Diverse structural types of natural products and their mimics have served as targets of opportunity in our laboratory to inspire the discovery and development of new methods and strategies to assemble polyfunctional and polycyclic molecular architectures. Furthermore, our efforts toward identifying novel compounds having useful biological properties led to the creation of new targets, many of which posed synthetic challenges that required the invention of new methodology. In this Perspective, selected examples of how we have exploited a diverse range of natural products and their mimics to create, explore, and solve a variety of problems in chemistry and biology will be discussed. The journey was not without its twists and turns, but the unexpected often led to new revelations and insights. Indeed, in our recent excursion into applications of synthetic organic chemistry to neuroscience, avoiding the more-traveled paths was richly rewarding.
In conjunction with receiving the Ernest Guenther Award in Natural Products Chemistry for 2017, I was asked to write a Perspective article summarizing some of our research that led to this award, which is indeed a great honor. However, this award is really a tribute to a team effort that recognizes the outstanding achievements of members of my research group over the years, and I am deeply indebted to all of them; I was merely their conductor. This Perspective is thus a partial account of their accomplishments, and I apologize to those whose stories I was unable to tell.
Natural products have long played a major role in medicine and science. For example, naturally occurring compounds are arguably the single most important source for new drugs to treat human disease.1 Efforts directed toward their synthesis have led to the discovery of new chemistry and reactivity and to the invention and development of new strategies for generating skeletal frameworks and for forming new chemical bonds. The remarkable diversity of natural products offers a virtually limitless playing field for discovery in chemistry, biology, and medicine, and we have explored only a small fraction of that space.
In thinking about presenting how we have used natural products and their mimics as targets of opportunity to address chemical and biological problems, I decided to organize the discussion primarily along thematic lines, but with a chronological suborganization. Accordingly, I will first present some results in the area of oxygenated natural products, which comprises work directed toward the syntheses of macrolide antibiotics and related compounds, C-aryl glycoside natural products, and polycyclic xanthone natural products. One theme of work in this area is the use of substituted furans and pyrans as building blocks. Next, some results in the synthesis of alkaloid natural products, arguably the cornerstone of our efforts over the years, will be discussed. Work in this area features developing methods for the synthesis of quaternary carbon atoms and the use of Diels–Alder, vinylogous Mannich, ring-closing metathesis, and dipolar cycloaddition reactions to construct subunits common to a wide variety of alkaloid natural products.
We have not restricted our attention to compounds that fit the common definition of a natural product as being a secondary metabolite. Rather, we have long believed in a broader definition that regards a natural product as being any compound of natural origin. In that context, we have focused on the design and development of natural product mimics as tools to interrogate biology. We thus became interested in phospholipid analogues and cyclopropane-derived analogues of peptides. At the outset, we were attracted to these natural product mimics as potential enzyme inhibitors, but we pivoted over time to using peptide mimics to probe complex questions of energetics and structure in protein–ligand interactions. Finally, an interesting journey will be discussed that commenced with solving a problem in alkaloid synthesis and ended with the design of a general platform for the synthesis of a diverse array of functionalized heterocyclic scaffolds. This program evolved to the discovery of new compounds that show significant promise as potential therapeutic leads to treat neurodegenerative and neurological conditions.
Synthesis of Oxygenated Natural Products
Furans and Hydropyrans as Templates
In the early days at The University of Texas, we were drawn to the challenges associated with developing short and efficient approaches to the synthesis of oxygenated natural products, including those having functionally and structurally complex frameworks. Some oxygenated natural products that captured our interest included Prelog–Djerassi lactone,2 3-deoxy-d-manno-2-octulosonic acid [(+)-KDO],3 1-deoxycastanospermine,4 tirandamycin A,5 macbecin I,6 herbimycin A,7 and erythromycin B (Figure 1).8,9
Some representative oxygenated natural products.
Toward developing a general approach to solve the stereochemical problems presented by these natural products, it occurred to us that the oxidative transformation of substituted furans 1 by an Achmatowicz or related reaction would provide hydropyrans of the general form 2 (Scheme 1). We reasoned that these hydropyrans 2 would nicely serve as conformationally rigid templates that could be modified by stereoselective reactions to introduce new substituents. These elaborated hydropyrans would then serve as key intermediates in the synthesis of oxygenated natural products, such as those depicted in Figure 1.
One use of a furan as a starting material to prepare oxygenated natural products is exemplified by our synthesis of (+)-KDO, a higher monosaccharide that forms the link between lipid A and the hydrophilic polysaccharide subunits in the outer membrane lipopolysaccharides of Gram-negative bacteria (Scheme 2).10 Notably, analogues of (+)-KDO had been developed as antibacterial agents. The point of departure for the synthesis of (+)-KDO was 3, which was prepared in 53% yield in a one-pot operation from Garner’s aldehyde. The derived intermediate 4 was then oxidatively processed into the hydropyran 5 together with 14% of its anomer. Stereoselective hydride reduction of the enone followed by an iodonium ion-initiated cyclization of a carbamate led to the key intermediate 7. The yield of this cyclization was severely compromised by the fact that considerable amounts of recovered starting material were isolated, and despite extensive efforts, it was not possible to identify conditions that led to high conversion. Nevertheless, the starting material 6 could be recycled, so the overall process was reasonably efficient. Sequential removal of the iodo and benzyl groups led to 8, which was converted by Swern oxidation and hydrolysis of the cyclic carbonate to 9. Removal of the acetonide protecting group furnished (+)-KDO. This short synthesis of (+)-KDO highlights the utility of our approach for the preparation of higher monosaccharides and densely hydroxylated hydropyrans.
Herbimycin A is a representative member of the ansamycin antibiotics that exhibits a broad spectrum of biological activities, including antiangiogenic and antitumor properties. Our synthesis of this novel antibiotic, which is outlined in Schemes 3–5,11 is exemplary of how we typically tried to use natural products as targets of opportunity to discover and develop new chemistry. Namely, we first employed our strategy to use furans and hydropyrans derived therefrom as templates to create the C3–C8 and C9–C15 subunits of the natural product. Our plan then called for coupling these two fragments by stereoselective formation of the C8–C9 double bond (Figure 2). Because methodology for such constructions was severely limited, we knew it would be necessary to develop a new process for the stereoselective synthesis of trisubstituted olefins to address this deficiency. Along the way, we also encountered several other unexpected challenges that required the development of new methods.
