Utilization of Metabolite Identification and Structural Data to Guide Design of Low-Dose IDO1 Inhibitors

Abstract

Herein the discovery of potent IDO1 inhibitors with low predicted human dose is discussed. Metabolite identification (MetID) and structural data were used to strategically incorporate cyclopropane rings into this tetrahydronaphthyridine series of IDO1 inhibitors to improve their metabolic stability and potency. Enabling synthetic chemistry was developed to construct these unique fused cyclopropyl compounds, leading to inhibitors with improved pharmacokinetics and human whole blood potency and a predicted human oral dose as low as 9 mg once daily (QD).

Utilization of checkpoint inhibitors such as pembrolizumab, which targets programmed cell death protein 1 (PD-1), has revolutionized cancer treatments over the past few years.^1,2 However, not all patients respond well to the α-PD-1 therapies, partially because of upregulation of other immunosuppressive pathways in the tumor microenvironment.^1,3 Therefore, discovering immunosuppressive targets that can synergize with these paradigm-shifting immunotherapies will be key to increasing the overall response rate to these treatments.

Indoleamine 2,3-dioxygenase-1 (IDO1) is an enzyme involved in the catabolism of tryptophan into N-formylkynurenine, which undergoes further transformations to produce kynurenine and ultimately other downstream catabolites.⁴ IDO1 is a known immunosuppressive enzyme that potentially affects a variety of ailments,⁵ and the overexpression of IDO1 in cancer tissue has been correlated with poor prognosis in a variety of cancers such as melanoma and lung.⁴ IDO1 is believed to lead to an immunosuppressive environment in two different ways. First, the depletion of tryptophan in the tumor microenvironment can lead to a lowering of T-cell activation, as tryptophan is an essential nutrient for T-cell viability. Second, kynurenine is hypothesized to be an immunosuppressive metabolite through its induction of the aryl hydrocarbon receptor (AhR) and subsequent signaling pathways.⁴

The importance of IDO1 as a target in immuno-oncology has led to intense interest from the medicinal chemistry community. For example, heme-binding epacadostat and heme-displacing linrodostat (BMS-986205) are both IDO1 inhibitors that recently went into clinical trials.^6,7 Unfortunately, phase 3 clinical data with epacadostat did not recapitulate the promise of phase 1 and 2 data. The reasons for the phase 3 failure are not fully understood, although insufficient target engagement in the tumor microenvironment and potential induction of AhR by epacadostat are potential leading causes.⁸ Thus, interest in this target persists, and several clinical trials are ongoing with linrodostat in combination with nivolumab in muscle-invasive bladder cancer. We recently reported the discovery of potent heme-displacing diamide IDO1 inhibitors based on hits identified from an ALIS screen,⁹ and optimization of these inhibitors led to the discovery of compounds with excellent overall profiles and low predicted human doses.¹⁰ We also reported the discovery of another structurally distinct series of heme-displacing IDO1 inhibitors with a tetrahydronaphthyridine core, as shown in compound 1 (Figure 1A).¹¹ Despite the promising potency and rat pharmacokinetics (PK) data for 1, the short half-life in dog was a concern since data from this preclinical species were utilized for human PK and dose prediction. Furthermore, general improvements to the half-life would also help to mitigate any safety-related concerns that can arise from high peak-to-trough ratios for compounds with shorter half-lives.¹² Finally, improvements to the ligand efficiency of these compounds were desired too, since this type of enhancement often correlates with higher-quality lead compounds.¹³

Figure 1.

(A) Data on compounds 1 and 2, our recently disclosed IDO1 inhibitors. Units: Cl, mL min^–1 kg^–1; V_dss, L/kg; t_1/2 and t_1/2eff, h. LLE is based on hWB potency. Doses were predicted using allometry targeting IC₇₅ hWB at trough. (B) Strategy to block oxidative metabolism of compound 1.

Initial efforts to improve compound 1 by altering the pyrimidine ring, such as switching the CF₃ group to a methyl group as noted in 2, led to a shorter half-life in rat and lower HeLa cell and human whole blood (hWB) potency, although the lipophilic ligand efficiency (LLE) did improve modestly (Figure 1A). Changing the pyrimidine moiety to other heteroaromatic ring systems did not lead to the desired improvements in PK and potency either.¹¹ Therefore, to design IDO1 inhibitors with an improved profile, a metabolite identification (MetID) study was conducted on 1. In rat, dog, and human hepatocyte incubations, the methyl group at the benzylic position and the tetrahydronaphthyridine ring were identified as positions of oxidative metabolism (Figure 1B).

Our strategy to improve the half-life across species relied on the intentional incorporation of metabolically stable groups at sites of oxidative metabolism of compound 1.^14,15 Designing compounds to lower the clearance without decreasing the volume to a similar extent could have a positive effect on both parameters pertaining to half-life. One way this could be accomplished is by substituting the methyl group in 1 for another small and stable alkyl group. A recent matched molecular pair analysis showed that beyond methyl groups, cyclopropane rings were the alkyl groups least susceptible to oxidative metabolism.¹⁶ The low level of oxidative metabolism of cyclopropane rings is due to the high C–H bond dissociation energy, and other examples demonstrate that the appropriate placement of a cyclopropyl group can lead to increased half-lives.¹⁷ Consequently, replacing the methyl group in 1 with a cyclopropyl moiety became a top priority (Figure 1).

