n→π* Interactions Modulate the Disulfide Reduction Potential of Epidithiodiketopiperazines

Abstract

Epithiodiketopiperazines (ETPs) are a structurally complex class of fungal natural products with potent anticancer activity. In ETPs, the diketopiperazine ring is spanned by a disulfide bond that is constrained in a high-energy eclipsed conformation. We employed computational, synthetic, and spectroscopic methods to investigate the physicochemical attributes of this atypical disulfide bond. We find that the disulfide bond is stabilized by two n→π* interactions, each with large energies (3–5 kcal/mol). The n→π* interactions in ETPs substantially decrease the disulfide reduction potential, endowing stability in physiological environments in a manner that impacts their biological activity. These data reveal a previously unappreciated means to stabilize a disulfide bond and highlight the utility of the n→π* interaction in molecular design.

Graphical Abstract

INTRODUCTION

Organisms are engaged in an incessant race to evolve strategies against selective pressures.¹ Among fungi such as Chaetomium spp., the strategy to avoid predation or abate competition is manifested in the form of natural products such as epithiodiketopiperazines (ETPs).² ETPs comprise a structurally diverse and biologically active family of fungal alkaloids characterized by a disulfide (or polysulfide) that bridges a 2,5-diketopiperazine (Figure 1A).^3,4 The combination of the unique and challenging molecular architecture of ETPs^3,4 and their potent biological activity² has captured the attention of scientists across a wide range of disciplines.^5–7

Figure 1.

Properties of ETPs. (A) Chemical structures of representative ETPs. (B) Depiction of the biological origin and cellular entry of an ETP. (C) Graph of the distribution of cystinyl C_β–S_γ–S_γ′–C_β′ dihedral angles in high-resolution (<3.0 Å) protein crystal structures (black; Table S1), dependence of the energy of the disulfide bond in dimethyl disulfide on the C–S–S–C dihedral angle (red; Table S2), and range of C–S–S–C dihedral angles in crystalline ETPs (blue; Table S3). (D) Overlap of the n_S,p and π_C=O* orbitals (green) in a model ETP derived from N,N-dimethylalanine. Calculations were at the M06–2X/6–311+G(d,p) level of theory.

Structure–activity relationships have revealed that the activity of ETPs is dependent upon reduction of the epidisulfide bond (Figure 1B) that spans the diketopiperazine (DKP) ring.^4,5,8 Unlike those in proteins, the disulfide bonds in ETPs are locked in an eclipsed conformation (Figure 1C).⁹ In dimethyl disulfide, a prototypical disulfide, values for the C–S–S–C dihedral angle near 0° correspond to ~10 kcal/mol of strain energy (Figure 1C). We reasoned that a compensatory force must exist within ETPs to ameliorate the instability imposed by the eclipsed conformation.

Here, we report on the physicochemical underpinnings of the disulfide bond of ETPs. We began by considering ETP natural products and used quantum mechanical methods to search for the origin of the stability of the disulfide bond. These investigations revealed that n→π* interactions¹⁰ that arise from the overlap of the p-type lone pairs of the sulfur atoms with the π* orbitals of the amide carbonyl groups (Figure 1D) are an integral component of ETP alkaloids. We then synthesized and structurally analyzed a series of C4-substituted bisprolyl-ETPs that were designed to manipulate the energetics of the n→π* interaction, and we measured the reduction potential of these synthetic ETPs. We find that C4-substitution perturbs the n→π* interaction and correlated parameters, including the disulfide reduction potential. Our data support a role for n→π* interactions in tuning the reduction potential of ETPs and thus their biological activities.

RESULTS AND DISCUSSION

n→π* Interactions in ETPs.

To investigate the chemical forces that stabilize the strained disulfide bridge, we examined a structurally diverse catalog of natural ETPs with known crystal structures. Untethered disulfide bonds prefer a C–S–S–C dihedral angle near |θ| = 90° (Figure 1C).¹¹ In ETPs, the value of |q| is <20° (Figure 2A), which is near an energy maximum (Figure 1C). We suspected that the strength of the two n→π* interactions in an ETP could compensate for the strain energy of its eclipsed disulfide bond and that evidence for this compensation would be apparent in the X-ray structures of the natural products. In the 1970s, pioneering crystallographic analyses of Bürgi and Dunitz revealed that the optimal angle for nucleophilic attack at a carbonyl group occurs at ~107°.¹² Approach at other angles leads to less efficient orbital overlap, resulting in a smaller donation of electron density. We find that values of the S···C=O angle in ETPs are indeed close to the Bürgi–Dunitz trajectory, having a range of 119°–128° (Figure 2B).

