|Home | About | Journals | Submit | Contact Us | Français|
Full details of the development of a direct coupling of catharanthine with vindoline to provide vinblastine are described along with key mechanistic and labeling studies. Following an Fe(III)-promoted coupling reaction initiated by generation of a presumed catharanthine radical cation that undergoes a subsequent oxidative fragmentation and diastereoselective coupling with vindoline, addition of the resulting reaction mixture to an Fe(III)–NaBH4/air solution leads to oxidation of the C15′–C20′ double bond and reduction of the intermediate iminium ion directly providing vinblastine (40–43%) and leurosidine (20–23%), its naturally occurring C20′ alcohol isomer. The yield of coupled products, which exclusively possess the natural C16′ stereochemistry, approaches or exceeds 80% and the combined yield of the isomeric C20′ alcohols is >60%. Preliminary studies of Fe(III)–NaBH4/air oxidation reaction illustrate a generalizable trisubstituted olefin scope, identified alternatives to O2 trap at the oxidized carbon, provides a unique entry into C20′ functionalized vinblastines, and affords initial insights into the observed C20′ diastereoselectivity. The first disclosure of the use of exo-catharanthine proceeding through Δ19′,20′-anhydrovinblastine in such coupling reactions is also detailed with identical stereochemical consequences. Incorporating either a catharanthine N-methyl group or a vindoline N-formyl group precludes Fe(III)-promoted coupling, whereas the removal of the potentially key C16 methoxy group of vindoline does not adversely impact the coupling efficiency. Extension of these studies provided a total synthesis of vincristine (2) via N-desmethylvinblastine (36, also a natural product), 16-desmethoxyvinblastine (44) and 4-desacetoxy-16-desmethoxyvinblastine (47) both of which we can now suggest are likely natural products produced by C. roseus, desacetylvinblastine (62) and 4-desacetoxyvinblastine (59), as well as a series of key analogues bearing systematic modifications in the vindoline subunit. Their biological evaluation provided additional insights into the key functionality within the vindoline subunit contributing to the activity and sets the foundation on which further, more deep-seated changes in the structures of 1 and 2 will be explored in future studies.
Vinblastine (1)1 and vincristine (2) are the most widely recognized members of the class of bisindole alkaloids as a result of their clinical use as antitumor drugs (Figure 1, Velban® or Velbe® and Oncovin®, respectively). Originally they were isolated in trace quantities (0.00025% of dry leaf weight for vinblastine) from the leaves of Catharanthus roseus (L.) G. Don,2 and their biological properties were among the first to be shown to arise from inhibition of microtubule formation and mitosis that today is still regarded as one of the more successful targets for cancer therapeutic intervention.3
In addition to being among the first natural products whose structures were determined by X-ray crystallography, they were also among the first for which X-ray analysis of a heavy atom derivative was used to establish their absolute configuration.4 Both vinblastine and vincristine possess the identical velbanamine upper subunit and nearly identical vindoline-derived lower subunits differing only in the dihydroindole N-substituent. Despite this small structural difference, vinblastine and vincristine differ in their antitumor properties and dose-limiting toxicities.1,3 Two additional semisynthetic vinca alkaloids, vindesine (3, Eldesine®)5 and vinorelbine (4, Navelbine®)6 have been developed as antitumor drugs and a third, vinflunine (5),7 is in late-stage clinical trials in Europe. Notably, the latter two derivatives incorporate a ring-contracted norvelbanamine upper subunit.
Vincristine (2) is used in combination therapy to treat acute leukemias and lymphomas and constitutes an important component of the regime that has been so successful in treating childhood leukemias. Vinblastine (1) is often used in combination to treat bladder and breast cancers and is an integral part of the curative treatment regime for Hodgkin's disease. Vinorelbine was approved for use in Europe (1991) and the US (1995) for the treatment of non small-cell lung cancer and vindesine has been approved for the treatment of melanoma. Neurotoxicity (vincristine) or myelosuppression (vinblastine) are the main side effects of administration and neutropenia is the principal dose-limiting toxicity of the vinca alkaloids, but recovery occurs following treatment. However, the major limitation to the continued use of the vinca alkaloids is the emergence of drug resistance derived principally from overexpression of phosphoglycoprotein (Pgp), an efflux pump that transports many of the major drugs out of the cell. In fact, vinblastine represents one of the most studied prototypical substrates for Pgp efflux responsible for multidrug resistance (MDR). Thus, in addition to identifying vinblastine and vincristine analogues that may address the current dose-limiting toxicities, the development of a modified vinca alkaloid that is not a substrate for Pgp efflux and is efficacious against MDR tumors would constitute a major advance. Additionally, the emerging evidence that the vinca alkaloids also possess antiangiogenic activity that may contribute to their in vivo antitumor activity, especially in combination with other drugs, may provide additional future clinical applications.8
Due to the pharmaceutical importance and low natural abundance of vinblastine and vincristine, C. roseus has become one of the most extensively studied medicinal plants serving as a model for biotechnological studies of plant secondary metabolism. Their biosynthesis involves the participation of at least 35 intermediates, 30 enzymes, 30 biosynthetic and 2 regulatory genes and 7 intra- and intercellular compartments.9 Presently, the clinical supplies of 1 and related drugs are derived from natural sources. Fortunately, the doses are so small that the production amounts are manageable even with the trace natural abundance of 1 (0.01%) or 2 (0.0003%) in the source plants. Nonetheless, the effort required even for this limited quantity suggests that an efficient synthetic approach might provide a viable alternative. Even the development of an effective coupling protocol starting with the more abundant naturally occurring (+)-catharanthine (6)10 and (−)-vindoline (7)2,10 may supplant the direct use of plant produced vinblastine or vincristine. Interestingly, only C. roseus produces catharanthine and does so with an absolute configuration enantiomeric with structurally related alkaloids also found in C. roseus and related alkaloids found in nature. More significantly, an effective synthetic approach would provide access to analogues that incorporate deep-seated structural changes that have not yet been explored.3,11 Typically, it has been semisynthetic derivatives of the natural products that have been examined, restricting the structural sites and opportunities to improve on the properties of 1 or 2.
