Aminoisoxazoles as Potent Inhibitors of Tryptophan 2,3-Dioxygenase 2 (TDO2)

Abstract

Tryptophan 2,3-dioxygenase 2 (TDO2) catalyzes the conversion of tryptophan to the immunosuppressive metabolite kynurenine. TDO2 overexpression has been observed in a number of cancers; therefore, TDO inhibition may be a useful therapeutic intervention for cancers. We identified an aminoisoxazole series as potent TDO2 inhibitors from a high-throughput screen (HTS). An extensive medicinal chemistry effort revealed that both the amino group and the isoxazole moiety are important for TDO2 inhibitory activity. Computational modeling yielded a binding hypothesis and provided insight into the observed structure–activity relationships. The optimized compound 21 is a potent TDO2 inhibitor with modest selectivity over indolamine 2,3-dioxygenase 1 (IDO1) and with improved human whole blood stability.

One of the critical functions of the immune system is to detect and eliminate malignant cells as they arise and develop antigenic mutations. For tumor cells to survive, the host immune system must be rendered functionally tolerant to otherwise immunogenic tumor-associated antigens. To accomplish this, tumors have been shown to enlist an array of endogenous host regulatory pathways.¹ One of these pathways is mediated by catabolism of the essential amino acid tryptophan (Trp) into kynurenine (Kyn), a process driven by the rate-limiting enzymes indoleamine-2,3-dioxygenase 1 (IDO1) and Trp-2,3-dioxygenase 2 (TDO2).² These enzymes catalyze the oxidation of l-tryptophan to N-formyl kynurenine (NFK), which rapidly converts to kynurenine. Overexpression of these enzymes has been observed in several cancers and is thought to mitigate the immune response toward these tumors.^3,4

The enzymatic functions of IDO1 and TDO2 lead to multiple immune-related responses. IDO1-mediated Trp depletion activates GCN2 kinase,⁵ which in turn phosphorylates and inactivates the transcription factor eIF2α leading to the blockade of protein translation and inhibition of cell proliferation. Kyn accumulation, however, activates the aryl hydrocarbon receptor (AhR),⁶ a transcription factor implicated in cancer progression, invasion, and immune suppression.² The overall result of IDO1 activity is induction of the regulatory T cell (Treg) compartment of the immune system, with concomitant inhibition of the effector T cell (Teff) compartment. IDO1-deficiency (gene ablation or small molecule inhibition) has been shown to boost Teff and decrease Treg function in several autoimmunity and in vivo tumor models.⁷⁻⁹

The physiological function of TDO2 includes the regulation of systemic Trp levels as well as liver Kyn levels. In vitro and in vivo studies suggest that TDO2 plays a role in promoting tumor cell survival and motility, and that TDO2-expressing tumors have a lower degree of T cell infiltrate that allows them to escape immune rejection.^4,10 Genetic knockout or treatment of mice with a tool enzymatic inhibitor of TDO2 activity resulted in increased sensitivity of mice to endotoxin-induced shock, consistent with a role for TDO2 in suppressing inflammation and maintaining immune homeostasis.¹¹ In tumors, TDO2 activity is highly expressed in liver cancer and glioblastoma and upregulated in several other cancers.^4,12 Thus, IDO1 and TDO2 inhibitors may represent a useful therapeutic intervention that restores the immune response against tumors, especially when combined with immune checkpoint inhibitors. Indeed, nonclinical data demonstrates that IDO1 and anti-PD-L1 dual inhibition shows markedly greater efficacy in activating the immune system and inhibiting tumor growth than either treatment alone, with a response associated with more pronounced activation of intratumoral CD8+ T cells, including proliferation and cytokine production.^13,14 Preliminary data of a phase I/II clinical trial in melanoma where the IDO1 inhibitor epacadostat is being tested in combination with anti-PD1 has shown promising results in terms of response rates and depth of responses, compared to anti-PD-1 data.¹⁵

While several classes of IDO1 inhibitors have been disclosed in the literature,¹⁶⁻²⁰ at the onset of our program, TDO2 inhibitors had been explored to a far lesser extent.²¹ Several patent applications describing either TDO-selective or IDO/TDO dual inhibitors have appeared since.²²⁻²⁶ TDO2 inhibition alone or in combination with IDO1 inhibition may allow fuller suppression of the kynurenine pathway across a range of tumor types. Here, we report our efforts in identifying and optimizing aminoisoxazole-based TDO2 inhibitors.

