**3. Synthesis and Revision of the Absolute Configuration of the Cytotoxic Mandelalides from the South African Marine Ascidian,** *Lissoclinum* **sp.**

The encrusting colonial didemnid ascidian *Lissoclinum* sp. (Figure 1b) collected by SCUBA from Algoa Bay, on the southeast coast of South Africa, afforded sub-milligram (0.5–0.8 mg) quantities of the glycosylated, polyketide macrolides, the mandelalides A–D (**12a**, **13**–**15**, Figure 4). Mandelalides A and B exhibited potent low nanomolar cytotoxicity (IC50 12 and 44 nM, respectively) against NCI-H460 lung cancer cells [18]. The relative configuration of the macrolide rings in **12a**, **13**–**15** was established through integration of ROESY data with homonuclear (3*J*HH) and heteronuclear ( 2,3*J*CH) coupling constants, while the absolute configurations of **12a** and **13** were extrapolated from the hydrolysis and subsequent chiral GC-MS analysis of the respective monosaccharide residues (2-*O*-methyl-6-dehydro-α-L-rhamnose and 2-*O*-methyl-6-dehydro-α-L-talose). The paucity of **12a**, **13**–**15** isolated from the MeOH-CH2Cl2 extract of *Lissoclinum* sp. (*i.e.*, 0.8 mg of **12a**) and difficulties encountered in the further supply of these compounds from their natural source implied that the synthesis of **12a** (the most active compound in the mandelalide series) would provide sufficient quantities of **12a** to explore the mechanism of *in vitro* cytotoxicity exhibited by this compound. As described below, the synthesis of **12a** and the diastereomer **16** by Willwacher *et al.* and Xu, Ye and co-workers revealed errors in the original assignment of the absolute configuration at positions C17, C18 C20, C21 and C23 and resulted in the correction of the chemical structure of mandelalide A (**12a**) to **16** [19,20].

In 2014, two years after the isolation of the mandelalides was first reported [18], Willwacher and Fürstner reported the first total synthesis of **12a** in a 4.5% overall yield [21]. They also noted the structural similarities between **12a** and madeirolide A (**17**, Figure 4), an equally scarce metabolite previously isolated from a marine sponge, *Leiodermatium* sp. [22,23], and ascribed the absence of anti-proliferative activity against pancreatic cancer cells reported for **17** to the structural differences between these two compounds. Anticipating in their proposed synthesis of **12a** that final closure of the macrolide ring, concomitant with insertion of the Δ<sup>14</sup> *Z*-olefin, could be achieved with ring closing alkyne metathesis (RCAM), Willwacher and Fürstner successfully synthesized the two main building blocks (**18**) and (**19**) emerging from their retrosynthetic analysis of **12a**. Cobalt-catalyzed carbonylative epoxide opening and iridium-catalyzed two-directional Krische allylation were identified as key synthetic steps required for the synthesis of **18** and **19**, respectively, while the RCAM protocol would not have been possible without the use of a highly-selective molybdenum alkylidene complex catalyst; the first time that this catalyst has been successfully incorporated into a natural product total synthesis [21]. Finally, regioselective trimethylsilyl trifluoromethanesulfonate (TESOTf)-catalyzed rhamnosylation of the mandelalide aglycone proceeded smoothly to afford mandelalide A with the chemical structure **12a**, originally proposed by McPhail and co-workers [18].

**Figure 4.** Chemical structures of compounds **12a**–**19**.

Comparing the spectroscopic data acquired for their synthetic product with those of naturally-occurring mandelalide A (**16**), Willwacher and Fürstner noted significant chemical shift and coupling constant differences between the NMR datasets of synthetic **12a** and the natural product (Tables 1 and 2). Initially, given the relative magnitude of the observed differences, their attention was focused on differences in the 13C chemical shift assigned to the C11 methine and the C25 methyl carbon atoms (Table 1) and discrepancies in the <sup>3</sup>*J*11,12 coupling constants (Table 2; *J*11,12 = 9.7 and 7.6 Hz, respectively, for naturally-occurring mandelalide A and **12a**, respectively) associated with the C11 stereogenic center. However, synthesis of the 11-*epi*-diasteromer of **12a** (**12b**) did not provide clarity on the source of spectroscopic differences between the synthetic and natural products, and Willwacher and Fürstner were at a loss to explain where the anomalies resided in the proposed structure of **12a** [21]. Reddy *et al.* [24] also tackled the synthesis of the aglycone of **12a**, reflected in the original structure proposed by McPhail and co-workers [18]. Their 32-step synthesis afforded the putative mandelalide A aglycone in a 6.3% overall yield, and the spectroscopic data acquired for the synthetic product was consistent with the analogous data for the aglycone of **12a** synthesized by Willwacher and Fürstner.