Key disconnections in herbimycin A.
The synthesis of the vinyl iodide 15, which comprises C3–C8 of herbimycin A, commenced with the addition of 2-lithiofuran to the protected aldehyde 10 to give 11 (Scheme 3). Processing of 11 led to the bicyclic intermediate 12, which is conformationally biased to enable highly stereoselective reduction of the keto group leading to 13. The bridged bicyclic framework had thus served its purpose as a rigid template to enable stereochemical control at C6, so 13 was converted in three operations to the monocyclic hydropyran 15.
The C9–C15 subunit was then prepared from furfuraldehyde (16) via an Evans aldol reaction to deliver 17, which was oxidatively transformed into the hydropyranone 18 (Scheme 4). The stereoselective introduction of the requisite methyl group at C14 was achieved by a cuprate addition providing 19. Because of the axial orientation of the methyl group at C14 of 19, it was not possible to stereoselectively reduce the ketone moiety to give the requisite equatorial alcohol at C12. However, treatment of 19 with sodium borohydride gave an intermediate lactone that was reduced to give a diol, the primary alcohol of which was selectively protected to give 20. Although a seemingly straightforward transformation, inversion of the alcohol at C12 of 20 was unexpectedly problematic. We ultimately discovered that a modified Mitsunobu reaction in which p-nitrobenzoic acid was used as the nucleophile proceeded smoothly and in excellent yield to invert the stereochemistry at C12 leading to 21,12 which was transformed in two straightforward steps to the aldehyde 22. This new method for effecting the inversion of secondary alcohols is widely applicable, especially for sterically encumbered alcohol substrates, and it has been broadly utilized by others.
As mentioned previously, our plan to assemble the C3–C15 subunit of herbimycin A entailed joining these two subunits stereoselectively to form an E-trisubstituted olefin. Although the Julia olefination and other reactions work well in conjunctive processes to form E-disubstituted olefins, they tend to proceed in low yields and diastereoselectivities when applied to the construction of trisubstituted alkenes. Knowing in advance that the methodology for such transformations was severely limited, we developed a general and effective procedure for preparing trisubstituted olefins that is illustrated by the conversion of 22 and 15 into 24 via the allylic alcohol 23; no detectible amounts of the Z-olefin were observed (Scheme 5).13 The key intermediate 24 was converted into 26 by selective unmasking of the C15 aldehyde in 24 followed by a reaction with the aryllithium 25. In related work in the ansa antibiotic arena, the aniline moiety of compounds related to 25 had been protected as a 2,5-dimethylpyrrole. Because the procedure to convert such N-arylpyrroles to anilines requires somewhat vigorous conditions, we developed an alternative way of diprotecting anilines as their 1-aryl-2,5-bis-triisopropylsilyloxypyrrole derivatives, which are easier to remove.14 The intermediate 26 is closely related to a compound Tatsuta previously converted into herbimycin A.15 At this juncture, it made little sense to simply repeat these reactions, so we terminated our efforts with the synthesis of 26.
From the very outset of our work in the area of oxygenated natural products, we were attracted to the considerable challenges associated with the total synthesis of the erythromycin antibiotics. These important antibiotics owe their activity to their ability to inhibit ribosomal-dependent protein biosynthesis by binding to the 50S ribosomal subunit;16 erythromycin A remains in clinical use for treating bacterial infections. In addition to using a furan as a key building block for the stereoselective creation of a subsection of the macrocyclic backbone, we were also interested in developing an “abiotic” approach to these compounds. Namely, all of the contemporaneous approaches to macrolide antibiotics were patterned after their biosynthesis, which involves formation of a macrolide ring, followed by the introduction of the requisite carbohydrate groups onto pendant hydroxyl groups by glycosyl transfer enzymes. Indeed, in the landmark synthesis of erythromycin A in 1981, the Woodward group employed this strategy.17 During the course of these efforts, the Woodward group explored in some detail the conformational features of the seco acid backbone that were required for cyclization. We took the opposite approach. Namely, we queried whether a glycosylated seco acid derivative, such as 27, might undergo macrolactonization to give an intermediate that could be transformed after appropriate refunctionalizations into erythromycin B (Scheme 6). If such a strategy were feasible, the cladinose residues at C3 and a desosamine group at C5 of 27 might then nicely serve as surrogates of hydroxyl protecting groups, thereby resulting in a shorter synthesis because traditional protection of these hydroxyl groups would be unnecessary.
In order to establish the underlying feasibility of this rather speculative plan, we set to the task of establishing proof-of-principle for the macrocyclization in a critical model study.18 Accordingly, erythromycin B was degraded into the seco acid derivative 28, drawing largely on considerable literature precedent in the field. Gratifyingly, we discovered that cyclization of 28 using the Yamaguchi cyclization conditions afforded 29 (Scheme 7);19 other standard tactics to induce macrocyclizations were less effective.
As a prelude to the eventual implementation of this strategy, we engaged in a variety of support studies that revealed important information about a number of steps. One of these efforts led to the successful synthesis of seco acid derivatives of erythromycin A and erythromycin B.20 In our first attempt to prepare erythromycin B according to the plan in Scheme 6, we prepared a protected seco acid derivative having an intact carbon backbone, but all attempts to introduce a cladinose residue onto this framework were unsuccessful.8 Given the late-stage nature of this failure, we opted at the time for an expedient solution that led to the first total synthesis of erythromycin B via a classical approach in which the two carbohydrate residues were introduced onto a preformed macrocyclic lactone. This synthesis of erythromycin B required only 30 steps in the longest linear sequence. As a fringe benefit, we also finished a 23-step synthesis of 9(S)-dihydroerythronolide B, one of the shortest to date.