The initial cyclopropyl analogue 3 had worse cell and hWB potency and increased clearance compared with compound 1 (Figure 2) and also suffered from time-dependent inhibition (TDI) of CYP3A4. On the basis of compounds 1 and 2, switching from the CF₃ group to the methyl group on the pyrimidine ring improved the LLE, and we envisioned that a similar change in compound 3 could lead to an enhanced on-target profile,¹³ lessening any CYP-related issues. Gratifyingly, compound 4 bearing the 2-methylpyrimidine group showed improved hWB potency and LLE compared with compound 3. Also, no CYP3A4 TDI was noted with compound 4. Further, compound 4 had an improved half-life in dog compared with 1 (t_1/2 = 3.3 h for 4 vs 1.6 h for 1; Figures 1 and 2). By the use of allometry-based dose prediction targeting IC₇₅ (hWB) at trough, the predicted once daily (QD) dose for 4 was 53 mg.^18,19 Although this was an excellent starting point, the team wanted access to lower-dose inhibitors to better position these assets to potential changes in target engagement while still maintaining a low dose.²⁰

Figure 2.

Key data on compounds 3 and 4. Units: Cl, mL min^–1 kg^–1; V_dss, L/kg; t_1/2 and t_1/2eff, h. LLE is based on hWB potency. Doses were predicted using allometry targeting IC₇₅ hWB at trough. The CYP3A4 TDI ratio was obtained as the IC₅₀ with 30 min preincubation divided by IC₅₀ without preincubation. na, not available; nd, not determined.

To further lower the dose, additional improvements in half-life and hWB potency were desired. In order to make efficient alterations to molecule 4 to realize these improvements, MetID data and structural information were needed. As shown in Figure 3, a crystal structure of compound 4 bound to hIDO1 was obtained. The protein side chains, the tetrahydronaphthyridine ring, and the amides in the A pocket are relatively unchanged in the crystal structure of 4 when it is overlaid with our previously disclosed amide crystal structure (Figure 3A).¹¹ Interestingly, however, the pyrimidine that was utilized as a carbonyl isostere in the C pocket does not make a water-mediated interaction between Arg-343 and His-346.¹¹ However, 4 still maintains a hydrogen bond with the side-chain hydroxyl of Ser-167 and a van der Waals interaction with the side chain of Ala-264.²¹ Also, the cyclopropyl ring in 4 projects into a lipophilic pocket consisting of residues such as Ile-217 and Phe-163 (Figure 3A), affording another opportunity to maximize binding in this pocket. MetID data for 4 showed that metabolism still occurred on the tetrahydronaphthyridine ring in all species, and surprisingly, oxidation of the cyclopropyl group was a predominant metabolic pathway in rat hepatocyte incubations (Figure 3B). This could explain the observed high clearance in rat and the observed large species difference in clearance in dog. On the basis of the structural information it appeared that a methyl group could be added to the cyclopropyl ring to further probe the lipophilic pocket and potentially block metabolism. Also, on the basis of the molecular modeling results shown in Figure 5A, it appeared that a small fused cyclopropane ring system could be added on the tetrahydronaphthyridine ring to reduce oxidative metabolism and fill in an empty space around the ring (Figure 3B).

Figure 3.

Structural and MetID data for compound 4. (A) X-ray cocrystal structure of 4 bound to hIDO1 (yellow, PDB ID 6X5Y) overlaid with the previously reported amide structure (purple, PDB ID 6WPE). (B) Proposal of compounds 5 and 6 to block oxidative metabolism and increase the potency.

Figure 5.

Docked pose and key data on compound 6. (A) Compound 6 (yellow) docked pose overlaid with the crystal structure of 4 (cyan). (B) Units: Cl, mL min^–1 kg^–1; V_d, L/kg; t_1/2 and t_1/2eff, h. LLE is based on hWB potency. The dose was predicted using allometry targeting IC₇₅ hWB at trough.

Gratifyingly, further extension into the lipophilic pocket occupied by the cyclopropyl group by the addition of a methyl group to the cyclopropyl ring led to an 8-fold increase in hWB potency (8 nM for compound 5 vs 64 nM for compound 4) as well as improved LLE (Figure 4).²² In addition, compound 5 displayed an almost 2-fold improvement in the half-life in dog (6.7 h for 5 vs 3.8 h for 4). The combined enhancements led to a 3-fold improvement in human dose prediction for compound 5 (vs compound 4) down to 16 mg QD.

Figure 4.

Key data on compound 5. Units: Cl, mL min^–1 kg^–1; V_dss, L/kg; t_1/2 and t_1/2eff, h. LLE is based on hWB potency. The dose was predicted using allometry targeting IC₇₅ hWB at trough.

