Catalysis-Enabled Access to Cryptic Geldanamycin Oxides

Abstract

Catalytic, selective modifications of natural products can be a fertile platform for not only unveiling new natural product analogues with altered biological activity, but also for revealing new reactivity and selectivity hierarchies for embedded functional groups in complex environments. Motivated by these intersecting aims, we report site- and stereoselective oxidation reactions of geldanamycin facilitated by aspartyl-peptide catalysts. Through the isolation and characterization of four new geldanamycin oxides, we discovered a synergistic effect between lead peptide-based catalysts and geldanamycin, resulting in an unexpected reaction pathway. Curiously, our discoveries would likely not have been possible absent the attractive noncovalent interactions intrinsic to both the catalysts and the natural product. The result is a set of new “meta” catalytic reactions that deliver both unknown and previously incompletely characterized geldanamycin analogues. Enabled by the catalytic, site-selective epoxidation of geldanamycin, biological assays were carried out to document the bioactivities of the new compounds.

Short abstract

We report site and stereoselective oxidation reactions of geldanamycin facilitated by aspartyl-peptide catalysts.

Introduction

Catalysis in chemistry and biochemistry is founded upon accelerating reaction rates and achieving control over some selectivity issue.^1,2 Accordingly, the field of “asymmetric catalysis” continues to blossom as an ever more powerful way of producing stereochemically homogeneous building blocks with high efficiency.³ Directing chiral catalysts toward complex substrates, such as bioactive natural products, creates a further challenge in addition to enantioselectivity, involving the differentiation of functional groups within the same compound.^4,5 If more than one copy of the same reactive functional group exists in the structure, then the issue of stereoselectivity is compounded by that of site selectivity, resulting in a demanding landscape of partitioned pathways that lead to many different products to traverse. As a pragmatic solution to these multifaceted selectivity challenges, the application of a diverse catalyst library to a given scaffold offers the opportunity to simultaneously assess (a) multiproduct reaction outcomes and (b) to achieve analogues that might not be available in a straightforward manner by either biosynthetic or chemical methods. In addition, and perhaps most alluringly, subjecting complex molecules to catalyst libraries also creates the opportunity to unveil unexpected reactivity, leading to compounds that might not be targeted at all based on canonical reactivity patterns. This paper discloses findings along all of these lines, wherein we have observed an unanticipated cooperativity between a complex substrate of interest and various catalysts. Accordingly, our study has provided a number of perhaps “cryptic” natural product analogues.

In terms of project design, we were curious about expanding the breadth of aspartic acid-based peptides in the context of natural product derivatization. In the field of asymmetric catalysis, for small molecule functionalization, we had previously demonstrated that Asp-containing peptides are efficient and selective catalysts for both enantioselective alkene epoxidation (Figure 1a),^6,7 as well as Baeyer–Villiger oxidation^8,9 with then-unprecedented catalyst-controlled reversals of intrinsic migratory aptitude tendencies (Figure 1b). Mechanistically, the key has been a catalytic shuttle between the aspartic acid catalyst and its reactive aspartyl peracid form (1), which transfers the O atom in each scenario. We had also shown that a predictive choice of catalyst (i.e., peptide-sequence-selected) could be employed to predetermine whether epoxidation or Baeyer–Villiger oxidation would occur, at least with a carefully designed, admittedly “rigged” substrate 2 (Figure 1c).⁹ Accordingly, a critical next step for the advancement of catalyst-dependent, site-selective diversification of complex natural products with the Asp-based catalyst paradigm is to demonstrate feasibility with genuinely complex bioactive structures. For this purpose, we chose to explore the catalyst-dependent diversification of geldanamycin using Asp-containing peptides (Figure 1d).

Figure 1.

Overview of aspartyl-peptide catalyzed reactions and geldanamycin.

