Introduction

The 90 kDa heat-shock protein (HSP90) is a member of a class of evolutionarily conserved molecular chaperones that have an integral role in the maintenance of protein homeostasis within cells. These chaperones assist in the folding, stabilization, activation, and proteolytic turnover of newly synthesized “client” proteins [1,2,3]. Among these client proteins are those involved in aberrant cellular functions responsible for the hallmarks of cancer, including angiogenesis and glucose metabolism [4, 5], growth factor independence [2], cell cycle progression [6], tissue invasion and metastases [7], avoidance of apoptosis [8], and acquired drug resistance [9]. By protecting the aberrant survival of these proteins, HSP90 enables tumor progression and resistance to therapies. Inhibition of HSP90 accordingly results in the degradation of these aberrant proteins through the ubiquitin–proteasome pathway, resulting in tumor cell death [10].

Upregulation of the heat-shock response, in particular HSP70 and HSP27, occurs in response to HSP90 inhibition, and is postulated as a biomarker of HSP90 inhibition and target engagement [11]. Increasing evidence implicates this compensatory upregulation of HSP70 as a resistance mechanism to HSP90 inhibition based on the anti-apoptotic activity of HSP70 at the pre-mitochondrial, mitochondrial, and post-mitochondrial level [12]. HSP70 has been shown to block the recruitment of procaspase-9 to the Apaf-1 apoptosome [8, 13] and to directly antagonize apoptosis-inducing factor (AIF) [14]. Induction of HSP70 has been linked to the activity of the positive transcription elongation factor (P-TEFb), a multiprotein complex containing the catalytic subunit CDK9 that phosphorylates the C-terminal domain of RNA polymerase II and facilitates elongation of the HSP70 transcript [15]. Several in vitro studies have shown that ablation of HSP70 activity enhances HSP90 inhibitor-mediated cell death [16,17,18,19].

Onalespib is a non-ansamycin small molecule inhibitor of HSP90, with an IC50 = 0.7 nM that has demonstrated modest anti-tumor activity in patients with advanced solid tumors using both twice-weekly and once-weekly schedules [19,20,21]. AT7519 is a small molecule inhibitor of CDKs 1,2,4,5, and 9 (IC50 range 11–220 nM) that has demonstrated preliminary activity in patients with advanced solid tumors, chronic lymphocytic leukemias, and mantle cell lymphomas [22,23,24].

In vitro studies supporting the rationale for the combination in a colorectal cancer cell line show dose-dependent inhibition of onalespib-induced HSP70 upregulation by AT7519 (Supplementary Fig. 1A). Combination treatment with onalespib + AT7519 results in a rapid reduction of phosphorylated RNA polymerase II, a result of CDK9 inhibition, suggesting an impaired cellular survival mechanism. In these studies, downstream modulation of client proteins, particularly HER-2 and pAKT, were seen (Supplementary Fig. 1B). Similar data were confirmed in in vitro and in vivo models of NSCLC driven by ELM4-ALK, mutant EGFR, and mutant KRAS [25]. Based on extensive preclinical evidence supporting synergism of the combination, we conducted a phase I trial of the HSP90 inhibitor onalespib in combination with the CDK inhibitor AT7519, in patients with advanced solid tumors to determine the safety profile and maximally tolerated dose and to assess the pharmacodynamic (PD) effects on HSP70 expression in patient-derived PBMCs and plasma.

Patients and methods

Study population

Patients with advanced solid tumors without approved curative therapy or effective palliative therapy, ≥ 18 years of age, with an Eastern Cooperative Oncology Group (ECOG) performance status 0–1, and evaluable or measurable disease per Response Evaluation Criteria in Solid Tumors (RECIST) version 1.1 were eligible for the study. All patients were required to have adequate organ function defined by absolute neutrophil count ≥ 1.5 × 109/L, platelet count ≥ 100 × 109/L, total bilirubin ≤ 1.5 × institutional upper limit of the reference range (ULRR), aspartate aminotransferase (AST) and alanine aminotransferase (ALT) ≤ 2.5 × ULRR, creatinine clearance ≥ 50 mL/min by Cockcroft–Gault formula, creatinine phosphokinase within ULRR, QTc < 450 msec for men or < 470 msec for women by Fridericia correction, and left ventricular ejection fraction ≥ 50% on echocardiogram. Patients were required to have completed previous cancer therapy at least 3 weeks prior to study entry, and adverse events from prior therapy must have recovered to eligibility levels with the exception of alopecia. There were no limits on the number of prior therapies. Exclusion criteria included untreated brain metastases or carcinomatous meningitis, pregnancy or lactation, pre-existing retinal disease on an ophthalmologic exam, grade ≥ 2 peripheral neuropathy, or HIV status requiring antiretroviral therapy due to the potential for PK drug interactions.

