Abstract
Treatment approaches using Hsp90 inhibitors at their maximum tolerated doses (MTDs) have not produced selective tumor toxicity. Inhibition of Hsp90 activity causes degradation of client proteins including those involved in recognizing and repairing DNA lesions. We hypothesized that if DNA repair proteins were degraded by concentrations of an Hsp90 inhibitor below those required to cause nonspecific cytotoxicity, significant tumor-selective radiosensitization might be achieved.
Tandem mass tagged-mass spectrometry was performed to determine the effect of a subcytotoxic concentration of the Hsp90 inhibitor, AT13387 (onalespib), on global protein abundance. The effect of AT13387 on in vitro radiosensitization was assessed using a clonogenic assay. Pharmacokinetics profiling was performed in mice bearing xenografts. Finally, the effect of low-dose AT13387 on the radiosensitization of three tumor models was assessed.
A subcytotoxic concentration of AT13387 reduced levels of DNA repair proteins, without affecting the majority of Hsp90 clients. The pharmacokinetics study using one-third of the MTD showed 40-fold higher levels of AT13387 in tumors compared with plasma. This low dose enhanced Hsp70 expression in peripheral blood mononuclear cells (PBMCs), which is a biomarker of Hsp90 inhibition. Low dose monotherapy was ineffective, but when combined with radiotherapy, produced significant tumor growth inhibition.
This study shows that a significant therapeutic ratio can be achieved by a low dose of Hsp90 inhibitor in combination with radiotherapy. Hsp90 inhibition, even at a low dose, can be monitored by measuring Hsp70 expression in PBMCs in human studies.
At least 20 Hsp90 inhibitors have been developed by academic groups and pharmaceutical companies. However, tolerable monotherapy doses of all the Hsp90 inhibitors have shown only limited efficacy. This preclinical study demonstrates that a subcytotoxic concentration of an Hsp90 inhibitor can inhibit the DNA repair process and sensitize tumors to radiotherapy. These results show that while Hsp90 inhibition alone does not seem to be an effective approach in cancer therapy, significant improvement can be achieved without systemic toxicity if a lower dose of Hsp90 inhibitory therapy is combined with radiotherapy.
Introduction
Hsp90 serves as an ATP-dependent molecular chaperone that aids in cellular protein maturation in a highly regulated process and stabilizes its clients during various stressful conditions. Over 600 identified clients of Hsp90 (http://www.picard.ch/downloads/Hsp90facts.pdf) are involved in a multitude of cellular processes (e.g., cell-cycle control and proliferative/antiapoptotic signaling). Several clients are crucial for normal cell growth, while many of them are oncogenic (1). Hsp90 often exerts cytoprotective functions by preventing the degradation of proteins involved in the recognition and removal of a diverse array of DNA lesions in normal tissues (2, 3). It has been considered as a promising therapeutic target since the realization that several mutated or aberrantly expressed oncogenes depend on Hsp90 for their stabilization (4).
Hsp90 cycles with client proteins in different ways. The classic client proteins, such as proteins involved in DNA repair, undergo cycles of protein·Hsp90 heterocomplex assembly/disassembly to produce complexes that are stabilized such that they can be analyzed by normal biochemical separation techniques. Client proteins that undergo cycles of stable heterocomplex assembly are very unstable in the absence of cycling with Hsp90, and treatment with Hsp90 inhibitors is followed by rapid client protein degradation. Other client proteins are more inherently stable, undergo more dynamic cycles of Hsp90 assembly, and exhibit slower degradation when Hsp90 is inhibited. Thus, the consequences of Hsp90 inhibition may be more severe with classic clients than with dynamically cycling clients (5).
However, this putative chemotherapeutic benefit of Hsp90 inhibitors has not been fully realized for patients in clinical trials. The tolerability profile in human subjects remains unclear, despite Hsp90's druggability, with data from 20 inhibitors and more than 70 clinical trials (completed/withdrawn/ongoing) with 11 specific inhibitors. The main difficulty lies in finding a therapeutic window in which efficacy outweighs toxicity. This is likely due to the critical normal functions that Hsp90 regulates in normal cells. In clinical studies, where these inhibitors have been used at the maximally tolerated or effective dose levels to inhibit the target, they also accentuate night blindness and severe hematologic, cardiac, and hepatotoxicities (6–9). So far, none of these inhibitors has achieved FDA approval.
Inhibition of Hsp90 activity promotes degradation of client proteins that include a spectrum of oncogenes involved in cell survival, such as EGFR, ErbB2, Src, and cMet, as well as molecules involved in DNA repair, such as ATM, ATR, RAD51, and DNA-PK (4, 10–13). Degradation of these proteins correlates with the induction of γH2Ax, inhibition of DNA repair, cell death, and profound radiosensitization (14). Several groups have conducted preclinical studies and shown that inhibition of Hsp90 activity would be an ideal approach to sensitize tumor cells to localized ionizing radiotherapy (12, 15–21). We have listed 70 clinical trials (Supplementary Table S1) and observed only two trials (results not posted) that combined an Hsp90-targeted agent (ganetespib and AT13387) with chemo (cisplatin, paclitaxel, and carboplatin) and radiotherapy. On the basis of our understanding from published results, we can postulate a possibility of dose-related severe toxicities in human subjects receiving a combination of systemic therapy (both molecularly targeted agents and chemotherapy) and radiotherapy.
Thus, the main objective of this study was to determine a method of using Hsp90 inhibitors that limits the toxicities previously observed in clinical studies while still sensitizing cancer cells to radiotherapy. As DNA repair proteins are among the Hsp90 clients, which would be involved in repair after irradiation, we hypothesized that low concentrations/doses of Hsp90 inhibitors, which are not toxic on their own could selectively radiosensitize by causing degradation of some DNA repair proteins. Therefore, we designed a study to determine whether DNA repair proteins are affected by low concentrations/doses of Hsp90-targeting agents. Radiosensitization was assessed after we found selective degradation of DNA repair proteins. We found that noncytotoxic concentrations of AT13387 radiosensitized a panel of head and neck squamous cell carcinoma (HNSCC) cell lines. Thus, we combined systemic administration of lower doses of an Hsp90 inhibitor prior to localized radiotherapy in xenograft models and found substantial radiosensitization of tumors, with minimal toxicity. We also examined potential biomarkers for validation of Hsp90 inhibition for future clinical studies.
