Feasibility and safety of targeting mitochondria for cancer therapy – preclinical characterization of gamitrinib, a first-in-class, mitochondriaL-targeted small molecule Hsp90 inhibitor

ABSTRACT Mitochondria are key tumor drivers, but their suitability as a therapeutic target is unknown. Here, we report on the preclinical characterization of Gamitrinib (GA mitochondrial matrix inhibitor), a first-in-class anticancer agent that couples the Heat Shock Protein-90 (Hsp90) inhibitor 17-allylamino-geldanamycin (17-AAG) to the mitochondrial-targeting moiety, triphenylphosphonium. Formulated as a stable (≥24 weeks at −20°C) injectable suspension produced by microfluidization (<200 nm particle size), Gamitrinib (>99.5% purity) is heavily bound to plasma proteins (>99%), has intrinsic clearance from liver microsomes of 3.30 mL/min/g and minimally penetrates a Caco-2 intestinal monolayer. Compared to 17-AAG, Gamitrinib has slower clearance (85.6 ± 5.8 mL/min/kg), longer t1/2 (12.2 ± 1.55 h), mean AUC0-t of 783.1 ± 71.3 h∙ng/mL, and unique metabolism without generation of 17-AG. Concentrations of Gamitrinib that trigger tumor cell killing (IC50 ~1-4 µM) do not affect cytochrome P450 isoforms CYP1A2, CYP2A6, CYP2B6, CYP2C8 or ion channel conductance (Nav1.5, Kv4.3/KChIP2, Cav1.2, Kv1.5, KCNQ1/mink, HCN4, Kir2). Twice weekly IV administration of Gamitrinib to Sprague-Dawley rats or beagle dogs for up to 36 d is feasible. At dose levels of up to 5 (rats)- and 12 (dogs)-fold higher than therapeutically effective doses in mice (10 mg/kg), Gamitrinib treatment is unremarkable in dogs with no alterations in clinical-chemistry parameters, heart function, or tissue histology, and causes occasional inflammation at the infusion site and mild elevation of serum urea nitrogen in rats (≥10 mg/kg/dose). Therefore, targeting mitochondria for cancer therapy is feasible and well tolerated. A publicly funded, first-in-human phase I clinical trial of Gamitrinib in patients with advanced cancer is ongoing (ClinicalTrials.gov NCT04827810)


Introduction
The rewiring of metabolic pathways, including in mitochondria, 1 is a ubiquitous cancer trait important for disease progression. 2 Accordingly, exploitation of mitochondrial bioenergetics, 3 buffering of reactive oxygen species (ROS), 4 and inhibition of cell death pathways 5 have been implicated in tumor growth, acquisition of metastatic competence, and resistance to conventional and molecular therapy. 6 On this basis, mitochondria may provide a unique therapeutic target for cancer, 7,8 suitable to disable key pathways for tumor maintenance, regardless of genetic makeup or driver mutation(s). As a result, multiple mitochondrial-targeted anticancer agents, or mitocans, 9 have been developed, and mitochondrial protease ClpP agonist, ONC201, 10 Complex I inhibitor metformin/ME-344, 11 antagonist of mitochondrial translation, tigecycline, 12 and PDH/KDH blocker CPI-613 13 have entered clinical trials in humans.
In addition, there is evidence that mitochondria may be uniquely 'wired' in cancer compared to normal tissues, potentially enabling a broader therapeutic window. 14 One example is the ubiquitous dependence or 'addiction' of tumor mitochondria to a heightened protein folding environment, 15 essential to buffer the proteotoxic stress invariably associated with tumor growth, in vivo. 16 Mechanistically, this is accomplished by the selective accumulation of molecular chaperones, including Heat Shock Protein-90 (Hsp90) and its homolog TNF compared to cytosol by mass spectrometry of isolated subcellular fractions. 30 Once in mitochondria, Gamitrinib triggers acute proteotoxic stress, 20,23 shuts off multipleorganelle functions, including bioenergetics, 19 and delivers potent anticancer activity with IC 50 of 0. 16-29 µM in an NCI 60 cell-line screen, including cell lines representative of common malignancies, such as colon adenocarcinoma (IC 50 , 0.35-29 µM), breast adenocarcinoma (IC 50 , 0.16-3.3 µM), and melanoma (IC 50 , 0.36-2.7 µM). 31,32 Strong anticancer activity was also seen in a combination of regimens of Gamitrinib plus molecular therapy [33][34][35] in models of epithelial and hematopoietic malignancies.
Here, we report the preclinical characterization of Gamitrinib as the first, subcellularly directed antagonist of mitochondrial proteostasis. Based on these findings, a publicly funded, first-in-human phase I clinical trial of Gamitrinib in patients with advanced cancer is currently ongoing (ClinicalTrials.gov, NCT04827810).

