Novel Synthesis of 8-Deaza-5,6,7,8-tetrahydroaminopterin Analogues via an Aziridine Intermediate

An efficient method for the construction of the tetrahydrofolate skeleton is described. Starting from pterin analogues and aromatic amines, 8-deaza-5,6,7,8-tetrahydroaminopterin derivatives and the heterocyclic benzoyl isosteres were synthesized via a novel aziridine intermediate. Following this method, the byproducts of carbon-nitrogen bond hydrogenolysis in traditional synthetic strategy can be completely avoided.


Introduction
Tetrahydrofolate plays a central role in one-carbon metabolism involving DNA biosynthesis, and the biologically active cofactor forms are substrates for at least 15 relevant enzymes [1][2][3][4]. Hence, tetrahydrofolate analogues are widely used as inhibitors of folate-dependent enzymes for chemotherapy. Early in 1960s, tetrahydrohomofolate ( Figure 1) was found to be a potent inhibitor against TS (thymidine synthase) [5]. After then, more tetrahydrofolate analogues were developed as antitumor drugs, including lometrexol [6] as a specific inhibitor of GARFT (glycinamide ribonucleotide formyl transferase), AG2034 [7] and LY309887 [8] as next-generation GARFT inhibitors, side chain modified 5-deazatetrahydrofolate analogues [9] as potential dual inhibitors of OPEN ACCESS FPGS (folylpolyglutamate synthetase) and GARFT as therapeutic agents, compound 1 and its analogues as MS (methionine synthase) and DHFR (dihydrofolate reductase) inhibitors [10]. In our research on new antifolate inhibitors, we have identified N- [4-{(2-[2,4-diamino-5-(2,3-dibromopropane)-5,6,7,8-tetrahydropyrido(3,2-d) pyrimidin-6-yl]methyl)amino}-3-bromobenzoyl]-Lglutamic acid (ZL033) [10,11], which inhibits methionine synthase (IC 50 : 1.4 ± 0.4 µmol/L) and proliferation of HL-60 cells (IC 50 : 2.0 ± 1.0 µmol/L). Despite the remarkable biological importance of tetrahydrofolate analogues, not many synthetic approaches have been reported. The common synthesis strategy [12][13][14][15] is based on a critical step which is catalytic hydrogenation of folate derivatives, and was the strategy we used in synthesis of ZL033 and a series of novel 8-deazatetrahydrofolate derivatives as DHFR and MS inhibitors (Scheme 1). The reductive hydrogenation is a critical step in the entire synthetic scheme due to the benzylic hydrogenolysis of the C 9 -N 10 bond as a side reaction. Careful adjustment of reaction solution acidity and protection of N 10 with an electron-withdrawing substituent before reduction didn't work well either. Against this background, we now report a practical and novel method for synthesizing 8-deazatetrahydrofolate 1 by ring-opening of an aziridine intermediate. Following this method, ZL033 and its heterocyclic benzoyl isosteres can be synthesized. And the modification in the side chain of 8-deazatetrahydrofolate can lead to more optional tetrahydrofolate analogues. Scheme 1. Reported route for compound 1.

Retrosynthetic Analysis of 1
Our synthetic strategy is displayed in Scheme 2. Compound 1 is a key intermediate which can be converted into ZL033 easily by the same method in Scheme 1, so we focused on the convenient synthesis of 1. We envisaged that target compound 1 and its derivatives could be constructed by ring-opening of aziridine 5 with aromatic amines. The aziridine 5 would be synthesized via cyclization of compound 6, which could be obtained by catalytic hydrogenation of compound 3. Accordingly, we started our synthetic studies by seeking a general and practical method for preparing the reduced compound 6. Scheme 2. Synthetic strategy.