As a consequence, a number of pioneering studies have defined methods for coupling the lower half, vindoline (7), with appropriate precursors to the upper velbanamine subunit. These include the seminal Potier12 and Kutney13 disclosures of a coupling protocol enlisting a Polonovski reaction of catharanthine N-oxide (8) in which its embedded olefin controls the regioselectivity and coupling efficiency of the resulting iminium ion and necessarily provides anhydrovinblastine (9), Scheme 1. Conducting the reaction at low temperature was found to improve the C16′ coupling diastereoselectivity (>5:1 at −78 °C vs 1:1 at 0 °C),11 and the subsequent conversion of anhydrovinblastine to vinblastine was addressed by conversion to14 or with direct generation15 of the enamine 11 that in turn was oxidized to the C20′ alcohol. This indirect conversion of anhydrovinblastine to vinblastine via the enamine was developed as a result of the preferential α versus β face delivery of reagents to the Δ15′,20′ -double bond and the competitive reactivity of 9 toward electrophilic reagents required of most olefin oxidation methods. The resulting overall conversions, requiring 814 or 515 steps, range from 10–40%.
Alternative approaches enlisting chloroindolenine intermediates derived from indole C3 electrophilic chlorination of precursors to the velbanamine subunit were slower to develop. Following the disclosures that carbomethoxycleaveamine (12) couples with vindoline to provide the epimeric C16′ diastereomer 13,12,16 both Magnus17 and Kuehne–Bornmann18 described protocols that predominantly17 or exclusively18 provide the correct C16′ diastereomer, Scheme 2. These enlist velbanamine precursors that proceed through larger indole-fused ring systems or those lacking the velbanamine piperidine requiring post coupling assemblage of the intact upper subunit. The most effective of these approaches detailed by Kuehne18 requires 4 steps post coupling that proceeded in ca. 60% overall yield. Most recently and based on key observations of Fritz,19 Fukuyama disclosed a diastereoselective coupling of an even more advanced and larger ring velbanamine precursor 14 incorporating the C20′ alcohol permitting access to 1 in 4 steps and ca. 50% overall yield.20
Herein, we report the use of a single-step biomimetic coupling of catharanthine (6) and vindoline (7) to directly provide vinblastine in yields competitive with the best of the past protocols. Not only was this used to extend our 11-step total synthesis of natural (−)- and ent-(+)-vindoline21 to a 12-step total synthesis of natural (+)- and ent-(−)-vinblastine, but we also report the application of this modified coupling protocol to the total synthesis of vincristine (2) and a systematic series of vinblastine analogues, many of which constitute even more minor naturally occurring constituents of the periwinkle C. roseus than 1 itself. These analogues constitute a key series in which a single feature in the lower vindoline subunit has been altered and the results of their examination provide the foundation on which our future studies will be based. In the course of these studies, we conducted detailed mechanistic and labeling studies of the Fe(III)-coupling reaction responsible for formation of the critical C16′ stereocenter as well as the subsequent low valent iron-mediated oxidation reaction utilized for C20′ hydroxylation. We also report initial studies on the extension of the oxidation protocol for the introduction of alternative C20′ functionality.
With the completion of a first generation total synthesis of vindoline that was extended to a series of related analogues,21 we have begun examining protocols for their incorporation into vinblastine and its analogues. In these studies, we found that modification of the direct Fe(III)-promoted biomimetic coupling between catharanthine and vindoline first described by Kutney22 to produce anhydrovinblastine (9) and modification of a subsequent Fe(III)-based oxidation procedure described by Sakamoto23 can provide vinblastine in a single step in conversions competitive with those of the multistep protocols.24 Our investigations began with an examination of the Kutney coupling protocol for the generation of anhydrovinblastine.
Anhydrovinblastine (9) is itself a naturally occurring vinca alkaloid first demonstrated in biosynthetic labeling studies25 in C. roseus and later by its extraction from the plant.26 Not only is it an important intermediate in the synthesis of the natural vinca alkaloids but, as the most accessible product from the coupling of vindoline and catharanthine, it is also the semisynthetic precursor to the two non-naturally occurring drugs vinorelbine (4) and vinflunine (5). Using improved conditions for a coupling first disclosed by Kutney,22 treatment of a mixture of catharanthine (6, 0.022 M) and vindoline (7, 0.022 M) with FeCl3 (5 equiv, 23 °C), presumably generating the catharanthine amine radical cation which undergoes a subsequent oxidative fragmentation, leads to biomimetic coupling providing the iminium ion 10 exclusively possessing the natural C16′ stereochemistry, Scheme 3. Reduction with NaBH4 produces anhydrovinblastine (9) in superb conversion (90%) provided CF3CH2OH, which solubilizes the reactants,24 is used as a cosolvent with the aqueous 0.1 N HCl reaction solution.27 Although this cosolvent effect provides what might appear to be only a modest improvement for vindoline itself (90% vs 77%),22,27 this modification is much more significant with a representative less soluble vindoline analogue 17 that was examined at the same time, Scheme 3. The use of MeOH as a cosolvent provided a homogeneous reaction solution and improved the conversions, but it also provided the MeOH addition product with catharanthine and appeared to promote the self-coupling of 7 or 17 arising from their oxidation. Replacing MeOH with the non nucleophilic cosolvent CF3CH2OH provided a homogeneous reaction mixture, superb conversions with both vindoline (90%) and 17 (90%), and little or no self coupling products derived from either 7 or 17. Finally, conducting the reaction under Ar in degassed solvents appeared to subtly improve the conversions (by 5–10%) although reactions conducted without such precautions still proceed superbly.
