Practical synthesis of the therapeutic leads tigilanol tiglate and its analogues

Results and discussion

Anticipating that esters at C12 and C13 could be exchanged by late-stage diversification and would minimize interference with B-ring modifications, we opted to start with the simple diacetate 3, which is prepared from phorbol (2) with an 82% yield via t-butyldimethylsilyl protection at C20 followed by acetylation at C12 and C13 and a desilylative workup (Fig. 3). Efforts to convert this two-step process into one step via triacetylation and selective deprotection of the C20 acetate gave lower yields (~65%) and was not utilized on large scale.

Chemo-, regio- and stereoselective oxidation at C5 of diacetate 3 in the presence of potentially oxidizable allylic sites at C8, C10, C19 and C20 was efficiently realized using a photosensitized singlet oxygen ene reaction with Rose bengal as the photosensitizer, green light-emitting diodes (λ = 535 nm) as the photon source^39,40 and methanol-d₄ as the solvent, which minimizes singlet oxygen destruction⁴⁰. In situ reduction of the resultant hydroperoxide initially produced allylic alcohol 4 in moderate yield (66%). Although this reaction can be routinely performed batch-wise on small scales (<500 mg), large-scale batch reactions suffered from light penetration issues and raised concerns about the accumulation of the potentially unstable hydroperoxide intermediate³⁹. To address these scalability problems, we assembled a cyclic flow photoreactor that utilized a peristaltic pump and Tygon tubing (Supplementary Figs. 3 and 4)⁴¹. Using this apparatus, we produced the ene product 4 on a decagram scale (for example, 19 g) in an 88% yield as determined by quantitative NMR. Although further purified for characterization purposes, this compound was sufficiently pure to be used directly in the following step thereby avoiding chromatographic purification.

It was envisioned that 4 could be converted into the C5 alcohols 6 or 7 via rhenium-catalysed allylic transposition^42,43. However, the reaction of 4 using literature conditions was sluggish and provided only minor amounts of the undesired C5α-hydroxy-C6,C20 alkene. As an effective alternative route to 6, we found that epoxidation of the C5,C6 alkene in 4 with m-chloroperbenzoic acid (mCPBA) proceeded preferentially from the sterically more accessible β-face to give epoxide 5, with the desired C5β-O bond, in 77% yield. N-methylimidazole (NMI)-catalysed chemoselective tosylation of the primary C20 alcohol and subsequent reaction with sodium iodide gave exclusively the desired β-C5 alcohol 6 in 88% yield⁴⁴.

On treatment with catalytic Re₂O₇, the bis-allylic alcohol 6 underwent a highly chemoselective 1,3-allylic alcohol transposition^42,43 to afford C5β-hydroxy phorbol diacetate 7 in a 76% yield (90% based on recovered 6), which serves as a diversification node^21,24 to access unexplored B-ring analogues of 1. Other rhenium catalysts led to complex mixtures or a lower conversion (Supplementary Table 1).

Subsequent epoxidation of the C6,C7 alkene of 7 occurred only on the sterically more accessible (undesired) β-face, as expected from our previous work^45,46. Although chiral catalyst-controlled epoxidation might address this selectivity problem⁴⁶, we found a more effective solution; specifically, the facial selectivity exhibited by 7 can be dramatically reversed by the conversion of 7 into its acetonide 8 (92%). Models suggest that the acetonide between the C5 and C20 alcohols induces a conformational change in the B-ring that makes the β-face more sterically encumbered and the α-face less so. Additionally, this protection of the C5 and C20 alcohols serves to simplify subsequent functionalization of the C12 and C13 alcohols. Although the initial epoxidation of acetonide 8 with mCPBA under a variety of conditions (Supplementary Table 2) was slow and low yielding, we found that treatment with the sterically smaller and more reactive dimethyldioxirane (DMDO) stereoselectively gave α-epoxide 9 in a 63% yield. This substrate-controlled facially selective epoxidation is unprecedented for this class of compounds and thus provides a potentially general method to access other structurally similar and biologically active tigliane and daphnane diterpenoids^30,34.

