opening the ligand binding site, and subsequently transfers it to the middle and C-terminal domains of HSP90, where the final maturation takes place22–24. Only the transfer

Immobilized silibinin (µM)

* **

**

N (ppm)

120 130

**

**

64 µM immobilized silibinin

110 120 130

10

9

8

7

10

1

9

8

7

1

H (ppm)

H (ppm)

e

d

Specific H-Dex binding (cpm)

0 4 8 16 32 64 128 256

0

**

c 110

15

0 37 111 333 1,000

**

50

N (ppm)

17-AAG (µM)

0 222 667 2,000 6,000

100

Novobiocin (µM)

0 37 111 333 1,000

150

of GR requires ATP22. Disruption of the catalytic cycle at this early step by 17-AAG, which blocks ATP hydrolysis, or by Bag1, a protein that increases ADP release from HSP70, results in the release of GR from the chaperone complex in an immature, unfolded state that either aggregates or is degraded, which reduces steroid ligand binding15,16,22. To investigate whether silibinin treatment causes GR to lose steroid binding activity, we characterized its effects on 3H-dexamethasone binding. Unexpectedly, we observed that silibinin increased the number of GR sites (control, without silibinin: 36,240 ± 588 receptors per cell; silibinin: 58,740 ± 1,855 receptors per cell) that bind 3H-dexamethasone with high affinity (control Kd = 2.77 ± 0.37 nM; silibinin Kd = 2.49 ± 0.28 nM) (Fig. 2e). This binding is

b

Competition Silibinin (µM)

15

a Hsp90 CTD binding (%)

10,000 8,000

Silibinin Control

B/F

3

Input NI 0 5 150 500

Silibinin Novobiocin Figure 2 Silibinin promotes the release of mature GR in (µM) (µM) 6,000 AtT-20 cells through direct binding to the C-terminal domain of Hsp90. (a) ELISA of the percentage of HSP90 4,000 0.6 CTD bound to increasing concentrations of immobilized 0.4 IB: Hsp90 2,000 IP: 0.2 silibinin (left) or in the presence of competing silibinin, GR IB: GR 0 0 novobiocin, or 17-AAG (right). Means ± sd; student’s t test; 0 3,000 6,000 9,000 Bound (cpm) *P ≤ 0.05, **P ≤ 0.01, competing compound versus IB: GR IP: 0 5 10 15 20 25 30 35 control. (b,c) Superposition of 1H and 15N heteronuclear Hsp90-α IB: Hsp90 3 H-Dex (nM) single quantum coherence (HSQC) NMR spectra of 200 µM 15N-labeled human HSP90 CTD, showing (b) free HSP90 CTD (black), 500 µM silibinin (green); and (c) free HSP90 CTD (red), 500 µM (green) and 1 mM novobiocin (blue). (d) Co-immunoprecipitation of the GR–Hsp90 complex after incubation of AtT-20 cell lysates with silibinin or novobiocin. NI, non-immune IgG; IP, immunoprecipitation; IB, immunoblot. (e) Binding of 3H-dexamethasone (3H-Dex) to GR in AtT-20 cells after treatment with 30 µM silibinin (inset, Scatchard plot; B/F, bound/free). Error bars in a,e show means ± sd of one experiment representative of three. cpm, counts per minute. Input NI 0 5 150 500

© 2015 Nature America, Inc. All rights reserved.

Cell proliferation (Absorbance, % control)

e

HSF1 β-actin

c

Fold expression of Hsp90-β

NP

Hsp90-α

0 25 50 100 150

b

NFPA

0 10 20 40 60

CPA

Fold expression of Hsp90-α

NP

0 0.25 0.50 0.75 1.0

a

advance online publication nature medicine

letters

d

e

30

40

Dex (nM): 6

Figure 3 Silibinin enhances GR activity in AtT-20 cells. (a) Transcriptional activity of GR on the MMTV–Luc reporter * after treatment with 17-AAG, silibinin, or novobiocin alone or 4 20 in combination with dexamethasone (Dex). AU, arbitrary units; ** ** * Cpd, compound. (b) Transcriptional activity of wild-type (WT) ** and P548A/T549A/V551A–mutant (AAA) GR overexpressed in ** ** 2 10 AtT-20 cells under 30 µM silibinin treatment. (c) Response to Dex in AtT-20 cells transiently transfected with NurRE–Luc or Pomc-Luc and treated with silibinin as indicated. (d) Secretion 0 0 of ACTH by AtT-20 cells under treatment with silibinin alone Silibinin (µM): Silibinin (µM): or in combination with Dex. Error bars in a–d show means ± sd Dex (nM): Dex (nM): for one representative experiment with at least two replications; student’s t test; *P ≤ 0.05; **P ≤ 0.01, inhibitor plus Dex versus Dex alone. (e) Secretion of ACTH by primary cultures of rat normal anterior pituitary cells after the indicated treatment. Error bars show means ± sd for one representative experiment with at least two replications; student’s t test; *P ≤ 0.05, **P ≤ 0.01, Dex versus control).