Geldanamycin exhibits profound biological activity, targeting Hsp90,^10,11 a molecular chaperone responsible for folding, stabilization, and maturation of many client proteins, and has shown promise as an anticancer therapeutic.¹²⁻¹⁵ However, geldanamycin exhibits poor bioavailability and hepatotoxicity,¹⁶ stimulating the pursuit of new analogues to address these issues via total^17,18 and semisynthetic methods.¹⁹⁻²² In fact, geldanamycin derivatives, such as 17-allylamino-17-demethoxygeldanamycin (17-AAG),²³ have reached clinical trials for the treatment of many types of cancer, including colorectal, breast, ovarian, lung, multiple myeloma, and leukemia.²⁴ In addition, geldanamycin is a challenge for confronting chemoselectivity, as it contains two alkenyl regions for epoxidation—a diene and an isolated trisubstituted alkene—along with quinoid functionality replete with unsaturation (red bonds, Figure 1d). Therefore, multiple alkene sites may be subject to functionalization, along with the consideration of π-facial selectivity that could lead to various diastereomers from the epoxidation of each alkene site. Finally, practical considerations were also favorable, such as accessibility to the compound and solubility in a variety of reaction-compatible solvents.

Results and Discussion

Our survey of aspartyl peptide-based catalysts produced interesting and unexpected results immediately. As is discussed in detail below, four principle mono-oxide products of geldanamycin are observed in these reactions, and their ratios may be extensively modulated with catalyst-dependent kinetic control. Two of the compounds resolve long-standing ambiguities regarding diastereomerism of the geldanamycin-derived epoxides,^14,21 one of which is efficiently accessed for the first time based on a unique, peptide-based catalyst. Two other compounds, each also unveiled for the first time as a result of a specific aspartyl peptide-based catalyst, are the products of an unanticipated reaction pathway. Moreover, observation of these compounds emerges from an unanticipated cooperation between the complexity of the peptide-based chiral catalysts and the embedded functional group arrays within the macrocyclic molecular scaffold of geldanamycin. This symbiosis, between complex chiral catalysts and complex substrates, enables explicit biological testing of these new and fully characterized compounds, several of which represent structures that were outside of our original conception, grounded in either rational design or traditional scaffold diversification analysis.

Chemistry

Initiating our investigations, geldanamycin oxidation was tested using m-CPBA to establish inherent reactivity and selectivity, as these results are likely to be minimally biased by outer sphere noncovalent interactions between the reactive species and substrate, beyond the traditional Henbest paradigm.²⁵ Reaction of geldanamycin with m-CPBA cleanly yielded two mono-oxidized products in a 15:85 ratio (A:B, Figure 2, entry 1), strongly favoring compound B. Upon isolation and characterization by NMR spectroscopy, it was determined that A and B are diastereomeric epoxides, derived from reaction of the 8,9-trisubstituted alkene. These compounds have been reported previously, albeit in the absence of either a relative or absolute stereochemical assignment.^14,21 Nor had either been produced with catalyst-dependent selectivity, to our knowledge. In our hands, we were delighted to achieve X-ray quality crystals of A. Analysis of the X-ray crystal structure revealed the complete stereochemistry of the compound, displaying that epoxidation occurs from the “bottom” face, as drawn, to yield A, implying that the m-CPBA-favored product B is the result of epoxidation of the “top” alkene face.

Figure 2.

Initial peptide evaluation and structure determination. Data is an average from two trials. Determined using UPLC from crude reaction mixtures with an internal standard and calibration curve. Product distributions are corrected to total 100%. See Supporting Information for yields. (abbreviations: Boc = tert-butyloxycarbonyl; Aic = 2-aminoindane carboxyl; Z = Cbz = carboxybenzyl; Acpc = 1-aminocyclopropane carboxyl; EBA = ethylbenzylamine; trt = trityl).

