SNX2112, a Synthetic Heat Shock Protein 90 Inhibitor, Has Potent Antitumor Activity against HER Kinase–Dependent Cancers

Heat shock protein 90 (Hsp90) is a protein chaperone that functions to promote the maturation and conformational stabilization of a subset of cellular proteins important in transducing proliferative and survival signals. Hsp90 clients include protein kinases (e.g., HER2, Raf-1, Akt, and Cdk4), steroid receptors (e.g., androgen receptor and estrogen receptor), and transcription factors (e.g., Hif1α; refs. 1–7). A number of mutant oncoproteins also require Hsp90 function, including v-Src, mutant epidermal growth factor receptor (EGFR), and mutant B-Raf, whereas their wild-type counterparts are either not dependent or only weakly dependent on Hsp90 (8–12). Given the critical roles played by Hsp90 clients in tumor growth and maintenance, inhibition of Hsp90 has emerged as a possible strategy for the treatment of advanced cancers.

Several natural products, including the ansamycin geldanamycin, inhibit Hsp90 chaperone function by binding to an ATP pocket in the NH₂-terminal domain of the protein. Geldanamycin proved too toxic for human use, but a 17-carbon position derivative, 17-(allylamino)-17-demethoxygeldanmycin (17-AAG), is now being tested in ongoing phase 1 and 2 clinical trials. Although antitumor activity has been observed in early-stage clinical trials of 17-AAG, this agent is poorly soluble and has limited oral bioavailability. The poor solubility of 17-AAG has necessitated the use of DMSO and cremaphor-based formulations that likely contribute to the toxicities observed in the clinical trials of this agent. Furthermore, the requirement for i.v. dosing has also likely limited the efficacy of 17-AAG in patients by placing practical limitations on the schedules of administration that can be evaluated. Accumulating data with non-ansamycin Hsp90 inhibitors also suggests that the dose-limiting hepatotoxicity of 17-AAG may be in part “off target”, attributable to the chemical reactivity of its benzoquinone group and not a direct consequence of Hsp90 inhibition (13). For these reasons, orally bioavailable Hsp90 inhibitors that lack a quinone moiety may be more efficacious and less toxic than 17-AAG. Finally, expression of P-glycoprotein and loss or mutation of the NQO1 gene, which is required for the bioreduction of 17-AAG to the more potent hydroquinone 17-AAGH₂, have been proposed as mechanisms of de novo or acquired resistance to 17-AAG (14, 15). Therefore, Hsp90 inhibitors that are not substrates for P-glycoprotein and do not require NQO1 metabolism may be more effective clinical agents than 17-AAG.

To identify novel inhibitors of Hsp90, a compound library was screened against the purine-binding proteome to identify novel scaffolds that selectively bind to the ATP pocket of Hsp90. Specifically, a purine-based affinity resin was used to capture purine-binding proteins. Compounds that displaced Hsp90 family members from this column were then identified by mass spectrometry (MS) sequencing. Using this technology, SNX-2112 was identified as a compound that selectively binds to the ATP pocket of Hsp90 family members (Hsp90α, Hsp90β, Grp94, and Trap-1). The SNX-2112 scaffold is unrelated in structure to any of the natural product–based Hsp90 inhibitors (including the geldanamycins, radicicols, and macbesins) and to the purine-based PU series (16).

We now show that SNX-2112 displays the antitumor profile of the natural product Hsp90 inhibitors: degradation of Hsp90 clients including HER2, the Rb-dependent G₁ cell cycle arrest of cancer cells, and induction of morphologic differentiation of MCF-7 cells. HER2 degradation by SNX-2112 in HER2-dependent breast cancer cells resulted in potent inhibition of the Akt and extracellular signal-regulated kinase (Erk) pathways and inhibition of tumor cell proliferation both in vitro and in xenograft models. Furthermore, SNX-5542, a water-soluble and orally bioavailable prodrug of SNX-2112, displayed a favorable pharmacodynamic profile with a single oral dose administered to tumor-bearing mice, resulting in preferential tumor accumulation and greater inhibition of the Erk and Akt pathways in tumor compared with normal tissues. These effects were seen at nontoxic doses, which could be delivered chronically on a daily or five times per week schedule. These data suggest that SNX-2112 represents a novel inhibitor of Hsp90 with pharmacologic advantages over the natural product Hsp90 inhibitors and form the basis for the human clinical testing of this agent in patients with breast cancer and other advanced malignancies.