Figure 2.

Graphs showing measured and calculated parameters of natural ETPs. (A) C–S–S–C dihedral angles. (B) Angle of the sulfur donor to the carbonyl acceptor. (C) Cumulative energy of n→π* interactions versus sulfur–sulfur bond length (R² = 0.52). (D) Energy of an n→π* interaction versus its sulfur to carbonyl-carbon distance (R² = 0.53). Energies on the ordinate were computed at the M06–2X/6–311+G(d,p) level of theory. Chemical structures and data are in Tables S3 and S4.

Next, we used quantum chemistry to investigate the chemical forces that stabilize the strained disulfide bridge of ETPs. We found that the disulfide 3p lone pairs engage intimately with the amide carbonyl groups of the diketopiperazine scaffold. We then employed second-order perturbation theory calculations within the Natural Bond Orbital theory formalism to investigate the strength of the interaction.¹³ We found that n→π* interactions between the sulfur 3p lone pair (n) and carbonyl π antibonding orbital (π*) have energies of 3–5 kcal/mol (Table S4). Compared to n→π* interactions studied previously in proteins (0.3–0.7 kcal/mol), these n→π* interactions are among the strongest observed to date.¹⁴

Anticipating that smaller disulfide bond lengths, r_S–S, would be consistent with reduced electron–electron repulsion between the 3p orbitals of the two sulfur atoms due to donation of electron density from each n_S,p into the π* orbital of a carbonyl group, we looked for evidence of n→π* interactions in the structure of natural ETPs. Specifically, we measured values of r_S–S in natural ETPs, calculated values of ΣE_n→π*, and found a correlation (Figure 2C). Likewise, measured values of the donor–acceptor distance, d_S···C=O, decrease as values of E_n→π* increase (Figure 2D). Although neither of these correlations is strong, they are consistent with an n→π* interaction.

Design of ETP Model Systems.

To examine the physicochemical underpinnings of the ETP substructure in greater detail, we designed a symmetrical ETP model that reduces the complexity of the disulfide exchange equilibria while providing opportunities to examine the impact of substituents on the disulfide bond. We envisioned bisprolyl-ETP 5a (Figure 3A), and the C4-substituted derivatives 5b–e as an ideal platform for rigorous structural and physicochemical analyses.¹⁵ Substitution of pyrrolidine rings at C4 is known to influence ring puckering via a gauche effect (Figure 3B). Specifically, R-configured electron-withdrawing groups at the C-4 position in a proline residue favor C4-exo ring puckering, and S-configured groups favor C4-endo ring puckering.¹⁶ Thus, we chose to introduce flouro, hydroxy, and acetoxy groups in R or S configurations at C4 of bisprolyl-ETPs, giving rise to compounds in which the 4-substituent and disulfide bond are on the same face of the fused rings (cis) or on opposite faces (trans). Our calculations revealed that, as intended, substitution at the C4-position induces conformational changes in ETP 5 that, in turn, modulate orbital overlap and thus the energy of n→π* interactions (Figures 3C and 3D). In particular, the cis and trans configurations differ by ~1 kcal/mol.

Figure 3.

Design principles for ETP model systems. (A) C4-substituted series of symmetric bisprolyl-ETPs 5. (B) Newman projection of ETPs illustrating the gauche effect in cis and trans configurations. (C) Symmetrical n→π* interactions in model ETPs and their calculated energies. (D) Overlap of n_S,p and π_C=O* orbitals in bisprolyl-ETP 5a. Calculations were at the M06–2X/6–311+G(d,p) level of theory.

Synthesis of C4-Substituted Bisprolyl-ETPs.