The coin of the realm in studies of site-selective catalysis is the catalyst-dependent reversal of the intrinsic selectivity patterns observed with simple reagents and catalysts. Thus, we then transitioned to the exploration of Asp-based catalysts. Rather surprisingly, results divergent from those observed with m-CPBA surfaced immediately. Using the monomeric Boc-Asp-OMe, the m-CPBA-favored product B is still favored over A, although to a lesser degree, and a new product (C) is now observed (Figure 2, entry 2). Interestingly, C results from epoxidation of the quinone at the 16,17-position, a result also confirmed by X-ray crystallography analysis. Further evaluation of catalysts (vide infra) led to the observation of product D, which also results from quinone epoxidation at the 16,17-position, yet from the alternate π face in comparison to C. This structure was also established by X-ray crystallography analysis. In general, other minor products, such as from bis-oxidation or diene epoxidation, may be produced, yet low yields (<5%) excluded these from our analysis. These observations alone struck us as quite noteworthy. The most preliminary survey of Asp-based catalysts rapidly delivered three alternative products relative to the major compound produced by m-CPBA. Two of these geldanamycin oxides resolved long-standing structural ambiguity (A and B), and two (C and D) proved to be new geldanamycin scaffolds, unknown prior to the present studies, to our knowledge. Overall, all four compounds provided new insight into reactivity trends of both geldanamycin and Asp-containing peptide catalysts.

To assess the extent to which these catalyst-dependent product ratios could be varied, a library of 14 Asp-containing peptides was selected from previous projects and was then evaluated (Figure 2, select results shown; for full list, see Supporting Information). Low selectivities were measured for predictable β-turn-biased and/or β-hairpin-biased sequences²⁶⁻²⁸ (entries 3–4), including two peptides with C-terminal amide caps (entries 5–6). A legacy catalyst from our group for allylic alkene epoxidation⁷ slightly favors product C (entry 7, P2). Both reactivity and selectivity for 8,9-epoxidation improved significantly when Boc-Asp-Aic-d-Asp(t-Bu)-Val-OMe (P8) was employed (entry 8). The formation of products C and D is largely suppressed with this more reactive peptide, resulting in a distribution of 68%, 25%, 5%, and 1%, respectively (A, B, C, and D, entry 8). In contrast to the reaction using m-CPBA, which yielded A and B in a 1:5.7 dr, P8 provides a coveted reversal of diastereoselectivity, delivering a 2.7:1 dr (A:B). Finally, legacy catalyst P1, which was previously reported to promote Baeyer–Villiger oxidation (Figure 1b),⁸ provides the most striking catalyst-dependent diversion from conventional reactivity, exhibiting 98% selectivity for quinone epoxidation, affording products C and D in a 3.5:1 diastereomeric ratio (entry 9).

Thus, our initial examination of Asp-based peptides (Figure 2), yielded two catalysts that favor two distinct products, alternative to that favored by m-CPBA: P8, favoring 8,9-epoxide A; and P1, favoring 16,17-epoxide C. Moreover, the screening provides catalyst-dependent access to the unique epoxyquinones C and D, which are not detected to an appreciable extent with m-CPBA. As these catalysts provide stark contrasts to the outcomes induced by m-CPBA, we wished to understand further the basis of these catalyst-controlled outcomes in hopes of some degree of further optimization.

To this end, several new catalysts were explored in the context of attempted optimization for product A (Figure 3). Relative to catalyst P8, exchanging Aic at the i+1 position for Acpc, d-Phe, or d-Val resulted in similar product distributions (entries 1–3, respectively). However, P10 with Acpc in the i+1 position does provide a modest enhancement of A:B ratio (3.1:1 dr), which could be further optimized (vide infra). On the other hand, if the i+1 residue is converted to an l-amino acid (Phe, Val, or Pro), the selectivity for A consecutively diminishes (entries 4–6, respectively). For example, with P15 (Pro at i+1), almost equal amounts of A, B, and C are observed (entry 6), as is seen with other peptides containing the sequence Asp-Pro (P2, P4, P6, Figure 2).

