The PI3K pathway is considered a master regulator for cancer due to its frequent activation, making it an attractive target for pharmacologic intervention. While substantial efforts have been made to develop drugs targeting PI3K signaling, few drugs have been able to achieve the inhibition necessary for effective tumor control at tolerated doses. HSP90 is a chaperone protein that is overexpressed and activated in many tumors and as a consequence, small-molecule ligands of HSP90 are preferentially retained in tumors up to 20 times longer than in normal tissue. We hypothesize that the generation of conjugates that use a HSP90-targeting ligand and a payload such as copanlisib, may open the narrow therapeutic window of this and other PI3K inhibitors. In support of this hypothesis, we have generated a HSP90-PI3K drug conjugate, T-2143 and utilizing xenograft models, demonstrate rapid and sustained tumor accumulation of the conjugate, deep pathway inhibition, and superior efficacy than the PI3K inhibitor on its own. Selective delivery of T-2143 and the masking of the inhibitor active site was also able to mitigate a potentially dose-limiting side effect of copanlisib, hyperglycemia. These data demonstrate that by leveraging the preferential accumulation of HSP90-targeting ligands in tumors, we can selectively deliver a PI3K inhibitor leading to efficacy in multiple tumor models without hyperglycemia in mice. These data highlight a novel drug delivery strategy that allows for the potential opening of a narrow therapeutic window through specific tumor delivery of anticancer payloads and reduction of toxicity.
The PI3K pathway is one of the most frequently dysregulated signaling cascades in cancer and is implicated in a wide range of tumor types (1). The PI3K–AKT–mTOR pathway is an intracellular signaling pathway that regulates cellular processes such as increased cell growth, proliferation, and differentiation, which in combination with its common dysregulation in cancer, makes it an attractive target for pharmacologic intervention (2). Given the importance of this pathway in cancer, more than 40 compounds that target the PI3K–AKT–mTOR pathway have been tested in clinical trials involving patients with a wide range of tumor types. However, many of these agents have not advanced beyond early-stage clinical testing due to the inability to achieve deep pathway inhibition in tumor tissue while avoiding dose-limiting toxicities in the patient (3, 4). Despite numerous PI3K–AKT–mTOR inhibitors entering clinical trails, only five pathway inhibitors, temsirolimus (mTOR), everolimus (mTOR), idelalisib (PI3K), alpelisib (PI3K), and copanlisib (PI3K) have been approved by the FDA for clinical use in the treatment of a number of different cancers, with the majority of approvals in liquid tumors, demonstrating a clear need for novel approaches to achieving deep PI3K–AKT–mTOR pathway inhibition while hyperglycemia in the solid tumor setting.
Lack of site-specific drug delivery, particularly delivery of cytotoxic agents to tumors, has led to the emergence of tumor-selective drug delivery systems over the last 30 years. Tumor-selective drug delivery systems include the use of targeting moieties, such as mAbs, peptides, and synthetic polymers, chemically linked to an active cytotoxic agent to form a drug conjugate (5, 6). While antibody–drug conjugates (ADC) have demonstrated efficacy in the treatment of solid tumors, with five being currently marketed, they have potential deficiencies that include the slow speed of diffusion and poor tumor penetration, potentially limiting effectiveness in the solid tumor setting (7). The increased time that an ADC is in circulation can lead to the prolonged release of payload in the bloodstream causing potential systemic toxicities (8). Given the liabilities of antibody-based therapies, miniaturized drug conjugates have the potential to target tumor cells allowing for rapid accumulation of potent payloads in tumor cells deep into the core of the tumor and slowly release the payload over time giving sustained delivery of the drug to the tumor while being quickly cleared from the plasma, resulting in a decrease in overall toxicity. PEN-866, currently in phase I/IIa (NCT03221400) applies these fundamental principles and is a synthetic small-molecule drug conjugate comprised of a heat shock protein 90 (HSP90) targeting moiety attached through a cleavable linker to SN-38, the active metabolite of the topoisomerase I inhibitor, irinotecan (approved worldwide for the treatment of patients with colorectal cancer and other cancers). HSP90 is a highly conserved and widely expressed molecular chaperone protein that regulates a diverse range of cellular functions such as the folding, stability, and degradation of many proteins. As a result, HSP90 exerts marked effects on normal biology and disease processes (9, 10). Although HSP90 is highly expressed in most cells, it has been shown to be upregulated and activated in a variety of tumor cells compared with normal healthy tissue.
The activated form of HSP90 found in cancer tissues is present in a highly complexed state with cochaperone proteins with high ATPase activity, with these complexes functioning as a network to enhance cellular survival, regardless of tissue of origin or genetic context. The consequence of this high activation state is a propensity to selectively bind with high affinity to HSP90 inhibitors as compared with the latent form of HSP90 found in normal tissue rendering HSP90 inhibitors selective for cancer cells (11, 12). The selectivity of HSP90 inhibitors for the activated form of HSP90 present in the cancer cells allows for preferential killing of cancer cells compared with normal cells (13, 14). While PEN-866 has been designed to limit HSP90 inhibition, it still demonstrates selective targeting of the activated cancer form of HSP90, and is able to penetrate, accumulate, and be retained in the tumor cells leading to superior antitumor activity when compared with an equimolar SN-38 dose of irinotecan and to small cell lung cancer (SCLC) standard-of-care agents, cisplatin, and topotecan (15, 16).