Although we were not able to implement our original plan as set forth in Scheme 6, we developed an alternative embodiment of that approach that was successful. The synthesis commenced with an Evans aldol reaction of 5-ethylfurfuraldehyde (30) leading to 31 (Scheme 8).9 Oxidative transformation of the diol 31 led to the conformationally biased bicycle 32. Stereoselective introduction of the two methyl groups at C8 and C6 was easily achieved by additions of Me2CuLi and then MeLi under Luche conditions to give 33 with high stereochemical efficiency. Having served its key role as a stereochemically biased template, the bicyclic ketal 33 was unraveled to give the thioketal 34, albeit with some unavoidable erosion of the stereochemistry at C8.
Refunctionalization of 34 led to 35, and this ketone underwent a highly diastereoselective aldol reaction to give 36 (Scheme 9). Stereoselective hydride reduction of the C9 ketone followed by acetal formation and removal of the cyclic carbonate protecting group led to 38. Introduction of a protected desosamine residue at C5 using the glycosyl donor 39 proceeded with poor regioselectivity giving substantial quantities of the C6 glycosylated derivative 41 in addition to the desired 40. When we tried to obviate glycosylation at C6 by protection of the C6 tertiary hydroxyl group, glycosylation of the C5 hydroxyl group was unsuccessful; we were thus obliged to accept this result. It is noteworthy that the introduction of a desosamine residue at C5 in macrocyclic intermediates in our first synthesis of erythromycin B was not accompanied by significant amounts of glycosylation at C6.8
With 40 in hand, completion of the erythronolide backbone remained. Refunctionalization of 40 led to 42, setting the stage to introduce the remaining three carbon atoms via stereoselective crotyl stannylation, followed by oxidative cleavage of the terminal olefin to give 43 (Scheme 10). Initial attempts to introduce the cladinose residue at the C3 hydroxyl group in 43, or its immediate olefin precursor, were unsuccessful. Because material was in short supply, we opted to see whether we might be able to induce macrolactonization on a substrate lacking the cladinose moiety. Toward this goal, selective removal of the C13 hydroxyl protecting group and cyclization of the intermediate hydroxy acid according to the Yamaguchi protocol led to lactone 44.19 Reaction of 44 with the glycosyl donor 45 furnished the bisglycosylated lactone 46. This synthesis of erythromycin B was completed in four steps that were developed during our first synthesis8 and involved global deprotection and selective oxidation of the hydroxyl group at C9. This route to erythromycin B, which required only 27 steps in the longest linear sequence, is three steps shorter than our first synthesis wherein the carbohydrate residues were introduced after macrolide formation. Moreover, this synthesis represents the first time any macrolide antibiotic had been prepared by an “abiotic” approach in which a sugar residue was appended as a surrogate hydroxyl protecting group prior to the macrolactonization step.
Furans as Building Blocks for C-Aryl Glycoside Synthesis
Another area of inquiry arose because we recognized the significant challenges associated with synthesizing natural products of the C-aryl glycoside class.21 This important family of compounds has long attracted interest because of their broad range of biological activities and their resistance to enzymatic hydrolysis. Some C-aryl glycosides that were of interest included galtamycinone22 and vineomycinone B2 methyl ester,23 two representative members of the Group II C-aryl glycosides, 5-hydroxyaloin A,24 a Group I C-aryl glycoside, as well as kidamycin and its isomer isokidamycin,25 which belong to the Group III C-aryl glycoside family (Figure 3).
Representative C-aryl glycoside antibiotics.
In keeping with our broad synthetic objectives, a central goal was to develop a unified approach to the major classes of C-aryl glycosides that was concise and general, so it could be broadly applied to the syntheses of any member of this family of natural products. The strategy that thus evolved is illustrated in Schemes 11–13. The essence of the approach features the cycloaddition of furfuryl glycosides with substituted benzynes to give cycloadducts that undergo acid-catalyzed rearrangement to provide C-aryl glycosides.26 For example, furfuryl glycosides such as 48 are readily available from furan (47). The cycloaddition of 48 with benzyne 49, which was generated in situ by deprotonation/elimination of 2-chloro-1,4-dimethoxybenzene, yields 50, analogues of which were known to undergo facile acid-catalyzed reorganization to provide 1-naphthol derivatives.26 In the present case, this rearrangement would deliver 51, a Group I C-aryl glycoside, or 52, a Group II C-aryl glycoside, depending upon the orientation of the sugar residue on 50 (Scheme 11).
Similarly, the bis-glycosylated furan 53 can be envisioned as a precursor of 55, a representative Group III C-aryl glycoside (Scheme 12) via cycloaddition with 49 and subsequent rearrangement of the intermediate 54. We also envisioned the possibility of inducing ring opening reactions of furan-benzyne cycloadducts such as 56 leading to compounds such as 52, thereby offering an alternative route to Group II C-aryl glycosides (Scheme 13).
Inspection of the sequences shown in Schemes 11–13 highlights an important feature of this approach. Namely, the introduction of the C-aryl glycoside moiety is performed in tandem with the annelation of a new aromatic ring. Because the formation of the C-aryl glycoside is coupled with an increase in skeletal complexity, we envisioned that this new entry to C-aryl glycosides would lead to more concise syntheses of these natural products.
With an overall strategy in mind, it remained to demonstrate proof-of-principle. The ease with which furfuryl glycosides can be assembled is nicely illustrated by the conversion of the readily available 57 into 58 (Scheme 14).27 Cycloaddition of 58 with benzyne 49 provided 59, which underwent acid-catalyzed rearrangement to generate the exemplary Group I C-aryl glycoside 60.
Similarly, the furfuryl glycoside 62 can be readily assembled from 61 by sequential addition of 3-furyllithium, followed by stereoselective hydride reduction (Scheme 15). Transformation of 62 via cycloaddition with benzyne 49 and acid-catalyzed rearrangement of the intermediate cycloadduct 63 delivered the model Group II C-aryl glycoside 64.
Examination of the processes of Schemes 14 and 15 reveals a limitation of this approach to C-aryl glycosides. In particular, the benzyne intermediate 49 in each of these examples is symmetrical. However, a mere glance at the natural products in Figure 3 reveals that the benzyne precursors that would be required to synthesize any of these C-aryl glycosides must necessarily be unsymmetrical. Although unsymmetrical benzynes are known to undergo regioselective cycloadditions with unsymmetrical furans,28 the process is not general.29 Hence, the question arose: How can one control the regiochemistry in the cycloaddition of furfuryl glycosides with unsymmetrical benzynes? We briefly examined the possibility of placing a bulky protecting group on one of the oxygen atoms of the hydroquinone precursor of the benzyne; however, it became quickly apparent that the regiochemical course of cycloadditions of unsymmetrical benzynes could not be controlled through simple steric effects. Accordingly, it was necessary to develop an alternate strategy to solve this regioselectivity issue.
