Evolution of a gram-scale synthesis of (+)-discodermolide

Article abstract of DOI:10.1021/ja0015287

An efficient, highly convergent, stereocontrolled total synthesis of the potent antimitotic agent (+)-discodermolide (1) has been achieved on gram scale. Key elements of the successful strategy include (1) elaboration of three advanced fragments from a common precursor (CP) which embodies the repeating stereochemical triad of the discodermolide backbone, (2) σr-bond installation of the Z trisubstituted olefin, exploiting a modified Negishi cross-coupling reaction, (3) synthesis of a late-stage phosphonium salt utilizing high pressure, and (4) Wittig installation of the Z disubstituted olefin and the terminal (Z)-diene.

Full text of DOI:10.1021/ja0015287

8654
J. Am. Chem. Soc. 2000, 122, 8654-8664
Evolution of a Gram-Scale Synthesis of (+)-Discodermolide
Amos B. Smith III,* Thomas J. Beauchamp, Matthew J. LaMarche, Michael D. Kaufman,
Yuping Qiu, Hirokazu Arimoto, David R. Jones, and Kaoru Kobayashi
Contribution from the Department of Chemistry, Monell Chemical Senses Center, and Laboratory for
Research on the Structure of Matter, UniVersity of PennsylVania, Philadelphia, PennsylVania 19104
ReceiVed May 2, 2000
Abstract: An efficient, highly convergent, stereocontrolled total synthesis of the potent antimitotic agent (+)-
discodermolide (1) has been achieved on gram scale. Key elements of the successful strategy include (1)
elaboration of three advanced fragments from a common precursor (CP) which embodies the repeating
stereochemical triad of the discodermolide backbone, (2) σ-bond installation of the Z trisubstituted olefin,
exploiting a modified Negishi cross-coupling reaction, (3) synthesis of a late-stage phosphonium salt utilizing
high pressure, and (4) Wittig installation of the Z disubstituted olefin and the terminal (Z)-diene.
In 1990, Gunasekera and co-workers1 at the Harbor Branch
of Bristol-Myers Squibb.) Significantly, the binding affinity to
microtubules is higher.6
Oceanographic Institute reported the isolation of (+)-discoder-
molide (1), an architecturally novel antitumor agent derived from
the deep-water marine sponge Discodermia dissoluta. The
structure of (+)-discodermolide comprises a linear polypropi-
onate backbone punctuated by (Z)-olefinic linkages at C(8,9),
C(13,14), and C(21,22). Structure elucidation entailed extensive
spectroscopic studies, including a combination of 1-D and 2-D
NMR techniques; the relative stereochemistry was subsequently
revealed by X-ray crystallography.1 The absolute configuration
however remained undefined until Schreiber and co-workers2
synthesized both antipodes. Interestingly, the unnatural (-)-
antipode also displays significant cytotoxicity.3
Initial biological studies revealed that (+)-discodermolide
suppresses both the two-way mixed-lymphocyte reaction and
the concanavalin A-induced mitogenesis of murine splenocytes
in vitro (IC50 0.24 and 0.19 mM, respectively), with no
associated cytotoxicity.4 (+)-Discodermolide also inhibits the
graft vs host-splenomegaly response induced by injection of
parental splenocytes into F1 recipient mice, with potency
intermediate between those of cyclosporin A and FK506.5 These
findings stimulated considerable interest in discodermolide as
a possible immunosuppressant.
Subsequently, ter Haar et al. reported that (+)-discodermolide
is a potent cell growth inhibitory agent, which arrests cell
development at the M phase by binding and stabilizing mitotic
spindle microtubules.6 Thus, the ability of (+)-1 to inhibit cell
growth was suggested to resemble the clinically proven anti-
cancer agent Taxol (2, Figure 1). (Taxol is a registered trademark
(1) Gunasekera, S. P.; Gunasekera, M.; Longley, R. E.; Schulte, G. K.
J. Org. Chem. 1990, 55, 4912. Correction: Ibid. 1991, 56, 1346.
(2) Nerenberg, J. B.; Hung, D. T.; Somers, P. K.; Schreiber, S. L. J.
Am. Chem. Soc. 1993, 115, 12621. (b) Hung, D. T.; Nerenberg, J. B.;
Schreiber, S. L. Chem. Biol. 1994, 1, 67. (c) Hung, D. T.; Nerenberg, J. B.;
Schreiber, S. L. J. Am. Chem. Soc. 1996, 118, 11054.