Here we report on the preclinical characterization of an HSP90-PI3K conjugate. By leveraging the preferential accumulation of HSP90-targeting ligands in tumors, as seen with PEN-866, and by masking of the PI3K inhibitor active site though conjugation to an HSP90-targeting ligand, our HSP90-PI3K conjugate selectively accumulates and is retained in tumors leading to deep pathway inhibition over time with a decrease in a commonly observed toxicity, hyperglycemia, for PI3K alpha inhibitors. These proof-of-concept experiments include demonstration of extended intratumoral drug exposures, extended inhibition of the PI3K–AKT–mTOR pathway, and a superior therapeutic window as compared with the PI3K inhibitor alone. Together, these results demonstrate a superior tumor-selective drug delivery system that can deliver payloads directly to tumors, achieve sustained and deep pathway inhibition while mitigating hyperglycemia
Materials and Methods
The investigational compounds T-2026, T-2143, and T-2212 were synthesized according to the procedures described in the Supplementary Material and were isolated and utilized as their respective trifluoroacetate salts. T-2026 is the PI3K inhibitor payload that was incorporated into the conjugates. T-2143 is the HSP90-binding conjugate of T-2026 and T-2212 is the nonbinding control analogue of T-2143 where part of the HSP90-binding pharmacophore has been blocked with the addition of a methyl group. Copanlisib as the dihydrochloride salt was purchased from MedKoo Biosciences, Inc. The doses used for all of the compounds in all in vivo experiments represent free base equivalent doses.
HSP90-binding activity assay
Assays were carried out at BPS Biosciences. Briefly, the reactions were conducted at room temperature for 3 hours in a 100 μL mixture containing assay buffer, 5 nmol/L FITC-labeled geldanamycin and the test compound. Each compound was run in duplicate, 10-point curve starting at a concentration of 10 μmol/L. Fluorescence intensity was measured at an excitation of 485 nm and an emission of 530 nm using a Tecan Infinite M1000 microplate reader. Fluorescence intensity is converted to fluorescence polarization using the Tecan Magellan6 software. The fluorescence polarization data were analyzed using the computer software, Graphpad Prism. The fluorescence polarization (FPt) in absence of the compound in each dataset was defined as 100% activity. In the absence of protein and the compound, the value of fluorescent polarization (FPb) in each dataset was defined as 0% activity. The percent activity in the presence of the compound was calculated according to the following equation: % activity = (FP-FPb)/(FPt-FPb) × 100%, where FP = the fluorescence polarization in the presence of the compound. For IC50 measurements, the values of percent activity versus a series of compound concentrations were then plotted using nonlinear regression analysis of Sigmoidal dose-response curve generated with the equation Y = B+(T-B)/1+10((logEC50-X)×Hill Slope), where Y = percent activity, B = minimum percent activity, T = maximum percent activity, X = logarithm of compound, and Hill Slope = slope factor or Hill coefficient. The IC50 value was determined by the concentration causing a half-maximal percent activity.
All cells lines were purchased from the ATCC. The following cell lines were maintained in the recommended media: NCI-H460 (ATCC HTB-177, RRID: CVCL_0459), HTC-116 (ATCC CCL-247, RRID: CVCL_0291), LS174T (ATCC CL-188, RRID: CVCL_1384), BT-474 (ATCC HTB-20, RRID: CVCL_0179), and SKOV3 (ATCC HTB-77, RRID: CVCL_0532).
In vitro cellular proliferation
The following cell lines were seeded at 2,500 cells per well in a 96-well plate: HCT116, LS174T, and BT-474. NCI-H460 cells were seeded at 500 cells per well. All cell lines were seeded in their respective medium. Plated cells were incubated overnight before treatment. Cells were treated with a dose range of T-2212 and T-2143 which were dissolved in DMSO (Sigma, D4540). The starting concentration was 10 μmol/L for each compound which was diluted by 3-fold for a total of 10 data points. The number of viable cells was determined by CellTiter-Glo 2.0 Cell Viability Assay (Promega G9243) through quantifying ATP following the manufacturer's protocol. All samples were run in duplicate. Inhibitory concentration of 50% growth was determined using GraphPad statistical software, GraphPad Prism 6.
Cell-free enzymatic assay
The determinations of inhibitory concentration of 50% of PI3K3CA, PIK3CB, PIK3CD, and PIK3CG activity with T-2143, copanlisib, or T-2026 were carried out at ProQinase GmbH using the Free Choice Kinase assay. All samples were run in duplicate.
In vivo studies
All studies were conducted in accordance with the Tarveda Therapeutics Institutional Animal Care and Use Committee. All mice were treated in accordance with the OLAW Public Health Service Policy on Human Care and Use of Laboratory Animals and the ILAR Guide for the Care and Use of Laboratory Animals. Female athymic nude mice (CrTac:NCr-Foxn1nu) were purchased from Taconic (IMSR, catalog no. TAC:ncrnu, RRID: IMSR_TAC:ncrnu). Infectious disease testing on the cell lines was done by Charles River Laboratories by PCR using the Mouse Essential Clear Panel, within 3 years of cell line use. Cells were only used if they had been passaged a total of 25 times or less. Mice were 8-11 weeks old and weighed approximately 20 grams at the time of dosing. T-2143, T-2212, copanlisib, and T-2026 were all dosed intravenously. Animals were euthanized once tumors reached a maximum of 2,000 mm3 or at the end of study, whichever came first. At the end of each study, the animals were euthanized via CO2 inhalation followed by cervical dislocation to ensure death.