Cell lines
Caco-2 and HEK293 cells were obtained from the American Type Culture Collection (ATCC, Manassas, VA) and grown in culture according to the supplier's specifications. In some experiments, HEK293 cells were stably transfected with hERG cDNA, and polyclonal cultures were maintained in the presence of 250 µg/ml geneticin (G418). Cells were maintained in a medium containing 10% fetal calf serum at 37°C in 5% CO 2 and plated on 35 mm dishes at least 24 h prior to the experiment.

Chemical synthesis and analysis of Gamitrinib
The complete chemical synthesis, HPLC profile, and mass spectrometry of Gamitrinib (GA mitochondrial matrix inhibitor) have been described previously. 28 The structure of Gamitrinib is combinatorial and contains the Hsp90 inhibitor 17-AAG linked to triphenylphosphonium as a mitochondrial-targeting carrier. The bulk Gamitrinib powder is stored at −20°C in the dark.

Gamitrinib pharmacokinetics (PK) in rat plasma
Sprague-Dawley rats (n = 3) were administered with a single intravenous (IV) dose of 1 mL Gamitrinib (5 mg/kg) formulated as described above without added microfluidization. Blood samples were collected via lateral tail vein using K 2 EDTA as an anticoagulant at 0.083, 0.25, 0.5, 1, 2, 4, and 24 h post-dose, chilled on ice, and centrifuged at 5000 rpm for 10 min. Aliquots (0.2 mL) of each plasma sample were stored frozen at −20°C. For analysis, 50 µL aliquots of the study samples were mixed with 50 µL of ACN/H 2 O (50:50), extracted by protein precipitation using ACN containing the internal standard Gamitrinib-d 15 (100 ng/mL), and the aliquots of the supernatant were mixed with deionized water for Gamitrinib determination. Batch samples under analysis included a calibration curve, a matrix blank (blank rat plasma), a reagent blank, a control zero (blank rat plasma spiked with internal standard), and duplicate QC samples at three concentration levels (low, medium, and high) in addition to the study samples. Within each batch, the study samples were bracketed by calibration standards or QC samples. The lowest calibration standard served to evaluate system suitability at the beginning of each batch. In all batches, the system suitability samples displayed adequate separation and acceptable peak shapes, retention times, and signal-to-noise ratios. Overall, the concentration of Gamitrinib was measured in 21 rat plasma samples in 1 analytical batch using LC-MS/MS data acquisition on a Shimadzu Nexera LC system coupled with an AB Sciex Triple Quad 5500 mass spectrometer. Chromatograms were integrated using Analyst 1.6.2 software. A weighted (1/x 2 , x = concentration) linear regression was used to generate the calibration curve for Gamitrinib. The concentration of Gamitrinib was calculated using the peak area ratio of analyte to internal standard based on the standard curve. The mean, standard deviation, precision, accuracy, and assay variability were calculated using Microsoft Excel.