The additional feature of this reaction that we wish to report is that Fe2(SO4)3 (5 equiv) also effects the coupling of vindoline with catharanthine to provide anhydrovinblastine in good yields (71%) under the same conditions. Although this unoptimized conversion is not yet as superb as that observed with FeCl3, it represents an Fe(III) reagent that, unlike FeCl3, effectively supports the subsequent oxidation of anhydrovinblastine to vinblastine without further additives. Moreover and unlike Fe2(ox)3 (below), which supports the anhydrovinblastine oxidation but not the vindoline/catharanthine coupling, Fe2(SO4)3 is an iron reagent capable of supporting both the initial coupling and subsequent oxidation.
As first observed by Sundberg,27b N-methylcatharanthine (19) fails to couple with vindoline upon treatment with FeCl3 even under our modified reaction conditions, equation 1. Although Sundberg observed that 19 is consumed (but not coupled) with vindoline present, and is recovered unchanged in absence of vindoline, we observed the consumption of 19 whether vindoline was present or not. Analogous observations have been made with Potier's Polonovski reaction with N-methyl catharanthine N-oxide that also failed to initiate coupling with vindoline leading simply to fragmentation.28 The implication of this requirement for a free indole NH has been discussed in detail by others,28 and suggests azabenzfulvene participation in the catharanthine fragmentation and its subsequent coupling with vindoline. However and unlike the Polonovski-based coupling protocol of Potier, the Fe(III)-promoted catharanthine oxidative fragmentation proceeds with exclusive formation of the natural C16′ stereochemistry even at 25 °C, which corresponds to clean inversion of the stereochemistry at the reacting C16′ center of the C16′–C21′ bond undergoing cleavage. Although at odds with invoking an azabenzfulvene intermediate and not offered as a mechanistic rationale, this stereochemical outcome follows expectations of vindoline attack at C16′ concurrent with fragmentation.
Although oxidation of the Δ15′,20′ -double bond of anhydrovinblastine has been observed and explored with a full range of oxidants including air (O2),29,30 none do so preferentially from the β-face. Most such efforts provide the corresponding α-epoxide (leurosine), and none provide vinblastine or its naturally occurring C20′ alcohol isomer leurosidine (20) in appreciable amounts. The exception to this is the oxidation protocol of Sakamoto.23 This oxidation entails the direct Fe-mediated (FeCl3, 30–500 equiv) conversion of anhydrovinblastine (9) to vinblastine and leurosidine in the presence of air (O2) at 0–25 °C in aqueous buffer containing carboxylic acid additives (e.g. ammonium oxalate) upon addition of a reducing agent including NaBH4 (30 min). As a prelude to our more detailed studies combining an Fe(III)-promoted oxidation of anhydrovinblastine with the Kutney Fe(III)-mediated coupling of vindoline and catharanthine, we briefly examined this reaction, Scheme 4. Simply submitting anhydrovinblastine (9) to the oxidation conditions (30 equiv FeCl3, air, 0 °C, 1 h; 20 equiv NaBH4) in the Kutney 0.1 N HCl–glycine buffer only provided trace amounts of vinblastine. By using ammonium oxalate (60–120 equiv) as an additive, oxidation with FeCl3 (30–60 equiv)/NaBH4 (20–50 equiv) in our 0.1 N HCl–CF3CH2OH solvent system at 0 °C under progressively dilute conditions (0.22 mM) provided improved conversions (ca. 35–40%) being derived in part from the increasing O2 content that is limited by its solubility (concentration). Finally, by utilizing Fe2(ox)3 under heterogeneous (30 equiv) or homogeneous (10 equiv, see below) conditions in place of the FeCl3–additive combinations, superb conversion to vinblastine (50%) and leurosidine (20%) was achieved. Notably, no reaction is observed in the absence of the initiating reductant (NaBH4). Further, NaBH4 proved superior to the limited number of alternative reagents examined (e.g. NaCNBH3, LiBH4, NaBH(OAc)3, BH3, Bu3SnH), olefin reduction is observed in the absence of air (O2), other conventional oxidants (e.g. H2O2, t-BuOOH) failed to support the reaction, and the reaction proceeds directly from 9 to 1/20 without the intermediacy of the isomerized enamine 1131 or oxidized iminium ion 1032 (see labeling studies). Moreover, and as detailed in subsequent studies, this trisubstituted olefin oxidation reaction (C20′-OH derived from O2) is not unique to anhydrovinblastine and does not require the allylic amine. Interestingly and although not investigated in detail, the seemingly analogous Mukaiyama Co(II)-mediated olefin oxidation (Co(acac)2, O2, PhSiH3) failed to provide either vinblastine or leurosidine from 9.33
The preceding studies set the stage for combining the initial Fe(III)-promoted vindoline/catharanthine coupling reaction with the subsequent Fe(III)-based oxidation reaction to directly provide vinblastine (1) in a single step, Scheme 5.23 Thus, FeCl3 (5 equiv) mediated coupling of vinblastine with catharanthine using our solvent modification on the Kutney procedure (0.1 N HCl– CF3CH2OH) without reductive workup is followed by addition of the reaction mixture containing the iminium ion 10 to a second Fe(III) solution (soluble Fe2(ox)3, 10 equiv) cooled to 0 °C and saturated with air. Subsequent addition of NaBH4 (20 equiv) initiates both reduction of the intermediate iminium ion and selective oxidation of the Δ15′,20′-double bond with installation of the C20′ alcohol to provide vinblastine (1, 41%), its naturally occurring C20′ alcohol isomer leurosidine2 (20, 21%), along with anhydrovinblastine (9, 10%). The yield of identifiable coupled material exceeds 80–90% with the combined yield of C20′ alcohols being 62–69% (2:1 β:α). This one step coupling reaction was conducted to provide natural (+)-vinblastine as well as with synthetic ent-(+)-vindoline21 and ent-(−)-catharanthine to provide ent-(−)-vinblastine ([α]D23 −38 (c 0.05, CHCl3)).