(3) The mode of action involves blocking the S phase of the cell cycle;
see ref 2.
(4) Gunasekera, S. P.; Cranick, S.; Longley, R. E. J. Nat. Prod. 1989,
52, 757.
(5) (a) Longley, R. E.; Caddigan, D.; Harmody, D.; Gunasekera, M.;
Gunasekera, S. P. Transplantation 1991, 52, 650. (b) Longley, R. E.;
Caddigan, D.; Harmody, D.; Gunasekera, M.; Gunasekera, S. P. Ibid. 1991,
52, 656. (c) Longley, R. E.; Gunasekera, S. P.; Faherty, D.; McLane, J.;
Dumont, F. Ann. N. Y. Acad. Sci. 1993, 696, 94.
Figure 1.
That the cell growth inhibitory effects of (+)-discodermolide
and Taxol are indeed similar was confirmed by Day and co-
workers.7 (+)-Discodermolide (1) also displays potent activity
against multi-drug-resistant (MDR) carcinoma cell lines, includ-
(6) (a) ter Haar, E.; Kowalski, R. J.; Hamel, E.; Lin, C. M.; Longley, R.
E.; Gunasekera, S. P.; Rosenkranz, H. S.; Day, B. W. Biochemistry 1996,
35, 5, 243. (b) Hung, D. T.; Chen, J.; Schreiber, S. L. Chem. Biol. 1996, 3,
287.
(7) Balachandran, R.; ter Haar, E.; Welsh, M. J.; Grant, S. G.; Day, B.
W. Anti-Cancer Drugs 1998, 9, 67.
10.1021/ja0015287 CCC: $19.00 © 2000 American Chemical Society
Published on Web 08/26/2000

Gram-Scale Synthesis of (+)-Discodermolide
J. Am. Chem. Soc., Vol. 122, No. 36, 2000 8655
ing Taxol-resistant lines.8 In this regard, Horwitz et al.9 reported
that 1 was unable to replace Taxol in a newly discovered Taxol-
dependent human lung carcinoma cell line (A549-T12), whereas
the natural products epothilone A and B (3, 4) and eleutherobin
(5) were able to serve as substitutes for Taxol and thus maintain
viability of the cell line.9 Importantly, the Taxol-dependent cell
line proved to be 20-fold more sensitive to 1 in the presence of
low concentrations of Taxol than in its absence.9 The signifi-
cance of these results is 2-fold: (1) (+)-discodermolide may
bind to microtubules at a site distinct from that of Taxol, the
epothilones (3, 4), and eleutherobin (5) and (2) the combination
of (+)-discodermolide and Taxol may constitute a promising
synergistic cancer chemotherapeutic treatment regime. Unfor-
tunately, the current scarcity of natural material (0.002% w/w
from frozen sponge) has precluded further evaluation as an
anticancer agent.
element, while also exploiting the dithiane coupling and σ- and
π-bond construction tactics that had proven so effective in our
syntheses of FK506 and rapamycin.12 Thus, disconnections at
C(8,9), C(14,15), and C(21,22) generated fragments A, B, and
C, each envisioned to arise from a common precursor (CP)
possessing the recurring triad of contiguous stereogenic centers.
Union of A and B via either cuprate chemistry13 or palladium-
catalyzed cross-coupling14 would then lead to an AB fragment
possessing the C(13,14) trisubstituted (Z)-olefin; in turn the ylide
derived from the AB fragment would be coupled via Wittig
olefination with aldehyde C to generate the C(8,9) cis-alkene.
Introduction of the terminal Z-diene would also exploit a Wittig
olefination or like process. Completion of the discodermolide
venture would then only require elaboration of the C(1) lactone
carbonyl, selective incorporation of a carbamate moiety at C(19),
and global deprotection.
The Common Precursor. The critical role of the common
precursor CP demanded development of a route that would be
amenable to large-scale production. To orchestrate the triad of
contiguous stereogenic centers, we selected the Evans syn-aldol
protocol.15 Given that the absolute stereochemistry of (+)-
discodermolide was unknown at the outset (early 1993), we
arbitrarily chose the commercially available (R)-(-) antipode
of methyl 3-hydroxy-2-methylpropionate (6) as starting material.
