STX-478 is a wild-type-sparing, oral, CNS-penetrant, novel allosteric inhibitor of mutant phosphatidylinositol-3 kinase α (PI3Kα), targeting a cryptic pocket near the ATP-binding site. PI3Kα plays a central role in many cancers, and has been recently highlighted in coverage of the 2021 Molecule of the Year nominee and PI3Kα degrader, inavolisib. Currently approved PI3Kα modulators are limited by their off-target activities on WT PI3Kα and other kinases, leading to significant side effects including hyperglycemia and rash.  The approved agents like alpelisib also have low brain penetration, limiting their therapeutic utility in treating brain metastases.  

Figure 1. Select examples of clinical PI3Kα modulators.

STX-478 selectively targets PI3Kα helical- and kinase-domain mutations that are prevalent across tumors. Remarkably, STX-478 is devoid of WT PI3Kα inhibition, which is known to cause metabolic dysfunction. STX-478’s selectivity profile allows targeting a range of PI3Kα-mutant cancers including gynecological, and head and neck cancers, while potentially being sufficiently well-tolerated to be used in earlier treatment settings such as the adjuvant setting for breast cancer. STX-478 is currently in Ph. I/II trial (NCT05768139) to evaluate its safety and efficacy in cancer patients with certain PI3Kα mutations, and provides an excellent case study in kinase drug discovery.

Mutant PI3Kα Pathway Activation Drives Cancers but WT PI3Kα Inhibition Causes Hyperglycemia 

PI3Kα, a lipid kinase, is a well-established driver of cancer from numerous preclinical and clinical studies, and is highly mutated in many significant types of cancer including endometrial (>30%), breast (>30%), bladder (>20%), colorectal carcinoma (>17%), and head and neck squamous cell carcinoma (>15%). More than 80% of PI3Kα mutations are located within the helical domain (E542K and E545K) or the kinase domain (H1047X). These hotspot mutations like H1047X drive oncogenic transformation through well-characterized mechanisms, but the wild-type version is important both for healthy tissue homeostasis. Non-selective inhibition of the WT PI3K/AKT pathway blocks insulin action, preventing glucose uptake via the GLUT4 transporter, leading to glycogen breakdown in the liver and hyperglycemia. 

Hyperglycemia Caused by FDA-Approved PI3Kα Inhibitor Alpelisib Limits Its Benefits

Alpelisib (Novartis), a PI3Kα orthosteric inhibitor, was approved in 2019 for the treatment of ER+/HER2-, PI3Kα-mutated, advanced, or metastatic breast cancer in combination with fulvestrant (ER degrader). Alpelisib continues to be studied in the same indication with other ER degraders such as elacestrant. However, the hyperglycemia that comes with inhibition of WT PI3Kα is an undesired side effect, preventing prescriptions in the ~31% of breast cancer patients with poorly controlled, diabetes, and also leads to increased insulin secretion which induces PI3Kα activity and can theoretically limit the efficacy of PI3Kα inhibitors. 

In the SOLAR-1 Ph. III trial (NCT02437318), alpelisib induced on-target hyperglycemia (any grade - 64%), prompting discontinuation in 6.3% of patients. In a real-life assessment of alpelisib effectiveness, over 60% of patients required a dose reduction or permanent discontinuation due to toxicities, including those caused by WT PI3Kα inhibition. Additional effects like rash have been observed with alpelisib (any grade - 36%, prompted discontinuation in 3.2%) as well as with inavolisib and inhibitors of other  PI3K-isoforms. Alpelisib-induced rash was associated with increased eosinophils, but it’s still unclear whether this effect is mediated by PI3Kα inhibition or another off-target. While WT PI3Kα inhibition may not be the only contributor to alpelisib’s toxicities, sparing WT PI3Kα and maintaining selectivity across the kinome is desirable from both safety and efficacy perspectives.

Selective PI3Kα Inhibitors Are Emerging to Overcome WT PI3Kα-Mediated Hyperglycemia - Patent Highlights

Though selectivity for mutant forms of PI3Kα are desirable, identifying selective inhibitors has been challenging. Only a handful of molecules in clinical development are stated to be mutant-selective PI3Kα modulators, including Genentech’s inavolisib and Scorpion’s STX-478.

