the correct functionalization pattern to the
B-ring (Fig. 3). This required hydroxylation
of C5 and installation of the fifth quaternary
stereocenter at C8. [4+2]-Photooxygenation
of the diene system in the B-ring with singlet
oxygen gaveendo-peroxide 14 as a single dia-
stereomer. The selectivity in the course of the
[4+2] reaction from 13 to 14 originates from
a preferredexo-attack (front face) of oxygen
versus anendo-attack (back face), together with
complete shielding of theendo-face by the Me–
18-group (marked blue in 13 ; Fig. 3). Cleavage
of theendo-peroxide proved to be challenging:
Kornblum–DeLaMare rearrangement ( 22 )of
theendo-peroxide resulted in poor yields even
upon extensive screening of conditions, where-
as reductive cleavage of the peroxide was ac-
companied by decomposition under a range of
common conditions. After extensive experi-
mentation, we found that treatment of 14 with
2,6-di-tert-butyl-4-methylphenol (BHT) reduc-
tively cleaved theendo-peroxide bond cleanly
to yield 71% of hydroxyenone 15 (the proposed
mechanism is discussed in the supplementary
materials). Compound 15 required inversion
of stereochemistry at C5 to match the config-
uration of the alcohol in canataxpropellane
( 2 ) at this position. For this purpose, selec-
tive allylic oxidation to the ketone at C5 was
performed with 2-iodoxybenzoic acid (IBX)
and subsequent directed reduction of this
ketone with sodium triacetoxy borohydride
[NaBH(OAc) 3 ] assisted by the neighboring
hydroxyl group at C20 proceeded by inter-
mediate 16 (Fig. 3) and established the cor-
rect configuration at C5 in compound 17.
However, additional 1,4-reduction of 16 to
give 18 was observed under these conditions.
In both compounds 17 and 18 ,thestereo-
center at C5 now matched the configuration in
the natural product, as confirmed by single-
crystal x-ray analysis of 17. Compounds 17 and
18 were converged to 19 under the conditions
described below (Fig. 4). Protection of the 1,3-
diol of enone 17 as the benzylidene acetal
followed by 1,4-reduction afforded 19 in 95%
yield (Fig. 4). Ketone 18 was directly protected
as the benzylidene acetal to afford 19 .Elabo-
ration of the B-ring wasaccomplished by con-
verting the ketone at C8 to the vinyl triflate
with Comin’sreagent(A) and subsequent
palladium-catalyzed carboxymethylation, af-
fordinga,b-unsaturated ester 20 in 77% yield.
Dissolving metal reduction with Mg in meth-
anol gave the corresponding saturated ester,
and subsequenta-alkyation with methyl iodide
exclusively installed the desired configuration
at the quaternary stereocenter of C8 to give
21 in 83% yield. The stereoselective intro-
duction of this methyl group is attributed to
an attack from theexo-side (see Fig. 3). Up to
this point, 1 g of compound 21 was obtained
based on 9 g of the dienone building block 7.
With the B-ring fully established, we turned
our attention to the final C–Cbond-forming
event, pinacol coupling. Selectively forming
the desiredtrans-pinacol (out of four possi-
ble diastereomers) while avoiding radical
fragmentation of the cyclobutane system
posed a substantial challenge. For this pur-
pose, 21 was reduced with LiAlH 4 and theTBS group was removed with tetrabutylammo-
nium fluoride to deliver diol 22 .Swernoxida-
tion of 22 gave the corresponding dialdehyde.
After extensive experimentation, we were
pleasedtofindthattheuseoftitaniumte-
trachloride (TiCl 4 )/Zn ( 23 )resultedinthe
formation of pinacol 23 as a single and desiredSchneideret al.,Science 367 , 676–681 (2020) 7 February 2020 2of5
Fig. 2. Retrosynthetic analysis of (–)-canataxpropellane (2).Fig. 1. Comparison of the Taxol core with the complex taxane core and key features of
(–)-canataxpropellane (2).RESEARCH | REPORT