Toxicity in rats
Ninety-six male and female Sprague-Dawley rats (6-7 weeks old) were received from Charles River Laboratories (Raleigh, NC) catheterized via a femoral vein prior to arrival. Animals were acclimated in individual stainless-steel cages with access to water and Certified Rodent Diet #2014 C (Envigo RMS) ad libitum for 6 d prior to study initiation. Environmental controls were set to maintain a temperature range of 20 to 26°C, a relative humidity range of 30 to 70%, eight or greater air changes/h, and a 12-h light/dark cycle. Animals were infused with sterile isotonic (0.9%) saline for a minimum of 4 d -(0.20 mL/h for males or 0.15 mL/h for females) during the predose phase. At initiation of dosing, animals were 8 to 9 weeks old, and body weights ranged from 283 to 341 g for males and 188 to 238 g for females. Gamitrinib formulated as indicated above was administered by IV infusion on d 4,8,11,15,18,22,25, and 29 of the dosing phase at 1, 10, 25 mg/kg/ dose and dose volume of 5 mL/kg. The vehicle control administered with the same schedule contained 2.5% DMSO, 0.125% polysorbate 80, 0.031% lecithin, 1.25% sucrose and 4.375% dextrose. Animals were checked twice daily throughout the duration of the study for mortality, abnormalities, and signs of pain or distress. Detailed observations were conducted for each animal up to two times during the predose phase and for each toxicity animal prior to dosing on d 1, 8, 15, 22, and 29 of the dosing phase and on d 1, 7, and 14 of the recovery phase. Blood samples were collected on d. 1, 4 and 29 of the dosing phase 5 min, 1 h and 24 h post-dose, maintained on chilled cryoracks and centrifuged within 1 h of collection. Tissue samples harvested from each animal were embedded in paraffin, sectioned, and slides were prepared and stained with hematoxylin and eosin.

Toxicity in beagle dogs
Eighteen male and female (5-6 mo old) purebred beagle dogs (Cumberland, VA) were acclimated for 47 d (20°C to 26°C, relative humidity 30-70%, 10 or greater air changes/h, and a 12-h light/dark cycle) prior to study initiation. Animals were given water ad libitum and Certified Canine Diet #5007 (PMI Nutrition International Certified LabDiet®) for 4 to 5 h each day. At study initiation, animal body weights ranged from 8.6 to 11.5 kg for males and 6.0 to 9.4 kg for females. At least 1 week prior to initiation of dosing, animals were fasted overnight, anesthetized, and a catheter attached to a subcutaneous vascular access port was surgically implanted into a jugular vein. Animals were acclimated to infusion jackets prior to catheter implantation surgery. Animals were infused with sterile saline at a dose rate of 5 mL/h when the vascular access ports were accessed and when animals were connected to the infusion system (except when dosed with the test or vehicle control article). The patency of each catheter was checked as needed. Prednisolone tablets were administered orally at 2 mg/kg the night prior to dosing. Diphenhydramine (5 mg/kg; IM injection 0.1 mL/kg, 50 mg/mL) was administered prior to dosing and post the start of infusion (d 1 only). Male and female dogs received dose levels of Gamitrinib of 1.25, 3.33 and 6.25 mg/kg/dose formulated as indicated above during a 1 h infusion on d 1, 8,11,15,18,22,25,29,32, and 36 at a volume of 2 mL/kg. The vehicle control was as above. During dosing on d 1, multiple animals, including controls, were noted with clinical observations of swollen head/body, hypoactivity, twitching, red skin, and/or vocalization due to vehicle components. Dosing was stopped, and 23 of 24 animals were administered additional diphenhydramine, 11 were administered Flunixin meglumine and buprenorphine, and 6 were administered acepromazine.
Symptoms subsided in all animals after completion of these interventions. As a result of these observations, the polysorbate levels, which were increased from 0.025% (w/v) to 0.125% (w/v) to minimize precipitation, were returned to the original concentrations; 2 mg/kg prednisone was administered orally the night prior to dosing; and the diphenhydramine pretreatment dose was updated to 5 mg/kg 15

Statistical analysis
Levene's test was used to test for equality of variances between groups. Where Levene's test was not significant (P > .05), ANOVA was conducted; where Levene's test was significant (P ≤ .05), a rank transformation was applied before the ANOVA was conducted. Where the group effect from the ANOVA was significant (P ≤ .05), comparisons between each treated group and the control were made using Dunnett's t-test. If the ANOVA was not significant (P > .05), no further analyses were conducted.

