Pre-mRNA Splicing-Modulatory Pharmacophores: The Total Synthesis of Herboxidiene, a Pladienolide-Herboxidiene Hybrid Analog and Related Derivatives

Abstract

Herboxidiene is a natural product that has previously been shown to exhibit antitumor activity by targeting the spliceosome. This activity makes herboxidiene a valuable starting point for the development of anticancer drugs. Here, we report an improved enantioselective synthesis of herboxidiene and the first report of its biologically active totally synthetic analog: 6-norherboxidiene. The synthesis of the tetrahydropyran moiety utilizes the novel application of inverse electron-demand Diels-Alder chemistry and the Ferrier-type rearrangement as key steps. We report, for the first time, cytotoxicity IC₅₀s for synthetic herboxidiene and analogs in human tumor cell lines. We have also demonstrated that synthetic herboxidiene and its analogs can potently modulate the alternate splicing of MDM-2 pre-mRNA.

The prominence of pre-mRNA splicing in gene expression in higher eukaryotes makes it an attractive target for therapeutic intervention and recently this process has emerged as a potentially important therapeutic target in cancer.^1–4 The pre-mRNA splicing process involves the removal of introns (noncoding sequences) from pre-mRNA, which is followed by ligation of exons (coding sequences).¹ Alternative splicing is the mechanism by which different forms of mature mRNAs are generated from the same gene. Commonly, alternative splicing patterns determine the inclusion or exclusion of portions of the coding sequence in the mRNA, giving rise to protein isoforms that differ in their peptide sequence and hence in their chemical and biological activity.¹ Alternative splicing plays important roles in the development of multicellular organisms and in numerous pathologies, including cancer.⁶ This splicing process is catalyzed by a macromolecular complex called the spliceosome, which is composed of five small nuclear ribonucleoproteins (snRNPs) (U1, U2, U4, U5 and U6) and over 150 associated proteins.

Several bacterial natural products such as herboxidiene (1 also known as GEX1A),^7–9 pladienolide B (2),^10–13 FR901464^14–16 and the thailanstatins¹⁷ have been shown to effect splicing by targeting the splicing factor 3b (SF3b) subunit, an essential component of the spliceosome. Several of these natural products also induce cell cycle arrest at the G1 and G2/M phase and show potent antitumor activity in human tumor cell lines.^{11, 14,18} Importantly, several of these natural products have also been reported to show potent in vivo activity in tumor xenograft models.^11,23 Ongoing synthetic work^23–27 have also provided novel synthetic spliceosome modulators such as 1-deoxy FR901464,²³ and the meayamycins^23,24 (all analogs of FR901464), as well as new structure-activity relationships for FR901464 analogs.²³ Indeed, one of the semi-synthetic derivatives of 2 (E7107) has shown remarkable pre-clinical tumor regression efficacy and advanced to Phase I clinical trials;¹² which stimulated considerable interest in the potential of splicing modulators as potential therapeutic agents for the treatment of cancer.^4,27,28 Very recently interesting active-analogs of another natural product (FD-895), which is structurally related to 2, have also been reported.²⁷

RESULTS AND DISCUSSION

We have also been engaged in a successful effort to design new effective and highly active synthetic analogs of FR901464 and 2 by the application of a consensus pharmacophore hypothesis.^29–31 We believed that we could extend our approach by the generation of a new structurally simplified scaffold based on herboxidiene. Thus, using the guidance of our pharmacophore hypothesis, we designed a hybrid molecule (3) from herboxidiene (1) and pladienolide B (2), (Figure 1). This hybrid 3 has the tetrahydropyran core of 1 (red) and side chain of 2 (blue). Based on our pharmacophore model we anticipated that compound 3 could incorporate the correct molecular geometry as well as all of the key SF3B1 interaction features.²⁹

Figure 1.

The structures of splicing modulators: herboxidiene (1), pladienolide (2) and our synthetic hybrid molecule (3)

For the synthesis of the hybrid analog 3 we coupled aldehyde 16b with the Julia-Kocienski reagent 4 that has been used in the total synthesis of pladienolide, which was prepared as reported (Scheme S1).³² We were initially surprised when we tested the cytotoxicity of our hybrid molecule 3 in various human tumor cell lines and it was found inactive (IC₅₀ > 20 MM). In these assays we used natural 1 (prepared by bacterial fermentation, purchased from Cfm Oskar Tropitzsch e.K) as a standard. For comparison we also referred to the reported activity of 2 (IC₅₀s in the nanomolar range).^10,33 Based on these data we hypothesized that the hydrogen-bond donor (OH) at C18 represents an additional new pharmacophore feature in herboxidiene and that this key functionality is required for the activity of this class of compounds, in order to effectively interact with SF3B1 (see Figure 2). To investigate our hypothesis, we next turned our attention towards the synthesis of herboxidiene and its analogs that maintain this hydrogen-bond donor feature.