Chemicals. 17-AAG was obtained from the Drug Synthesis and Chemistry Branch, Developmental Therapeutics Program, National Cancer Institute, and was dissolved in DMSO to yield 50 mg/mL and 10 mmol/L stock solutions and stored at −80°C. SNX-2112 and SNX-5542 were obtained from Serenex, Inc. SNX-2112 was also dissolved in DMSO for in vitro studies, whereas SNX-5542 was formulated in dextrose 5% in water for in vivo studies.

Cell culture. BT-474, SKBr-3, SKOV-3, MCF-7, and MDA-468 were obtained from the American Type Culture Collection. Cells were maintained in DMEM-F12 medium supplemented with 100 units/mL penicillin, 100 mg/mL streptomycin, 4 mmol/L glutamine, and 10% heat-inactivated fetal bovine serum and incubated at 37°C in 5% CO₂. H1650 was also obtained from American Type Culture Collection and grown in RPMI supplemented with 1 mmol/L pyruvate, 1.5 g/L sodium bicarbonate, 4.5 g/L glucose supplemented with 100 units/mL penicillin, 100 mg/mL streptomycin, 4 mmol/L glutamine, and 10% heat-inactivated fetal bovine serum. Cell viability was determined by seeding 2,000 to 5,000 cells per well in 96-well plates and treating with drug 24 h after plating in complete medium (200 μL). Each drug concentration was tested in eight wells. Cells were assayed using the Alamar blue viability test after a 96-h incubation. Flow cytometry was done using nuclei stained with ethidium bromide and isolated via the Nusse protocol (17).

Animal studies. Four- to 6-week-old nu/nu athymic BALB/c female mice were obtained from the National Cancer Institute-Frederick Cancer Center and maintained in pressurized ventilated caging at the Sloan-Kettering Institute. All studies were done in compliance with Institutional Animal Care and Use Committee guidelines. Before BT-474 cell inoculation, 0.72 mg sustained release 17β-estradiol pellets were placed s.c. with a 10 g trocar. Tumors were established by injecting 1 × 10⁷ cells suspended 1:1 (volume) with reconstituted basement membrane (Matrigel, Collaborative Research). For efficacy studies, mice with established tumors were selected. Fourteen days after inoculation, mice were treated with SNX-5542 using the indicated doses. Tumor dimensions were measured with vernier calipers and tumor volumes were calculated using the formula π/6 × larger diameter × (smaller diameter)². For pharmacodynamic studies, mice with well-established tumors were treated with SNX-5542 and sacrificed pretreatment, 3, 6, 10, 24, and 48 h posttreatment (two mice per time point).

Immunoblotting. Lysates were prepared by homogenizing tumors in SDS lysis buffer [50 mmol/L Tris-HCl (pH 7.4), 2% SDS], boiling for 10 min, followed by brief sonication. Lysates were then cleared by centrifugation at 14,000 × g for 10 min, and the supernatant was collected. Lysates from cells in culture were prepared by washing twice in cold PBS followed by lysis with NP40 lysis buffer (50 mmol/L Tris-HCl; 1% NP40; 40 mmol/L NaF; 150 mmol/L NaCl; 10 μmol/L/mL Na₃VO₄/phenylmethylsulfonyl fluoride/DTT; and 1 mg/mL leupeptin, aprotinin, and trypsin inhibitor). Protein concentration of each sample was determined using the BCA kit (Pierce Chemical) as per the manufacturer's instructions. Twenty-five or 50 μg of protein were resolved by SDS-PAGE and transferred onto nitrocellulose membranes. Blots were probed overnight at 4°C with primary antibodies (all from Cell Signaling except the following: HER2, Upstate Biotechnology; phosphatidylinositol 3-kinase–p85, Upstate Biotechnology; cyclin D1, Santa Cruz Biotechnology; P-HER2, Upstate Biotechnology) to detect proteins of interests. After incubation with horseradish peroxidase–conjugated secondary antibodies, proteins were visualized by chemiluminescence (ECL, Amersham Corp.).