A high-throughput screen of the Genentech compound library identified a number of related aminoisoxazoles as inhibitors of TDO2. Inhibitory potency was measured in both biochemical and cellular assays of IDO1 and TDO2. In the biochemical assays, inhibitors were incubated with recombinant enzymes in the presence of the substrate tryptophan. The production of the N-formylkynurenine (NFK) was detected either in a fluorescence- or mass spec-based assay, and inhibitory potency was expressed as IC₅₀ values. In the cellular assays, inhibitors were incubated with either SW48 cells (in which endogenous TDO2 is highly expressed) or A172 cells (in which IDO1 is highly expressed upon induction by IFNγ). The production of NFK released from cells into the media was detected by a fluorescence-based assay, and the inhibitory potency was expressed as EC₅₀ values. The cytotoxicity of compounds was also measured for the same treated cells with a luminescence-based assay detecting cellular ATP levels. Since TDO2 biochemical potency was poorly predictive of cellular potency for the aminoisoxazole compounds (see Figure S1 in Supporting Information), our medicinal chemistry designs were consequently guided by cellular potency. As a counter screen to identify inhibition as a result of cytotoxicity, we assessed cell viability of SW48 and A172 cells by measuring intracellular ATP concentration. We found that the aminoisoxazoles generally do not affect intracellular ATP concentrations up to 25 μM inhibitor concentration in a cell viability assay, indicating that the cellular potency was not caused by cytotoxicity of the inhibitors.

Aminoisoxazole 1 was the most potent hit found (cellular EC₅₀ of 85 nM, Table 1). We first investigated the importance of the amino group on the isoxazole moiety. Removal of the amino group (2) resulted in complete loss of potency. Addition of a methyl group on the nitrogen atom (3) resulted in a modest ∼8-fold reduction of potency. Replacement with a primary amide (4), secondary amide (5), hydroxymethyl group (6), or addition of a methyl group to the isoxazole 3-position (7) all resulted in significant reduction of TDO2 inhibitory potency.

Table 1. TDO2 Cellular and Biochemical Potencies of the Aminoisoxazole Modifications.

^aCellular potency.

^bBiochemical potency. IC₅₀ and EC₅₀ values are the averages of at least two independent experiments.

We next investigated substitution and replacement of the phenyl moiety of compound 1. Introduction of p-Cl (8), p-F (9), m-Cl (10), or m-methoxy (11) on the phenyl ring reduced TDO2 cell potency (Table 2). Replacing the phenyl group with a cyclopentyl (12) resulted in ∼4-fold reduction of potency, while a cyclohexyl (13) reduced the potency by 18-fold. Since more hydrophilic compounds tend to have better metabolic stability, we were interested in introducing groups that are more polar than the phenyl group. Pyridyl analogs (14 and 15) resulted in more than 10-fold reduction of potency. Between the two thiophenyl regioisomers, cellular potency of 3-thiophenyl 17 was much better than 2-thiophenyl 16. The unsubstituted pyrazole (18) was ∼2.6-fold less potent than the corresponding phenyl analog 1, while addition of a methyl group at either the C3 or C4 of the pyrazole (19 and 20) resulted in further erosion of TDO2 potency. However, introduction of a fluorine substituent at C4 of the pyrazole led to the most potent TDO2 inhibitor 21 within the aminoisoxazole series, with a cellular EC₅₀ of 77 nM.

Table 2. TDO2 Cellular and Biochemical Potencies of Phenyl Ring Substitution or Replacements.

^aCellular potency.

^bBiochemical potency. IC₅₀ and EC₅₀ values are the averages of at least two independent experiments.

Because the isoxazole core was found to exhibit poor stability in whole blood (vide infra), we investigated replacement of the 5-aminoisoxazole moiety with alternate heterocycles. Given that the amino group is required in the 5-aminoisoxazole for TDO2 potency (Table 1), the amino group was retained in all the designed heterocyclic targets. However, all azoles tested, including 4-aminoisoxazole (22), 3-aminopyrazoles (23 and 24), 3-amino-1,2,4-triazole (25), 4-amino-1,2,3-triazole (26), 4-aminoisothiazole (27), 5-aminoisothiazole (28), 5-amino-1,2,3-thiadiazole (29), and 4-amino-1,2,3-thiadiazole (30), were inactive against TDO2 (Table 3). We concluded that the 5-aminoisoxazole is a structurally unique core that is required for TDO2 inhibition.

Table 3. TDO2 Cellular and Biochemical Potencies of Isoxazole Replacement.

^aCellular potency.

^bBiochemical potency. IC₅₀ and EC₅₀ values are the averages of at least two independent experiments.