Figure 2.

The hypothetical pharmacophore features for herboxidiene.⁴ The features are F1: Donor (Don) and acceptor (Acc), F2: Acc, F3: Hydrophobic (Hyd), F4: Acc (epoxide), F5: Acc, and F6: Don & Acc. The pharmacophore is calculated using the Molecular Operating System 2011.10 from the Chemical Computing Group, Inc.

Herboxidiene (1) is a bacterial secondary metabolite, originally isolated from the Streptomyces chromofuscus strain A7847 by Isaac et al and initially evaluated as a herbicide.⁷ Subsequently Edmunds et al. determined the absolute configuration of 1,⁸ which was further confirmed by the first total synthesis by Blakemore and Kocienski.³⁴ Importantly, a study by Horiguchi et al also reported the in vivo antitumor activity of 1 in a murine tumor model.²³ Of particular interest to our lab was the recent photoaffinity-labeling study which found that the SF3b1 (SAP155) protein is also the major target of 1.⁹ Because 1 has a simpler structure, when compared to the other known splicing modulators, it is an especially attractive starting point for the development of structure-activity relationship (SAR) studies. Therefore 1 represents a unique opportunity for the discovery of drug-like synthetic agents for SF3b modulation.

In the realm of herboxidiene total synthesis, impressive progress has been made recently.^35–40 However, to the best of our knowledge, no study on the evaluation of the biological activity of synthetic 1 and its analogs in human cancer cell lines has been reported to date. A recent publication by Koide and co-workers provides the only report on the biological activity of synthetic 1 and this is in a non-tumor (HEK293) cell line.⁴¹ Therefore as a first step in exploring totally synthetic analogs of 1, as potential antitumor therapeutics, we embarked on the total synthesis of the natural product herboxidiene and its close analogs.

Our assembly of 1 and its analogs took advantage of a convergent strategy in which tetrahydropyran cores 16a/16b would be coupled with a herboxidiene side-chain precursor through the versatile Julia-Kocienski olefination reaction. We envisioned the synthesis of aldehydes 16a/16b from 14a/14b, which in turn could be derived via Ferrier-type rearrangement of compounds 9a/9b. We planned to construct dihydropyranones 9a/9b using an inverse electron-demand Diels-Alder reaction between (S)-2-(benzyloxy)propanal (7) and diene 8a or 8b. The sulfone 19 can be derived from alcohol 18, which is conveniently prepared by an efficient method reported by Urpi and co-workers.³⁵

Our tetrahydropyran synthesis commenced with the preparation of diene 8a/8b (as a 10:1 mixture of Z and E isomers) using a method reported by Li et al.⁴³ The MgBr₂ mediated inverse electron-demand Diels-Alder reaction of diene 8a or 8b and (S)-2-(benzyloxy)propanal (7) then provided dihydropyranones 9a and 9b as single diastereomers.⁴² The reduction of the ketone moiety was achieved under Luche conditions to selectively provide the alcohols 10a and 10b. The latter in turn were acetylated with Ac₂O and Et₃N to provide the corresponding acetates 11a and 11b. In an effort to obtain the tetrahydropyran core, we initially investigated application of the Ireland-Claisen rearrangement of 11a, which has been used in the construction of analogous tetrahydropyrans.⁴⁴ We anticipated that Ireland-Claisen conditions, followed by esterification with CH₂N₂ would lead to 12a. Unfortunately, in our hands this reaction sequence failed to provide the desired products in acceptable yields. We therefore pursued the alternative route to these intermediates that installed the desired stereochemistry and ester moiety in one step via a Ferrier-type rearrangement using a silylketene acetal.^45–47 The stereochemical outcome of this reaction has previously been shown to be highly dependent on the conditions employed.⁴⁹ We developed conditions for this transformation that provided a 2:1 mixture of cis- and trans-fused substituted dihydropyrans 12a (in 72% yield) and 12b (in 70% yield). This ratio is identical to that reported in a related C-glycolsylation reaction of this type.⁴⁷ We found that cis- and trans- isomers of 12a and 12b were readily separated by flash chromatography to give the stereochemically pure products. The relative stereochemistry of both isomers was established by NOESY experiments and by single crystal X-ray structures of 4-nitrophenyl ester derivatives (see Figure 3 and SI). At this point, the simultaneous reduction of the alkene moiety and benzyl ether cleavage of 12a and 12b was accomplished by catalytic hydrogenation. Next, the oxidation of the resulting alcohols 13a and 13b yielded the corresponding ketones 14a and 14b. Grignard reaction of 14a and 14b with vinylmagnesium bromide then furnished the tertiary alcohols 15a and 15b as a mixture of diastereomers. Finally, PCC oxidation of 15a or 15b afforded the desired aldehydes 16a and 16b as inseparable mixtures of E:Z isomers (7:2). (We found however that the final products could be separated to give pure E derivatives, see Scheme 3 and discussion below). This concise route towards the tetrahydropyran core of 1 represents a significant advancement in the syntheses of the key intermediate 14a.^35–39

Figure 3.

