Autoxidation of a C2-Olefinated Dihydroartemisinic Acid Analogue to Form an Aromatic Ring: Application to Serrulatene Biosynthesis

Abstract

Dihydroartemisinic acid (DHAA) is a plant natural product that undergoes a spontaneous endoperoxide-forming cascade reaction to yield artemisinin in the presence of air. The endoperoxide functional group gives artemisinin its biological activity that kills Plasmodium falciparum, the parasite that causes malaria. To enhance our understanding of the mechanism of this cascade reaction, 2,3-didehydrodihydroartemisinic acid (2,3-didehydro-DHAA), a DHAA derivative with a double bond at the C2-position, was synthesized. When 2,3-didehydro-DHAA was exposed to air over time, instead of forming an endoperoxide, this compound predominantly underwent aromatization. This olefinated DHAA analogue reveals the requirement of a monoalkene functional group to initiate the endoperoxide-forming cascade reaction to yield artemisinin from DHAA. In addition, this aromatization process was exploited to illustrate the autoxidation process of a different plant natural product, dihydroserrulatene, to form the aromatic ring in serrulatene. This spontaneous aromatization process has applications in other natural products such as leubethanol and erogorgiaene. Due to their similarity in structure to antimicrobial natural products, the synthesized compounds in this study were tested for biological activity. A group of the tested compounds had minimum inhibitory concentration (MIC) values ranging from 12.5 to 25 μg/mL against the bacterial pathogen Staphylococcus aureus and the fungal pathogen Cryptococcus neoformans.

Graphical Abstract

Dihydroartemisinic acid (DHAA, 1) is the biosynthetic precursor to artemisinin (2), an endoperoxide-containing plant natural product used to treat malaria (Figure 1). When DHAA is exposed to air, a cascade reaction involving (i) ene reaction to form an allylic hydroperoxide, (ii) C–C bond cleavage, (iii) incorporation of a second molecule of oxygen, and (iv) polycyclization to form artemisinin occurs.^1,2 Due to the important biological activity of the endoperoxide bridge,^3–7 there have been a number of previous studies^8–11 dedicated to elucidating the mechanism of endoperoxide formation in artemisinin biosynthesis. Furthermore, there have been numerous studies focused on the topic of artemisinin synthesis^12–14 and biosynthesis.^15,16 Either singlet oxygen¹⁷ or ozone¹⁸ has been reported to lead to the construction of the endoperoxide.

Figure 1.

Dihydroartemisinic acid (DHAA, 1) was previously shown to convert to artemisinin (2), and this reaction was promoted by light.^1,2

The purpose of this study was to probe the mechanism of this spontaneous cascade reaction (Figure 1) by synthesizing a 2,3-olefinated DHAA analogue (Figure 2, 3, 2,3-didehydro-dihydroartemisinic acid or 2,3-didehydro-DHAA) and determining whether or not this compound would undergo the same endoperoxide-forming process as DHAA (Figure 2A). The introduction of a double bond in the C2-position of DHAA would allow us to test the hypothesis of an alternative endoperoxide formation process through a [4+2] cycloaddition or the similar endoperoxide-forming process observed with DHAA (Figure 1). Based on other naturally occurring small molecules, the [4+2] cycloaddition product between oxygen and the diene was a valid hypothesis. For instance, a similar conversion of a different diene-containing natural product, α-terpinene, reacts with singlet oxygen to form a [4+2] product to yield the endoperoxide-containing natural product ascaridole.¹⁹ However, in this study, when 2,3-didehydro-DHAA was left open to air over time, neither endoperoxide moieties presented in Figure 2A were formed (4 or 5). Instead, 2,3-didehydro-DHAA underwent aromatization to afford compound 6 (Figure 2B, 3 to 6). In the latter part of this study, we applied the knowledge of this spontaneous aromatization process to explore the biosynthesis of the aromatic ring formation of a different natural product, serrulatene (Figure 3B).

Figure 2.

Summary of the results for the aromatization of 2,3-didehydro-dihydroartemisinic acid (2,3-didehydro-DHAA) (3 to 6) presented in this report.

Figure 3.

(A) Oxidation of α-terpinene to ascaridole and p-cymene by singlet oxygen (7 to 8 and 9). (B) Autoxidation of dihydroserrulatene to serrulatene (10 to 11). (C) Oxidation of miltiradiene to abietatriene (12 to 13). (D) Structures of aromatic ring-containing natural products erogorgiaene, leubethanol, and 7-hydroxycalamanene (14, 15, and 16).