specific and Hsp90 dependent, as it is displaced by unlabeled dexa methasone and the selective GR antagonist RU-38486 and eliminated by 17-AAG (Supplementary Fig. 4), which is in line with previous studies15,16,22. This shows that silibinin, in contrast to 17-AAG, increased the pool of fully mature GR that was able to bind ligand with high affinity, which constitutes a novel biochemical mechanism of action for C-terminal HSP90 inhibitors. To investigate how this mechanism influences GR transcriptional activity, we used MMTV–Luc, a reporter construct in which firefly luciferase is under the transcriptional control of the promoter of the mouse mammary tumor virus25. In transiently transfected AtT-20 cells, the combination of silibinin or novobiocin with low concentrations of dexamethasone potentiated GR transcriptional activity (Fig. 3a). This increase of GR transcriptional activity was completely blocked by RU-38486 (Supplementary Fig. 5a). Similarly, silibinin in combination with dexamethasone enhanced the activity of the specific GR reporter GRE2–TK–Luc, whereas it did not affect the activity of the minimal TK–Luc construct (Supplementary Fig. 5b). This shows that silibinin and novobiocin specifically increased GR transcriptional activity. As expected, 17-AAG, which induced GR protein degradation (Fig. 1e), also abolished GR transcriptional activity (Fig. 3a). Furthermore, the GR mutation P548A/T549A/V551A, which alters the interaction between GR and HSP90 (ref. 26), interfered with the effects of silibinin, indicating that these effects depend on Hsp90 (Fig. 3b). We next tested whether silibinin enhances the GR-mediated suppression of Pomc, a key endogenous target gene of GR in corticotroph cells. We investigated the trans-repression of a reporter driven by Nur77, a major regulator of Pomc transcription27, and of the complete Pomc promoter28. Whereas no effects on the basal activities of NurRE–Luc and POMC–Luc were observed, silibinin enhanced the transcriptional repression of both constructs produced by dexa methasone (Fig. 3c). In corticotroph cells, the POMC peptide is cleaved into ACTH. Consistent with the results obtained with POMC–Luc, silibinin


100 0 100 40 100 80

10 0 10 40 10 80

1.0 0 1.0 40 1.0 80

0 0 0 40 0 80

10 0 10 10 10 20 10 40

1.0 0 1.0 10 1.0 20 1.0 40

0 0 0 10 0 20 0 40

ACTH (ng ml–1)

ACTH (ng ml–1)


0 Silibinin (µM):

20

10 100 1,000 10,000 Dex (nM)

10

1.0

40

10

150 25 50 100 150

0.1

** **

0

0

**

10

1,000

0 10 10 10 10

40 5 10 20 40 0 10 10 10 10

Dex (nM):

0 0 0

0 Cpd (µM):

1.00 100 0.125 100 0.25 100 0.5 100 1.00

**

40

80

10

**

80

2,000

100

120

3,000

NurRE–Luc Pomc-Luc

120

0

**

** **

c

WT, silibinin WT, control AAA, silibinin AAA, control

4,000

40

**

b

0

160

**

Novobiocin

10

**

Luc (% control)

Silibinin

MMTV–Luc (AU)

17-AAG 200

0 10 100

MMTV–Luc (AU)

a

enhanced the effects of low concentrations of dexamethasone on endogenous ACTH secreted into the supernatant of AtT-20 cells (Fig. 3d). As expected, 17-AAG reduced the effect of dexamethasone on ACTH secretion (Supplementary Fig. 6). To examine whether silibinin may alter normal corticotroph cell function, we treated primary cultures of rat anterior pituitary cells. Silibinin did not influence the basal or dexamethasone-induced suppression of ACTH production in normal cells (Fig. 3e). In contrast, it enhanced the suppression of endogenous ACTH elicited by dexamethasone in five primary cultures of human corticotroph adenomas (out of six total samples) from subjects with Cushing disease (Fig. 4a), thus restoring glucocorticoid sensitivity in vitro, which confirmed our results with AtT-20 cells. We next studied the effects of silibinin on the growth, hormone secretion, and symptoms of corticotroph adenomas in an allograft model of Cushing disease, consisting of nude mice with AtT-20 cells injected subcutaneously25,29. Silibinin was orally administered daily at 300 mg kg−1 to male nude mice with AtT-20 cell tumors, which resulted in significantly lower tumor growth compared to the vehicle treatment (P ≤ 0.05; Fig. 4b). Mice under silibinin treatment had lower plasma ACTH and corticosterone concentrations than the vehicle-treated animals (Fig. 4c,d), which is consistent with our results in AtT-20 cells and primary cultures of human corticotroph adenoma cells. In addition, hypercortisolemic mice implanted with AtT-20 cells showed a distinctive, abnormal accumulation of fat that reflected the ‘moon face’ and ‘buffalo hump’ typical of Cushing disease in humans3,29. These symptoms were noticeably alleviated in silibinin-treated mice in comparison to the vehicle-treated animals (Fig. 4e). We propose that silibinin reverses a pathogenic mechanism in which the overexpression of HSP90 restrains the release of mature GR and leads to partial glucocorticoid resistance. This mechanism is different from—but might act in addition to—the dissociation of transcriptional complexes9,10 and other mechanisms that contribute to glucocorticoid resistance30,31. Thus, we also propose that the

letters

2

3

4

5

6

b

80

** 40

0 Silibinin: + – + Dex: – + +

*

*

Tumor volume (mm3)