Ultimately this choice led to the unnatural (-)-antipode of 1.11a
As detailed here, this strategy is equally amenable to the
synthesis of (+)-discodermolide, utilizing the commercially
available (S)-(+) antipode of 6. Although some of the trans-
formations described in this paper were developed during our
initial synthesis of (-)-discodermolide, for clarity of presenta-
tion, all transformations are depicted in the optical series of the
natural product. The absolute stereochemistries with associated
chiropic properties of all intermediates are available in the
Supporting Information.
Not surprisingly the remarkable biological activity and novel
structure, in conjunction with the scarcity of the natural material,
has led to considerable interest in discodermolide as a synthetic
target.10 To date, six total syntheses of discodermolide, including
our first-generation synthesis of the unnatural levorotatory
congener, have been reported.2,11 In this, a full paper, we
describe the evolution of a synthetic strategy which recently
culminated in the preparation of 1 g of the natural congener.11b
Synthetic Plan. Analysis of the discodermolide structure
revealed a repeating triad of contiguous stereocenters (Scheme
1), separated by (Z)-olefinic linkages at C(8,9) and C(13,14).
From the synthetic perspective we sought an efficient, highly
convergent approach taking full advantage of this structural
Scheme 1
Our point of departure for the construction of the common
precursor CP required for natural (+)-discodermolide entailed
protection of hydroxy ester (+)-6 as the p-methoxybenzyl
(PMB) ether (Scheme 2), exploiting the Bundle trichloroimidate
protocol.16 Reduction with LiAlH4 and distillation provided
(11) (a) Smith, A. B., III; Qiu, Y.; Jones, D. R.; Kobayashi, K. J. Am.
Chem. Soc. 1995, 117, 12011. (b) Smith, A. B., III; Kaufman, M. D.;
Beauchamp, T. J.; LaMarche, M. J.; Arimoto, H. Org. Lett. 1999, 1, 1823.
(c) Haried, S. S.; Yang, G.; Strawn, M. A.; Myles, D. C. J. Org. Chem.
1997, 62, 6098.(d) Marshall, J. A.; Lu, Z.-H.; Johns, B. A. J. Org. Chem.
1998, 63, 7885. (e) Paterson, I.; Florence, G. J.; Gerlach, K.; Scott, J. P.
Angew. Chem., Int. Ed. 2000, 39, 2, 377. (f) Halstead, D. P. Ph.D. Thesis,
Harvard University, Cambridge, MA, 1998.
(12) (a) Smith, A. B., III.; Condon, S. M.; McCauley, J. A. Acc. Chem.
Res. 1998, 31, 35. (b) FK506 formal synthesis: Smith, A. B., III; Chen,
K.; Robinson, D. J.; Laakso, L. M.; Hale, K. J. Tetrahedron Lett. 1994, 35,
4271. Also see: Smith, A. B., III; Hale, K. J. Tetrahedron Lett. 1989, 30,
1037. Smith, A. B., III.; Hale, K. J.; Laakso, L. M.; Chen, K.; Rie´ra, A.
Tetrahedron Lett. 1989, 30, 6963. (c) Rapamycin and 27-demethoxyrapa-
mycin total syntheses: Smith, A. B., III; Condon, S. M.; McCauley, J. A.;
Leazer, J. L., Jr.; Leahy, J. W.; Maleczka, R. E., Jr. J. Am. Chem. Soc.
1997, 119, 947. Smith, A. B., III; Condon, S. M.; McCauley, J. A.; Leazer,
J. L., Jr.; Leahy, J. W.; Maleczka, R. E., Jr. J. Am. Chem. Soc. 1997, 119,
962. Smith, A. B., III; Condon, S. M.; McCauley, J. A.; Leazer, J. L., Jr.;
Leahy, J. W.; Maleczka, R. E., Jr. J. Am. Chem. Soc. 1995, 117, 5407.
Also see: Smith, A. B., III; Condon, S. M.; McCauley, J. A.; Leahy, J.
W.; Leazer, J. L., Jr.; Maleczka, R. E., Jr. Tetrahedron Lett. 1994, 35, 4907.
Smith, A. B., III; Maleczka, R. E., Jr.; Leazer, J. L., Jr.; Leahy, J. W.;
McCauley, J. A.; Condon, S. M. Tetrahedron Lett. 1994, 35, 4911.
