Discovery of Alternative Binding Poses through Fragment-Based Identification of DHODH Inhibitors

Abstract

Dihydroorotate dehydrogenase (DHODH) is a mitochondrial enzyme that affects many aspects essential to cell proliferation and survival. Recently, DHODH has been identified as a potential target for acute myeloid leukemia therapy. Herein, we describe the identification of potent DHODH inhibitors through a scaffold hopping approach emanating from a fragment screen followed by structure-based drug design to further improve the overall profile and reveal an unexpected novel binding mode. Additionally, these compounds had low P-gp efflux ratios, allowing for applications where exposure to the brain would be required.

Acute myeloid leukemia (AML) is an aggressive hematological cancer with an estimated >20 000 new cases in the U.S. in 2023.¹ This cancer of the blood and bone marrow is characterized by an accumulation of myeloblasts which lack the ability to differentiate into healthy hematopoietic cells and results in uncontrolled proliferation.² Improved therapies are of high interest as AML can rapidly progress if left untreated, leading to low survival rates (5 year, ∼30%) associated with current standards of care.

Seminal reports by Sykes et al.^3,4 and others⁵⁻⁸ demonstrated that inhibition of dihydroorotate dehydrogenase (DHODH) in AML cells induced differentiation and apoptosis and reduced proliferation in both in vitro and in vivo AML models. DHODH is a mitochondrial enzyme responsible for catalyzing the fourth of six steps in the de novo pyrimidine synthesis converting dihydroorotate to orotate.⁹⁻¹¹ It is ubiquitously expressed and is not known to be mutated or overexpressed in cancer. However, the importance of the de novo pyrimidine pathway is increased in rapidly proliferating or malignant cells to satisfy the increasing demand for nucleic acid precursors.

With the recent discovery of the role of DHODH in AML there has been a renewed interest in DHODH as a hematologic drug target.¹²⁻¹⁵ In particular, AML has been a target indication for DHODH inhibitors with compounds such as Brequinar,¹⁶ BAY2402234,¹⁷ PTC-299,¹⁸ and ASLAN-003¹⁹ progressing to clinical trials. Herein is described continued efforts^20,21 in the identification of potent DHODH inhibitors through a scaffold hopping approach starting from a fragment hit and using structure-based drug design (SBDD) to further improve the overall profile that also revealed an unexpected novel binding mode distinct from the original fragment hit.

Prior work by our group applied a virtual screening campaign alongside SBDD to identify potent DHODH inhibitors with a chemically distinct N-heterocyclic pyridyl core (Figure 1).¹⁵ From this effort, compound 1 was profiled as a potent lead in biochemical and cellular assays. Compound 1 displayed favorable physicochemical properties and showed efficacy in vivo in a preclinical model of AML. However, one notable feature of the compounds in this series was that they were all substrates for the efflux transporter P-glycoprotein (P-gp) with an efflux ratio above 30. Compounds with higher efflux ratios can display lower oral bioavailability as well as reduced potency against P-gp-expressing cell lines. Furthermore, P-gp substrates cannot be used for applications that require brain penetration such as glioblastoma where DHODH inhibition appears to be a promising metabolic vulnerability.^22,23 Building on the learnings from this initial work, a fragment screen was explored to identify alternative binders and address the P-gp liability.

Figure 1.

Prior work progressing a virtual screening hit to lead 1.

Fragment-based drug discovery (FBDD) is now a widely used approach to identify small molecule binders of a specific target,²⁴⁻²⁷ with several FDA-approved therapeutics originating from a fragment screen.²⁸⁻³⁶ These fragments typically start off with improved lipophilic ligand efficiencies (LLEs) and occupy better physicochemical space that can then be optimized to a more potent inhibitor or incorporated into a known scaffold to improve activity.³⁷ In this work, an internal fragment screen using surface plasmon resonance (SPR) revealed compound 2 as a weak binder (K_D = 130 μM) with an IC₅₀ of 82 μM in the hDHODH enzymatic assay (LLE = 1.5).³⁸ The structure of 2 bound to DHODH was solved by cocrystallization and showed the fragment binding in the same ubiquinone site as the lead series with clear density for the acetanilide portion of the fragment.³⁹ The amide carbonyl of the fragment forms a hydrogen bond to Y356, similar to that observed with the carbonyl of the triazolone ring of compound 1. Interestingly, 2 engages in a water mediated interaction with R136, a residue not observed to engage with the triazolone moiety of 1 (Figure 2).

Figure 2.

Crystal structure of 2 (PDB entry 8VHM) and lead 1 (PDB entry 8DHG) bound to DHODH.

