Armin
Bauer
* and
Mark
Brönstrup†
*
Sanofi-Aventis Deutschland GmbH, R&D LGCR/Chemistry, Industriepark Höchst G838, 65926 Frankfurt am Main, Germany. E-mail: armin.bauer@sanofi.com; mark.broenstrup@sanofi.com
First published on 21st October 2013
Covering: up to March 2013
In addition to their prominent role in basic biological and chemical research, natural products are a rich source of commercial products for the pharmaceutical and other industries. Industrial natural product chemistry is of fundamental importance for successful product development, as the vast majority (ca. 80%) of commercial drugs derived from natural products require synthetic efforts, either to enable economical access to bulk material, and/or to optimize drug properties through structural modifications. This review aims to illustrate issues on the pathway from lead to product, and how they have been successfully addressed by modern natural product chemistry. It is focused on natural products of current relevance that are, or are intended to be, used as pharmaceuticals.
(i) How can structural variations be performed in a time- and cost-efficient manner to modulate and enhance their biological activities and to optimize pharmacodynamic, pharmacokinetic and safety (ADMET‡) properties of the lead structure?
(ii) How can a large-scale supply of drug substance in GMP (Good Manufacturing Practices)-quality be realized?
In the current review, we would like to present recent highlights from natural product research and development that illustrate how the lead optimization and/or compound supply challenge has been successfully solved through modern natural product chemistry. For a comprehensive overview on natural products in clinical development, we would like to refer to recent review articles.6,7
To tackle the above-mentioned challenges, a growing arsenal of methods has been developed in organic chemistry, biochemistry, molecular biology, molecular genetics, and combinations thereof. Innovations in synthetic methodology have always been closely mixed with the total synthesis of natural products; they induced several paradigm shifts in natural product chemistry, as outlined in a recent review.8 In fact, an effective synthetic access to a natural product lead continues to be the critical determining factor for the successful accomplishment of medicinal chemistry optimization programs.9
Understanding and exploiting the biosynthesis pathways of natural products has an increasing impact on accelerating their discovery and development process.10 Turn-around times for strain improvement may be considerably reduced by rational approaches based on knowledge of the genetic regulation of secondary metabolite production and of quantitative fluxes of precursors. Progress in genetic engineering techniques like precursor-directed biosynthesis, combinatorial biosynthesis and mutasynthesis approaches11,12,13,14 are the foundations of “biosynthetic medicinal chemistry”,15,16 opening up a structural space that is often orthogonal to classical natural product derivatisation or total synthesis approaches.
The extended range of options has led to the intriguing situation that there is no single recipe for success. Instead, we observe revisions of long-prevailing standards for product realization:
– Plant metabolites have been provided as drug substances through horticulture, followed by extraction and purification processes for centuries, with various inherent drawbacks. The example of artemisinin production (chapter 5.3) demonstrates that heterologous expression in genetically engineered standard hosts, followed by semisynthesis, has now become a low-cost, high quality alternative.
– The multistep total synthesis of polyketides or peptides, mostly seen as a purely academic discipline, has become a viable option for industrial production, in particular for drugs with treatment regimens that are limited in time or require low dosages. This is exemplified by eribulin, the GLP-1 analogs and ziconotide (chapters 2.3, 3.3, and 5.1, respectively).
– For several generations, fermentation/semisynthesis was the only practical method to overcome bacterial resistance against the ‘mature’ tetracycline class of antibiotics (chapter 2.4). A recently developed, highly convergent total synthesis enabled a broad structure–activity-relationship (SAR) study of structural parts that have not been amenable to semisynthesis, and led to the identification of a novel candidate that is profiled in Phase II clinical trials.
– The rapamycin scaffold (chapter 3.4) exhibits such unique biological properties that improved, next generation congeners were generated through subtle structural modifications by semisynthesis or genetic engineering. In stark contrast, the optimization of sodium–glucose co-transporters inhibitors (chapter 2.1) or an S1P1 antagonist (chapter 2.2) through medicinal chemistry led to pronounced structural changes, rendering the original natural product lead behind the final compounds being hardly discernible.
– Drug researchers traditionally had to decide whether to address a given biological target either with a biological macromolecule (e.g. a monoclonal antibody, or an oligonucleotide) or a small molecule (e.g. a natural product or a synthetic compound). Recently, it has been proven in a clinical setting that both formats can be combined in a single molecule, a so-called immunoconjugate (chapter 4), through the covalent linkage of natural product effectors to a monoclonal antibody, while retaining key advantages of each format. This approach will be illustrated by ado-trastuzumab emtansine and brentuximab vedotin.
– Is there still space for the classical semisynthetic modification of natural product scaffolds obtained by fermentation to discover drugs? Definitely yes. In most cases, the approach remains the most efficient to explore an SAR in the short timescales of drug discovery, and it is also the most cost-effective, sustainable and reliable supply of complex natural product scaffolds for drug development and production.17 Its versatility is demonstrated by the successful development of next generation glycopeptides, echinocandins, and rapamycins (chapters 3.1, 3.2 and 3.4, respectively).
Thus, today's natural product chemists have a rich and growing arsenal of methods at hand. In the following chapters, successful applications of this arsenal to optimize natural products by total synthesis (chapter 2), semisynthesis (chapter 3), and immunoconjugation (chapter 4), or to produce them in unmodified forms (chapter 5) will be presented.
The history of the development of SGLT inhibitors began with phlorizin 1, a dihydrochalcone glycoside that was first isolated in 1835 from the root bark of the apple tree.211 played a key role in (i) uncovering the function of renal glucose re-absorption, (ii) the discovery of the SGLTs and (iii) as a structural starting point for lead optimization.
In the late 19th century, it was observed that 1 induced glucosuria after oral administration, and it was soon recognized that diabetes mellitus symptoms like persistent excretion of glucose and polyuria were mimicked by chronic administration of this compound to dogs.22 As early as 1899, 1 was used for the first time experimentally in a diabetes patient to lower serum glucose levels.18,23
Subsequently, 1 served as a valuable tool compound for a better understanding of kidney function24 and for uncovering the link between hyperglycemia and insulin resistance.25 Studies on the mechanism of action on the cellular level started in the 1950s. They revealed that 1 blocked sugar transport in both the small intestine and the kidney,26 and finally led to the identification of the sodium–glucose transporters (SGLTs) responsible for the renal re-absorption of glucose.27 The two isoforms SGLT1 and SGLT2 have been cloned from several species,28 and 1 turned out to be a non-selective inhibitor of both human isoforms.29 Since blocking of SGLT1 was thought to be associated with side effects, in particular through affecting gastrointestinal function, selective inhibition of SGLT2 was considered as an important prerequisite for the development of a safe drug. Further issues of 1 included a low metabolic stability due to rapid hydrolysis by glucosidases, resulting in poor pharmacokinetic profiles (in particular, low oral bioavailability and short plasma half lives), and toxic effects associated with the liberated aglycon phloretin 2.
