Using the heat-shock response to discover anticancer compounds that target protein homeostasis

Abstract

Unlike normal tissues, cancers experience profound alterations in protein homeostasis. Powerful innate adaptive mechanisms, especially the transcriptional response regulated by Heat Shock Factor 1 (HSF1), are activated in cancers to enable survival under these stressful conditions. Natural products that further tax these stress responses can overwhelm the ability to cope and could provide leads for the development of new, broadly effective anticancer drugs. To identify compounds that drive the HSF1-dependent stress response, we evaluated over 80,000 natural and synthetic compounds as well as partially purified natural product extracts using a reporter cell line optimized for high-throughput screening. Surprisingly, many of the strongly active compounds identified were natural products representing five diverse chemical classes (limonoids, curvularins, withanolides, celastraloids, and colletofragarones). All of these compounds share the same chemical motif, an α,β-unsaturated carbonyl functionality, with strong potential for thiol-reactivity. Despite the lack of a priori mechanistic requirements in our primary phenotypic screen, this motif was found to be necessary albeit not sufficient, for both heat-shock activation and inhibition of glioma tumor cell growth. Within the withanolide class, a promising therapeutic index for the compound withaferin A was demonstrated in vivo using a stringent orthotopic human glioma xenograft model in mice. Our findings reveal that diverse organisms elaborate structurally complex thiol-reactive metabolites that act on the stress responses of heterologous organisms including humans. From a chemical biology perspective, they define a robust approach for discovering candidate compounds that target the malignant phenotype by disrupting protein homeostasis.

Figure 1

Glioma cells experience proteotoxic stress and are dependent on HSF1 for proliferation and survival. a) Immunoblot analysis of normal diploid cell line IMR90 and established glioma cell lines LN428, LN827 and U87 with antibodies to HSP27, HSP90, and HSF1. α-Tubulin, loading control. b) Immunoblot of LN428 cells 3.5 days following lentivirus-mediated knockdown of HSF1 with shRNA. shRNA-ha6 and ha9 target HSF1. shRNA-GFP and shRNA-SCR (scrambled) are controls that do not target HSF1. c) Viability of glioma cells 6 days following infection with indicated lentiviruses. % viability relative to uninfected control is plotted (six replicates per condition). d) Photomicrographs of glioma cells without lentivirus infection (control, top row) and following lentivirus-mediated HSF1 knockdown with shRNA-ha6 (bottom row). Photomicrographs at 6 days following infection.

Figure 2

Small-molecule natural products with an α,β-unsaturated carbonyl moiety identified in a screen for heat-shock response activators.

Figure 3

Thiol-reactive natural products induce a heat-shock response. a) Heat map of HSE-eGFP induction compared to vehicle only. Determinations were in triplicate. b) Immunoblot analysis of LN428 following treatment with indicated amounts of WA (6) for 16 hours. c) UV absorption spectra of WA (6) and pubesenolide (10) with or without Lcysteine. d) NMR spectra of WA (6) with or without glutathione. e) Heat-shock reporter activation by WA alone and in the presence of a 50x molar excess of n-acetylcysteine (NAC). Dashed line at 100% indicates baseline level of reporter signal. Decrease below 100% reflects cytotoxicity.

Figure 4

Thiol-reactive natural products are cytotoxic to glioma cell lines. Heat maps of glioma cell survival in the presence of thiol-reactive natural products and analogs. LN428 cells have wild-type PTEN and a p53 V173M/R282W mutation. LN827 cells have an Exon 3 splice acceptor mutation of PTEN and a p53 mutation. U87 cells have an Exon 3 splice acceptor mutation of PTEN and a wild-type p53. BT70 cells have a G44V mutation in PTEN and R273C mutation in p53. Cells were treated with the indicated concentrations of the compounds and viability was measured using resazurin 48 h later (in triplicate).

Figure 5

WA has anti-cancer activity in a glioma stem cell-based orthotopic xenograft model in mice. a) Schematic depiction of the previously characterized BT70 glioma orthotopic model. The cells were derived from a human glioblastoma (example H&E shown). Tumor spheres are immunoreactive for MIB1 (red) and OLIG2 (green) and form infiltrative tumors (H&E). b) Kaplan-Meier analysis of mice bearing orthotopic BT70 xenografts treated with WA (12 mg/kg) or vehicle control. c) HMOX1 (HSP32) mRNA expression in orthotopic BT70 xenografts is modulated by WA treatment. Measurements were in duplicate using the Ncounter system (Error bars SD, p=0.09 for 6 mg/kg WA, p<0.001 for 12 mg/kg WA).