Cells were harvested from tissue culture and made into a final suspension by mixing with matrigel (BD Biosciences, catalog no. CB-40234; 1:1 ratio). All cells were implanted subcutaneously in the right flank of mice. LS174T (H1047R) and NCI-H460 (E545K) cells were implanted at a concentration of 2.5 × 106 cells per mouse and BT-474 (K11N) cells were implanted at a concentration of 5.0 × 106 cells per mouse. Tumor growth was monitored throughout the study and measured twice weekly. Tumor volume was measured in two dimensions using calipers and volume was calculated using the formula: (w2 × l)/2 = mm3, assuming 1 mg is equivalent to 1 mm3 of tumor volume. The health of the mice was monitored, and noteworthy clinical observations were recorded. Acceptable toxicity was defined as group mean body weight (BW) loss of less than 20% during the study and not more than one treatment-related death among 10 treated animals. Mice were randomized into treatment groups of 10 animals per group and therapy began when tumor volumes were approximately 200 mm3 usually 7-14 days postimplantation.
For mouse in vivo studies, T-2143 and T-2212 were formulated in 5% DMSO (Alfa Aesar, High-performance liquid chromatography (HPLC) grade) and 95% of freshly prepared 10% Solutol (Kolliphor HS 15) in 5.2% Dextrose (USP grade, Sigma-Aldrich). Copanlisib HCl (MedKoo Biosciences, Inc.) was prepared in 5% Mannitol (Pearlitol PF; Roquette) in water for injection (HyPure, GE Hyclone). T-2026 dosing solution was prepared in 15% DMSO:Solutol (Alfa Aesar, HPLC grade, Kolliphor HS 15) in 85% Dextrose (USP grade, Sigma-Aldrich, Batch). All dosing solutions were prepared fresh and transferred through 0.22 μm PTFE filters (Millex-LG, Merck Millipore) in a sterile hood and the target doses were verified by established HPLC methods.
Mice were dosed intravenously, three animals per each time point were anesthetized using CO2 and approximately 1.0 mL of blood sample was collected via cardiac puncture using a needle (BD Biosciences, catalog no. 14-826-88). While animals were still anesthetized, the animals were euthanized via cervical dislocation. Blood was transferred to K2EDTA sample collection tube (BD Biosciences, catalog no. 02-669-33), inverted four times and centrifuged for 10 minutes at 10,000× g, plasma harvested, transferred into labeled 96-well plate, and stored at −80°C.
Mice bearing LS174T or BT-474 tumors were treated with 10.6 mg/kg of copanlisib, 25 mg/kg of T-2143, or 25 mg/kg of T-2212 (N = 3). Tumors were collected 1, 24, or 72 hours postdose. Xenograft tumors were processed for lysate using Cell Lysis Buffer 10× (Cell Signaling Technology, 9803) diluted to 1×. Proteins were separated by SDS-PAGE on NuPAGE 4%–12% Bis-Tris Protein Gels, 1.5 mm, 15-well (Thermo Fisher Scientific, NP0336BOX). Proteins were transferred from the gel to a nitrocellulose membrane using iBlot Transfer Stack, nitrocellulose, regular size (Thermo Fisher Scientific, IB301001). After the transfer, the membrane was blocked in PBS Blocking Buffer (LI-COR, 927-40000). This was followed by overnight incubation of 1:2,000 Phospho-Akt (Ser473; D9E) XP Rabbit mAb (Cell Signaling Technology, 4060, RRID:AB_2224726) and 1:1,000 GAPDH (D4C6R) Mouse mAb (Cell Signaling Technology, 97166, RRID:AB_2756824) at 4°C. After the primary incubation, the membrane was washed with Pierce 20× TBS Tween 20 Buffer (Thermo Fisher Scientific, 28360) and incubated further for one hour at room temperature with the secondary antibodies: 1:10,000 IRDye 680RD Goat anti-Rabbit IgG (LI-COR 926-68071) and 1:10,000 IRDye 800CW Goat anti-Mouse IgG (LI-COR, 926-32210). Then the membrane was imaged on the LI-COR Odyssey Imaging System. Bands were quantified on the basis of their size and brightness using LI-COR Odyssey imaging software.
All doses were scaled to the BW of the individual animals at a dose volume of 10 mL/kg. Treatment regimens and dosages in each experiment are described in the Results and in the Figure Legends. Each treatment group consisted of 10 animals. Percent tumor growth inhibition (%TGI) was defined as the difference between the mean tumor volume (MTV) of the vehicle and the MTV of the drug-treated group, expressed as a percentage of the MTV of the vehicle group. Statistical analysis was done using one-way ANOVA and Tukey multiple comparisons in GraphPad statistical software, GraphPad Prism 6. A single efficacy study is represented in the graphs presented in the paper; however, for both the BT-474 and LS174T studies, separate graphs are shown to comparing T-2143 with either the negative control, T-2122, or the components that collectively represent T-2143 copanlisib and ganetespib, respectively, as shown in Figs. 2 and 5.