This unique interaction observed in the crystal structure prompted an investigation to incorporate different moieties into a previous lead series typified by compound 1 to maintain potency while addressing other issues such as P-gp efflux. To mitigate any synthetic challenges that could arise with the aminopyridine core and better mimic the fragment hit, phenyl scaffold I-1 (see Supporting Information) was chosen to explore the structure–activity relationship (SAR) of various amides and ureas in a rapid manner. The amide series compounds 4–8 were prepared via reaction of the readily available acid chlorides with intermediate I-1 (Scheme 1). Compound 10 was prepared via amide bond formation with the requisite ester building block using trimethylaluminum followed by deprotection of the TBS group. Subsequent hydrogenation afforded compound 9 in a 66% yield. Urea 11 was obtained in 16% yield using the corresponding isocyanate. While compound 18 used a slightly modified route,⁴⁰ the majority of the urea compounds were prepared via carbamate and displacement with the corresponding amine.

Scheme 1. Synthesis of Compounds: (a) Acid Chloride, Et₃N, DCM, −10 °C; (b) Phenyl Chloroformate, Pyridine, DCM, 0 °C to r.t., 12 h, 89%; (c) Et₃N, DCM, r.t.; (d) Isocyanatoethane, Et₃N, DCM, 0 to 40 °C, 12 h then DCE, 80 °C, 12 h, 16%; (e) AlMe₃, DCM, 10 to 60 °C, 12 h, 88%; (f) TBAF, THF, 15 °C, 1 h, 71%; (g) 10% Pd/C, EtOAc, 15 psi, 15 °C, 2 h, 66%.

Based on fragment hit 2, acetamide 4 was prepared and tested in the hDHODH enzymatic assay and cellular proliferation assay with MOLM-13 AML cells,⁴¹ resulting in an IC₅₀ of 100 nM and 370 nM, respectively (Table 1). As a reference, compound 3 containing the triazolone and pyridyl core was 0.24 and 0.16 nM in the enzymatic and cellular assay, respectively. Though replacement of the triazolone resulted in a significant potency loss, this result demonstrated that alternative substituents were tolerated at that position and provided an opportunity for further SAR development. Simply elongating the methyl to an ethyl (5) led to a 10-fold increase in cellular potency (IC₅₀ = 35 nM) while the n-propyl analog 6 and cyclopropyl analog 7 resulted in roughly equipotent compounds in terms of cellular activity.⁴² A branched alkyl substituent (8) did not provide any advantage, remaining equipotent to the linear alkyl chains. Exploring substitution possessing a primary ROH to potentially engage with T360 through a water mediated H-bond led to hydroxy-substituted compound 9. While this resulted in a loss of activity in the cellular assay (209 nM), rigidifying through incorporation of an olefin in the form of 10 regained some of the loss (64 nM). Notably, compound 6 maintained good permeability (Papp A > B (+inh) = 22 cm/s × 10^–6); however, human and mouse microsomal stability was extremely poor with a t_1/2 of 31 and <4 min, respectively (data not shown).

Table 1. Amide Substitution SAR.

^aValues represent the average of n ≥ 2 experiments. Interassay variability <30%.

With moderately potent amide compounds identified, the urea comparators were next evaluated (Table 2). Gratifyingly, compound 11, the urea equivalent of compound 6, maintained equal cellular activity, demonstrating that urea insertion was viable at this position. The N,N-dimethyl and methyl,ethyl urea derivatives 12 and 13 provided a ∼25-fold boost in activity, yielding compounds with 2.0 and 0.82 nM activity in the cellular assay, respectively. The urea comparator of compound 8, diethyl urea 14 revealed an exceptionally potent compound with an IC₅₀ of 0.63 nM in the cellular assay. This subnanomolar activity is now within the same range as parent compound 3 showing how a simple fragment inspired urea was equivalent in potency to the more chemically complex triazolone moiety. Keeping one ethyl group of the urea constant, other functionality was explored to further investigate effects on potency and address the poor metabolic stability observed for these compounds. Cyclopropyl (15), difluoroethyl (16), and hydroxyethyl (17) provided single digit nanomolar compounds in the cellular assay; however, only modestly improved human and mouse metabolic stability was observed. Furthermore, disrupting the anticipated water-mediated R136 interaction suggested by Figure 2 by capping the NH of the urea with a methyl (18) did lead to an expected loss of potency and further erosion of microsomal stability. Notably, while triazolone 3 had an efflux ratio of >30, these urea derivatives were not P-gp substrates and encouraged us to explore the SAR further.

Table 2. Urea Substitution SAR with Alkyl Amines.

^aValues represent the average of n ≥ 2 experiments. Interassay variability <30%.

^bH/M LM = human/mouse liver microsomes.

^cEfflux ratio = Papp B > A/A > B.