To better understand the structure–activity relationships and develop a binding mode hypothesis, we constructed a model of 21 bound to human TDO2 (Figure 1) using the X-ray crystal structure, PDB 5TI9.²⁷ Our model predicts a binding mode with interactions similar to those of tryptophan. The primary amine, isoxazole ring nitrogen, and fluoropyrazole moiety in 21 correspond with the ammonium, carboxylate, and indole ring of tryptophan, respectively. In addition, the internal strain energy in this pose was found to be at the global energy minimum using quantum mechanical calculations (Figure S2). The strong interaction between the primary amine and the heme propionic acid in the model of 21 could explain the requirement for a primary or secondary amine as observed in Table 1. Small, hydrophobic replacements of the 4-fluoropyrazole in 21 would fit in the pocket (Table 2), while larger substituents such as the methoxy group in 11 would extend further than the tryptophan indole to experience a steric clash, thus explaining the increase in both IC₅₀ and EC₅₀ of 11. Finally, the model predicts important interactions between the isoxazole heteroatoms and Arg144, which could explain the general loss of potency when either of these atoms were replaced with hydrogen-bond donors or were methylated as in 22–24 and 26–27. The predicted interactions with Arg144 may also explain the measurable biochemical activities in thiazole 28 or thiadiazole 30. Triazole 25 maintained the two hydrogen bond acceptors of the isoxazole, but the nitrogen-linked ring was predicted to adopt an alternate low-energy conformation (Figure S2) compared to 1 or 21. The complete loss of cell potency of all isoxazole substitutions cannot be fully explained by the model. However, this model is consistent with our observation that the aminoisoxazole did not cause perturbation of the maximal UV absorption wavelength when incubated with IDO1 or TDO2 (data not shown), suggesting that it does not bind the heme iron directly. In addition, TDO2 biochemical IC₅₀ values increased when a higher tryptophan concentration was used, suggesting that isoxazoles are competitive with the tryptophan substrate. For example, biochemical IC₅₀ of compound 8 shifted from 0.83 to 4.21 μM when tryptophan concentration was increased from 0.2 to 1 mM. IC₅₀ of compound 10 shifted from 0.50 to 2.94 μM when tryptophan concentration was increased from 0.2 to 1 mM.

Figure 1.

Binding mode hypothesis for compound 21. X-ray structure (PDB: 5TI9) of human TDO2 (purple) is shown with heme (gray), tryptophan substrate (gray), and O₂ (red). A putative bound conformation for compound 21 (orange) was then modeled (see methods in the Supporting Information for details). Distances are in Angstroms.

Selected compounds were also tested in an IDO1 cell-based assay. As shown in Table 4, the compounds showed modest TDO2 selectivity over IDO1, ranging from 6- to 14-fold.

Table 4. TDO2 selectivity over IDO1.

Compound	TDO2 selectivitya (fold)
1	13×
3	12×
17	14×
21	6×

^aCalculated as IDO1 EC₅₀/TDO2 EC₅₀.

As TDO2 is a heme-containing enzyme, we tested the compounds in cytochrome P450 (CYP) inhibition assays and found that the 5-aminoisoxazole analogs did not inhibit common drug-metabolizing CYP enzymes. For instance, the most potent compounds 1, 17, 18, and 21 all had IC₅₀s greater than 10 μM for all the CYP isoforms that were tested (CYP3A4, CYP1A2, CYP2D6, CYP2C9, and CYP2C19).

We also investigated the stability of select compounds in whole blood to predict the in vivo stability (Table 5). The original HTS hit 1, the N-methylated analog 3, and thiophene 17 demonstrated poor stability in rat, dog, and human whole blood. The 3-methylisoxazole analog 7 showed improved whole blood stability but with significantly reduced TDO2 potency. However, pyrazole analog 21 demonstrated improved human whole blood stability while maintaining TDO2 potency. The cause of the observed whole blood instability is not well understood at this time.²⁸

Table 5. Whole Blood Stability^a.

	% remaining	% remaining	% remaining
Compound	rat	dog	human
1	0	0	0
3	0.1	0.1	0.3
17b	3	3	NT
7	50	65	84
21	NT	NT	91

^aCompounds (2 μL, 0.5 mM) were incubated in 100 μL of fresh blood for 2 h unless otherwise stated. Parent remaining is analyzed by LC–MS/MS and reported as a percentage. NT: not tested.

^bIncubated for 30 min.

In summary, high-throughput screening identified a series of 5-aminoisoxazole compounds as potent TDO2 inhibitors. Medicinal chemistry investigations revealed that both the amino group and the isoxazole moiety were important for TDO2 inhibitory activity, and a hypothetical binding mode was developed through molecular modeling, providing insight into the observed SAR. Optimized compound 21 is a potent TDO2 inhibitor with modest selectivity over IDO1, no observed CYP inhibition, and improved human whole blood stability over the original HTS hit. However, the mechanism of whole blood instability for this series, and relationship with structure, remained unclear. As a result, this series was deprioritized for further optimization in light of alternative, more promising chemical matters.

Glossary

ABBREVIATIONS

IDO

indoleamine 2,3-dioxegenase

TDO

tryptophan 2,3-dioxegenase

NFK

N-formyl kynurenine

Kyn

kynurenine

AhR

aryl hydrocarbon receptor

Treg

regulatory T cell

Teff

effector T cell

P450

cytochrome P450

Supporting Information Available

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acsmedchemlett.7b00427.

• Experimental procedures, cell versus biochemical potencies, and quantum mechanical torsional energy analysis (PDF)

Author Contributions

The manuscript was written through contributions of all authors. All authors have given approval to the final version of the manuscript.

The authors declare no competing financial interest.