Toward that end, we developed a tethering strategy in which the benzyne precursor and the furan are linked together with a removable tether prior to generation of the benzyne.30 For example, regioselective deprotonation of the furan ring of 58, followed by appending a functionalized silyl group led to 65 (Scheme 16). Coupling 65 with the phenol 66 via a Mitsunobu reaction provided 67. Generation of a benzyne from 67 was easily accomplished using tert-BuLi, leading to the cycloadduct 68. Removal of the temporary silyl tether and sequential protection and acid-catalyzed rearrangement furnished the model Group I C-aryl glycoside 70.
An application of a palladium-catalyzed ring-opening reaction of benzyne–furan cycloadducts to generate C-aryl glycosides (see Scheme 13) is exemplified for the formation of the model Group III C-aryl glycoside 73 (Scheme 17).27 In the event, ring opening of 59 with the iodo glycal 71 proceeded with high regioselectivity, presumably directed by steric effects, to give 72, and catalytic hydrogenation of the enol ether moiety in 72 yielded 73 as a single diastereomer.
We prepared galtamycinone22 and a precursor of 5-hydroxyaloin A24 using this benzyne–glycosyl furan cycloaddition strategy, but the syntheses of vineomycinone B2 methyl ester23 and isokidamycin25 best illustrate the scope and potential of this novel approach to C-aryl glycoside natural products. Vineomycinone B2 methyl ester is a degradation product of vineomycin B2, which was isolated from a culture of a Streptomyces and found to exhibit anticancer and antibiotic activity.31 Examination of its structure (Figure 4), reveals that the central quinone ring is flanked by two aromatic rings, bearing on one side a carbohydrate residue and a hydroxy ester group on the other. It occurred to us that we might be able to simultaneously annelate both of the benzene rings in vineomycinone B2 methyl ester concomitant with introducing the appropriate side chains using a variation of the general plan depicted in Scheme 11.
Furan-derived subunits of vineomycinone B2 methyl ester.
As the first step toward implementing this plan, the furfuryl glycoside 75 was prepared from the readily available lactone 74, which was an intermediate in our synthesis of galtamycinone (Scheme 18).22 Introduction of the silyl tethering group onto the furan ring of 75 was easily achieved by regioselective metalation, followed by appending a functionalized silyl side chain to furnish 77.
The synthesis of the furan 81 commenced with the protection and Sharpless dihydroxylation of the commercially available homoallylic alcohol 78 (Scheme 19). Conversion of 79 into 80 featured a step involving epoxide formation and ring opening. Although it was not possible to regioselectively metalate 80, it did undergo regioselective bromination, and subsequent metal halogen-exchange and introduction of the silyl tethering moiety provided 81.
Furan intermediates 81 and 77 were then sequentially attached to the tetrabromohydroquinone 82 leading to 84 (Scheme 20). In the key step of the synthesis, 84 was treated with excess n-BuLi to deliver 85 in a single operation that involved two intramolecular benzyne–furan cycloadditions. Although 85 was produced as a complex mixture of stereoisomers, removal of the tethering groups and ring opening of the two bicycloheptadiene rings furnished 86. It then remained to remove the protecting groups and oxidize the primary alcohol to deliver vineomycinone B2 methyl ester. This synthesis highlights the ease with which complex skeletal systems can be rapidly assembled using our benzyne–furan cycloaddition methodology to produce a C-aryl glycoside moiety in tandem with forming a new aromatic ring.
We then turned to the significantly greater challenge of the synthesis of kidamycin, which is a member of the pluramycin class of C-aryl glycoside antibiotics and displays a broad range of antibacterial, antifungal, and anticancer activities.32 Given its planar tetracyclic core structure, it is not surprising that kidamycin, like other pluramycin antibiotics, binds to DNA leading to single strand cleavage.33
Because of the unsymmetrical nature of kidamycin, it was again necessary to employ a tethering approach to control regiochemistry in the benzyne–furan cycloaddition (Scheme 21).25 Toward this end, the first stage of the synthesis involved making the glycosyl furan 88 from the known carbohydrate derivative 87, which was prepared from d-rhamnal according to the protocol reported by Brimble.34 Refunctionalization of 88 led to 89, and introduction of the silyl tethering group to give 90 followed previous work (see Scheme 18).
The highly substituted naphthalene 91 was then coupled with 90 via a Mitsunobu reaction to generate 92, which underwent an intramolecular benzyne–furan cycloaddition upon treatment with n-BuLi to provide 93 (Scheme 22). Processing of 93 via cleavage of the silyl tether, acid-catalyzed ring opening of the oxabicycle, and a series of refunctionalizations then led to 96. We had originally intended to convert 96 into the enynone 98 by a carbonylative cross coupling reaction we had developed in ancillary work and had already successfully applied to the synthesis of luteolin.35 Unfortunately, this method could not be applied to the conversion 96 → 98, and we were forced to adopt a stepwise strategy. In the event, reaction of the anion generated from 96 by metal–halogen exchange with the ynal 97 followed by oxidation of the intermediate alcohol gave 98.
We then turned our attention to forming the pyranone ring. We had originally anticipated that removal of the MOM protecting group followed by cyclization via a 6-endo-digonal pathway would lead to 99; however, our optimism would not be rewarded. We discovered that treating 98 with Lewis and Brønsted acids led to the formation of a five-membered ring benzofuranone via a 5-exo-digonal cyclization, not the desired hydropyranone 99. Although similar cyclizations were also known to give benzofuranones,36 there was precedent suggesting that this undesired pathway might be avoided by first converting the ynone into a vinylogous amide.37 Fortunately, we discovered that the vinylogous diethylamide derived from 98 did undergo acid-catalyzed cyclization to form the hydropyranone ring in 99.
