The First Chemical Synthesis of Limonin: Supply Chain Implications for Future Clinical and Toxicological Testing of Limonin Glucoside

            In a previous series of posts at this site, we highlighted a placebo-controlled human study with purified limonin glucoside (LG) in which significant health benefits were observed. The disclosure of these exciting results by the U.S. Department of Agriculture (USDA) provoked a thought exercise concerning the downstream supply chain issues that this development program will likely face. The USDA team led by Darshan Kelley was able to obtain about 100 grams of high-purity LG through development and execution of an optimized extraction process starting from molasses, a by-product obtained from citrus processing waste streams. However, in order to conduct additional clinical and toxicological testing of this unique limonoid natural product, multiple hundreds of grams (and likely kilograms) of API will be required at various stages. Given the architectural complexity of limonin and related substances derived thereof, does partial or total chemical synthesis provide a viable sourcing alternative? In this post, we will examine the recently disclosed first total synthesis of racemic limonin and compare the productivity of this process with that of the optimized isolation from citrus molasses that was previously developed by the USDA.

            We should begin by noting that the completion of the first total synthesis of limonin is an exceptional accomplishment, given that this limonoid has been known as the bitter principle derived from citrus for over 150 years. Limonin, an oxidatively fragmented 17-furylandrostane, is a highly complex small molecule, on par with other notable limonoid family members including azadirachtin and libiguin A. The detailed molecular structure of limonin remained unknown until 1960, when a determination was finally made using chemical derivatization and X-ray diffraction methods. More than fifty years later, in 2015, we now can begin to discuss the state-of-the-art laboratory methods that enabled the first and only chemical synthesis of limonin. The total synthesis of limonin, reported by the Japanese team led by Shuji Yamashita and Masahiro Hirama, can roughly be broken down into two synthetic “phases,” as outlined in the Figure above. In phase 1, the 4,4,8-trimethylandrostane steroidal framework is constructed from readily available geraniol. The critical synthetic operations that led to the successful execution of phase 1 include a tandem radical polycyclization using Barry Snider’s manganese(III)-mediated chemistry, followed by a standard Robinson annulation to forge the steroid A-ring. This portion of the route provides access to the requisite androstane system containing the appropriate methylation and oxidation patterns for the completion of the target structure. However, 17 synthetic steps are required to produce the desired intermediate for initiation of phase 2 of the route, oxidative fragmentation and elaboration of the eastern D-ring and western A-ring. The identification of an abundant, plant-derived sterol that contains oxygenated functionality at carbon positions 3, 7 and 19, to be used as a starting material for a semisynthetic process, would significantly improve the practicality of the overall approach by reducing the step count. Phase 2 of the Tohoku University process is truly innovative and provides a pioneering blueprint for late-stage synthetic manipulations involving complex limonoid systems.

            The overwhelming lack of synthetic progress reported towards the limonin system prior to Yamashita and co-workers’ 2015 disclosure is striking. In the mid-1970s and early 1980s, a German group at ETH Zürich led by Walter Graf published the synthesis of a model compound (shown above) incorporating the eastern substructure of limonin, albeit lacking the 14,15-epoxy functionality. Almost two decades later, in the late 1990s, the research group of Ernesto Suárez (Spain) reported the synthesis of a similar model system bearing the western limonin system. The photochemical reaction that was developed by Suárez is a fragmentation process wherein irradiation of a hemiketal hydroxyl with visible light in the presence of oxidants promotes peroxidation of an initially formed carbon-centered radical to furnish a peroxy radical. The peroxy radical reacts further with iodine to give an alkoxy radical and iodoxyl radical. The alkoxy radical finally engages a proximally situated, unactivated C-H position in intramolecular functionalization through the intermediacy of a six-membered cyclic transition state to produce the desired limonin tetrahydrofuran. To the best of my knowledge, these are the only published protocols for limonin synthesis that were available to the Tohoku University team at the outset of their studies. The group did indeed make use of the Suárez precedent (see below), providing a fascinating late-stage example of the tandem fragmentation/C-H functionalization process on, without question, the most complex substate reported to date.