Figure 3A–C show other examples of natural products bearing a diene functional group. α-Terpinene is known to undergo a [4+2] reaction with molecular oxygen to form ascaridole and a minor aromatization pathway to form p-cymene (Figure 3A, 7 to 8 and 9).²⁰ An isomer of α-terpinene, α-phellandrine, is also known to undergo endoperoxide formation as well as multiple ene reactions to form a mixture of allylic hydroperoxides and an aromatization to form p-cymene.²¹ Dihydroserrulatene has been suggested to undergo aromatization to serrulatene (Figure 3B, 10 to 11).²² Miltiradiene has been postulated to autoxidize to form the aromatized compound 13 (Figure 3C, 12 to 13).²³ Other natural products hypothesized to form an aromatic ring from a diene precursor include erogorgiaene,^24–26 leubethanol,²⁷ and 7-hydroxycalamanene²⁸ (Figure 3D, 14, 15, and 16). Due to the antibacterial activity of serrulatene and its derivatives, the syntheses of these compounds have been previously reported.^29–40

RESULTS AND DISCUSSION

Synthesis of [3-²H]-2,3-Didehydro-DHAA (19) and 2,3-Didehydro-DHAA (6).

In our previous study with synthesizing 3,3-dideuterodihydroartemisinic acid, a key step involved the deuteration of an enone intermediate with LiAlD₄ and AlCl₃, which resulted in the isolation of diene 17 (Scheme 1A). This diene intermediate (17) with the C12-alcohol was oxidized to the aldehyde (18) with Dess-Martin periodinane, and subsequently to the carboxylic acid (19) under Pinnick oxidation conditions.

Scheme 1.

(A) Synthesis of 3-Deutero-2,3-didehydro-DHAA (19) to Test the Hypothesis of Endoperoxide Formation (Figure 2A); (B) Synthesis of 2,3-Didehydro-DHAA (3) from the Allylic Alcohol (20)

Furthermore, to form the nondeuterated version of the diene (Figure 2, compound 3), the allylic alcohol intermediate (20)¹ was treated with methanesulfonyl chloride and triethylamine to directly yield the eliminated product (21) (Scheme 1B). The resulting acetate was deprotected with methyllithium, and the alcohol (22) was oxidized with Dess-Martin periodinane followed by the Pinnick oxidation to afford 2,3-didehydro-DHAA (22 to 23 to 3). The crystal structure of compound 22 is shown in Figure 4.

Figure 4.

Crystal structure of compound 22.

NMR Time Course Shows Aromatization of the Diene (3 and 19) to Form the Aromatic Ring (6 and 24).

Initially, the 2,3-didehydro-DHAA and 3-deutero-2,3-didehydro-DHAA derivatives (3 and 19) were used to probe whether the endoperoxide-forming cascade reaction occurs (e.g., Figure 2A, 3 to 6). These compounds were placed under identical conditions as previously performed with DHAA;¹ low milligram quantities were dried down in clear glass vials and placed by the window. The contents of the vials were monitored over time using ¹H NMR spectroscopy (see SI, Part 1). Notably, the appearance of aromatic protons coincided with the loss of the vinylic protons of the starting material for both 3 and 19 after 7 days, confirming their conversion to 6 and 24, respectively (SI, Part 1). The monodeuterated aromatic compound 24 was subsequently used as an internal standard to quantify the aromatization process as described in the following section.

LCMS Time Course Enables Quantification of Aromatization of Diene 3 to Aromatic Compound 6.

An assay using liquid chromatography–mass spectrometry (LC-MS) was developed to determine whether this new aromatization process was dependent on light (Figure 6A). Three different conditions were tested with vials containing the C2-olefinated DHAA derivative (3): (i) UV-C light, (ii) dark, and (iii) UV-AB light (Table 1). Under the UV-AB light, the lamp caused the temperature to be above ambient temperature. In all three conditions, diene 3 converted to aromatic compound 6, and this process was quantified with the use of deuterated compound 24 as an internal standard (Figure 6). The carboxylic acid derivatives (6 and 24) were measured through negative mode electrospray ionization (ESI). A possible allylic hydroperoxide (25) was detected with m/z 265 with a retention time (t_R) of 3.09 min (Figure 6B). This hydroperoxide was also observed when diene 3 was exposed to ¹⁸O₂ (see SI, Part 6).

Figure 6.

(A) LCMS time course showing conversion of diene 3 to aromatic compound 6 under UV-C light. The formation of a possible endoperoxide intermediate is observed, which rearranges to an allylic hydroperoxide and then the aromatic compound. Monodeuterated aromatic compound 24 was used to quantify aromatization of the diene. Sample LC-MS trace showing extracted ion chromatograms of compounds 6, 24, and 25 (m/z 231, 232, and 265: top, middle, bottom), respectively, at (B) 24 h and (C) 550 h.

Table 1.