One of the most important features to emerge from the optimization efforts, which included a more detailed accounting of all characterizable products, was the effect of utilizing solublized Fe2(ox)3 in the second oxidation reaction. In the course of our early work and the optimization of the second stage of the reaction (Supporting Information Table S1), Fe2(ox)3 was found to be more effective than most iron salts for promoting the olefin oxidation with installation of the C20′ alcohol. Like the Sakamato protocol,23 the stoichiometry was high resulting in heterogeneous reaction conditions where not all the iron salt was soluble at the onset of the reaction (addition of NaBH4). Completely solubilizing the initial Fe2(ox)3 solution by stirring the suspended salt solution for an extended period (2 h) before the onset of the reaction (air bubbling for 10 min, then addition of NaBH4) resulted in a much more efficient and equally effective oxidation reaction, Figure 2. In fact, the use of solubilized Fe2(ox)3 (30 equiv) led to over oxidation of vinblastine, a result that was improved by reducing the amount of Fe2(ox)3 (10 equiv). Reoptimization of the amount of added NaBH4 to initiate the reaction (20 equiv) and a reexamination of the reaction concentration (0.2 mM) which insures an adequate supply of dissolved O2 provided a superb and highly reproducible single step coupling protocol.
Although we did not conduct an exhaustive survey of Fe(III) salts, both Fe(III) citrate and FeF3 (at 60 equiv, 40 equiv NaBH4) failed to support the oxidation reaction (both insoluble), Fe(NO3)3 provided only a modest level of oxidation, whereas Fe2(SO4)3 proved nearly as effective as Fe2(ox)3. In addition and because of its solubility properties (soluble in H2O–EtOH), the use of Fe2(SO4)3 permitted the exploration of additional cosolvents, Figure 3. Here, the use of even 1:1 H2O–EtOH did not shut down the oxidation reaction although the conversion to vinblastine was more modest (27 vs 40%). These observations indicate that studies of alternative Fe(III) reagents as well as the incorporation of additional reaction cosolvents, especially those that readily dissolve O2, might serve to further improve the overall conversions. Significantly, the Fe(II) reagent FeSO4 failed to support the oxidation reaction. Additionally, small improvements in the yield (lutidine: 43% 1; DBU: 44% 1) or diastereoselectivity (2,2′-bipyridine: 3:1 1:20) have been observed when the oxidation with Fe2(ox)3 is run in the presence of an organic base suggesting further optimization may still be possible (Supporting Information Table S2).
Finally, we were able to demonstrate that a single Fe(III) reagent (Fe2(SO4)3) is capable of supporting both the initial coupling and subsequent oxidation. Although limited optimization efforts were conducted revealing that systematic reductions in the amount of Fe2(ox)3 and NaBH4 led to progressively lower conversions, our initial trials do suggest conditions likely could be developed that may entail simply sequential addition of air (10 min bubbling) and NaBH4 following the 2 h coupling reaction, Figure 4.
