|Home | About | Journals | Submit | Contact Us | Français|
The total syntheses of the PKC inhibitors (+)-calphostin D, (+)-phleichrome, cercosporin, and 10 novel perylenequinones are detailed. The highly convergent and flexible strategy developed employed an enantioselective oxidative biaryl coupling and a double cuprate epoxide opening, allowing the selective syntheses of all the possible stereoisomers in pure form. In addition, this strategy permitted rapid access to a broad range of analogs, including those not accessible from the natural products. These compounds provided a powerful means for evaluation of the perylenequinones structural features necessary to PKC activity. Simpler analogs were discovered with superior PKC inhibitory properties and superior photopotentiation in cancer cell lines relative to the more complex natural products.
Several years ago we initiated a program to explore the use of the Cu-catalyzed enantioselective oxidative biaryl coupling1 in the syntheses of enantiopure perylenequinones and demonstrated that systems possessing only helical stereochemistry can be configurationally stable.2 With the development of a key aldol cycloaddition utilizing a dynamic stereochemistry transfer, we further completed the first total synthesis of hypocrellin A (4; Figure 1).3 Prior to our efforts the total syntheses of the (−)- calphostins A–D (1a–d) and (+)-and (−)-phleichrome (2) were reported involving diastereoselective biaryl couplings.4 Unfortunately, these couplings afforded mixtures with the wrong diastereomer usually predominating; additional steps were required to establish the correct stereochemistry. Furthermore, cercosporin (3) with a bridging seven-membered ring remained a challenging synthetic target. Although the structurally related 1 and 2 are atropisomerically stable, the additional seven-membered ring in cercosporin lowers the atropisomerization barrier, allowing 3 to readily atropisomerize at 37 °C (Eq 1).5
Herein, we report the first total syntheses of (+)-1d and 3 exploiting a novel double cuprate epoxide opening. The combination of enantioselective oxidative biaryl coupling, double cuprate epoxide opening, and decarboxylative functionalization provides a potentially general means for constructing a diverse array of perylenequinone analogs with complete control of the helical and centrochiral stereochemical elements. As a result, we describe the synthesis of 10 new perylenequinone analogs as well as (+)-1d, (+)-2, and 3 from a common chiral binaphthyl precursor that can be generated readily in multigram batches. The routes to these new compounds are discussed with respect to chemical efficiency and stereochemistry. Among these new compounds, we identified several with longer wavelengths of absorption, potentially leading to superior photosensitizers. We also report IC50 values for all of the analogs against protein kinase C (PKC) establishing which elements are the most crucial to inhibition of the regulatory domain. Finally, we report CC50 values for selected analogs against cancer cell lines.
The perylenequinone family of natural products (Figure 1) is characterized by a helical chiral extended oxidized pentacyclic core combined with C7,C7′-substitution containing centrochiral stereocenters.6 The perylenequinone portion confers several novel features to these compounds including tautomeric forms that rapidly interconvert, low barriers to atropisomerization, and low barriers to photoexcitation.
Calphostin D, (−)-1d, and phleichrome, (−)-2, are isolates of the Cladosporium fungi – Cladosporium cladosporioides and Cladosporium phlei, respectively.7,8 Cercosporin, 3, was first isolated by Kuyama and Tamura from Cercospora kikuchii,9 a fungus responsible for the “purple speck disease” of soy beans as well as damage to a host of plant species worldwide.10 The structures were elucidated through a series of chemical transformations and subsequent spectroscopic analysis.7,8,11,12 The X-ray structure of 3 was used to assign the absolute and relative configuration.5 Correlation of the CD and NMR spectra allowed determination of the helical configurations and relative stereochemistries of the remaining natural products. Related natural products, hypocrellin A (4),13 hypocrellin (ent-4),14 elsinochromes (5),15 and scutiaquinone (6)16 have also been isolated from several different fungi.
From a synthetic perspective, 1–3 contain the same stereochemical elements – helical chirality and stereogenic C7,C7′-2-hydroxypropyls. Beyond these elements, the perylenequinones exhibit intriguing architectural aspects, namely keto-enol tautomerism17,18 and potential atropisomerization of the helical perylene core (Figure 2). The barrier to atropisomerization of the helical configuration, which entails rotation of the C2,C2′- and C7,C7′-groups past one another (Figure 2), varies substantially for the compounds in this series. Whereas the calphostins and phleichrome are atropisomerically stable, the additional seven-membered ring in cercosporin lowers the barrier, making it a particularly challenging synthetic target (see Table 1 and discussion below).5,19 On the other hand, the hypocrellin and hypocrellin A atropisomerize rapidly at ambient temperature presenting two sets of sharp peaks in the NMR spectrum.18
The perylenequinones exhibit light induced activity in biological systems making them photodynamic therapeutic candidates.20 Photodynamic therapy is a treatment that uses a small molecule, called a photosensitizer or a photosensitizing agent, oxygen, and a light source.21 After the photosensitizers are delivered to target cells/tissues, they are exposed to a specific wavelength of light forming excited states. In turn, these species transfer the energy to molecular oxygen to produce singlet oxygen as well as other reactive oxygen species that kill cells. Generation of photosensitizers that can be activated by longer wavelength light, which can penetrate deeper into tissue, is crucial. The nine photosensitizer compounds (none perylenequinones) currently in clinical use as photoactivated anticancer agents encompass three related structural classes: porphyrins, chlorins, and phthalocyanins.21 While optimization of current photosensitizers is warranted, the multiple criteria of desirable photosensitizers makes discovery of new classes critical to furthering the practice of photodynamic therapy.