ACTH (% control)

Silibinin

300

*

** + – + – + +

+ – + – + +

+ – + – + +

+ – + – + +

+ – + – + +

d

c

Vehicle

300

200

* 100

250 200 150

*

100

0 0

10 20 Time (d)

30

Corticosterone (mg ml–1)

1

120

ACTH (pg ml–1)

a Adenoma

5 4 3 2

0

Control Vehicle Silibinin


Figure 4 Effects of silibinin on primary cultures of human corticotroph adenomas and on a mouse allograft model of Cushing disease. (a) ACTH secreted by primary cultures from human corticotroph adenomas treated with silibinin (black bars), Dex (dark gray), or silibinin plus Dex (light gray). Error bars, mean ± sd of percent change over control for each adenoma; student’s t test; *P ≤ 0.05, **P ≤ 0.01, silibinin plus Dex versus Dex alone. Six adenomas were tested in total. (b) Tumor growth in Naval Medical Research Institute (NMRI) mice implanted with AtT-20 cell allografts under daily treatment with 300 mg kg −1 silibinin (n = 12) as compared to vehicle (n = 12). Error bars show means ± sem; treated/vehicle tumor volume = 0.55; student’s t test: *P ≤ 0.05. (c,d) Plasma concentrations of (c) ACTH and (d) corticosterone 24 h after the last treatment. Error bars in c,d show means ± sem; student’s t test: *P ≤ 0.05, silibinin versus vehicle; naive NMRI mice served as control. (e) Symptoms of representative animals in the vehicle and treatment groups (total of 12 animals per group). Arrows, fat pads; arrowheads, tumor sites.

effects on corticotroph adenoma cell proliferation and cell cycle that involve kinases such as Cdc2 are independent of the action of Hsp90 on GR. N-terminal HSP90 inhibitors block the entry of client proteins into the chaperone cycle by inhibiting ATP hydrolysis and thereby obstructing their transfer from HSP70 to HSP90 (ref. 22). This prevents the subsequent ATP-independent folding of client proteins by the middle and C-terminal domains of HSP90, which results in the degradation of unfolded GR and client kinases14. In contrast to this mechanism, we suggest that C-terminal HSP90 inhibitors induce the release of GR after its final folding promoted by the C-terminal Hsp90 domain. In this way, these C-terminal inhibitors do not influence the total amount of GR protein but instead increase the number of GR receptors that are stable and have high ligand binding affinity. Further, these results indicate that the disruption of the unfolding–refolding cycle of GR in intact (living and uncontrolled by any external manipulation) cells at a late step—by inhibitors acting directly on the C-terminal Hsp90 domain without interfering with the early ATPdependent functions of the N-terminal Hsp90 domain—promotes GR release from the chaperone cycle and into its final cellular fate as a regulator of gene transcription. Our results are in agreement with the mechanism of action proposed for novobiocin, in which it binds to the C-terminal domain of HSP90 and induces a conformation that prompts the release of kinases before they achieve maturation11,17,19,20,32. With regard to GR, however, novobiocin and silibinin, acting through the same mechanism, trigger its release from the Hsp90 complex in a fully mature state. The dissimilarity in maturation between GR and client kinases might be due to the fact that they interact with separate sites on HSP90 (refs. 23,33,34) and require different co-chaperones35, which causes the various client proteins to be at different stages of folding when HSP90 is inhibited at its C-terminus. Our results contrast with two earlier studies in which novobiocin inhibited the maturation of GR and targeted it for degradation36,37. This discrepancy is explained by the fact that novobiocin in high concentrations binds to HSP70 (ref. 36) and the N-terminal ATP binding pocket of HSP90, in addition to the more sensitive C-terminal binding site38 involved in the effects reported here. Therefore, novobiocin in high concentrations might directly inhibit ATP hydrolysis during the initial steps of the HSP90 catalytic cycle by acting like an N-terminal HSP90 inhibitor.

*

1

Control Vehicle Silibinin

e

Vehicle

Silibinin

Silibinin, novobiocin, and other HSP90 inhibitors have therapeutic effects in numerous types of cancer13,39. Silibinin in particular has an outstanding safety profile in humans, and it is currently used for the treatment of liver disease and poisoning40. Its effects on hormone secretion in the allograft model of Cushing disease are similar in magnitude to those produced by another experimental treatment that has been translated successfully into human studies25. Our results suggest that silibinin might be used to restore glucocorticoid sensitivity in subjects with Cushing disease, and our study opens the opportunity to explore a new pharmacological mechanism to regulate GR maturation in other diseases that involve resistance to glucocorticoids. Methods Methods and any associated references are available in the online version of the paper. Note: Any Supplementary Information and Source Data files are available in the online version of the paper.