First reports on improved SGLT2 inhibitors based on 1 were reported in the late 1990s by researchers at Tanabe.30 As a result, the O-glycoside analog T-1095 3 emerged as one of the first clinical development candidates.31 The toxicity problems were addressed by removing or modifying hydroxy groups in the aglycon part based on earlier reports on structure–activity relationships of 1 and 2.32 In order to prevent hydrolysis by β-glucosidases in the gastrointestinal tract, the glucose moiety was modified as a methyl carbonate prodrug. Although 3 showed promising effects in several diabetes models, its selectivity vs. SGLT1 was moderate (only 4-fold). Its development was discontinued in phase II.33 Subsequently, several companies developed O-glycoside SGLT2 inhibitors, and many of them entered clinical trials (like sergliflozin etabonate 4 and remogliflozin 5).34,35 However, the development of all “first-generation” compounds has been abandoned, likely due to the intrinsic metabolic lability of the O-glycoside bond, especially in humans, resulting in insufficient plasma half lives.35 The fact that the O-glycoside bond is more stable in rodents may have contributed to the (too) late recognition of this liability.
This issue could be solved with the C-glycoside-based “second-generation” SGLT2 inhibitors. Initial attempts at replacing the O-glycoside bond by a methylene bridge led to a drastic loss of activity. In contrast, the direct C-aryl glucosides, initially discovered at BMS by coincidence as unanticipated side products of O-glycosylation reactions, turned out to be potent and selective SGLT2 inhibitors after systematic variation of the aryl substitution pattern.36 Dapagliflozin 6, co-developed by BMS and AstraZeneca, represents the frontrunner of this approach,37 quickly followed by other C-glycosides34,38 like canagliflozin 7,39 empagliflozin 8,40 ipragliflozin 941 and tofogliflozin 10.42 All of them have advanced to phase III clinical trials. It turned out that this new class of oral drugs was generally well tolerated and revealed beneficial effects, such as reductions in fasting and postprandial glucose and HbA1c levels.43 In addition, favorable add-on effects like body weight reduction or blood pressure reduction were observed.44 The European Medicines Agency (EMEA) recommended the approval of Forxiga™ (dapagliflozin) 6 for the treatment of type 2 diabetes in April 2012, whereas the FDA has requested further studies assessing the risk–benefit profile before taking a final decision on approval in the US. In March, 2013, canagliflozin 7 (trade name: Invokana™) was approved for the US market to treat adults with type 2 diabetes. The development of SGLT-inhibitors is a success story for medicinal chemistry, but arguably even more for natural product research. Phlorizin has not only been the lead structure for all advanced compounds to date, but its application as a tool compound was essential for disclosing an intriguing area of human biology and medicine.
These remarkable effects triggered various investigations on the mechanism of action of 11. The immunosuppressive effect of 11 could be ascribed to a strong decrease of the number of peripheral blood lymphocytes (T and B cells). The analysis of the lymphocyte distribution in blood, lymph and several secondary lymphoid organs (SLO) revealed that sequestration of lymphocytes into the SLOs was enhanced, while egress into peripheral blood and lymph was decreased,51 thereby preventing immune cells from reaching inflammatory tissue.52 As 11 displayed a close structural homology with sphingosine, further studies were focused on putative interactions with intracellular sphingolipid metabolism.53 They demonstrated that 11 was converted to fingolimod phosphate 15 by sphingosine kinases (Scheme 1). The target of 15 was identified as a new class of transmembrane, G protein-coupled receptors (GPCRs) termed sphingosine 1-phosphate 16 (S1P) receptors.54 This disclosure led to intensified research into the whole field of lysophospholipid receptors as a pharmacological relevant target class.5515 modulated the function of S1P1, leading to internalization of the receptor and down-regulation at the level of gene expression.56 By this process, the S1P-induced signal that lymphocytes required to egress from the lymph system into inflammatory tissue was quenched. As a result, the attack of myelin antigens in the CNS by autoreactive lymphocytes, a characteristic event in progressing multiple sclerosis, was ultimately reduced.57 The discovery of fingolimod impressively illustrates the value of natural products for uncovering intriguing biological mechanisms. Myriocin was in addition the lead structure, notably with features that are remote from lead- or drug-like molecules according to textbook rules. We speculate that it would not survive the majority of today's filtering processes applied in the active-to-lead phase. A reduction of structural complexity of natural products, e.g. through the removal of stereogenic centers, is mostly associated with the loss or drop of biological activity. In contrast, a drastic simplification has been possible along the optimization of 12 to 11, making the final product hardly discernible as one derived from nature.
In further studies, 18 has been shown to interact with tubulin to create non-productive tubulin aggregates, resulting in the suppression of microtubule assembly without disruption of existing tubulin architecture. This interaction led to arrest in the G2-M phase of the cell cycle and subsequently to apoptosis. Further investigations revealed that 18 bound to the vinblastine site of tubulin in a non-competitive fashion and had no effect on colchicin binding.62,63 This unprecedented mechanism of action strongly contributes to the therapeutic value of the halichondrin family of antimitotics: activity is still displayed on cancer cells resistant to other antimicrotubule agents like taxanes.64 Compound supply was the major hurdle for clinical development since the beginning. The initial amount to start development activities was calculated to be around 10 g of halichondrin B 18 and the future commercial demand of drug substance was estimated to be 1–5 kg year−1.65 Due to the rare abundance of organisms producing 18 and the extremely low yield of isolated compound, collection from the wild was soon ruled out. The supply issue for the start of preclinical studies was partially solved by aquaculture of Lyssodendorix, which led to the provision of 310 mg of 18 from one ton of sponge.66 Nevertheless it soon became obvious that the only viable alternative would be total synthesis. The first successful syntheses of both halichondrin B 18 and norhalichondrin were reported in 1992 by the Kishi group at Harvard.67 However, this route and others developed subsequently68 could not be regarded as cost-effective.
The highly convergent and modular synthesis approach was used by the Kishi group and medicinal chemists at Eisai in order to produce simpler analogs and to identify the minimum pharmacophore of 18. The testing of intermediates from the total synthesis efforts revealed that growth inhibitory activity on DLD-1 cells could be traced back to the right half macrolactone fragment 19.65
However, 19 was devoid of activity in human tumor xenograft models – in contrast to the parent natural product. Based on the hypothesis that the right half intermediate 19 only induced a reversible mitotic block under the dosing schemes of the in vivo experiments, several regions of the polyether moiety were systematically modified, and the resulting analogs were tested in a sophisticated flow cytometric assay in order to characterize their ability to induce a complete mitotic block after drug washout.69 Subsequent studies guided by this assay revealed that simple furan or pyran fragments could replace the western fused pyran moiety. Finally, further functional group exploration and the replacement of the ester linkage in the macrolactone ring by a ketone led to the identification of eribulin 17 (E7389),70 an analog that displayed the desired potency on a panel of cancer cell lines without reversing mitotic block. Due to its remarkable biological profile, 17 entered development at Eisai. Although the molecular architecture of 17 is significantly simpler than that of 18, the compound still contains 19 stereogenic centers and can be regarded as the most structurally complex drug substance on the market to date which is prepared by total synthesis.