The perylenequinone core is a potent chromophore that permits the use of these compounds as photosensitizers. Upon exposure to light, perylenequinones initiate the generation of reactive oxygen species with high quantum efficiency.20 Furthermore, these compounds selectively bind the C1 regulatory domain of protein kinase C (PKC),22,23 some forms of which are upregulated in cancer cells and associated with growth.24 For these reasons, the perylenequinones are candidates for photodynamic therapy of cancer;20 light-induced activity has been seen with the natural products against selected tumor cell lines.25 In addition, they have displayed antiviral26 and immunotherapeutic27 properties. The development of new classes of PDT agents of defined molecular structure, with lower aggregation tendency, longer absorption wavelengths, high quantum yields, greater stability, and greater selectivity against cancer cells is highly desirable.
The perylenequinones are thought to cause cell apoptosis by two pathways: 1) light-induced production of singlet oxygen and 2) PKC inhibition. Considerable attention has been devoted to the development of potent and specific PKC inhibitors.24 The catalytic domain-acting PKC inhibitors, such as staurosporin (PKC, IC50 = 9 nM) and other indolocarbazoles, are highly potent but not specific for PKC, since they also inhibit other protein kinases.28 Since the discovery of staurosporin, the structurally related catalytic site inhibitors ruboxistaurin (PKCβI, IC50 = 4.7 nM; PKCβII, IC50 = 5.9 nM)29 and enzastaurin (PKCβI, IC50 = 30 nM; PKCβII, IC50 = 30 nM)30 have been found to be more selective for the PKCβ isozyme relative to other PKC isozymes and kinases leading to important new diabetes treatments.31 Significantly, the perylenequinones have demonstrated to be both potent and specific for PKC [Calphostin C (1c): PKC, IC50 = 0.05 – 0.46 μM; PKA, IC50 = > 100 μM; PPK, IC50 = > 100 μM; Cercosporin (3): PKC, IC50 = 0.6 – 1.3 μM; PKA, IC50 = > 500 μM; PPK, IC50 = > 180 μM] most likely because they act on a regulatory domain which is unique to PKC.22,23 To date, evaluation of the perylenequinones as PKC inhibitors has been confined to natural product derivatives or simple analogs due to limitations of available synthetic methods.20,23,32,33 Novel perylenequinones will allow determination of the structural features necessary for PKC inhibition as well as generation of structures with improved photodynamic properties.
We desired a synthetic strategy to permit a flexible approach to all of the perylenequinone natural products 1–6 (Figure 1) as well as several derivatives. Examination of prior approaches to the clphostins4 revealed a common theme in that the stereochemistry of the C7-subsitution was generated first and then diastereoselective dimerizations were undertaken (Scheme 1, path b). The moderate diastereoselection (1:1 – 1:8) afforded the newly formed C1-C1′ biaryl bond presumably arises from the distance between this center and the C7,C7′-substituents. Furthermore, the predominant diastereomer did not usually correspond to the calphostin array although it does match that needed for phleichrome.
In devising a plan to selectively synthesize all of the possible stereoisomers in pure form, we elected to pursue a different strategy that mimics the likely biosynthesis.11c,34 By establishing the axial stereochemistry first (path a in Scheme 1), the corresponding helical stereochemistry can be generated with complete stereocontrol. The helical stereochemistry can in turn be utilized to control the C7,C7′-stereochemistry with good fidelity as demonstrated in our synthesis of hypocrellin A (Scheme 2).3 Notably, a synthesis of hypocrellin relying on path b in Scheme 1 would require oxidation of the initially formed alcohols resulting in loss of this stereochemical information.
Alternately, the C7,C7′-stereochemistry can be introduced from an external source, a gambit that permits selective synthesis of all the possible stereoisomers of calphostin D, phleichrome, and cercosporin. We selected an epoxide opening reaction to achieve this goal (Scheme 2). The net result is that all of the target structures (1–5) devolve onto a common synthetic intermediate, chiral biaryl 13 or its enantiomer, ent-13 (Scheme 2). Furthermore, bisiodide 13 provides a potential avenue for the synthesis of diverse C7,C7′-analogs by coupling using Heck, Suzuki, and related methods.