(13) Lipshutz, B. H.; Sengupta, S. In Organic Reactions; Paquette, L.
A., Ed.; John Wiley: New York, 1992; Vol. 41, Chapter 2.
(8) Kowalski, R. J.; Giannakakou, P.; Gunasekera, S. P.; Longley, R.
E.; Day, B. W.; Hamel, E. Mol. Pharm. 1997, 52, 613.
(9) Martello, L. A.; McDiad, H. M.; Regl, D. L.; Yang, C. H.; Meng,
D.; Pettus, T. R.; Kaufman, M. D.; Arimoto, H.; Danishefsky, S. J.; Smith,
A. B., III; Horwitz, S. B. Clin. Cancer Res. 2000, 6, 1978.
(10) For synthetic approaches to discodermolide, see: (a) Paterson, I.;
Wren, S. P. J. Chem. Soc., Chem. Commun. 1993, 1790. (b) Clark, D. L.;
Heathcock, C. H. J. Org. Chem. 1993, 58, 5878. (c) Golec, J. M. C.; Jones,
S. D. Tetrahedron Lett. 1993, 34, 8159. (d) Evans, P. L.; Golec, J. M. C.;
Gillespie, R. J. Ibid. 1993, 34, 8163. (e) Golec, J. M. C.; Gillespie, R. J.
Ibid. 1993, 34, 8167. (f) Yang, G.; Myles, D. C. Ibid. 1994, 35, 1313. (g)
Yang, G.; Myles, D. C. Ibid. 1994, 35, 2503. (h) Paterson, I.; Schlapbach,
A. Synlett 1995, 498. (j) Miyazawa, M.; Oonuma, S.; Maruyama, K.;
Miyashita, M. Chem. Lett. 1997, 1191. (j) Miyazawa, M.; Oonuma, S.;
Maruyama, K.; Miyashita, M. Chem. Lett. 1997, 1193. (k) Evans, D. A.;
Allison, B. D.; Yang, M. G. Tetrahedron Lett. 1999, 40, 4457-4460. (l)
Evans, D. A.; Halstead, D. P.; Allison, B. D. Tetrahedron Lett. 1999, 40,
4461-4462. (m) Hoffman, H. M. R.; Misske, A. M.; Tetrehedron 1999,
55, 4315. (n) Filla, S. A.; Song, J. J.; Chem, L.; Masamune, S.; Tetrahedron
Lett. 1999, 5449.
(14) (a) Hayashi, T.; Konishi, M.; Kobori, Y.; Kumada, M.; Higuchi,
T.; Hirotsu, K. J. Am. Chem. Soc. 1984, 106, 158. (b) Negishi, E.; Valente,
L. F.; Kobayashi, M. Ibid. 1980, 102, 3298. (c) Miyaura, N.; Ishiyama, T.;
Sasaki, H.; Ishikawa, M.; Satoh, M.; Suzuki, A. Ibid. 1989, 111, 314.
(15) Evans, D. A.; Bartroli, J.; Shih, T. L. J. Am. Chem. Soc. 1981, 103,
2127.

8656 J. Am. Chem. Soc., Vol. 122, No. 36, 2000
Smith et al.
Scheme 2
Scheme 3
alcohol 7; Swern oxidation17 then furnished aldehyde 8, which
was used without purification in an Evans aldol condensation
with oxazolidinone 9.18 Weinreb amide formation19 completed
construction of the common precursor CP.
was intended to permit eventual late-stage deprotection of the
C(21) and C(19) hydroxyls, facilitating selective introduction
in turn of the terminal diene and carbamate moieties. Accord-
ingly, treatment of CP with DDQ furnished crystalline acetal
13. Controlled reduction of the Weinreb amide23 provided
crystalline aldehyde 14 which was directly coupled with
oxazolidinone 9 via an Evans aldol reaction to provide crystal-
line alcohol 15 (83% yield, two steps), possessing the requisite
five contiguous stereogenic centers of subunit A. The structures
of 13 and 15 were confirmed by single-crystal X-ray analyses.21
Protection of the secondary hydroxyl as the TBS ether and
removal of the chiral auxiliary (LiBH4, EtOH, THF)24 then
furnished crystalline 17 (82% yield, two steps). Alcohol 17 was
then converted to iodide A (95% yield), completing the synthesis
of the C(17-21) subunit A in six steps and 56% overall yield
from CP.25 The remarkable number of crystalline intermediates
en route made this sequence well suited for the routine
production of A on a large scale.26
While preparing for the large-scale synthesis of natural
discodermolide (vide infra), we became aware of a report by
Walkup and co-workers20 describing aldol adduct 12 as a
crystalline solid. Since the phenylalaninol-derived oxazolidine
10 is noncrystalline and thus purified by silica gel chromatog-
raphy, the prospect of utilizing 11 in place of 9 appeared
attractive for the large-scale production of CP. Gratifyingly,
the dibutylboron enolate of 11, readily prepared from (1S,2R)-
norephedrine,15 did indeed produce a crystalline adduct with
aldehyde 8. The yield of 12 was 64-70% on a 60-70 g scale.