Our data suggest that while Hsp90 activity inhibition monotherapy may not be effective in patients because of dose-related toxicities, lower, nontoxic doses of these inhibitors may be effective in combination with radiotherapy. This radiosensitization may relate to inhibition of DNA repair in tumors receiving targeted radiation.
Materials and Methods
Materials
AT13387 (Astex Pharmaceuticals) was kindly provided by the Cancer Therapy Evaluation Program, NIH (Bethesda, MD). EGFR (sc-03), Rad51 (sc-398587), Chk1 (sc-8408), and CyclinB1 (sc-245) antibodies were procured from Santa Cruz Biotechnology. Antibodies for Hsp70 (catalog no., 4872), ATR (catalog no., 13934), ATM (catalog no., 2873), DNA-PK (catalog no., 12311), Caspase3 (catalog no., 9662), cMet (catalog no., 3127), and GAPDH (catalog no., 2118) were purchased from Cell Signaling Technology. γH2Ax (catalog no., 05-636) antibody was purchased from Millipore. Rad51 (Ab133534), pT1989-ATR (ab227851), PT68-Chk2 (ab85743), and pS345-Chk1(ab47318) were bought from Abcam. pS1981-ATM (NB100-306) was purchased from Novus.
Cell culture
Late and early passages of several human HNSCC cell lines, UMSCC1, UMSCC10B, UMSCC11B, UMSCC29, UMSCC47, UMSCC74B, and UMSCC92 were kindly provided by Dr. Thomas Carey (University of Michigan, Ann Arbor, MI). The SCCVII, MiaPaCa-2, Panc 1, and BxPC3 cell lines were purchased from the ATCC. All cell lines were grown in RPMI 1640 medium, except for UMSCC92, MiaPaCa-2, and Panc 1, which were grown in DMEM, supplemented with 10% cosmic calf serum. Cell lines were regularly tested for Mycoplasma. Although the donating laboratory and ATCC confirmed the genotypes, the genotypes of selected human cell lines were validated by the UM genotyping core using AMPFLSTR Identifier Plus Assay and Gene Mapper Software, Version 5 (Applied Biosystems). A keratinocyte cell line (BS4 mutant) was kindly provided by Dr. Stefan Stoll (University of Michigan, Ann Arbor, MI) and grown in EpiLife Medium from Gibco (catalog no., S0125).
Protein identification and relative quantitation by tandem mass tag labeling and MultiNotch MS3 tandem mass spectrometry
UMSCC74B cells (1.5 × 106) were treated with either 100 nmol/L or 3 μmol/L AT13387, and cells were harvested at the beginning of the treatment (0 hours) and after 9 and 24 hours (n = 2 at each time point for both concentrations). Protein identification and relative quantitation were performed by the tandem mass tag (TMT)-mass spectrometry approach as described by several groups (22, 23). “The Database for Annotation, Visualization and Integrated Discovery v6.8” and “The Protein Analysis Through Evolutionary Relationships-v.15.0” were used to understand the unbiased biological relevance of significantly (P ≤ 0.05) up- or downregulated proteins upon 100 nmol/L AT13387 treatment. The detailed method used to analyze the enriched/overrepresented pathways has also been provided in the Supplementary Materials and Methods section.
Irradiation
Irradiation was carried out at a dose rate of either 2 Gy (for SCCVII and all HNSCC cell lines or xenografts) or 1.8 Gy (pancreatic cancer cell lines or xenograft)/minute using a Philips RT250 (Kimtron Medical) in the University of Michigan Comprehensive Cancer Center Experimental Irradiation Core (Ann Arbor, MI). All animals were anesthetized with isoflurane before tumor irradiation and received irradiation only at flank tumors, while the rest of the body was shielded from radiation. The dosimetry was carried out using an ion chamber that is directly traceable to a National Bureau of Standards. All irradiations were conducted at room temperature.
Immunoblotting
Cells were plated in 60-mm dishes at a density of 3 × 105 cells per dish and incubated overnight or to 70% confluence. The cells were treated either with the vehicle (DMSO) or AT13387 (100 nmol/L). Media were changed immediately, and cells were harvested at various time points. To validate TMT-LC/MS-MS data, UMSCC74B cells were harvested at 0, 1, 3, 9, and 24 hours, and the rest of the lines were harvested at 0, 9, and 24 hours following AT13387 treatment. The pellets were washed twice with ice-cold PBS and resuspended in lysis buffer for 30 minutes. After sonication, particulate material was removed by centrifugation at 13,000 rpm for 10 minutes at 4°C. The soluble protein fraction was heated to 95°C for 5 minutes and then applied to a 4%–12% Bis-Tris Precast Gel (Invitrogen) and transferred onto a polyvinylidene difluoride membrane. Membranes were incubated for 1 hour at room temperature in a blocking buffer consisting of 5% BSA and 1% normal goat serum in TBS [137 mmol/L NaCl, 20 mmol/L Tris-HCl (pH 7.6), and 0.1% (v/v) Tween 20]. Membranes were subsequently incubated overnight at 4°C with the primary antibody in blocking buffer, washed, and incubated for 1 hour with horseradish peroxidase–conjugated secondary antibody. After three additional washes in TBS, bound antibody was detected by enhanced chemiluminescence plus reagent. For quantification of relative protein levels, immunoblot films were scanned and analyzed using ImageJ 1.32j software. The relative protein levels shown represent a comparison with untreated controls.
Clonogenic cell survival assay
Clonogenic assays were performed using a standard technique described previously (24, 25). Briefly, 250–10,000 cells were plated in 60-mm dishes in triplicate, and the next day, cells were treated with 100 nmol/L of AT13387 or 17-AAG. After 24 hours of treatment, cells were irradiated at doses from 0 to 8 Gy. Immediately after irradiation, the media were changed, and the cells were maintained in normal culture conditions. Eight to 15 days later, colonies were fixed with acetic acid/methanol (1:7, v/v), stained with crystal violet (0.5%, w/v), and counted using a stereomicroscope. Drug cytotoxicity (surviving drug-treated cells) was measured and normalized to the survival of the untreated control cells. Cell survival curves were fitted using the linear-quadratic equation, and the mean inactivation dose was calculated according to the method of Fertil and colleagues (26). The cell survival enhancement ratio was calculated as the ratio of the mean inactivation dose under control conditions divided by the mean inactivation dose after drug exposure. A value significantly greater than 1 indicated radiosensitization.