X-ray structure of nitrobenzoyl derivative of 13a.

Scheme 3.

Synthesis of Herboxidiene (1) and its Analogs. Reagents and conditions: a) Diisopropyl azodicarboxylate (DIAD), 5-mercapto-1-phenyltetrazole, PPh₃, THF, rt, 3 h, 89%; b) (NH₄)₆Mo₇O₂₄, 30% H₂O₂, EtOH, rt, 24 h, 88%; c) KHMDS, 16a or 16b, THF, −78 °C, 1.5 h, (E:Z = 7:2); d) 0.16 M HCl, MeOH, e) VO(acac)₂, t-BuOOH, CH₂Cl₂; f) KOSiMe₃, THF; g) Dess-Martin periodinane (DMP), CH₂Cl₂, 55%. PT= phenyltetrazole

We found that the method shown in Scheme 2 could provide compound 14b in 6 steps in an overall yield of 25% (see SI). Our improved route shown in Scheme 2 was also more efficient for the synthesis of aldehyde 16b. In addition to the route shown in Scheme 2 we also successfully implemented a known approach to 14b via oxa-Michael cyclization followed by functionalization from the resulting cis-fused pyran using a published procedure (see SI, Scheme S2).⁴⁸ We were also able to confirm the relative and absolute stereochemistry of 14a and 14b using single crystal X-ray crystallography of derivatives (see SI: Schemes S3, S4, Figure S1 and Figure S2).

Scheme 2.

Synthesis of Tetrahydropyran 16a/16b

Reagents and conditions: a) MgBr₂, THF, rt, 12 h; b) CeCl₃·7H₂O, NaBH₄, MeOH, 30 min; c) Ac₂O, Et₃N, CH2Cl2, 0 °C, 1 h; d) Ti(i-OPr)₂Cl₂, toluene, −78 °C to rt, 16 h, (cis:trans = 2:1); e) H₂, Pd/C, EtOAc, 1 h; f) Dess-Martin periodinane (DMP), CH₂Cl₂, 30 min; g) vinylmagnesium bromide, CH₂Cl₂, −78 °C, 24 h) Pyridinium chlorochromate (PCC), CH₂Cl₂, 40 °C, 18 h, E:Z = 7:2.

We then turned our attention to the acyclic side chain of 1. Recently, Urpi and co-workers reported an efficient method for the synthesis of alcohol 10 starting from the commercially available ester 17 in nine linear steps.²³ Using this procedure, with some minor modifications, we were able to synthesize compound 18 on a multigram scale (see SI). The resulting substrate 18 was subjected to a Mitsunobu reaction with 1- phenyl-1H-tetrazole-5-thiol, followed by ammonium molybdate-catalyzed oxidation, to provide sulfone 19. With sulfone 19 and aldehyde 16a and 16b in hand, we next focused on the convergent coupling to provide 20a and 20b. These products were obtained in moderate yields in the presence of KHMDS via the Julia-Kocienski olefination. Removal of the tert-butyldimethylsilyl group in 20a and 20b with HCl, followed by C18 hydroxyl directed selective epoxidation of alcohols 21a and 21b with catalytic VO(acac)₂ provided the pure diastereomers 22a and 22b in good yields, as expected based on the literature precedence.^35–38 The purification step at this point also readily removed the minor C9,C10-Z isomer as shown in Scheme 3 (see SI), which had been carried over from 16a/16b. Hydrolysis of the methyl esters 22a and 22b under basic conditions afforded pure herboxidiene (1) and C6-norherboxidiene (23). This series of reactions also provided important analogs for SAR studies, such as 21a, 22a and 22b. Additionally we prepared the C18 ketone derivative 24 by oxidation of 22b with DMP, in order to confirm the importance of the C18 hydrogen-bond donor feature in the herboxidiene pharmacophore.

Next, we evaluated the cytotoxicity of both natural and synthetic 1 and its analogs on six human cancer cell lines (Table 1). The IC₅₀ values of synthetic 1 were found to be in the low nanomolar range (4.5–22.4 nM) which was comparable to the results with the natural 1 (4.3–46.3 nM). Surprisingly, the methyl ester precursor of 1 (compound 22a) also displayed highly potent cytotoxicity (IC₅₀= 6.2–15.8 nM). However, a great loss in activity (>200-fold) was observed for triene 21a as compared to its epoxide derivative 22a. This result is consistent with the early reports of the herbicidal activity dependence on an intact epoxide group in several semisynthetic analogs of herboxidiene.⁷ C6-Norherboxidiene (23) showed a loss in activity (~10-fold) indicating that this methyl group is contributing to the potency of 1. Most importantly, when the ketone analog 24 was evaluated, it was found to be essentially inactive, with IC₅₀ values in micromolar range (> 5–25 MM). This low activity of 24 clearly supports our hypothesis that the hydrogen-bond donor (OH) interaction feature at C18 is required for the potency of this class of compounds in addition to the other features previously described (Figure 2).^4,29 This result is also consistent with the lack of activity seen in the hybrid molecule 3.