Catalytic Hydrogenation of 3
Compound 3 was prepared in the same way as before [16,17], and was catalytically hydrogenated directly before linking with the side-chain amine to avoid C 9 -N 10 hydrogenolysis (Scheme 3). As there is no fragile C 9 -N 10 bond in 3, stronger acidic conditions could be used compared to the similar hydrogenation in Scheme 1. An initial screening of the influence of different acidic conditions (Table 1) revealed that moderate acidity gave better results. The reduction reaction didn't happen without acid added ( Table 1, entry 1). Weak acids such as acetic acid gave product in 50~60% yield, even when excess acid was used as solvent (Table 1, entries 2~4), while the yield increased obviously with hydrochloric acid (Table 1, entries 5~7). Treatment of 3 with 8 equivalents of HCl for 24 h gave compound 6 in 93% yield, but with 16 equivalents of HCl, the yield was only 78%. The reason is that the hyroxymethyl group at 6-position of compound 3 could be reduced to 6-methyl group over PtO 2 under strong acidic conditions. We actually got the 6-methyl isoster of compound 6 in 12% yield under the conditions of entry 7.

Scheme 3. Catalytic hydrogenation of 3.
Variation of solvent also affected the yield strongly (Table 1, entries 6 and 8~9). Ethanol was a better solvent under these reaction conditions than methanol and acetone. The best conditions we found featured eight equivalents of hydrochloric acid in ethanol (Table 1, entry 6), resulting in a higher yield (93%) and the workup is easier than with the reduction reaction in Scheme 1.

Optimized Synthesis of 5
Before trying this synthetic strategy, we attempted to directly synthesize 1 from 6 in several ways, but none of them succeeded. The hypothetical Mitsunobu reaction between 6 and diethyl N-(p-aminobenzoyl)-L-glutamate 4 wouldn't proceed, and methods of transferring the hydroxymethyl group in 6 to aldehyde or halidemethyl group couldn't improve this condensation reaction.

Scheme 4. Two-step synthesis of compound 5.
Compounds 7 and 5 were both identified by 1 H-NMR, 13 C-NMR and ESI-MS. The 1 H-NMR (DMSO-d 6 ) spectrum of 7 revealed a multiplet signal at δ 4.06~3.92 ppm assignable to CH 2 Cl, which obviously differed from the CH 2 OH group signal (δ 4.97~4.90 ppm) of compound 6. The 13 C-NMR (DMSO-d 6 ) spectrum of 7 also displayed a different signal at δ 47.35 ppm assignable to CH 2 Cl with the CH 2 OH group signal at δ 51.79 ppm for 6. The ESI-MS result of 7 gave the intuitive proof of a chloride atom by a pair of isotopic peaks (214.1 and 216.1, 3:1). The 1 H-NMR (DMSO-d 6 ) spectrum of 5 displayed two doublet signals at δ 2.14 ppm (d, J = 5.2 Hz, 1H) and 1.56 ppm (d, J = 4.0 Hz, 1H) which were readily assigned to the two hydrogen atoms attached at methylene position of the aziridine ring and the 13 C-NMR (DMSO-d 6 ) spectrum of 5 displayed a new signal at δ 33.33 ppm assignable to NCH 2 in aziridine ring and a signal at δ 35.23 ppm assignable to NCH which gave both a relatively high-field shift compared to compounds 7 or 6. High resolution MS provided further confirmation for the two products.
The reaction above was somewhat low-yielding and its work-up procedure was complicated. After further investigation, we found an alternative method to obtain 5 in one step by treatment of 6 with p-toluenesulfonyl chloride, aqueous NaOH and tetrabutylammonium iodide in a two-phase system (Scheme 5). The yield increased to about 50%, and the work-up was simplified.

Synthesis of 1
Compounds 5, 4 and actived silica gel were mixed and shaken thoroughly under solvent-free conditions (Scheme 6). The desired product 1 was obtained in 90% yield, and the small amount of impurities could be removed easily by column chromatography. Compared with usual ring-opening reactions of aziridines using Lewis acid catalysis [22][23][24][25], the silica gel-induced method was convenient, green and cheap. Product 1 was verified by 1 H-NMR and ESI-MS, and it was compared with the same compound prepared by Scheme 1 to prove the structural consistency. Scheme 6. Synthesis of 1 under solvent-free conditions.