In order to clarify the steps involved in the coupling and oxidation conversions to vinblastine, labeling experiments (NaBD4, D2O, 18O2) were conducted to distinguish between otherwise attractive mechanistic possibilities, Scheme 6. By utilizing MS, and 1H and 2H NMR characterization of the products (Supporting Information), the number, site, and stereochemistry of deuterium incorporation obtained from use of NaBD4 in the coupling preparation of anhydrovinblastine indicate one D incorporation at α-C21′ consistent with the report of Kutney in his analysis of the Polonovski coupling,32a a clean key and diagnostic single D incorporation at α-C15′ in the oxidation of anhydrovinblastine to vinblastine, and clean D incorporation at each α-C15′ and α-C21′ in the combined coupling and oxidation reaction. In the oxidation of anhydrovinblastine to vinblastine, reactions run in D2O led to no deuterium incorporation, 18O2 labeling studies indicate the C20′ alcohol oxygen originates with O2 (oxidation reaction, >95% by MS) and not solvent water (hydration reaction), and reactions run in the absence of O2 lead to selective reduction of the Δ15′,20′-double bond providing C20′-deoxyvinblastine and its diastereomer C20′-deoxyleurosidine (1:1.5–2). This latter reduction diastereoselectivity (1:2) is reversed from that of the 2:1 diastereoselectivity observed in the oxidation reaction providing 1 and 20, and the use of NaBD4 indicates clean deuterium (two D) incorporation with one D at only α-C15′ and C20′. Although it is conceivable that the C20′ oxidation arises from 1,4-reduction of iminium ion 10 followed by enamine 11 oxidation, the labeling studies for the oxidation of anhydrovinblastine (9) to 1 (50%) and its C20′-isomer leurosidine (15–20%) under the Fe(III)–NaBH4/O2 conditions indicate that it can do so without the intermediacy of 10 indicating this is not necessary to achieve C20′ hydroxylation. Thus, the labeling studies (NaBD4) not only rule out the reoxidation to and intermediacy of iminium ion 10 in the conversion of anhydrovinblastine (9) to vinblastine, but they also rule out an Fe-catalyzed isomerization of 9 to enamine 11 and its resulting oxidation (no C21′ D incorporation). Additionally, reactions that should promote a 1,4-reduction of iminium ion 10 (e.g. NaCNBH3)14 result in subsequent enamine 11 protonation and reduction, not C20′ oxidation providing a mixture of anhydrovinblastine (9, 48% from 1,2-reduction) and 20′-deoxyleurosidine (31% from 1,4-reduction). Finally, subjection of either 20′-deoxyvinblastine or 20′-deoxyleurosidine to the oxidation conditions does not produce either vinblastine or leurosidine.
Without belaboring a discussion of all mechanistic possibilities, the labeling studies are most consistent with the oxidation of anhydrovinblastine to vinblastine by an Fe-mediated hydrogen atom radical addition to the trisubstituted Δ15′,20′-olefin initiated by NaBH4 treatment of an appropriate Fe(III) salt followed by reaction of the resulting carbon centered tertiary radical with O2 and subsequent reduction of the resulting hydroperoxide. In the one step coupling and subsequent oxidative conversion of vindoline and catharanthine to vinblastine, the NaBH4 serves to both reduce the iminium ion 10 to anhydrovinblastine (9) and as the controlled source of the Fe-promoted hydrogen atom radical generation for initiation of the olefin oxidation. Further supporting these interpretations are the observations below including the demonstration that the Fe(III)–NaBH4(air) oxidation reaction is not unique to substrates bearing the tertiary allylic amine. Finally, it is worth noting that the chemistry derived from the NaBH4 reduction of Fe(III) salts is complex,34 presumably distinct from the behavior of the well characterized hydridocarbonyliron complexes,35 and little precedent is available to draw mechanistic insights from. However, it is notable that the radical-mediated oxidation of an olefin (styrene) by a phthalocyanine Fe(III)–NaBH4 system in the presence of O2,36 Fe(II)- or Fe(III)-mediated alkene and alkyne reductions enlisting LiAlH4 in the absence of O2, and FeCl3–NaBH4 mediated sulfoxide and β-amino-enone reductions have been reported. 37
As part of the mechanistic probe of the Fe(III)– NaBH4(air) oxidation reaction, we examined the generality of the reaction with β-citronellol (21), a water soluble substrate bearing a trisubstituted double bond lacking the complexity of anhydrovinblastine and its tertiary allylic amine. Subjection of 21 to the oxidation conditions cleanly provided the tertiary alcohol 22 (68%) in good yield requiring less reagent (5 equiv of Fe2(ox)3, 6 equiv of NaBH4) than enlisted for the oxidation of anhydrovinblastine, and the use of NaBD4 provided the analogous single deuterium incorporation, Scheme 7. Inclusion of TEMPO (3 equiv) in the reaction mixture afforded the TEMPO adduct 23 in good yield (44%) along with 22 (44%), providing a superb combined yield of oxidation products (88%) and confirming that a radical trap is effectively incorporated at the oxidized carbon. Alternatively and without efforts at optimization, the inclusion of NaN338 (10 equiv) in a reaction run under Ar provided 24 (75%) in high yield whereas the use of NaNO239 (60 equiv) afforded the nitroso adduct 25 (41%). Thus, not only is the Fe(III)–NaBH4(air) oxidation reaction general for additional substrates containing a trisubstituted olefin, but alternatives to O2 may be used to functionalize the oxidized carbon.
These latter studies, in addition to providing key mechanistic insights into the anhydrovinblastine oxidation reaction, also introduce a unique opportunity for the synthesis of vinblastine analogues bearing modified C20′ functionality.40 To date, such studies have been limited to semisynthetic modifications of vinblastine itself and largely restricted to O-acylation (OH→OCOR), alcohol elimination and subsequent olefin reduction (OH→H) or acid/superacid catalyzed-additions including the reaction utilized to prepare the corresponding C20′ acetamido vinblastine analogue (OH→NHAc).41 Thus, starting with either anhydrovinblastine (9) or, more significantly, the coupling of vindoline and catharanthine followed by in situ generation and functionalization of 9, vinblastine analogues containing additional C20′ functionality may be accessed. As a demonstration of this potential and the scope of the Fe(III)–NaBH4 alkene addition reactions, the C20′ derivatives 26–29 bearing TEMPO, azido, and nitroso C20′ substituents were prepared, Scheme 8.