At this juncture, it was necessary to introduce the remaining carbohydrate residue, and we briefly considered the possibility of a glycal-induced ring opening reaction of 94 in analogy with chemistry developed in Scheme 17. However, we quickly realized such a tactic would not be applicable to the task at hand because this procedure would likely have led to the β-anomeric C-aryl glycoside, not the α-anomer as required. Accordingly, we planned to introduce the remaining carbohydrate residue via an O → C glycosyl transfer process that had been developed by Suzuki and that we believed would furnish the correct α-anomer via a kinetically controlled rearrangement.38 Reaction of 99 with 100, which was prepared in four steps (54% yield) from l-vancomycin, in the presence of Sc(OTf)3 furnished a single diastereomeric product (Scheme 23). Our initial excitement that the O → C glycoside rearrangement had occurred quickly evaporated, however, when we discovered that the product was the β-anomer 101, not the desired α-anomer 102. Although it is possible that 102 was kinetically formed and underwent rapid epimerization to 101 under the reaction conditions, we never observed any trace of 102 in the mixture. A detailed discussion may be found in our original paper,25 but suffice it to say we now believe that the intermediate oxonium ion formed during the rearrangement likely exists preferentially in a twist boat conformation, not a half chair conformation as originally predicted. If the O → C glycosyl transfer process occurs via a twist-boat transition state, the observed β-anomer would be expected. Although it is possible that an alternate protecting group strategy for the vancosamine residue might favor formation of the desired α-anomer, we did not perform any experiments to address this question.
Having been given a lemon, we resorted to making lemonade, and 101 was transformed in six steps, largely involving refunctionalization operations, into isokidamycin, the structure of which was verified by comparison of its spectra with those of an authentic sample. This achievement represents the first total synthesis of a bis-C-aryl glycoside natural product in the pluramycin family.
Furans as π-Nucleophiles in Vinylogous Aldol Reactions
Over the years, we have explored the use of furan intermediates in a number of applications, some of which have been presented previously. We have also been interested in the use of furans as π-nucleophiles, especially in vinylogous Mannich reactions, which will be discussed in more detail later. However, in planning an approach to 6,7-dideoxysqualistatin H5,39 we had occasion to use a furan in an intramolecular vinylogous aldol reaction.40 6,7-Dideoxysqualistatin H5 is a natural product related to the zaragozic acids and squalistatins, which had attracted considerable attention because of their activity as squalene synthase inhibitors.41
In order to set the stage for the pivotal vinylogous aldol reaction in our synthesis of 6,7-dideoxysqualistatin H5, 103, which was prepared in three steps from dimethyl d-tartrate, was esterified with the known acid 104, and removal of the tetrahydropyranyl protecting group followed by oxidation gave 105 (Scheme 24).42 The key cyclization of 105 proceeded efficiently using TiCl4 as the catalyst to give 106 together with less than 5% of other diastereomers. The spirocyclic lactone 106 was then converted into 107, thereby setting the stage for coupling with the side chain subunit 108 to produce 109. Treatment of 109 with methanolic acid then furnished a separable mixture of 110 and 111. The undesired ketals 111 could be re-equilibrated to give additional quantities of 110. The conversion of 110 into 6,7-dideoxysqualistatin H5 was then simply achieved by oxidation of the primary alcohol function in 110, followed by global saponification of the methyl ester groups.
General Route to Polycyclic Xanthone Natural Products
A more recent foray into the arena of oxygenated natural products was directed toward developing a general approach to polycyclic xanthone natural products. Such compounds range in structural complexity from the tricyclic tetramethoxyxanthone43 to the polycyclic representatives IB-00208,44 citreamicin η,45 and kibdelone C (Figure 5, xanthone rings highlighted in blue). Many compounds of this class exhibit potent antibacterial and anticancer activity. As was often the case in our group, we were drawn by both the structural features and the biological activities of members of this class of natural products.
Representative polycyclic xanthone natural products.
Toward developing a general entry to polycyclic xanthone natural products, we were intrigued by the possibility of exploiting a novel variant of the well-known Moore rearrangement.46 In particular, we envisioned that the Moore reaction of 112 would provide an intermediate quinone 113 that would undergo cyclization to generate the xanthone 114 (Scheme 25).43 We reasoned that the requisite disubstituted acetylene 112 might be rapidly assembled by an “acetylide stitching” process in which a squaric acid derivative and either an aldehyde or activated carboxylic acid derivative would serve as electrophilic partners in reactions with acetylide anions.
We first performed a series of simple model experiments to validate the underlying feasibility of this new approach to substituted xanthones, and we then applied it to a facile synthesis of tetramethoxyxanthone (Scheme 26).43 The key intermediate 118 was readily assembled by the acetylide stitching process using the aldehyde 115 and dimethyl squarate (117). Rearrangement of 118, followed by Jones oxidation of the secondary alcohol provided 119, which underwent facile cyclization upon acid-catalyzed removal of the PMB protecting group; regioselective methylation of the less hindered phenolic hydroxyl group then furnished tetramethoxyxanthone.
Although the synthesis of tetramethoxyxanthone was sufficient to establish proof-of-principle, it remained to demonstrate the applicability of the approach in a more complex setting. We thus set to the challenging task of synthesizing IB-00208, which displays strong antibiotic activity against Gram-positive bacteria and potent anticancer activity against several cancer cell lines.47 In order to set the stage for the pivotal Moore rearrangement, it was first necessary to prepare 128 (Scheme 27). We envisioned that 128 could be prepared by joining the squaric acid derivative 126 with the aldehyde 127 via an acetylide stitching process similar to that shown in Schemes 25 and 26. However, methods for preparing angularly fused benzocyclobutenones such as 126 were not available, thus demanding the development of new methodology that would feature a new application of ring-closing metathesis (e.g., 125 → 126).