Study design and treatment administration

This study was conducted through the National Cancer Institute’s Experimental Therapeutics Clinical Trials Network (ETCTN) and opened at four sites. The primary objective was to determine the safety and tolerability of the combination. Secondary objectives were to describe the PK and to evaluate the PD effects of the addition of AT7519 to onalespib on HSP70 expression in PBMCs and plasma as proof-of-principle of target effect. The trial followed a 3 + 3 dose-escalation design evaluating sequential stepwise dose escalation of each agent, incorporating a 7-days lead-in (cycle “0”) of onalespib alone administered on days 1 and 4, to accommodate PK and PD evaluation of onalespib single-agent activity versus the combination. Thereafter, both agents were administered as a 1-h intravenous infusion twice weekly (days 1, 4, 8, and 11) every 21 days. The first dose level evaluated onalespib at 40 mg/m2 + AT7519 at 21 mg/m2 based on published monotherapy pharmacokinetics of the agents [19, 22]. Starting doses were below the monotherapy RP2D due to concerns for the potential for overlapping liver toxicities. The dose-escalation strategy favored increasing onalespib first based on the observation that higher doses of onalespib led to greater HSP70 induction in these studies, whereas the primary role of AT7519 was to limit the transcription of HSP70 through reduction in RNA polymerase II activation.

Dose-limiting toxicity definitions and study assessments

Safety was assessed via monitoring of toxicities during the lead-in cycle 0 + cycle 1 according to National Cancer Institute Common Terminology Criteria for Adverse Events (NCI-CTCAE) v.4.03. Dose-limiting toxicities (DLT) were defined as Grade 4 neutropenia > 5 days or febrile neutropenia, Grade 4 thrombocytopenia, and any Grade 3–4 non-hematologic toxicities related to study drug and occurring during the lead-in and/or cycle 1. Grade 3 ≥ nausea, vomiting, diarrhea, or electrolyte derangements were considered dose-limiting if refractory to management and not improved to Grade ≤ 2 within 48 h. Any Grade 2 or higher ocular toxicity that did not resolve to Grade 1 or less within 2 weeks of withholding treatment was considered dose-limiting. Additionally, more than one individual dose omission during the DLT window as a result of drug-related toxicity was considered dose limiting. A Safety Monitoring Committee comprised of the Study Chair and the Principal Investigators at each participating site determined the dose escalations according to the toxicity observed in the previous cohorts. The MTD was defined as the dose level at which no more than one of six evaluable patients experienced a DLT during the lead-in period and the first cycle of treatment. At the MTD, up to ten patients could be enrolled in an expansion phase to further evaluate safety, tolerability, and PD endpoints.

A physical examination, assessment of vital signs, pertinent tumor biomarker assessments, hematology and chemistry assessments were performed at screening, day 1 of cycle 0, prior to each treatment (days 1, 4, 8, and 11) and day 15 of cycle 1, and days 1 and 8 of each 21-days cycle starting with cycle 2. Electrocardiograms were performed at the start of each cycle. Based on published ocular and cardiac toxicities of the agents [19, 22] ophthalmologic exam and echocardiogram were performed at screening, after two cycles, and as clinically indicated. Radiologic assessments by computerized tomography or magnetic resonance imaging were performed at screening and every 2 months during the first four cycles, and every three cycles thereafter if patients continued to receive treatment.