Synthesis of ZL033
Compound 1 can be converted to ZL033 in the same reaction sequence of allyl substitution, bromination and hydrolysis as before [11] (Scheme 7). Scheme 7. Synthesis of ZL033 from 1.

Synthesis of Other Tetrahydrofolate Analogues
Aside from ZL033, many other tetrahydrofolate analogues can be synthesized from intermediate 1 by simply modifying the substituents on N 5 or N 10 . Furthermore, the strategy above can be used for synthesis of various compounds, such as heterocyclic benzoyl analogues and modification in the side chain of tetrahydrofolate. Earlier research proved that substitution of N 10 -phenyl by aromatic heterocycles benefited inhibitory activity against several kinases [26][27][28][29]. The synthetic route of heterocyclic benzoyl analogues was similar with that of ZL033, including the ring-opening of aziridine by aromatic heterocyclic amines (Scheme 8). Compounds 10, 11 and 12 were obtained in acceptable yields by the solid-phase reaction catalysed by silica gel.  Compounds 10, 11 and 12 were transformed into analogues of ZL033 by substitution and hydrolysis reactions. This novel synthetic route is flexible to a certain extent. With various pteridines and aniline analogues, more tetrahydroaminopterin and tetrahydrofolate analogues were prepared in this novel method. The studies of these heterocyclic analogues of 8-deazatetrahydrofolate on the antifolate activity and the structure-activity relationship will be published in the future.

General
All reactions were monitored by TLC analysis (pre-coated silica gel plates with fluorescent indicator UV254, 0.2 mm) and visualized with 254 nm UV light. The products were purified by column chromatography on silica gel (silica gel 230-400 mesh or silica gel H, produced by Qingdao Ocean-Chem Inc., Qingdao, China), or preparative thin layer chromatography on silica gel (silica gel GF 254 , 0.5 mm, also produced by Qingdao Ocean-Chem Inc., Qingdao, China). 1 H-NMR spectra were measured on Bruker avance III 400 spectrometer (400 MHz, Bruker Co., Fällanden, Switzerland) or JEOL JNM-AL300 spectrometer (300 MHz, Joel, Japan). Data were reported as follows: Chemical shifts in ppm from tetramethylsilane as an internal standard in CDCl 3 or DMSO-d 6 , multiplicity (s = singlet, d = doublet, t = triplet, q = quartet, m = multiplet, br = broad), coupling constants (Hz), integration. 13 6-choloromethyl-5,6,7,8-tetrahydropyrido [3,2-d]pyrimidine (7): A stirred solution of compound 6 (502 mg, 5.1 mmol) in POCl 3 (10 mL) was refluxed at 110 °C for 2 h. The resulting solution was evaporated to a minimized volume of about 1 mL under reduced pressure. Then it was diluted with 50 mL dichloromethane and evaporated to a minimized volume again. This procedure was repeated twice with diluting solvent changing successively to ethanol and methanol. Acetone (50 mL) was added to the concentrated sticky liquid, and the resulting white precipitate was filtered, washed with acetone (20 mL) and dried under vacuum to afford 221 mg (40.2% yield) of 7 as an off-white solid; m.p. 228~232 °C. 1 .60 mL, 4.1 mmol). The reaction was stirred at room temperature for 1 h, then concentrated and isolated by preparative TLC with a chloromethane and methanol mixture (9:1) as eluent. Compound 5 (210 mg) was obtained as a faint yellow solid, the yield was 85.7%, while the two-step overall yield from compound 6 was 34.5%.

Conclusions
In conclusion, we have developed a novel synthetic route providing the biologically important compound 1 in good yields via an aziridine intermediate, thus avoiding the benzylic hydrogenolysis of the carbon-nitrogen bond of earlier methods. It also represents a convenient preparation of tetrahydroaminopterin or tetrahydrofolate analogues.