The synthesis of the benzocyclobutenone 126 commenced with the transformation of 120 into 121, wherein the MOM group not only serves as a protecting group but also the eventual source of the C1 carbon atom in the hydropyran ring (Scheme 27).44 Because both enantiomers of propylene oxide are commercially available, we would be able to introduce the correct stereochemistry at C3 once the absolute configuration in IB-00208 was known. After removal of the two O-methyl groups from 121, acid-catalyzed cyclization furnished 122, which was converted via a Duff formylation, followed by protection and Wittig olefination to give 123. Union of 123 and 124 led to 125, which underwent ring-closing metathesis in the presence of Grubbs II catalyst to provide 126. This approach to 126 appears to be general and has been applied to the synthesis of other angularly fused benzocyclobutenones.44b
With 126 in hand, acetylide stitching with 127 followed by cleavage of the dimethyl acetal delivered 128, which underwent Moore rearrangement upon heating to give the tetracyclic intermediate 129. Following oxidation of 129 to introduce a methoxy group at C1, removal of the TBS protecting group and cyclization furnished the spirocyclic product 130 rather than the desired fused ring system. This mode of ring closure did not occasion surprise because we had previously observed that 2,6-disubstituted benzoyl substrates similar to 129 underwent kinetically controlled cyclizations to give spirocyclic products that could be isomerized to the desired fused ring systems.43 Processing 130 via oxidation at C1 and thermal rearrangement then provided 131. Removal of the MOM-protecting group gave an inseparable mixture of 132, the aglycone of IB-00208, and its tautomer 133. In retrospect, the formation of 132 and 133 is perhaps not unexpected because the redox potentials of the two compounds might be predicted to be similar. Because all efforts to separate these compounds failed and because they underwent facile interconversion as well as transformation by unknown decomposition pathways, we were unable to characterize these compounds individually nor were we able to convert 132 into IB-00208. Nevertheless, our general strategy for the synthesis of polycyclic xanthone natural products generally worked as planned, and we have applied a similar approach to the synthesis of a pentacyclic precursor of citreamicin η.45
Synthesis of Quaternary Carbon Atoms
Another significant challenge that captured our attention in the early days at The University of Texas involved the formation of quaternary carbon atoms, a structural motif that occurs widely in a diverse array of natural products of biological interest. The methodology for forming fully substituted carbon atoms was rather limited at the time,48 so we sought to discover general strategies for assembling quaternary carbon centers. Part of the motivation to invent and develop such procedures arose from our specific interest in spirocyclic sesquiterpenes such as acorone,49 as well as in several Amaryllidaceae and related alkaloids that included O-methyljoubertiamine,50,51 mesembrine,51 lycoramine,52 crinine,53 and pretazettine54 (Figure 6).
Selected natural products having quaternary carbon atoms.
In one appealing approach to the construction of quaternary carbon atoms, we envisioned replacing both of the carbon–oxygen bonds of a ketone 134 with carbon–carbon bonds leading to 135 (Scheme 28). In this novel process for geminal acylation–alkylation, the first step corresponds to a carbonyl homologation reaction,55 whereas the second step involves the introduction of an electrophile at the carbon-atom α to the newly added aldehyde group. It occurred to us that the reaction of 134 with the phosphonate anion 136 would generate an enamine 137 that could be elaborated by reaction with an electrophile, followed by an aqueous acid workup to produce 135.56 This procedure for effecting the geminal acylation–alkylation of ketone was successfully applied to a synthesis of O-methyljoubertiamine,50,51 but not unexpectedly, the limitations of this approach for the synthesis of more complex alkaloids became quickly apparent because enamines react with only a limited number of reactive electrophiles.
Citation: Arbesman S, Laughlin G (2010) A Scientometric Prediction of the Discovery of the First Potentially Habitable Planet with a Mass Similar to Earth. PLoS ONE 5(10): e13061. https://doi.org/10.1371/journal.pone.0013061
Editor: Enrico Scalas, University of East Piedmont, Italy
Received: July 16, 2010; Accepted: September 1, 2010; Published: October 4, 2010
Copyright: © 2010 Arbesman, Laughlin. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: The authors have no support or funding to report.
Competing interests: The authors have declared that no competing interests exist.
The search for a habitable extrasolar planet has long interested scientists, but only recently have the observational tools become available to search for such planets . Beginning in 1995 with the discovery of an extrasolar planet around Pegasi 51, a star much like our own , the number of confirmed extrasolar planets has expanded into the hundreds. We now have a panoply of physical and orbital data on planets outside the solar system, and with it an increase in understanding of the formation of planetary systems. However, the holy grail of extrasolar planetary research – an Earth-like planet – has yet to be discovered. This search has more recently become even more intense with such large-scale surveys as NASA's Kepler mission .
While many, astronomers included, have speculated about when the first habitable planet might be discovered , no quantitative scientometric analysis has been performed. In order to do this, we develop a metric of habitability for all discovered planets, and use this to arrive at a prediction for when the first habitable planet is expected to be discovered.
Of course, predicting future scientific and technological progress is a slippery and difficult process. While there have been many such successes, such as Moore's Law , history is littered with predictions that are far off the mark .
Here too there are many difficulties. Estimating the habitability of planets is itself a complicated process with many parameters , but most research dwells on the combination of two properties of a planet: surface temperature necessary for liquid water, and planetary mass . Using these guidelines, we constructed a simple habitability metric. Using the habitability time series of previously discovered exoplanets, we created a bootstrap method to predict when the first Earth-like planet would be discovered. We predict the announcement of its discovery with a high probability by mid-year 2011.
Materials and Methods
Using the calculated mass and temperature of a planet, we constructed a simple habitability metric, , for a given planet , where is uninhabitable, and is an Earth-like planet. is defined as the following:(1)where is the product of three sub-measures, each themselves on a scale of 0 to 1:(2)(3)(4)
The formulas for the submeasures that makes up are below, where is one Earth mass, is the approximate width of half the acceptable temperature range (here we used ), is the midpoint of the range (), (for , and can either be (semi-major axis) or (semi-minor axis) of planet ):(5)(6)
Each sub-measure is simply the product of two opposing modified logistic curves with some rescaling, which yield functions of the forms seen in Fig. 1. Note that while there is a positive probability assigned to negative masses, this is simply a byproduct of the symmetric nature of the function, and does not affect the calculations, as there are no such planets with negative mass. For to be 1, all sub-measures must themselves be 1. Due to the “step-function-like” nature of these sub-measures, is likely to be either very low near 0 or very close to 1. Additional separate conditions (such as presence or absence of atmosphere) would make the model more restrictive and necessarily lower , so our model is erring optimistically on habitability, is an upper bound on .
