We first set out to determine the roles and timing of the several redox enzymes in the fma gene cluster. We individually deleted five genes, Af510, Af490, Af480, Af470, and Af440 in A. fumigatus CEA17 akuB^KU80 strain (pyrG89, ΔakuB^KU80), which is deficient in nonhomologous end joining (Supporting Information).¹⁷ In comparison to the isogenic control strain, the metabolic profile of ΔAf510, ΔAf480 and ΔAf470 showed complete abolishment of the formation of 1 (Figure 3B), while ΔAf440 retained production of 1. The ΔAf490 strain showed decreased titers of 1 (86% decreasing in the culture of 4 days using CYA medium). The nonessential role of Af440 is in accordance with that reported by Wiemann et al., in which Af440 (which corresponds to psoF) was shown to be involved in pseurotin biosynthesis.¹⁶ However in contrast to the Wiemann report, in the analysis of the ΔAf470 profile, we found the accumulation of 4 with m/z 471 [M+Na]⁺ (Figure 3B), therefore suggesting a role for Af470 in the oxidative cleavage of the terminal alkene of the dodecapentaenoate side chain into the carboxylic acid present in 1. On the other hand, deletion of Af480 led to the accumulation of a polyene-containing compound 5, with maximum UV absorbance at 333 nm and m/z 451 [M+Na]⁺ (Figure 3B). Compound 5 was isolated from a 6-day liquid culture of A. fumigatus ΔAf480 mutant strain and structurally characterized. On the basis of the ¹H, ¹³C and 2D NMR analyses, 5 was elucidated as 6-demethoxyfumagillin, in contrast to an aldehyde shunt product reported by Wiemann et al.¹⁶ The structure of 5 revealed that the nonheme iron-dependent oxygenase encoded by Af480 catalyzes hydroxylation of C6 in the biosynthesis of 1.

The only remaining unassigned oxygenase in the gene cluster was therefore the P450 gene, Af510, which shows the strongest sequence homology to the multifunctional P450 OrdA¹⁸ in aflatoxin biosynthesis (47% protein identity). Wiemann et al. reported that deletion of Af510 in A. fumigatus led to no detectable accumulation of any intermediates. While the same phenotype was observed here using LC–MS assay conditions, a more careful extraction with hexane and analysis of the hydrocarbon extract by GC-FID revealed the presence of β-trans-bergamotene (3) (Figure 3C). In the isogenic control, no accumulation of 3 was observed. This result suggested that the initial oxidation of 3 is catalyzed by the enzyme encoded by Af510.

The enzyme encoded by Af470 (Fma-ABM) was initially revealed to harbor a DUF4188 conserved domain of unknown function by NCBI Conserved Domain Search. A further homology modeling search with Phyre2¹⁹ showed, however, that Fma-ABM appears to be related to the cofactor-independent ABM superfamily monoxygenases with a ferrodoxin-like fold.²⁰ To further test the role of Fma-ABM, we cloned the intron-less Af470 (for reconstruction of the gene from mRNA, see Supporting Information) into a yeast 2 μm expression vector which was transformed into S. cerevisiae strain BJ5464-NpgA. Upon supplementing 4 to the yeast culture expressing Af470, the biotransformation into 1 was detected (Figure 4A). On the basis of protein structure prediction (Supporting Information, Figure S7A), Fma-ABM was predicted to be a membrane-bound protein. Upon purification of the microsomal fractions of the yeast strain expressing Fma-ABM and incubation with 4, the same conversion to 1 was observed (Figure 4B). These results confirmed the role of Fma-ABM in the oxidative cleavage of 4 to 1.

