Synthesis of 3,3'-di-O-methyl ardimerin and exploration of its DNA binding properties.

The 3,3'-di-O-methyl derivative (15) of the bis-C-aryl glycoside natural product ardimerin (1) has been synthesized in 11 steps from 2,3,4,6-tetrabenzylglucose (2) and 1,2,3-trimethoxybenzene (3). Key steps in the synthesis involve a Lewis acid mediated Friedel-Crafts type glycosylation and a Yamaguchi lactonization under Yonemitsu conditions. 3,3'-Di-O-methyl ardimerin aggregates in aqueous solutions at concentrations greater than 1 μM, and both UV and fluorescence binding studies indicate that 15 has a low affinity for duplex DNA.

Plants used in traditional Chinese medicine have yielded a wealth of chemical constituents with important biological activities.[1] Ardimerin (1a, Figure 1), a dimeric lactone with radical scavenging activity, was isolated from Ardisia japonica by Ryu et al. in 2002.[2] Subsequently, ardimerin digallate (1b) was isolated from the same species, along with the flavonoid quercitrin and the terpenoids friedelin, epifriedelinol, baurenol, and baurenyl acetate.[3] The digallate derivative of ardimerin was shown to inhibit HIV-1 and HIV-2 RNase H in vitro with IC50 values of 1.5 and 1.1 μM, respectively.

Figure 1

Structures of ardimerin and ardimerin digallate.

C-Aryl glycosides are an important class of naturally occurring compounds endowed with remarkable stability toward acid and enzymatic hydrolysis;[4] this affords them a sufficient intracellular lifetime to allow trafficking to the nucleus, where they bind DNA to form stable complexes.[5] The bis-C-aryl glycoside altromycin B has been shown by NMR studies to associate with DNA via a helix-threading mode of binding, with carbohydrate moieties positioned in opposite grooves of the duplex.[5e] Given that ardimerin is a symmetrical bis-C-aryl glycoside, we envisioned that, despite the nonplanarity of its aglycone,[6] it might also be capable of the recognition of nucleic acids by a threading mode of intercalation, with the glucosyl substituents positioned in both the major and minor grooves of DNA. To assess this possibility, we decided to undertake its synthesis and investigate its DNA binding properties.

We envisioned (Figure 2) that the C–C linkage between carbohydrate and aromatic moieties could be fashioned by a Lewis acid mediated Friedel–Crafts type C-glycosylation reaction between protected glucose 2 and 1,2,3-trimethoxybenzene (3).[9] Aromatic ring carbonylation and selective ortho methoxy group deprotection would then provide 7, a crucial substrate for esterification with the derived carboxylic acid monomer 8. Oxidation, macrocyclization, and protecting group removal would then provide the natural product.

Figure 2

Retrosynthetic analysis of ardimerin.

The carbohydrate coupling partner (2) required for the C-glycosylation reaction may be prepared in 72% overall yield from dextrose as previously described.[7] Treatment of 2a with a 1:1 solution of trifluoroacetic anhydride and CH2Cl2 for 30 min, followed by evaporation and combination with commercially available 1,2,3-trimethoxybenzene (1.5 equiv) and BF3·OEt2 (1.1 equiv) in CH2Cl2 at room temperature for 30 min, afforded coupled product 4 (>20:1 β:α at C.1) in 62% yield (Scheme 1).[8] Interestingly, 2-O-benzyl-1,3-dimethoxybenzene (3b) was not a suitable partner for the C-glycosylation reaction, undergoing rapid decomposition in the presence of either BF3·OEt2 or TMSOTf. However, the 2-O-tert-butyldiphenylsilyl derivative 3c coupled efficiently with 2b in the presence of TMSOTf as a Lewis acid promoter to provide the C-aryl glycoside product in 72% yield.