DNA damage marker γH2Ax focus formation assay
Cells were plated on sterile glass coverslips in 100-mm dishes. The next day, cells were treated with either vehicle (DMSO) or 100 nmol/L AT13387. Twenty-four hours after AT13387 treatment, cells were irradiated (2 Gy) and the culture medium was replaced immediately. The coverslips were removed after 30 minutes, 1 hour, 2 hours, and 6 hours postirradiation, and cells were fixed in 4% paraformaldehyde in PBS for 15 minutes at room temperature, followed by a wash in PBS. The cells were then permeabilized in 0.5% Triton X-100 for 10 minutes and blocked in (5% goat serum and 1% BSA, in TBS) for an hour at room temperature. The cells were sequentially incubated with the γH2Ax antibody at 1:1,500 dilutions overnight at 4°C and Alexa Fluor 594–conjugated secondary antibody at 1:300 dilutions for 1 hour at 37°C in PBS with 1% BSA. After three washes in 1 × TBS to remove the nonspecific antibody binding, the nuclei were counterstained with 4-diamidino-2-phenylindole dihydrochloride (DAPI). Fluorescence images were obtained using a DS-Fi1 (Nikon) camera fitted on an Olympus LX71 Microscope. Fields were chosen at random based on DAPI staining. For quantitation of γH2Ax foci and γH2Ax-stained cells, at least 100 cells from each of four independent experiments were visually scored for each condition. Cells with 10 or more γH2Ax foci were scored as positive.
The percentage of γH2Ax-positive cells in each treatment group (AT13387 only, radiation only, and combination of AT13387 + radiation) was compared via a linear model that included the effect for the time elapsed from radiation (minutes) and the treatment group. A quadratic time effect was incorporated into the model based on plots of the average γH2Ax foci per cell. Models were fit separately for each cell line. Differences between treatment groups were assessed using two-sided t tests with a significance level of 0.05. All analyses were conducted in R version 3.6.2.
Cell-cycle analysis
UMSCC74B, SCCVII, and MiaPaCa-2 cells (0.5 × 106) were grown overnight in 100-mm dishes and treated with 100 nmol/L AT13387 for 24 hours. Cells were harvested with 1 × ice-cold PBS and fixed in ice-cold 70% ethanol for 30 minutes at 4°C. Ethanol was removed, and fixed cells were washed twice with 1 × PBS by centrifuging at 850 × g for 5 minutes. Each sample was incubated with 2 μg/mL RNase for 1 hour at 37°C, followed by staining with propidium iodide (final concentration 1.8 μg/mL) at 4°C in the dark for 2 hours. Cell acquisition and data analysis were done by ZE5 Cell Analyzer (Bio-Rad) and Everest Software, respectively. In each case, 10,000 cells were analyzed in the Flow Cytometry Core of the University of Michigan Rogel Cancer Center (Ann Arbor, MI).
Pharmacokinetics and pharmacodynamics analysis
Pharmacokinetic/pharmacodynamic parameters were determined after 40 mg/kg i.p. administration of AT13387 to athymic BALB/c mice bearing subcutaneous UMSCC74B xenografts. In one group, blood and tumor samples were collected at 0.1, 2, 24, and 96 hours after the first administration. In another group, the second dose of AT13387 was administered 96 hours later, and blood and tumor samples were drawn at 98, 120, and 192 hours from the first intraperitoneal injection. Blood was drawn from the heart into a heparinized syringe. Samples were stored at −80°C until analysis. Both tumor tissue homogenates and plasma were processed in the same manner for LC/MS at the core facility of the University of Michigan (Ann Arbor, MI). A detailed protocol has been provided in the Supplementary Materials and Methods section.
Isolation of peripheral blood mononuclear cells
Mouse peripheral blood mononuclear cells (PBMCs) were isolated from whole blood by density gradient centrifugation using Ficoll, as described elsewhere (27, 28). Briefly, 2 mL of diluted blood (1:1 in PBS; 1 × PBS) was carefully layered over 2-mL Histopaque-1077 and centrifuged at 3,000 rpm for 30 minutes without brake and acceleration. After centrifugation, the interphase was collected and centrifuged again at 1,800 rpm for 10 minutes. Subsequently, the pellet was resuspended and centrifuged at 1,000 rpm for 10 minutes, two times with 1 × PBS. Finally, the pellet was assessed for immunofluorescence as well as immunoblotting.
Tumor growth studies
To generate tumor xenografts, 1 × 106 UMSCC74B or MiaPaCa-2 cells were transplanted into the flanks of athymic nude Fox n1/nu mice (Harlan Laboratories). SCCVII cells were transplanted into the flanks of immunocompetent C3H/HeJ mice. When the tumors reached a volume of approximately 100 mm3 (∼5–10 days), the mice were randomized into four groups (one vehicle control and three experimental groups for AT13387, irradiation, and the combination of both AT13387 and irradiation), and treatment was initiated. Tumors (>100 mm3) were separately harvested at various time points for pharmacodynamic studies. On the basis of the pharmacokinetics/pharmacodynamics data, we further reduced the dose of AT13387 to 30 mg/kg, and the mice were treated twice a week (Monday and Thursday). In the irradiation-only arm, UMSCC74B and SCCVII tumors were irradiated with 2 Gy fractions from Monday to Friday for 2 weeks (total 2 × 5 × 2 = 20 Gy). MiaPaCa-2 tumors were irradiated with 1.8 Gy fractions from Monday to Friday for 2 weeks due to their higher radiosensitivity (total 1.8 × 5 × 2 = 18 Gy). In the combination group, the mice were treated with AT13387, 2 hours prior to irradiation (only on Monday and Thursday). Body weight and other signs of toxicity were monitored daily, and the tumor volumes were recorded using calipers. Tumor volumes were calculated as follows: volume (mm3) = (L × W2)/2. Animals from all four groups were euthanized when tumors reached 800 mm3 volume. Time to second tumor doubling (TD2) was calculated as the time in days from the start of treatment to the second instance of tumor doubling. TD2 was summarized using the Kaplan–Meier method, and survival curves were compared using the log-rank test (29, 30). The effects of treatment on Hsp70 were analyzed by immunoblotting, IHC, and immunofluorescence.