The addition of the fused cyclopropane ring on the tetrahydronaphthyridine system increased the hWB potency over 3-fold (16 nM for 6 vs 64 nM for 4; Figure 5). As noted in the docked pose of 6, it appears that the fused cyclopropyl ring points toward a lipophilic area with Tyr-126, which could be one of the reasons for the enhanced potency (Figure 5A). The addition of the fused cyclopropyl ring further improved the dog PK profile as well, resulting in an almost 3-fold increase in half-life (9.0 h for 6 vs 3.8 h for 4). Maintaining good PK while increasing the potency for compound 6 led to a >5-fold improvement in dose (compared to compound 4) down to 9 mg QD.

Enabling chemistry routes to access cyclopropylbenzamides 4, 5, and 6 were key to being able to progress these compounds in an efficient manner. The synthesis of compounds 4 and 5 started with a palladium-catalyzed cyanation reaction of 7 (Scheme 1). With the key pyridyl nitrile 8 in hand, a modified Kulinkovich reaction was utilized to form the desired cyclopropylamine intermediates 9 and 10.²³ Then reaction with an acid chloride followed by deprotection afforded amino intermediates 11 and 12. Initial attempts to install the pyrimidine group via a palladium-catalyzed aryl amination reaction were met with very low yields.²⁴ However, it was discovered that compounds 4 and 5 could be obtained in high yields by means of the acid-catalyzed S_NAr reaction noted below. The (S,S) absolute stereochemistry for compound 5 was assigned using extensive NMR and vibrational circular dichroism (VCD) studies as detailed in Supporting Information (SI) section 10.

Scheme 1. Synthesis of Compounds 4 and 5.

To access the fused cyclopropane ring in compound 6, compound 13 was subjected to a Corey–Chaykovsky reaction (Scheme 2). This reaction proceeded well to afford 14 in good yield.²⁵ Then reduction, a protecting group swap, and oxidation afforded the key pyridine N-oxide intermediate 15. With this in hand, we were able to do a directed cyanation reaction α to the pyridine to afford pyridyl nitrile intermediate 16. Then a modified Kulinkovich reaction was carried out followed by trapping of the amine with 4-fluorobenzoyl chloride to furnish 17. A final deprotection and acid-catalyzed S_NAr reaction afforded the final compound 6. It is worth noting that this is the first report of a fused cyclopropane–tetrahydronaphthyridine system.²⁶ The absolute stereochemistry of compound 6 was unambiguously determined using extensive NMR and VCD studies as detailed in SI section 10.

Scheme 2. Synthesis of Compound 6.

Reaction conditions: (b) BH₃·Me₂S, 60 °C, 98% (16); (c) TFA, anisole; (d) LiHMDS, Boc₂O, 78% over two steps (17); (e) m-CPBA, DCM; (i) TFA, DCM, 0 °C; (j) 4-chloro-2-methylpyrimidine, TsOH, NMP, 65 °C, 50% over two steps.

To further show the utility of these IDO1 inhibitors, a pharmacokinetics/pharmacodynamics (PKPD) study was conducted with compound 4. The PKPD study was conducted in CT26 mice expressing humanized IDO1.²⁷ As shown in Figure 6, administration of a single dose of 4 at 3 mg/kg led to a profound lowering of kynurenine levels in the tumor 4–8 h postdose compared with the vehicle control.²⁷ The magnitude of in vivo kynurenine reduction (∼80%) at the plasma concentration achieved in this study agrees nicely with the in vitro hWB potency for compound 4, providing further support for our human dose prediction strategy.

Figure 6.

In vivo effects of compound 4 on tumor kynurenine levels in CT26-tumor-bearing mice after a single 3 mg/kg dose. Approximately 79% inhibition of kynurenine production was observed at 2 h and 4 h with plasma drug concentrations of 0.21 and 0.09 μM, respectively; 70% inhibition of kynurenine production was observed at 8 h with a plasma drug concentration of 0.03 μM. Kynurenine levels rebounded to baseline at 16 h as the drug concentration dropped to below the limit of quantification.

In conclusion, very low dose projected IDO1 inhibitors were developed. Guided by MetID and structural information, compounds were designed with appropriately placed cyclopropane rings to improve the potency, reduce oxidative metabolism, and increase the half-life. Enabling synthetic routes allowed for efficient access to these cyclopropyl analogues for further testing. The improvements noted above afforded compounds with projected human doses as low as 9 mg QD. In vivo PKPD data showed that compounds of this nature also led to a lowering of tumor kynurenine levels. Overall, the compounds described herein are poised to incontrovertibly test the mechanism of IDO1 inhibition in the clinic.

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsmedchemlett.1c00265.

• Synthetic procedures and analytical data for selected compounds, conditions for biological assays, IDO1 hWB assay protocols, in vivo PK protocol, prediction of human dose for compounds 4–6, PKPD protocol, and absolute stereochemistry determination for compounds 5 and 6 using NMR and VCD (PDF)

Accession Codes

The crystal structure of human IDO1 in complex with 4 has been deposited in the Protein Data Bank with accession code 6X5Y.

Author Contributions

The manuscript was written through contributions of all authors. All of the authors approved the final version of the manuscript.

The authors declare no competing financial interest.