As described above, we pursued the synthesis of the cis- and trans-C4-substituted bisprolyl-ETPs 5 (Figure 3A) to evaluate the structural parameters that modulate the reduction potential of the disulfide bond. The unsubstituted bisprolyl-ETP 5a (Figure 2, C4-H) had been used previously as a reference in cytotoxicity studies.^15,17

The general approach we used to access bisprolyl-ETPs 5 is outlined in Scheme 1. We envisioned that dithiepanethione 4 could serve as an effective precursor to ETP 5 as we have demonstrated en route to related systems.^4,7 We anticipated that the late-stage introduction of the trithiocarbonate to the bisprolyl-framework could provide access to both cis- and trans-dithiepanethiones 4 from a common precursor dioxasilane 3 (Scheme 1). The silyl bridge of dioxasilane 3 derived from silylation of DKP-diol 2 was anticipated to offer superior solubility in organic solvents, facilitating the sulfidation step. We expected that DKP-diol 2 could be prepared using our permanganate-mediated dihydroxylation¹⁸ of DKP 1, which itself is a cyclodipeptide derivative of the commercially available and inexpensive trans-4-hydroxy-L-proline. As described below, we found this synthetic approach to be advantageous in accessing C4-substituted derivatives through late-stage diversification.

Scheme 1.

Retrosynthesis of Designed Bisprolyl-ETPs (5)

As illustrated in Scheme 2, we targeted cis- and trans-C4-silyloxy ETPs 5c as common intermediates en route to other C4-substituted epidisulfides 5 (Figure 3A). Our synthesis commenced with the known silylation of trans-C4-hydroxy-L-proline followed by aryl boronic acid-catalyzed dehydrative-dimerization to corresponding DKP 1.¹⁹ The permanganate-mediated hydroxylation^4,18 of DKP 1 using bis(pyridine)silver(I) permanganate²⁰ in a pyridine–α,α,α-trifluorotoluene mixture (1:1) afforded DKP-diol 2 in 40% yield.²¹ We found that exposure of diol 2 to monosodium p-methoxybenzyl trithiocarbonate, a reagent for cis-sulfidation of DKP-diols,^8b and trifluoroacetic acid in dichloromethane led to the formation of cis-and trans-dithiepanethiones 4 (d.r. 2.2:1) in 61% and 19% yield, respectively. Nonetheless, as shown in Scheme 2, we found it advantageous to utilize dioxasilane 3 as the substrate for the sulfidation reaction since it afforded the cis-and trans-dithiepanethiones 4 (d.r. 1.4:1) in 51% and 39% yield, respectively. These dithiepanethiones were readily separated and efficiently converted to the corresponding cis-and trans-C4-silyloxy epidisulfides 5c in 90% and 71% yield, respectively, upon aminolysis followed by oxidative disulfide formation with potassium triiodide.^22,4

Scheme 2.

Synthesis of Substituted ETPs (5c)^a

^aConditions: (a) Py₂AgMnO₄, Pyridine, PhCF₃, 23 °C, 2 h 20 min, 40%. (b) i-Pr₂SiCl₂, NEt₃, DMAP, DMF, 0 → 23 °C, 95%. (c) sodium p-methoxybenzyl trithiocarbonate, TFA, CH₂Cl₂, 23 °C, 1.75 h, 51% (cis-4) + 39% (trans-4). (d) ethanolamine, acetone, 0 → 23 °C, 30 min; then KI₃, pyridine, CH₂Cl₂, 90%. (e) ethanolamine, acetone, 0 → 23 °C, 45 min; then KI₃, pyridine, CH₂Cl₂, 71%. Si = Si(Ph)₂t-Bu.

Having secured access to epidisulfides 5c, it was necessary to determine their relative and absolute stereochemistry prior to advancing towards the water-soluble derivatives for our study (vida infra). As shown in Scheme 3, cis- and trans-C4-silyloxy ETPs 5c were desilylated upon exposure to hydrogen fluoride in pyridine–THF mixture (1:9) to give cis- and trans-C4-OH ETPs 5b in 81% and 82% yield, respectively. We were able to obtain crystals of cis- and trans-C4-OH ETPs 5b suitable for X-ray diffraction by slow evaporation from dichloromethane–methanol (10:1). We found structural information such as S–S bond length and S···C=O angle of approach observable in the solid state corelated to both the strength of n→π* interactions as well as the reduction potential of C4-substituted bisprolyl-ETPs (vida infra). Subsequent treatment of cis- and trans-C4-OH ETPs 5b with acetyl chloride and pyridine in dichloromethane led to the isolation of cis- and trans-C4-OAc ETPs 5d in 78% and 92% yield, respectively.