During preclinical development, the major obstacles that had to be overcome in the context of this unprecedented complex structure were the cost of goods and synthetic feasibility. Continuous scientific contributions from the Kishi group and considerable efforts at Eisai chemical development finally led to a cost-effective, scalable synthesis of 17 in 62 steps,71 which opened the way for the first clinical phase I trials in collaboration with the NCI. The positive outcome of a phase III study72 in patients with metastatic breast cancer previously treated with established drugs in a setting where most treatment schemes fail finally led to the approval of 17 as Halaven™ in 2010 in the US and 2011 in Europe.73
Further semisynthetic modifications led to the “second-generation” analogs doxycycline 23 (Vibramycin™) and minocycline 24 (Minocin™) launched in the 1960s and 1970s, which have played a dominant role in clinical practice until today. The discovery, development and application of the first- and second generation tetracyclines has been extensively reviewed.76 In order to overcome resistance problems created by the widespread use of tetracyclines in human and veterinary medicine, a third generation is currently being developed. These efforts resulted so far in the launch of tigecycline 25 (Tygacil™) as the first representative of glycyl-modified tetracyclines in 2005.77
The high complexity and functional density of the tetracyclines has attracted organic chemists for a long time. The first racemic total synthesis of a tetracyline was published by the Woodward group in 196278 and served as a basis for the synthetic strategies of other approaches realized in the following decades,79 in particular by the Muxfeldt group:80 the construction of the ring system is performed in a linear manner in the D → A direction. However, the inefficiency of this approach has hampered the application of total synthesis for providing clinical candidates (with one notable exception, the analog (±)-6-thiatetracycline 2681), leaving semisynthesis as the practical method of choice to generate modified analogs with improved antimicrobial and pharmacokinetic profiles – until recently.
The situation changed with a report of a new, highly convergent approach to the tetracyclines by the Myers group in 2005.82 As outlined in Scheme 2, their strategy is based on the parallel preparation of D ring and AB ring fragments (27 and 28, respectively), which are assembled to the protected intermediate 29 at a very late stage of the synthesis with simultaneous formation of the C ring of the tetracyline system 30.83 This approach is approximately 10 steps shorter than previously reported syntheses, rendering it much more efficient and practical for the preparation of sufficient amounts of compounds for biological characterization. After further refinement,84 the coupling of building blocks 28 and 27 turned out to be extremely flexible and led to the synthesis of more than 3000 new tetracyclines.85
Many of those derivatives would have been extremely difficult or impossible to prepare by semisynthesis approaches, in particular heterocyclic D ring analogs or compounds with additional cycles condensed to the D ring. Modifications at this particular region are regarded to be particularly valuable, since X-ray studies with tetracyclines bound to the target – the 30S subunit of the bacterial ribosome – suggested the D ring to be the most promising site for structural variations.86 Several of the new analogs displayed an outstanding profile, breaking through resistance vs. tetracyclines and other antibiotics.
Systematic variation of this area has led to the identification of eravacycline 31 (TP-434), the first representative of a new sub-class of tetracylines that are fluoro-substituted at the D-ring, the so-called fluorocyclines.87 Eravacycline 31 showed a promising broad-spectrum antibiotic profile and was claimed to overcome common tetracycline resistance mechanisms.88 In addition, an oral administration route is principally feasible; its efficacy is currently being investigated. The compound has successfully completed a phase II clinical trial in patients with complicated intra-abdominal infections. The efficiency of the synthesis seems to pave the way for a cost-effective scale-up for the drug substance requirements of antibiotic therapy.
The eravacycline case demonstrates how a novel synthetic route enabled the discovery of new, promising drug candidates in a seemingly mature, well-exploited area.
The molecular basis of vancomycin resistance is now well understood. Inhibition of bacterial cell wall synthesis is mediated by tight binding of the antibiotic to the integral precursor peptidoglycan peptide terminus D-Ala-D-Ala. Remodeling of this cell wall precursor to D-Ala-D-Lac is the underlying mechanism for the two most prominent types of resistance, VanA and VanB.92
Glycopeptides like 32 and teicoplanin 33 display a fascinating and complex molecular architecture. It consists of a heptapeptide backbone that is bent to a concave shape through biphenyl and diphenylether cross-links of aromatic amino acids residues. In the case of 32 it took nearly three decades from the first isolation in 195693,94 until the complete disclosure of its structure95 and total syntheses were not accomplished until 1998.96 With a total synthesis at hand, it became possible to break the VanA and VanB resistance through a rational re-design of the vancomycin peptide backbone. Repulsive interactions between the oxygen atom of the (D-Ala-)D-Lac moiety of resistant bacteria and the carbonyl group of the central, substituted phenylglycine amino acid were eliminated through a ‘surgical’ replacement of the carbonyl by a methylene or an imine group, thereby restoring the affinity towards the target.97
However, all derivatives in advanced drug development have been generated by semisynthetic modification of the natural products. For obvious practical reasons, the root cause of resistance, i.e. the binding pocket of the peptidoglycan peptide terminus, was not modified, but the periphery was decorated with hydrophobic and hydrophilic moieties. Surprisingly, the new derivatives successfully restored sensitivity of resistant bacteria, as discovered by scientists at Lilly in the 1990s with a series of N-alkyl derivatives of vancomycin.98 A fine-tuning of ADME properties of derivatives based on the closely related scaffolds of 32, chloroeremomycin and A40926, a member of the teicoplanin family, led to the development of telavancin 34,99 oritavancin 35100 and dalbavancin 36, respectively, as the most advanced representatives of the second-generation glycopeptides (Fig. 2).94
Telavancin 34 (Vibativ™) was launched in 2009,101 and oritavancin 35 and dalbavancin 36 are currently undergoing phase III clinical trials.102 Due to the hydrophobic side chains that characterize the new glycopeptides, they were placed into a separate sub-class called “lipoglycopeptides”. The structural modifications introduced two favorable properties: 1) a membrane anchoring of the lipid chain positioned the glycopeptide close to its target, and it led to disruption of membrane potential, increased permeability and finally, cell lysis. This additional mode of action also rendered the lipoglycopeptides rapidly bactericidal, whereas vancomycin displayed a bacteriostatic profile. 2) The side chains enhanced the homodimerization of the lipoglycopeptides. Due to multivalency, the homodimers exhibited a much tighter binding to the peptidoglycan, which translated to largely improved in vitro activities.103 In addition to enhanced activities against VRE or VRSA, 34–36 also exhibit strongly (10–100-fold) improved activities against S. pneumonia compared to vancomycin.104 The expansion of the indication areas to nosocomial pneumonia is under current review of regulatory authorities in the EU and the US.