Tissue distribution. Studies on the distribution of SNX-2112 were done on 40 to 100 mg of tissue flash frozen in liquid nitrogen at the indicated time points. Tissues were homogenized using the Bio-Plex cell lysis kit, and lysates were prepared using the Bio-Plex phosphoprotein detection reagent kit according to the manufacturer's instruction. Bead sets tested included the phospho-Akt (S473) and phospho-Erk1/2 (R202/Y204, R185/Y187). Extracts were prepared with a glass Dounce homogenizer in 100% acetonitrile containing an internal standard. Samples were analyzed by liquid chromatography–tandem MS using a Shimadzu high-performance liquid chromatography and an Applied Biosystems 4000 Q Trap.

ATP displacement assay. For the protein affinity–displacement assay, a purine-based affinity resin was generated by incubating ATP-linked Sepharose with Jurkat cell lysate (flash frozen and homogenized in saline) at 4°C (18). This was then incubated with test compounds (e.g., SNX-2112 or 17-AAG) for 90 min. Proteins eluted by drug were then resolved by SDS-PAGE, visualized with silver staining, and excised from the gel for MS-based identification. Briefly, after destaining and trypsin digestion (described elsewhere), peptides were extracted with μC18 ZipTips (Millipore) and then eluted and spotted directly to a conventional stainless steel matrix-assisted laser desorption/ionization target with a saturated solution of α-cyano-4-hydroxycinnamic acid (Sigma) in 50% acetonitrile (VWR), 0.15% formic acid (Sigma; refs. 19, 20). Mass spectra were then acquired using a MALDI-TOF/TOF 4700 Proteomics Analyzer (Applied Biosystems). MS spectra were acquired (1,000 shots per spectrum), and the three peaks from each with the greatest signal-to-noise ratio were automatically submitted for tandem MS analysis (3000 shots per spectrum). The collision energy was 1 keV. Air was used as the collision gas. Protein identification was done from the MS and tandem MS data using GPS Explorer software (Applied Biosystems) with the integrated Mascot database search engine.

Effect of SNX-2112 on Hsp90 client proteins. The physiochemical properties of the ansamycins responsible for their deficiencies as drugs have led to widespread efforts to develop soluble inhibitors of Hsp90 that are orally bioavailable and lack the benzoquinone scaffold of the geldanamycin derivatives. SNX-2112 was identified using a protein affinity–displacement assay that used a purine-based affinity resin to capture purine-binding proteins derived from cell lysates. Screening hits were those compounds that eluted one or more proteins in a concentration-dependent manner from the resin. The primary screen thus provided data on both target affinity and selectivity for proteins within the purine-binding proteome (21, 22). Using this screen, SNX-2112 was observed to bind to Hsp90α and Hsp90β with low nanomolar affinity. With the exception of the Hsp90 family members Grp94 and Trap-1, SNX-2112 did not displace any of the >2,000 other purine-binding proteins included within the screen. Using this screen, SNX-2112 was observed to bind to Hsp90α and Hsp90β with a K_a of 30 nmol/L, compared with 88 nmol/L for geldanamycin and 1,039 nmol/L for 17-AAG (Table 1). Moreover, several analogues of SNX-2112 were shown by X-ray crystallography to bind the amino-terminal ATP site of Hsp90 (data not shown).

Exposure of tumor cells to the natural product inhibitors of Hsp90, including geldanamycin and radicicol, causes degradation of a distinct subset of client proteins, Rb-dependent G₁ arrest, and, in Rb-deficient cells, mitotic block and apoptosis (23). We therefore evaluated whether SNX-2112 could phenocopy these effects of the natural product Hsp90 inhibitors and whether it had other effects that could not be attributed to Hsp90 inhibition. One consequence of Hsp90 inhibition in cancer cells is degradation of Hsp90 client proteins, including HER2 and Akt. We therefore compared the change in HER2 and Akt expression in BT-474 cells in vitro after exposure to 17-AAG or SNX-2112. Treatment of cancer cell lines with 1 μmol/L 17-AAG or SNX-2112 resulted in down-regulation of HER2 expression within 3 to 6 h of drug exposure with near-complete loss of HER2 expression by 10 h (Fig. 1A). Treatment with SNX-2112 also resulted in a decline in total Akt expression; however, as observed with 17-AAG, the kinetics of Akt degradation differed from that of HER2 (maximal effect noted after 24-48 h versus 10 h for HER2). The expression of Erk and p85 phosphatidylinositol 3-kinase, two proteins that do not require Hsp90 for stability, were unaffected by either SNX-2112 or 17-AAG. 17-AAG and SNX-2112 displayed similar potencies in BT-474 cells with down-regulation of HER2 observed with both compounds between 25 and 100 nmol/L (Fig. 1B).