Additional methods. Detailed methodology is described in the Supplementary Methods. Acknowledgments This work was partly supported by a grant from the German Research Foundation (SFB1035 to M.S.), the Bayerisches Staatsministerium für Wirtschaft, Infrastruktur, Verkehr und Technologie (m4 Award to F.H. and M.P.-P.), Federal Ministry of Education and Research (BMBF; PersoMed to M.P.-P. and F.H.); and postdoctoral fellowships to L.F. from the European Commission (FP7-PEOPLE-20112011-IIF 301193, Hsp90NMR) and the European Molecular Biology Organization (EMBO ALTF 1255-2011). The expression plasmids for the HSP90-α CTD and GR were gifts from U. Hartl (Max Planck Institute of Biochemistry) and S. Simons Jr. (US National Institutes of Health), respectively. We thank J. Stalla and T. Kloss for technical assistance. We thank M. Theodoropoulou and K. Lucia for valuable comments on the manuscript. AUTHOR CONTRIBUTIONS M.R. and M.P.-P. conceived of and designed the experiments. M.B. collected biopsies and diagnosed subjects. M.R., C.K., L.F., and M.P.-P. carried out experiments. M.R., M.S., F.H., G.K.S., and M.P.-P. analyzed the data. M.R., F.H., G.K.S., and M.P.-P. wrote the manuscript. COMPETING FINANCIAL INTERESTS The authors declare competing financial interests: details are available in the online version of the paper. Reprints and permissions information is available online at http://www.nature.com/ reprints/index.html.

advance online publication nature medicine


letters 1. Keller-Wood, M.E. & Dallman, M.F. Corticosteroid inhibition of ACTH secretion. Endocr. Rev. 5, 1–24 (1984). 2. Dahia, P.L.M. & Grossman, A.B. The molecular pathogenesis of corticotroph tumors. Endocr. Rev. 20, 136–155 (1999). 3. Newell–Price, J., Bertagna, X., Grossman, A.B. & Nieman, L.K. Cushing’s syndrome. Lancet 367, 1605–1617 (2006). 4. Tritos, N.A., Biller, B.M. & Swearingen, B. Management of Cushing disease. Nat. Rev. Endocrinol. 7, 279–289 (2011). 5. Karl, M. et al. Cushing’s disease preceded by generalized glucocorticoid resistance: clinical consequences of a novel, dominant-negative glucocorticoid receptor mutation. Proc. Assoc. Am. Physicians 108, 296–307 (1996). 6. Lamberts, S.W.J. Glucocorticoid receptors and Cushing’s disease. Mol. Cell. Endocrinol. 197, 69–72 (2002). 7. Pratt, W.B. & Toft, D.O. Steroid receptor interactions with heat shock protein and immunophilin chaperones. Endocr. Rev. 18, 306–360 (1997). 8. Picard, D. et al. Reduced levels of Hsp90 compromise steroid-receptor action in vivo. Nature 348, 166–168 (1990). 9. Kang, K.I. et al. The molecular chaperone Hsp90 can negatively regulate the activity of a glucocorticosteroid-dependent promoter. Proc. Natl. Acad. Sci. USA 96, 1439–1444 (1999). 10. Freeman, B.C. & Yamamoto, K.R. Disassembly of transcriptional regulatory complexes by molecular chaperones. Science 296, 2232–2235 (2002). 11. Marcu, M.G., Chadli, A., Bouhouche, I., Catelli, M. & Neckers, L.M. The heat shock protein 90 antagonist novobiocin interacts with a previously unrecognized ATPbinding domain in the carboxyl terminus of the chaperone. J. Biol. Chem. 275, 37181–37186 (2000). 12. Zhao, H., Brandt, G.E., Galam, L., Matts, R.L. & Blagg, B.S.J. Identification and initial SAR of silybin: An Hsp90 inhibitor. Bioorg. Med. Chem. Lett. 21, 2659–2664 (2011). 13. Trepel, J., Mollapour, M., Giaccone, G. & Neckers, L. Targeting the dynamic Hsp90 complex in cancer. Nat. Rev. Cancer 10, 537–549 (2010). 14. García-Morales, P. et al. Inhibition of Hsp90 function by ansamycins causes downregulation of cdc2 and cdc25c and G2/M arrest in glioblastoma cell lines. Oncogene 26, 7185–7193 (2007). 15. Whitesell, L. & Cook, P. Stable and specific binding of heat shock protein 90 by geldanamycin disrupts glucocorticoid receptor function in intact cells. Mol. Endocrinol. 10, 705–712 (1996). 16. Segnitz, B. & Gehring, U. The function of steroid hormone receptors is inhibited by the Hsp90-specific compound geldanamycin. J. Biol. Chem. 272, 18694–18701 (1997). 17. Matts, R.L. et al. Elucidation of the Hsp90 C-terminal inhibitor binding site. ACS Chem. Biol. 6, 800–807 (2011). 18. Young, J.C., Hoogenraad, N.J. & Hartl, F.U. Molecular chaperones Hsp90 and Hsp70 deliver preproteins to the mitochondrial import receptor Tom70. Cell 112, 41–50 (2003). 19. Marcu, M.G., Schulte, T.W. & Neckers, L. Novobiocin and related coumarins and depletion of heat shock protein 90–dependent signaling proteins. J. Natl. Cancer Inst. 92, 242–248 (2000). 20. Yun, B.G., Huang, W., Leach, N., Hartson, S.D. & Matts, R.L. Novobiocin induces a distinct conformation of Hsp90 and alters Hsp90–cochaperone–client interactions. Biochemistry 43, 8217–8229 (2004). 21. Dittmar, K.D., Demady, D.R., Stancato, L.F., Krishna, P. & Pratt, W.B. Folding of the glucocorticoid receptor by the heat shock protein (Hsp) 90-based chaperone machinery: the role of p23 is to stabilize receptor–Hsp90 heterocomplexes formed by Hsp90-p60-Hsp70. J. Biol. Chem. 272, 21213–21220 (1997).