Pharmacokinetic assessments

Blood samples were collected in 4-mL plastic Vacutainer tubes with spray-dried K2EDTA (Becton, Dickinson and Co., Franklin Lakes, NJ) prior to dosing and designated timepoints after onalespib infusion on cycle 0 day 1, and on cycle 1 days 1 and 11 after sequential infusion of onalespin and AT7519. Blood collection tubes were centrifuged to harvest the plasma which was stored in cryovials at − 80 °C. Concentrations of onalespib and AT7519 in plasma samples were determined concurrently by ion-pair reversed-phase high-performance liquid chromatography with tandem mass spectrometric detection. The assay was validated as recommended by the US Food and Drug Administration Guidance for Industry, Bioanalytical Method Validation (May 2018). Plasma samples (50 µL) were spiked with a solution of the internal standard and prepared for analysis by vigorously mixing with three volumes of acetonitrile and centrifuging to pellet the precipitated proteins. The supernatant was removed and diluted with 1.5-volumes of 0.2% (v/v) heptafluorobutyric acid in water. An aliquot (10 µL) of the final sample solution was injected onto a Phenomenex (Torrance, CA) Luna 5 µm C18(2) analytical HPLC column maintained at 40 °C and separated by gradient elution using a mobile phase composed of 0.2% (v/v) heptafluorobutyric acid in acetonitrile and water delivered at 1.0 mL/min. An Agilent Technologies (Santa Clara, CA) model G6410B triple quadrupole mass spectrometer with an electrospray ionization interface was operated in the multiple-reaction monitoring mode to detect the m/z 410.2 → 132.1, m/z 382.1 → 84.2, and m/z 324.2 → 293.2 transitions for onalespib (retention time 5.4 min), AT7519 (4.6 min), and the internal standard (4.0 min), respectively. Quantitation was based upon integrating the extracted ion chromatogram for each transition to provide peak areas and calculating the ratio of the analyte-to-internal standard peak area for each study sample, quality control sample, and calibration standard.

Time points were determined as the difference between the blood sample collection time and the starting time of the infusion for each drug. Plasma concentration (free base equivalent)–time data were analyzed by noncompartmental methods using model 202 for constant infusion drug input in WinNonlin Professional version 5.0.1 (Pharsight Corp, Mountain View, CA). Pharmacokinetic parameters are reported as the geometric mean (geometric %CV) of the values for individual patients at each dose level. GraphPad Prism for Windows, version 8.3.0 (GraphPad Software, La Jolla, CA) was used for the statistical comparison of mean pharmacokinetic parameters using the paired or unpaired two-tailed t test, as appropriate, after logarithmic transformation of the data. P < 0.05 was the criterion for statistical significance.

Pharmacodynamic assessments

PBMCs and plasma were isolated from peripheral blood prior to onalespib during cycle 0 (C0D1), 2–4 h after completion of onalespib administration during cycle 0 day 4 (C0D4), and 2–4 h after completion of onalespib and AT7519 administration during cycle 1 day 11 (C0D11) based on optimal sampling time of putative biomarkers in preclinical modeling. The level of HSP70 protein expression was measured by western blot in PBMCs as reported previously [26]; soluble HSP70 protein level was measured by ELISA in plasma (Fig. 1). The significance of changes between C0D1, C0D4, and C1D11 was evaluated using the Wilcoxon sign-rank test. Correlation between change in plasma levels and change in PBMCs was tested using Pearson correlation. Additional exploratory PD assays that were planned included evaluation of HSP90 inhibition in tumor biopsies by assessing transcriptional changes and expression of HSP70 as a comparison to PBMC and plasma expression levels, based on previously published data suggesting differential expression in PBMCs versus tumor [27]. An expansion cohort with paired tumor biopsies was not pursued, however, due to internal re-prioritization by the sponsor to discontinue further development of both agents necessitating early closure of the study.

Fig. 1
figure 1

HSP70 expression in patient-derived plasma and peripheral blood mononuclear cells (PBMCs). Peripheral blood samples were collected at baseline, 2–4 h after completion of onalespib on C0D4, and 2–4 h after completion of the combination of onalespib + AT7519 on C1D11, and separated into plasma and PBMC pellets. The level of soluble HSP70 protein expression was measured by ELISA in plasma (Panel a); HSP70 protein expression was measured by western blot in PBMCs and quantitated (Panel b) as reported previously [26]

Full size image

Results

Patient disposition and characteristics

A total of 29 patients enrolled in the study between September 2016 through January 2019, including a 10-patient expansion cohort at the MTD for evaluation of pharmacodynamic endpoints (Table 1). Twenty-eight patients received at least one treatment, one patient was determined to be ineligible after enrollment, and accordingly was not treated. All patients who received at least one treatment were evaluable for safety. Twenty-three patients were evaluable for response by RECIST v1.1. Three patients were inevaluable for response due to disease progression after completion of the lead-in cycle, necessitating change in treatment to include radiotherapy, surgical intervention, or transition to hospice. Two patients were removed from the study during the first cycle due to dose-limiting toxicities. Eight patients remained on study for greater than four cycles, two patients remained on study for ten cycles.