            The portion of the inaugural limonin synthesis that is truly worthy of highlighting is the ‘phase 2’ oxidative fragmentation and elaboration of the androstane D-ring, a 13a-cyclopentenone, into the intact eastern limonin substructure. The synthetic technology depicted in the Scheme above will undoubtedly guide and enable future synthetic studies targeting limonoid systems that were previously considered synthetically inaccessible due to excessive chemical complexity. The overall conversion is initiated by installation of a butenolide system via a Stille coupling protocol that has been used previously to synthesize cardenolide natural products. Singlet oxygen [4+2] cycloaddition then simultaneously functionalizes the steroidal C13 and C17 positions in a single operation. As originally reported by Karel Wiesner in the 1980s, the butenolide is a latent synthetic form of a furan ring system and, notably, this reductive conversion could be accomplished without harming the sensitive endoperoxide. Fortunately, upon exposure to a ruthenium(II) catalyst as described by the Nobel laureate Noyori in the late 1980s, the endoperoxide functionality was successfully fragmented to an oxy radical intermediate that underwent further conversion to a bis-epoxide. Then, upon exposure of the intermediary bis-epoxide to mildly acidic silica gel, a suprafacial 1,2-hydride shift ensued, giving rise to an epoxy ketone albeit with the undesired stereochemical configuration at C17. Base-mediated epimerization of the C17 position followed by a Baeyer-Villiger ring-expansion of the five-membered D-ring finally completed the stereocontrolled construction of a fully elaborated limonin eastern substructure, suitably functionalized for late-stage oxidative manipulation of the A-ring.

            As noted above, in the course of their synthetic endgame, Yamashita and Hirama et al directly applied the Suárez protocol, as reported in the 1989 J. Org. Chem. manuscript, without noticeable modification. This remarkably complex example of the Suárez reaction, demonstrated on 4.4-milligram scale, behaved ‘as advertised,’ and the desired oxidative fragmentation product was obtained and processed further in crude form. The final two synthetic operations, involving deprotection and redox manipulation of C7, ultimately furnished 600 micrograms of purified limonin that was identical in all respects to material derived from natural sources. The first total synthesis of (+/-)-limonin proceeded in 35 total steps starting from geraniol. Nearly half of those synthetic operations were required for total synthesis of the steroid framework (i.e. ‘Phase 1’). The oxidative rearrangement of the steroid A- and D-rings, which I have referred to here as ‘Phase 2,’ is indeed groundbreaking and will likely facilitate the synthesis other similarly complex, bioactive limonoid architectures.

            So how does this work impact the supply chain question that was advanced at the top of this post? As I have stated elsewhere, the metric for the degree of success of a synthetic project can be quantified by answering the question: ‘How does this technology compare with extraction from natural sources or microbial production by an optimized fermentation process?’ Academic synthetic work can be both virtuosic in its educational value for students, and, at the same time, completely impractical as judged by industrial standards. The best synthetic work is academically and aesthetically satisfying, but also useful in the lab on a meaningful scale. In the case of limonin, similar to the infamous example provided by azadirachtin, the state-of-the-art in synthetic chemical methodology falls short when compared with isolation of drug substance from natural sources. As noted previously, using the optimized USDA extraction process and estimating conservatively, approximately two liters of citrus molasses can be processed in four to six weeks to produce about 18 grams of high-purity limonin glucoside, suitable for human studies. Moreover, the molasses feedstock is a worthless by-product of citrus processing that juice producers are happy to dispose of. On the other hand, the chemical synthesis of limonin in racemic form, in all likelihood, required several man-years of effort in order to produce less than one milligram. This comparison is, admittedly, harsh and perhaps unfair, given that the two research groups involved in the work obviously had entirely different research goals. However, the comparison is nonetheless instructive and should be taken into consideration before it is asserted that organic chemistry has ‘matured,’ and that any molecule can be synthesized, given the appropriate time and resources. The challenge presented by a chemical synthesis of limonin disputes that fundamentally flawed notion.

Isolation of High-Purity Limonin Glucoside from Molasses, a By-Product Obtained from Citrus Processing Waste Streams

            In the previous post, we highlighted a study from the USDA that examined the health benefits of limonin glucoside (LG) supplementation, as indicated by circulating levels of blood lipids, glucose and biomarkers of chronic inflammatory diseases, in human subjects. As can be seen from the complicated HPLC trace shown above, recovery of pure limonoid glucosides from natural sources in order to fund these types of human studies is not a trivial undertaking. Moreover, given the molecular complexity of most limonoids, chemical synthesis may not necessarily provide a viable supply chain alternative. In my rough estimation, the USDA study required the sourcing of about one hundred grams of limonin glucoside containing very low levels of its corresponding aglycone, limonin. While limonin glucoside is tasteless, limonin tastes extremely bitter and beverage concentrations of >6 mg/liter render citrus juice unacceptable to consumers. In fact, the original isolation of LG by the USDA produced material that was insufficiently pure. Taste tests of the resulting formulated beverages indicated that they were too bitter for human consumption due to limonin aglycone contamination. In this post, we will outline and discuss the optimized citrus extraction process that was eventually developed in order to support the human LG supplementation study.

            Citrus molasses, a thick, viscous liquid that is dark brown to almost black in color, is a major by-product of citrus processing. The fresh pulp obtained after pressing the fruit is mixed with lime (calcium oxide) and pressed to remove moisture. The resulting liquid is then filtered to remove the larger particles, sterilized by heating and concentrated to produce molasses. Citrus molasses is often fed to animals or fermented to produce ethanol, yielding meager profit margins for juice producers. Interestingly, HPLC and LC/MS analyses (see Figure above, left-hand panel) have shown that limonoid glucosides are present in significant amounts in molasses. A preparative scale method for recovery of LGs from citrus molasses was disclosed in 2002 by Thomas Schoch, Gary Manners and Shin Hasegawa at the USDA.