Significantly, 12 could be isolated directly from the reaction
mixture by crystallization. In practice, the first three intermedi-
ates were carried forward without purification, and 12 was
isolated by crystallization (55%, four steps, 60 g scale). The
structure of 12 was confirmed by single-crystal X-ray analysis.21
Subsequent conversion of 12 to the common precursor CP
proceeded as before without complication (98% yield). Purifica-
tion of CP was facilitated by crystallization of the oxazolidinone
auxiliary from the reaction mixture (80-90% recovery). The
readily recovered auxiliary further augmented the efficiency of
this sequence.
Fragment B. As outlined in Scheme 1 our synthetic plan
called for construction of a vinyl halide possessing the (Z)-
configuration to couple with fragment A. We selected bromide
23 as our initial target (Scheme 4). Beginning again with CP,
protection of the hydroxyl as the TBS ether, followed by DIBAL
reduction, afforded aldehyde 19 in 89% overall yield. Unfor-
Fragment A. The polypropionate structural motif of fragment
A suggested a second asymmetric aldol reaction (Scheme 3).
Initial generation of the p-methoxybenzylidene (PMP) acetal22
(22) Oikawa, Y.; Yoshioka, T.; Yonemitsu, O. Tetrahedron Lett. 1982,
23, 889.
(16) (a) Nakajima, N.; Horita, K.; Abe, R.; Yonemitsu, O. Tetrahedron
Lett. 1988, 29, 4139. (b) Iversen, T.; Bundle, D. R.; J. Chem. Soc., Chem.
Commun. 1981, 1240.
(17) Mancuso, A. J.; Swern, D. Synthesis 1981, 165.
(18) Gage, J. R.; Evans, D. A. Org. Synth. 1990, 68, 77.
(19) (a) Levin, J. I.; Turos, E.; Weinreb, S. M. Synth. Commun. 1982,
12, 989. (b) Evans, D. A.; Bender, S. L.; Morris, J. J. Am. Chem. Soc.
1988, 110, 2506.
(20) Walkup, R. D.; Kane, R. R.; Boatman, P. D., Jr.; Cunningham, R.
T. Tetrahedron Lett. 1990, 31, 7587.
(21) Complete crystallographic data for 12, 13, 15, 49, 89, and 1 have
been deposited in the Cambridge Crystallographic Data Centre.
(23) Nahm, S.; Weinreb, S. M. Tetrahedron Lett. 1981, 22, 3815.
(24) Penning, T. D.; Djuric, S. W.; Haack, R. A.; Kalish, V. J.; Miyashiro,
J. M.; Rowell, B. W.; Yu, S. S. Synth. Commun. 1990, 20, 307.
(25) Alcohol 17 could also be converted to iodide A in 94% overall
yield via tosylation (TsCl, pyridine) followed by displacement with sodium
iodide in DMF.
(26) We also explored the use of the norephedrine-derived oxazolidinone
11 employed in the synthesis of CP. Unfortunately the corresponding aldol
product was obtained as an oil. In addition, reductive removal of the auxiliary
resulted in competitive opening of the oxazolidinone ring; thus, oxazoli-
dinone 9 was the auxiliary of choice for the construction of fragment A.

Gram-Scale Synthesis of (+)-Discodermolide
J. Am. Chem. Soc., Vol. 122, No. 36, 2000 8657
Scheme 4
Scheme 5
and alcohol 26 in 83% yield. Presumably, the observed
selectivity is due to the higher oxidation potential of the acetal
which contains two electron-withdrawing oxygens. With these
results in hand, we proceeded with confidence.