In applying such an approach to cercosporin, the sequencing of the transformations is critical due to the relatively low atropisomerization barrier. From measurements that we performed, simple perylenequinones such as 16 enjoy only a moderate degree of atropisomeric stability. Since the C2,C2′,C7,C7′-groups cannot be coplanar due to steric interactions, a helical twist is imposed on the central perylenequinone unit (Figure 2). The strain results in a barrier to atropisomerization for 16 of 28.3 kcal/mol (Table 1, entry 3), substantially lower than that seen with BINOL (14) (37.1 kcal/mol, Table 1, entry 1).35 Larger C7,C7′-substituents would be expected to sterically hinder atropisomerization and to confer a higher degree of atropisomeric stability to the calphostins and phleichromes. Thus, we measured the atropisomerization rate of ent-1d, which has the same (M)-helicity as 16, in the same solvent (benzene) and, as expected, a higher barrier of 30.2 kcal/mol (Table 1, entry 4) was found relative to 16. As a result, the atropisomerization barriers for ent-calphostin D (ent-1d) and phleichrome (2) allows retention of the helical stereochemistry up to 60 °C (Table 1, entries 4–5).19 The addition of the methylidene bridge to the perylenequinones as seen in cercosporin (3) is expected to impose additional strain and lower the atropisomerization barrier further. For example, the addition of a similar bridge to BINOL (14) to provide 15 reduces this barrier from 37.135 to 3336 kcal/mol (Table 1, entries 1–2).37
Reports in the literature for cercosporin indicated that rapid atropisomerization occurs in boiling ether (37 °C)5 whereas a slower isomerization was observed in DMSO (t1/2 = 2 d at 60 °C).19 To determine if there is indeed a solvent dependence, we measured the atropisomerization rate in DMSO and benzene at 65 °C. Indeed, the rate was substantially faster in benzene corresponding to a barrier of 27.4 kcal/mol vs 28.2 kcal/mol in DMSO (Table 1, entry 6). These experiments revealed that at 60 °C or higher would result in unacceptable levels of atropisomerization. From a synthetic standpoint, these results mean that the combination of the perylenequinone and the methylidene bridge must be reserved to as late as possible. Furthermore, all subsequent steps must avoid even moderate temperatures in both the reaction media and purification.
Two distinct approaches fulfill the above requirements. In the first (bottom of Scheme 3), the perylenequinone (i.e., 7) is constructed prior to formation of the methylidene bridge in 3. For this strategy to succeed, the conditions for installation of the methylidene bridge need to be relatively mild since the product 3 can undergo significant atropisomerization at 60 °C (Table 1, entry 6). Since model studies indicated temperatures of 65 °C were needed to form the methylidene bridge, we turned to the reverse sequence (top of Scheme 3). Here the methylidene bridge would be installed early (i.e., 19), also obviating the need for three sets of orthogonal protecting groups in intermediate 8 (bottom of Scheme 3). For this strategy to succeed, the bridged binaphthyl compound needs to be configurationally stable, which is supported by the results from 15 (Table 1, entry 2) which can readily withstand temperatures up to 90 °C. Furthermore, the reaction conditions for formation of the perylenequinone (17 to 3) are very mild (room temperature and lower),3,4c,d which should prevent any atropisomerization of the sensitive 3 (Table 1, entry 6).
With the general strategy defined, the retrosynthesis evolved to that in the top of Scheme 3. Specifically, helical chiral 3 would arise from axial chiral 17 by means of low temperature reactions to suppress the ready atropisomerization of 3 (Figure 2). Retrosynthetic installation of the C3,C3′-esters in 18 provides a means of protecting the C3,C3′-positions for the next operation and provides functionality needed for highly selective naphthol coupling and rapid naphthalene synthesis. Deoxygenation to 19 and subsequent removal of the methylidene bridge furnishes 20 with a substitution pattern readily generated via biaryl coupling.2,38 The epoxide opening disconnection was applied at this stage to yield an even simpler biaryl 21. This stratagem reserved diversification of the synthetic sequence to the latest stage possible providing access to the calphostins, phleichrome, and cercosporin. However, the retron to 21 requires an unprecedented epoxide opening; namely, the combination of a complex biscuprate with the requirement for two epoxide alkylations. Biaryl 21, in turn, arises from C2,C2′,C4,C4′-protection of 13 (Scheme 2). Thus, common intermediate 13, available from our previous studies,2,3 could be used for the construction of all the perylenequinone natural products reducing the total number of steps that would need to be optimized. Due to the ability to independently generate the two stereochemical elements, the 2-hydroxypropyl group and the helical chirality, this synthetic approach would permit the syntheses of the complete diastereomeric series of 1, 2, and 3, including compounds not isolated from natural sources. In addition, novel perylenequinone structures would be easily accessible from advanced intermediates.
In prior syntheses of the calphostins and phleichrome, the formation of the C7,C7′-stereochemical array has been a central element. In all cases, incorporation of this element was intimately coupled to formation of the naphthalene skeleton prior to dimerization. Our strategy (Scheme 2) reserves introduction of these substituents until after naphthalene formation and dimerization, a feature which greatly streamlines production of the naphthalene portion and allows for maximum diversity in the synthesis of the natural products and analogs.
To install the stereogenic 2-hydroxypropyl groups we investigated three routes: 1) diastereoselective reduction of diketone 24; 2) asymmetric methyl-addition to dialdehyde 25; 3) organocopper epoxide-opening with 21 (Scheme 4). Unfortunately, both reduction of the diketone and methyl addition to the dialdehyde proved to be unproductive, proceeding with low yield or diastereoselectivity. Apparently, the biaryl axis is too distant to exercise a high degree of stereocontrol over the additions to the carbonyls of the C7,C7′-substitutents, an outcome reminiscent of the poor stereocontrol observed in the opposite direction (path b, Scheme 1). The last pathway, the organocopper epoxide-openings (Scheme 4) was particularly attractive in that it allowed access to all of the diastereomers with full stereocontrol without having to rely on the restrictions of relative diastereoselection. Furthermore, such a pathway permits greater flexibility in analog synthesis. For example, other epoxides or aziridines could lead to a host of analogs not available from the natural products. However, the unprecedented double alkylation of a highly complex biscuprate with two epoxides made this venture risky.