Subsequently, the effects of cyclizing the urea substituents by tethering them together to form a saturated ring was explored (Table 3). This modification seemed to be well-tolerated with simple pyrrolidine compound 19 revealing similar cellular potency (IC₅₀ = 1.3 nM) as compared to the diethyl compound 14. Exploring further substitution on the pyrrolidine ring such as difluoro 20 and methoxymethyl 21 led to a slight loss in potency. However, we were pleased to observe that polar functionality was tolerated. Consequently, efforts to further reduce lipophilicity and improve metabolic stability led to the synthesis of 22, which was equipotent to the unsubstituted pyrrolidine yet still only provided modest stability (t_1/2 = 56 min (h), 45 min (m)). The 2-substituted azetidine analogue 23 maintained activity and further reduced the clogP; however, both human and mouse microsomal stability were reduced. Substitution of the azetidine at position 3 in 24 instead led to a marked improvement in metabolic stability but at the expense of biological activity. Last, the pyrrolidinone 25 maintained potency while also exhibiting an improvement in metabolic stability with human and mouse liver microsomes t_1/2 of 62 and 106 min, respectively. Moreover, these rings with substitution at the 2-position (22, 23, and 25) continued to display low efflux ratios.

Table 3. Urea Substitution SAR with Cyclic Amines.

^aValues represent the average of n ≥ 2 experiments. Interassay variability <30%.

^bH/M LM = human/mouse liver microsomes.

^cEfflux ratio = Papp B > A/A > B.

^dcLogP calculated from BioByte.

With the improved metabolic stability observed for compound 25, the mouse pharmacokinetic profile was obtained (Table 4). Compound 25 demonstrated high oral bioavailability (>100%), a large volume of distribution, and a half-life of 3 h. The clearance was also relatively low (∼20% of hepatic blood flow). Additionally, the plasma protein binding (PPB) was high with 1.2% and 0.7% free fractions in the human and mouse, respectively. Overall, this shows that good pharmacokinetic properties and high potency can be achieved with derivatives devoid of the triazolone moiety. Furthermore, these ureas showed low efflux ratios, thus allowing for potential applications in areas that may require distribution into the brain.

Table 4. Mouse Pharmacokinetics of 25^a.

compound	F (%)	t1/2 (h)	Cl (mL/min/kg)	Vdss (L/kg)	PPB h,m (% free)
25	>100	3	17.7	3.6	1.2, 0.7

^aC57 mice (n = 6). Oral dose 10 mg/kg, PEG400/water (70:30) vehicle; IV dose = 2 mg/kg, PEG400/water (70:30) vehicle.

In FBDD, the optimized inhibitor incorporating the fragment moiety is often assumed to conserve the same geometric interactions as the initial hit.⁴³⁻⁴⁷ To confirm the anticipated mode of binding for the urea substituted analogs, the structure of compound 17 bound to DHODH was solved via X-ray crystallography (Figure 3). Surprisingly, as compared to fragment hit 2 and previous lead 1 (Figure 2), the carbonyl moiety did not make the presumed interactions with Y356 but instead rotated ∼180° to re-engage the side chain of Q47. The oxygen of the hydroxyethyl substituent on the urea makes a hydrogen bond to the side chain of T360 and an intramolecular hydrogen bond to the nitrogen of the urea. Considering the urea moiety as a simple extension of the amide from the fragment, this novel binding pose was unexpected and now provides a new opportunity for additional SAR exploration using structure-based drug design.

Figure 3.

Crystal structure of 17 (PDB entry 8VHL) bound to DHODH.

In summary, we described the application of a fragment-based screen approach and scaffold hopping methods to identify DHODH inhibitors with differentiated profiles. SAR exploration drove potency to levels previously reported for optimal triazolone-containing derivatives, and a lead molecule exhibited a good mouse pharmacokinetic profile. Moreover, these analogs displayed a much lower efflux ratio, thus allowing for potential applications where exposures in the brain would be required. Interestingly, the crystal structure of the optimized urea scaffold displayed an unexpected “flip” anticipated from fragment 2, leading to the loss of the interaction with Y356. Therefore, ongoing designs to re-engage this residue in this new series are ongoing and will be reported in due course.

Glossary

Abbreviations

DHODH

dihydroorotate dehydrogenase

AML

Acute Myeloid Leukemia

DHO

L-dihydroorotate

ORO

orotate

LLE

Lipophilic Ligand Efficiency

Cl_int

intrinsic clearance

HLM

human liver microsome

MLM

Mouse Liver Microsome

CYP

cytochrome P450

hERG

human Ether-à-go-go-Related Gene

%F

oral bioavailability

V_dss

volume of distribution steady state

t_1/2

half-life

C_max

maximum concentration

t_max

time at which maximum concentration is achieved

AUC

Area Under the Curve

TGI

Tumor Growth Inhibition

SPR

Surface Plasmon Resonance

Papp

apparent permeability

FBDD

Fragment-Based Drug Discovery

P-gp

P-glycoprotein

IC₅₀

half-maximal inhibitory concentration

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsmedchemlett.3c00543.

• Experimental procedures and characterization of final compounds (PDF)

Author Contributions

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

The authors declare no competing financial interest.