Fragment C. Our point of departure was TBS ether 18
(Scheme 6), previously prepared during the synthesis of
Scheme 6
tunately the one-step protocol of Smithers27 to introduce the
vinyl halide via an R-bromoalkylylide was not synthetically
useful (e.g., 40% yield, 4:1 Z/E). We therefore adopted a
stepwise approach, whereby 19 was first converted to the Z
R-bromo unsaturated ester 20 (Ph3PdCBrCO2Et,28 PhH, reflux;
75% yield). Reduction, followed by mesylation and displace-
ment with hydride (LiBHEt3) then furnished the (Z)-vinyl
bromide 23 in 77% yield from 20.
In seeking to verify the olefin geometry in 23 by sequential
1
metalation, protonation, and H NMR analysis, we discovered
that the initially formed vinyl anion was prone to 1,5-silyl
migration, affording 24 when the lithium-halogen exchange
(n-BuLi) was performed in THF.29 Although the unexpected
retro-[1,5]-Brook rearrangement30 provided evidence that bro-
mide 23 indeed possessed the desired (Z)-configuration (vide
infra), our synthetic plan was now in jeopardy, unless the
migration could be suppressed. Toward this end, halogen
exchange with t-BuLi at -78 °C in THF resulted in only 20%
rearrangement, while treatment with t-BuLi in Et2O (-78 °C)
completely circumvented the problem. We attribute the observed
difference in reactivity to stronger solvation of the organolithium
aggregate in THF, an effect that renders the carbanion more
nucleophilic.31 Protonation of the vinyl anion then led to 25 in
94% yield; the trans-olefinic geometry was confirmed by the
vinylic coupling constant (e.g., J ) 15.3 Hz).32
While the protecting groups in subunits A and B were
carefully chosen to exploit most efficiently CP, our long-term
strategy demanded removal of the primary PMB ether (derived
from 23) in the presence of the PMP acetal in the AB coupled
product (Scheme 1). It thus appeared prudent to explore the
feasibility of this unprecedented tactic before committing to this
protecting group scheme. Accordingly, exposure of an equimolar
mixture of PMB ether 23 and PMP acetal 17 to DDQ (1.1 equiv)
in CH2Cl2/H2O22 (Scheme 5) furnished acetal 17 largely intact
fragment B (vide supra). Oxidative removal of the PMB group
(DDQ, CH2Cl2, H2O) provided an unstable alcohol, 27, in yields
ranging from 60% to 86%, accompanied by significant amounts
of the corresponding δ-lactone, which could be recycled.33 Better
results were obtained by hydrogenolysis of 18 with Pearlman’s
catalyst, which provided 27 in 95% yield. Oxidation with SO3‚
pyridine34 then furnished aldehyde 28 (98% yield); the two steps
were routinely performed without purification of 27. Dithiane
generation via modification of the Evans protocol35 [e.g.,
(TMSSCH2)2CH2, ZnCl2, Et2O] minimized elimination of the
TBS ether, permitting efficient production of dithiane 29 (79%).
DIBAL reduction (91% yield), followed by dimethyl acetal
(27) Smithers, R. H. J. Org. Chem. 1978, 43, 2833.
(28) Boeckman, R. K., Jr.; Ko, S. S. J. Am. Chem. Soc. 1980, 102, 7146.
(29) For leading references on silicon migration from oxygen to carbon,
see: (a) Jung, M. E.; Nichols, C. J. J. Org. Chem. 1996, 61, 9065. (b)
Lautens, M.; Delanghe, P. H. M.; Goh, J. B.; Zhang, C. H. J. Org. Chem.
1995, 60, 4213. (c) Marumoto, S.; Kuwajima, I. J. Am. Chem. Soc. 1993,
115, 9021. (d) For a related 1,4-O f sp2C silyl rearrangement, see: Kim,
K. D.; Magriotis, P. A. Tetrahedron Lett. 1990, 31, 6137.
(33) The Weinreb amide was regenerated from the δ lactone in 89%
yield [HN(OMe)Me‚HCl, CH2Cl2, Me3Al].
(34) Parikh, J. R.; Doering, W. von E. J. Am. Chem. Soc. 1967, 89, 5505.
(35) ZnCl2 was used instead of ZnI2; see: Qui, Y. Ph.D. Thesis,
University of Pennsylvania, Philadelphia, PA, 1997. Evans, D. A.; Truesdale,
L. K.; Grimm, K. G.; Nesbitt, S. L. J. Am. Chem. Soc. 1977, 99, 5009.
(30) Brook, A. G. Acc. Chem. Res. 1974, 77.