The use of Grignard-derived cuprates in epoxide alkylation has enjoyed a rich history.39 While complex cuprates have been used successfully 40 and simple biscuprates have been employed in epoxide alkylation,41,42 we could locate no reports of a highly functionalized dianion effecting two ring-openings. Our primary concerns were: 1) the metal-halogen exchange on an electron-rich system and in the presence of the C3-methyl esters; 2) the stability of the electron-rich bisarylcuprate, and 3) the requirement that the biscuprate be the limiting reagent rather than in excess as is typical.
To fully evaluate the proposed epoxide-opening reaction, several naphthalenes were investigated revealing that a C7-iodide permitted the best coupling and that robust methyl protecting groups needed to be employed instead of acetates. Hindered bases (t-BuLi, i-PrMgBr) were found to attenuate addition to the C3-methyl ester. Encouragingly, the initial epoxide-opening with 26 yielded modest amounts of the desired product (40%) with complete regioselectivity (Table 2). Rigorously air- and water-free conditions drastically decreased the amount of protodemetallation, improving the yield of 27 to 77% (entry 1a). An examination of the cuprate addition revealed that additives such as TMSCl or HMPA or lower temperatures (−78 °C) provided no improvement. Upon screening the copper reagent and solvents (Table 2, entries 1a-d), CuI and THF were found to be optimal. Careful purification of the CuI by recrystallization proved to be the most critical element providing product 27 in 87% yield (Table 2, entry 1e).
Due to the requirement for a C4-acetate during biaryl coupling2 (see below) which is not tolerated in the epoxide opening, the epoxide opening reaction was reserved until after biaryl coupling in order to reduce the number of protecting group steps. While the task of forming a biscuprate and undertaking two epoxide alkylations on such a complex dianion was daunting, the opportunity for convergence as late as possible in the synthesis was appealing allowing analogs to be produced from one late intermediate. To examine epoxide opening as late as possible, 28, in which the oxidation of the C5,C5′-positions had already been accomplished, was utilized (Table 2, entry 2). An arene with acetate groups intact was the major product of this reaction arising from protodemetallation, which indicates that either cuprate formation or epoxide opening was sluggish. Presumably, steric gearing from the C5,C5′-positions hinders the C7,C7′-Grignards slowing subsequent reactions at these positions. Moving the epoxide alkylation prior to C5,C5′-oxidation but after formation of the methylidene bridge was examined next (Table 2, entry 3). While the desired product 31 did form in 40% yield, the methylidene bridge caused the C3,C3′-esters to be less hindered resulting in a competing halodecarboxylation. To alleviate this problem, the double epoxide alkylation was investigated with different C2,C2′-substitution (Table 2, entry 4). The C2,C2′-bismethoxy, bisbenzyloxy, and bisisopropoxy analogs all reacted well providing the corresponding products with two new stereocenters in 65–75% yield, corresponding to 81–87% yield for each of the two individual coupling reactions. Furthermore, reaction of the same M-chiral biaryl 32a with the enantiomeric epoxide provided 34, the diastereomer of 33a in 74% yield (Table 2, entry 5). While different rates might be expected via matched and mismatched reactions of the enantiomeric epoxides (R)- and (S)-propylene oxide with (M)-32a, no difference in the reaction rate was seen here. Apparently, the stereochemistry of the biaryl is distant from the reacting centers and exerts little control, in line with previous observations (see Scheme 4). Together, these results validated the use of the epoxide dialkylation as a key bond construction providing a highly flexible synthetic route. For example, biaryl 32 can be used to furnish 33a corresponding to ent-phleichrome (ent-2) and 34 corresponding to ent-calphostin D (ent-1d) as well as 33b and 33c which can either provide analogs with different C2,C2′-subsititution or serve as precursors toward cercosporin (3).
Our total synthesis efforts commenced with the simpler perylenequinone compounds lacking the methylidene bridge. As outlined in Scheme 5, the bisiodide 13 was generated in a straightforward manner using the protocol that we reported for its enantiomer.3 Thus, the M-biaryl was produced in good yield and enantioselectivity (81% ee) utilizing the (S,S)-diaza-cis-decalin copper catalyst.1 The high crystallinity of bisiodide 13 provided a simple means of enantioenrichment with one trituration giving >99% ee material and excellent mass recovery. The C3-ester is integral to the synthesis of the naphthalene and also provides a necessary coordination site for the biaryl coupling catalyst. Large quantities of common intermediate 13 could be generated, which was used in the synthesis of three natural products and ten analogs.