Figure 2. RLY-2608 and arbitrarily chosen PI3Kα inhibitors claimed by Relay Therapeutics and Loxo Oncology from the patent literature, for illustrative purposes only. MCF10A = Cell-based PIK3CA kinase assay (CisBio Phospho-AKT (Ser473) HTRF assay) to measure the degree of PIK3CA-mediated AKT phosphorylation on MCF10A cells overexpressing hotspot PIK3CA mutations (including H1047R, E542K, and E545K)

Relay Therapeutics has published several patent applications that claim  PI3Kα inhibitors. Arbitrarily chosen examples representative of patent scaffolds include example I-121 from WO2021222556A1, example I-923 from WO2023060262A1, and example I-51 from WO2023039532A1 (Figure 2). RLY-2608 is an allosteric, pan-mutant, and isoform-selective inhibitor of PI3Kα and is currently in Ph. I (NCT05216432). Preliminary results so far have shown no signs of grade 3 hyperglycemia or rash, but have shown promising initial anti-tumor activity in breast cancer patients. Loxo Oncology is the owner of at least three patent applications (invented by Petra Pharma and acquired by Loxo Oncology at Lilly for at least $333M) disclosed chromenone-based PI3Kα inhibitors. Arbitrarily chosen, representative examples include example 366 from US20230286960A1, example 7 from US20230147285A1, and example 251 from US20230096175A1 (Figure 2). LOXO-783 (structure not disclosed) is a mutant-selective and brain-penetrant allosteric PI3KαH1047R inhibitor that is currently in Ph. I (NCT05307705) to evaluate its safety and efficacy in solid tumors. While details about RLY-2608 and LOXO-783 are limited at this stage, making comparison with STX-478 challenging, these examples highlight industry interest in the space and more data about the utility of these molecules to selectively target mutant PI3Kα will be closely watched.

Affinity Selection Mass Spectrometry (ASMS) Leads to A Starting Point for STX-478

The discovery of STX-478 started with an affinity selection mass spectrometry (ASMS)-based screen, which led to the identification of compound 1 whose binding was confirmed by surface plasmon resonance (SPR). Compound 1 was subject to chiral separation, and testing of the separated enantiomers and diastereomers showed that the isomer, compound 2, retained activity over the most common variant PI3KαH1047R (Figure 3). 

Figure 3. The chiral separation of compound 1 revealed the active enantiomer.

Wortmannin, a well-known potent PI3K inhibitor that binds to the ATP-binding site, potentiated compound’s 2 activity, suggesting that compound 2 binds to a different, allosteric site. Compound 2 was notably devoid of WT PI3Kα activity at 100 μM, making it an attractive starting point for SAR exploration.

SAR Exploration Around Compound 2 Led to the Discovery of STX-478 

Different approaches were followed to improve compound 2’s potency on PI3Kα^H1047R. SAR exploration started by seeking alternatives to the methylated linker, which led to the identification of compound 3 with a trifluoromethyl group that demonstrated an excellent mutant activity with acceptable potency on PI3Kα^H1047R (ACS SF 2023).

Table 1. SAR exploration of compound 2 at different regions led to the discovery of STX-478. - = not reported

Surprisingly, an aniline (a structural alert) was explored as an option to replace the cyclohexanol, leading to compound 4 with improved potency on mutant PI3Kα; however, it also increased logD, which prompted replacing aniline with the more polar and less bioactive aminopyrimidine in compound 5. Aminopyrimidine reduced logD but led to decreased metabolic stability in HLM. Different substitutions on the benzofuran were explored to further improve potency against PI3Kα^H1047R and metabolic stability. Finally, a potential metabolic soft spot on the benzofuran was blocked with fluorine to yield the more potent and metabolically stable clinical candidate STX-478, a rare example of a benzofuran-containing clinical compound.

