Article pubs.acs.org/jmc
Discovery of N‑(6-Fluoro-1-oxo-1,2-dihydroisoquinolin-7-yl)-5[(3R)‑3-hydroxypyrrolidin-1-yl]thiophene-2-sulfonamide (LSN 3213128), a Potent and Selective Nonclassical Antifolate Aminoimidazole-4-carboxamide Ribonucleotide Formyltransferase (AICARFT) Inhibitor Effective at Tumor Suppression in a Cancer Xenograft Model Kevin R. Fales,*,† F. George Njoroge,† Harold B. Brooks,† Stefan Thibodeaux,† Alicia Torrado,‡ Chong Si,† James L. Toth,† Jefferson R. Mc Cowan,† Kenneth D. Roth,† Kenneth J. Thrasher,† Kwame Frimpong,† Matthew R. Lee,† Robert D. Dally,† Timothy A. Shepherd,† Timothy B. Durham,† Brandon J. Margolis,† Zhipei Wu,† Yong Wang,† Shane Atwell,† Jing Wang,† Yu-Hua Hui,† Timothy I. Meier,† Susan A. Konicek,† and Sandaruwan Geeganage† †
Lilly Research Laboratories, Eli Lilly and Company, Indianapolis, Indiana 46285, United States Centro de Investigación Lilly, S. A., Avda. de la Industria 30, 28108 Alcobendas, Madrid, Spain
‡
S Supporting Information *
ABSTRACT: A hallmark of cancer is unbridled proliferation that can result in increased demand for de novo synthesis of purine and pyrimidine bases required for DNA and RNA biosynthesis. These synthetic pathways are frequently upregulated in cancer and involve various folate-dependent enzymes. Antifolates have a proven record as clinically used oncolytic agents. Our recent research efforts have produced LSN 3213128 (compound 28a), a novel, selective, nonclassical, orally bioavailable antifolate with potent and specific inhibitory activity for aminoimidazole-4-carboxamide ribonucleotide formyltransferase (AICARFT), an enzyme in the purine biosynthetic pathway. Inhibition of AICARFT with compound 28a results in dramatic elevation of 5-aminoimidazole 4-carboxamide ribonucleotide (ZMP) and growth inhibition in NCI-H460 and MDA-MB-231met2 cancer cell lines. Treatment with this inhibitor in a murine based xenograft model of triple negative breast cancer (TNBC) resulted in tumor growth inhibition.
■
formylation to generate α-N-formylglycinamide ribonucleotide (FGAR, 3). Subsequent transformation steps provide 5-aminoimidazole 4-carboxamide ribonucleotide (ZMP, 4), the substrate for the bifunctional human enzyme aminoimidazole-4-carboxamide ribonucleotide formyltransferase/inosine monophosphate cyclohydrolase (ATIC). The ATIC enzyme consists of two domains with separate functions. The first domain, aminoimidazole-4-carboxamide
INTRODUCTION Cancer requires both DNA and RNA to support unbridled proliferation.1 As components in the synthesis of DNA and RNA, the purines adenine and guanine are critical for cell proliferation. Generation of these key elements is directly dependent on onecarbon metabolic processes involving folate-dependent enzymes (Figure 1) that were first elucidated in 1959.2 The purine biosynthetic sequence initiates with conversion of 5-phosphoribosyl-1-pyrophosphate (PRPP, 1) to glycinamide ribonucleotide (GAR, 2). The folate-dependent enzyme glycinamide ribonucleotide formyltransferase (GARFT) then adds a single carbon via © 2017 American Chemical Society
Received: July 17, 2017 Published: October 26, 2017 9599
DOI: 10.1021/acs.jmedchem.7b01046 J. Med. Chem. 2017, 60, 9599−9616
Journal of Medicinal Chemistry
Article
Figure 1. Purine biosynthetic pathway.