Chemical synthesis and formulation development of Gamitrinib
The chemical synthesis of Gamitrinib containing the Hsp90 inhibitor 17-AAG linked to the mitochondrial-targeting carrier, triphenylphosphonium via a hexylamine linker has been described. 28 Clinical-grade (GMP compliant) Gamitrinib synthesized as described in Supplementary Figure S1A 28 has the chemical formula C 52 H 65 F 6 N 3 O 8 P 2 (>99.5% purity by UPLC), is purple solid (TM.795) and crystalline by X-ray powder diffraction, with a molecular weight of 1036.03. A 500 MHz 1 H NMR spectrum (DMSO-d6), 125 MHz 13 C NMR spectrum (DMSO-d6) and 282 MHz 19 F NMR spectrum (DMSO-d6) are all consistent with structure. The water content is 0.7% (Karl Fischer analysis), and the residual solvent concentrations (methanol, DCM, MTBE, and DIPEA) are all below limit of quantification (BLOQ, <3000, <600, <5000, and <3000 ppm, respectively). Gamitrinib is formulated using a three-step dilution process described in the Materials and Methods section. A sterile, Gamitrinib Injectable Suspension (GIS) is prepared for clinical use using microfluidization with the schematic flowchart shown in Supplementary Figure S1B. The resulting GIS has average particle size of 154 nm and D(0.9) size of 229 nm, pH 6.0. When stored frozen at −20°C, the GIS shows no significant changes in stability or particle size distribution upon analysis at 1, 2, 4, 8, 12, and 24 weeks after manufacturing (Table 1).
To further characterize the potential cardiac toxicity of Gamitrinib, electrocardiography studies were carried out on beagle dogs administered IV Gamitrinib at dose levels of 1.25, 3.3, and 6.25 mg/kg/dose twice weekly for 36 d plus a 14-d recovery period (Supplementary Figure S3). In this analysis, one out of five male dogs administered Gamitrinib at 6.25 mg/kg/dose exhibited a small (7%) prolongation of QTc interval (17 msec) on d 32 of the dosing phase, which reversed during the recovery phase. No Gamitrinib-related prolongation of QTc interval was observed in female dogs administered 6.25 mg/kg/dose or in both sexes administered 1.25 or 3.33 mg/kg/dose (Supplementary Figure S3). No Gamitrinib-related ECG changes in PR interval, QRS duration, QT interval, or heart rate were observed on d 32 of the dosing phase in animals administered 1.25, 3.33, or 6.25 mg/kg/dose or on d 11 of the recovery phase in animals administered 6.25 mg/kg/ dose. No other rhythm abnormalities or qualitative ECG changes were observed (Supplementary Figure S3).

PK studies
After the IV administration (5 mg/kg) to Sprague-Dawley rats (n = 3), the mean Gamitrinib C max was 1175.807 ng/mL (Figure 2(a)), with a mean volume of distribution at steady state (V ss ) of 65.471 L/kg, medium to slow clearance at 85.656 ± 5.856 ml/min/kg, and mean terminal-phase half-life (t 1/2 ) of 12.25 ± 1.55 h ( Table 2). Mean AUC 0-t and AUC INF values were 783.199 and 976.002 h•ng/mL, respectively ( Table 2). Gamitrinib metabolism in rats did not generate detectable levels of 17-(amino)-17-demethoxygeldanamycin (17-AG) (Figure 2(b), Supplementary Table S2), a key metabolite of 17-AAG processing, in vivo. 36 The IV administration of Gamitrinib to Sprague-Dawley rats at dose levels of 1, 10, or 25 mg/kg/dose twice weekly for 29 d resulted in increased C max values from 1 to 25 mg/kg/dose followed by a bi-exponential decline (Figure 2(c)). CL SS values ranged from 84.83 to 131.33 mL/min/kg and V SS values from 12.2 to 90.0 L/kg for d 4 and d 29. Gamitrinib C max and AUC 0-24 values were similar on d 4 and d 29, indicating no drug accumulation after multiple doses. Accumulation ratio values ranged from 0.045 to 1.04 for C max and from 0.344 to 1.33 for AUC 0-24 .

Plasma protein binding, stability, microsome clearance, and intestinal penetration
Gamitrinib was heavily bound to plasma proteins (99.3 ± 0.07%) with an average free fraction of 0.  Table S1).