To confirm the role of the enzyme encoded by Af480 (Fma-C6H), the 6xHis-tagged enzyme was solubly expressed from E. coli BL21 (DE3) and purified to homogeneity (Supporting Information, Figure S5). Bioinformatics analysis showed that Fma-C6H shares a similar sequence with the members of the phytanoyl-CoA dioxygenase (PhyH) superfamily, thereby requiring Fe (II) and α-ketoglutarate (α-KG) for its activity. Incubation of recombinant Fma-C6H with α-ketoglutarate, sodium ascorbate, and substrate 5, however, did not lead to the formation of any hydroxylated product. We therefore reasoned that Fma-C6H may function upstream in the pathway prior to attachment of the polyene portion. To obtain such a substrate, the hydrolysis of 5 under basic condition was performed to afford 6-demethoxyfumagillol (6) (Supporting Information). When recombinant Fma-C6H and 6 were incubated under the same condition as above, complete conversion of 6 to 6-hydroxylfumagillol (7) was obtained (Figure 4C). Furthermore, coincubation of Fma-C6H with the recombinant methyltransferase (Fma-MT) encoded by Af390–400 in the presence of the necessary cofactors (α-ketoglutarate, sodium ascorbate, and SAM) and 6 resulted in the formation of the expected 2 (Figure 4C). Collectively, these studies confirmed the function of Af480 and revealed that 6 (or a related compound) may be a key intermediate in the biosynthetic pathway of 1. To test this hypothesis, feeding of 6 to A. fumigatus blocked in bergamotene formation (deletion of Fma-TC, ΔAf520) restored biosynthesis of 1 as well as the production of 5 (see compilation of chemical complementation traces in Figure 7). With this information in hand, we then turned to investigation of the possible routes of oxidative modification of 3 into 6.

The transformation of 3 into 6 requires at least three oxidation steps to introduce the one C5 hydroxyl group and the two epoxide functionalities. In addition, the cleavage of the C–C bond between C5 and C8 must also take place. Several proposals for this mechanistically unresolved transformation have been advanced, including desaturation of the C5–C6 bond followed by concomitant cleavage of the C5–C8 bond.²¹ In light of the functional assignment of Af440, Af470, and Af480 described above, it appeared likely that the P450 encoded by Af510 (Fma-P450) is a multifunctional P450 that is responsible for most, if not all, of the oxidative steps between 3 and an advanced intermediate such as 6.

We previously showed that expression of Fma-TC alone in S. cerevisiae can lead to accumulation of β-trans-bergamotene (3).¹⁵ To investigate the activity of Fma-P450 toward 3, we cloned the intron-less Af510 into a yeast 2 μm expression vector for microsomal expression. To equip Fma-P450 with the optimal redox partner, we also cloned the A. fumigatus cytochrome P450 oxidoreductase (AfCPR) for coexpression.²² All three genes (encoding Fma-TC, Fma-P450, and AfCPR) were placed under the ADH2 promoter and transformed into BJ5464-NpgA. After four days of culturing followed by extraction with hexanes/EtOAc (1:1), we observed a series of sesquiterpene compounds (6, 8–19) that are derived from 3 (Figure5A). Of these compounds, 10, 11, 14, and 17 were the major products (0.6–1.1 mg/L); 8 and 12 were minor products (0.3–0.5 mg/L); and the remaining (6, 9, 13, 15, 16, 18, and 19) were present at less amounts (<0.3 mg/L). The same set of metabolites was also observed when 3 was directly supplied to the yeast culture expressing only Fma-P450 and AfCPR (Figure 5B). In the absence of 3 or Fma-P450, none of these 13 compounds was observable in the extract. To elucidate the structures of these products, large scale (12 L) fermentation of the triply transformed yeast strain was performed, followed by isolation and characterization of each compound (Figure 5D and Supporting Information).

The array of products recovered from S. cerevisiae clearly showcased the multifunctional capability of Fma-P450 (hydroxylation, epoxidation, rearrangement) in the oxidation of β-trans-bergamotene (3). The large number of shunt products, however, may be a result of the weak expression of fungal membrane-bound P450 in yeast. As a result of the low concentration of the P450, reactive intermediates may be readily intercepted or hydrolyzed to yield off pathway products such as 14 and 17. Therefore, to analyze the function of Fma-P450 under more controlled conditions, we prepared microsomal fractions from the yeast strain that overexpressed Fma-P450 and AfCPR for in vitro assay.

When 3 was incubated with 10 mg/mL of microsomal fractions and NADPH overnight, we detected 11 as essentially the single product in the extract (Figure 6A). Other major compounds observed in vivo, such as 10, 14, and 17, etc. are only present at very low levels when searched for using selective ion monitoring. Similarly, when the assay is repeated using 8 as the substrate, near complete conversion of 8 to 11 was observed with nearly no side products (Supporting Information, Figure S8). To analyze one of the individual steps catalyzed by the multifunctional P450, we assayed the epoxidation of 9 to 6 using the same microsomal extract. As shown in Figure 6B, we could observe incomplete conversion of 9 to 6 only in the presence of Fma-P450, thereby confirming the ability of this enzyme to catalyze the specific formation of epoxide.