used with great success for decades.12 The biophysical properties of classical antifolates require parenteral administration, followed by active uptake in cells via reduced folate carrier (RFC), folate receptors (FRα, FRβ, FRγ), or the protoncoupled folate transporter (PCFT), and subsequent conversion to the corresponding polyglutamates via folylpolyglutamate synthase (FPGS).13 While the resulting sequestration of the drugs within the cells dramatically enhances potency at the targeted enzyme, a potentially undesirable result of polyglutamation is enhanced activity toward other folate dependent enzymes.14 Also, a major mechanism of resistance to classical antifolates is through genetic mutation leading to impairment of the transport and polyglutamation mechanisms.15 In contrast, nonclassical antifolates penetrate cells through passive diffusion, require no intracellular modification in order to exert their effects, and thus can overcome resistance due to impaired active uptake.16 As noted above, inhibition of AICARFT results in elevation of ZMP, a key intermediate in the biosynthesis of purines for a wide variety of species. In addition to its role for the synthesis of DNA and RNA, effects of ZMP have been reported in diverse areas of pharmacological research including oncology, inflammation, autoimmune disease, and diabetes.17 Due to its broad effects, ZMP has also been touted as a master regulator of one-carbon metabolism in bacteria, having been shown to activate a conserved riboswitch that regulates expression of associated genes.18
ribonucleotide formyltransferase or AICARFT, facilitates formyl transfer from N10-formyl-tetrahydrofolate (N10-CHO-THF) to ZMP, producing 5-formamidoimidazole-4-carboxamide ribotide (FAICAR, 6). Subsequent cyclization of FAICAR in the second domain of ATIC, inosine monophosphate cyclohydrolase (IMPCH), results in inosine 5′-monophosphate (IMP, 7). Inhibitors of AICARFT (Figure 2) that have been reported include pemetrexed (9, Ptx), the dihydrofolate reductase (DHFR) inhibitor methotrexate (Mtx, 10), N-[(4-{[(2amino-4-oxo-3,4-dihydroquinazolin-6-yl)amino]sulfonyl}phenyl)carbonyl]-L-glutamic acid (11), 5- and 6-substituted pyrrolo[2,3-d]pyrimidines reported by Gangjee et al., a series of multisubstrate adduct inhibitors including β-DADF, an inhibitor of ATIC homodimerization, and a series of nonclassical antifolates reported by Agouron.3−8 Ptx is a classical antifolate with primary activity for inhibition of thymidylate synthase (TS) in the pyrimidine biosynthetic pathway and has been approved by the FDA with vitamin supplementation per label for use in combination with cisplatin as a first line treatment of advanced nonsquamous non-small-cell lung cancer. Ptx is also a multitargeting inhibitor, with activity against dihydofolate reductase (DHFR) and GARFT.9 More recently, Moran et al. reported that a secondary target of Ptx is AICARFT, resulting in accumulation of ZMP and activation of AMP-activated protein kinase (AMPK) with additional downstream effects.10,11 As classical antifolates, Ptx and likewise Mtx are compounds from a class of drugs that have been extensively studied and 9600
DOI: 10.1021/acs.jmedchem.7b01046 J. Med. Chem. 2017, 60, 9599−9616
Journal of Medicinal Chemistry
Article
and sustained elevation of ZMP, resulting in anticancer therapeutic activity. Strategy and Initial Leads. In 2004, Beardsley et al. reported the structure of ATIC in complex with the bisarylsulfonamide 11, in which the crystal structure revealed four distinct binding regions as highlighted in Figure 2 and described here as the pteridine, oxyanion hole, ZMP π-stack, and glutamate binding regions.20 To reiterate from Beardsley’s paper regarding AICARFT, (1) there was no sequence homology with other folatedependent enzymes, providing a basis for design of selective inhibitors, (2) the oxyanion hole was a major factor in determining the position of the sulfonyl-containing antifolates within the AICARFT active site, and (3) exploitation of the oxyanion hole for design of selective inhibitors was possible since other folate dependent enzymes (GARFT, TS, DHFR, etc.) did not contain this feature. Compound 11 was reported to be potent enzymatically (Ki = 6 nM),20 but this activity did not translate well in vitro as the reported IC50 for this compound was between 1 and 5 μM. This relative loss of activity could possibly be due to poor free permeability of the compound due to bis-carboxylic acids present in its glutamate side chain. Activity of this compound for cellular uptake by folate transporters or the reduced folate carrier was not reported. Use of this compound in vivo also has not been described. The background provided through the work reported by Beardsley et al.20 encouraged us with the possibility of utilizing structure based design in pursuit of AICARFT inhibitors. We began with an in-house screening effort of ∼100 000 compounds that ultimately yielded compound 12 (Figure 3), which demonstrated low inhibitory activity (IC50 = 35 μM, Table 1). Table 1. In Vitro Activity of Early SAR Compoundsa Figure 2. Reported AICARFT inhibitors and binding regions of 11. (a) Crystal structure of 11 was obtained from PDB (code 1pl0; BW2315U89UC).