The C2,C2′,C4,C4′-phenols of 13 were methylated by means of excess NaH in undistilled DMF with excess MeI (Scheme 5). The small amount of water in DMF was converted to NaOH, which cleaved the acetates. Subsequently, MeI alkylated all four positions with high efficiency (94% yield). At this stage, formation of the biscuprate and double epoxide alkylation were undertaken following the method described above (see Table 2 and accompanying discussion). Both (R)- and (S)-propylene oxide were employed to provide the diastereomers (M,R,R)-33a and (M,S,S)-34 in good yield. After benzylation of the newly formed alcohols, the C5,C5′-oxidation was undertaken. We found that our previously reported protocol3 for this transformation using Kita’s method [(PhI(OCOCF3)2, (CF3)2CHOH, NaOAc]43 was very sensitive to any changes in the oxidation potential of the substrate requiring reoptimization for each new precursor. As a result, we devised a more robust C5,C5′-oxidation utilizing a palladium-catalyzed O-arylation. While these transformations have been reported for a host of systems,44 we aimed to determine their applicability to highly functionalized, hindered, electron-rich systems. Thus, chlorination was effected with high efficiency (>90% yield) using sulfuryl chloride to provide (M,R,R)- 36 and (M,S,S)-36. Here, the C3,C3′-esters again proved critical; while not present in the natural products, they serve to protect the C3,C3′-positions from halogenation. Disappointingly, direct Oarylation of the bischloride with benzyl alcohol and other alcohols failed, likely due to the hindered nature of the C5,C5′-position. Gratifyingly, hydroxylation using KOH and a palladium catalyst45 proceeded extremely well and was found to be remarkably general for a range of substrates. While the product bisphenol was not stable, immediate protection with benzyl bromide furnished (M,R,R)-37 and (M,S,S)-37. A control reaction using KOH without palladium catalyst did not produce any of the bisphenol indicating that an addition/elimination pathway of the conjugate ester was not responsible for the results.
The next task was removal of the C3,C3′-esters which were immune to aqueous basic hydrolysis even at elevated temperatures, due to their steric hindrance. We discovered that an SN2 displacement of the methyl esters could be accomplished with NaCN in DMSO to provide the bisacid.3 However, conventional aromatic decarboxylation protocols46 required temperatures that would atropisomerize the biaryl. As a consequence we developed a palladium-catalyzed decarboxylation protocol,47 which worked well in these systems.3 Unfortunately, we discovered a competing C-H insertion reaction occurred with the methylidene bridge series en route to cercosporin (see below). Thus, we undertook development of a new decarboxylation strategy with 37. Reexamination of other decarboxylation methods48 yielded no success so our attention turned to decarbonylation.49 Formation of the requisite dialdehyde proved to be more efficient by overreduction using DIBAL and then oxidation with IBX (Scheme 5). Decarbonylation with two equivalents Wilkinson’s catalyst to (M,R,R)-38 and (M,S,S)-38 then proceeded smoothly provided that rigorously oxygen-free conditions were employed.
The remaining steps of the synthesis could all be conducted under very mild conditions. Removal of the four benzyl ethers occurred readily and was followed by oxidative cyclization with MnO2 to reveal the central perylenequinone core.4c,d Directed deprotection of the C4,C4′-methyl ethers with MgI2 4b,e,f furnished the enantiomers of the natural products phleichrome (ent-2) and calphostin D (ent-1d), respectively (Scheme 5). Each synthesis proceeded in a total of 17 steps with ent-2 being formed in an overall yield of 5.3% (average of 87% per step) and ent-1d being formed in an overall yield of 5.2% (average of 87% per step). With the enantiomers of the natural products available, it was now possible to assess the impact of the helical and centrochiral centers on the biological activity.
Starting from common intermediate 13 (see Scheme 5), orthogonal C2,C2′-benzyl ethers were installed via a Mitsunobu reaction. A subsequent one-pot deacylation/methylation of the C4,C4′-positions afforded 32b with high yield and scalability (Scheme 6). The double alkylation of the biscuprate of 32b with (R)-propylene oxide furnished 33b in high yield as a single diastereomer. The newly formed 2-hydroxypropyl groups were then masked as benzyl ethers furnishing 39. Subsequent chemoselective debenzylation of the more labile C2,C2′-benzyl ethers of 39 was efficiently accomplished using Pd/C poisoned with pyridine. Installation of the methylene bridge was not trivial, presumably due to disfavorable entropy in forming the seven-membered ring combined with ring strain. Of all the potential methylidene equivalents, ring formation was most facile with BrCH2Cl vs BrCH2I or BrCH2Br. Apparently, there is a balance between the leaving group ability of the second halide and its size with larger halides greatly decelerating the first alkylation event. Even so, formation of the seven-membered ring of 40 required heating to 65 °C, which was well below the atropisomerization threshold (see Table 1, entry 2).
As described above (see Scheme 5, 33a/34 to 37 and accompanying text), we elected to use a halogenation/O-arylation reaction to install the C5,C5′-oxygenation. The C5,C5′-chlorination proceeded uneventfully, but the chloro to alkoxy interchange was difficult. After optimization of the reaction conditions, we found that the catalyst system derived from Pd2dba3 and the X-phos(t-Bu) ligand proved to be effective in the coupling of 40 with KOH to provide the desired bisphenol. Unfortunately, the resulting product was highly unstable and could not withstand a one-pot alkylation protocol. Isolation of the bisphenol under carefully controlled conditions, however, proceeded well. Immediate exposure to our standard benzylation procedures (BnBr, NaH, DMF) furnished key intermediate 41 in 70% yield.