Toxicity in Sprague-Dawley rats
Male rats administered Gamitrinib IV (1 h infusion) at dose levels of 1, 10, or 25 mg/kg/dose twice weekly on d 1,4,8,11,15,18,22,25, and 29 (dosing phase) exhibited a small, fully recoverable and not adverse reduction in mean body weights at 10 (−5.5%) or 25 (−5.7%) mg/kg/dose (Figure 3). Gamitrinibrelated clinical observations involved animals administered ≥10 mg/kg/dose and included inguinal swelling, piloerection, hypoactivity, and sensitivity to touch at the infusion site. This is correlated with microscopic findings of mixed cell inflammation at the catheter/infusion site, which increased in incidence and/or severity in animals administered ≥10 mg/kg/dose (both sexes) and persisted through recovery. Alterations in clinical chemistry parameters, such as mildly to moderately higher neutrophil, and platelet (Plts) counts, minimally prolonged partial thromboplastin time (PT), lower albumin, higher globulin, and alkaline phosphatase concentrations were observed at the highest Gamitrinib dose level tested (Figure 3(a)) and likely related to inflammation, accompanied by histologic evidence of spleen and liver extramedullary hematopoiesis. Minimally to mildly higher serum urea nitrogen (UN) and creatinine concentrations in animals receiving Gamitrinib at 25 mg/kg/dose (Figure 3(a)) correlated with increased kidney weight and microscopic findings of tubular degeneration/regeneration (Figure 3(b)), which persisted to the end of the recovery phase. No effects on urinalysis or ophthalmic changes were identified. Gamitrinib-related mortality due to severe inflammation and marked hemorrhage at the infusion site occurred in two males, one female administered 25 mg/kg/dose and one male administered 10 mg/kg/dose. All other toxicity animals survived to their scheduled sacrifice. Gamitrinib-related mortality also occurred in two toxicokinetic females administered 10 mg/kg/dose and one toxicokinetic female administered 25 mg/kg/dose. Based on these findings, the non-observed adverse effect level (NOAEL) of Gamitrinib in rats is 1 mg/kg/dose, corresponding to C max and AUC 0-24 values of 87.1 ng/mL and 174 ng•h/mL, respectively, on d 29 of dosing. Due to non-severely toxic effects or mortality in fewer than 10% of the animals administered 10 mg/kg/ dose, the severely toxic dose in 10% of the animals (STD 10) is 10 mg/kg/dose, corresponding to C max and AUC 0-24 values of 311 ng/mL and 1300 ng•h/mL, respectively, on d 29.

Toxicity in beagle dogs
Male and female beagle dogs administered IV Gamitrinib at dose levels of 1.  Table  S3). Bone marrow and liver parameters were unremarkable in all group levels, and only a trend of increased serum urea nitrogen and creatinine was observed in animals (both sexes) receiving the highest dose level of Gamitrinib of 6.25 mg/kg/ dose ( Figure 4). This correlated with microscopic findings of slight to moderate kidney tubular degeneration/regeneration, which was reversible during the recovery period. In addition, similar findings were present in one recovery sacrifice control male, making their relationship to Gamitrinib uncertain. No Gamitrinib-related changes in organ weight were observed and electrolyte, calcium, and phosphorus levels were unchanged in the various groups ( Figure 4). Catheter and infusion site findings were similar in control and Gamitrinib-treated animals. Based on these findings, the NOAEL of Gamitrinib in dogs was 3.33 mg/kg/dose (C max , 560 ± 404 ng/mL; AUC 0-24 , 1740 ± 713 ng•h/mL on d 36 of dosing, both sexes). In the absence of effects on the overt well-being of the animals and evidence of reversibility of Gamitrinib-related findings, 6.25 mg/kg/dose is considered the highest non-severely toxic dose (HNSTD). This dose level corresponds to C max and AUC 0-24 values of 1260 ± 556 ng/mL and 3290 ± 1090 ng•h/mL, respectively (both sexes), on d 36 of the dosing phase.