22. Kirschke, E., Goswami, D., Southworth, D., Griffin, P. & Agard, D. Glucocorticoid receptor function regulated by coordinated action of the Hsp90 and Hsp70 chaperone cycles. Cell 157, 1685–1697 (2014). 23. Genest, O. et al. Uncovering a region of heat shock protein 90 important for client binding in E. coli and chaperone function in yeast. Mol. Cell 49, 464–473 (2013). 24. Lorenz, O.R. et al. Modulation of the Hsp90 chaperone cycle by a stringent client protein. Mol. Cell 53, 941–953 (2014). 25. Páez-Pereda, M. et al. Retinoic acid prevents experimental Cushing syndrome. J. Clin. Invest. 108, 1123–1131 (2001). 26. Kaul, S. et al. Mutations at positions 547–553 of rat glucocorticoid receptors reveal that Hsp90 binding requires the presence, but not defined composition, of a seven–amino acid sequence at the amino terminus of the ligand binding domain. J. Biol. Chem. 277, 36223–36232 (2002). 27. Philips, A. et al. Antagonism between Nur77 and glucocorticoid receptor for control of transcription. Mol. Cell. Biol. 17, 5952–5959 (1997). 28. Liu, B., Hammer, G.D., Rubinstein, M., Mortrud, M. & Low, M.J. Identification of DNA elements cooperatively activating proopiomelanocortin gene expression in the pituitary glands of transgenic mice. Mol. Cell. Biol. 12, 3978–3990 (1992). 29. Leung, C.K., Paterson, J., Imai, Y. & Shiu, R. Transplantation of ACTH-secreting pituitary tumor cells in athymic nude mice. Virchows Arch. A Pathol. Anat. Histol. 396, 303–312 (1982). 30. Bilodeau, S. et al. Role of Brg1 and HDAC2 in GR trans–repression of the pituitary POMC gene and misexpression in Cushing disease. Genes Dev. 20, 2871–2886 (2006). 31. Kino, T., De Martino, M.U., Charmandari, E., Mirani, M. & Chrousos, G.P. Tissue glucocorticoid resistance/hypersensitivity syndromes. J. Steroid Biochem. Mol. Biol. 85, 457–467 (2003). 32. Donnelly, A. & Blagg, B.S.J. Novobiocin and additional inhibitors of the Hsp90 C-terminal nucleotide-binding pocket. Curr. Med. Chem. 15, 2702–2717 (2008). 33. Fang, L., Ricketson, D., Getubig, L. & Darimont, B. Unliganded and hormonebound glucocorticoid receptors interact with distinct hydrophobic sites in the Hsp90 C-terminal domain. Proc. Natl. Acad. Sci. USA 103, 18487–18492 (2006). 34. Vaughan, C.K. et al. Structure of an Hsp90–Cdc37–Cdk4 Complex. Mol. Cell 23, 697–707 (2006). 35. Röhl, A., Rohrberg, J. & Buchner, J. The chaperone Hsp90: changing partners for demanding clients. Trends Biochem. Sci. 38, 253–262 (2013). 36. Kanelakis, K.C., Shewach, D.S. & Pratt, W.B. Nucleotide binding states of Hsp70 and Hsp90 during sequential steps in the process of glucocorticoid receptor–Hsp90 heterocomplex assembly. J. Biol. Chem. 277, 33698–33703 (2002). 37. Allan, R.K., Mok, D., Ward, B.K. & Ratajczak, T. Modulation of chaperone function and cochaperone interaction by novobiocin in the C-terminal domain of Hsp90: evidence that coumarin antibiotics disrupt Hsp90 dimerization. J. Biol. Chem. 281, 7161–7171 (2006). 38. Söti, C., Rácz, A. & Csermely, P. A nucleotide-dependent molecular switch controls ATP binding at the C-terminal domain of Hsp90: N-terminal nucleotide binding unmasks a C-terminal binding pocket. J. Biol. Chem. 277, 7066–7075 (2002). 39. Ramasamy, K. & Agarwal, R. Multitargeted therapy of cancer by silymarin. Cancer Lett. 269, 352–362 (2008). 40. Saller, R., Meier, R. & Brignoli, R. The use of silymarin in the treatment of liver diseases. Drugs 61, 2035–2063 (2001).

ONLINE METHODS We obtained all chemicals from Sigma-Aldrich unless stated otherwise.