Summary of Time Course Studies of the Aromatization of 2,3-Didehydro-DHAA (3) (Concentrated) Using LC-MS

entry	time	conditions	IS?	m/z 231a	m/z 232a	6 amt	conversionb	yield
1	24 h	dark	yes	50 318 822	15 538 211	10 μg	41%	15%
2	24 h	dark	no	57 261 352	6 309 792
3	24 h	UVAB	yes	45 619 186	15 952 665	8.5 μg	97%	13%
4	24 h	UV-AB	no	59 315 504	6 798 393
5	24 h	UV-C	yes	31 639 892	18 138 733	4.4 μg	93%	6.6%
6	24 h	UV-C	no	35 660 089	4 050 716

^aThe numbers indicate the areas under the corresponding peaks. The masses for the nondeuterated and deuterated aromatic compounds 6 and 24 are m/z 231 and 232, respectively. IS: internal standard. (See SI, Part 10, for raw data.)

^bConversion calculated by comparing the area of diene 3 (m/z 233) for each time point with the area of diene 3 at time point 0 h.

At 24 h, 4.4 μg of aromatic compound 6 was formed under UV-C light (Figure 6B and Table 1, entry 5), which was less than the 10 μg of aromatic compound 6 product detected when the sample was left in the dark (Table 1, entry 1). On the other hand, when diene 3 was left under UV-C light for a longer period of time (550 h), 6.3 μg of aromatic compound 6 was detected (Figure 6C; also see SI, Part 10).

Therefore, unlike the endoperoxide formation in the conversion of DHAA to artemisinin (Figure 1, 1 to 2),^1,2 there is no rate acceleration in the presence of light in the spontaneous aromatization process of 2,3-didehydro DHAA to the aromatic compound (i.e., Figure 5, 3 to 6). However, similar to transformation of DHAA to artemisinin,² there was a concentration dependence in the rate of autoxidation of diene 3 to compound 6 in that more of the aromatic ring was formed when there was more diene 3 in the vial (see SI, Part 10, 68 μg vs 6.8 μg). In addition, we tested whether the C12-carboxylic acid moiety of 3 plays a role during this aromatization process by subjecting alcohol 22 and aldehyde 23 in the identical aromatization conditions (SI, Part 2). Based on the comparison of the rates of autoxidation, the carboxylic acid may possibly play a role in accelerating the aromatization process (i.e., see SI, Part 2, in 24 h, aromatic product formed from alcohol 22, aldehyde 23, and carboxylic acid 3 in the dark: 0.22, 3.2, and 10 μg, respectively).

Figure 5.

¹H NMR time course showing aromatization of (A) 2,3-didehydrodihydroartemisinic acid (3) and (B) 3-deutero-2,3-didehydrodihydroartemisinic acid (19). See SI, Part 1, for NMR spectra.

To further investigate the “inhibiting” effects of UV-C light in the aromatization process (Figure 6A), diene 3 was left in milligram quantities in a vial and left under UV-C light over time. After 287 h, the contents of the vial were dissolved in CDCl₃ and analyzed by proton NMR spectroscopy (see SI, Part 8). The crude NMR spectrum showed complete consumption of the diene starting material. The NMR spectrum indicated the proton signals for the aromatized compound and a second set of protons (SI, Part 8), which corresponded to a more polar product (lower R_f value based on TLC analysis). Similarly, when diene 3 was left under UV-AB light, the diene was completely consumed after 21 h, and the new set of proton signals appeared (SI, Part 8). In contrast, when the same amount of diene starting material was left in the dark in a vial, the amount of the aromatic compound formed was similar, but there was still a significant amount of the diene starting material remaining (SI, Part 9).

Moreover, the decrease in the detection of aromatic compound 6 over time (Figure 6B to C, 24 to 550 h) suggested that the UV-C light converts the aromatized compound 6 into other products. In order to confirm this possible degradation of compound 6 under UV-C light, compound 6 was left under UV-C irradiation over 21 days (see SI, Part 11). In this time course, two different vials were used: (i) one vial (vial 1) was used to consecutively monitor the different time points (i.e., 23 h, 5 days, 9 days, 20 days, and 21 days) by dissolving the sample in CDCl₃ (0.7 mL for each time point) and after the NMR spectrum was obtained, the solution was dried down in the same vial under N₂ flow, and (ii) for the second vial (vial 2), the sample stayed irradiated under UV-C light until the final time point (21 days) without perturbing the system.

Irradiating compound 6 under UV-C light resulted in the formation of new proton signals as determined by NMR (SI, Part 11). In addition, the vial that had been perturbed throughout five different time points (i.e., vial 1) had degraded the aromatic compound 6 at a faster rate compared to the vial that remained undisturbed (i.e., vial 2) (see SI, Part 11). This slower degradation of aromatic compound 6 in the latter vial suggests that the solid residue of the aromatized compound 6 forms a lattice and the top layer of the lattice blocks the effects of the UV-C light on the rest of the material. A similar effect was observed when a solid residue of DHAA was left in a vial under UV light.²

Singlet Oxygen Promotes Aromatization of Diene 3 to Compound 6 in Solution.