Scheme 3.

Synthesis of C4-Substituted ETPs 5b and 5d^a

^aConditions: (a) HF-pyridine, pyridine, THF, 0 → 23 °C, 40 h, 81%. (b) AcCl, pyridine, CH₂Cl₂, 0 → 23 °C, 6 h, 78%. (c) HF-pyridine, pyridine, THF, 0 → 23 °C, 18 h, 82%. (d) AcCl, pyridine, CH₂Cl₂, 0 → 23 °C, 8 h, 92%. In the ORTEP representation of ETPs 5b, the thermal ellipsoids are drawn at 50% probability.

We were intrigued by the possibility of late-stage introduction of the fluoro-substituents to access cis- and trans-C4-F ETPs 5e by deoxofluorination of the available ETPs 5b–d.²³ Yet, despite a wealth of precedent for stereoinvertive deoxofluorination on C4-OH-substituted proline derivatives,²⁴ there are no reported examples of deoxofluorination in the presence of a disulfide.²⁵ We found that exposure of cis-C4-silyloxy ETP 5c to triethylamine trihydrofluoride (10 equiv) and triethylamine (5 equiv) in dichloromethane at 23 °C for 21 h afforded cis-C4-OH ETP 5b, and that exposure of the reaction mixture to morpholinodifluorosulfinium tetrafluoroborate (XtalFluor-M, 4 equiv) at −78 °C, followed by warming, provided ent-trans-C4-F ETP 5e in 4% yield. Similarly, exposure of trans-C4-silyloxy ETP 5c to identical conditions gave ent-cis-C4-F ETP 5e in 7% yield. Alternatively, we sought to pursue a complementary approach to C4-F-ETPs 5e via the use of the corresponding C4-F DKP 6 (Scheme 4).²⁶ Condensation of N-Boc-trans-4-F-L-proline with trans-4-F-L-proline-OMe hydrochloride²⁶ followed by trifluoroacetic acid-promoted cyclization of the resulting dipeptide gave the desired DKP 6 in 87% yield.²² Permanganate-mediated dihydroxylation of DKP 6 was not optimal due to incomplete dihydroxylation and complications arising from the water solubility of the resulting DKP-diol.

Scheme 4.

Synthesis of cis- and trans-C4-F ETPs 5e^a

^aConditions: (a) NaHMDS, S₈, THF, 23 °C, 2 h; NaBH₄, THF, EtOH, 0 → 23 °C, 2 h; KI, I₂, pyridine, DCM, 23 °C, 5 min, 19% (cis-5e) + 5% (trans-5e). In the ORTEP representation of ETPs 5e, the thermal ellipsoids are drawn at 50% probability.

Reasoning that the observed challenges were in part due to the inductive influence of the C4-F substituent,^4,18a we examined the use of base-promoted electrophilic sulfidation. Exposure of DKP 6 to a solution of elemental sulfur and sodium hexamethyldisilazide (NaHMDS) in tetrahydrofuran^15,27 followed by sequential reduction with sodium borohydride and oxidation with potassium triiodide gave separable mixtures of cis-C4-F and trans-C4-F epipolysulfides. The polysulfide mixtures were subjected separately to another round of reduction and oxidation to give cis- and trans-C4-F epidisulfides 5e, offering an alternative approach to access ETPs 5e for our planned study. Importantly, we obtained crystals suitable for X-ray diffraction of cis-C4-F epidisulfide 5e by recrystallization from acetone-hexanes (3:1) and of trans-C4-F epidisulfide 5e by slow evaporation of a saturated solution in acetone-hexanes (1:7), providing further opportunities for detailed structural analysis (Scheme 4).