As alluded to above (see Scheme 5, 37 to 38 and accompanying text), decarboxylation of 41 proved difficult. We had developed a palladium catalyzed protodecarboxylation for this application which succeeded in related systems.47 With 41, however, this procedure failed due to a competing C-H insertion reaction of the methylidene in the seven-membered ring. Fortunately, decarbonylation of the corresponding dialdehyde with excess Wilkinson’s catalyst proceeded well, provided that rigorously airfree conditions were employed. Furthermore, the relatively high temperatures (85 °C in diglyme) required for this process were well-tolerated in accord with our expectations (Table 1, entry 2) and no atropisomerization was observed.
After deprotection of 42, the bisphenol was oxidized by MnO2 to afford perylenequinone (Scheme 6).4c,d As discussed in the retrosynthetic analysis, this final cyclization step was held until the end of the synthesis because formation of the perylenequinone reduces the dihedral angle between the upper and lower portions from 50° in axial chiral 42 to 20° in helical chiral 3, further facilitating atropisomerization. Directed deprotection of the C4,C4′-methyl ethers was accomplished selectively with MgI2,4b,e,f completing the first total synthesis of cercosporin (3) in 20 steps and 2.8% overall yield (86% average yield per step). Notably, none of the diastereomer of 3 arising from the facile atropisomerization was observed in this sequence. The synthetic material possessed NMR and CD spectra identical to those from a sample of the natural product.
The two perylenequinone apoptosis pathways allow for the therapeutic effects to be localized in cancer therapy two distinct ways: 1) via direction of a specific wavelength of light to the region of the tumor and 2) via sequestration in cells with high levels of PKC. Since most tumor cells appear to have multiple growth and survival mechanisms, the two apoptosis pathways and means of therapeutic localization render the perylenequinones especially promising targets. As a result, we targeted derivatives to both increase the photodynamic activity and to also probe PKC binding. To increase the photo-response in the photodynamic window (600–900 nm), the perylenequinone chromophore was extended through C3,C3′-substitution. Additionally, we designed and synthesized perylenequinones in a systematic manner to examine the effects of the helical chirality, C2,C2′-substitution, C3,C3′-substitution, and C7,C7′-substitution, which have thus far seemed to be the most integral to PKC inhibition.23 Most perylenequinone analogs examined to date are derived from the natural product themselves;24 generation of the above variants from the natural products is difficult or not feasible. The lack of information has limited the conclusions drawn between structure and PKC inhibition. On the other hand, our synthetic strategy allows rapid and convergent synthesis of a large number of novel analogs for biological and photophysical evaluation.
The versatility of our perylenequinone strategy (Scheme 2) enabled the syntheses of the stereoisomers of 1–4, including the natural products themselves (Figure 3). Utilizing the enantioselective biaryl coupling we can access both enantiomers of the common intermediate (P-13 and M-13). In addition, the copper-mediated epoxide-openings are integral to this evaluation, allowing entry to the unnatural isomers of 1, 2, and 3, which would not be possible if a stereochemistry transfer reaction was used in the construction of the stereogenic 2-hydroxypropyl groups. Thus, our approach permits a systematic assessment of the helical stereochemistry and the C7,C7′-stereochemistry on PKC inhibition.
Docking studies were performed in order to determine what structural modifications might lead to higher activity.50 Starting with the PKCδ C1 regulatory domain obtained from a cocrystal structure with a phorbol ester, ent-phleichrome (ent-2), which contains the helical and centrochiral stereochemistries that led to the best binding (see below), was docked into the phorbol binding domain utilizing Autodock3.51 The lowest energy binding mode places the C2,C2′-methoxy groups into a hydrophobic pocket of the binding site. Since there was unfilled space in this pocket, we proposed to modify the C2,C2′-substitution to include the more hydrophobic isopropoxy and n-propoxy groups (48a and 48b). Docking of these two analogs of (ent-2) indicated superior binding and is expected to lead to higher affinity.52 As seen in Scheme 7, these novel perylenequinones could be synthesized by selective alkylation of an advanced intermediate (44) available from the cercosporin synthesis (step 11, Scheme 6). The remaining steps followed the protocol developed above for calphostin/phleichrome (Scheme 5) and cercosporin (Scheme 6) providing unnatural congeners 48a and 48b in a straightforward manner in 9 steps and 9 steps, respectively (Scheme 7). The yield for the last three steps was lower than those obtained above for cercosporin due to the larger C2,C2′-n-propoxy and –iso-propoxy groups which slow the perylenequinone forming reaction.