Note that while a nominal temperature, exceeds the boiling point of water, this value is representative only of the simple blackbody equilibrium temperature at the substellar point. Actual surface temperatures for potentially habitable planets will be controlled by a host of effects, some well-understood, some entirely speculative: The stellar flux is intercepted by an area , but must warm a planet of surface area . Our assigned limits on potential habitability represent correspond to a habitable zone whose outer radius is a factor of 2.6 times larger than its inner radius. Given the uncertainties on what constitutes the range of potentially habitable environments, this ratio is intended to be somewhat optimistic. Estimates by Mischna et al. (2000) , for example, advocate a ratio of 2.1, whereas the influential study of Kasting et al. (1993)  found a ratio of bounding distances equal to 1.76. The atmosphere may provide a significant greenhouse warming effect. The planet may have endogenous sources of energy, and cloud cover can provide a significant reflective albedo. For a recent detailed discussion of the geophysical and atmospheric factors relevant to potentially habitable planets, see, e.g. Lammer et al. (2010) .
Calculations of planetary surface temperature involve assumptions of stars on the Main Sequence and black body radiation models, and were as follows :(7)
Using these metrics, was calculated from readily available data  for all 370 planets in the dataset.
Discovery Date Prediction
In order to create a robust estimate of the date of discovery of the first Earth-like planet, the following factors were considered:
- The extrasolar planets considered were detected by two methods: radial velocity (RV) and transit. The radial velocity method detects a planet using the Doppler effect to determine the motion of the planet's star, and the transit method detects a planet using changes in brightness of the star due to the planet's transit in front of it. While the transit method provides accurate estimates of the mass, of a planet, the radial velocity method yields an estimate for , where is an unknown inclination for the planetary system, thereby only giving a lower bound for .
- Any estimate of the detection of the first Earth-like planet will necessarily be dependent on the vagaries of those planets which were previously discovered. If other stars were examined, a different set of of planets, and therefore values, would have been found.
A bootstrap analysis was conducted, which accounts for the estimates for planets discovered by the radial velocity method, and provides a robust estimate of the date of discovery. Each realization consisted first of calculating for each RV-detected planet in the dataset by drawing an inclination randomly chosen from the surface of the sphere. We then sampled 370 planets, each with its year of discovery, from the complete planetary data set with replacement, in order to create a bootstrapped time series of values of exoplanet discovery. The date of discovery chosen for each extrasolar planet was the mid-point of the year of discovery (including for 2010), due to the lack of more precise data.
To predict when the first habitable planet () will be discovered, we examined the upper envelope of each realization's values by year (the points described by the highest habitability metric for each year). Since the upper and lower bound of are known to be and , respectively, fitting a logistic equation is appropriate, and is similar to many other discovery curves, such as the number of mammalian species or number of chemical elements , . A logistic best fit of of the upper envelope of over time, , where the parameters and were allowed to vary, is as follows:(8)
The logistic curves of best-fit were performed using non-linear least squares fits on the sampled values. Due to the step-like function of , the best fit curves were extremely sensitive to initial conditions. While parameters close to the final fits were chosen for precision, variation in the results can be introduced by choosing different initial conditions. To determine the date as a fraction of the year, the value at which the logistic function first reached was calculated. To test the robustness of using this assumption, was used, which predicted early April 2011, and which predicted early June 2011.
Other robustness checks were performed as well. For either a bounds of or for the habitability equation, a prediction of discovery in early May 2010 was found. It is likely then that the precision of the fit, along with the data available, are what drive the prediction. Assuming an error of a month in either direction is therefore reasonable. In order to convert this to days and months of a year, non-leap years were assumed.
Results and Discussion
We examined 370 exoplanets, all of which have well-characterized properties. Doing so, yields for the majority of exoplanets. Notably, Gliese 581 d, thought to be in the habitable zone, yields the highest value, with about , and is still quite low. While some authors (e.g. Wordsworth et al. 2010 ) actually argue that Gliese 581 d is potentially habitable, we are of the opinion that its measured leads to an expected mass close to 10 Earth masses, and a possibly water-dominated composition more akin to an ice giant planet such as Uranus or Neptune than to a terrestrial planet like the Earth.
We conducted 10,000 realizations of the bootstrap method, where the data could successfully be fit to a curve, in order to arrive at a distribution of dates of discovery of the first habitable planet. This distribution is heavy-tailed, as seen in Fig. 2, with a median date of discovery of early May 2011 (2011.34). Additionally, detecting an Earth-like planet by the end of 2013 has a probability, we reach 75% in 2020, and don't achieve 95% likelihood until 2264.
An example realization is shown in Fig. 3, where the best-fit logistic curve arrives at (the upper border), in the first half of the year 2011. More precisely, it reaches at about one-third () through the year, which is early May 2011.
Figure 3. A single realization of the habitability of extrasolar planets over time.
values for the extrasolar planets are plotted, with those of the upper envelope (maximum for a given year of discovery) indicated in black. The black curve is the logistic best-fit curve of the upper envelope, using a nonlinear model, where and . The horizontal grey line indicates the maximum value of , the presence of an Earth-like habitable planet.
Additionally, we conducted the same bootstrap analysis for subsets of the planetary dataset up to the end of the years 2001–2010 (prior to this, there is not enough data to yield robust estimates). This allows us to determine what the likeliest date of the discovery of an Earth-like planet would have been predicted to be, if the analysis were conducted throughout the previous decade. The median dates of discovery are shown in Fig. 4.
The creation of a single metric of habitability, , allows for quantitative prediction of when the first Earth-like planet is expected to be discovered – in this case, a date of early May 2011. Of course, this prediction of when the discovery of the first Earth-like planet will be announced has ignored technological advancement entirely, as well as many other factors. However, technological progress can often be well-described by a functional form independent of the processes underlying its advancement . Similarly, it is likely that the multiple methods of extrasolar planet discovery (such as radial velocity and transit methods) combine to yield relatively smooth curves on the march towards further discovery. The testable prediction given here is likely found to be accurate in the coming months, given the recent launch and ongoing results of many projects.