Human samples. We obtained corticotroph pituitary adenomas (n = 26, 74% female 18–65 years old, 26% male 12–60 years old) and non-functioning pituitary adenomas (n = 17, 53% female 24–65 years old, 47% male 55–76 years old) from subjects that had undergone transphenoidal surgery. The specimens were stored in sterile cell culture medium at 4 °C for up to 24 h, processed for primary cultures, or frozen at −80 °C for protein extraction or cryosections. The clinical diagnosis was confirmed by routine immunohistochemistry for pituitary hormones. Postmortem human anterior pituitary samples (n = 9, 45% female 35–79 years old, 55% male 43–72 years old) were obtained from autopsies performed 5–16 h after accidental death, and frozen at −80 °C. All experiments with human material were performed under approval of the local Ethics Committee of the Ludwig Maximilian University of Munich, Germany. Informed written consent was received from each subject whose pituitary adenoma tissue was used in the study. Immunohistochemistry. We fixed cryosections of human corticotroph pituitary adenomas in 4% paraformaldehyde, and subsequently immunostained for HSP90-α with clone EPR3953 (Epitomics). Diaminobenzidine was used as substrate for the biotinylated secondary antibody anti-rabbit IgG (Vector Laboratories, B-1000) after loading with the ABC kit (Vector Laboratories). We counterstained nuclei with toluidine blue. Cell culture. We obtained the mouse anterior pituitary corticotroph adenoma cell line AtT-20/D16v-F2 recently from ATCC and grew it in what will hereafter be referred to as cell culture medium: DMEM (Gibco, 41965) supplemented with 10% FCS (Gibco, 10270; heat-inactivated for 1 h at 55 °C) and 100 U ml−1 penicillin/streptomycin (Biochrom) at 37 °C in a humidified atmosphere with 95% air and 5% CO2. AtT-20 cells were tested for mycoplasma (ATCC) and validated based on their response to dexamethasone using ACTH RIA as described below. Primary cell cultures. We only considered primary cultures with cell vitality above 95% and in which ACTH production was confirmed, as previously described25. We seeded cells in 96-well plates in cell culture medium with 10% charcoal-stripped FCS for 48 h before treatment. Male Sprague–Dawley rats from Charles River Laboratories (6–7 weeks of age) were allowed to acclimate for 4 d before sedation with CO2, decapitation, and dissection of the anterior pituitary. We seeded cells at a density of 2 × 104 cells per well in 96-well plates in cell culture medium with 10% charcoal-stripped FCS and MEM vitamins (Biochrom) for 48 h before treatment. Proliferation assays. We seeded AtT-20 cells at 2 × 103 cells per well in 96-well plates and left them to attach for 24 h. After that, we changed the medium to cell culture medium with 2% FCS with the indicated drugs. The DMSO (Roth) concentration was maintained constant for all conditions at 0.2% (vol/vol). After 96 h incubation, cell viability was determined using the WST-1 assay (Roche) according to the manufacturer’s instructions, or cells were counted at the indicated time points. Immunoblotting. Protein extraction from human tissue was done in RIPA buffer (50 mM Tris pH 8.0, 150 mM NaCl, 1% NP-40, 0.5% sodium deoxycholate, and 0.1% SDS) supplemented with Complete protease inhibitor cocktail. We centrifuged the homogenized lysate at 16,000g for 15 min and subjected the supernatant to Bradford protein assay (Bio-Rad). We performed immunoblotting using standard procedures. In short, 10–15 µg total protein boiled in Roti-Load 1 (Roth) were separated on a 10% acrylamide gel, transferred to PVDF membranes (Millipore), and incubated with antibodies to HSP90-α (ADI-SPA-840 at 1:2,000, Enzo Life Sciences), HSP90-β (clone E296 at 1:2,000, Abcam), HSF1 (MA5-14632 at 400 ng ml−1, Thermo Scientific), and β-actin (clone 8H10D10 at 1:4,000, Cell Signaling), followed by incubation with HRP-conjugated secondary antibodies at a dilution of 1:5,000 (Cell Signaling 7074S or 7076S) and detection with clarity ECL (Bio-Rad).