The increased hydrophobicity also led to higher protein binding compared to vancomycin, resulting in longer in vivo half-lives of the drug and thus longer dosing intervals (of up to one weekly injection in the case of 36) in patients.105,106
In summary, semisynthetic modifications carefully tuned the compound's physicochemical properties. This led to an additional cellular mode of action and in consequence, restored efficacy against resistant pathogens and improved pharmacokinetic properties were obtained as clinical benefits.
There are several key features that make them a valuable addition to the existing treatment options for fungal infections: compared with azoles and polyenes, echinocandins display an enlarged spectrum for Candida spp. beyond C. albicans, an improved renal and hepatic safety profile and reduced cytochrome-mediated drug–drug interactions. These favourable profiles can be partly assigned to the fact that the echinocandins – in contrast to other antifungals – act on a target that is not present in mammals, since they inhibit β-(1,3)-D-glucan synthase, an essential enzyme complex in the formation of the fungal cell wall.108
The development of all echinocandins launched today can be traced back to a few closely related representatives of a class of large lipopeptides identified from fungal sources in the 1970s and 1980s. The echinocandins are amphiphilic cyclic hexapeptides comprised of rather hydrophilic amino acids and a fatty acid side chain. Echinocandin B 40, a fermentation product of Aspergillus nidulans and the starting point for the development of 39, was discovered in 1974 as the first representative of the class.10940 showed potent activity against Candida albicans and Candida tropicalis, and in vivo activity was demonstrated in an animal model of candidiasis. On the other hand, low solubility, poor oral bioavailibility and potential hemolytic side effects had to be overcome before clinical development.110 It soon became evident that the lipophilic side chain was an important structural feature that modulated both activity and toxicity and offered potential for the fine-tuning of pharmacokinetic properties of this substance class. The development of a biotransformation process for the deacylation of 40 by Actinoplanes utahensis turned out to be key for access to a broad variety of side-chain modified echinocandins.111 As shown in Scheme 3, removal of the linoleoyl unit from 40 led to a cyclic hexapeptide core (deacyl-echinocandin B 41) devoid of antifungal activity. The first new derivative to enter clinical trials, cilofungin 42, was selected by Lilly from several analogs obtained by the re-acylation of this core.
Structure–activity relationship studies revealed that the acyl moiety had to be a fatty acid chain of at least 12 carbon atoms to restore anti-fungal activity. Shorter and/or more polar acyl motifs yielded inactive compounds. These observations led to the assumption that a lipophilic moiety of sufficient length is necessary for attachment to the fungal membrane, preceding the interaction with the glucan synthase as the molecular target; a similar structural requirement and mechanism was discussed above in the context of increased activity of lipoglycopeptides in comparison to vancomycin.112 Although clinical studies with cilofungin 42 gave at first encouraging results for some types of Candida infections, the product had to be discontinued: cilofungin still displayed the low water solubility of its parent echinocandin B; finally, the formulation that had to be used for a parenteral administration caused adverse effects.111,113
Despite this first setback, inhibition of glucan synthesis remained an attractive approach for antifungals, and further activities in the field were spurred by the increased medical need for antifungal agents in the context of the growing number of immunocompromised patients, in particular those with secondary infections associated with AIDS. New whole-cell based assays for the detection of glucan synthesis inhibitors were developed and led to the discovery of a series of new antifungal lipopeptides.114 Lilly scientists focused on further exploration of structure–activity relationships around 40 and 42. A comparison of the lipophilicities of the side chain fragments revealed that calculated logP values of >3.5 were required for antifungal activity. Furthermore, side chains with linear, rigid moieties (such as [1,1′;4′,1′′]-terphenyls) were most effective in increasing the antifungal potency, and incorporation of such side chains resulted in compounds with long half-lives.115 The development candidate LY303366 was the outcome of these efforts. LY303366 was effective and well tolerated in clinical studies. In contrast to 41, no major adverse effects were observed, and the successful development culminated with the launch of the compound as anidulafungin 39 (Eraxis™).108
The starting point for the development of caspofungin 37 (Cancidas™) was a new sub-class of echinocandin-like lipopeptides, named pneumocandins, which were discovered at Merck in the 1980s in a broad screening campaign for cell-wall targeting antifungals. The pneumocandins had a particular attraction since they displayed an enlarged spectrum against several Candida pathogens. Moreover, they were active on Pneumocystis carinii, an opportunistic organism causing life-threatening pneumonia in AIDS patients, and showed no hemolysis in vitro. Pneumocandins were produced by the fungus Zalerion arboricola, with pneumocandin A043 being the major secondary metabolite in the original culture, accompanied by pneumocandin B044 and D045 as the most abundant minor fermentation products.116 These compounds differ from 40 by the fatty acid side chain and one or two amino acid residues.
Initially 43 was selected for further optimization by medicinal chemists to overcome some limitations still associated with the natural product. Improvements in antifungal spectrum, potency, chemical stability and water solubility were required. However, after careful profiling of the minor fermentation products, it turned out that 44 was considered as a more suitable starting point for chemical optimization. The limitations of compound supply as a prerequisite for a large program of semisynthetic modifications was overcome by developing a producer strain with increased titers of 44 (now as the main product) and reduction of fermentation by-products. This producer strain was obtained after isolating further related strains which displayed higher titers of 44, followed by mutant selection and media variations.117 In contrast to the Lilly approach, the optimization strategy of the Merck scientists did not follow a biotransformation-based removal–reacylation strategy for a variation of the lipophilic side chain, but focused on modifications of the peptide core structure, leaving the fatty acid chain untouched. Synthetic modifications turned out to be challenging since the natural product 44 was unstable under non-neutral conditions and displayed limited solubility in a variety of solvents. Nevertheless the semisynthesis efforts resulted in a first development candidate which solved the solubility issue: a phosphate ester prodrug 46 was obtained from 44 through esterification of the phenolic hydroxyl group of the homotyrosine moiety.118 In parallel, additional positions on the nucleus 44 were modified. This work was partly guided by the total synthesis of echinocandin D119 and simplified analogs that helped to figure out the minimum structural requirements for antifungal activity.120 Both the modification of the hemiaminal group as well as the reduction of the carboxamide function of the hydroxy glutamine moiety to a primary amine (hydroxy ornithine) turned out to be favourable: The 2-aminoethyl-substitution on the aminal conferred improved stability, solubility and higher activity against Candida and Aspergillus in vivo, while the reduction of the glutamine led to even further improved anti-Candida activity.121,122 Finally, the replacement of the oxygen in the hemiaminal moiety by nitrogen gave the aza analogue MK-0911, now caspofungin 37, with an improved pharmacokinetic and safety profile. MK-0911 proved to be superior to the pneumocandin B0 phosphate ester prodrug 46 and was therefore chosen for clinical development.110