A great deal of current interest is focused on NASA's ongoing Kepler mission . The Kepler spacecraft employs the photometric transit method to detect planet candidates, and it is an open question as to whether this method can achieve the first detection of a planet with . While the initial results of Kepler were released on June 15, 2010, the Kepler team has delayed publication of 400 of the most promising extrasolar planetary candidates until February 2011. Within this large pool of withheld candidates, it is virtually certain that some have radii that are observationally indistinguishable from Earth's radius. It is likely, however, that because of the limited time base line of the mission to date, the Kepler planet candidates to published in February 2011 may be too hot to support significant values for .
In order to determine how useful Kepler will be in the search for a habitable planet, we reran our prediction analysis using only those planets discovered using the transit method. And it turns out that the method is unable to converge on a likely date of discovery, due to the paucity of the data (62 planets) and the low values for these planets. No doubt Kepler will increase this number of planets, but this provides a counter-balance to the assumption that the Kepler team will discover the first habitable planet.
It must be noted that by publicizing our prediction, there is a concern that it will become accurate, simply due to the well-studied Hawthorne Effect . However, due to the large number of observations and long periods of time required to confirm an extrasolar planet discovery, it is unlikely that our prediction at this time will appreciably affect the announcement of the discovery of an Earth-like planet.
Therefore, it is reasonable to use the habitability metric curve as a rough prediction for when the first potentially habitable planet will be discovered, in this case, as early as May 2011, and likely by the end of 2013.
We would like to thank Jukka-Pekka Onnela for reading drafts of this manuscript and David Charbonneau for the initial conversation which prompted this investigation.
Conceived and designed the experiments: SA GL. Performed the experiments: SA. Analyzed the data: SA. Contributed reagents/materials/analysis tools: SA GL. Wrote the paper: SA GL.
- 1. Santos N, Benz W, Mayor M (2005) Extrasolar planets: Constraints for planet formation models. Science 310: 251–255.N. SantosW. BenzM. Mayor2005Extrasolar planets: Constraints for planet formation models.Science310251255
- 2. Mayor M, Queloz D (1995) A jupiter-mass companion to a solar-type star. Nature 378: 355–359.M. MayorD. Queloz1995A jupiter-mass companion to a solar-type star.Nature378355359
- 3. Koch DG, Borucki WJ, Basri G, Batalha NM, Brown TM, et al. (2010) Kepler mission design, realized photometric performance, and early science. Astrophysical Journal Letters 713: L79–L86.DG KochWJ BoruckiG. BasriNM BatalhaTM Brown2010Kepler mission design, realized photometric performance, and early science.Astrophysical Journal Letters713L79L86
- 4. Laughlin G (2010) Systemic: Characterizing Planetary Systems. G. Laughlin2010Systemic: Characterizing Planetary Systems.Available: http://oklo.org/. Available: http://oklo.org/.
- 5. Moore G (1965) Cramming more components onto integrated circuits. Electronics 38: 114–116.G. Moore1965Cramming more components onto integrated circuits.Electronics38114116
- 6. Grossman L (2004) Forward thinking. L. Grossman2004Forward thinking.TIME October 11, 2004. TIME October 11, 2004.
- 7. Vogel G (1999) Planetary systems: Expanding the habitable zone. Science 286: 70–71.G. Vogel1999Planetary systems: Expanding the habitable zone.Science2867071
- 8. Schilling G (2007) Exoplanets: Habitable, but not much like home. Science 316: 528b.G. Schilling2007Exoplanets: Habitable, but not much like home.Science316528b
- 9. Mischna MA, Kasting JF, Pavlov A, Freedman R (2000) Influence of carbon dioxide clouds on early martian climate. Icarus 145: 546–554.MA MischnaJF KastingA. PavlovR. Freedman2000Influence of carbon dioxide clouds on early martian climate.Icarus145546554
- 10. Kasting JF, Whitmire DP, Reynolds RT (1993) Habitable Zones around Main Sequence Stars. Icarus 101: 108–128.JF KastingDP WhitmireRT Reynolds1993Habitable Zones around Main Sequence Stars.Icarus101108128
- 11. Lammer H, Selsis F, Chassefière E, Breuer D, Grießmeier JM, et al. (2010) Geophysical and atmospheric evolution of habitable planets. Astrobiology 1: 45–68.H. LammerF. SelsisE. ChassefièreD. BreuerJM Grießmeier2010Geophysical and atmospheric evolution of habitable planets.Astrobiology14568
- 12. Survey CP (2010) Exoplanet Orbit Database. CP Survey2010Exoplanet Orbit Database.Available: http://www.exoplanets.org/: Accessed June 14, 2010. Available: http://www.exoplanets.org/: Accessed June 14, 2010.
- 13. Arbesman S (2010) Quantifying the ease of scientific discovery. Scientometrics. S. Arbesman2010Quantifying the ease of scientific discovery.ScientometricsDOI:10.1007/s11192-010-0232-6. DOI:10.1007/s11192-010-0232-6.
- 14. Price D (1986) Little science, big science– and beyond. New York: Columbia University Press. D. Price1986Little science, big science– and beyondNew YorkColumbia University Press
- 15. Wordsworth R, Forget F, Selsis F, Madeleine JB, Millour E, et al. (2010) Is gliese 581d habitable? some constraints from radiative-convective climate modeling. R. WordsworthF. ForgetF. SelsisJB MadeleineE. Millour2010Is gliese 581d habitable? some constraints from radiative-convective climate modeling.Technical Report arXiv: 1005.5098. Technical Report arXiv: 1005.5098.
- 16. Koh H, Magee C (2006) A functional approach for studying technological progress: Application to information technology. Technological Forecasting and Social Change 73: 1061–1083.H. KohC. Magee2006A functional approach for studying technological progress: Application to information technology.Technological Forecasting and Social Change7310611083
- 17. Borucki WJ, Koch D, Basri G, Batalha N, Brown T, et al. (2010) Kepler planet-detection mission: Introduction and first results. Science 327: 977–980.WJ BoruckiD. KochG. BasriN. BatalhaT. Brown2010Kepler planet-detection mission: Introduction and first results.Science327977980
- 18. Landsberger H (1958) Hawthorne Revisited. Ithaca: Cornell University. H. Landsberger1958Hawthorne RevisitedIthacaCornell University