Aminoimidazole-4-carboxamide ribonucleoside (AICAr, compound 5, Figure 1) is the corresponding des-phosphorylated analogue of ZMP and as such is freely permeable and orally bioavailable. Once absorbed, AICAr is rapidly phosphorylated by cells yielding ZMP and has been widely used to study systemic metabolic effects of AMPK activation by ZMP. AICAr was also studied clinically as an oncolytic under the trade name Acadesine, for treatment of childhood acute lymphoblastic leukemia (ALL), but required high doses to maintain elevated levels of ZMP, due to rapid metabolic progression of this intermediate through the purine biosynthetic pathway (Figure 1).19 Encouraged by the potential pharmacological benefit of extended elevation of ZMP, we hypothesized that selective inhibition of AICARFT with an orally bioavailable, nonclassical antifolate would lead to durable
compd
enzyme IC50 ± SD (nM)
cell IC50 ± SD (nM)b
12 13 14 15 16 25
34800 90000 ± 5445 >100000 608 ± 286 18100 ± 12000 723 ± 241
nd nd nd 567 ± 289 nd 1811 ± 910
a
With the exception of compound 12, all values represent average results from replicates of two or more measurements. bConcentration at 50% maximum inhibition in NCI-H460 cells, grown in low folate media as described in the biological methods.
Subsequent explorations resulted in 13, potentially a minimum pharmacophore since removal of the primary amide from 13 generating 14 resulted in no inhibitory activity even at 100 μM.
■
RESULTS AND DISCUSSION Our approach in order to identify more potent and drug-like compounds led us to investigate modifications in four distinct
Figure 3. Initial leads. 9601
DOI: 10.1021/acs.jmedchem.7b01046 J. Med. Chem. 2017, 60, 9599−9616
Journal of Medicinal Chemistry
Article
Scheme 1
Contrary to our expectations, AICARFT enzyme IC50 of 16 was found to be only 18 ± 12 μM, representing a 30-fold loss in potency versus 15. Oxyanion Hole. Although crystal structure analysis indicated a sulfonamide group would be ideal for binding to the oxyanion hole in AICARFT, in silico calculations indicated improvements in drug-like properties (solubility, permeability) could be achieved through replacement of this functionality. We investigated numerous reported sulfonamide replacements including common isosteres such as carboxamide and urea, as well as simpler two-atom linkers like alkylamines, each designed to engage key residues in the oxyanion hole (as reviewed below for sulfonamide). In every case the compounds produced were devoid of activity in the enzyme assay (i.e., IC50 > 100 μM, structures and data not shown). In order to enhance the druglike properties of the bis-arylsulfonamide in compounds like 15, F for H substitution on C6 of the neighboring isoquinolone as in compound 25 (Scheme 2) was targeted, expecting to lower the pKa of the neighboring sulfonamide, leading to improved aqueous solubility at or above pH 6. In addition to improved solubility, we hypothesized that intramolecular hydrogen bonding between the C6 fluorine and the sulfonamide NH might lead to improved permeability. 7-Amino-6-fluoroisoquinolone 24 was prepared by in situ generation of a carbamoyl chloride from phenylethanamine 21 and triphosgene, followed by cyclization under Friedel−Crafts conditions to provide dihydroisoquinolone 22. Selective nitration afforded the 7-nitro intermediate that was oxidized to yield isoquinolone 23 by heating overnight in dichloroethane with 10 equiv of MnO2. Subsequent hydrogenation provided amine 24 as a brown solid that was readily crystallized from water. Similar to the synthesis of 15 (Scheme 1), coupling of 24 and 18 yielded the corresponding 6-fluoro isomer 25. As predicted, introduction of fluorine at C6 resulted in pKa reduction to 5.97 for 25 in comparison with 6.93 for 15 (Table 2) with concomitant decrease in log D at neutral pH and above. No significant difference in solubility was found between 15 and 25, and in contrast to our hypothesis, measured permeability in MDCK cells found reduced permeability (0.35 vs 0.55% A-B). This reduction could be due to the increased polarity of the NH bond in 25 as reflected in the lower pKa for this compound. Nonetheless, this modification did provide a pharmacokinetic benefit in vivo since after a 10 mg/kg oral dose a 3-fold enhancement of plasma exposure in mice with 25 was achieved, yielding an AUC of 55 800 ng·h/mL and 79% bioavailability in comparison to an AUC of 16 100 ng·h/mL and 43% bioavailability for 15. No differences in enzymatic inhibition were detected between compounds 15 and 25, although a slight loss in cell based activity was measured with 25. In subsequent SAR fluorine at quinolone position C6 often
