Discovery of 5′-Substituted 5-Fluoro-2′-deoxyuridine Monophosphate Analogs: A Novel Class of Thymidylate Synthase Inhibitors

5-Fluorouracil and 5-fluorouracil-based prodrugs have been used clinically for decades to treat cancer. Their anticancer effects are most prominently ascribed to inhibition of thymidylate synthase (TS) by metabolite 5-fluoro-2′-deoxyuridine 5′-monophosphate (FdUMP). However, 5-fluorouracil and FdUMP are subject to numerous unfavorable metabolic events that can drive undesired systemic toxicity. Our previous research on antiviral nucleotides suggested that substitution at the nucleoside 5′-carbon imposes conformational restrictions on the corresponding nucleoside monophosphates, rendering them poor substrates for productive intracellular conversion to viral polymerase-inhibiting triphosphate metabolites. Accordingly, we hypothesized that 5′-substituted analogs of FdUMP, which is uniquely active at the monophosphate stage, would inhibit TS while preventing undesirable metabolism. Free energy perturbation-derived relative binding energy calculations suggested that 5′(R)-CH3 and 5′(S)-CF3 FdUMP analogs would maintain TS potency. Herein, we report our computational design strategy, synthesis of 5′-substituted FdUMP analogs, and pharmacological assessment of TS inhibitory activity.