Impressively, these were generated starting from vindoline and catharanthine using the one pot coupling and subsequent in situ oxidation protocol even at a stage that the procedure was not yet optimized and directly provided 26–29. Notably, the TEMPO trap did not alter or improve the intrinsic 2:1 diastereoselectivity favoring the vinblastine versus leurosidine C20′ stereochemistry for an oxygen substituent, the azide of 28 was introduced with remarkable efficiency (47% overall) directly from vindoline/catharanthine even without optimization, and both the azide and nitroso groups were introduced providing exclusively the leurosidine versus vinblastine C20′ stereochemistry. In each instance, the C20′ stereochemical assignments were made based on diagnostic 1H NMR chemical shifts of three distinct signals: 3′-Hβ (br t), 6′-Hβ (dd), and 17′-Hβ (d). Each proton exhibits an unusual and distinct chemical shift and each is further downfield for the vinblastine versus leurosidine C20′ stereochemistry (Supporting Information Table S3). Concerned that this correlation, as strong as it appeared, might not prove accurate for establishing the C20′ stereochemistry with 28 and 29 where only a single isomer was observed, we unambiguously established the structure and stereochemistry for azide 28 by single-crystal X-ray analysis.40 Not only did this confirm the structure and assigned C20′ stereochemistry of 28, but it also served to reaffirm the C16′ stereochemistry set in the initial coupling reaction between vindoline and catharanthine. Finally and not surprisingly, azide 28 was also produced in a comparable conversion by treatment of anhydrovinblastine (9) with Fe2(ox)3–NaBH4 (10 equiv/30 equiv) in the presence of NaN3 (60 equiv). No doubt, many other C20′ functionalized derivatives may be accessed by using this approach and these will be investigated in due course.
As an alternative to coupling catharanthine with vindoline and in efforts to improve the diastereoselectivity of the C20′ alcohol introduction, we prepared and examined the substrate 32, Scheme 9. This catharanthine isomer incorporates an exocyclic versus endocyclic double bond, but it is still positioned to provide the necessary control of regioselectivity for the oxidative fragmentation of 32 and its radical cation required for coupling with vindoline. The question was would this impact the diastereoselectivity of the olefin oxidation reaction enlisted for direct introduction of the C20′ alcohol. Without optimization, substrate 32 was prepared from catharanthine (6) using our Fe2(ox)3/NaBH4(air) conditions for direct oxidation of the trisubstituted olefin providing a near 1:1 mixture of the tertiary alcohol 30 and the lactone 31 derived from the isomeric alcohol. Subsequent acid-catalyzed elimination of water provided the desired substrate 32 (46%) and additional lactone 31 (21%). The direct single step coupling of 32 with vindoline (7) with in situ oxidation of the C20′ center proceeded with a diastereoselectivity that was not distinguishable from the use of catharanthine itself (single natural C16′ diastereomer, 2:1 C20′ alcohol diastereomers) and in unoptimized conversions that approached (30–32% vinblastine, 16–20% leurosidine), but did not surpass those optimized for 6. Similarly, oxidation of the intermediate Δ19′,20′-anhydrovinblastine (33)41a provided the same 2:1 mixture of C20′ alcohol diastereomers when exposed to the Fe2(ox)3/NaBH4(air) oxidation conditions. Nonetheless and despite the disappointment that the use of 32 or 33 did not alter this 2:1 diastereoselectivity, their success provides synthetic access to vinblastine or its analogues through a coupling that has not been previously examined, provides an alternative synthesis of Δ19′,20′-anhydrovinblastine for semisynthetic modifications, and suggests that both 9 and 33 or 32 and 6 proceed through a common intermediate in the oxidation to provide vinblastine/leurosidine.
The remainder of our efforts focused on defining the scope of the Fe(III)-promoted coupling reaction and its extension to a series of key analogues bearing modifications in the lower vindoline subunit. The most significant of these is vincristine (2) bearing a N-formyl group in the lower subunit that necessarily impacts its nucleophilic reactivity in the coupling with the electrophilically activated catharanthine. Although it is well known that N-formylvindoline (30) fails to participate in the Polonovski reaction with catharanthine N-oxide (8),1 we are unaware of reports of its reactivity in the Fe(III)-promoted coupling reaction. Therefore, both 34 and a simplified model 37 (Supporting Information)42 were examined and both failed to couple with catharanthine under our modified coupling conditions, Scheme 10. Thus, an indirect approach enlisting the coupling of (–)-N-desmethylvindoline (35)21,43–45 with catharanthine (6) to first provide N-desmethylvinblastine (36), a natural product in its own right and useful naturally occurring precursor to vincristine,2 and its subsequent formylation was adopted. Thus, both 35 and 38 reacted effectively with catharanthine (6) under our one step protocol for both coupling and subsequent oxidation to directly provide 36 (42%) or 39 (35%) and their corresponding 20′ alcohol diastereomers (22% and 21%, respectively). Formylation of 36 and 39 (HCO2H, Ac2O, 23 °C, 2 h) provided vincristine (2, 94%) and 40 (70%) in good overall conversions.