Tissue microarray
A tissue microarray (TMA) was used to study multiple human cancer/normal tissues in a single assay. Each tumor sample from all three xenograft models was collected at the indicated time point (6 hours following irradiation or 30 hours following AT13387 treatment in the combination arm). The TMA was developed, stained, and analyzed at the Research Histology and Immunohistochemistry Core, University of Michigan Rogel Cancer Center (Ann Arbor, MI). To understand the in vivo mechanisms of radiosensitization, γH2Ax, Rad51, pChk1/2, pATM, and pATR were analyzed using these TMA samples. Statistical analysis (one-way ANOVA, P < 0.05; **, P < 0.001; ***, P < 0.0001) was used to measure the level of significance.
Immunostaining
The Tissue and Histology Core of the University of Michigan Rogel Cancer Center (Ann Arbor, MI) and the Pathology Core for Animal Research at the University of Michigan (Ann Arbor, MI) assisted in preparing specimens for IHC. After slides were deparaffinized in xylene and rehydrated using serial ethanol dilutions, antigen unmasking was performed by immersing slides in citrate buffer (pH 6.0) for 20 minutes at high pressure and temperature using a pressure cooker. Slides were then washed in PBS, blocked for 1 hour, and incubated in the primary antibody at 4°C overnight. Slides were then washed again in PBS, incubated in secondary antibody for 1 hour, rewashed, and mounted with a coverslip. Images were acquired using a DS-Fi1 (Nikon) camera fitted on an Olympus LX71 Microscope.
Ethics statement
Here, we confirm that all the animal studies conducted here were approved by the University Committee on Use and Care of Animals of the University of Michigan (Ann Arbor, MI, protocol No. PRO00006797).
Results
Mechanisms of AT13387-mediated radiosensitization
We began by determining the effect of a cytotoxic concentration (3 μmol/L) of AT13387 (treated cells failed to form colonies at this concentration) on the global protein abundance using high-resolution TMT-LC/MS-MS in UMSCC74B cells after a 24-hour exposure (9-hour data, Supplementary Table S2). This concentration and time were selected on the basis of the peak plasma concentration observed in a phase I clinical study (26). A total of 5,883 proteins with high confidence peptide-spectrum match were detected. A standard volcano plot was developed using a total of 5,086 proteins which show a low FDR (Fig. 1). This analysis showed that the abundance, defined as log2 of treated versus untreated condition when the limit was set to ≤−0.5- or ≥0.5-fold, of more than 20% of the proteins was significantly reduced. These proteins belong to a multitude of cellular processes including those involved in the regulation of cell survival (e.g., EGFR), proliferation (transcription factor, jun-B), signaling (tyrosine-protein kinase, JAK1), cell cycle (cyclin-dependent kinases), transcription (CCAAT/enhancer-binding protein beta), genome/mitochondrial integrity (histone proteins), and DNA repair. This analysis also showed significant upregulation of more than 15% of detected proteins involved in protein degradation machinery (ubiquitin-conjugating enzyme E2), cell-cycle progression (aurora Kinase A), cell stress (e.g., Hsp cochaperones), and cell death pathway (e.g., annexins and DnaJ proteins). The change in the global proteome at 3 μmol/L AT13387 in UMSCC74 cells is listed in Supplementary Table S2. The magnitude observed on protein levels (total 2,145 significantly affected protein of 5,086) suggests that the 3 μmol/L AT13387 concentration would not be selective. Therefore, when given at a dose producing such drug levels in patients, it seems unlikely that the cumulative effect of the drug would be selective for cancer treatment.
Next, we set out to determine whether a correlation exists between the inhibition of Hsp90 activity and the upregulation in Hsp70 protein expression. Hsp70 is a cochaperone protein and its upregulation is known to serve as a biomarker of Hsp90 inhibition. UMSCC74B cells were treated for 2 hours with a range of AT13387 concentrations from 1 nmol/L to 10 μmol/L. Cell lysates were resolved on SDS-PAGE and change in the Hsp70 protein level was assessed by immunoblotting (Supplementary Fig. S1). We noted an approximately twofold increase in the Hsp70 protein levels at 100 nmol/L AT13387 concentration. This effect was only minimally enhanced at the higher concentrations. This concentration is 30-fold less than the Cmax achieved in a phase I clinical study (31), and is expected to be safe in patients. Also, a 24-hour exposure to this concentration resulted in greater than 70% survival of normal keratinocytes (Supplementary Fig. S9), demonstrating a relative lack of toxicity. Therefore, we repeated the proteomic experiment as described in Fig. 1 using 100 nmol/L AT13387. We hypothesized that this concentration would affect only the classic clients of Hsp90 that are primarily regulated by Hsp90 chaperone activity.