Deacetylation of diester 9 provided the corresponding C12,C13 diol 10 in an 86% yield as determined by quantitative NMR. Although further purified for characterization purposes, this compound was sufficiently pure to be used directly in the following step thereby avoiding chromatographic purification. Diol 10 serves as a second point of diversification for C12,C13 derivatization, now with the desired C5β-hydroxy-C6α,C7α-epoxy B-ring in place^11,24,36. From 10, EBC-46 was prepared on gram scale via selective diesterification⁴⁷ and acidic deprotection of the acetonide. In our laboratory, this overall route and greatly improved final esterification sequence delivered over 2.5 g of EBC-46 (Supplementary Fig. 7). All the steps were performed by two or more investigators to ensure reproducibility. Collectively, our synthetic strategy provides access to B-ring analogues from intermediate 7, A-ring analogues from intermediates 7–12 and C-ring analogues from intermediate 10.

Although a comprehensive analysis of the binding, selectivity and biological activities of various analogues will be disclosed separately, it is noteworthy that even modest structural changes dramatically affect PKC affinity and selectivity. To begin our investigation into the role of the C5β-hydroxy-C6α,C7α-epoxy functionality and the C12,C13 esters in determining PKC affinity and selectivity, we prepared an initial series of analogues (Fig. 4a). These analogues, along with EBC-46, were tested for their affinity to PKC-β_I and PKC-θ, representative conventional and novel isoforms of PKC, respectively (Fig. 4b). Specifically, to determine the role of the C6α,C7α-epoxide in PKC binding and selectivity, we synthesized a C6,C7-alkene analogue (13, SUW400), otherwise inaccessible from EBC-46. Interestingly, this analogue exhibited a nearly identical binding affinity and selectivity to PKC-β_I and PKC-θ when compared with that of EBC-46. This finding suggests that the C6α,C7α-epoxide is not necessary for the isoform-selective binding exhibited by EBC-46. Similarly, to determine the role of the C5β-alcohol in the PKC binding and selectivity, we synthesized a C5-deoxy-C6,C7-alkene analogue (15, SUW402). This analogue showed a stronger but less selective binding than that of both EBC-46 and SUW400. This finding suggests that the C5β-alcohol plays an important role in isoform binding selectivity. Finally, to begin investigating the role of the C12,C13 esters in PKC binding and selectivity, we synthesized a diacetate analogue (14, SUW401). This analogue showed a very low PKC binding affinity when compared with that of EBC-46. This finding suggests that the C12,C13 esters also play an important role in PKC binding. Given the potent PKC affinity of EBC-46, SUW400 and SUW402, these compounds were tested in vitro for their ability to permeate CHO-K1 (Chinese hamster ovary factor K1) cells and translocate, in real time, an optically tagged PKC fusion protein (PKC-GFP; GFP, green fluorescent protein) from the cytosol to the membrane–the hallmark of PKC activation¹ (Supplementary Fig. 8). The details and experimental procedures for this assay were published previously⁴⁸. EBC-46 showed a modest translocation of PKC-β_I-GFP at low (200 nM) concentrations and a robust translocation at high (1,000 nM) ones (Fig. 4c). Interestingly, the more synthetically accessible SUW400 and SUW402 showed a comparable translocation to that of EBC-46 at low (200 nM) as well as high (1,000 nM) concentrations. Future studies on these and other analogues, readily accessible from our synthetic route, are directed at the elucidation of the structural basis for isoform-selective PKC modulation and the role of isoform selectivity in human immunodeficiency virus and AIDS latency reversal, tumour ablation, antigen enhancement for antigen-targeted antibody and chimeric antigen receptor cell therapies, suppression of T-cell exhaustion and neurological disorders.

In summary, we describe a scalable laboratory preparation of tigilanol tiglate (1, EBC-46), an approved veterinary therapeutic and a human clinical lead for cancer and other indications. Previously, tigilanol tiglate was considered synthetically inaccessible and only available from a limited natural source, the latter raising environmental concerns. Our synthetic strategy also enables access to numerous biologically active tiglianes, daphnanes and their analogues. This strategy will accelerate future studies directed at the structural basis for PKC isoform selectivity and its role in mode of action and disease-specific activities.

Methods

As the hazard of new compounds is unknown, all the procedures were conducted with full personal protective equipment in a way that avoids exposure. CHO-K1 (ATCC) was the cell line used for translocation experiments. No commonly misidentified cell lines were used in this study. None of the cell lines used were authenticated. No statistical methods were used to predetermine sample sizes, but our sample sizes are similar to those reported in previous publications⁴.

Supplementary Material

Data availability

The data supporting the findings of this study are available within the article and its Supplementary Information. The X-ray structure of phorbol-13-acetate bound to the PKC-C1 domain was obtained from the structure reported by Hurley (Protein Data Bank: 1PTR).