In order to clarify the potential role of singlet oxygen in the aromatization process of diene 3 to compound 6, diene 3 was exposed to singlet oxygen (Scheme 2). The diene was dissolved in CH₂Cl₂, and singlet oxygen was generated in solution in the presence of methylene blue, which was irradiated with a flood lamp. Under these conditions, the diene (3) had aromatized to compound 6 in 30 min. Furthermore, in the presence of singlet oxygen, a minor amount of endoperoxide 4 was formed.

Scheme 2.

Singlet Oxygen Promoted Aromatization of Diene 3 to Form Compound 6 and Also Formed a Minor Amount of Endoperoxide 4^a

^aReaction time: 70 minutes; within this time, all starting material (3) was consumed. Percentages are isolated yields.

Mechanistic Proposal for Aromatization of Diene 3 to Compound 6.

Based on the experimental results with the conversion of diene 3 to compound 6 presented above, a mechanism is proposed for the aromatization reaction (Figure 7).

Figure 7.

Mechanistic proposals to compound 6 from diene 3 through reaction with molecular oxygen. The carboxylate of a neighboring molecule of 3 or the conjugate base of H₂O₂ could be the base that deprotonates any of the intermediates (e.g., 3A or 3C).

The main result excluding the possibility of an endoperoxide intermediate to form the aromatized compound was the isolation of a stable endoperoxide during singlet oxygen conditions (Scheme 3, 3 to 4). The aromatization of this cyclic diene (3 to 6) is in contrast to a different system: when the natural product α-terpinene (a monocyclic diene) is exposed to singlet oxygen, the major product is the endoperoxide ascaridole (see SI, Part 8, and Figure 3A, 7 to 8).

Scheme 3.

Synthesis of Acetate 33 from Dihydroartemisinic Alcohol 26

There are three possible allylic peroxides that can be formed through removal of C6–H (3A), C1–H (3B), or C15–H (3C) (Figure 7). The elimination of H₂O₂ will yield the aromatized product (6). The observation of an allylic hydroperoxide intermediate is supported in experiments that involved ¹⁸O₂ where an intermediate with the mass of [M + 36] (m/z 269) was observed (see SI, Part 6). Furthermore, the photooxygenation of a similar diene system that lacks a carboxylic acid moiety has been reported to yield allylic hydroperoxide intermediates by Sasson and Labovitz.⁴¹

Synthesis of 9,10-Dihydroserrulatene (38) to Investigate Aromatization to Serrulatene (11).

Since 2,3-didehydro-DHAA aromatized to compound 6 (Figure 7, 3 to 6), this autoxidation process was exploited toward understanding the biosynthesis of serrulatene from dihydroserrulatene (Figure 3B, 10 to 11). Instead of the known biosynthetic precursor dihydroserrulatene (i.e., 8,9-dihydroserrulatene, 10), we used our strategy to eliminate a mesylate intermediate to eventually access 9,10-dihydroserrulatene (38). The synthesis of 9,10-dihydroserrulatene began with dihydroartemisinic alcohol 26 (Scheme 3 and SI, Part 3).

Because of the reactivity of the 1,3-diene system (i.e., autoxidation of diene 3 to aromatic compound 6), the diene was introduced after the side chain was extended from the dehydrodecalin system of dihydroartemisinic alcohol (26) as the acetate 33 (Scheme 3).

Dihydroartemisinic alcohol was oxidized to the aldehyde (27) with Dess-Martin periodinane (Scheme 3). The resulting aldehyde was treated with 3 equiv of the Wittig reagent to yield methyl enol ether 28, which was isolated as an E:Z (59/41) mixture as determined by the integration of the proton signals and assignment based on the J value of the vinyl proton (J = 13 or 6.5 Hz for the E or the Z isomer, respectively). Furthermore, the PPh₃ from the excess Wittig reagent coeluted with the product after purification via silica gel column chromatography. Therefore, the purified material was stirred in iodomethane/THF (30 mL/20 mL), which resulted in the formation of the phosphonium salt that crashed out as a white solid and was subsequently removed with a short pad of silica gel. This protocol to remove the PPh₃ was used in the forthcoming steps involving the Wittig reaction (29 to 30 and 37 to 38). Methyl enol ether 28 was subsequently hydrolyzed with HCl in THF and H₂O to yield aldehyde 29. Aldehyde 29 was subjected to the second sequence of the Wittig reagent to yield methyl enol ether 30, which was isolated as an E:Z (48/52 based on ¹H NMR integration) mixture. Methyl enol ether 30 was hydrolyzed to yield aldehyde 31. The aldehyde was reduced with NaBH₄ (31 to 32), and the resulting alcohol (32) was protected with acetyl chloride and Et₃N in CH₂Cl₂ to give acetate 33.