Table 1 Patient demographics
Full size table

Adverse events

The most common adverse events (AEs) attributed to treatment and occurring in ≥ 10% of patients are summarized in Table 2. Of these AEs, the majority were Grades 1 and 2 (92.3%), 7% were Grade 3, and 0.7% were Grade 4. Across all grades, the most common AEs experienced by patients were gastrointestinal and included diarrhea (22/28; 79%), oral mucositis (16/28; 57%), vomiting (14/28; 50%), nausea (13/28; 46%), and dry mouth (11/28; 39%). Fatigue was also a recurring AE, occurring at a frequency of 54% (15/28). Additional adverse events included hyperglycemia (10/28; 36%), increase in alkaline phosphatase (8/28; 29%), and anemia (10/28; 36%). At the 80 mg/m2 dose of onalespib, two patients (7%) reported blurry vision and “floaters;” however, no abnormalities were detected on ophthalmologic exams. Visual disturbances resolved with holding treatment and did not recur with rechallenge of treatment. Central ECG recordings revealed no change in cardiac repolarization or prolongation of QTc intervals. Additionally, no changes in echocardiogram ejection fraction were detected after two cycles in any patients. Two patients (7%) had Grade 1 increase in creatine phosphokinase consistent with previously reported toxicity of AT7519. The two dose-limiting toxicities at DL3 (onalespib 80 mg/m2 + AT7519 27 mg/m2) were Grade 3 increase in cardiac troponins and Grade 3 oral mucositis. Both patients were removed from further study treatment at physician discretion. No additional dose escalation was pursued, and the MTD was declared at DL2 with onalespib at 80 mg/m2 + AT7519 at 21 mg/m2.

Table 2 Summary of adverse events attributed to study treatment
Full size table

Pharmacokinetic analyses

Mean values of the pharmacokinetic parameters for onalespib and AT7519 for 17 patients evaluated at the MTD of the combination (DL2) are presented in Table 3. The mean maximum plasma concentration (Cmax) of onalespib on day 1 (331 ng/mL) and day 11 (355 ng/mL) of cycle 1 was comparable to the mean Cmax for the cycle 0 dose (349 ng/mL), as expected, because these samples were all obtained near the end of the onalespib infusion and just before the AT7519 infusion was started during cycle 1. The mean total body clearance (CL) of onalespib was 53.8 L/h/m2 on cycle 0 when given alone, 57.5 L/h/m2 on day 1 and 62.5 L/h/m2 on day 11 when given together with AT7519 in cycle 1; these differences were not statistically significant (P = 0.49 for C1D1 vs. C0; P = 0.22 for C1D11 vs. C0). The mean CL of onalespib for the five patients treated with the 40 mg/m2 dose in DL1 was comparable to the corresponding values of the mean CL for the 80 mg/m2 dose. Finally, there was no marked difference between the mean CL of onalespib 80 mg/m2 when given together with 21 or 27 mg/m2 doses of AT7519 in dose levels 2 and 3, respectively.

Table 3 Mean pharmacokinetic parameters for onalespib and AT7519
Full size table

Mean values of the pharmacokinetic parameters for AT7519 21 mg/m2 were similar on days 1 and 11 of cycle 1 for patients evaluated at DL2. The mean Cmax was 459 ng/mL on day 1 and 509 ng/mL on day 11. The mean CL was 9.1 and 8.6 L/h/m2 on days 1 and 11, respectively. Mean values of the pharmacokinetic parameters for the 21 mg/m2 dose of AT7519 in four of the patients that concurrently received onalespib 40 mg/m2 in dose level 1 were similar to the values shown for the dose given together with onalespib 80 mg/m2.

Pharmacodynamic analyses

HSP70 expression in serial plasma and PBMC samples analyzed by ELISA and quantified western blot are shown in Fig. 1. Increased HSP70 protein expression was demonstrated in both plasma and PBMCs across all dose levels after onalespib treatment, consistent with downstream signaling after target engagement in response to HSP90 inhibition. Decreases in HSP70 expression with the addition of AT7519 to onalespib were variable across all samples. Statistical analyses were limited by the small number of patient samples (5/28) who met the criteria for comparative analysis: treatment completion through C1D11, collection of all three samples at C0D1, C0D4, and C1D11, and adequate collection sampling for isolation of both plasma and PBMCs for each timepoint. Pearson correlation of change in HSP70 levels in plasma and PBMC at time points C0D1 (P value = 0.89), C0D4 (P value = 0.12), and C1D11 (P value = 0.08) did not achieve statistical significance.