Animals were weighed on the days indicated in the graphs (Supplementary Fig. S3A–S3C). For each treatment group, weights were averaged and compared with the average weight of the group on first day of treatment to obtain the percent change in weight. Statistical analysis was done using one-way ANOVA and Tukey multiple comparisons in GraphPad statistical software, GraphPad Prism 6.
Blood glucose study
Nontumor-bearing mice were dosed intravenously with equimolar concentrations of the vehicle, 10.8 mg/kg copanlisib, 25 mg/kg T-2143, and 10.8 mg/kg T-2026. Each treatment group consisted of N = 5. Mice were monitored for any significant health changes. For the T-2143 and vehicle control groups, tails were pricked 1, 4, 6, and 24 hours postdosage. For all other groups, blood was sampled 1-hour postdosage and blood glucose readings were taken with the AlphaTRAK 2 Blood Glucose Monitoring System (Zoetis CFGW210-M1673).
Tumor and plasma pharmacokinetic analysis
In vivo sample preparation
Standard calibrations were prepared by serially diluting 3 mmol/L methanol (Thermo Fisher Scientific, catalog no. A456-4) stock solutions of T-2143, T-2212, and T-2026 in pooled BALB/C female mouse Lithium Heparin plasma (Bioreclamation IVT, catalog no. MSE02PLLHXNN). Mouse tumors were weighed and homogenized in PBS (Thermo Fisher Scientific, catalog no. 14040133) containing 100 nmol/L of dichlovous (Sigma, catalog no. 45441) using a handheld tissue homogenizer (Omni). A protein precipitation procedure was performed with 100 μL of blank plasma standards, homogenate and sampled mouse plasma, crashed with 300 μL of acetonitrile (Thermo Fisher Scientific, catalog no. A998SK-4) containing 0.1% formic acid (Thermo Fisher Scientific, catalog no. A117-50) and 100 ng/mL of glyburide (Sigma, catalog no. G2539) into a 96-well plate. The samples were capped, vortexed to mix, then centrifuged at 3,500 × g for 5 minutes. A total of 100 μL of the supernatant was then transferred to a new 96-well plate before LC/MS-MS analysis.
LC/MS-MS for quantification
An LC/MS-MS method was developed for investigational molecules; T-2143, T-2212, and T-2026. Leap technologies HTS PAL autosampler was used to inject 10 μL of sample volume to ABSciex 4000 series tandem mass spectrometer coupled with a Shimadzu LC10AD VP LC system. An ACE 3 C18 3 μmol/L, 50 × 2.1 mm2 column (PN ACE-111-0502) was used to chromatographically separate analytes. Mobile phase consisted of acetonitrile containing 0.1% formic acid (A) and water containing 0.2 mol/L of ammonium acetate (Thermo Fisher Scientific, A637-500) and 0.1% formic acid. A gradient elution was employed, starting at 5% A, holding for 0.5 minutes, increasing to 95% A over 3 minutes, holding for 1.5 minutes then reequilibrating at 5% A for 1.1 minutes. Mass spectrometric detection was performed using electrospray ion source operating in positive mode. The mass spectrometer was operated in multiple reactions monitoring mode, monitoring m/z for analytes, T-2143, T-2212, T-2026, and glyburide (internal standard). Analyst software version 1.6.2 (ABSciex) was used to fit T-2143, T-2212, and T-2026 standard curves with a linear 1/× weighted fit.
Noncompartmental pharmacokinetic analysis was performed using Phoenix WinNonlin version 8.1 (Certara).
HSP90 binding is essential for miniaturized drug conjugate activity
Our HSP90-PI3K conjugates are comprised of an HSP90-targeting moiety, cleavable linker, and an anticancer therapeutic payload. Linkers and payloads of the Pentarin platform are designed and optimized in such a way as to mask the payload's active site through linker attachment to the payload, rendering the payload inactive until intracellular cleavage releases the active payload within the tumor cell. The pharmacophore of the T-2143 is based off of the HSP90 inhibitor, ganetespib. For our proof-of-concept experiments the PI3K inhibitor copanlisib was chosen as the model payload. Copanlisib was modified by converting the terminal morpholine to a piperazine to produce T-2026. This incorporated an amine that could be used for the purposes of linking. The piperazine analogue of copanlisib was shown to be highly potent across PI3K subtypes, similar to copanlisib (Table 1). The conjugate of this piperazine was constructed with a carboxyesterase cleavable linker to the HSP90-binding pharmacophore as shown for the structure of T-2143 (Fig. 1A). It was found that linking to the payload in this way diminished inhibition of PI3K subtypes (Table 1). A nonbinding control of T-2143, T-2212, was generated through blocking the HSP90-binding pharmacophore with a single methyl group addition to one of the resorcinol hydroxyl groups. The binding affinity to HSP90α of the molecules was determined using a competitive binding assay, where the competitive molecule was a fluorescently labeled version of the potent HSP90-binding molecule, geldanamycin. The addition of a methyl group proved to be sufficient to completely block the binding affinity as compared with T-2143 with an IC50 of 84 nmol/L and Kd of 1.4 nmol/L as compared with no appreciable inhibition of activity by T-2212 up to concentrations of 10 μmol/L (Fig. 1B). Furthermore, the Kd of T-2143 is equivalent to ganetespib, with a binding Kd of 0.5 nmol/L. This conservative molecular modification has very little impact on the overall properties of T-2212 and has been shown to not have an impact on the mouse pharmacokinetic profile relative to T-2143 (Table 2; Supplementary Fig. S1).