Scientists at Fujisawa focused on water solubility as a major selection criterion for optimization starting points in their search for β-(1,3)-D-glucan synthase inhibitors.123 These efforts resulted in the discovery of FR901379 47 from the fungus Coleophoma empetri F-11899.124 FR901379 47 had an excellent water solubility of 50 mg mL−1 (compared to 0.008 mg mL−1 for 40) due to the presence of a sulfate group on the homotyrosine moiety. The compound was more potent against Candida in vitro than 40. On the other hand, a pronounced hemolytic activity was observed, which was assigned to the linear aliphatic fatty acid moiety.125 Optimization was therefore guided by the knowledge obtained in the context of echinocandins and cilofungin chemistry.126 Thus, FR901379 47 was deacylated with Actinoplanes utahensis and re-acylation studies confirmed – as in the case of 42 and its analogs – that rigid, linear aromatic moieties were ideally suited to obtain improved compounds.127 Finally, a compound with a side chain closely related to that of 39 was chosen as a development candidate: FK463, now micafungin 38, can be regarded as a close analog of anidulafungin 39, the main differential structural features being the sulfate group on the homotyrosine unit and the replacement of the central phenyl ring of the terphenyl side chain by an isoxazole.128
However, the use of unmodified human GLP-1 48 is limited by its rapid degradation through dipeptidyl peptidase IV (DPP-IV), resulting in an extremely short plasma half-life of 2–5 min.134 The development of modified GLP-1 agonists resistant to degradation by DPP-IV enabled the translation of the incretin mimic concept into clinical practice.135 Two principles were followed (Fig. 4): (i) in the case of exendin-4 50 (exenatide, marketed as Byetta™), an altered amino acid sequence originating from an animal homolog (see below) was a much weaker substrate of DPP IV. 50 was the first GLP-1 agonist on the market, launched in the US in 2005 and in Europe in 2007. Lixisenatide 51 (launched as Lyxumia™ in 2013) followed the same principle, but achieved further prolonging of plasma half-life through the attachment of a hexa-lysine residue at the C-terminus. (ii) Liraglutide 49 (launched as Victoza™ in 2010) was directly derived from the human GLP-1 sequence and modified with a palmitate side chain linked through a γ-L-glutamoyl residue. The fatty acid moiety exhibits a strong, reversible binding to human serum albumin, thereby conferring protection against degradation.136
The discovery of 50 was initially unrelated to the “incretin concept”. 50 is a linear, 39-amino acid peptide that shares about 50% homology with human 48. It was isolated in 1992137 from the salivary gland of the lizard Heloderma suspectum, known as the Gila monster originating in Arizona and New Mexico. The discovery was based on earlier observations that the venom of Heloderma suspectum stimulates adenylyl cyclase activity in guinea pig pancreatic acinar cells.138 In a first hypothesis it was suggested that 50 activated a putative exendin receptor to increase pancreatic acinar cAMP.137
After human GLP-1 was recognized as a close structural homolog,139 it was demonstrated that 50 stimulated insulin secretion in murine insulinoma cells through activation of the GLP-1 receptor, a Gi coupled GPCR that indeed enhances intracellular cAMP levels.140 These findings triggered clinical investigation of 50 in type 2 diabetes.
Since the isolation of 50 from an animal source to ensure a sustainable supply was completely out of the question, alternatives had to be elaborated. While the peptide portion of liraglutide 49 is currently being produced in recombinant yeast,13650 is manufactured by solid-phase peptide synthesis (SPPS) employing the 9-fluorenyl-methoxycarbonyl (Fmoc) strategy.141 Despite the intrinsic drawbacks of SPPS (high reagent costs, considerable generation of waste) and the need for reversed-phase chromatography for final purification, the drug substance can be provided in a cost-effective manner: since the dose for this hormone-like drug is as low as 5–10 μg twice daily, 3.7–7.3 mg of active pharmaceutical ingredient is sufficient to treat a patient for a year.142
Rapamycin 52 was originally discovered in 1975 as a secondary metabolite from a Streptomyces hygroscopicus strain due to its antifungal activity. The compound was named rapamycin after the origin of its producer strain, which had been isolated from a soil sample collected at Rapa Nui (Easter Island).143 The complete structural assignment of 52 was accomplished a few years later based on X-ray crystallography and spectroscopic data.14452 was one of the first complex polyketide synthetase products whose biosynthesis was elucidated.145,146 Although 52 showed excellent anti-Candida activity,147 the compound was not developed as an anti-fungal agent. Based on further biological profiling experiments that revealed its potent immunosuppressive activity,14852 was at first developed without further structural modifications as the oral immunosuppressant drug sirolimus (Rapamune™, approved for prevention of rejection in organ transplantation in 1999).149 Considerable efforts have been made to unravel the details of the mode of action of 52 after functional effects of the compound on T cells had been studied in more detail.150 The molecular target was identified in a gene complementation assay in yeast. Rapamycin-resistant strains were selected, and the gene that restored sensitivity to rapamycin was cloned and named TOR for “target of rapamycin”.151 Shortly thereafter, the mammalian homolog referred to today as mTOR (mammalian target of rapamycin) was identified.152 It was found that 52 had the remarkable ability to occupy two hydrophobic binding sites at the same time (Fig. 5). It inhibited mTOR through binding initially to the immunophilin FKBP-12 (one of the FK506-binding proteins), and then forming a ternary complex with mTOR.153 mTOR is a serine-threonine kinase that plays a central role integrating signals from growth factors, nutrients, stress, hormones and mitogens to regulate survival, proliferation and cell growth. Disruption of mTOR activity led to inhibition of protein translation and blocking of the cell cycle transition from the G1 to the S phase.154
Fig. 5 Structures of rapamycin 52 and its analogs currently approved for human use or in clinical trials. The FKBP12 and mTOR binding regions are highlighted. |
In addition to its application in immunological indications, 52 was also considered as an anticancer agent due to its antiproliferative effect. Further evaluations were at first discouraged by the finding that for an antiproliferative effect in mouse cancer models, much higher doses than for the prevention of allograft rejection were needed, raising concerns that rapamycin would not be tolerated in humans at the high doses estimated to be required for anti-tumour activity.15552 gained renewed interest when further studies revealed that some tumour cell lines were highly sensitive to 52 and that the compound displayed a unique pattern of reactivity in the NCI-60 cancer cell line panel.156 Since 52 was poorly soluble and underwent extensive first-pass metabolism, resulting in potentially variable and low absorption and exposure,157 and since solubility and stability problems prevented the development of a parenteral formulation,158 analogs with improved pharmaceutical and pharmacokinetic properties were sought after. This effort has led to the development of several rapalogs, which are now in clinical use for a variety of indications in immunology, oncology and cardiology. To date, 5 distinct total syntheses of 52 have been reported.159,160 Nonetheless all derivatives that are currently in clinical use or late stage development are based on semisynthetic approaches. The most-explored site of derivatization was the unhindered hydroxyl function on the cyclohexyl moiety. Although the pharmacodynamic effects of the rapalogs seemed to be quite similar, the compounds differed in formulation, application and dosing schemes, thereby altering drug exposure.