■ INTRODUCTION
−7 Under physiological conditions, the TS enzyme plays an essential role in catalyzing the de novo synthesis of deoxythymidine monophosphate (dTMP) from deoxyuridine monophosphate (dUMP) in the presence of the 5,10-CH 2 -tetrahydrofolate (mTHF) cofactor and a TS active site cysteine residue. 8,9−12  5-FU additionally generates two other cytotoxic metabolites, 5-fluoro-2′-deoxyuridine 5′-triphosphate (FdUTP) and 5-fluorouridine 5′triphosphate (FUTP), which incorporate into growing oligonucleotides as substrates for DNA and RNA polymerases, respectively (Figure 1). 1 Although 5-FU effectively treats a fraction of cancer patients with these unique mechanisms of action, its widespread clinical utility is significantly limited by catabolic degradation in the liver (up to 85%) by dihydropyrimidine dehydrogenase (DPD) as well as by poor oral bioavailability. 13,14−17 However, the productive metabolism of these prodrugs to FdUMP proceeds via the generation of 5-FU, subsequently resulting in DPDmediated degradation and systemic toxicity, among other undesirable consequences (Figure 2).−21 While this approach offers several potential advantages over 5-FU and approved prodrugs thereof, NUC-3373 requires intravenous administration and is still responsible for the production of FdUTP and FUTP, which indiscriminately target both healthy tissue and tumor tissue.
Given the shortcomings discussed above, there is a need to develop novel analogs and/or prodrugs of FdUMP that limit the formation of promiscuous triphosphate metabolites while efficiently releasing FdUMP (or efficacious analogs thereof) and avoiding 5-FU generation.Previous results from our laboratory using cell-based hepatitis C virus (HCV) replicon assays revealed that functionalization of 2′-C-methyluridine at the 5′-carbon (i.e., 5′-CH 3 ) significantly compromised antiviral activity of the corresponding mono-and diphosphate prodrugs (Figure 3). 22In this assay, the antiviral activity of nucleoside mono-and diphosphate prodrugs is dependent upon: (1) conversion of prodrug to the corresponding nucleoside monophosphate or diphosphate; (2) conversion of nucleoside monophosphate or diphosphate to the corresponding nucleoside triphosphate; and (3) substrate recognition by and concomitant inhibition of HCV RNA-dependent RNA polymerase (RdRP).Importantly, conversion of these prodrugs to the corresponding mono-and diphosphates relies on enzyme-mediated activation at positions relatively distal from the 5′ carbon.Therefore, the lack of antiviral activity of 5′-CH 3 derivatives is much more likely to be controlled by slow conversion of mono-and diphosphate metabolites to the corresponding triphosphates and/or poor substrate recognition of these triphosphates by HCV RdRP.
While these results indicated that our 5′-functionalization strategy would clearly not be effective for nucleoside-based therapeutics requiring efficient conversion to (and target recognition of) the corresponding triphosphates, FdUMP is uniquely active at the monophosphate stage. 2 Accordingly, we hypothesized that 5′-functionalized FdUMP analogs could maintain potent inhibition of TS and associated antitumor efficacy while reducing undesired side effects and systemic toxicity by limiting the generation of 5-FU and the corresponding nucleoside triphosphates (Figure 3).We report herein our approach to testing this hypothesis using a combination of molecular modeling, organic synthesis, and biochemical techniques.
Initially, molecular modeling was employed to assess whether 5′-substituted FdUMP analogs could be well accommodated within the human TS (hTS) binding pocket.Utilizing the available X-ray cocrystal structure of hTS in complex with FdUMP (PDB ID 6QXG), several FdUMP analogs featuring small 5′-substituents (Table S1) were docked (Schrodinger's Glide-SP) into the FdUMP binding pocket.Initial binding energy calculations using Prime MM-GBSA suggested that three different substitutions, 5′(R)-CH 3 , 5′(S)- CHF 2 , and 5′(S)-CF 3 , would maintain TS inhibitory activity comparable to FdUMP.The predicted binding poses of FdUMP analogs featuring these substitutions also aligned very closely (Table S2) with the conformation of FdUMP cocrystallized in this pocket (PDB ID 6QXG).We further investigated relative binding free energy values using the FEPbased methodology (Desmond − D. E. Shaw research group). 23,24The resulting FEP-calculated binding energies suggested that two of these three substitutions, 5′(R)-CH 3 and 5′(S)-CF 3 , would effectively inhibit TS function (Table S3).Consistent with these results are poor conformational alignments and poor FEP-predicted binding energies of FdUMP analogs featuring 5′(S)-CH 3 , 5′(CH 3 ) 2 , and 5′(R)-CF 3 motifs, relative to FdUMP and the 5′(R)-CH 3 -and 5′(S)-CF 3 -substituted analogs.As illustrated in Figure 4, a general observation from these computational experiments can be described: Small 5′-substituents occupying the green region of the TS catalytic site were much more likely to be tolerated compared to small 5′-substituents occupying the red region (Figure 4C).This is highlighted by the predicted stereochemical preference of hTS for 5′(R)-CH 3 and 5′(S)-CF 3 substituents (better relative binding energies, green region) over their epimeric counterparts, 5′(S)-CH 3 and 5′(R)-CF 3bearing congeners (poorer relative binding energies, red region).This predicted diastereopreference can be ascribed to unfavorable intramolecular steric interactions as well as clashes with the catalytic Cys195, ultimately leading to an unsuitable binding conformation.In contrast, the green region  highlights a small pocket that could nicely accommodate 5′substituents, such as 5′(R)-CH 3 and 5′(S)-CF 3 .It is also notable that, even with the preferred stereoconfiguration, larger substituents (e.g., ethyl) were disfavored (Table S3) relative to smaller substituents (e.g., methyl).
To synthesize the epimeric 5′(S)-CH 3 FdUMP analog (2, Scheme 1), key intermediate 7 was first subjected to ATH conditions using the same metal catalyst with the opposite enantiomeric ligand, RuCl(p-cymene)[(S,S)-Ts-DPEN] (10).Although this procedure generated the 5′(S)-CH 3 alcohol 11 in 83% yield, the diastereoselectivity (7:3 dr) was suboptimal.In an attempt to enhance stereoselectivity, a small set of chiral transition metal catalysts was screened, 25 albeit with limited success, as each of these conditions failed to substantially improve 5′(S)-CH 3 diastereoselectivity, affording either predominantly the 5′-(R)-isomer or a diastereomeric mixture (Table S4).These results suggested that a mixture of substratecontrolled and catalyst-controlled diastereopreferential reactivity is operative.We then hypothesized that reducing the steric bulk of the 3′-OH protecting group from TBS to acetyl (Ac) would allow catalyst-controlled reactivity to predominate.Accordingly, key intermediate 7 was converted in 47% yield over two steps to its 3′-O-acetyl derivative, which was then subjected to ATH using Noyori's chiral catalyst 10 to deliver the desired 5′(S)-CH 3 alcohol 12 in 67% yield and 99:1 dr.The absolute configuration of the 5′(S) stereocenter was determined by single crystal X-ray analysis (Table S5).Alcohol 12 was then converted to the corresponding monophosphate 2 using a three-step procedure in which 12 was first treated with POCl 3 to give the monophosphate in 20% yield.Next, removal of the 3′-O-acetyl protecting group using aqueous NH 3 , followed by counterion exchange using Dowex-Li + afforded 5′(S)-CH 3 FdUMP 2 as the bis-lithium salt in 27% yield over two steps.The 5′-gem-(CH 3 ) 2 analog 3 was synthesized in three steps from the intermediate 6.The acid 6 was first converted to its methyl ester in 86% using TMSCHN 2 , followed by Grignard methylation using methylmagnesium bromide to give the tertiary alcohol in 72% yield.Subsequent phosphorylation conditions involving POCl 3 additionally facilitated contaminant TBS deprotection to give the monophosphate in 14%.Counterion exchange using Dowex-Li + afforded the 5′-gem-(CH 3 ) 2 FdUMP 3 in 12% isolated yield.Synthesis of 5′(R)-CF 3 and 5′(S)-CF 3 substituted FdUMP analogs are described in Scheme 1b.First, the FdU 3′-OH group was protected as the corresponding benzyl ether 13 using a three-step protocol involving: (1) protection of the 5′-OH group as a trityl ether in 78% yield using trityl chloride in pyridine under MW irradiation at 100 °C; (2) protection of the 3′-OH as the corresponding benzyl ether in 67% yield using benzyl bromide and sodium hydride; and (3) removal of the trityl group using 80% aqueous acetic acid to afford intermediate 13 in 81% yield (42% yield over 3 steps).Next, alcohol 13 was oxidized to aldehyde 14 using DMP.−29 Fortunately, this diastereomeric mixture could be cleanly resolved using silica gel column chromatography to deliver 5′(S)-CF 3 (15, 27% yield) and 5′(R)-CF 3 (16, 18% yield) derivatives, respectively.The absolute stereochemical assignment was determined using X-ray crystallography, indicating that alcohol 15 featured the 5′(S)-CF 3 motif (Table S5).These intermediate alcohols were further converted to the corresponding monophosphates in two steps: (1) treatment of 15 and 16 with POCl 3 in pyridine gave the 3′-OBn monophosphates; and (2) hydrogenolysis of the 3′-OBn groups using Pd(OH) 2 on carbon afforded 5′(S)-CF 3 FdUMP analog 4 in 12% yield and 5′(R)-CF 3 FdUMP analog 5 in 6% yield over two steps.