|.||Inhibition of PI3K IC50 (nmol/L) .|
|Isoform .||T-2143 .||Copanlisib .||T-2026 .|
|.||Inhibition of PI3K IC50 (nmol/L) .|
|Isoform .||T-2143 .||Copanlisib .||T-2026 .|
Note: Conjugation of payload to HSP90-targeting moiety results in masking of the payload's active site. The ability of T-2143 and copanlisib to inhibit four isoforms of PIK3CA was evaluated in a cell-free enzymatic assay.
|Parameter .||Unit .||T-2143 .||T-2212 .|
|Parameter .||Unit .||T-2143 .||T-2212 .|
Note: T-2143 and T-2212 exhibit similar plasma pharmacokinetic profiles in mice. Mice were treated with a single intravenous bolus of 25 mg/kg of T-2143 or T-2212. Mice were sacrificed and plasma was collected at 0.83, 0.5, 2, 4, and 24 hours postdose. Plasma samples were analyzed for T-2143 or T-2212 levels. T1/2 is elimination half-life, Cmax is maximum plasma drug concentration, AUC24 is area under the plasma concentration curve from the time zero to time t, VZ is volume of distribution during terminal phase, Vss is volume of distribution at steady state, and CL is total body clearance of drug from plasma.
PIK3CA mutations occur in about 15% to 30% of breast, endometrial, and colon cancers. A panel of cell lines harboring common gain-of-function activating PIK3CA mutations, and reported to be sensitive to pathway inhibition, were chosen for HSP90-PI3K conjugate testing. The in vitro antiproliferative activities of T-2143 and T-2212 were tested in comparison with the PI3K inhibitor alone, copanlisib (17–22). In the cell lines selected, potent cell kill activity similar to copanlisib could only be achieved with T-2143, not T-2212. Furthermore, in three of the five cell lines tested, 50% cell growth inhibition could not be achieved with T-2212, whereas T-2143 and copanlisib demonstrated a potent effect on cell viability, showing that HSP90 targeting is essential for delivery of the PI3K inhibitor payload from our miniaturized drug conjugate (Table 3).
|.||Inhibition of proliferation IC50 (nmol/L) .|
|Cell line .||T-2143 .||T-2212 .||Copanlisib .|
|.||Inhibition of proliferation IC50 (nmol/L) .|
|Cell line .||T-2143 .||T-2212 .||Copanlisib .|
Note: T-2143 inhibits in vitro cellular proliferation similar to copanlisib, while the nonbinding control, T-2212, does not. Cells were treated with T-2143, T-2212, or copanlisib for 72 hours and then measured for proliferation (N = 3).
*NA = 50% inhibition not achieved at highest concentration tested of 10 μmol/L.
To further investigate the dependence of HSP90 targeting for tumor-specific payload delivery of our HSP90-PI3K conjugate, tumor pharmacokinetic and efficacy studies were carried out in LS174T colon cancer xenograft-bearing mice comparing T-2143 against T-2212. Initial experiments defined the MTD of T-2143 and T-2212 at 25 mg/kg given once weekly. In support of the MTD data, the BWs for all efficacy studies are shown in Supplementary Fig. S3A–S3C and demonstrate tolerability of the compounds at doses up to 25 mg/kg given once weekly. The validity of T-2212 as an appropriate control was further evaluated through plasma pharmacokinetic studies in mice, described in more detail below. Briefly, the plasma concentration versus time curves showed a high degree of similarity for T-2143 and T-2212 and resulted in closely matched half-lives and AUC, suggesting that any differences in in vivo activity would not be a result of differences in plasma pharmacokinetic profiles, but due to differences in HSP90-binding ability (Table 2; Supplementary Fig. S1). Following a single bolus injection of T-2143 or T-2212, at their MTD 25mg/kg, LS174T tumors and plasma were harvest 24, 48, and 72 hours postdose. Tumors were analyzed for total HSP90-PI3K conjugate levels. In agreement with the in vitro data that suggest the targeting is important for activity of the drug, T-2143 tumor levels were significantly higher than T-2212 tumors levels with approximately 6-fold higher concentration at both the 24- and 48-hour time points (Fig. 2A). T-2143 and T-2212 were also tested for differences in effects on tumor growth inhibition (TGI). Mice bearing LS174T or BT474 tumors were treated with T-2143 or T-2212 and evaluated for their effects on TGI after two weekly MTD doses. In the LS174T model, T-2143 treatment led to a TGI of 87% while T-2212 treatment resulted in a minimal TGI of 29%, demonstrating a statistically significant increase in tumor growth inhibition with T-2143 as compared with nontargeted control (P = 0.006; Fig. 2B) Results in the BT474 model demonstrated similar results with T-2143 treatment leading to a 72% in TGI, while T-2212 was only able to achieve a TGI of 16% (Fig. 2C). Together, these data demonstrate that HSP90 binding is essential for successful tumor delivery of the PI3K inhibitor payload from our HSP90-PI3K miniaturized drug conjugate.