Because complete stereochemical assignment of cis- and trans-ETPs 5b–e is critical to our analysis of the structural features impacting the reduction potential of the epidisulfide-bridge, prior to obtaining the X-ray structure for ETPs 5b and 5e, we conducted detailed nuclear magnetic resonance (NMR) studies of related derivatives (Scheme 5). We anticipated differentiating the diastereomeric pairs of cis- and trans-epidisulfides 5b and 5e by reductive methylation and selective nuclear Overhauser effect (nOe) NMR experiments relative to the C4-α-stereochemistry encoded within trans-4-hydroxy-L-proline.^28,8a Accordingly, the reductive methylation of ETPs 5c and 5e using sodium borohydride and methyl iodide afforded bis(methylthioether) DKPs 7 and 8, respectively.²² The stereochemistry of the C2 methyl sulfide was secured via nOe correlations from the S-methyl to the C4-stereocenter through the C3H_α/β and C5H_α/β protons as illustrated in Scheme 5. Thestereochemical assignments of cis- and trans-sulfides 7 and 8 were consistent with the X-ray crystal structures of the corresponding cis- and trans-ETPs 5b and 5e, respectively.

Scheme 5.

Stereochemical Assignment of Sulfides 7–8

Gauche Effect Modulates n→π* Interaction in Model ETPs.

With the model compounds in hand, we were poised to deconvolute relationships between the structural and physicochemical properties of ETPs. Consistent with our hypothesis about the role of n→π* interactions in ETPs, we found that trends in the synthetic bisprolyl-ETPs mirror those of natural ETPs (cf.: Figures 2 and 4). The most striking trend is depicted in Figure 4D, where a change in S···C=O distance of a range of 0.07 Å corresponded with a change in E_n→π* interaction of 2 kcal/mol.

Figure 4.

Graphs showing measured and calculated parameters of synthetic ETPs with known crystal structures. (A) C–S–S–C dihedral angles (○). (B) Angle of the sulfur donor to the carbonyl acceptor (○). (C) Cumulative energy of n→π* interactions versus sulfur–sulfur bond length (R² = 0.82). (D) Energy of an n→π* interaction versus its sulfur to carbonyl-carbon distance (R² = 0.90). In panels A and B, data for natural ETPs (●) are shown again for comparison. Energies on the ordinate were computed at the M06–2X/6–311+G(d,p) level of theory. Data are listed in Table S5.

Crystallographic Signature of n→π* Interactions.

Following the precedent of Bürgi and Dunitz,¹² the most compelling experimental signature of an n→π* interaction has become the pyramidalization of the acceptor carbonyl group towards the electron-pair donor.²⁹ That signature is evident in the crystal structures of both ETP natural products and model ETPs 5b and 5e (Figure 5). Moreover, the interaction is coupled with hydrogen bonding in the crystal structures of gliotoxin, sporidesmin A, and epicorazine, which can lead to aberrant pyramidalization.

Figure 5.

Graph showing two measures of the pyramidalization of carbonyl-group acceptors toward sulfur donors in natural (●) and synthetic (○) ETPs in known crystal structures. Values of Δ were determined with the CCDC program Mercury. Value of χ_C were determined from the ω₁ and ω₂ dihedral angles with the equation: χ_C = ω₁ – ω₂ + π(mod 2π).³⁴ The origin of outlying points is indicated; red: carbonyl groups that accept hydrogen bonds in the crystal structure.

Disulfide Photophysics Enables Electrochemical Measurements.

The C–S–S–C dihedral angle (θ) correlates with the UV–vis absorption maximum of the disulfide.¹¹ For example, oxidized lipoic acid (E°′ = −288 mV³⁰) has C–S–S–C dihedral angle near θ = 45° and an absorption maximum of 330 nm (Figure S1),¹¹ which is distinct from those of ETPs (Table S6), though there is some spectral overlap (Figures S2 and S3). Using this assay, we were able to detect the concentration of lipoic acid in an ETP ⇄ lipoic acid equilibrium to a concentration as low as 1 μM.