(31) McGarrity, J. F.; Ogle, C. A. J. Am. Chem. Soc. 1985, 107, 1805.
(32) Eliel, E. L.; Wilen, S. H. Stereochemistry of Organic Compounds;
John Wiley: New York, 1994; Chapter 9.

8658 J. Am. Chem. Soc., Vol. 122, No. 36, 2000
Smith et al.
formation, then gave 31 (99%). Coupling of the dithiane with
benzyl (S)-(+)-glycidyl ether 32 exploited the conditions
developed earlier in our laboratory in connection with our
FK506 and rapamycin syntheses.12 This sequence provided
alcohol 33 in 79% yield. Unmasking of the ketone [(CF3CO2)2-
IPh, 80%]36 followed by Evans directed reduction37 furnished
the anti-diol 35 (97%), which embodied all of the stereogenic
centers required for fragment C. Acid-catalyzed cyclization then
provided methoxy pyran 36 in 87% yield, as a separable mixture
of R and â anomers (2:1). The large coupling constant in 36R
(J4,5 ) 9.8 Hz), diagnostic of the diaxial arrangement of the
C(4) and C(5) hydrogens, confirmed the selectivity of the
directed reduction.38 The configuration of the methoxy group
was deduced by a series of NOE experiments.39 The epimeric
mixture 36 was then converted to the corresponding TBS ethers
37 (R:â ) 2:1, 97% yield).
(2 equiv of t-BuLi, -78 °C) followed by the addition of MgBr2,
ZnCl2, or B-methoxy-9-BBN furnished the requisite organo-
metallic species.42 Not surprisingly, the initial lithio species also
proved susceptible to silyl migration to form 41, a significant
side reaction at -45 °C in Et2O or at -78 °C in THF (cf.
Scheme 4). Best results were obtained by metalation with t-BuLi
at -78 °C for ca. 10 min followed by addition of MgBr2.
Although AB was only obtained in 14% yield, we were
encouraged by the observation that most of the vinyl bromide
could be recovered. The high recovery of 23 suggested slow
oxidative insertion of the palladium into the carbon-bromine
bond. We therefore examined the more reactive vinyl iodide
46 (Scheme 8). Initially, iodide 46 was prepared in one step
Scheme 8
In our initial synthetic strategy, we had planned to mask the
C(1) carbonyl in fragment C as the mixed methyl acetal.
However, under a variety of conditions, we were unable to
hydrolyze the acetal moiety in the presence of the TBS ethers,
a required endgame transformation. To circumvent this obstacle
we converted 37 to the ethylthio derivative 39, a tactic first
reported by Schreiber and co-workers in their discodermolide
syntheses.2 Toward this end, hydrogenolysis of 37 afforded
alcohol 38 quantitatively; exposure to ethanethiol and MgBr2
in Et2O40 then furnished a separable mixture of hemithioacetal
39 and the â anomer (6:1) in 83% yield.39 Although in principle
both epimers could be carried forward, it was more expedient
to work with homogeneous material. Swern oxidation of 39 led
to 40 (86% yield), our first-generation fragment C. The overall
sequence proceeded in 14 steps from CP and in 24% overall
yield (average yield/step 90%).
Union of Fragments A and B. At the outset, we had planned
to employ a vinyl cuprate to effect σ-bond construction of the
C(13-14) Z trisubstituted olefin.13 Unfortunately, attempts to
couple iodide A (Scheme 7) with the cuprate prepared from
(86% yield) from bromide 23 via Nickel-catalyzed bromine-
iodine exchange.43 The overall length of this sequence from CP
however was not acceptable (i.e., seven steps). We therefore
explored the Zhao one-step protocol.44 Treatment of aldehyde
19 with the ylide derived from 4344 afforded 46 in 40-46%
overall yield from CP.45 The Z/E selectivity ranged from 8:1
to 17:1. In practice, the sequence required only a single
chromatographic purification, and could be performed routinely
on a 30 g scale.