We also viewed the coupling of (–)-vindorosine (41) with catharanthine (6) as an additional test of the scope of the Fe(III)-promoted coupling reaction and one that also provides the key vinblastine analogue 44 not yet examined. Not only would this further define the scope of the chemistry achievable using the methodology, but the subsequent evaluation of 44 would define the importance and potential role of the vinblastine C16 methoxy subustituent. Moreover and as a result of our earlier synthetic efforts that provided the natural product (–)-vindorosine46 and a range of naturally occurring (42)21 or synthetic (43) analogues,47 both 42 and 43 were also available to examine. The former provides access to 4-desacetoxy-16-desmethoxyvinblastine (47), which along with 44 we can now suggest likely constitute natural products produced by C. roseus awaiting isolation and characterization. Thus, vindorosine (41), 42, and 43 were subjected to the one step Fe(III)-promoted coupling reaction with catharanthine (6) and subsequent Fe2(ox)3–NaBH4(air) oxidation reaction to provide the corresponding C16-desmethoxyvinblastine and leurosidine analogues in the ca. 2:1 diastereoselectivity favoring the vinblastine C20′ stereochemistry along with 15–20% of the corresponding anhydrovinblastine analogues 46, 49, and 52, respectively (not shown), Scheme 11.
Thus, removal of the C16 methoxy substituent from vindoline or related substrates does not negatively impact the coupling reaction promoted by Fe(III). This is in contrast to observations made enlisting the Polonovski coupling conducted at 0 °C where the yield was substantially reduced (18% and 81% recovered vindorosine) and only the unnatural C16′ stereochemistry (R vs S) was observed.12b Significantly, 44 and the related bisindole alkaloids exhibited substantially altered spectroscopic (1H NMR) properties. Most notably, the 1H NMR is typically less crisp and the characteristic C17′ and C6′ signals and their chemical shifts diagnostic of the natural C16′ and C20′ stereochemistry for 1 (δ 4.51 and 4.04, respectively) move to higher field (< δ 3.80). These observations suggest that the conformation of 44–52 is substantially altered or, more precisely, that they may access many more available conformations consistent with expectations that the presence of the C16 methoxy group restricts rotation about the critical C15–C16′ bond linking the two halves of such molecules. Accordingly and as detailed herein, the removal of this C16 methoxy group results in a substantial, but not complete, loss in cytotoxic potency. Nonetheless, the observation of modest biological activity with the desmethoxyvinblastine derivatives would seem to confirmed that the coupling generates the natural C15–C16′ stereochemistry; an assignment we still regard as tentative as a result of the unusual optical rotations [46 (–), but 49 and 52 (+) vs 1 and related analogues (+)].
As a result of our prior efforts, a series of additional key analogues (53–58) of vindoline were prepared for incorporation into the corresponding vinblastine analogue, Scheme 12.21,48 Each constitutes the iterative removal of key substituents found in the vindoline subunit of vinblastine (e.g. C4-Ac, C4-OAc, C5-Et, C6–C7 double bond), at least two (59 and 62) constitute natural products in their own right,2 and their evaluations were expected to continue to help define the individual substituent roles. 4-Desacetylvinblastine (62) is the penultimate biosynthetic precursor to 1 and the precursor to a series of potent and efficacious semisynthetic C4 O-acyl vinblastine derivatives.49 4-Desacetoxyvinblastine (59) is its immediate biosynthetic precursor and constitutes an even more minor (10-fold) naturally occurring bisindole alkaloid than 1, which itself is found in trace amounts (≤ 0.00025% of dry leaf weight). It has been reported to possess equally efficacious, but less potent, antitumor activity compared to vinblastine and was not pursued due to its even lower natural abundance.5 Because these studies were conducted at an early stage of our efforts even before we had the opportunity to examine the single step coupling and subsequent oxidation reaction in detail, many of the preparations were conducted enlisting our original non optimal oxidation conditions (30 equiv FeCl3, 60 equiv H4NCOCONH4, air, 20 equiv NaBH4, 30 min; see Scheme 4) providing modest conversions by our present standards. Although not evident from the presentation above, the concise syntheses of many of the vindoline analogues [e.g. 56 (7-steps), 58 (6-steps), 55 (8-steps), 53 (10-steps)],21,46-48 their direct chromatographic resolution, and their single step incorporation into the corresponding natural product or vinblastine analogue provides a remarkably concise synthetic entry into this class of bisindole antitumor agents (<10 steps) that was inconceivable to us as we began our studies. Not only does this simplify the synthetic access to additional analogues for further study, but it also suggests that a totally synthetic analogue is economically well within reach for use in the clinic.
Sundberg has reported that the Fe(III)-promoted coupling of racemic catharanthine (2 equiv) with natural (–)-vindoline (1 equiv) provided recovered racemic catharanthine (1:1 mixture of enantiomers), indicating that the reaction proceeds without a kinetic diastereoselectivity (resolution) favoring the natural product.27b Since the two products were not characterized, we elected to additionally examine this reaction, but using (+)-catharanthine (1 equiv) and the readily available racemic vindoline derivative 56 (2 equiv) with characterization of the resulting anhydrovinblastine derivatives. Interestingly and perhaps remarkably, the Fe(III)-promoted coupling reaction provided 70 (natural, 29%) and 77 (unnatural, 51%) in a 1:1.75 ratio with both products possessing the correct C16′ stereochemistry, but slightly favoring the coupling of the unnatural enantiomer of 56, Scheme 13. Independently, the natural enantiomer of 56 (90%, Scheme 3) and the unnatural enantiomer of 56 (77%) were coupled with (+)-catharanthine providing 70 and 77, respectively, confirming the product assignments derived from the coupling of racemic 56.