Tumor retention of HSP90-PI3K conjugate leads to deep and durable PI3K pathway inhibition
The pharmacokinetic profile of T-2143 was evaluated following administration of a single intravenous dose in nontumor-bearing mice. T-2143 exhibited a low volume of distribution, a plasma clearance rate of 2.62 mL/kg/minute, and a half-life of 7.4 hours, demonstrating that T-2143 is relatively quickly cleared from circulation reducing the chances for normal tissue exposure and toxicity (Fig. 3). Levels of released payload, T-2026, from T-2143 were also measured in the plasma. Very minimal amounts of T-2026 were detectable in the plasma with a Cmax of 0.3 μmol/L while T-2143 was able to achieve a Cmax of 121 μmol/L, demonstrating that while T-2143 is in circulation, minimal amounts of payload are released decreasing the chances for toxicity from payload exposure. As discussed above, the plasma half-life of T-2212 was similar to T-2143 and the similarities extended to the other pharmacokinetic parameters (Table 2). Next, pharmacodynamic experiments were conducted to investigate conjugate accumulation and release the payload in tumor over time despite being cleared from the plasma.
PIK3CA mutations and amplifications are known to be drivers of AKT pathway activation, with AKT being central to the activity of the PI3K/AKT/mTOR pathway (23–26). As a result of AKT's importance to the pathway, we investigated the ability of T-2143 to effectively target the tumor, be retained, and deliver the PI3K inhibitor payload over time. Pharmacodynamic analysis was performed in PI3K-mutated BT-474 breast cancer xenograft-bearing mice where T-2143 was administered as a single intravenous bolus injection of T-2143 and tumors were collected at 1, 24, 48, and 72 hours postdose. Tumors were split then analyzed for T-2143 tumor retention and pAKT (S473) levels over time. T-2143 was rapidly taken up by the tumor with 1.2 μmol/L measured at the 1-hour time point. The highest tumor levels were measured 24 hours postdose with 2.3 μmol/L of T-2143 measured at this time point. T-2143 was well retained in the tumor with 0.55 μmol/L still measured 72-hour postdose. Tumor payload levels increased over time with 2.1 μmol/L still detectable at the 72-hour time point (Fig. 3B). Tumor accumulation and sustained release of the payload translated into durable PI3K pathway inhibition. While at the 1-hour time point, pAKT levels only decreased 33% relative to the vehicle control and at 24-hour postdose, a 75% reduction was observed, which was sustained out to 72 hours with an 83% reduction measured at this time point (Fig. 3C). Together, these data suggest that although the conjugate is rapidly cleared from the plasma, T-2143 can effectively target and accumulate in tumors and that accumulation of T-2143 leads to sustained release of the payload, and PI3K pathway suppression over time.
To inform the schedule of efficacy dosing, more extended pharmacodynamic and pharmacokinetic experiments were carried out in the LS174T xenograft model. Mice were administered a single dose of T-2143 or its vehicle control and tumors were collected at 1, 24, 48, 72, and 168 hours postdose. As in the BT-474 model, T-2143 rapidly accumulated and was retained in LS174T tumors with 0.431 μmol/L of T-2143 and 1.45 μmol/L of the payload detected at the 72-hour time point, which translated into an 82% reduction in pAKT levels relative to the payload (Fig. 4). By the 7-day (168-hour) time point, only 0.09 μmol/L of T-2143 was detectable in the tumor and the pAKT signal had returned to levels similar to that of the vehicle, suggesting that T-2143 could be dosed on a once a week schedule to achieve deep pathway inhibition.
Masking of payload active site increases therapeutic window
Hyperglycemia is a known and potentially dose-limiting side effect of PI3K inhibitors due to their interaction with the insulin-glucose regulatory axis (22). While this effect is usually transient due to the compensatory insulin release from the pancreas, it has been shown that this insulin feedback can reactivate PI3K–mTOR signaling in tumors, compromising the inhibitor's effectiveness (27). Our HSP90 miniaturized drug conjugates are designed with the payload binding site blocked by the linker to prevent binding to the target until the payload is released. Payload activity is blocked until the HSP90 ligand targets the conjugate to tumor cells and the payload is released through linker cleavage preferentially in the cancer cell. In addition, HSP90 conjugates are also designed to have short plasma circulation allowing for rapid accumulation and penetration into the tumor while potentially limiting the overall exposure to normal tissue. By incorporating tumor-specific delivery and masking of the T-2143 payload while in circulation, we hypothesized that with our conjugate we could mitigate an increase in glucose, resulting in a larger therapeutic window.