The acetate (33) was oxidized with SeO₂ to yield allylic alcohol 34 (Scheme 4). The resulting alcohol was eliminated with MsCl to give the diene 35. The acetate was deprotected with MeLi to afford primary alcohol 36. The primary alcohol was converted to dihydroserrulatene (38) through oxidation with Dess-Martin periodinane to furnish aldehyde 37. The resulting aldehyde (37) was treated with the Wittig reagent to yield 9,10-dihydroserrulatene (38).

Scheme 4.

Synthesis of 9,10-Dihydroserrulatene 38 from Acetate 33

Spontaneous Aromatization of 9,10-Dihydroserrulatene (38) to Serrulatene (11).

Once 9,10-dihydroserrulatene (38) was synthesized, two conditions were tested to monitor the aromatization process: (i) singlet oxygen in solution and (ii) dry conditions in the dark. Under singlet oxygen conditions, the alkene in the side chain did not stay intact; instead, a dioxetane (40) formed (Figure 8). The observation that the olefin in the side chain of 38 reacts with singlet oxygen (Figure 8) suggests that singlet oxygen does not play a role in the biosynthesis of the aromatic ring of serrulatene, leubethanol, or erogorgiaene.

Figure 8.

Spontaneous oxidation of 9,10-dihydroserrulatene to serrulatene (38 to 11). Singlet oxygen formed dioxetane 40. The numbering of 38 follows the same numbering for erogorgiaene.²⁵

Under conditions where 9,10-dihydroserrulatene (38) was left to spontaneously oxidize in the dark (Figure 9, over 110 days), serrulatene was successfully formed (38 to 11), presumably through the allylic peroxide intermediate 39 (Figure 8). The NMR signals for serrulatene (see SI, Part 1) matched the previously reported values of the synthesized standard.³¹ Furthermore, an authentic standard of serrulatene (11) was synthesized in this study and the ¹H NMR signals of the authentic standard matched the autoxidation product of 9,10-dihydroserrulatene (see SI, Part 3).

Figure 9.

¹H NMR time course showed aromatization of compound 38 to serrulatene (11) over time (see SI, Part 1).

Cascade Diels–Alder/Autoxidation Transformation between Two Diene Moieties to Yield a Heterodimer of Artemisinin Covalently Bound to DHAA.

In order to clarify whether aromatization of diene 3 was dependent on different wavelength ranges of light, diene 3 was placed under three conditions: (i) irradiation with a UVA and UVB full spectrum (280–400 nm) sun lamp (75 W), (ii) irradiation with a UVC light (100–280 nm), and (iii) with no light. For all conditions, a mass that corresponded to a dimerization and oxidation product of diene 3 was detected at m/z 513.2858 in the negative electrospray mode (ESI negative mode; see SI, Part 8). The new mass is speculated to arise from a Diels–Alder reaction between two molecules of diene 3. After the dimerization to 41, one of the components resembles DHAA, which in turn undergoes oxidation to artemisinin to form 42.

Furthermore, evidence supporting this artemisinin–DHAA dimer is the observation of the light-driven rearrangement of the endoperoxide (Figure 10, 43), which ionizes as the reduced acetonitrile adduct in the positive electrospray ionization (ESI positive) mode at m/z 560.3582 (46 or 47) (see SI, Part 8).

Figure 10.

Observation of a possible Diels–Alder adduct between two molecules of diene 3 that undergoes spontaneous oxidation and rearrangement to 42 and 43. The structures are proposed by high-resolution mass spectrometry (see SI, Part 8).

Antimicrobial Activity of Compounds from this Study.

Since erogorgiaene, leubethanol (Figure 3D, 14 and 15), and other serrulatene derivatives^42–48 have reported antibacterial activities, we tested the antimicrobial activity of some of the synthesized derivatives reported in this paper. Specifically, we sought to determine the range of antimicrobial activities of these compounds against both Gram-positive and Gram-negative bacteria as well as fungal pathogens. For this determination, we selected a representative human pathogen for each of these three categories of microbes. Pseudomonas aeruginosa was selected as the representative Gram-negative pathogen due to its ability to cause problematic acute and chronic infections on multiple sites of the human body.⁴⁹ Staphylococcus aureus was selected as the representative Gram-positive pathogen because this microbe is a natural component of the human microbiome that can cause infections in most of the same host niches as P. aeruginosa, including lung and wound infections.⁵⁰ Cryptococcus neoformans was selected as the representative fungal pathogen due to its ability to cause devastating human infection in numerous sites of the body including urine, bloodstream, and cerebrospinal fluid, especially in patients with a prior immunocompromised status.⁵¹ Antimicrobial resistance is an emerging threat in each of these three pathogens.^52–54 Therefore, identification of new antimicrobials targeting these problematic microorganisms is critical. We compared the activity of our synthesized compounds with various control drugs (ciprofloxacin, CIP; novobiocin, NOV; gentamicin, GEN; and amphotericin B, AMP) that are commonly being used to treat the pathogens mentioned above. For example, ciprofloxacin and gentamicin are active against P. aeruginosa and S. aureus infection,^55–57 whereas novobiocin is used to treat methicillin-resistant S. aureus.⁵⁸ Amphotericin B is a well-known antifungal agent used to treat fungal infections for several decades.⁵⁹

Resazurin dye was used to measure cell viability after exposure of the cells to each compound.⁶⁰ While no compounds exhibited effects against P. aeruginosa at any of the tested concentrations, certain compounds did exhibit promising anti-staphylococcal and/or anti-cryptococcal effects. These results are summarized in Figures 11 and 13.