Collectively, these in vivo and vitro results confirmed the multifunctional role of Fma-P450, including C5 hydroxylation, coupled or sequential 4e-oxidative cleavage, and rearrangement to yield epoxycyclohexanone, and the tandem epoxidation. At the same time, it is also evident that Fma-P450 alone is insufficient to generate the correct stereoisomer of 5R-demethoxyfumagillol 6, instead the 5S diastereomer 11 being formed in yeast or using yeast microsomes. Therefore, given that a ketone intermediate 12 could be isolated, the transformation of 3 most likely involves a C5 ketone intermediate that is stereoselectively reduced into the 5R isomer.

Several ketoreductases (KRs) such as the 3-KR in sterol biosynthesis and 3-KR for long-chain fatty acid synthase have been reported in S. cerevisiae,²⁵ hence it is possible that an endogenous yeast KR could catalyze the reduction of the C-5 ketone to form the 5S-hydroxy moiety observed in compounds 10, 11, and 13. To test this hypothesis, we prepared both soluble and microsomal extracts of the untransformed host BJ5464-NpgA. We first tested if the isolated ketone 12 could be selectively reduced to 13. As shown in Supporting Information, Figure S9, complete reduction of 12 to 13 was observed. The R-isomer of 13 may be present at a very low concentrations as suggested by ion-monitoring, but the amount was insufficient for purification and characterization. To further test whether this endogenous yeast KR activity is responsible for formation of the other 5S-containing products shown in Figure 5D, we chemically prepared the three ketone substrates, 5-keto-cordycol (21), 5-keto-demethoxyfumagillol (22), and 5-keto-fumagillol (23) (Figure 7D, spectroscopic data in Supporting Information, Tables S17–S19 and Figures S81–S86). These compounds were synthesized from 10, 11, and 2 by PCC oxidation, respectively. When either 21 or 22 was added to the control yeast extract (soluble or microsomal), we observed the corresponding ketoreduction to 10 or 11 (Supporting Information, Figure S10), respectively, thereby validating the involvement of endogenous yeast KRs in the formation of compounds with the 5S hydroxyl functional group.

With the ketones 21–23 in hand, we performed chemical complementation experiments by feeding each of these compounds individually to the ΔAf520-blocked mutant. In each case, complete restoration of biosynthesis of 1 was observed (Figure 7A). Therefore, the C5-ketone moiety in each of the compounds can be processed correctly. Furthermore, chemical complementation of ΔAf510 mutant by 22 restored fumagillin production, indicating that Fma-P450 is not involved in the biosynthetic pathway beyond 22 and is not essential for reduction of the C5 ketone. Lastly, complementation of 23 to the ΔAf480 mutant efficiently restored fumagillin production, hinting that reduction of C5 may occur as the last step in the formation of 2.

We hypothesized that a KR encoded in the Fma gene cluster might be responsible for the stereospecific 5R reduction of the ketone during the biosynthesis of 1. Although no standalone KR is found in the Fma cluster shown in Figure 3A, Af490 encodes a partial PKS in which only the DH-KR domains are present. Upon initial discovery of the fma cluster, the pseudo/partial PKS was assumed to be an inactive enzyme that had no likely role in the biosynthetic pathway. Although deletion of Af490 did not completely abolish the production of 1, we did notice an 86% decrease in the titer of fumagillin (1) (Figure 3B). The incomplete abolishment of the biosynthesis of 1 may be due to the presence of endogenous KRs in A. fumigatus that have overlapping functions to that encoded by the pathway enzyme Af490 (Fma-KR). To investigate the role of Fma-KR, the 110 kDa protein was therefore expressed as a 6xHis tag fusion and purified from S. cerevisiae (Supporting Information, Figure S6). Upon individual incubation of each of the ketone derivatives 21, 22, or 23 with recombinant Fma-KR, we observed complete conversion to the corresponding 5R-hydroxy-containing compounds, 9, 6, or 2, respectively (Figure 8A–C).