Discussion
In this study, we have shown that Gamitrinib, a first-in-class mitocan inhibitor of mitochondrial protein folding, can be synthesized as a clinical-grade material, formulated as a sterile, stable injectable suspension and administered IV to two animal species for up to 36 d. In line with its unique mechanism of action of subcellular organelle targeting, Gamitrinib drug-like properties differ from those of nonmitochondrial-targeted 17-AAG (Tanespimycin) with slower clearance, longer half-life, and unique metabolism without generation of 17-AG. Finally, prolonged IV administration of Gamitrinib is essentially unremarkable in beagle dogs and causes occasional inflammation at the infusion site and modest alterations of kidney function in Sprague-Dawley rats.
Although rewiring of metabolic pathways is a key tumor driver, 1 and mitochondria provide an actionable, multifunctional therapeutic target, 7,8 the (pre)clinical development of mitocans is mostly in infancy. 9 The most visible conclusion of these studies is the encouraging preclinical safety of Gamitrinib, the first mitocan to disrupt a Hsp90 mitochondrial proteome in cancer. 19 When administered to dogs, concentrations of Gamitrinib up to 12.8-fold higher than therapeutically active doses in mice (10 mg/kg) did not elicit the extensive alterations of bone marrow, liver, and gastrointestinal tract functions seen with non-mitochondrial targeted 17-AAG 37 and its derivatives, 17-DMAG, 38 and IPI-504. 39 The only reportable observation in our study was that the highest dose level of Gamitrinib of 6.25 mg/kg caused only modest elevation of urea nitrogen and creatinine concentrations, with microscopic evidence of kidney tubule degeneration/regeneration, which was reversed during the 14-d recovery phase. The potential cardiac liability of Gamitrinib, prompted by the role of Hsp90 in hERG protein folding 40 was also unremarkable. Gamitrinib IC 50 for hERG conductance in patch-clamp studies (3.5 µM) was up to 20-fold times higher than the IC 50 of inhibition of tumor growth observed in an NCI 60-cell line screen, 28 and only 1 out of 10 animals exhibited a small, 7% prolongation of QTc interval (17 msec) fully reversed during the recovery phase. In the rat study, Gamitrinib concentrations (25 mg/kg) up to 5-fold higher than the efficacious antitumor dose in mice, produced only occasional inflammation at the infusion/catheter site, accompanied by mild elevation of urea nitrogen and evidence of kidney tubule degeneration/ regeneration.
The molecular basis for the significantly more favorable safety of Gamitrinib compared to non-mitochondrialtargeted Hsp90 inhibitors [37][38][39] remains to be elucidated. Previous data have suggested that Hsp90 may have greater affinity in tumors compared to normal tissues, 14 which when combined with the selective accumulation of TRAP1 and Hsp90 in tumor mitochondria, 17 may make the therapeutic inhibition of mitochondrial chaperones 15 more efficient in cancer than normal tissues. It is also possible that the rapid intracellular transfer of Gamitrinib followed by its stable  27 In addition to improved safety, Gamitrinib exhibited unique drug-like properties compared to 17-AAG. These included slower clearance (85.65 ± 5.85 mL/min/kg), much longer terminal-phase half-life (t 1/2 , 12.25 ± 1.55 h), and no generation of 17-AG, a key metabolite of 17-AAG processing. 36 The structural basis for the increased Gamitrinib stability in vivo is currently not known. However, these data are reminiscent of the greater stability of 17-DMAG compared to 17-AAG, 41 and it is possible that the presence of the triphenylphosphonium side chain reduces the oxidative metabolism of Gamitrinib, 41 causing slower drug clearance and longer t 1/2 .
Despite the promise of Hsp90 as a hub of many oncogenic pathways, 42 the clinical response to GA or non-GA Hsp90 antagonists has been disappointing. 27 Although feasible, Hsp90-directed therapy has shown marginal, if any, patient responses as a single agent or in combination, hampered by unacceptable toxicity, including treatment-related deaths. [37][38][39] These negative results have brought into question the validity of Hsp90 as a therapeutic target. Our data, combined with the preclinical anticancer activity of Gamitrinib, [31][32][33][34][35] suggest a different scenario, where only a few, selected chaperone functions out of likely hundreds, are exploited for tumor growth. 42 In this context, "untargeted," global Hsp90 inhibition as previously pursued in the clinic, 27 may exacerbate organ and tissue toxicity while paradoxically narrowing the therapeutic window. The pool of Hsp90 compartmentalized in tumor mitochondria, 19 which escapes inhibition by "untargeted" Hsp90 antagonists, 28,30 may provide one such key chaperone function exploited for tumor growth: preserving a multifunctional mitochondrial proteome highly vulnerable to oxidative and proteotoxic stress.
Based on the findings presented here, the feasibility and safety of Gamitrinib are currently being evaluated in a firstin-human, phase I clinical trial in patients with advanced cancer as part of a publicly funded academic effort (NCT04827810). In addition to validating mitochondrial proteostasis and Hsp90 as actionable therapeutic targets, the clinical development of Gamitrinib suggests that other mitochondrial functions exploited in cancer, such as onecarbon metabolism 43 and DNA transcription, 44 may also be druggable using a similar, "targeted" approach of subcellular drug delivery. 7,8