Whereas the docking studies described above indicated a possible interaction of the C2,C2′-substituents with a hydrophobic pocket in the regulatory domain site, the much higher affinity of calphostin A (Figure 1, 1a, R1 = R2 = COPh, IC50 = 0.25 μM) compared to calphostin D (1d, R1 = R2 = H, IC50 = 6.4 μM) points to interaction of the C7,C7′-portion with the binding site.22b While it may be possible that the perylenequinone lies in a sufficiently deep pocket to undergo interactions with both the western and eastern portions of the structure, we expect that C2,C2′- and C7,C7′-analogs will be invaluable in elucidating the relative importance of each of these interactions. Our first step was to determine whether the stereogenic hydroxyl groups on the C7,C7′-substituents of the natural products were critical to binding. Thus, the simpler C7,C7′-propyl substitution was incorporated into analog 16 (Scheme 8). Based upon the improved activity of this compound relative to the parent natural products (ent-1d and ent-2) with the C7,C7′-2-hydroxypropyl substitution (see below), a series of derivatives (53- 57) incorporating the C7,C7′-propyl groups were designed to probe the effect of substitution at the C3,C3′-positions. Specifically, bromo (57), ester or acid (53, 55, 56), and vinyl (54) C3,C3′-substitution was proposed to increase the absorption wavelength of the perylenequinone chromophore (Scheme 8). The use of the C3,C3′-methyl ester in our synthetic strategy proved very useful at this juncture allowing all of these structures to be quickly accessed either directly or by decarboxylative functionalization (Scheme 8).
The synthesis of 16 and 53–57 (Scheme 8) commenced with racemic bisiodide 13, a byproduct of the trituration to provide M-13 (see Scheme 5). Cross coupling with the iodides of 13 is a versatile and simple means to introduce any desirable C7,C7′-substitution. In this case, Suzuki coupling with pinacol allylboronate provided bisallyl 49.3 Subsequent hydrogenation yielded the bis-n-propyl compound. The lack of functionalization of the C7,C7′-groups allowed facile C5,C5′-hydroxylation with Kita’s reagent43 to provide 50 following our first generation protocol.3
Compound 50 represents the first branchpoint intermediate in the analog syntheses; a small amount was subjected to MnO2 followed by MgI2 to effect oxidative cyclization2 and deprotection furnishing analog 53. The remainder of branchpoint intermediate 50 was subjected to benzyl bromide and NaH to protect the C5,C5′-naphthols in preparation for ester hydrolysis to provide the next key branchpoint intermediate, biascid 51. Ester hydrolysis here was surprising facile relative to congeners 37 (Scheme 5) and 41 (Scheme 6). Presumably, the smaller C7,C7′-groups alleviate gearing interactions that sterically hinder the C3,C3′-esters.
Bisacid 51 was subject to three different fates. In the first pathway, a palladium catalyzed decarboxylative Heck reaction of 51 followed by perylenequinone formation provided bisstyryl derivative 54.53 In the second pathway, the C5,C5′-benzyl ethers were cleaved and the more acidic carboxylic acids were then selectively benzylated using BnBr and K2CO3. This tactic was necessary because the carboxylic acids did not survive the MnO2 mediated oxidative cyclization, nor was selective hydrogenation of the bisbenzyl ester of 51 successful. Subsequent perylenequinone formation yielded 55, which could then be further hydrogenated to yield the perylenequinone bisacid 56. In the third pathway, protodecarboxylation using a palladium source furnished 52,47 which was then transformed to perylenequinone 1654 following protocols described above. Perylenequinone 16 could, in turn, be treated with Br2 to provide bisbromide 57. Subsequent enzyme assays with this analogs series revealed high activity for 16, stimulating us to also undertake a synthesis of M-16 since we had established that the M-isomers are more potent (see below).
The activity of the calphostins and hypocrellin A are thought to arise in part by modulation of protein kinase C (PKC), a family of C1 domain containing kinases. The PKCs are membrane bound proteins that play important roles in intracellular signal transduction effecting proliferation, differentiation, and apoptosis and have been implicated in a variety of disease states including diabetes, cancer, heart disease, and cognitive disfunction.24,55 Because the C1 domain is found in only a small subset of the human kinome,56 selectivity relative to the many other biologically important kinases may be achievable. In light of this information, the newly synthesized analogs were tested for their ability to inhibit PKC phosporylation, providing a benchmark comparison of a broad range of the natural products and analogs to establish the most critical elements for PKC inhibition.
As seen in Table 3, there are clear trends between the perylenequinone architecture and PKC-inhibitory activity. Notably, these results represent the first systematic evaluation of helical chirality on PKC inhibition. We found a direct correlation between helical chirality and inhibition; the M-perylenequinones were 1.8–20 times more potent than the corresponding P-isomers with the same C7,C7′-stereochemistry [(+)-1 vs (−)-2, (+)-2 vs (−)-1, and 3 vs epi-3]. For the hypocrellins (4 and ent-4), the helical stereochemistry atropisomerizes rapidly at room temperature and the compounds are 4:1 equilibrium mixtures of the two atropisomers.18 In line with the above results, the greatest PKC inhibition was observed for the compound where the M-helicity predominates (ent-4). Particularly noteworthy is that the unnatural isomers (+)-1 and (+)-2 were more potent than the natural products (−)- 1 and (−)-2. With respect to the stereochemistry of the C7,C7′-substitution, the R,R-array was found to offer slightly better (1.7-fold) activity than the S,S-array [(+)-1 vs (+)-2, (−)-1 vs (−)-2].
Docking models of the perylenequinones indicated that larger hydrophobic groups at the C2,C2′-positions better fill a pocket in the binding site (see above).52 Analogs of the most active stereochemical array (M,R,R) were thus constructed to probe the role of the C2,C2′-substitution (Scheme 7). Pleasingly, the bisisopropyl 48a (0.8 μM) and bis-n-propyl 48b (1.5 μM) were more potent than the parent compound (+)-2 (3.5 μM) providing support for the docking model.