Allosteric Binding Alone Does Not Explain STX-478 Selectivity for PI3Kα^H1047R Over PI3Kα^WT

A co-crystal structure of PI3Kα^H1047R  with STX-478 shows that it occupies a cryptic allosteric site (PDB: 8TGD). The urea forms a bifurcated hydrogen bond (one HBD for two HBAs) with the L911 backbone carbonyl and another suboptimal hydrogen bond with G912 carbonyl that positions the central linker with pyrimidine closer to the protein surface. The benzofuran and trifluoromethyl moieties form hydrophobic interactions with the hydrophobic and aromatic residues at the deepest point of the pocket. Surprisingly, STX-478 binds to PI3Kα^WT at this site as well, so the site alone does not explain the mutant selectivity. Furthermore, no significant differences are seen on comparison of the co-crystal structures of STX-478 bound to the allosteric sites of PI3Kα^H1047R (PDB: 8TGD) and PI3Kα^WT (PDB: 8TDU).

STX-478 Mutant Selectivity is Driven by Differential Binding Kinetics Differences Between  PI3Kα^H1047R and PI3Kα^WT

Subsequent binding kinetics studies (Table 2) suggest that STX-478 mutant selectivity is actually driven by differences in binding kinetics between PI3Kα^H1047R and PI3Kα^WT. STX-478 has both a faster association constant (Kon) for PI3Kα^H1047R versus PI3Kα^WT, and a slower dissociation (Koff) rate from PI3Kα^H1047R versus PI3Kα^WT, leading to a  higher drug-target residence time on the mutant. 

Table 2. Binding kinetics of STX-478 explained its PI3Kα mutant selectivity

These data suggest that the allosteric site is more accessible in the mutant enzyme versus the WT. This provides yet another recent example of binding kinetics playing a critical role in kinase selectivity (see recent case studies on FGFR2 inhibitor RLY-4008 and JAK3 inhibitor ritlecitinib).

STX-478 is More Selective Than Alpelisib Across In Vitro Assays

In vitro profiling shows that STX-478 is potent and more selective for kinase-domain PI3Kα mutants including PI3Kα^H1047L, PI3Kα^M1043X, and the most common variant PI3Kα^H1047R across biochemical and cellular settings (Table 3). 

Table 3. Activity and selectivity data of STX-478 and alpelisib from Cancer Discov. and Scorpion Therapeutics Corporate Presentation. * kinome profiling was conducted at K_m for ATP (Eurofins).

STX-478 showed 14- and 9-fold selectivity for PI3Kα^H1047R over PI3Kα^WT in the biochemical and cell-based assays, respectively. In contrast, alpelisib was almost equipotent against mutant and wild-type PI3Kα in both the biochemical and cell-based assays. STX-478 retained activity, albeit with reduced potency, against the helical domain mutants E545K (IC₅₀ = 71 nM) and E542K (IC₅₀ = 113 nM). In the CTGlo cell viability assay, STX-478 was ~13-fold more potent at inhibiting the proliferation of PI3Kα^H1047R compared to PI3KαWT. Notably, STX-478 was 12-fold less potent at inhibiting cell viability of PI3Kα^WT cells compared to alpelisib. Finally, STX-478 displayed a kinome-wide selectivity over 373 kinases representing ~70% of the human kinome, including other PI3Kα isoforms. The cellular selectivity data helps validate STX-478’s ability to differentiate from alpelisib in the clinic.

STX-478 is Effective in Human Tumor Xenografts while Devoid of Metabolic Dysfunction in Higher Species

STX-478 demonstrated robust efficacy in Cal33 patient-derived xenograft (PDX) models at 100 mg/kg and was more efficacious than alpelisib at the alpelisib dose estimated to match clinical coverage levels of alpelisib in humans (20 mg/kg). The tumor growth inhibition (TGI) of STX-478 at 100 mg/kg QD was similar or superior to 50 mg/kg QD alpelisib in different CDX studies (Table 4). 