In a cell-free PIK3CA enzyme assay, the HSP90-PI3K conjugate T-2143 is 116-fold less active than copanlisib but retains potent cell kill activity demonstrating the ability to mask the payload's active site until it is delivered directly to the tumor cells (Table 1). To test this in vivo, mice were treated with a single intravenous bolus dose of 25 mg/kg T-2143, an equimolar dose of copanlisib to the payload amount of T-2143 (10.8 mg/kg), or the T-2143 vehicle alone. One-hour postcompound administration, copanlisib treatment led to 4-fold increase in blood glucose levels in comparison with the vehicle control. The average blood glucose level in the vehicle control group was 128 mg/dL while the copanlisib group increased to 553 mg/dL and the T-2143 group average level stayed at 124 mg/dL (Fig. 5). The effect of T-2143 on blood glucose levels was additionally measured at 4, 6, and 24 hours postdose administration. No increase in blood glucose levels relative to the vehicle control was detected with T-2143 administration, demonstrating the ability of the HSP90-PI3K conjugate to reduce target specific toxicity, hyperglycemia, through payload active site masking, potentially increasing the therapeutic window (Supplementary Fig. 2A).
Superior tumor growth inhibition observed with HSP90-PI3K conjugate in comparison with copanlisib alone
Antitumor activity of T-2143 was evaluated in comparison with copanlisib alone in mouse xenograft models of human cancer all reported by ATCC to have mutations in the PIK3CA gene that result in gain-of-function activity of the pathway. The three models, LS174T (H1047R), BT-474 (K11N), and NCI-H460 (E545K), were treated with T-2143, copanlisib, ganetespib, the combination of copanlisib and ganetespib, or a vehicle control. Ganetespib was chosen to represent the effects of the HSP90-targeting moiety alone because it is structurally similar to the ligand used in T-2143 and the molecules have similar binding affinities as described above (28). Copanlisib and ganetespib doses were chosen on an equimolar basis to either the amount of payload or ligand used to create T-2143. All treatments were dosed on a once per week basis and given as a single bolus intravenous injection. In all three models tested, T-2143 treatment was well tolerated with no significant BW loss observed (Supplementary Fig. S3A–S3C). In the NCI-H460 non–SCLC xenograft model, T-2143 was the only treatment able to produce statistically significant tumor growth inhibition as compared with vehicle with a TGI of 48% (P = 0.043). Copanlisib alone was only able to achieve a TGI of 15%, while the combination of copanlisib achieved a TGI of 8% (Fig. 6A). In the BT-474 breast cancer xenograft model, T-2143 treatment resulted in a TGI of 74%, while copanlisib alone, or the combination of copanlisib and ganetespib, was only able to achieve a TGI of 22% as compared with vehicle alone. Furthermore, there was a statistically significant difference between the treatment of T-2143 when compared with the activity of the combination treatment (P = 0.02), demonstrating a superior efficacious response in this model over the payload or combination treatment (Fig. 6B). Results in the LS174T colon cancer model were similar to that of the BT-474 model, where T-2143 was not only able to achieve superior antitumor activity with a TGI of 87% compared with vehicle, but it was statistically significant from the treatment of copanlisib alone (TGI, 31%; P = 0.009) or the combination treatment (TGI, 33%; P = 0.01; Fig. 6C). The superior efficacy of T-2143 compared with copanlisib is also consistent with the ability of T-2143 to inhibit the signaling more potently and more sustainably as compared with companlisib (Supplementary Fig. S4). Together, these results demonstrate the ability of T-2143 to effectively target the tumor and release the payload overtime resulting in superior tumor growth inhibition in comparison with the payload alone.
Lack of tumor-specific delivery of anticancer agents presents a number of challenges to their ability to be efficacious. Poor tumor specificity can lead to high toxicity and lack of sufficient target engagement, all resulting in a low therapeutic window. While ADCs have provided a means of site-specific delivery of anticancer agents, they have traditionally only been used to deliver well-characterized cytotoxic agents to cancer cells and have potential limitations in treating solid tumors (29, 30). In cancer, HSP90 is expressed in an activated and highly complexed state, and it has been demonstrated that HSP90 inhibitors have a propensity to selectively bind with high affinity to the activated from as compared with the latent form of HSP90 found in normal tissue rendering HSP90 inhibitors selective for cancer cells (11, 12). This tumor selective delivery has also been demonstrated for the HSP90 conjugate, PEN-866, which has potent binding affinity and tumor selectivity, without significantly inhibiting HSP90 activity (15). Here we have demonstrated a way to leverage preferential HSP90 binding and selectively deliver anticancer payloads beyond cytotoxic agents, through a set of proof-of-concept experiments.
The PI3K/AKT/mTOR signaling pathway is one of the most important intracellular pathways and is dysregulated in a wide spectrum of human cancers. Although considered a master regulator of cancer, only five pathway inhibitors, temsirolimus (mTOR), everolimus (mTOR), idelalisib (PI3K), alpelisib (PI3K), and copanlisib (PI3K) have been approved to date for clinical use with the majority of approvals in liquid tumors. On the basis of the clear need for novel approaches to achieving deep PI3K–AKT–mTOR pathway inhibition while decreasing toxicity in the solid tumor setting, we utilized the Pentarin platform to generate a HSP90 conjugate containing an inhibitor of the PI3K–AKT–mTOR pathway, T-2143. Given the FDA approval status and preclinical activity of copanlisib, this drug was chosen as a model payload to demonstrate our ability to selectively deliver payloads from this class directly to tumor cells (19).