Scheme 1

Synthesis of C-Glycoside Monomers 7 and 8

Several methods were explored to introduce the aldehyde moiety on the aromatic ring. Attempted Vilsmeier formylation (DMF, POCl3, toluene, 100 °C) of 4 resulted in minimal conversion, even after an extended reaction time (24 h).[9] Directed ortho-lithiation[10] with n-BuLi/TMEDA and trapping with DMF gave only low (<10%) yields of carbonyl-containing products, likely due to intramolecular protonation of the aryllithium species by the benzyl ether protecting groups on the carbohydrate.[11] Bromination (NBS, CHCl3, reflux) of 4 resulted in the formation of aryl bromide 5 in 90% yield. Lithium–halogen exchange with n-BuLi, followed by rapid quenching with DMF, again led to a hydrodebrominated product arising from the aforementioned intramolecular proton transfer process. However, magnesiate formation according to the protocol of Oshima (i-PrMgCl, 2 equiv of n-BuLi, THF, 0 °C to −78 °C, then 5) followed by quenching with DMF led to a 71% yield of the desired aldehyde 6.[12]

Selective deprotection of the methoxy group ortho to the aldehyde initially proved to be problematic. Treatment of 6 with 1–3 equiv of BCl3 in CH2Cl2 at −60 °C (1 h) or room temperature (overnight) led to significant substrate decomposition. The combination of 6 with AlCl3 in benzene at 80 °C or in CH2Cl2 at room temperature gave only poor yields of the desired hydroxyl aldehyde. Finally it was discovered that treatment of 6 with 1.1 equiv of AlCl3 and 1.5 equiv of NaI in CH3CN (0.25 M) at 80 °C for 1 h gave hydroxy aldehyde 7 in 90% yield. Subsequent benzyl ether formation and oxidation with NaClO2 gave a 50% overall yield of carboxylic acid 8. Interestingly, all attempts to cleave the methoxy group meta to the aldehyde of 7 (corresponding to the C.3/C.3′ position of the natural product) by extended exposure to AlCl3/NaI (80 °C) resulted in substrate decomposition.

With both 7 and 8 in hand, we set out to identify conditions for the construction of the eight-membered diolide (Scheme 2). Attempts to directly dimerize the model compound 2,3-dimethoxysalicylic acid (SOCl2, dilute toluene, reflux; DCC or EDC, DCM, rt; TFAA, DCM, 0 °C) failed, producing only uncharacterized oligomers in low yields. In line with literature precendent,[13] coupling of the known acid 9a(14) and aldehyde 10a(15) was accomplished via direct addition of sodium alkoxide 10b to acid chloride 9b in THF at room temperature; subsequent oxidation (NaClO2, NH2SO3H, acetone/water) afforded carboxylic acid 11. Harris has demonstrated[13] that subjection of ortho-benzyl protected carboxylic acid substrates similar to 11 to refluxing thionyl chloride leads directly to eight-membered diolides of the type 12, arising from acid chloride formation, in situ benzyl ether cleavage, and macrolactonization of the hydroxy acid chloride; however, we observed that refluxing 11 in SOCl2 for 3 h led only to the intermediate hydroxy acid chloride, which was sufficiently stable to survive aqueous reaction workup. Instead, the acid chloride intermediate was diluted in benzene or toluene (0.01 M) and treated with 3 equiv of DMAP and stirred at room temperature overnight. In this way, diolide 12 could be secured in 50–60% yield.