Table 4. STX-478 leads to tumor growth inhibition (TGI) across PI3K mutant xenograft models without signs of insulin induction, in contrast to alpelisib which clearly increases insulin levels in nearly all models. TGI and insulin levels reported 1-hour post-dose. NS: not significant. Negative TGI values indicate % regression. HNSCC: head and neck squamous cell carcinoma.

Similar to alpelisib, STX-478 showed robust and durable tumor regression in ER+/HER2- xenograft tumor models in combination with fulvestrant and/or palbociclib, supporting the clinical combinations in humans. Insulin tolerance tests (ITT) and oral glucose tolerance tests (OGTT) were conducted in non-tumor-bearing, BALB/c nude mice following five days of repeat dosing of both alpelisib at 50 mg/kg and STX-478 at 100 mg/kg QD. STX-478 was well-tolerated with no effect on the rate of glucose disposal, body weight, or fasting plasma glucose. In comparison, alpelisib caused overt insulin resistance in most of the xenograft models (Table 4 - △ insulin). Remarkably, STX-478 is able to reach dog exposures exceeding efficacious plasma AUC (mouse CDX) by 10-fold with no effect on serum insulin levels (at 3 mg/kg in dogs, 14*QD). This is in sharp contrast to alpelisib, which interfered with glucose and insulin homeostasis at all doses tested in dogs (0.2, 1.0, 5.0 mg/kg/day). These data suggest that STX-478 has equal or superior anti-tumor activity versus alpelisib with significantly lower risk of PI3Kα^WT-mediated hyperglycemia.

STX-478 has Best-in-Class Potential with High Brain Exposure 

In addition to the aforementioned selectivity, STX-478 demonstrated promising preclinical safety, isoform selectivity, and favorable PK profiles, as well as high in vivo CNS penetration with low efflux (Table 5). 

Table 5. Highlights from STX-478’s preclinical characterization

STX-478 has Excellent Preclinical PK in Higher Species 

STX-478 showed excellent preclinical PK, with exceptionally low clearance and high bioavailability in higher species (dogs and monkeys) (Table 6).

Table 6. STX-478 Preclinical PK. Preclinical brain exposure data were not reported.

STX-478 is a Molecule to Watch in Ph. I/II for Breast Cancer

STX-478 is under investigation in a Ph. I/II trial (NCT05768139) to assess its safety, tolerability, pharmacokinetics, and anti-tumor activity in participants with advanced solid tumors expressing PI3Kα H1047X, E542/E545 mutations or other kinase domain mutations. Part 1 evaluates STX-478 as monotherapy for breast cancer and different solid tumor types, while Part 2 examines combination therapy with fulvestrant in breast cancer patients. The trial is expected to be completed in 2026, and the data will be eagerly watched. With many new emerging treatment options for breast cancer patients, including next-generation ER degraders, a safer agent than alpelisib for use in combination may be highly valuable to patients.

Chemical Synthesis

The patent medicinal chemistry synthesis of STX-478 started with the alkylation of  the phenol of 1-(3,5-difluoro-2-hydroxyphenyl)ethan-1-one with methyl-2-bromoacetate, followed by the intramolecular cyclization in the presence of DBU to yield the benzofuran core. The methyl ester was then reduced with LiAlH₄ to afford the alcohol intermediate. The alcohol was oxidized to the aldehyde using IBX and treated with TMSCF₃ in basic conditions to produce the α-trifluoro alcohol intermediate. The alcohol intermediate was again oxidized with IBX to form the ketone, and a reductive amination was performed with NH₂OH, followed by Raney-Ni to yield the racemic primary amine. This free amine was primed to react with phenyl (2-aminopyrimidin-5-yl)carbamate to generate the final urea compound as a racemic mixture. The resulting enantiomers were separated by chiral HPLC to afford STX-478 as the R-enantiomer.

Figure 4. Synthesis of STX-478

Relevant Patent Documentation

“Urea derivatives which can be used to treat cancer.” WO2022265993A1 (2022).

Comments?

Want to share your scientific insight or contribute to the conversation about this compound? Find out how here.

Updated [11/21/2023]: The section referencing RLY-2608 was updated to reflect the recent disclosure of RLY-2608 by Relay Therapeutics.