One approach led to the highly soluble ester prodrug temsirolimus 53 (Torisel™, approved in 2007), which was designed for intravenous administration in oncology in order to achieve high exposures with low inter-individual variabilites in well-controlled chemotherapy settings.161 Temsirolimus 53 was mainly metabolized to 52; through parenteral administration of 53, higher plasma concentrations of the parent compound 52 were achievable than after direct oral administration of 52. Two other non-prodrug derivatives have been developed for cancer treatment: everolimus 54 (Afinitor™, approved in 2009)162 for oral administration and ridaforolimus (deforolimus) 55 for both oral and intravenous applications (currently in late stage clinical trials). In the structure of 54, the hydroxy group at the cyclohexyl moiety of rapamycin is replaced by a 2-hydroxy ethoxy group, while 55 carries a dimethyl-phosphinic acid ester at this position. 54 had a lower plasma protein binding and higher oral bioavailibility compared with 52.156 Yet another variation yielded zotarolimus 56 (ABT-578), a rapamycin derivative obtained by replacing the hydroxy function by a tetrazole under inversion of stereochemistry. Zotarolimus 56 is used exclusively in drug-eluting stents for the treatment of coronary artery disease. It is characterized by a high lipophilicity and a shorter terminal half life than 52, properties that were particularly desired for the application in stents to obtain a slow release of the drug directly into the wall of the coronary artery. Lipophilicity was considered to be favorable for crossing cell membranes to inhibit neointimal proliferation of the target tissue, and a short half life led to rapid clearance of the drug once released into the blood stream, thereby reducing systemic toxicity.163
Since the mTOR pathway is up-regulated in several malignant diseases, and the dysregulation of this pathway also involves upstream targets, several approaches beyond the development of rapalogs have been pursued. In particular, from directly targeting the active site of the kinase function of mTOR and the upstream kinases PI3K and Akt, several small molecule inhibitors – non-rapalogs – have emerged which are currently undergoing clinical trials.156,164,165
Like other immunophilin ligands, such as cyclosporin A and FK506, 52 may be regarded as a privileged template for the modulation of protein–protein interactions mediated by small molecules.166 Therefore both semisynthetic and biosynthetic approaches to new rapamycin analogs have been pursued to address other targets and indications beyond those related to the mTOR pathway. In the early 1990s it was reported that 52 and FK506 exhibited neurotrophic and neuroprotective effects.167 The finding that these effects were partially independent of inhibition of mTOR or calcineurin led to efforts towards uncoupling neuroprotective from immunosuppressive effects. In the case of 52 this was achieved by semisynthetic modifications in the mTOR binding region while leaving the FKBP (immunophilin) binding region untouched. The derivatives WYE-592 57 and ILS-920 58 were obtained after Diels–Alder reaction of the triene moiety of rapamycin 52 with nitrosobenzene and subsequent hydrogenation, respectively. Both compounds displayed significantly reduced immunosuppressive activity and were active in a rodent model of ischemic stroke.146,168 More recent efforts to access novel biologically active rapalogs have been based on genetic manipulation, precursor-directed biosynthesis and mutasynthesis.14,16,146,169,170
In the past three decades, a bold strategy has been pursued to combine the strengths of the two formats summarized above: The covalent coupling of a targeting monoclonal antibody to a natural product effector was expected to combine the specificity and tolerance of mAbs with the cell-killing efficiency of natural products in a single molecule, a so-called immunoconjugate or antibody–drug conjugate (ADC). After binding of the antibody to its cancer cell surface antigen, the ADC is internalized into the cell. Inside the cell, the linker between antibody and effector is cleaved through a chemical or enzymatic mechanism (see below), thereby releasing the ‘free’ natural product for cell killing.
The approach offers a great opportunity to turn extremely active natural products (with potential new modes of action) into useful drugs that might have been otherwise excluded from development due to high toxicity and safety issues. In fact, it is now well recognized that suitable toxic payloads for antibodies should have a potency in the range of 10–100 pM on a cellular level, which is 2–3 orders of magnitude higher than optimum potencies for conventional chemotherapeutic agents.171 While any class of compounds can serve as an effector in principle, almost exclusively natural products have been applied to fulfil this strict design criterion. As immunoconjugates can be regarded as highly sophisticated drug delivery systems, a careful optimization of all components – the monoclonal antibody, the cytotoxic payload and the chemical linker – is crucial for therapeutic efficacy.172
Gemtuzumab ozogamicin 60 (Mylotarg™) was the first antibody–drug conjugate to be approved. It is based on the highly cytotoxic agent calicheamicin γ1I59, which is linked to an antibody directed against the CD33 antigen present on leukemic myeloblasts.173 The acid-labile hydrazone linker is cleaved in the lysosome, thereby releasing the calicheamicin moiety in 60 that acts through a unique mechanism: DNA is targeted through interaction of the sugar moiety with selected base pairs in the minor groove. Upon reductive activation of the disulfide moiety a thiolate is formed, which then attacks the enone function of the enediyne moiety, thus triggering a Bergman cyclisation which generates a diradical. Reaction of this diradical with the DNA backbone leads to double strand cuts, ultimately resulting in cell death.174
Despite positive clinical results that led to the accelerated approval of gemtuzumab ozogamicin 60 in 2000, the drug was withdrawn in 2010 due to concerns about safety and clinical benefit that were raised after subsequent phase III trials. Another calicheamicin-based immunoconjugate, inotuzumab ozogamicin targeting the CD22 antigen expressed on malignant B cells, is still in late stage clinical development for acute lymphoblastic leukaemia.171 An altered biodistribution of CD33+ antigen expressing cells leading to reduced efficacy was discussed as one reason for the failure of 60: High loads of the CD33 antigen in acute myelogenous leukemia (AML) circulating in the blood stream resulted in an increased presence of bound immunoconjugate in the plasma compartment, leading to reduced levels of 60 in CD33+ cells in the bone marrow.172,175
The case of 60 demonstrates that for the development of safe and effective immunoconjugates, many parameters influencing pharmacokinetic and pharmacodynamics must be controlled. It was not until 2011 that the next antibody–drug conjugate was launched.176
Brentuximab vedotin 63 (Adcetris™) was approved for the treatment of Hodgkin and systemic anaplastic large cell lymphoma. It is based on a fully synthetic analog of dolastatin 10 61 linked to an anti-CD30 antibody. 61 was first reported in 1987 after its extremely challenging isolation from the sea hare Dolabella auricularia.177 It belonged to a family of very potent antineoplastic pentapeptid-like structures acting as tubulin polymerization inhibitors. Only two years later a convergent total synthesis of dolastatin 10 61 was published,178 which resolved any resupply issues and paved the way to combinatorial routes that allowed for the synthesis of thousands of analogs, the auristatins. This flexible synthesis route enabled a systematic fine-tuning of physicochemical properties, stability and potency of new derivatives aimed at conjugation to antibodies. A simplified dolastain 10 analog, monomethyl auristatin E (MMAE) 62 was the result of this work. MMAE 62 is characterized by a built-in site for stable linker attachment (a secondary amine) and displayed high potency, solubility and stability under physiological conditions. Brentuximab vedotin 63 was produced by conjugating MMAE 62 to the chimeric anti-CD30 monoclonal antibody cAC10 through a protease-cleavable linker containing a valine-citrulline dipeptide motif.179 This dipeptide sequence was found to be stable in plasma, but easily cleaved by cathepsin B, a lysosomal protease encountered after binding to the CD30 antigen and internalization by clathrin-mediated endocytosis. After proteolytic cleavage, 62 is released into the cytosol and inhibits microtubule polymerization, leading to cell cycle arrest at the G2-M transition phase and ultimately to apoptosis in CD30-expressing lymphoma cells.180