As shown in Fig. 2, we found significant changes in only 3.6% of the global proteins (211/5,086 proteins, P < 0.05) at the 100 nmol/L concentration at 24 hours after treatment in UMSCC74B cells. Hsp70 was the most upregulated protein at this concentration as well (log2 of fold change, 1.86), confirming data shown in Supplementary Fig. S1 that this low concentration inhibits Hsp90 chaperone activity (the entire list of identified proteins in 100 nmol/L AT13387-treated UMSCC74 cell lysate is also provided in the Supplementary Table S2). The amplitude of the effects of 100 nmol/L AT13387 on the global proteome was significantly lower than that of 3 μmol/L AT13387. Interestingly, except for Hsp70, the abundance of any other protein did not reach beyond the 1.5-fold abundance (log scale) level. The majority of proteins either up- or downregulated are known to be associated with several biological processes like cell–cell adhesion, protein folding, DNA damage/repair, chromatin remodulation, transcription, RNA editing/splicing, mitochondrial function, oxidative stress, or immunomodulation pathways. We demonstrate enriched biological processes (genes-to-terms) associated with significantly affected (log2 fold change ≥ 0.5; P ≤ 0.05) proteins in Supplementary Table S2. A higher enrichment score for a group indicated that the group members are involved in more important (enriched) roles. Genes associated with the term “DNA damage or repair” appeared later in the list (cluster 8, 17, 19, etc.). However, several other proteins in this list are also indirectly involved in DNA damage or repair processes especially those enriched at the top. A functional annotation table for significantly downregulated proteins is included as Supplementary Table S4. In Fig. 2B, selected up- or downregulated proteins (levels were changed either significantly or insignificantly in TMT-LC/MS-MS) are illustrated via a heatmap. These include proteins associated with DNA repairs such as ATR, ATM, DNA-PK, and RAD50, cell cycle (e.g., cyclin B), survival/proliferation (e.g., EGFR and cMet), and death (caspase 3 and LC3B I/II). The effect of the low concentration of AT13387 on these proteins in UMSCC74B, SCCVII, and MiaPaCa-2 cell lines was validated by immunoblotting (Fig. 2C, i; Supplementary Fig. S2) and was compared with that of an immortal normal-tissue cell line, keratinocytes (Fig. 2C, ii).
At this low concentration of AT13387, chaperone function of Hsp90 was affected. But the abundance of Hsp90 clients involved in the regulation of cell survival or cell-cycle progression such as cyclin B, AKT, cyclin A2, cyclin B1, Cdk2, or Cdk7 was either minimally or not affected. Also, proapoptotic markers such as caspase 3 or 7 both remained unchanged at this concentration, suggesting that 100 nmol/L AT13387 may not affect cell-cycle progression or cause apoptotic cell death. We anticipated that this concentration would radiosensitize cells due to its significant effect on several proteins involved in promoting DNA damage and repair (32).
A subtoxic concentration of AT13387 sensitizes HNSCC cells to irradiation
The effect of AT13387 and 17-AAG on the in vitro radiation sensitivity was assessed. The effect of AT13387 on the enhancement ratio in eight HNSCC cell lines and three pancreatic cell lines is tabulated in Fig. 3A, and the radiosensitizing effect of 17-AAG in UMSCC74B, SCCVII, and MiaPaCa-2 is tabulated in Supplementary Fig. S3C. AT13387 treatment (100 nmol/L) caused radiosensitization in all 11 cell lines used in this study, irrespective of the presence of multiple driver mutations. The data shown in Supplementary Fig. S3A and S3B suggest that 100 nmol/L AT13387 promotes better radiosensitization than 30 nmol/L. A 24-hour exposure of 30 or 100 nmol/L AT13387 resulted in more than 80% and 70% survival of immortalized keratinocytes, respectively, indicating the modest effect of the drug as a single agent (Supplementary Fig. S9), demonstrating a relative lack of toxicity. Although 100 nmol/L concentration of both AT13387 and 17-AAG showed radiosensitization effect in all tested cancer cell lines, 17-AAG was cytotoxic in normal-tissue keratinocytes. Particularly, UMSCC74B, UMSCC10B, and UMSCC1 showed enhancement ratios of 2.87 ± 0.52, 2.3 ± 0.48, and 1.8 ± 0.24, respectively. A moderate sensitizing effect of AT13387 was observed in SCCVII, UMSCC29, UMSCC92, and UMSCC47 (enhancement ratios of 1.7 ± 0.3, 1.7 ± 0.42, 1.67 ± 0.37, and 1.4 ± 0.22, respectively; Fig. 3A). Line plots demonstrate the survival fraction of UMSCC74B cells (Fig. 3B) and SCCVII cells (Fig. 3C) following increasing doses of irradiation alone or in combination with 100 nmol/L AT13387. Altogether, the data suggest that a subtoxic concentration of AT13387 can cause radiosensitization regardless of the specific oncogenic driver(s) present in the cell lines.
Combination of AT13387 and irradiation results in enhanced DNA damage (γH2Ax foci formation) and delay in repair (γH2Ax foci clearance)
Upon determination that a subtoxic concentration of AT13387 radiosensitizes HNSCC and pancreatic cancer cell lines, we assessed whether radiation-induced DNA damage repair was affected by a subtoxic concentration of AT13387. This was done by evaluating the γH2Ax foci formation, a biomarker for DNA damage, using an immunofluorescence assay. Representative images were captured at 6 hours following irradiation (or 30 hours following AT13387 treatment). In both the UMSCC74B and SCCVII cells, we observed a significant enhancement in γH2Ax foci formation in the cells treated with AT13387 before irradiation (Fig. 4A and B), which was maintained for up to 6 hours after radiation (Fig. 4C), but not at 24 hours (Supplementary Fig. S10). The expected mean number of γH2Ax foci per cell (Fig. 4D) was approximately 11.4 [P = 0.009; 95% confidence interval (CI), 3.3–19.5] and 13.8 (P = 0.016; 95% CI, 2.9–24.6) higher in the combination treatment group for SCCVII and UMSCC74B, respectively, when compared with the AT13387-only treatment group. The expected mean number of γH2Ax foci per cell was approximately 6.3 (P = 0.113; 95% CI, 1.7–14.4) and 4.3 (P = 0.410; 95% CI, 6.6–15.1) higher in the combination treatment group for SCCVII and UMSCC74B, respectively, when compared with the radiation-only treatment group. Immunoblotting data also confirmed a two- to threefold increase in γH2Ax induction in the combination treatment group and a longer time to resolve the damage (Supplementary Fig. S4). These results suggested that exposure to subtoxic concentration of AT13387 promoted greater DNA damage and delayed the irradiation-induced repair response.