Lastly, to examine the product profile of Fma-P450 in the presence of Fma-KR, we repeated the biotransformation of 3 in S. cerevisiae by coexpression of Fma-P450, Fma-KR, and AfCPR. As shown in Figure 8D, the product profile is altered considerably to reflect the presence of the dedicated C5 reductase. The 5R-hydroxy isomers 6 and 9 are now clearly major products of the assay, while the 5S-hydroxy diastereomers 10 and 11 have essentially disappeared. The same results were obtained in the in vitro assay when Fma-P450-containing yeast microsomes were coincubated with purified Fma-KR. Whereas 11 was formed as the single product in the previous assay shown in Figure 6A, inclusion Fma-KR led to the production of the natural diastereomer 6 (Figure 8E).^21b,26

On the basis of the new enzymes characterized and the new compounds isolated and identified in this work, we can now deduce a complete biosynthetic pathway for fumagillin (1), as shown in Figure 9. The pathway begins with the conversion of FPP to β-trans-bergamotene (3) by a membrane-bound Fma-TC.¹⁶ The initial oxidation of 3 by Fma-P450 involves C–H hydroxylation at the bridgehead C5 position to yield 8. Subsequently, a four electron oxidation initiated at C-9 coupled to cleavage of the cyclobutane C5–C8 bond of the bicyclo[3.1.1] core yields the on-pathway intermediate epoxyketone 21. An additional epoxidation reaction also catalyzed by Fma-P450 then furnishes the characteristic bisepoxide ketone 22. A possible mechanism for the Fma-P450-catalyzed conversion of 3 to 22 is shown in Figure 10 and will be elaborated further in the Discussion section.

The diepoxyketone 22 is then subjected to successive C-6 hydroxylation and O-methylation by Fma-C6H and Fma-MT, respectively, to yield 23, which is then stereoselectively reduced by Fma-KR to 5R-hydroxy-seco-sesquiterpene 2. Acylation catalyzed by the Fma-AT with the dodecapentaenoyl group on Fma-PKS to yield 4 has been previously demonstrated.¹⁵ Finally, oxidative cleavage of 4 by the oxygenase (Fma-ABM) encoded by Af470 arrives at 1.

We have also shown that Fma-KR can stereospecifically reduce each of the three potential ketone intermediates 21–23 into the corresponding 5R-hydroxyl compounds, thereby implying that this ketoreduction step may occur at different stages during the biosynthesis of 2. In the proposed pathway, however, we have assigned the enzyme to catalyze reduction of the most advanced intermediate 23. We have shown that compounds 6 and 9, which can be produced by reduction of 22 and 21, respectively, can each be acylated efficiently by Fma-AT to yield the analogues 5 and 20. For example, feeding 6 to the ΔAf520 strain resulted in the production of both 5 and 1 (Figure 7A), with a relative ratio around 4:1. This indicated that C-6 hydroxylation by Fma-C6H may be slower than the C-5 acylation catalyzed by Fma-AT. The formation of 5 and 20 also demonstrates that Fma-AT has significant tolerance toward the substitution pattern of the terpene substrate, and can intercept these 5R-reduced products in vivo. In contrast, we have established that Fma-C6H is highly specific toward the unacylated intermediate 6. Therefore, 5 and 20 cannot be oxidized to 1. However, no trace of 5 or 20 is recovered from A. fumigatus, thereby suggesting that neither 6 nor 9 is an intermediate in vivo. Hence, we proposed that reduction of the C5 ketone takes place after the 6-methoxyl function has been installed in 23, thereby preventing premature C5 acylation of the terpenoid by Fma-AT.

A. fumigatus strain used in this study was A. fumigatus CEA17 akuB^KU80 strain (pyrG89, ΔakuB^KU80), which is deficient in nonhomologous end joining¹⁷ and was maintained on Czapek-Dox agar or glucose minimal agar (GMM).⁴⁰ Details on AFUA_8G00390 and AFUA_8G00470–510 deletion mutants are shown in the Supporting Information. E. coli TOP10 (Invitrogen) and XL1-Blue (Stratagene) were used for DNA manipulation, and BL21 (DE3) was used for protein expression. Saccharomyces cerevisiae strain BJ5464-NpgA (MATα ura3–52 his3- Δ200 leu2- Δ1 trp1 pep4::HIS3prb1 Δ1.6R can1 GAL) was used as the yeast expression host.

A. fumigatus genomic DNA was prepared using CTAB isolation buffer as described elsewhere.⁴¹ Polymerase chain reactions were performed using Phusion DNA Polymerase (New England Biolabs) or Platinum Pfx DNA polymerase (Invitrogen). DNA restriction enzymes were used as recommended by the manufacturer (New England Biolabs). RNA extraction was performed using a RiboPure Yeast Kit (Ambion), and ImProm-II Reverse Transcription System for RT-PCR (Invitrogen) was used to synthesize complementary DNA (cDNA) from total RNA. PCR products were subcloned to a pCR-Blunt vector (Invitrogen) and confirmed by DNA sequencing. Primers used to amplify the genes were synthesized by Integrated DNA Technologies and are listed in Supporting Information, Table S2. In vivo yeast recombination cloning was performed by transforming the S. cerevisiae BJ5464-NpgA with DNA fragments with >35 bp overlaps and includes a 2 μm plasmid backbone (derived from YEplac195 or YEplac112) using an S.c. EasyComp Transformation kit (Invitrogen).