Prior reports indicated that acylation of the C7,C7′-hydroxypropyl group provided superior activity (i.e. calphostin A, 1a, R1 = R2 = COPh, 0.25 μM vs calphostin D, 1d, R1 = R2 = H, 6.4 μM; Scheme 1).22b To determine if this result arises from a simple increase in hydrophobicity and removal of the two hydroxyl H-bond donors, the C7,C7′-n-propyl analogs were synthesized (Scheme 8). Notably, this compound possessed significant potency with racemic 16 exhibiting an IC50 of 1.2 μM indicating that hydrophobic C7,C7′-groups are indeed optimal (Table 3). Based upon our discovery (see above) that the M-helicity provided superior inhibition, enantiopure M-16 was generated and was found to exhibit an IC50 of 0.4 μM. Based on these data, P-16 is indeed a poorer inhibitor (2.0 μM) supporting our prior conclusions. Enantiopure M-16 is a 9–25 times better inhibitor than the nearest congeners, (+)-1 and (+)- 2, and 16–28 times better than the corresponding natural products, (−)-1 and (−)-2.
The finding that the stereogenic C7,C7′-substitution is not necessary for potency and that architecturally simple structures such as 16 can be employed, motivated us to examine C3,C3′-analogs of the C7,C7′-n-propyl series (Scheme 8). In doing so, compounds (53–57) with esters, acids, styryls, or bromo groups as the C3,C3′-positions were designed to alter the chromophore to improve the absorption profile. Analogs 53–57 displayed increased absorption in the 600–800 nm range compared to the natural products. Assessing the effects on PKC inhibition, the bisacid 56 was found to decompose rapidly. On the other hand, 53-4 and 57 were found to be 2–4 times less potent than the parent analog 16, having the same architecture but no C3-substitution.
The assays described above were performed with PKC containing regulatory and catalytic domains. To determine whether these perylenequinones were indeed inhibiting phosporylation by acting on the regulatory domain, assays were performed with PKC containing only the catalytic subunit. Over 85% of the catalytic subunit activity was retained even when 50 μM 16 was employed which is well above the IC50 (1.2 μM) seen with the whole enzyme.
To determine whether the structural effects on PKC inhibition are related to anticancer activity, nine cancer cell lines were screened against the most potent M-isomers: ent-phleichrome [(+)-2], cercosporin (3), and hypocrellin (ent-4). In addition, hypocrellin A (4) with the P-stereochemistry was assayed to assess the effects of the helical stereochemistry in cellular systems. Finally, the most potent analog, 16, was included. All of the compounds exhibited micromolar or submicromolar CC50 values for growth inhibition in line with their light activated mechanism of action. Surprisingly, hypocrellin (ent-4) and hypocrellin A (4) exhibited nearly identical activity indicating that the helical stereochemistry was unrelated to the growth inhibition. This result implies that PKC inhibition is not the sole cause of growth inhibition, which is expected due to the phototoxicity of the compounds. In organisms, binding to PKC is expected to localize the perylenequinones to cancer cells whereupon directed light irradiation would initiate the perylenequinone phototoxic response. Even so, differences were seen between the compounds and between different cell lines. Overall, cercosporin (3) was the most active compound in every cell line exhibiting CC50 values of 0.12–0.27 μM across the series. The remaining compounds exhibited similar activity across all of the cell lines with the exception of simple analog 16, which was a poorer inhibitor of HT29 and a superior inhibitor of A2780 and JHU-012 relative to all the compounds except cercosporin (3). Given the selectivity of cercosporin (3) and simple analog 16 for the head and neck cancer line JHU-012, the compounds were further examined without light exposure in this line. The photopotentiation was strongest with cercosporin (4.9-fold) among the natural products. Remarkably, the photopotentiation was even more pronounced for simple analog 16 (5.8-fold).
In summary, first total syntheses of (+)-1 and 3 and 10 perylenequinone analogs have been accomplished in an efficient manner beginning with a common chiral binaphthalene. The approach employing enantioselective oxidative biaryl coupling, double cuprate epoxide opening, and decarboxylative functionalization permits access with complete control of the helical and centrochiral stereochemical elements of these photoactive compounds. The analogs provided an assessment of the structural features important to PKC inhibition with M-helicity and hydrophobic C7,C7′-substitution being identified as advantageous. Substitution at the C2,C2′- and C3,C3′-positions was well tolerated and led to compounds simpler than the natural products, but possessing superior absorption profiles which may provide an avenue to superior photosensitizers. Finally, cercosporin (3) and simple analog 16 provide superior inhibition of growth activity in cancer cell lines and the latter displayed higher levels of photopotentiation than any of the natural products. Further biological evaluation of these and additional compounds, including selective inhibition amongst the PKC isoforms, is underway and will be reported in due course.
We are grateful to the NIH CA-109164 for financial support and to Novartis and the American Chemical Society for graduate fellowships (B.J.M.). Special thanks to E. M. Daub and Quest Pharmaceuticals for providing samples of cercosporin and hypocrellin, respectively. We thank James Ianni for assistance with the docking studies, Feng Gai and Virgil Percec for assistance with CD measurements, and Barry Cooperman for assistance with the enzyme assays.