T-2143 was designed to circumvent the limitations of currently available PI3K inhibitors. The pharmacokinetic and pharmacodynamic profile of T-2143 demonstrated the ability to selectively target and rapidly penetrate into solid tumors, have a significant effect on pathway inhibition, and mitigate a known dose-limiting toxicity of PI3K inhibitors that inhibit the alpha PI3K isoform and suggests the potential to mitigate toxicities of drugs that inhibit this isoform (27). The short plasma circulation time of T-2143 limits overall exposure; however, tumor specific delivery of T-2143 was not compromised. As shown in the experiments presented here, T-2143 was able to rapidly accumulate in tumors and achieve deep PI3K/AKT/mTOR pathway inhibition as evidenced by the inhibition of phospho-AKT. Due to the sustained release of the payload over time, T-2143 is able to achieve deeper pathway inhibition in comparison with copanlisib alone which has been demonstrated to inhibit the pathway early but not in a sustainable fashion as evidenced by the signal rebounding over time (19).
Here, we demonstrated that through optimized linker chemistry, the copanlisib active site of T-2143 was masked while the conjugate remains in circulation, with selective cleavage and release of the PI3K payload occurring within the tumor. Hyperglycemia is a known and potentially dose-limiting side effect of PI3K inhibitors due to their interaction with the insulin-glucose regulatory axis (22). The masking of the payload in T-2143 allowed for the mitigation of a spike in blood glucose in contrast to copanlisib which led to a significant increase in blood glucose levels. Together, these data demonstrate that as a result of the thoughtful design of T-2143, we were able to achieve deep tumor penetration, sustainable pathway inhibition, and minimize toxic side effects seen with other alpha targeting-pathway inhibitors thereby increasing the payloads therapeutic window.
The data presented here also demonstrated that the targeting moiety of T-2143 is necessary for tumor accumulation. The HSP90-binding ability of T-2143 leads to accumulation of the payload in vivo, when compared with the nonbinding control, demonstrating that the HSP90-binding capacity of T-2143 is necessary for intratumoral accumulation of the payload. Further evidence was provided by demonstrating that not only the HSP90 binding was necessary for tumor accumulation, but also that it was sufficient to deliver the PI3K payload and achieve deep pathway inhibition over time. This was validated by persistent drug activity and suppression of PI3K/AKT/mTOR pathway, which translated into remarkable effects on tumor growth inhibition in comparison with copanlisib alone. In all models tested, T-2143 was able to achieve tumor growth inhibition greater than copanlisib alone, demonstrating that site-directed delivery and sustained release of the payload translates into greater activity in vivo. Furthermore, each of the three models tested are reported to have a different PIK3CA mutational status, showing that T-2143 can be effective regardless of the PIK3CA mutational background. Tumor growth inhibition achieved with T-2143 was also superior to treatment with a combination of ganetespib and copanlisib, providing evidence that our improved antitumor activity is unlikely a result of the dosing of both of the payload and HSP90-targeting moiety but rather a direct result of sustained intratumoral release of the payload from HSP90-PI3K conjugate accumulation. Together, the data demonstrates that our miniature drug conjugates are designed to leverage the preferential expression of HSP90 in cancer, deliver payloads directly to tumors, and achieve deep pathway inhibition overtime greater than the payload alone.
In summary, we have presented proof-of-concept experiments demonstrating a unique way of leveraging the selective overexpression and activation of HSP90 in solid tumors to deliver payloads, beyond traditional cytotoxic agents used in other tumor-selective delivery systems. Using this platform, we have the potential to target payloads to selectively accumulate to solid tumors and destroy the tumor cells while minimizing exposure of healthy tissue to the payloads, thereby increasing the payloads therapeutic window.
Disclosure of Potential Conflicts of Interest
B. Moreau is a research leader at Tarveda Therapeutics. A. Cirello is a scientist II at Tarveda Therapeutics. K. Kriksciukaite is an associate director at Tarveda Therapeutics. M. Robinson is a research associate at Tarveda Therapeutics. S. Movassaghian is a senior scientist at Tarveda Therapeutics. T. Cipriani is a scientist 1 at Tarveda Therapeutics. R. Wooster is a CSO and has ownership interest (including patents) in Tarveda Therapeutics. M.T. Bilodeau is a CSO at Tarveda Therapeutics. K.A. Whalen is a senior director at Tarveda Therapeutics. No potential conflicts of interest were disclosed by the other authors.
Conception and design: S. Perino, B. Moreau, B.H. White, T. Cipriani, R. Wooster, M.T. Bilodeau, K.A. Whalen
Development of methodology: S. Perino, B. Moreau, J. Freda, A. Someshwar, S. Movassaghian, T. Cipriani, R. Wooster, M.T. Bilodeau, K.A. Whalen
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S. Perino, B. Moreau, J. Freda, J.M. Quinn, J. Romagnoli, M. Robinson
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S. Perino, B. Moreau, J. Freda, A. Cirello, K. Kriksciukaite, J. Romagnoli, T. Cipriani, K.A. Whalen
Writing, review, and/or revision of the manuscript: S. Perino, J. Freda, A. Cirello, K. Kriksciukaite, S. Movassaghian, T. Cipriani, M.T. Bilodeau, K.A. Whalen
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): S. Perino, J. Freda, M. Robinson, T. Cipriani
Study supervision: S. Perino, M.T. Bilodeau, K.A. Whalen
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.