Another explanation for inhibition of the DNA repair process could be cell-cycle redistribution as inhibition of Hsp90 is known to cause either G1-phase or G2–M-phase cell-cycle arrest. The mass spectrometry data suggested an insignificant or minimal effect on the proteins associated with cell-cycle regulation following a subtoxic exposure to AT13387. Nevertheless, we investigated the effect of 100 nmol/L AT13387 on the cell-cycle distribution by assessing DNA content using flow cytometry. UMSCC74B (wild-type TP53), SCCVII, and MiaPaCa-2 (mutant TP53) cells were treated with either 100 nmol/L AT13387 or vehicle, and cell-cycle analysis was performed at 24 hours using flow cytometry. The UMSCC74B cell line showed an increase in the G1-phase population (11%; P < 0.05), whereas a minimal change in the G1-phase population was observed in the SCC VII and MiaPaCa-2 cells (P > 0.05). This difference could be due to the difference in TP53 status. However, this treatment had minimal effect on the G2–M-phase populations in all the cell lines, which is the most radiosensitive phase of the cell cycle, suggesting that the radiosensitization is largely independent of cell-cycle redistribution by low concentration of AT13387 treatment (Supplementary Fig. S5).
Pharmacokinetics and biomarker studies to optimize dose for combination studies
These in vitro data suggested that a subtoxic concentration of AT13387 can cause significant radiosensitization of cancer cells. To translate these findings for animal studies, we wanted to determine a safe dose of AT13387 that can also be monitored using upregulation in the Hsp70 level, which is an established biomarker of Hsp90 inhibition (26). Therefore, we conducted pharmacokinetics and biomarker studies to evaluate whether a lower than MTD of AT13387 can reach xenograft tumors and whether such a dose can inhibit Hsp90 activity as monitored by upregulation of Hsp70 protein levels in the tumor cells. We also evaluated whether the effect of such a low concentration on Hsp90 activity can be monitored by Hsp70 upregulation in the PBMCs.
We began by assessing the pharmacokinetic effects of 40 mg/kg AT13387, given as either single or double intraperitoneal injections administered at 0 and 96 hours, in plasma and UMSCC74B xenograft tumors. About half of the MTD (i.e., 90 mg/kg) was selected in this study to determine whether the accumulation of AT13387 in the tumor would be enough for radiosensitization. The resulting concentrations of AT13387 in the plasma and tumor were determined at 0, 0.5, 2, 24, 96, 98, 120, and 196 hours after injection. A three-compartment model with first-order rate of absorption from the peritoneum into the plasma with transfer rate constants from the plasma to tumor provided an excellent fit [R2 = 0.998 (plasma) and R2 = 0.994 (tumor)] to the individual observed data (Supplementary Table S3). The median transfer rate constant from plasma to the tumor was twofold higher than that from tumor to plasma implying tumor retention. The model predicted median (min, max) Cmax in plasma (Tmax = 0.4 hour) and tumor (Tmax = 3.2 hour) were 299 (27.6, 354.5) ng/mL and 3,003 (1,426, 5,301) ng/g, respectively. Similarly, the tumor median (min, max) AUC0–infinity was 90,733 (24,336, 168,448) ng*hour/g, which was at least 40-fold more than the plasma median (min, max) AUC0–infinity, 2,263 (168.8, 31,787) ng*hour/mL (Fig. 5A; Supplementary Table S3). The results show that the xenograft tumor exposures achieved even at less than half the MTD dose of AT13387 were higher than that required for radiosensitization and that AT13387 was cleared rapidly from the plasma and retained for a longer duration in the tumor as shown by other groups.
Following the determination of the pharmacokinetics profile of AT13387, we sought to determine the duration for which AT13387 can exert its effects in vivo. To investigate these effects, a single 40 mg/kg dose of AT13387 was administered at 0 hours, and a second 40 mg/kg dose at 96 hours, to mice bearing UMSCC74B xenografts. The expression of Hsp70 was analyzed in the tumor at the various time points as described previously. Following administration, pharmacodynamics studies showed that the expression of chaperone protein Hsp70, our biomarker for Hsp90 activity, was upregulated after AT13387 treatment (Fig. 5B). This result was confirmed by densitometry analysis, which showed an eightfold upregulation of Hsp70 expression at 24 and 120 hours (Fig. 5B). Nearly two- to fivefold increased Hsp70 expression was demonstrated using IHC (Fig. 5C). This upregulation suggests that AT13387 treatment showed the desired inhibitory effect of Hsp90 inhibition in vivo.
Determination of Hsp90 inhibition in PBMCs
In this study, we were also interested in establishing a biomarker to quantify Hsp90 inhibition by a low dose of inhibitor, which we expected to be nontoxic systemically. It has been observed that Hsp90 inhibition correlates with an increase in Hsp70. Using immunofluorescence staining (Fig. 5D) and immunoblotting (Supplementary Fig. S6), we determined whether a dose equaling one-third of MTD could enhance Hsp70 expression in mouse PBMCs. At least twofold induction of Hsp70 in PBMCs isolated from the blood of treated mice was observed. This induction in Hsp70 correlated well with the radiosensitizing concentration determined from in vitro experiments (Supplementary Fig. S1) and suggests that a less than half the MTD dose will be enough for inhibiting Hsp90 activity, followed by radiosensitization of HNSCC cell lines. It also suggests the potential usefulness of Hsp70 from blood cells as a clinical biomarker in future studies.
Preclinical assessment of radiosensitization by a subtoxic dose of AT13387 in a pancreatic cancer, a head and neck human, and syngeneic model
Preclinical experiments were conducted to determine the radiosensitizing effects of low-dose AT13387 against several xenograft models. On the basis of the pharmacokinetics/pharmacodynamics data, we further reduced the dose of AT13387 to 30 mg/kg. Mice were randomized into four groups and treated as described in the Materials and Methods sections.
As expected, we observed a significant effect on both the tumor growth rate (Fig. 6A, D, and G) and the tumor doubling time (Fig. 6B, E, H) in the combination treatment group. In the data displayed as TD2, we observed a minimal effect on tumor doubling time with low-dose AT13387 as monotherapy in all the three tumor models used in this study. Radiotherapy alone was minimally effective in UMSCC74B and SCCVII tumor models, but had a substantial effect in MiaPaCa-2 tumors. The AT13387 treatment as monotherapy or in combination with radiation was well-tolerated (Fig. 6C, F, and I).