Antitumor activity

Twenty-three patients (82%) had measurable disease and were evaluable for the best response to therapy by RECIST v1.1. Sixteen patients (69.5%) had disease progression during the first two cycles, including 7 of 12 patients with stable disease (SD) by tumor metrics but with clinical progression in non-measurable sites. Two patients achieved a partial response (PR). One patient with an adenocarcinoma of the palate who achieved a confirmed PR remained on study for 15 cycles. One patient with a Sertoli–Leydig tumor showed marked tumor regression (Fig. 2). This response was not confirmed as a partial response as the patient withdrew from the study after six cycles without confirmatory scans, opting to pursue definitive resection of the remaining tumor. Two additional patients, a colon cancer patient and an endometrial cancer patient both remained on study with SD for a total of ten cycles each. The colon cancer patient progressed after ten cycles whereas the endometrial cancer patient withdrew consent for further treatment due to recurring intolerable Grade 2 fatigue.

Fig. 2
figure 2

Clinical response in a patient with Sertoli–Leydig tumor harboring AKT3 amplification and STK11 R304W mutation. Computed axial tomography (CT) coronal images are shown, yellow circles highlight the index mesenteric mass; baseline (left panel), following two cycles (middle panel), following six cycles (right panel)

Full size image

Discussion

In this study, we show onalespib and AT7519 can be given safely in combination with no clinically relevant pharmacokinetic interactions of the agents, and report preliminary antitumor activity with PD confirmation of target engagement and proof-of-principle of modulation of HSP70 expression with the addition of AT7519. Clinical development of first-generation HSP90 inhibitors has been challenging due to prominent ocular and liver toxicities. In vivo modeling suggests prolonged retinal HSP90 inhibition causes photoreceptor cell death and ocular toxicity, and is postulated to vary by individual agent retina-to-plasma exposure rather than a class-wide effect [28, 29]. No clinically significant ocular toxicities were reported in this study, including those patients receiving therapy for > 6 months, supporting the premise that ocular toxicity may be dependent on the drug-specific retinal elimination rate and less of a concern for second-generation agents such as onalespib [29]. Five patients (18%) in this study experienced low-grade rises in liver transaminases, though no Grade 3 or 4 events occurred. Although the majority of toxicities were Grade 1/2, the recurring nature of these toxicities resulted in the omission of the second treatment of the week or treatment delays for patients who remained on study for longer than six cycles. In our experience, twice-weekly administration of this combination proved to be a challenging treatment schedule.

This study additionally highlights the complexities in developing targeted combinatorial dosing strategies and early phase clinical trial design. The administration schedule of days 1, 4, 8, 11 every 21 days did not result in clinically relevant alterations in the plasma pharmacokinetics of either drug. Mean values of the CL and apparent biological half-life of onalespib were within the ranges previously reported for the initial phase I clinical trial of this agent given as monotherapy [19]. Additionally, AT7519 exhibits linear pharmacokinetics at doses ranging from 14.4 to 32.4 mg/m2 when given as a single agent to adult patients with solid tumors [22]. In our experience, administration of AT7519 as a 1 h infusion immediately after completing the 1 h infusion of onalespib has no effect on the plasma pharmacokinetics of the latter, at the doses evaluated in this study.

The DLT events reported in this study, mucositis in one patient and cardiac troponin increase in a second patient, are postulated to reflect a pharmacodynamic interaction as abrogation of HSP70 upregulation by the addition of AT7519 may result in enhanced onalespib-specific on-target effect and toxicity. Although the study was not designed to assess for dose-dependent pharmacodynamic interaction, the occurrence of both these events at the same dose level guided the decision to terminate escalation. This highlights the limitations of a 3 + 3 design and would argue in favor of a modified adaptive trial design or the time-to-event Bayesian optimal interval design in the identification of the optimal dose for combination studies [30]. Alternate schedule of administration and intermediate doses were not pursued due to the study sponsor’s internal decision to discontinue further development of both agents. The combinatorial MTD was accordingly declared at onalespib 80 mg/m2 + AT7519 21 mg/m2.