Figure 3.

Point mutations of A-selective catalyst. Data is an average from two trials. Determined using UPLC from crude reaction mixtures with an internal standard and calibration curve. Product distributions are corrected to total 100%. See Supporting Information for yields. (abbreviations: Boc = tert-butyloxycarbonyl; Acpc = 1-aminocyclopropane carboxyl; Dap = l-2,3-diaminopropionic acid).

As shown in Figure 3, additional changes at the i+2 position generally diminishes the selectivity for A. When d-Asp(t-Bu) is replaced with d-Asn(trt) at i+2, thus adding a hydrogen-bond donor in place of an acceptor, selectivity for A decreases (entry 7). Intriguingly, when the possibility for hydrogen-bonding at this position is removed through insertion of a d-Leu residue, selectivity for A is restored to a small degree (entry 8). Finally, low steric bulk at i+3 (Gly, entry 9) causes diminished selectivity for product A, yet a polar aromatic group (Tyr(t-Bu), entry 10) did not significantly modulate the product distributions in comparison to P10. However, including a hydrogen-bond donor (Dap(Boc)) at i+3 produces a lower amount of A, and a greater amount of C now emerges, affording almost equal yields of A, B, and C (entry 11). Overall, this series of peptide mutations revealed that most sequences, except those with additional hydrogen-bond donors, are likely accessing similar transition state ensembles, resulting in comparable energetic landscapes for the production of the major product, A.

To gain additional structural information concerning both A-selective catalyst P10 and C-selective P1, peptide truncation studies were performed. When A-selective catalyst P10 is shortened to the trimer Boc-Asp-Acpc-d-Asp(OMe)-OMe, ∼8–9% selectivity is lost for the formation of product A (Figure 4a, entry 2). However, the product ratios are well within the range observed for other tetrameric peptides, reinforcing the observation that the i+3 residue is not the most influential position in guiding selectivity profiles (e.g., P18 and P19 in Figure 3). The dimer, Boc-Asp-Acpc-OMe, reveals a much more significant perturbation on product selectivity, providing approximately equal amounts of A, B, and C (entry 3), suggesting that a β-turn-biased structure of P10 may be the source of the observed preference for A in the optimal tetrameric peptides. Indeed, the dimer results in similar product distributions in comparison to the monomer, Boc-Asp-OMe (entry 4). All of these data, including insight from the crystal structures, led us to consider a selectivity model, wherein the carbamate at C7 in geldanamycin could be engaged in a directing hydrogen-bond with the peptide backbone, guiding epoxidation to the “bottom” alkene face (Figure 4b).

Figure 4.

Mechanistic considerations for A-selective catalyst P10. Data is an average from two trials. Determined using UPLC from crude reaction mixtures with an internal standard and calibration curve. Product distributions are corrected to total 100%. See Supporting Information for yields.

The truncation study for C-selective P1, which favors quinone epoxidation (C and D), resulted in a much less drastic attenuation in product selectivity (Figure 5a). Surprisingly, each truncate resulted in similar product ratios to P1, all favoring the formation of C and D. Even the dimer Boc-Asp-Pro-OMe (entry 3) provides C and D in 69% and 27%, respectively, in comparison to P1, yielding C and D in 76% and 22%, respectively. A particularly remarkable comparison is that between two dimers, with only one residue changed: the dimer Boc-Asp-Acpc-OMe (Figure 4a, entry 3) favors 8,9-epoxidation overall (74:26 = A+B: C+D), while Boc-Asp-Pro-OMe remains selective for quinone oxidation (4:96 = A+B: C+D). Thus, it appears that the first two residues of P1 (Asp-Pro) are primarily responsible for directing the reaction toward the quinone epoxidation pathway.

Figure 5.