To explore the mechanism of AT13387-mediated radiosensitization, tumors were harvested following either a single dose of 30 mg/kg AT13387 at 24 hours, 2 Gy fraction of radiotherapy at 6 hours, or a combination. A TMA was constructed and the change in DNA damage repair markers such as γH2Ax, pChk1/2, Rad51, pATM, and pATR was assessed by counting the percentage of positive cells in the five random fields. The statistical analysis was performed using one-way ANOVA. Representative images (Supplementary Figs. S7 and S8) suggested that a single low dose of AT13387 and radiation induced DNA damage at varying degrees across the xenografts. The increase in γH2Ax levels showed a trend in the combination treatment more than the single treatment condition across the tumor types. In the SCCVII syngeneic tumor model, this increase was statistically significant. A similar pattern (nonsignificant) was observed in the pChk1/2, pATM, and pATR levels in the combination treatment group (Supplementary Fig. S8A–S8D).
Discussion
In this study, we evaluated a combination treatment regimen of AT13387 at a reduced dose to minimize potential off-target effects and yet preserve its radiosensitizing properties in preclinical models. We have found that an equivalent concentration (3 μmol/L) to the plasma Cmax of AT13387 achieved upon an MTD dose to a human is nonselective in vitro in terms of its effect on Hsp90 client proteins. A substantial reduction in the in vitro concentration of AT13387 to 100 nmol/L revealed a relatively selective effect on a subset of Hsp90 clients that are involved in DNA repair, while not affecting most other Hsp90 client proteins.
On the basis of the effect of 100 nmol/L AT13387 treatment on DNA repair proteins detected by the global proteome analysis, this concentration was used for each in vitro experiment. This concentration inhibited the Hsp90 chaperone function, causing nearly twofold upregulation of Hsp70 protein levels. Although a minimum percent inhibition of the Hsp90 activity requirement for radiosensitization remains unknown, we found that 100 nmol/L AT13387 had minimal effect on the glucocorticoid receptor folding in vitro (Supplementary Fig. S11). Because of these small effects on clients, 100 nmol/L AT13387 had minimal effect on the antiproliferative or proapoptotic pathways, but reduced the levels of DNA repair proteins such as ATR, DNA-PK, and RAD51. These data led to an assessment of the radiosensitizing potential of AT13387 against a panel of cancer cell lines with varying genetic backgrounds. This reduced AT13387 concentration remained a potent radiosensitizer across the eight HNSCC and three pancreatic cancer cell lines used in this study. However, normal-tissue keratinocytes remained unaffected at this low concentration. These in vitro studies also suggest the primary mechanism of radiosensitization appears to be independent of cell cycle and specific genetic mutation. Observed radiosensitization was accompanied by degradation of several DNA repair client proteins of Hsp90, enhanced DNA damage, and delayed γH2Ax foci clearance in vitro.
This study also aimed to find a dose, schedule, and frequency of administration of AT13387 that reduces Hsp90 activity in tumors to a level that can cause radiosensitization without affecting normal tissue. To understand these parameters, detailed pharmacokinetics and pharmacodynamics studies were performed in mice bearing UMSCC74B xenografts. Hsp90 inhibitors are distributed throughout the body, but are retained for an extended period in the tumors when compared with plasma or normal tissues, primarily due to the highly activated state of Hsp90 in tumors. To translate our in vitro findings into in vivo efficacy studies, we first assessed the effect of a dose of AT13387 that led to comparable compensatory upregulation of Hsp70 levels in mouse PBMCs and tumor cells as observed in vitro. We found that a single injection of 40 mg/kg AT13387, which is less than half of the MTD dose, induced Hsp70 expression by about twofold in PBMCs within 2 hours and lasted up to 24 hours. Similarly, two- to fourfold induction of Hsp70 expression was also observed in the xenograft tumors. Results from this study suggest that the upregulation of Hsp70 expression is a qualitative rather than quantitative biomarker of Hsp90 inhibition. In a phase I study, about 1.5-fold increase in Hsp70 expression in serum, PBMCs, or tumor was observed. This increase did not correlate with an increase in Hsp90 inhibitor from 100 mg to 800 mg dose level (33). This in vivo change in Hsp70 expression is similar to in vitro studies conducted at the low concentration (100 nmol/L) of Hsp90 inhibitor suggesting that this concentration of AT13387 inhibits Hsp90 and is expected to be relatively safe to normal tissues. Here, we also determined the pharmacokinetics profile of AT13387 injected once or twice 4 days apart at the reduced dose level. The resulting concentration of AT13387 in the plasma was compared with the xenograft tumors to determine tumor-specific accumulation, in addition to finding a potentially effective dose for radiosensitization. Further analysis of the pharmacokinetics parameters matched with each clinical observation even after using a low-dose of AT13387. Despite using a low dose, the inhibitor can accumulate in the xenograft tumors at least 10-fold above plasma levels. Drug levels in the tumors even at the half MTD dose were well over radiosensitizing concentrations. To our knowledge, this is the first study evaluating Hsp70 as an immediate, detectable, and qualitative biomarker for mice treated with a low dose of Hsp90 inhibitor.
Although Canella and colleagues demonstrated that long acting nature of AT13387 and its ability to cross the blood–brain barrier makes it effective as a single agent, several clinical studies supported the view that the treatment failures were due, in part, to the use of MTD (34). Available results indicated that the combination therapy with a much lower dose of Hsp90 inhibitor may be the future of cancer therapy. Our data also revealed a plasma AUC of 2,263 ng*hour/mL (5.52 μmol/L*hour) based on a 40 mg/kg dose in mice that confirms the allometric translatability of information in the mouse model to that in humans. On the other hand, we observed a 40-fold difference in the AUC of 90,733 ng*hour/mL (221.3 μmol/L*hour) in the UMSCC74B xenograft tumor. This accumulation of AT13387 in tumor inhibited Hsp90 activity as reflected by the sixfold upregulation of Hsp70 at 24 hours after injection. The observed concentration–time profile of AT13387 in this mouse model is consistent with the clinical trial experience with this agent. Previous dose-ranging studies in patients escalated the dose from 20 mg/m2 to 210 mg/m2 and included pharmacokinetics and safety assessments (35). These clinical data identified a linear plasma pharmacokinetics profile with an average AUC of 5.5–6 μmol/L*h in patients receiving 120 mg/m2, which was associated primarily with hematologic grade 1 adverse event. On the basis of in vivo modeling data, we estimated that a 30 mg/kg dose in mice would translate to doses of approximately 90 mg/m2 in patients, which is safe. This exploratory analysis supports the possibility of clinical benefits by combining low-dose HSP90 inhibitors with radiotherapy.