nature medicine

We seeded AtT-20 cells at 3 × 105 cells per well in 6-well plates and left them to attach for 24 h. After that, we replaced the medium with cell culture medium containing 2% FCS and added drugs at a constant DMSO concentration of 0.2%. After 48 h, we washed the cells in cold PBS and scraped them in RIPA buffer supplemented with Complete protease inhibitor cocktail. We centrifuged the lysate at 16,000g for 15 min and subjected the supernatant to Bradford protein assay. Membranes were incubated with antibodies to GR (M-20 at 1:1,500, Santa Cruz), Cdc2 (clone 1/Cdk1/Cdc2 at 1:1,500, BD Transduction Laboratories), HSP90-α (PA3-013 at 100 ng ml−1; Thermo Scientific), and β-actin (clone 8H10D10), followed by incubation with HRP–conjugated secondary antibodies (Cell Signaling 7074S or 7076S, or Sigma A9542, each diluted 1:5,000) and detection with clarity ECL (Bio-Rad). HSP90 binding assay. The cloning of amino acids 566–732 of human HSP90-α (HSP90 CTD) into pProEX HTa vector has been described previoulsy18. We induced expression in BL21 competent cells at OD600 = 0,5 with 600 µM IPTG until OD600 = 1 was reached and purified the His-tagged recombinant protein using a Ni-NTA column (Qiagen), followed by dialysis into HG buffer (20 mM HEPES, pH 8.0, 0.01% Triton X-100, and 10% glycerol) to a final concentration of 5 mg ml−1 protein. We immobilized silibinin to polystyrene plates (Costar, 3695) overnight at 4 °C in bicarbonate buffer (3.03 g Na2CO3, 6.0 g NaHCO3 in 1 l aqua bidest., pH 9.5) with constant 5% DMSO, followed by 1 h of blocking in PBS with 1% gelatin at room temperature (RT)41. Purified HSP90 was incubated for 1 h at RT in PBS with 1% BSA at a concentration of 5 µg ml−1, followed by three washes in PBS. Bound HSP90 was competed with increasing concentrations of silibinin or 17-AAG (Tocris) dissolved in PBS with 1% BSA and 0.05% Triton X-100 or novobiocin (Calbiochem) dissolved in TBS with 0.05% Tween-20 with DMSO at a constant concentration of 5% for 1 h at RT, followed by three washes in PBS. Primary (clone 68/Hsp90, BD Transduction Laboratories) and HRP-conjugated secondary antibody (Cell Signaling 7076S) were each diluted 1:1500 in PBS with 3% BSA plus 0.1% Tween-20 and incubated for 1 h at RT, followed by 3 washes in PBS. HRP substrate (R&D Systems, DY999) was added to the wells according to the manufacturer’s instructions, followed by colorimetry at 450 nm. NMR spectroscopy. NMR spectra were recorded on a Bruker AV800 spectrometer (Bruker Topspin 3.2, Bruker, Billerica, USA) at 32 °C. NMR experiments were processed using NMRPipe, and data analyzed using CcpNMR Analysis. Experiments were performed with 15N-labeled HSP90 CTD in PBS buffer supplemented with Triton X-100 (silibinin binding) or Tween-20 (novobiocin binding) in the presence or absence of silibinin or novobiocin. Stock silibinin was prepared at 30 mM in DMSO and stock novobiocin was prepared at 100 mM in PBS buffer. Additional spectra were recorded in the presence and absence of 1% DMSO to observe any effects from DMSO addition. Chemical shift perturbations (CSP) were based on 2D 1H, 15N water flip-back HSQC correlation experiments and calculated as a weighted chemical shift: ∆d N −H = (∆d1H × 10)2 + (∆d15N )2 NMR signals with a CSP 2σ greater than the mean were considered significant. Co-immunoprecipitation. We grew AtT-20 cells at 1 × 106 per 10-cm dish for 48 h in cell culture medium, and an additional 48 h in cell culture medium with 2% FCS. We then washed and scraped the cells in cold PBS. We dissolved the pooled cell pellet in 150 µl per dish of TEDGM (10 mM Tris, 50 mM NaCl, 4 mM EDTA, 1 mM DTT, 10% glycerol, and 10 mM Na–MoO4, pH 7.4) supplemented with Complete protease inhibitor cocktail, followed by two 1-s steps of sonication with a microtip at intermediate setting, and subsequent centrifugation for 15 min at 16,000g and 4 °C. We precleared the supernatant of 12 dishes with 10 µl per dish of Dynabeads–protein G (Invitrogen) for 30 min, rotating at 4 °C. We incubated aliquots of the supernatant with increasing concentrations of silibinin dissolved in TEDGM with 1% BSA and 0.05% Igepal630, or novobiocin dissolved in TEDGM with 0.1% Triton X-100, each with constant 3% DMSO for 1 h at 4 °C. Prior to drug incubation, we adjusted aliquots of the lysate to 0.5% BSA plus 0.05% Igepal-630 for silibinin, and

doi:10.1038/nm.3776

to 0.1% Triton X-100 for novobiocin. Per condition, we added 2 µl H-300 or 0.4 µg PA3-013 for 2 h at 4 °C. The immune complex was captured with 10 µl per condition of Dynabeads–protein G for 30 min at 4 °C, and washed 3 times with TEDGM with 0.1% Tween-20, and once with TEDGM without detergent, followed by boiling in Roti-Load 1 and immunoblotting for GR (M-20) and total HSP90 (clone H90-10 at 100 ng ml−1, Thermo Scientific).