Both SNX-2112 and 17-AAG inhibited phosphatidylinositol 3-kinase/Akt and Raf/Mek/Erk signaling as assayed by loss of the phosphorylated forms of Akt (Ser-473) and Erk, respectively. Consistent with our prior studies of 17-AAG, Akt inhibition was observed before loss of Akt protein expression. In breast cancer cells with HER2 amplification, inhibition of HER2-dependent activation of phosphatidylinositol 3-kinase/Akt signaling results in loss of D-cyclin expression and induction of an apoptotic program. Both 17-AAG and SNX-2112 induced these effects at comparable doses and with similar kinetics (Fig. 1).

Effect of SNX-2112 on proliferation. Several groups, including our own, have observed that breast cancer cell lines with HER2 amplification are more sensitive to 17-AAG than cell lines with low levels of HER2 expression (24). We therefore assessed the effects of 17-AAG and SNX-2112 on tumor cells with variable levels of HER2 expression using a panel of breast, lung, and ovarian cancer cell lines. In all cell lines tested, SNX-2112 inhibited cell proliferation with IC₅₀ values ranging from 10 to 50 nmol/L. In contrast to 17-AAG, the sensitivity of cancer cell lines to SNX-2112 in vitro did not correlate with the level of HER2 expression. Similarly, SNX-2112 sensitivity in vitro did not correlate with the expression or mutational status of ER, PTEN, or PIK3CA (Fig. 2A).

In BT-474 cells (HER2 amplified, breast cancer), the antiproliferative effects of SNX-2112 and 17-AAG were associated with hypophosphorylation of Rb, arrest in G₁, and modest levels of apoptosis. Both 17-AAG and SNX-2112 caused these effects on the cell cycle progression of HER2-amplified, Rb wild-type cells at a comparable drug concentration (Fig. 2B). We have previously shown that Hsp90 inhibitors do not induce a G₁ block in tumor cells in which the Rb protein is mutationally inactivated (23). Instead, these cells undergo mitotic arrest, with attendant massive apoptosis. As shown in Fig. 2B, SNX-2112 caused a profound mitotic block in MDA-468 (Rb-negative) cells. Previous work showed that geldanamycin, herbimycin A, radicicol, and the synthetic purine-like Hsp90 inhibitor PU24FCl also caused mitotic arrest in MDA-468 cells as well as other tumor cells with mutant Rb (23, 25). Thus, the cellular effects of SNX-2112 in these models are equivalent to those of other, well-characterized Hsp90 inhibitors. Notably, whereas all other Hsp90 inhibitors effectively block MDA-468 cells in mitosis, 17-AAG does not effectively induce a mitotic block at the concentrations studied (Fig. 2B). It has been proposed that reduction of 17-AAG to 17-AAGH₂ is required for maximal activity with 17-AAG, and 17-AAG lacks activity in MDA-468 because this model is deficient in DT-diaphorase function (14). Our data thus suggest that in contrast to 17-AAG, SNX-2112 retains activity in NQO1-deficient cells. In summary, the effects of SNX-2112 on the expression of Hsp90 client proteins, tumor cell proliferation, and cell cycle progression are consistent with those of the natural product Hsp90 inhibitors.

Pharmacodynamics of Hsp90 inhibition. Despite its drawbacks, 17-AAG effectively induces the degradation of Hsp90 client proteins in human tumor xenografts and intermittent administration has significant antitumor activity in murine models of HER2-dependent breast cancer (26, 27). SNX-5542 is a prodrug of the active metabolite, SNX-2112, with improved solubility and greater oral bioavailability. As shown in Fig. 3, oral administration of a single dose of SNX-5542 to mice bearing BT-474 resulted in down-regulation of HER2 expression 6 to 24 h after drug exposure. Concordantly, there was loss of the phosphorylated forms of Akt and Erk as well as cyclin D1 expression. During the interval of Akt and Erk inhibition, an increase in the cleaved form of poly(ADP)ribose polymerase (c-PARP) was noted, indicative of apoptosis. These effects were dose dependent as lower doses (e.g., 25 mg/kg) showed a less profound effect on client protein expression. A maximal biological effect, however, was seen at 75 mg/kg (data not shown). In mice treated with a single dose of SNX-5542 up to 150 mg/kg, no gross toxicity was evident. These data confirm that tolerable doses of SNX-5542 can inhibit Hsp90 function and induces degradation of the HER2 client protein in tumors in vivo.