We found that the reduction potentials of synthetic ETPs range from −221 to −267 mV (Table 1). Thus, each synthetic ETP would be reduced nearly completely upon cytosolic entry. The stability afforded by two n→π* interactions decreases the ETP E°′ values substantially (Table 1), enabling the disulfide bond to remain intact in a wide range of physiological environments, including the endoplasmic reticulum (which has E°′ ≈ −210 mV^33,32c). In the cytosol, however, the high ratio of reduced glutathione to oxidized glutathione leads to a reduction potential of E°′ ≈ −320 mV.³² Hence, the measured reduction potentials indicate that the n→π* interactions in ETPs are responsible for a balance between extracellular stability and intracellular activity. Without its n→π* interactions, the disulfide bond in an ETP would be much less stable than, for example, that in dithietane (C₂H₄S₂; E°′ = −239 mV³⁰), which has a disulfide bond within a 4-membered ring. Conversely, if the n→π* interactions were too strong, intracellular reduction to the active bisthiol form would not occur.

Table 1.

Reduction Potentials (E°′) of Synthetic ETPs

ETP	E°′ (mV)a	E°′−n→π* (mV)b
C4-H (5a)	−254 ± 5	—
cis-C4-OH (cis-5b)	−244 ± 7	−57 ± 7
trans-C4-OH (trans-5b)	−242 ± 6	−69 ± 6
cis-C4-OAc (cis-5d)	−230 ± 4	—
trans-C4-OAc (trans-5d)	−267 ± 3	—
cis-C4-F (cis-5e)	−264 ± 3	−121 ± 3
trans-C4-F (trans-5e)	−221 ± 5	−93 ± 5

^aValues (± SD) of E°′ were derived from the thiol and disulfide concentrations of a solution equilibrated with reduced and oxidized lipoic acid.³¹

^bValues of E°′_−n→π* were calculated at 25 °C from the measured values of E°′ and the calculated values of ΣE_n→π* for synthetic ETPs with known crystal structures (Figure 4C).

The n→π* interactions within ETPs make their reduction potential responsive to the environment. Specifically, n→π* interactions to a carbonyl group are stronger in protic environments because a hydrogen bond polarizes the carbonyl group, making it a superior acceptor of an n→π* interaction.³⁵ Thus, we anticipate that the disulfide bond in an ETP will be more vulnerable to reduction in a hydrophobic or otherwise desolvated environment, such as the ligand-binding site of a protein or the active site of an enzyme.

Energetic Basis for ETP Electrochemical Equilibria.

Finally, we sought to understand the reduction potentials of ETPs in light of the strain in their disulfide bonds. We found a correlation between two energies: the reduction potential of the disulfide bond and its UV–vis absorption maximum (Figure 6). Specifically, the reduction potential of an ETP is larger (which is indicative of being more easily reduced to the bisthiol) when the absorption maximum is larger (which correlates with a more eclipsed C–S–S–C dihedral angle¹¹). To our knowledge, this is the first example of a correlation between these two manifestations of energy: E°′ ∝ λ_max. This correlation is consistent with the intrinsic stability (Figure 1C) and photophysics¹¹ of disulfide bonds.

Figure 6.

Correlations between the absorption maxima and standard reduction potential of the disulfide bond in synthetic ETPs 5e. (A) Graph showing the absorption maximum and reduction potential. Values of λ_max were measured experimentally (black; R² = 0.53) or calculated with the M06–2X (blue; R² = 0.62) or PBE0 (red; R² = 0.62) DFT functionals. Data are listed in Table S6. (B) Natural transition orbitals for the S₀→S₁ transition of cis-C4-F ETP 5e (top) and trans-C4-F ETP 5e (bottom).

To investigate the electronic basis of the relationship between E°′ and λ_max, we computed the natural transition orbitals (NTOs) of synthetic model ETPs 5. At such eclipsed C–S–S–C dihedral angles, the ground state (S₀) is composed primarily of p-type lone pair density (Figure 6B).¹¹ Promotion of electrons to the first excited state (S₁) populates an NTO with antibonding character, consistent with the observed correlation of E°′ and λ max (Figure 6A).

CONCLUSIONS

We discovered an important aspect of ETP natural products: strong n→π* interactions in which electron density is donated from the sulfur atoms of the disulfide bond into the carbonyl groups of the diketopiperazine. Two strong n→π* interactions in each ETP nearly completely compensate for the ~10 kcal/mol of instability imposed by the eclipsed conformation of the disulfide bond. This discovery could elucidate structure–activity relationships of ETPs and inform the design of new ETPs with desirable properties.