Very recently, the first antibody drug conjugate acting on solid tumors successfully completed clinical development: ado-trastuzumab emtansine 64 (Kadcyla™) was approved in February 2013 by the FDA for use as a single agent for the treatment of patients with human epidermal growth factor receptor 2 (HER-2) positive metastatic breast cancer that had undergone prior therapy. 64 consists of the maytansinoid DM-1 68 attached to a recombinant humanized monoclonal HER2-binding antibody, trastuzumab (Herceptin™).181,182 The naked antibody is well established in the treatment of HER2-overexpressing breast cancer.183
The toxic payload DM-1 68, a semisynthetic derivative of the maytansinoid ansamitocin P-3 66, was attached through a non-reducible thioether linker.181,182 The eponymous first representative of this highly cytotoxic class of compounds, maytansine 65, was isolated from the Ethiopian shrub Maytenus serrata by means of bioassay-guided fractionation. Other congeners have not only been isolated from plants, but also from an Actinomycete and from mosses.18465 displayed a very potent activity on several carcinoma and leukemic cell lines. It bound directly to tubulin in a mode similar to the Vinca alkaloids and thus interfered with the formation of the mitotic apparatus at cell division.185 As also observed for other tubulin-binding anti-cancer agents, this interaction led to arrest in the G2-M phase of the cell cycle and subsequently to apoptosis.186 Early clinical trials with 65 were initiated in the 1970s. However, dose-limiting toxicity and lack of response in the majority of patients enrolled in phase II trials led to the discontinuation of 65 as a single agent for anti-cancer therapy. From the early 1990s on, the maytansinoids were re-investigated for their use as toxic payloads for antibodies. 66 was chosen as the starting material for semisynthesis, since it was readily available from fermentation. DM1 68 and other related maytansinoids suitable for conjugation (such as DM4 69) were synthesized from maytansinol 67, which is obtained by reductive cleavage of 66 at the acyloxy function at C3 with LiAlH(OMe)3.187 Re-esterification of maytansinol with several carboxylic acids mediated by DCC/ZnCl2 led to new derivatives displaying cytotoxic activity in the 10–90 pM range, which corresponded to an ideal potency for the use as toxic payloads in antibody conjugates.188
DM1 68 and DM4 69 were combined with various antibodies and linker types, resulting in several new investigational immunoconjugates that have entered clinical trials. The successful development of 65 as the first approved frontrunner of the maytansinoid immunoconjugates may be due to a relatively stable linker and its “dual” mode of action: 65 contains a non-reducible thioether linker, which was cleaved relatively slowly by proteases in the endosome of tumor cells after internalization, resulting in a longer-lasting active metabolite. The linker was highly stable in the extracellular environment and exhibited a relatively low clearance of the immunoconjugate (in contrast to conjugates with reducible disulfide linkers). In addition to the delivery of the toxic payload, ado-trastuzumab emtansine 65 still displayed the properties of the naked antibody, namely inhibition of HER-2 signaling, the therapeutic principle of trastuzumab (Herceptin™), and induced antibody-dependent cellular toxicity.181,189,212
In the past 3 years, impressive clinical results provided the proof-of-concept for the immunoconjugate approach. They triggered intense R&D activities that has already led to more than 30 clinical development candidates as of May 2013.190 The fact that most of them employ maytansinoids or 62 demonstrates that the establishment of a single natural product-based effector platform can enable the harvesting of multiple clinical drug candidates. It also suggests that the exploration of additional effector principles is a highly rewarding field for future natural product research.
The first two marketed marine natural product drugs ziconotide (Prialt™, Elan Pharmaceuticals) and trabectedin (Yondelis™, PharmaMar)191 did not have a sustainable biological production source; consequently, chemists have developed a viable access, by total- and semisynthesis, to the products (chapters 5.1 and 5.2, respectively). In order to become independent from the fluctuations of horticulture for the supply of the antimalarial product artemisinin, a combination of molecular biology, creating a novel producer strain for a central drug intermediate, with organic synthesis has been realized (chapter 5.3).
As the animal sources provided only minute quantities of peptides, total synthesis approaches for the conotoxins were required to ensure a sustainable supply for profiling. A first solid-phase peptide synthesis of 70 was accomplished in 1987 on p-methylbenzhydrylamine resin using the Merrifield protocol. Protection of side-chain amino acid functions was realized by standard methods for the Boc/Bzl strategy, in the case of the cysteine residues as p-methoxy benzyl (Mob) or N-acetyl aminomethyl (Acm) derivatives. After building the ω-conotoxin MVIIA sequence on the resin, the completely deprotected linear peptide with the 6 Cys residues in reduced sulfhydryl form was obtained following cleavage with hydrogen fluoride-based cocktails. Formation of the disulfide bond cross-linking in the correct connectivity was achieved by simple air oxidation at neutral pH in aqueous solution to give correctly folded 70. The yields of the folding step could be improved by the use of additives, such as cysteine and DTT.
In further studies, N-type voltage-sensitive calcium channel blocking was identified as the mode of action of 70.194 Development of the compound as an antinociceptive agent was initiated through its high effectiveness in a rat model of neuropathic pain.195 In 2005, ziconotide was launched in the USA for the treatment of severe chronic pain. Due to a therapeutic scheme with doses as low as 2.4–21.6 μg day−1,196 the multistep total synthesis on solid phase remains an economically viable supply of the drug substance.
Although the establishment of aquacultures of E. turbinata for the production of 71 proved to be very challenging and required considerable efforts, this approach has provided the batches required for clinical trials.203,204 It was nevertheless considered necessary to establish a sustainable, long-term drug supply capable of serving the needs for commercialization of the drug. A breakthrough was finally achieved by the development of a semisynthetic process starting from the readily available fermentation product cyanosafracin B 72, which can be obtained from Pseudomonas fluorescens in multi-kilogram quantities.205 As shown in Scheme 4, 72 represents a well-suited scaffold for the construction of the more complex 71 and its analogs, since its structure is comprised of two of the three fused tetrahydroisoquinoline rings present in the ecteinascidins and displays a favourable arrangement of functional groups. The transformation into 71 can be accomplished in 20 steps with a reported total yield of 1.14%.