Antitumor effects of SNX-5542. The antitumor properties of SNX-5542 were evaluated using a variety of doses and schedules in the BT-474 breast cancer xenograft model. Treatment of mice with SNX-5542 doses of 100 mg/kg or greater (150 mg/kg Monday-Wednesday-Friday, 200 mg/kg Monday-Wednesday-Friday, and 100 mg/kg Monday to Friday) resulted in complete tumor growth inhibition and in some mice partial tumor regressions (data not shown). However, chronic treatment of mice with SNX-5542 at doses of 100 mg/kg or greater was associated with an average weight loss of between 5% and 15% with several animal deaths occurring after 1 week of treatment. SNX-5542, at a dose of 50 mg/kg, on a daily schedule, also effectively induced client degradation and inhibited Akt and Erk pathway activity without associated weight loss or other gross toxicities, and therefore this dose and schedule was chosen for further evaluation. SNX-5542 at a dose of 50 mg/kg (five times per week) caused complete growth inhibition and partial regression of BT-474 xenografts (Fig. 4A). Tumor growth inhibition was durable in this model with no evidence of tumor regrowth noted 5 weeks after the last dose on day 35. We also tested the 50 mg/kg dose using a daily schedule (seven times per week) in mice bearing BT-474 xenografts and observed nearly identical results (data not shown). In summary, Hsp90 inhibition with SNX-5542 has significant antitumor activity in a HER2-driven breast cancer model, and daily administration was feasible in mice without apparent toxicity. This contrasts with 17-AAG where continuous dosing schedules for longer than 5 days proved toxic in mice (5).

In human clinical trials, the frequency with which 17-AAG has been administered has been limited by both toxicity and the practical limitations imposed by its i.v. formulation. In patients, weekly dosing of 17-AAG is likely sufficient to down-regulate HER2 expression enough to confer antitumor activity in patients, but intermittent dosing is likely much less effective in tumors driven by clients that are less sensitive to the drug. Because SNX-5542 can be administered daily, its effects on tumor growth were evaluated in H1650 xenografts, which express a mutant EGFR and are PTEN deficient. Activating mutants of EGFR are selectively sensitive to Hsp90 inhibitors compared with wild-type EGFR (10, 28). H1650 contains a copy of EGFR with a deletion between amino acids 746 and 750 but is insensitive to EGFR kinase inhibitors likely due to concomitant loss of PTEN expression in these cells. SNX-5542 administered five times per week had significant antitumor activity in mice with established H1650 xenografts (Fig. 4B). As 17-AAG is not well tolerated by mice when given on a daily or five times per week schedule, a dose of 17-AAG (100 mg/kg) that resulted in Hsp90 client degradation in vivo was given three times per week to mice bearing H1650 tumors. On this schedule, 17-AAG administration resulted in an ∼40% inhibition of tumor growth after 5 weeks of treatment compared with a more than 70% reduction with SNX-5542.

Tumor selectivity of Hsp90 inhibition. The basis for the therapeutic index of Hsp90 inhibitors has not yet been fully elucidated. Previous work has suggested that one basis for the selective sensitivity of cancer cells to Hsp90 inhibition may be a preferential accumulation of 17-AAG and other Hsp90 inhibitors in tumor tissues (29). To determine whether selective tumor uptake was also a property of SNX-5542, its tissue distribution after oral administration was assessed in nude mice bearing established BT-474 tumor xenografts. Figure 5 shows the biodistribution of SNX-2112 across a number of tissues in mice treated with a single oral dose (75 mg/kg) of SNX-5542. Of note, SNX-5542 is rapidly converted in vivo to SNX-2112, and measurable levels of SNX-5542 could therefore not be reliably detected in the serum or tumor. MS of homogenized tissues revealed preferential accumulation of the active metabolite, SNX-2112, in tumor tissues, particularly at the 24 and 48 h time points. For instance, at 24 h, there was a >10-fold excess of drug found in tumor tissue (5 μmol/L) compared with that in lung, small intestine, liver, skin, uterus, and kidney. The drug concentrations in muscle, brain, and heart were negligible (<100 nmol/L) at these time points. The latter is notable as cardiac toxicity has been a theoretical concern with this class of agents given the role played by chaperones in the maturation of the cardiac potassium channel HERG (30).