In an attempt to reduce the potential off-target effects while maintaining the ability of AT13387 to enhance radiosensitivity, the efficacy of AT13387 at a further reduced level (30 mg/kg) was determined against one pancreatic and two head and neck tumor models. As anticipated, treatment with a systemic lower dose of AT13387 was safe, but it was not effective as monotherapy against any of these tumor models. However, significant inhibition of tumor growth with acceptable toxicity was observed in combination with radiotherapy. This effect correlated with an increase in γH2Ax foci formation after the first cycle of combination therapy suggesting that inhibition of Hsp90 activity compromised DNA repair resulting in radiosensitization. Similar tumor radiosensitizing results were obtained using the MiaPaCa-2 xenograft model even at lower doses of radiation (1.8 Gy) with 30 mg/kg AT13387, which were ineffective as a monotherapy. Overall, these findings demonstrate that treatment with subcytotoxic concentrations of an Hsp90 inhibitor can selectively radiosensitize cancer cells in vitro and in xenograft models. These results suggest that treatment with Hsp90 inhibitors can have potential clinical benefits in combination with radiotherapy.
Clinical studies of Hsp90 inhibitors used as dose escalating monotherapies as well as in combination with chemotherapy have produced clinical responses, but, unfortunately, these responses have been accompanied by severe eye, liver, gastrointestinal, cardiac, or bone marrow toxicities. The toxicities of Hsp90 inhibitors may be made worse when chemotherapy was added to the treatment regimen (and vice versa; ref. 36). To enhance efficacy and reduce the toxicities observed in the combination of Hsp90 inhibitors with systemic chemotherapy, several preclinical or clinical studies have assessed the addition of other molecular targeted agents such as erlotinib, trastuzumab, and vemurafenib along with of Hsp90 inhibitors with or without chemotherapy. These efforts have also largely proven to be clinically ineffective, perhaps because Hsp90 inhibition targets the same clients (e.g., EGFR, Her2, and Braf) as that of the molecularly targeted agents. Therefore, clinical trials involving Hsp90 inhibitors have been either withdrawn before moving forward to phase III or failed to get FDA approval.
To the best of our knowledge, this is the first study that specifically demonstrates that subcytotoxic concentration of an Hsp90 inhibitor can enhance radiosensitization. Increased γH2Ax foci are often considered as a pharmacodynamics marker for radiation-induced DNA damage which correlates with radiosensitization (32). We also observed a trend of increased γH2Ax foci and other DNA repair markers across the tumor types in the combination group. After one cycle of treatment, the majority of γH2Ax foci were resolved, but the enhanced radiosensitization suggests that either the repair was “error prone” or that repeated treatment had a cumulative effect on the treatment outcome.
In addition to the direct effect on DNA repair, Hsp90 and its cochaperones such as Hsp70 also possess multiple immunomodulatory functions (37, 38). Inhibition of Hsp90 activity is known to cause immunosuppression in patients. A recent study found that nontoxic concentration of Hsp90 inhibition can trigger an antitumor immune response (39), which was absent at cytotoxic concentration. We noted that inhibition of Hsp90 activity by AT13387 treatment greatly radiosensitized SCCVII xenografts, developed in syngeneic C3H/HeJ mice background. Whereas AT13387-mediated radiosensitizing effect against UMSCC74B and MiaPaCa-2 xenografts, developed in immunocompromised nude mice, was less profound, indicating a potential role of the immune component in radiosensitization by AT13387. If these immunologic findings are maintained or potentiated by radiotherapy, it would further support the revival of Hsp90-targeting agents as potential clinical radiosensitizers.
Disclosure of Potential Conflicts of Interest
M.J. Schipper reports personal fees from Innovative Analytics outside the submitted work. Y. Osawa reports grants from NIH (noted in financial support) during the conduct of the study. M.K. Nyati reports grants from NIH, Rogel Cancer Center (support grant), and University of Michigan (Protein Folding Disease Research Initiative) during the conduct of the study; in addition, AT13387 was provided by the Cancer Therapy Evaluation Program (CTEP), NIH. No potential conflicts of interest were disclosed by the other authors.
Authors' Contributions
R.K. Mehta: Data curation, formal analysis, investigation. S. Pal: Data curation, software, formal analysis, validation, investigation, methodology, writing-original draft, writing-review and editing. K. Kondapi: Investigation. M. Sitto: Investigation. C. Dewar: Investigation. T. Devasia: Data curation, software, formal analysis. M.J. Schipper: Data curation, software, formal analysis. D.G. Thomas: Data curation, software, formal analysis, validation, investigation, visualization, methodology. V. Basrur: Data curation, visualization, methodology, writing-review and editing. M.P. Pai: Data curation, software, formal analysis, validation, writing-review and editing. Y. Morishima: Writing-review and editing. Y. Osawa: Writing-review and editing. W.B. Pratt: Writing-review and editing. T.S. Lawrence: Conceptualization, writing-review, and editing. M.K. Nyati: Conceptualization, resources, data curation, supervision, funding acquisition, project administration, writing-review and editing.
Acknowledgments
This article is dedicated to the memory of Dr. William B. Pratt.We thank the University of Michigan Pharmacokinetic, Proteomic, Flow cytometry, Histology, and DNA sequencing cores and animal facilities at the University of Michigan. AT13387 for this research was provided by the Cancer Therapy Evaluation Program at the NIH. We would also like to thank Dr. Meredith A Morgan for providing valuable inputs on the discussion section, Mary A Davis for editing, and Juan Cai for helping with the STR study. This work was supported by the NIH (R01 CA131290 to M.K. Nyati; U01CA216449 to T.S. Lawrence; and GM 077430 to Y. Osawa) and the Rogel Cancer Center support grant, P30CA46592. This project is a component of the Protein Folding Disease Research Initiative of the University of Michigan Medical School.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.