areas of the inhibitors, defining the parts in correspondence with the binding elements shown previously for 11 (Figure 2), including the pteridine, oxyanion hole, ZMP π-stack, and glutamate binding regions. Under the aforementioned designations, the exploration of pertinent features will be described systematically in subsequent sections. Pteridine Binding Region. With 13 as a design starting point (Figure 3), our first target was envisioned to include addition of two carbons and cyclization to generate the isoquinolone 15 (Scheme 1). While compounds 12−14 were available from commercial sources, 15 was synthesized by coupling of commercially available 7-aminoisoquinolin-1(2H)one 17 to 4-cyanobenzenesulfonyl chloride 18 (Scheme 1). Biochemical testing of 15 found that it provided a dramatic increase in AICARFT enzymatic inhibition versus 13 (IC50 = 608 nM vs 90 μM). Due to the enzymatic potency of 15, it was also tested in vitro using NCI-H460 cells. This cell line was chosen based on Moran’s identification of AICARFT as a secondary target for Ptx in NCI-H46O.10,11 Compound 15 was found to have enzymatic activity (IC50) of 608 ± 290 nM and comparable cell activity (IC50) of 567 ± 290 nM. It is relevant and important to note with regard to this cell activity as well as the in vivo studies discussed below that the levels of folate in media used for cell culture are often higher than circulating folate levels in humans and rodents. For example, Roswell Park Memorial Institute (RPMI) 1640 cell culture media contain 2 μM folic acid. Circulating serum folate levels have been reported for several species and are lower than RPMI 1640 media.21,22 For example, human serum folate levels are in the range of 20 nM, and rodents exhibit ∼10-fold higher levels of ∼200 nM. Since our inhibitors directly competed with N10formyltetrahydrofolate (the natural substrate for AICARFT, Figure 1), higher folate levels in cell culture media were expected to have a negative impact on antiproliferative response relative to normal in vivo conditions. For this reason, the cell culture assay was run using cells grown under low folate conditions as described in the biological methods. The reported structure of 11 (Figure 2)20 showed interactions of the aminoquinazolinone moiety deep within the pteridine pocket, including hydrogen bonds to Asn489, Asp546′, Asn547′, and Ile452′. These interactions were likely comparable to those made by the pteridine ring of the natural substrate. This along with the expectation that the isoquinolone of 15 similarly bound in the pteridine pocket led us to attempt further binding enhancement by conversion of the isoquinolone to be more pteridine-like through replacement of C4 with nitrogen and addition of an exocyclic amine at C3 (see Scheme 1 for numbering). Compound 16 was generated from the corresponding amine 20, prepared by reductive hydrogenation of the commercially available 2-amino-6-nitroquinazolinone 19. 9602
DOI: 10.1021/acs.jmedchem.7b01046 J. Med. Chem. 2017, 60, 9599−9616
Journal of Medicinal Chemistry
Article
saturation and alkylation of the pyridone ring in 29 resulted in 32a,b that likewise produced a 6- to 18-fold increase in measured plasma exposure. The unbound or free fraction in mouse plasma was also determined and was low for all compounds tested, although the dihydropyridones 32a and 32b were unique in the set for this parameter, returning unbound fractions that were below the limit of quantitation (