Data-guided mechanistic speculation into quinone oxidation to form C and D. Data is an average from two trials. Determined using UPLC from crude reaction mixtures with an internal standard and calibration curve. Product distributions are corrected to total 100%. See Supporting Information for yields.

Notably, any perturbation to the l-Asp-l-Pro dipeptide stereochemistry supports this assertion (Figure 5a, entries 5–7). In each case, any other stereochemical dyad results in diversion of the reaction pathway toward 8,9-epoxidation (formation of A and B). Furthermore, peptides containing the Asp-Pro sequence tend to provide a higher yield of products C and D, but the overall product distribution is dependent on the remaining residues (e.g., P2, P4, P6, P15). The subtle yet complex influence of the peptide sequence on the reaction outcome reinforces the synergistic nature of attractive interactions between the substrate and catalyst.

As quinone-selective catalyst P1 promotes a distinct reaction pathway, any selectivity model for the formation of C and D will differ substantially from that shown in Figure 4b. Thus, shown in Figure 5b is an alternative scenario that may account for the strict requirement for l-Asp-l-Pro chirality, with further selectivity tuning by distal residues. Accordingly, it is possible that the peracid and Pro carbonyl (blue circle) are oriented in the same direction,²⁹ such that interactions with the Pro(C=O) are responsible for this selectivity. The hydroxyl group at C11, which is situated near the back of the quinone, as drawn, could be involved in directing the peptide via hydrogen-bonding with Pro(C=O) toward product C. Formation of the minor epoxyquinone D could then result from conformational inversion of the quinone π-face, or even be directed by hydrogen-bonding with the amide at C1.

Regarding the plausible selectivity models presented in Figures 4b and 5b, it is absolutely essential to acknowledge that mechanistic rationalizations for selectivity outcomes with ratios at these levels must be realized with an abundance of mechanistic caveats. However, at the global level of these observed reversals of product selectivity, in situations in which multiple, non-stereoisomeric products are formed, we contend that the state of the art is limited and that these heuristic models may guide the field forward. Computational studies (and even empirical studies^29,30) of problems of this complex nature are also a frontier-level, extensive endeavor,³¹⁻³³ and ongoing efforts along these lines will no doubt add further insight.

As we did not anticipate formation of the epoxyquinones C and D from these studies, we wondered if their formation was a particular function of the catalysts, the conformation of geldanamycin and its epoxyquinone adducts, or a synergistic result of both features. Therefore, we examined the reactions of several model quinones (3–6; Figure 6) under analogous conditions. In all cases, no detectable reactions were observed using m-CPBA, and decomposition of the model quinones was detected under standard reaction conditions with P1, or when nucleophilic epoxidation conditions were tested (e.g., t-BuOOH, base). In contrast, compounds C and D seem to be quite stable when formed under conditions of catalytic P1. These realizations further bolster the symbiotic relationship between these catalysts and their substrates. These findings may well suggest that P1, originally discovered as a nucleophilic peracid and Baeyer–Villiger oxidation catalyst,⁸ may well be functioning once again in a peptide-tuned nucleophilic manner in the presence of the specific functionality-adorned geldanamycin scaffold. This assertion is supported by the fact that, subsequent to the discovery of P1, and based on its behavior, we also found the classical, stoichiometric nucleophilic epoxidation conditions (t-BuOOH, base; see Supporting Information for details) allow the formation of C in a comparable yield to catalytic P1 (D was observed in a much lower amount). For practical considerations, C may be efficiently accessed using simple nucleophilic epoxidation conditions as outlined in the Supporting Information. Furthermore, since these conditions fail to deliver efficient epoxidation of the model quinones further underscores the importance of the global functionality of the natural product, rather than simply its local structural environment.

Figure 6.

Evaluation of model quinones.