            A schematic flow diagram of the USDA process is reproduced in the Figure above. To begin, one to two liters of crude molasses is diluted with distilled water and the resulting solution is centrifuged. The supernantant thus obtained is then loaded onto a column containing Dowex 50WX4-100 resin and the outflow of the column is directly connected to a second column containing SP70 Sepabeads. Ethanol is used to elute the partially purified mixture from the second column. The Dowex column, a sulfonic acid cation-exchange resin, is intended to remove colored compounds present in molasses. The Sepabeads, which are a divinylbenzene copolymer (basically polystyrene), subsequently absorb and remove hydrophobic components. The eluent solution from the SP70 column is then adjusted to pH 6.5 with dilute aqueous sodium hydroxide in order to maintain consistent chemical reactivity and chromatographic behavior within the partially purified LG mixture.

            The LG solution is then loaded onto a Q-Sepharose column and the column is eluted with an aqueous sodium chloride gradient. The Q-Sepharose stationary phase is a quaternary ammonium resin, designed to bind negatively charged compounds. Limonin glucosides are relatively acidic (pK_a~2.8) and are therefore anionic at pH 6.5. The salt gradient elutes the negatively charged limonoids in order of decreasing pK_a, enabling separation of the LGs from things like flavonoids, which are phenolic and less acidic. The resulting LG-containing fractions are then subjected to a second Dowex/SP70 treatment and the solvent is finally removed by iterative co-evaporation with ethanol to furnish a golden-brown limonoid powder. The method outlined above can produce about ten grams of >60% pure limonoid glucoside feedstock per week. Further re-purification of the powder by recrystallization from warm water yields semi-pure limonin glucoside, contaminated predominantly with unacceptably high levels of the corresponding aglycone, limonin. In order to remove the bitter limonin impurity, an HPLC ‘polishing’ method utilizing food grade solvents (aqueous ethanol) was developed. Using a 75 x 300 mm C18 column, the method is capable of processing 20 grams of material per run in less than three hours. Recovery of the purified limonin glucoside following evaporation of solvent is nearly perfect (~94%) and the limonin content is consistently reduced to <0.10% by weight. Estimating conservatively, it looks like about two liters of molasses could be processed in four to six weeks to produce around 18 grams of high-purity limonin glucoside, suitable for human studies. By industrial standards, this type of productivity would be considered relatively low-throughput, given that a dose of 500 mg/day/patient is required for beverage supplementation studies. But, in the case of limonin or a related API of similar molecular complexity, does partial or total chemical synthesis offer a more scalable alternate supply chain approach to support downstream clinical programs? In a forthcoming post, we will examine the first and only reported chemical synthesis of limonin, disclosed in the summer of 2015 by researchers at Tohoku University in Japan. We will then directly compare the synthetic efficiency achieved by the world’s best organic chemists, using state-of-the-art methods, with that of recovery from citrus processing waste streams, in terms of practicality, cost and overall effort.

On Selective Functionalization of Polyhydroxylated Terpenoids: The First Chemical Synthesis of the Natural Insecticide, Ryanodine

            As I was working to wrap my head around the recent total synthesis of limonin in preparation for a forthcoming post, I was surprised to see that the first total synthesis of the naturally occurring insecticide ryanodine was published by Masayuki Inoue’s group at the University of Tokyo. One of the impressive aspects of this work is that it is an asymmetric chemical synthesis of (+)-ryanodine and not the related polyol, ryanodol. Ryanodine is the ester product derived from regioselective esterification of ryanodol with 1H-pyrrole-2-carboxylic acid. Importantly, ryanodine binds to the ryanodine receptor, an intracellular calcium channel, with high potency whereas ryanodol does not. Ryanodol was synthesized in racemic form by Pierre Deslongchamps’ group in the late 1970s (43 steps from carvone), but the group was not able to reach ryanodine because esterification of ryanodol occurs at an undesired position. Inoue’s synthetic strategy allows selective ‘blocking’ of the hemiketal -OH group along with 5 of the 6 hydroxyls of the ‘ryanoid’ polyol, such that the hindered C3-hydroxyl can be selectively functionalized en route to the target structure. Similar protecting group strategies were required in order to synthesize certain complex polyhydroxylated steroids including ouabain, ecdysone and, more recently, linckoside A. At any rate, the inaugural total synthesis of a natural product in 2015 that has been known since 1948, and which possesses important pharmacological activity, is indeed a great achievement. Ryanodine is clearly not a steroid so I won’t discuss its synthesis in any degree of detail here, but an overview of the successful protecting group strategy and endgame protocol is diagrammed above.