Figure 11.

Most active antimicrobial compounds that were identified using the resazurin dye assay in this study (22, 27, 28, 32, and 33). The numbers at the top, middle, and bottom are the MIC values in μg/mL against JE2 (S. aureus), H99 (C. neoformans), and PA14 (P. aeruginosa) cells, respectively.

Figure 13.

Summary of other compounds tested for antimicrobial activity. ^aMIC values in μg/mL concentration. Not shown: artemisinin (2) with MIC values of >100, 100, and >100 μg/mL against JE2, H99, and PA14 cells, respectively. See SI, Part 12, for a more detailed discussion.

Compounds 22, 27, and 33 had at least 25 μg/mL minimum inhibitory concentration (MIC) values against C. neoformans (Figure 11). Among the control drugs used in this study, AMP and GEN showed potential activity against C. neoformans with a MIC of <0.048 and 6.25 μg/mL, respectively. Compounds 28 and 32 had at least 12.5 μg/mL MIC values against S. aureus. In comparison, NOV, GEN, and CIP showed potential killing effects against S. aureus with a MIC of <0.19, 6.25–12.5, and 12.5–25 μg/mL, respectively (see SI, Part 12). None of the synthesized compounds in this study had potent activity against P. aeruginosa. On the other hand, the control drugs CIP and GEN showed potential efficacy against P. aeruginosa with a MIC of <0.04 and 0.39 μg/mL, respectively.

The most potent compounds against C. neoformans with MIC values of 12.5 μg/mL were compounds 27 and 33, while the most potent compounds against S. aureus with MIC values of 12.5 μg/mL were compounds 28 and 32 (Figure 11). These compounds all have one double bond between C4 and C5 in the cis-dehydrodecalin system. In contrast, the compounds with the 2,4-diene system (Figure 13, 3, 22, 23, and 36; also see SI, Part 12) generally had less potent antimicrobial activity.

Therefore, in order to understand the difference in the geometries of the bicyclic ring systems of the tested compounds (C4-monoalkene vs 2,3-diene), the crystal structures of 1, 22, and 48 were overlaid (Figure 12). Compounds 1 and 48 both have a C4-monoalkene system, while compound 22 has a 2,4-diene system. Furthermore, since one of the tested compounds (Figure 11, compound 34) bears a 3β-hydroxy substituent, we were interested in knowing the effect that this hydroxy group had on the geometry of the bicyclic ring system. The crystal structures of 1² and 48¹ were published in our previous studies. Based on the structural overlay, the 2,4-diene system clearly had a difference in the geometry when compared to the C4-monoalkene (Figure 12B and D). However, the presence of the C3β-hydroxy group did not affect the geometry of the bicyclic ring system (Figure 12C).

Figure 12.

Structural overlay of compounds 1, 22, and 48. (A) Structures of 1, 22, and 48. (B) Structural overlays of (B) 1 (red) and 22 (blue), (C) 1 (red) and 48 (green), and (D) 22 (blue) and 48 (green). Structural overlays of crystal structures were performed with Chem3D software (20.1.1 PerkinElmer).

Figure 13 shows a summary of the other compounds that were tested with less activity. Many of the tested compounds with an aromatic ring lacked potent activity (Figure 13, compounds 6, 11, 49, 51, and 52; MIC values ≥100 μg/mL). Considering the facts that (i) some of the aromatic ring-containing structures differ only by the saturation pattern in the ring system and (ii) the aromatic ring system is less potent than its saturated ring system counterpart (e.g., compound 52 vs diene 22, ≥100 μg/mL vs 25 μg/mL against C. neoformans, respectively, or compound 51 vs monoalkene 27, 100 μg/mL vs 25 μg/mL against C. neoformans, respectively), we can conclude that the three-dimensional geometry of the bicyclic ring system is important for some of the antimicrobial activities observed (see SI, Part 12, Figures S12–S24). Although the endoperoxide bridge gives artemisinin its antiplasmodial activity,³ the endoperoxide-containing compounds in this study (2 and 4, Figure 13) were less potent than the other derivatives that lack the endoperoxide (e.g., 22, 27, or 33 in Figure 11). One possible explanation of the decreased potency of the endoperoxide-containing compounds could be related to the lack (or low concentration) of heme (Fe²⁺) in the cells relative to Plasmodium falciparum. Heme is required to generate a radical species that leads to the biological activity.^61–63 A reasonable approach to enhance the potency of the endoperoxide-containing compounds could be to add an Fe²⁺ source.⁶⁴