**Table 1.** Comparative 13C NMR data (CDCl3, 175 † and 150 ‡ MHz) reported by McPhail and co-workers [18] for naturally-occurring **16**; by Willwacher *et al.* [20,21] for synthetic **12a**, **12b** and **16**; and by Xu, Ye and co-workers [19] for synthetic **12a** and **16**.


**Table 2.** Comparative 1H NMR data (CDCl3, 700 † and 600 ‡ MHz) reported by McPhail and co-workers [18] for naturally-occurring **16** and by Willwacher *et al.* [20,21] for synthetic **12a** and **16**.

A second synthesis of **12a** by Xu, Ye and co-workers [19] was published in Angewandte Chemie International Edition shortly after Willwacher and Fürstner's synthetic communication appeared in the same volume of the journal. Approaching the synthesis of the 24-membered macrocycle via a different route to that used by Willwacher and Fürstner; Xu, Ye and co-workers initially constructed the two sub-units (**20** and **21**, Figure 5) with Prins cyclization, providing diastereoselective access to the tetrahydropyran moiety in **20** and a Rychnovsky–Bartlett cyclization generating the tetrahydrofuran ring in **21**. As anticipated from the retrosynthetic analysis that guided this synthetic approach, both subunits were successfully assembled into the aglycone via Suzuki coupling and Horner–Wadsworth–Emmons macrocyclization. The synthesis of **12a** was concluded with the addition of a protected rhamnose moiety to the mandelalide aglycone via a Kahne glycosylation reaction followed by a single-step collective removal of the silyl protecting groups. Xu, Ye and co-workers also identified the incompatibility of the NMR datasets acquired for their synthetic product and naturally-occurring mandelalide A, including the significant chemical shift differences associated with the C11 and C25 carbon atoms (Tables 1 and 2). However, from direct comparison of the opposite configurations of the stereogenic centers in the northern hemisphere of **12a** and **17**, they correctly postulated that, assuming **12a** and **17** share a common biogenesis, the differences in absolute configuration would probably be confined to this part of the molecule and not C11, to which an *S*

configuration was assigned in both **12a** and **17**. The convergent synthetic approach to **12a** by Xu, Ye and co-workers enabled them to synthesize the diastereomer, **16**, with opposite configurations at positions C17, C18 C20, C21 and C23 to those initially reported for **12a** [18] and consistent with the configurations assigned to the analogous chiral centers in **17** [19]. Willwacher *et al.* also recently reported a further synthesis of **16** [20] and confidently postulated revised structures for mandelalide B–D **(22**–**24**, Figure 5). Comparison of the NMR data of naturally-occurring mandelalide A with those of **16** (Tables 1 and 2) confirmed that these two compounds were identical and that any uncertainties around the correct structure of mandelalide A had been successfully resolved.

**Figure 5.** Chemical structures of compounds **20**–**25**.

Although the chemical structure of mandelalide A has now been unequivocally established as **16**, the inconsistency in the cytotoxicity data reported for this compound remains unresolved. McPhail and co-workers reported that the naturally-occurring mandelalides A and B possessed potent cytotoxicity against human NCI-H460 lung cancer cells (IC50 12 and 44 nM, respectively) and Neuro-2A neuroblastoma cells (IC50 29 and 84 nM, respectively) [18]. These results are at variance with the reported lack of cytotoxicity exhibited by synthetic **16** when screened against a panel of ten cancer cell lines of different histological origin by Xu, Ye and co-workers [19] (Table 3). Willwacher *et al.* also noted the negligible cytotoxicity of **16** against cancer cell lines with the exception of a single human breast carcinoma cell line [20] (Table 3).

**Table 3.** Comparative IC50 data (μM) reported by McPhail and co-workers [18] for naturally-occurring **16** and by Xu, Ye and co-workers [19] and Willwacher *et al.* [20] for synthetic **16**.