Having identified the selective peptide catalysts P10 and P1 for geldanamycin oxidation, and in pursuit of ample quantities for characterization and biological evaluation, several optimized experiments were carried out (Scheme 1). Fortuitously, higher reactivity was observed as a function of higher reagent loadings/lower peptide catalyst loadings. With P10, A and B were favored in slightly higher ratios than at the microscale, resulting in a 3.9:1 dr of A:B (74:19:6:1 of A:B:C:D). Products A and B were isolated in 39% and 10% yields, respectively. With P1, higher conversion could be achieved through use of stoichiometric DMAP, resulting in an A:B:C:D ratio of 0:0:76:24, from which C and D were isolated in 32% and 14% yields, respectively. Taken together, a comparison of catalysts P10 to P1 on a somewhat larger scale reveals that the C8,9 (A+B) versus C16,17 (C+D) functionalization ratios can swing from 93:7 to 0:100 as a function of the peptide-based catalyst, of course with attendant tuning of the diastereoselectivity in each series.

Scheme 1. Isolation of Geldanamycin Analogues.

As 17-substituted geldanamycin derivatives have demonstrated improved bioactivity and bioavailability,^23,34,35 we sought to test these optimal peptide sequences and reaction conditions with 17-AAG to access additional analogues. We predicted that P10 would likely lead to the same mode of epoxidation, as the local environment near the 8,9-position is unchanged. However, we were uncertain if adjusting the electronic nature of the quinone would disrupt or enhance quinone epoxidation with P1. With the scaled reaction conditions shown in Scheme 1, P10 indeed affords the expected product, in fact with elevated selectivity compared to geldanamycin, resulting in 17-A in 93% selectivity and 33% isolated yield. The same epoxide diastereomer as in A was confirmed by X-ray crystallography analysis (see Supporting Information). Yet, m-CPBA was less effective with 17-AAG than geldanamycin, providing only 30% conversion and favoring the same product as P10 in a 2:1 ratio (see Supporting Information). Finally, reaction of 17-AAG with P1 resulted in only 10% conversion of the starting material and detection of one new mono-oxidized product by UPLC-MS (see Supporting Information). While we have not definitively assigned the structure of this compound, the low reactivity of this modified quinone core in 17-AAG reflects the significant influence of the substituents on accessing quinone oxidation. No further optimization or isolation of this product was attempted.

Biology

Intrigued by these novel epoxy-geldanamycin and 17-AAG analogues, their biological activity was investigated. As previously mentioned, geldanamycin is a well characterized Hsp90 inhibitor and has shown promising anticancer activity. Therefore, the antiproliferative activity manifested by these analogues was determined against two breast cancer cell lines, MCF-7 and SKBr-3. Only one of the new compounds inhibited the growth of these cancer cells at 10 μM. Compound B, which resulted from the reaction of geldanamycin and m-CPBA, retained the majority of the parent compound’s antiproliferative activity. In fact, compound B manifested an EC₅₀ value of 338 ± 160 and 28.7 ± 14.3 nM against MCF-7 and SKBr-3 cell lines, respectively. This activity was similar (albeit less active) to geldanamycin’s EC₅₀ values of 19.9 and 24 nM, respectively (NCI ID: 122750).

In addition to the antiproliferative activity demonstrated by these compounds, the molecules were evaluated for direct binding to Hsp90 isoforms. The affinity of these epoxide analogues was determined for the individual isoforms of Hsp90 to establish whether any selectivity resulted from this modification. Fluorescence polarization assays have revealed geldanamycin to exhibit a high affinity toward three of the four isoforms (Hsp90α, Hsp90β, Grp94), as compared to the mitochondrial localized isoform, TRAP1.¹⁵ Therefore, the affinity of these molecules for Hsp90α, Hsp90β, and Grp94 was determined using a competition fluorescence polarization assay. The results are summarized in Table 1.