Activity of A. annua extracts against Plasmodium growth in a mouse model was discovered in the early 1970's by Chinese scientists through the systematic investigation of more than 640 traditional herb preparations that might have antimalarial activity.208 In 1972, artemisinin 73 (qinghaosu) was isolated209 and structurally characterized210 as the active principle.
In China, artemisinin has been used for the treatment of resistant Plasmodium falciparum malaria, but its therapeutic value is limited by its low solubility both in water (around 60 μg mL−1) and in lipidic formulations.211 Artemisinin itself has a very short half-life in humans and is rapidly converted mainly to the active metabolite dihydroartemisinin 74.212 In order to improve the pharmaceutical and pharmacokinetic properties of artemisinin, several semisynthetic derivatives, such as artemether 75, arteether 76 and artesunate 77 were developed. All of them are accessible through 74 obtained by reduction of 73, and all of them are – as 73 – converted back to 74 as the active metabolite in humans.213 As artemisinin and its derivatives are the most effective antimalarials currently available,214 the use of 74, 75 or 77 in combination with other antimalarials in so-called “artemisinin-based combination therapies (ACTs)” is recommended by the WHO in their guidelines for the treatment of malaria.215 Whereas 75 and 76 have an improved solubility in lipidic formulations compared with 73,21677 displays a high water solubility (around 0.6 mg mL−1) due to its carboxylic acid function.212 It is indispensable for the successful treatment of severe P. falciparum infections in areas of low malaria transmission where high doses of an artemisinin-derived drug have to be administered intravenously.213,217
The first total synthesis of artemisinin starting from the chiral pool building block (−)-isopulegone was reported in 1983.218 Although several other improved total syntheses were published,219 the approach remains too expensive for commercialization. Today artemisinin is still produced from Artemisia annua cultivated in East Asia (mainly China and Vietnam) and to a lesser extent in Africa.220
The typical yields of artemisinin based on dried plant material are reported to be in the range of 0.8%. Like for other non-regulated commercial crops, the market for artemisinin experiences strong fluctuations: from 2003–2009 the price ranged between US$ 200–1100 per kilogram and the produced quantities have not always met the demand.213,219,221 Notable improvements of the production in plants have been made by conventional breeding and selection of high artemisinin genotypes; other attempts using plant cell and hairy root cultures had limited success so far.222
The biosynthesis of the isoprenoid 73 proceeds – like with almost all sesquiterpenes – via the cyclisation of farnesyl diphosphate 78 (FPP) generated in the mevalonate pathway (Scheme 5). Detailed knowledge of this pathway enabled metabolic engineering approaches in Artemisia annua.223 In one study, overexpression of endogenous farnesyl diphosphate synthase led to an increase of the artemisinin yield from 0.65% to 0.9% (based on dried plant material), indicating that the optimization of a single gene might not be sufficient for a multifold increase of yield.224
In different studies, a microbial production system for the biosynthetic intermediates amorphadiene 79 and artemisinic acid 82, followed by chemical conversion to 73, was explored as an alternative.225 The technology for this approach was developed in the Keasling laboratory.226 Initially, it was based on the biosynthesis of 79 in E. coli via heterologous expression of the mevalonate pathway. This pathway was split into two operons, the “upstream” operon MevT that converts Acetyl-CoA to mevalonate, and the “downstream” operon MBIS, transforming mevalonate further to FPP 78. Introduction of plasmids carrying these operons together with a plasmid carrying the gene for amorphadiene synthetase (ADS) into E. coli DH1 resulted in a strain that produced amorpha-4,11-diene 79 at an average titer of 27.4 g L−1 after careful optimization of the fermentation process.225,227 Ideally, a biotechnological process should comprise most of the biosynthesis pathway to the target product – in this case, artemisinic acid 82 – as the final result of the cascade of enzymatically controlled chemical transformations. However, CYP activities high enough for an efficient oxidation of amorpha-4,11-diene 79 to artemisinic acid 82 could not be realized in E. coli. As a consequence, the heterologous expression system was switched to baker's yeast. As outlined in Scheme 5, the artemisinic acid-producing strain Saccharomyces cerevisae EPY224 was created by combining an optimized endogenous pathway to 78 with the heterologous expression of the A. annua genes amorphadiene synthase (ADS), CYP71AV1 and its redox partner CPR (which oxidized 79 to 82).228 Key for a successful increase of product titers was to suppress carbon flux through competing pathways. In particular, high levels of 78 for the conversion to 79 were ensured by down-regulation of the gene ERG9 encoding for squalene synthetase, a pathway that led to depletion of 78. After careful optimization, a defined fed-batch process with titers of 25g L−1 of artemisinic acid 82 was obtained that was ready for industrialization.229 The heterologous expression process, seen as a prime example for a rising discipline coined ‘synthetic biology’, was followed by the downstream chemical conversion of 82 to 73. It can be achieved in 2 steps by photo-oxidation of dihydroartemisinic acid 83 obtained by reduction of 82.230 A scaleable process was established in a 3-step sequence via the reaction of 84, a mixed anhydride of 83, with singlet oxygen that has been generated by irradiation of the acidic reaction mixture in the presence of catalytic amounts of tetraphenyl porphyrin (Scheme 6 and Fig. 7).231,232 The sequence gave less by-products than the direct oxidation of 83, thereby completing the fully industrialized process for the production of 73. With 50–60 tons of 73 planned for 2014, the biotechnological and chemical process is supposed to complement the plant-derived processes, thereby minimizing fluctuations in supply and prices for this essential class of anti-malaria drugs.233 For orientation: the global demand for artemisinin in 2013 was estimated to be 180–200 tons during the Artemisinin Conference 2013 (January 15–16, Nairobi, Kenya).
Scheme 6 The scalable synthesis of artemisinin 73 from dihydro-artemisinic acid 83via photo-oxidation of mixed anhydrides. |
Recently, a continuous-flow process for the conversion of 83 into 73 was published by the Seeberger group,234 which might further reduce the costs for artemisinin supply, if it can be demonstrated ultimately that the process is amenable to an industrial scale.235
As outlined in this review, sophisticated total synthesis has to be considered as a realistic option for many, even highly complex structures. Significant contributions to natural product development are to be expected from genetic engineering and synthetic biology approaches; those can also be favorably combined or tightly entwined with synthetic steps. Novel, sophisticated delivery systems, such as antibody–drug conjugates, overcoming intrinsic toxicity and pharmacokinetic liabilities of valuable natural compounds, have recently provided a convincing clinical proof-of-concept and await a much broader and systematic exploration.
In summary, a broad and growing arsenal of methods is available today to advance natural products from leads to drugs, with no generic recipe for success and consequently no straightforward prediction of project duration and risk. This characteristic is regarded as a clear drawback from the perspective of planning-oriented operations, and at the same time as an attractive, adventurous challenge from an exploratory research perspective.
Footnotes |
† New address: Helmholtz Centre for Infection Research, Inhoffenstraße 7, 38124 Braunschweig. |
‡ Absorption, distribution, metabolism, excretion, toxicology |
This journal is © The Royal Society of Chemistry 2014 |