The studies detailed herein provided a wide range of analogues of 1 and 2 bearing single point changes in their structures permitting an assessment of their contributions to the cellular activity of the natural products (cytotoxic activity against L1210 and HCT116). Additionally, the analogues were examined for their susceptibility to multidrug resistance (MDR), resulting from overexpression and drug efflux by Pgp using a well-established companion vinblastine resistant cell line (HCT116/VM46).50 The result of these studies are summarized in Figure 5.
Several of the natural products, their simple derivatives, or analogues have been previously reported,2,3b,11,14,15,18 but in many instances without establishing their in vitro cytotoxic activity. For those reported (1–3, 9, 20, 33), the activity compares well with the potency disclosed and many of the overall trends highlight those already recognized: vinblastine > anhydrovinblastine (typically 10-fold) > leurosidine (typically 100-fold) analogue and that the 20′-deoxy derivatives are quite potent. Examination of the analogues derived from the systematic removal of the key vindoline substituents [C4-Ac (equipotent), C4-OAc (10-fold), N1-Me (10-fold), C5-Et (10-fold), C16-OMe (100-fold), C6– C7 double bond (100-fold), C20′-OH (8-fold)] indicate that the C16-OMe and C6–C7 double bond appear to be more important than the C4 or N1 substituents and that the latter sites constitute useful positions for modification or synthetic simplifications (removal) that remain quite potent (e.g., 59, 62, 36, 71). However, the introduction of more polar functionality at C3 appears to dampen the loss in activity due to such changes (e.g., 3/6,7-dihydro-3, 4–7 fold vs 100-fold for 1/65), suggesting that even the more significant structural simplifications may be tolerated within the appropriate derivatives. The effects of the substituent removal typically are cumulative although they are dampened as the analogues become increasingly less potent. Additionally, the isomeric Δ19′,20′-anhydrovinblastine (33) proved to be less potent than anhydrovinblastine (9, 2–10 fold), the unique 20′-nitrosoleurosidine (29, IC50 = 40–50 nM) was found to be remarkably potent and the trends established for this series (20′-NO > 20′-H > 20′-N3 > 20′-TEMPO ≥ 20′-OH > 20′-NH2) indicate that this is a site rich for modification using the technology detailed herein, the leurosidine 20′-TEMPO derivative (27, 3–7 fold) was surprisingly potent and more active than the corresponding vinblastine derivative (26), and the unnatural enantiomers examined (1, 9, 20) were inactive. None of the analogues exhibited substantially improved activity against the resistant cell line HCT116/VM46 indicating they are subject to Pgp efflux although those bearing modification at C20′ (e.g., deoxyvinblastine, anhydrovinblastine) displayed a diminished loss of activity against the resistant cell lines. These observations set the foundation on which further, more deep-seated changes in the structures of 1 and 2 will be explored in future studies.
With the introduction of the intramolecular cascade [4+2]/[3+2] cycloaddition reaction of 1,3,4-oxadiazoles51 that is ideal for the assemblage of the functionalized pentacyclic ring system of vindoline and its simpler variants,21 its extension to the total synthesis of vinblastine, vincristine, related natural products, and key analogues were initiated. Herein we provide full details of the development of a direct coupling of catharanthine with vindoline to provide vinblastine along with key mechanistic and labeling studies. Following an Fe(III)-promoted coupling reaction initiated by generation of a presumed catharanthine radical cation that undergoes a subsequent oxidative fragmentation and diastereoselective coupling with vindoline, addition of the resulting reaction mixture to an Fe(III)–NaBH4/air solution led to oxidation of the C15′–C20′ double bond and reduction of the intermediate iminium ion directly providing vinblastine (40–43%) and leurosidine (20–23%). The yield of coupled products, which exclusively possess the natural C16′ stereochemistry, approached or exceeded 80% and the combined yield of the isomeric C20′ alcohols was >60%. Preliminary studies of Fe(III)–NaBH4/air oxidation reaction illustrate a generalizable trisubstituted olefin scope, identified alternatives to O2 trap at the oxidized carbon, provided a unique entry into C20′ functionalized vinblastines, and afforded initial insights into the observed C20′ diastereoselectivity. The first disclosure of the use of exo-catharanthine proceeding through Δ19′,20′-anhydrovinblastine in such coupling reactions was also detailed with identical stereochemical consequences. Incorporating either a catharanthine N-methyl group or a vindoline N-formyl group precluded Fe(III)-mediated coupling, whereas the removal of the C16 methoxy group of vindoline did not adversely impact the coupling efficiency. Extension of these studies provided a total synthesis of vincristine (2) via N-desmethylvinblastine (36, also a natural product), 16-desmethoxyvinblastine (44) and 4-desacetoxy-16-desmethoxyvinblastine (47) both of which we can now suggest are likely natural products produced by C. roseus, desacetylvinblastine (62) and 4-desacetoxyvinblastine (59), as well as a series of key analogues bearing systematic modifications in the vindoline subunit. Their biological evaluation provide additional insights into the key functionality within the vindoline subunit contributing to the activity and sets the foundation on which further, more deep-seated changes in the structures of 1 and 2 will be explored in future studies.
We gratefully acknowledge the financial support of the National Institute of Health (CA115526 and CA042056) and the Skaggs Institute for Chemical Biology. We are also especially grateful to Paul Hellier of Pierre Fabre for the generous gift of vindoline, vindorosine, and catharanthine.