3H-dexamethasone

binding. Binding of 3H-dexamethasone to GR has been described previously26. In short, we seeded AtT-20 cells at 2 × 105 cells per well in 24-well plates in cell culture medium containing 10% charcoal-stripped FCS and left them to attach for 24 h. After that, we replaced the medium with cell culture medium containing 0.1% charcoal-stripped FCS and added vehicle or 30 µM silibinin, each with a total concentration of 0.2% DMSO. After 48 h, we replaced the medium with DMEM containing 3H-dexamethasone (PerkinElmer) with or without a 500-fold excess of unlabeled dexamethasone for 4 h at 4 °C, followed by careful washing in cold PBS and lysis with Passive Lysis Buffer (Promega). We added the lysate to Ultima Gold scintillation fluid (PerkinElmer) and counted it in a beta counter (Beckman). In parallel, control wells were subjected to immunoblotting for GR (M-20) and β-actin (clone 8H10D10), and additional wells were trypsinized to count viable cells for normalization. The dissociation constant (Kd) and maximal binding (Bmax) to calculate receptor binding sites was determined using nonlinear regression for single site saturation ligand binding (Sigma Plot). The data points depict specific binding in cpm per million cells (total binding minus unspecific binding). The Scatchard plot was derived from bound/free plotted over bound 3H-dexamethasone (cpm per million cells) after subtraction of unspecific binding (in the presence of a 500-fold excess of unlabeled dexamethasone) from total binding. Reporter assays. We seeded AtT-20 cells at 1 × 105 cells per well in 24-well plates in cell culture medium containing 10% charcoal-stripped FCS and left them to attach for 24 h. In each well, we transfected 0.4 µg empty pcDNA3.1+ (Invitrogen) and 0.3 µg reporter gene: MMTV–Luc as previously described25, NurRE–Luc27 or POMC–Luc28 plus 0.1 µg RSV-Gal (Addgene) for normalization for 3 h using Superfect (Qiagen) according to the manufacturer’s instructions. Similarly, for overexpression of rat GR or the AAA triple mutant (P548A/T549A/V551A of rat GR)26, we transfected 0.2 µg of expression vector together with 0.2 µg pcDNA3.1+ plus reporter plasmids as above. After that, we replaced the transfection mix with cell culture medium without FCS overnight. We then added HSP90 inhibitors (30 µM silibinin for overexpressed WT GR and AAA mutant) in cell culture medium with 2% charcoal-stripped FCS at a constant DMSO concentration of 0.2% for 30 h, and dexamethasone was added for additional 18 h. After washing in PBS, the cells were lysed in Passive Lysis Buffer, followed by measurement of reporter activity according to the manufacturer’s instructions (Promega). Relative Luc values (Luc/Gal) were used to normalize for transfection efficiency in all experiments.

doi:10.1038/nm.3776

ACTH radioimmunoassay (RIA). ACTH RIA was performed as described previously25. We seeded AtT-20 cells at 1 × 105 cells per well in 24-well plates in cell culture medium containing 10% charcoal-stripped FCS and left them to attach for 24 h. We treated them with silibinin in cell culture medium with 0.1% charcoal-stripped FCS for 24 h, followed by a washing step in the same medium. We added silibinin together with dexamethasone in cell culture medium with 0.1% charcoal-stripped FCS for an additional 24 h. We kept the DMSO concentration constant at 0.2% for all conditions. The supernatant was harvested and stored at −20 °C for RIA. We treated primary cultures of human corticotroph adenomas as described for AtT-20 cells, and the values refer to ACTH secreted by 1 × 104 cells per 24 h. Likewise, primary cultures of normal rat pituitary cells were treated as described for AtT-20 cells, except that the cell culture medium contained 1% charcoal-stripped FCS and MEM vitamins. In vivo experiments. We injected 24 male NMRI Foxn1nu/Foxn1nu mice from Harlan Laboratories (4–6 weeks) under isoflurane anesthesia unilaterally subcutaneously with 30 µl AtT-20 cells dissolved in PBS at a concentration of 5 × 106 per ml. We divided the animals using number-based simple randomization into vehicle (n = 12; 10 ml kg−1 d−1 aqua bidest.) and treatment group (n = 12; 300 mg kg−1 d−1 silibinin (Silicur, Hexal) in aqua bidest.). Seven days after injection, we started the administration of silibinin or vehicle per gavage to all animals (Time 0). In a non-blinded setting, we determined tumor volumes every 4 d with a caliper. We collected blood samples from isofluraneanesthetized animals by cardiac puncture 24 h after the last treatment. We used RIA to measure plasma ACTH (ImmuChem, MP Biomedicals Ltd., Germany) according to the manufacturer’s instructions, as well as plasma corticosterone (Rat/Mouse RIA, DRG Instruments Ltd., Germany). No animals were excluded from analyses. All animal experiments were conducted according to the guidelines of the German Animal Welfare Act and were approved by the Ethical Commission of the Albert Ludwig University Freiburg. Statistical analysis. P ≤ 0.05 was considered statistically significant in twotailed, unpaired Student’s t-tests. All the in vitro experiments included negative controls and reached significance in triplicates. For the allograft model in nude mice, we performed preliminary experiments on tumor growth and hormonal concentrations to estimate the effect size. From these data, we expected a strong effect and therefore chose the sample size of n = 12 animals per treatment group. For all data, violations of parametric t-test assumptions with 5% significance threshold were tested. The Kolmogorov–Smirnov test was applied to analyze normality of data distribution in the samples, followed by F-test to verify whether the assumption of equal variances between groups was fulfilled. 41. Morra, G. et al. Dynamics-based discovery of allosteric inhibitors: selection of new ligands for the C-terminal domain of Hsp90. J. Chem. Theory Comput. 6, 2978–2989 (2010).

nature medicine