Scheme 2

Model Study: Synthesis of Diolide 12

With a method to prepare the diolide core of ardimerin in hand, we proceeded to explore the similar union of aldehyde 7 and carboxylic acid 8 (Scheme 3). Treatment of compound 8 with oxalyl chloride in the presence of catalytic quantities of DMF gave rise to the corresponding acid chloride, which was added to the potassium salt of 7 in THF at 0 °C; the resultant crude aldehyde 13a was immediately oxidized under Pinnick–Lindgren–Kraus conditions[16] to afford the stable carboxylic acid 13b. Refluxing 13b in SOCl2 for 3 h gave rise to the corresponding benzyloxy acid chloride and not the desired hydroxyl acid chloride; further heating in SOCl2 overnight led only to extensive substrate decomposition. To effect removal of the benzyl ether before conversion to the acid chloride, compound 13b was treated with a 1:1 mixture of TFA and toluene at room temperature for 5 min.[17] The intermediate hydroxy acid was then treated with oxalyl chloride (cat. DMF, CH2Cl2), and a dilute solution (0.1 M) of the resulting acid chloride was then added dropwise to a refluxing solution of DMAP (3 equiv) in benzene. However, this condition gave rise to the hydroxyacid monomer resulting from DMAP-induced cleavage of the ester linkage. Standard Yamaguchi lactonization conditions,[18] involving slow addition of the mixed anhydride (seco acid, 2,4,6-trichlorobenzoyl chloride, Et3N, THF, rt) to a refluxing solution of DMAP in toluene, gave the same result. Gratifyingly, attempted macrolactonization under Yonemitsu conditions (2,4,6-trichlorobenzoyl chloride, Et3N, DMAP, and the seco acid in benzene, 1 × 10–3 M)[19] at room temperature gave rise to the desired diolide 14 in 50% yield. Hydrogenation of 14 over Pearlman’s catalyst gave 3,3′-di-O-methyl ardimerin 15 in 90% yield. Interestingly, attempted acylation of 15 (Ac2O, Pyr, rt, 16 h or Ac2O, i-Pr2NEt, DMAP)[20] led to none of the desired peracetate and the production of numereous side products. Ultimately, it was found that stirring 15 in neat acetyl chloride[21] overnight led to formation of peracetate 16 in 85% yield. The β-stereochemistry of the glucosyl moieties was indicated by the 8.5 Hz coupling constant of the C.1 proton, and the connectivity of the molecule was verified by 1H–1H COSY and NOESY experiments (see Supporting Information).

Scheme 3

Seco-acid Macrolactonization

In our attempts to access ardimerin by selective cleavage of the C.3 and C.3′ methyl ethers, treatment of 14 with BCl3/CH2Cl2 (rt, overnight), AlCl3/NaI (80 °C, CH3CN, 3 h), or MgI2 (50–80 °C, toluene)[22] initially led to no starting material conversion, but after a prolonged reaction and an increase in the number of equivalents of Lewis acid, extensive decomposition products, arising from diolide cleavage, were formed. Similarly, use of sodium ethanethiolate in DMF (100 °C, 2 h)[23] also led to dissolution of the bislactone moeity. These data indicate that removal of the requisite methyl ethers is likely to be successful only on substrates prior to formation of the diolide core of the natural product.

The binding of 15 to duplex DNA was explored by UV and fluorescence spectroscopies. A concentration-dependent red shift in the absorption at λmax = 214 nm in the ultraviolet spectrum of 15 suggested that self-association/aggregation was occurring in an aqueous buffer solution (10 mM Tris-EDTA) at concentrations >1 μM (Figure 3).[24] Thermal denaturation studies showed no significant shift in the T (68 °C) of salmon testes DNA in the presence of 15 at low ligand/DNA ratios.[25] Furthermore, compound 15 displayed relatively limited ability to displace bound ethidium bromide from calf thymus DNA as compared to control compound daunorubicin over the same concentration range (1 × 10–9 M to 4 × 10–7 M; see Supporting Information).[26] These data suggest that 15 has a low affinity for duplex DNA, perhaps indicative of the difficulty in accommodating the bulky chromophore-linked C-glycosyl moieties in the narrow minor groove, the initial site of small-molecule binding to DNA.[27]

Figure 3

UV absorption spectra of 15 (10 mM Tris-EDTA) at varying concentrations; [15] = (0.79, 1.18, 1.95, 2.70, 3.80, 5.56, 8.81, and 11.76) × 10–6 mol L–1 for curves 1–8, respectively.

In summary, we have developed an 11 step synthesis of the 3,3′-di-O-methyl derivative of the natural product ardimerin and have shown that this substance readily aggregates in aqueous solution and has a low apparent affinity for duplex DNA. Current efforts toward the completion of the synthesis of ardimerin are centered around deprotection of the C.3 methyl ether of aldehyde 7.