In addition, the C12 substituent also played a role in the different biological activities observed. With the exception of DHAA (Figure 11, compound 1, MIC 25–50 μg/mL against C. neoformans), compounds bearing a C12-carboxylic acid were generally not biologically active (Figure 13, compounds 3, 4, 6, and 53; MIC values ≥100 μg/mL). Similarly, compounds with a completely hydrophobic side chain at C12 did not possess antimicrobial activity (compounds 11 and 49; MIC values >100 μg/mL). Even though compound 28 and compound 30 are structurally similar—both have a monoalkene in the bicyclic system as well as a methyl enol ether in the side chain—their antimicrobial activities are very different. MIC values against JE2 (S. aureus) and H99 (C. neoformans) are 12.5 and 50 μg/mL for compound 28 (Figure 11) but >100 and >100 μg/mL for compound 30 (Figure 13). The extra methylene group at C12 in compound 30 seems to lower the potency of the compound, which could be correlated to the introduction of a more flexible side chain and a less rigid three-dimensional structure.

These observations suggest that compounds are not biologically active when the substituent at C12 is too polar (carboxylic acid) or too nonpolar (alkyl side chain) (see SI, Part 12). However, further investigation is warranted to test this hypothesis.

In conclusion, a 2,3-didehydro-DHAA analogue (3) was synthesized, and this compound was used to test for endoperoxide formation when exposed to air. Instead of endoperoxide formation, this 2,3-didehydro-DHAA derivative underwent aromatization through the desaturation of the Cl–C6 position to form aromatic compound 6. This study suggests that the monoalkene functional group of DHAA is a requirement for the spontaneous endoperoxide-forming reaction to occur to yield artemisinin (Figure 1, 1 to 2). Furthermore, these results have implications in enhancing the understanding of other natural products that contain aromatic rings, which are formed by diene precursors, erogorgiaene²⁴ and leubethanol,²² two natural products possessing antibacterial properties. A group of the compounds synthesized in this study were tested and had antimicrobial activity against S. aureus and C. neoformans with MIC values at 12.5 μg/mL (Figure 11, compounds 27, 28, 32, and 33); these compounds all possessed a C4-monoalkene in the bicyclic ring system, which have a different three-dimensional structure compared to the 2,4-diene (Figure 12) and the aromatic ring systems. Further synthesis of analogues will be explored to identify more potent antimicrobial derivatives.

EXPERIMENTAL SECTION

General Experimental Procedures.

Melting points of compounds were obtained on a Global Medical and Lab Solutions melting point apparatus (India). Optical rotations were measured on a JASCO P-1010 polarimeter (JASCO Inc. Easton, MD, USA). IR spectra were acquired on an FTIR instrument (Nicolet iS50 FT-IR spectrometer, Thermo Fisher Scientific, Waltham, MA, USA). The IR data were analyzed with OMNIC software (Thermo Fisher Scientific). A Bruker 500 MHz Avance III HD NMR spectrometer (Bruker, Billerica, MA, USA) was used to obtain NMR spectra of the synthesized compounds in this study. The NMR data were analyzed with Topspin software (Bruker). Liquid chromatography–high-resolution mass spectrometry (LC-HRMS) data were acquired on an Acuity UPLC system (Waters, Milford, MA, USA) connected to an LTQ Orbitrap XL mass spectrometer (Thermo Fisher Scientific). The UPLC was equipped with a photo diode array (PDA) detector to analyze the UV–vis spectra (wavelength range from 190 to 500 nm) of select compounds shown in the Supporting Information (SI, Part 11). The LCHRMS data were analyzed using Xcalibur 2.1 software (Thermo Scientific). Chem3D 20.1.1 (PerkinElmer, Waltham, MA, USA) was used to perform the structural overlays in Figure 12.

LCMS Time Course (Table 1 and Figure 6).

Diene (3, 5.5 mg) was dissolved in CH₂Cl₂ (4.1 mL) to make solution A. Solution A was divided into 10 amber glass (2 mL) vials and 20 clear glass (2 mL) vials, which had a diameter of 1 cm and a height of 3 cm (50 μL each vial or 67.1 μg of diene 3 for each vial). The solvent in each vial was evaporated with a stream of N₂. The vials were divided into three groups: the 10 amber glass vials were stored in a black plastic box in a cabinet (group 1), 10 of the clear glass vials were placed in a box under a UV–C lamp (group 2), and the other 10 clear glass vials were placed in a box under a UV-AB lamp (group 3). See also SI, Part 10.

An internal standard of d₁-carboxylic acid (24) was made (20 μg/mL in CH₃OH: solution A).