SNX-2112 concentrations in tumor tissue and liver were also correlated with the expression of Hsp90 clients in these tissues. As expected, degradation of HER2 and consequent inhibition of the Akt and Erk pathways (as assessed by measuring the phosphorylated forms of Akt and Erk) was observed between 3 and 24 h (data not shown). However, HER2 expression rose to baseline levels 48 h after dosing, despite persistently high levels of SNX-2112 (1.8 μmol/L) in the tumor. These data suggest that intratumoral drug levels correlate poorly with the level of Hsp90 inhibition in tumors. Such a result could be explained by intracellular compartmentalization of the drug, synthesis of new Hsp90, or shifting of Hsp90 from a low-affinity “latent” reservoir to a high-affinity, cochaperone-bound pool. The binding kinetics of SNX-2112 and 17-AAG (i.e., slow off rate) may contribute to this phenomenon. To determine whether the differential accumulation of drug in tumor versus normal tissues could explain, at least in part, the selective sensitivity of tumor versus normal cells to Hsp90 inhibitors, the effects of the drug on Akt and Erk pathway activity in liver and tumor tissues were examined. A single dose of SNX-5542 was administered, and lysates from liver and tumor tissue were prepared at various times after treatment and assessed for expression of phosphorylated Akt (Ser-473) or Erk by quantitative ELISA. The phosphorylated forms of Akt and Erk showed a precipitous decline in tumor tissue after SNX-5542 treatment, which lasted ∼24 h. In contrast, SNX-5542 treatment had only a minimal effect on the activity of these pathways in liver. We speculate that the lack of effect on Akt and Erk pathway activity in liver versus tumor cells is attributable to the presence of a highly sensitivity Hsp90 client (HER2) in BT-474 tumors, which is absent in liver.

Hsp90 is a protein chaperone that promotes the maturation and conformational stabilization of a subset of cellular proteins important in transducing proliferation and survival signals. Hsp90 clients include protein kinases (e.g., HER2, Raf-1, Akt, and Cdk4), steroid receptors (e.g., androgen receptor and estrogen receptor), and transcription factors (e.g., Hif1α; refs. 1–7). Given the critical roles played by these Hsp90 clients in tumor growth and maintenance, inhibition of Hsp90 has emerged as a possible strategy for the treatment of advanced cancers. Several mutant oncoproteins, including v-Src, mutant EGFR, and mutant B-Raf, are also Hsp90 clients whereas their wild-type counterparts are either not dependent or only weakly dependent on Hsp90 chaperone function (8–12). The dependence of these gain-of-function mutants on Hsp90 suggests that Hsp90 may be permissive for the development of tumors that express these oncogenes.

In this study, we characterized the antitumor effects of SNX-2112, a novel compound that binds selectively to the NH₂-terminal ATP pocket of Hsp90. The SNX-2112 scaffold was identified by screening the purine-binding proteome for nonquinone- and nonpurine-containing scaffolds that bind selectively to Hsp90. This compound is pan-selective for the Hsp90 family in that it binds to Hsp90α, Hsp90β, Grp94, and Trap-1. To determine whether binding of SNX-2112 to Hsp90 resulted in inhibition of Hsp90 chaperone activity, we compared the effects of SNX-2112 to those of the geldanamycin derivative 17-AAG using a panel of breast, ovarian, and lung cancer cell lines. We found that SNX-2112 potently down-regulated HER2 expression and inhibited Akt and Erk pathway activity in breast cancer cells with HER2 amplification. These effects occurred with a kinetics and potency similar to that of 17-AAG. We further showed that treatment of breast cancer cells in vitro with SNX-2112, like 17-AAG, resulted in marked growth inhibition and other hallmarks of the natural product Hsp90 inhibitors, including a Rb-dependent G₁ growth arrest and morphologic differentiation in selected models. These data suggest that the drugs have the same target activity (antagonizing Hsp90 activity) and comparable antitumor activities in vitro.