For production of 1 and other metabolites, A. fumigatus and mutant strains were cultured in CYA medium (Czapek-Dox agar supplemented with 5 g/L yeast extract). Total RNA for RT-PCR was extracted from A. fumigatus grown on Czapek-Dox liquid medium with 5 g/L yeast extract (CYB) after 4 days of cultivation. For chemical complementation of ΔAf520, ΔAf510, and ΔAf480 mutants, purified compounds are supplemented to the CYA medium at 0.2 mg/mL individually. LC–MS analyses of conversion of 7 to 8 by Af480 and 7 to 2 by Af480 and Af390–400 were performed on Prevail 3 μm, 2.1 × 100 mm² C18 reversed-phase column (Alltech) and separated on a 5–95% (v/v) CH₃CN linear gradient in H₂O supplemented with 0.05% (v/v) formic acid at a flow rate of 125 μL/min. All LC–MS analyses except the aforementioned were performed on a Shimadzu 2010 EV LC–MS (Phenomenex Luna, 5 μm, 2.0 × 100 mm², C18 column) using positive and negative mode electrospray ionization with a linear gradient of 5–95% MeCN-H₂O in 30 min followed by 95% MeCN for 15 min with a flow rate of 0.1 mL/min. ¹H, ¹³C, and 2D NMR spectra were obtained on Bruker AV500 spectrometer with a 5 mm dual cryoprobe at the UCLA Molecular Instrumentation Center.

Af390 and Af400 were obtained as one transcript by RT-PCR verifying that Af400 was misannotated (for revised annotation, see Supporting Information). The DNA fragemnt of Af390–400 and Af480 from cDNA were inserted into pET21 (Novagen) digested with NdeI and XhoI to yield pKW20174 and pKW20172, respectively. Primers used for the amplification and cloning are listed in Supporting Information, Table S2. Recombinant enzymes were expressed with C-terminal His₆-tagged in E. coli BL21 (DE3) and purified by nickel affinity chromatography. The cells were cultured at 37 °C, 250 rpm in 500 mL of LB medium with 35 μg/mL carbenicillin. Isopropylthio-β-d-galactoside (IPTG, 0.1 mM) to induce protein expression was added at OD₆₀₀ between 0.4 to 0.6 and the cells were further cultured for 12–16 h at 16 °C. The cells were then harvested by centrifugation (3500 rpm, 15 min, 4 °C), resuspended in ∼25 mL of lysis buffer (100 mM Tris-HCL, pH 7.4, 0.1 M NaCl, 20 mM imidazole), followed by lysed by sonication on ice. Cell debris was removed by centrifugation (15 000 rpm, 30 min, 4 °C). The His6-tagged proteins were purified by using Ni-NTA agarose (Qiagen) according to manufacturer’s instructions. Purified enzyme was concentrated and exchanged into buffer A (50 mM Tris-HCl, pH 7.9, 2 mM EDTA, 2 mM DTT) + 10% glycerol with the centriprep filters (Amicon) and stored at −80 °C for enzyme assays.

For in vitro synthesis of 2 and 7, 10 μM Fma-C6H480 was incubated with 20 μM substrate 6, 1 mM sodium ascorbate, 1 mM α-ketoglutarate, and 100 mM NaCl in 100 mM Tri-HCl (pH 7.4). For in vitro synthesis of 2, the assay was performed using the same condition as above with additional 1 mM SAM and 10 μM Fma-MT. The reaction was incubated at 1 h and extracted twice with ethyl acetate. The organic phases were dried and dissolved in 20 μL of MeOH and subjected for analysis by LC–MS as described in Chemical Analysis.