At each time point (0, 24, 48, 550, 1008 h), two vials from each group were taken. Of the two vials, one vial was dissolved in 100 μL of CH₃OH, and one vial was dissolved in 100 μL of solution A. The samples were analyzed by LC-HRMS. Table 1 shows the data for the 24 h time point (see SI, Part 10, for all data).

Conditions for Singlet Oxygen Generation to Aromatize Diene in Scheme 3.

Diene (3, 343 mg, 1.5 mmol) was dissolved in CH₂Cl₂ (43 mL), and the solution was split into two separate 11 dram vials (28 × 108 mm). To vial 1, 1 mL of a solution of methylene blue in CH₂Cl₂ (4 mg/mL) was added, and to vial 2, 1.6 mL of methylene blue in CH₂Cl₂ (2 mg/mL) was added. Vial 1 contained 4 mg of methylene blue (0.013 mmol, 0.02 mol equiv), while vial 2 contained 3.2 mg of methylene blue (0.010 mmol, 0.01 mol equiv). The reactions were cooled to 0 °C, and a flood lamp was used to irradiate the reactions. After 20 min, an aliquot was taken from each reaction and ¹H NMR spectra were obtained. After an additional 50 min, the reaction progress was checked by ¹H NMR, which confirmed complete consumption of the diene starting material. Preliminary experiments involved the use of CDCl₃ and an NMR tube as the reaction vessel; however, it was determined that the small opening of the NMR tube resulted in less exposure to oxygen. Therefore, a reaction vessel with a wider opening was used. See also SI, Part 1.

The two reactions were combined and purified to yield the aromatic compound 6 (93 mg, 0.40 mmol, 27%) and the endoperoxide 4 (17 mg, 0.063 mmol, 4.3%).

Testing of Antimicrobial Activity.

Broth microdilution assays were performed to evaluate the antimicrobial efficacy of the selected compounds following a published protocol.⁶⁵ Briefly, overnight grown cells were diluted in culture media to a final density of around 1 × 10⁵ CFU/mL. A 200 μL amount of the diluted cells was inoculated in 96-well plates and exposed to the selected compounds (stock solutions prepared in DMSO at 10 mg/mL) at concentrations ranging from 0.048 to 100 μg/mL (except compounds 56a and 56b, which were tested at 50 μg/mL maximum due to solubility issues, and novobiocin was tested at 800 μg/mL maximum) and incubated at 37 °C. Both the P. aeruginosa and S. aureus were incubated standing in lysogeny broth for 24 h, whereas C. neoformans was incubated standing in sabouraud dextrose broth for 48 h. Following incubation, cells were exposed to resazurin dye and further incubated (P. aeruginosa and S. aureus, 1 h; C. neoformans, 24 h) to observe the color change. Following the above-mentioned protocol, S. aureus (JE2 strain),⁶⁶ C. neoformans (H99 is a standard strain isolated by Dr. John Perfect, Duke University, from an individual with cryptococcal meningoencephalitis), and P. aeruginosa (PA14 strain)⁶⁷ were used to determine antimicrobial activity of the compounds. See also SI, Part 12.

Crystal Structure of Compound 22.

Single crystals of compound 22, C₁₅H₂₄O, were prepared by slow evaporation of a solution of EtOAc and hexanes (0.1 mL, 1:1, v/v). A suitable colorless block-like crystal, with dimensions of 0.196 mm × 0.173 mm × 0.125 mm, was mounted in Paratone oil onto a nylon loop. All data were collected at 100.0(1) K, using a XtaLAB Synergy/Dualflex, HyPix fitted with Cu Kα radiation (λ = 1.541 84 Å). Data collection and unit cell refinement were performed using CrysAlisPro software.⁶⁸ The total number of data were measured in the 7.96° <2θ < 153.2° range, using ω scans. Data processing and absorption correction, giving minimum and maximum transmission factors (0.296, 1.000), were accomplished with CrysAlisPro⁶⁸ and SCALE3 ABSPACK,⁶⁹ respectively. The structure, using Olex2,⁷⁰ was solved with the ShelXT⁷¹ structure solution program using direct methods and refined (on F²) with the ShelXL⁷² refinement package using fullmatrix, least-squares techniques. All non-hydrogen atoms were refined with anisotropic displacement parameters. All hydrogen atom positions were determined by geometry and refined by a riding model.

Overlaying 1, 22, and 48 in Chem3D.

The crystal structures of DHAA (1)² and diol 48¹ were previously reported, and the .cif files were used. The structure for diene 22 was obtained in this report.

For each structural overlay (Figures 12B, C, and D), the two .cif files were uploaded on Chem3D software, and double bonds in the ChemDraw panel were incorporated where present in the molecule. Two atoms corresponding to each structure were selected, and the Display Distance Measurement feature was activated. At least two more pairs of atoms were selected, and the Display Distance Measurement was shown. Under the Measurement window, the optimal distance was set to 0 Å for each pair of atoms. The Structural Overlay option was selected, and the Minimize Feature was selected.