One notable exception, however, was the breast cancer cell line MDA-468. In this model, SNX-2112 was markedly more potent than 17-AAG. Previous work has shown that 17-AAG is metabolized by DT-diaphorase to the more potent hydroquinone 17-AAGH₂ (14, 15). MDA-468 cells are resistant to 17-AAG because the gene encoding for this activity, NQO1, is mutated in MDA-468 cells. Transfection of NQO1 into MDA-468 cells, however, restores sensitivity of this model to 17-AAG, confirming that loss of DT-diaphorase expression can confer 17-AAG resistance (15). Our data thus suggest that SNX-2112 activity is independent of NQO1 activity and that SNX-2112 may therefore have a broader spectrum of antitumor activity than 17-AAG. Given that the purine-based synthetic Hsp90 inhibitor PU24FCl has a similar activity profile to SNX-2112 in these cells, we hypothesize that the dependence of MDA-468 cells on Hsp90 function is not accurately reflected by the lack of activity of 17-AAG in this model.

SNX-2112 has variable oral bioavailability and therefore with the goal of testing the utility of this compound in mice, several prodrugs of SNX-2112 were developed that display improved solubility and pharmacologic properties. We show that a single dose of SNX-5542, a water-soluble prodrug of SNX-2112, was sufficient to induce the degradation of HER2 in tumor-bearing xenografts. Following oral administration of SNX-5542, the drug was rapidly converted to SNX-2112 where it preferentially accumulated in tumor tissues. Notably, recovery of HER2 expression and the activity of its downstream effector pathways were observed at late time points (24 and 48 h) despite the continued presence in the tumor of SNX-2112 concentrations greater than those necessary to inhibit Hsp90 in vitro. We speculate that this may be due to either intracellular compartmentalization of the drug or induction of Hsp70 expression. In experiments where we rechallenged with daily doses of SNX-5542 in vivo, we did observe similar kinetics of client degradation and signal deactivation after a third consecutive dose. Thus, at least for three consecutive doses, we did not find significant tachyphylaxis to the effects of SNX-2112.

In mice with established BT474 (HER2-amplified) xenografts, daily oral administration of SNX-5542 resulted in partial tumor regressions. These data were comparable if not superior to the effects of i.p. administration of 17-AAG in this model system using intermittent dosing schedules (three times per week or days 1-5 every 2 weeks). In prior studies of 17-AAG, daily dosing was not, however, feasible in either mice or in patients due to hepatotoxicity (31–35). For this reason, 17-AAG is currently being tested in phase 2 trials using only intermittent dosing schedules: either weekly or days 1, 4, 8, and 11 every 21 days. Notably, in a pilot dog study, no significant hepatotoxicity was observed after 13 days of twice-daily dosing of SNX-5542 at the highest dose tested, 10 mg/kg. These data suggest that non-ansamycin Hsp90 inhibitors such as SNX-2112 may have a more favorable toxicity profile than 17-AAG, although human clinical trials will be necessary to test this hypothesis.

In addition to 17-AAG, several novel ansamycins are now in clinical development. These include the water-soluble and orally bioavailable geldanamycin derivative 17-(dimethylaminoethylamino)-17-demethoxygeldanamycin and IPI-504, a prodrug of 17-AAG with improved solubility and oral bioavailability (36, 37). Although these agents have superior pharmacologic profiles to 17-AAG in terms of solubility and oral bioavailability, they also contain the quinone species found in 17-AAG and would therefore be predicted to retain the hepatotoxicity characteristic of this class of agents. As a nonquinone-based Hsp90 inhibitor, SNX-2112 may therefore have potential toxicologic advantages over these geldanamycin derivatives.

Finally, it is important to note that Hsp90 inhibition has shown provocative activity in a variety of cancer types. In this report, we show that SNX-5542 has activity in a model of EGFR mutant non–small cell lung cancer. We find that SNX-5542 is superior to 17-AAG in this model. In contrast to the effects of SNX-5542 in mice with established BT-474 tumors, monotherapy with SNX-5542 was insufficient to induce complete growth inhibition in mice with established H1650 xenografts. It is unclear if the greater resistance of this tumor to SNX-5542 relates to intrinsic properties of the tumor model or deficiencies in target inhibition in this system. For example, H1650 cells contain not only an EGFR deletion mutant but also are PTEN deficient. Although tumor regression was not observed in this model with SNX-5542 alone, we have recently shown that the combination of 17-AAG and paclitaxel is synergistic in this model. Therefore, despite its advantages over 17-AAG, the use of combination strategies will likely still prove necessary in some systems despite the presence of a sensitive Hsp90 client oncoprotein. Nevertheless, given the potential advantages of the small-molecule platform, these studies underscore the impetus for the clinical testing of SNX-5542, an Hsp90 inhibitor with superior pharmacologic properties.