■ CONCLUSION
In summary, the mechanism of antitumor activity of fluoropyrimidine-based therapeutics, including 5-FU, floxuridine, and capecitabine, relies heavily on inhibition of TS by the common metabolite FdUMP.However, most of the drugs and prodrugs in this class also cause significant toxicities that are driven by high systemic concentrations of 5-FU.We reported herein a novel application of our previously reported nucleoside 5′-functionalization strategy.Although this approach previously failed to deliver active anti-HCV agents, application of this concept to FdUMP, which is uniquely active at the nucleoside monophosphate stage, proved to be fruitful.As predicted by our modeling efforts and empirically corroborated by synthesizing and evaluating the TS inhibitory activity of these novel nucleoside monophosphates, a strong stereochemical preference for 5′(R)-CH 3 -and 5′(S)-CF 3substituted FdUMP analogs is apparent.Ongoing efforts involve mechanism of action determination and detailed metabolism experiments designed to test the hypothesis that 5′-functionalization of nucleoside and nucleotide therapeutics limits conversion to the corresponding triphosphate metabolites.

■ ASSOCIATED CONTENT
* sı Supporting Information Supporting Information includes: The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsptsci.2c00252.
Molecular modeling protocol and associated data (Table S1, list of 5′-substitutions; Table S2, Prime MM-GBSA; Table S3, FEP); detailed synthetic schemes for compounds 1−5 (Schemes S1−S4); ATH catalyst screen (Table S4); small molecule X-ray crystal structures (Table S5); synthetic procedures, spectroscopic characterization data, and copies of NMR spectra for all purified compounds; in vitro hTS inhibition assay protocol, associated UV absorption spectra (Figures S1  and S2), and representative enzyme inhibition data (Figure S3); summary table comprising tested com- pounds with IC 50 potencies and FEP relative binding energy predictions (Table S6

Figure 6 .
Figure 6.Concentration−response curves of FdUMP (a) and active 5′-functionalized analogs (b and c) against recombinant hTS-mediated conversion of 2′-deoxyuridine monophosphate (dUMP) and 5,10-methylene tetrahydrofolate (mTHF) to deoxythymidine monophosphate (dTMP) and dihydrofolate (DHF), respectively.The final concentrations of hTS, dUMP, and mTHF in the reaction systems were 0.84 μM, 50 μM, and 250 μM, respectively.The reaction rates were determined by following the change in the optical absorbance at 340 nm, which corresponded to the formation of the coproduct, DHF.