For biotransformation in S.c. of Fma-P450 and AfCPR, S. cerevisiae strain BJ5464-NpgA harboring both Fma-P450 and AfCPR plasmid were inoculated to 4 mL of Yeast Synthetic Drop-Out medium without uracil and leucine. The cells were grown for 72 h with constant shaking at 28 °C. A 15 μL aliquot of the seed culture was inoculated with 2 mL of YPD (10 g yeast extract, 20 g peptone, and 950 mL of Milli-Q water) supplemented with 1% dextrose. 3, 8, and 9 (0.5 mg in 10 μL of DMSO) was added to the culture after 48 h at 28 °C with shaking and the cells were cultivated for another 24 h. The cultures were extracted by hexanes-ethyl acetate (1:1) twice, the organic layers were concentrated in vacuo and redissolved in 100 μL of MeOH. A 10 μL aliquot of samples was further analyzed by LC–MS with the method described in Chemical Analysis. For biotransformation in S.c. of Af470, the culture of S. cerevisiae strain BJ5464-NpgA harboring Af470 was prepared by a similar method above and 4 (0.1 mg in 10 μL of DMSO) was added to the culture after 48 h.

Details in preparation of Fma-P450 and AfCPR-containing microsomes for in vitro assay are shown in the Supporting Information. For in vitro microsomes assay, 10 mg/mL (wet weight) microsomal fractions containing Af510 and AfCPR, 1 mM substrates, 2 mM NADPH, and NADPH regeneration system (BD) solution A (5 μL) and B (1 μL), and 100 mM PBS, pH 7.4 were incubated in a 100 μL reaction. The reaction was incubated at room temperature for overnight and extracted with 100 μL of hexanes-ethyl acetate (1:1) twice. The organic phase was dried and redissolved in 20 μL of MeOH for analysis by LC–MS. The amount of protein in 10 mg/mL microsomes was calculated to be 180 μg/mL based on a modified Bradford assay against a BSA standard curve (protein samples were predenatured in 0.1 M NaOH).

S. cerevisiae BJ5464-NpgA was transformed with pHCfmaKR. For 1 L culture, the cells were grown at YPD (10 g/L yeast extract, 20 g/L peptone) supplemented with 1% dextrose and incubated at 28 °C with shaking for 72 h. The cells were harvested by centrifugation (3750 rpm at 4 °C for 10 min), and the cell pellet was resuspended in 20 mL of lysis buffer (50 mM NaH₂PO₄, 150 mM NaCl, 10 mM imidazole, pH 8.0) and lysed by sonication on ice in one minute intervals until homogeneous. To remove cellular debris, the homogeneous mixture was centrifuged at 17000 rpm for 1 h at 4 °C. Ni-NTA agarose resin was added to the supernatant (2 mL) and the solution was stirred at 4 °C overnight. Soluble Fma-KR was purified by gravity-flow column chromatography with increasing concentrations of imidazole in Buffer A (50 mM Tris-HCl, 500 mM NaCl, 20 mM–250 mM imidazole, pH 7.9). Purified protein was concentrated and buffer was exchanged into Buffer B (50 mM Tris-HCl, 2 mM EDTA, 100 mM NaCl, pH 8.0) using an Amicon Ultra-15 Centrifugal Filter Unit and stored in 10% glycerol. The purified Fma-KR was analyzed by SDS-PAGE (Supporting Information, Figure S6) and their concentration was calculated to be 8.7 mg/L, using the Bradford assay with BSA as a standard.

For in vitro synthesis of 6, 9, and 2, 10 μM Fma-KR was incubated with 1 mM substrates 21–23, respectively, 2 mM NADPH in 100 mM PBS, pH 7.4 in a total 100 μL reaction. The reaction was incubated at room temperature overnight and extracted with 100 μL of hexanes–ethyl acetate (1:1) twice. The organic phase was dried and dissolved in 20 μL of MeOH for analysis on LC–MS.

The method of oxidation by PCC followed that of Asami et al.⁴² A suspension of PCC (1 mg, 0.0046 mmol) and powdered molecular sieves 4A (3.0 mg) was added to a solution of 2 (2.3 mg) in 0.5 mL of CH₂Cl₂. The mixture was stirred in an ice–water bath and for 2.5 h at room temperature. Florisil and CH₂Cl₂ were added to the mixture, and the suspension was filtered through a combination of Celite and Florisil. The filtrate was concentrated in vacuo. The residue was purified by Silica gel plate (Merck, TLC Silica gel 60 F₂₅₄, glass plates) with 25% acetone-hexanes developed by two times to give 23 (1.1 mg). A similar method was used as above to obtain 21 (0.8 mg) and 22 (1.0 mg) from 10 (2.2 mg) and 11 (3.0 mg), respectively.