Table 1. Epoxy Analogues Apparent Affinity Data.

	apparent KD (μM)
CMPD	Hsp90α	Hsp90β	Grp94
A	3.41 ± 0.60	1.08 ± 0.073	2.55 ± 0.46
B	0.083 ± 0.029	0.054 ± 0.0001	0.13 ± 0.07
C	0.76 ± 0.26	0.395 ± 0.053	1.10 ± 0.49
D	0.32 ± 0.09	0.24 ± 0.04	0.54 ± 0.09
17-A	>10	>10	>10
17-B	0.90 ± 0.38	0.81 ± 0.09	1.13 ± 0.02

Consistent with the antiproliferative data, compound B displayed a high affinity for all three Hsp90 isoforms. In fact, compound B exhibited a 20-fold higher affinity toward Hsp90 as compared to diastereomer A. Intriguingly, compounds C and D exhibited no antiproliferative activity, but retained submicromolar affinity toward Hsp90.

In order to validate that compound B manifests its antiproliferative activity via Hsp90 inhibition, the ability of compound B to induce client protein degradation was investigated. Client protein degradation occurs as a result of Hsp90 inhibition, because the protein substrate is unable to reach conformation maturation and instead is degraded via the ubiquitin-proteasome pathway. Therefore, SKBr-3 cells were treated with increasing concentrations of compound B, and select Hsp90-dependent client proteins were evaluated via Western-blot analysis (Figure 7). Consistent with Hsp90 inhibition, the EGF receptor and HER2 levels were reduced in a dose-dependent manner. In addition, treatment of cells with geldanamycin is known to induce the heat shock response, which results in the increased expression of Hsp70 and/or Hsp90.¹⁶ As shown in Figure 7, compound B induced a dose-dependent increase in the expression of Hsp70 and Hsp90, consistent with Hsp90 inhibition in the cellular environment.

Figure 7.

Compound B client protein degradation assay. Effect of compound B on the maturation of Hsp90-dependent client proteins. SkBr-3 cells were treated with indicated concentration of compound or vehicle for 24 h. Cell lysates were evaluated for indicated protein levels via Western-Blot analysis. Representative actin loading control.

Historically, strategies to diversify geldanamycin have focused on the quinone moiety. However, these results suggest that it may be advantageous to develop analogues of geldanamycin through innovative chemical diversity methods such as that described herein, which provided access to analogues that were otherwise unobtainable.

Conclusions

These studies provide significant evidence that peptide modified catalytic functionalities can deliver dramatic reversals of stereo-, site-, and functional group selectivity in complex molecular settings. Complementary to studies of site-selective natural product diversification in settings where explicit targeting strategies may be derived, for example, the glycopeptide antibiotics,³⁶⁻³⁹ this report shows the generality of these ideas for application to both a new natural product scaffold (geldanamycin) and to a reaction class that might have seemed recalcitrant to the approach (functional group-selective epoxidation). Unique chemical discoveries emerged that highlight a potentially fertile intersection for discovery when the domains of complex catalysts and complex substrates are compelled to interact. Certainly in this case, the new products formed, and possibly the selective formation of previously unassigned geldanamycin oxides, would have been elusive absent the diversification of the geldanamycin scaffold with catalysts targeting outer sphere interactions.

Selective access to samples of A, B, C, and D also delivered new information about the biological potential resident in these new derivatives and present an enigma worthy of further study—that the quinone analogues C and D maintain Hsp90 affinity, while losing antiproliferative effects.⁴⁰ Accordingly, to the extent that these chemical studies unveiled previously unknown and resolved previously unassigned, geldanamycin analogues, perhaps these studies have also added incentive to search for additional presently cryptic catalyst-enabled structures, creating opportunities for resolving additional cryptic biological ambiguities.

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acscentsci.0c00024.

• General information. Peptide synthesis. Peptide screening conditions and results. Large scale reaction conditions. Liquid chromatography chromatograms. Model quinone synthesis and reactions. Biological testing. NMR spectra. X-ray crystallography data (PDF)

• Crystallographic information files (CIF1, CIF2, CIF3, CIF4)

The authors declare no competing financial interest.