Table 1.

Cytotoxicity of herboxidiene and its analogs in various human cancer cell lines.

Entry	Cell Linesa	IC50 (nM)b
		Natural 1	Synthetic 1	21a	22a	22b	23	24
1	JeKo-1	4.3 ± 1.5	4.5 ± 0.7	1306 ± 660	12.0 ± 1.8	310 ± 57	55.2 ± 7.2	>4700
2	HeLa	14.7 ± 2.5	6.8 ± 1.4	903 ± 137	11.0 ± 2.3	310 ± 37	74.7 ± 3.8	>7600
3	PC-3	46.3 ± 5.2	22.4 ± 4.2	2022 ± 562	18.9 ± 1.9	266 ± 41	256 ± 35.0	>28000
4	SK-MEL-2	34.0 ± 32.0	14.0 ± 11.0	>1800	20.0 ± 18	1243 ± 187	301 ± 293	>240000
5	SK-N-AS	12.7 ± 2.3	6.5 ± 0.9	859 ± 165	14.7 ± 2.0	646 ± 286	90.5 ± 8.9	>5000
6	WiDr	12.0 ± 1.9	7.2 ± 0.3	1271 ± 594	8.1 ± 1.2	246 ± 42	122 ± 20	>5400

^aCancer Type: JeKo-1 (Mantle cell lymphoma); HeLa (Cervical adenocarcinoma); PC-3 (Prostate adenocarcinoma); SK-MEL-2 (Malignant melanoma); SK-N-AS (Neuroblastoma); WiDr (Colorectal adenocarinoma).

^bCells were treated continuously with the compound for 72 h. Cytotoxicity was determined as the IC₅₀ values calculated from the percentage of viable cells remaining at 72h, measured with CellTiter-Glo reagent. The IC₅₀ values represent the average of three independent determinations ± SE.

Previously we have reported the use of an assay for the evaluation of the efficacy of alternative splicing modulator drugs.^31,49 This assay depends on our observation that the splicing modulator drugs that we have evaluated induce exon skipping in MDM2 pre-mRNA, which results in the formation of shorter isoforms that can readily be detected (see SI, Figure S3). In addition to our investigation of the cytotoxicity of these compounds we also performed this MDM2 alternative splicing assay for compounds 1, 22a, and 23, in order to determine whether these compounds modulate the splicing machinery. This assay was performed as previously described.⁴⁹ The results shown in Figure 4 demonstrate for the first time that both natural and synthetic 1 potently induce MDM2 alternate splicing of MDM2 pre-mRNA. Synthetic 1 appears to be slightly more potent in this induction (Figure 4), we believe this is due to the higher purity of the synthetic material when compared to the commercially available fermentation product (~92% pure). Although 22a and 23 are less potent than 1, it is clear that these two compounds are effective in the induction of alternative splicing of MDM2 pre-mRNA and that the potency of 22a is superior to sudemycin C1. All of these results are consistent with the pharmacophore features shown in Figure 2.

Figure 4.

Herboxidiene and its analogs modulate alternate splicing of MDM2 in Rh18 cells. The cells are exposed to drug for 8 h in these experiments (see SI for assay details).⁴⁹

In conclusion, the flexible and concise syntheses of herboxidiene and its previously unexplored analogs were accomplished by the orchestration of synthetic schemes that build on published chemistry, which include the novel application of the inverse electron-demand Diels-Alder reaction, a Ferrier-type rearrangement as well as the established Julia-Kocienski reaction, as key steps. Initial cytotoxicity assays of the synthetic compounds provided a good initial insight into the pharmacophore features for this class of compounds. We have also demonstrated that synthetic herboxidiene and its potent analogs efficiently modulate the alternative mRNA splicing. This study provides strong evidence that synthetic herboxidiene analogs may serve as good anticancer drug lead candidates in the discovery of new synthetic agents targeting the pre-mRNA splicing process.

METHODS

The methods are reported in the Supporting Information.

Scheme 1.

Synthesis of the pladienolide-herboxidiene hybrid molecule 3.

Reagents and conditions: a) 16b, KHMDS, THF, −78 °C, 1.5 h, 46%; b) TBAF, THF, 76%; c) KOSiMe₃, THF, rt, 79%.