Even at doses below the RP2D for each individual agent administered on a twice-weekly schedule, 120 mg/m2 for onalespib and 27 mg/m2 for AT7519 [19, 22], antitumor responses were seen. While we did not preselect for specific mutations due to the expansive nature of the HSP90 interactome which includes more than 200 client proteins, notable clinical responses of interest occurred in two patients with tumors harboring alterations in AKT on retrospective analysis of Next-Generation sequencing of archival tumor. One patient with endometrial cancer achieved durable clinical benefit with SD lasting > 10 months on study had an AKT E17K activating mutation. A second patient with a Sertoli–Leydig tumor was found to have both AKT amplification and STK11 R304W alteration on molecular profiling of her tumor. In vitro data suggest both STK11 and AKT form intracellular complexes with HSP90 and Cdc37 [31, 32]. In these studies, treatment with HSP90 inhibitors resulted in proteasome-mediated degradation of the respective protein. Changes in HSP70 expression could not be evaluated in these two patients as the C1D11 sample could not be collected on the patient with endometrial cancer and the baseline samples for the patient with Sertoli–Leydig tumor were too degraded for comparative analysis. Neither patient had progressed on study at the time of study discontinuation. The patient with endometrial cancer withdrew after 10 cycles due to recurring Grade 2 fatigue and the patient with Sertoli–Leydig tumor withdrew after achieving an unconfirmed PR, to pursue definitive resection of her remaining tumor.

The patient with palate adenocarcinoma who achieved a confirmed PR on study was found to have androgen receptor (AR) P392S alteration, a missense mutation which has been shown to predict for partial androgen insensitivity [33]. While this particular variant has not previously been identified in salivary cancers, other AR variants which have previously been implicated in resistance to androgen deprivation therapy (ADT) in prostate cancer have also recently been identified in salivary cancers [34]. HSP90 has been shown to play a critical role in stabilization of the AR and overexpression of HSP90 has been detected in prostate cancer cells [35]. It is conceivable to postulate that depletion of AR through HSP90 inhibition may explain this patient’s response and presents an opportunity to address ADT resistance in the treatment of salivary cancers [36].

For this patient, plasma analysis demonstrated an increase in soluble HSP70 expression after treatment with onalespib alone followed by a decrease in HSP70 expression after treatment with the combination of onalespib and AT7519 (Fig. 3). This is consistent with a proof-of-principle of transcriptional signaling downstream from target engagement following HSP90 inhibition with onalespib only, followed by downstream target effect of inhibition of CDK9-mediated transcription of HSP70 with combination therapy. Changes in HSP70 expression in PBMCs followed the same trend but were not statistically significant. Due to the small sample size of the study, definitive correlation of changes in HSP70 expression with response could not be verified in other patients. We additionally cannot rule out that variability of the timing of sample collection across multiple institutes may have contributed to the wide range in HSP70 expression across samples.

Fig. 3
figure 3

Clinical response a patient with adenocarcinoma of the palate with corresponding HSP70 measurements in plasma and PBMC. a Computed tomography scan demonstrating partial response of thoracic metastases (arrow) at baseline (Top) and at 13 cycles (Bottom). Right-sided panels show corresponding: b changes in soluble HSP70 protein expression in plasma as measured by ELISA at C0D1, C0D4 and C1D11; and c ImageQuant quantitated changes in HSP70 protein expression as measured by western blot in PBMCs at C0D1, C0D4 and C1D11

Full size image

In this study, we have demonstrated the biologic activity of onalespib and AT7519 on HSP70 expression. The variability of HSP70 expression in plasma and PBMCs seen in this study highlights not only the challenges of biomarker development for this class of agents but also adds to the ongoing debate of whether PBMCs are an optimal surrogate for HSP90 inhibition, as putative driver client proteins may be differentially expressed in PBMCs versus tumor [10]. The intriguing preliminary responses seen in this study warrant further evaluation in tumors harboring mutations in HSP90 client proteins of interest, particularly AKT and AR, with incorporation of pharmacodynamic markers of HSP90 inhibition in tumor samples. The decision to halt additional agent development and study closure precluded the collection of tumor biopsies in a planned expansion cohort. Future biomarker development will require further validation of on-target impact on relevant client proteins within tumor tissue given the expansive nature of the HSP90 interactome, variability of client protein degradation within tumor, and differential PK/PD profiles of HSP90 inhibitors in tumor versus normal tissue, to determine the robustness of this parameter as a surrogate biomarker [37, 38]. Further validation of intratumoral target engagement and development of biomarker profiling of HSP90 inhibition will be needed to support the successful development of these classes of agents.