Scheme 7
Upon scale-up, we noted the formation of significant amounts
of an olefin lacking the iodide. A similar result was recently
reported by Marshall and Johns.11 This side product 25 proved
to be the direct result of incomplete incorporation of iodine in
1
(39) Selected NOE values (%) and H NMR coupling constants for 36
and 39. We thank Dr. Peter Dormer for technical assistance.
vinyl bromide 23, via lithium halogen exchange followed by
the addition of a variety of Cu(I) salts [i.e., CuCN, CuI, (Th)-
Cu(CN)Li], by and large proved unsuccessful. Only the higher
(40) Kim, S.; Park, J. H.; Lee, S. Tetrahedron Lett. 1989, 30, 6697.
(41) Corey, E. J.; Kyler, K.; Raju, N. Tetrahedron Lett. 1984, 25, 5115.
(42) (a) Bailey, W. F.; Punzalan, E. R. J. Org. Chem. 1990, 55, 5404.
(b) Negishi, E.; Swanson, D. R.; Rousset, C. J. Org. Chem. 1990, 55, 5406.
(43) Takagi, K.; Hayama, N.; Inokawa, S. Chem. Lett. 1978, 1435.
(44) Chen, J.; Wang, T.; Zhao, K. Tetrahedron Lett. 1994, 35, 2827.
(45) While attempting to optimize the formation of 46 via the Zhao
protocol, we observed significant formation of methyl ketone and epoxide
byproducts. These results suggest that formation of an epoxyphosphonium
salt is competitive with Wittig olefination. A detailed discussion of the Zhao
chemistry has been published and will not be repeated here; see: Arimoto,
H.; Kaufman, M. D.; Kobayashi, K.; Qui, Y.; Smith, A. B., III. Synlett
1998, 765.
41
order cuprate exploiting the Corey reagent n-Bu4NCu(CN)2
furnished the desired AB segment, albeit in only 8% yield.
Attention quickly shifted to palladium-catalyzed cross-
coupling reactions.14 Initial studies targeted the coupling of
bromide 23 with organomagnesium, organozinc, or organoboron
reagents prepared from A.14 To this end, lithiation of iodide A
(36) Stork, G.; Zhao, K. Tetrahedron Lett. 1989, 30, 287.
(37) Evans, D. A.; Chapman, K. T.; Carreira, E. M. J. Am. Chem. Soc.
1988, 110, 3560.
(38) See ref 32, Chapter 11.

Gram-Scale Synthesis of (+)-Discodermolide
J. Am. Chem. Soc., Vol. 122, No. 36, 2000 8659
phosphonium salt 42, thereby leading to a mixture of ylides
upon deprotonation with NaHMDS. The poor iodine incorpora-
tion was in turn the result of the low solubility of iodine in
THF; increasing the amount of THF (0.1 M I2) during the
preparation of 43, in conjunction with vigorous stirring at
ambient temperature prior to cooling, circumvented this problem.
As anticipated vinyl iodide 46 proved superior to bromide
23 in the palladium-catalyzed cross-coupling reactions, although
not without significant initial experimentation. Best results
involved modification of a Negishi protocol (Scheme 9).14b
Scheme 10
Scheme 9
To understand better the nature of this reaction, we turned to
model alcohol 52 (Scheme 11), prepared from aldehyde 19 via
reaction with the ylide derived from isopropyltriphenylphos-
phonium iodide, followed by oxidative removal (DDQ) of the
PMB ether. Again attempts to generate the iodide 54 from 52
via the tosylate proved unsuccessful. However, a modification
Specifically, treatment of iodide A and 1 equiv of ZnCl2 in ether
with 3 equiv of t-BuLi at -78 °C, followed by warming to
ambient temperature and transfer to an intimate mixture of vinyl
iodide 46 and Pd(PPh3)4 (5 mol %), furnished AB after workup
in 66% yield (recrystallized). The efficiency of this cross-
coupling reaction is particularly noteworthy. In contrast to most
cross-coupling reactions which require a minimum of 1.5 equiv
of alkyl iodide for comparable efficiency, only 1.1 equiv of A
was required.46 Importantly, 3 equiv of t-BuLi is required for
complete consumption of A. Treatment with 2 equiv of t-BuLi,
followed by quenching with water, leads to a 1:1 mixture of
starting iodide and the expected alkane. Accordingly, we
speculate that a mixed tert-butyl-alkyl zinc intermediate (e.g.,
47) is the reactive alkyl donor in the coupling process.
Scheme 11
The AB Wittig Reagent. Conversion of AB to the requisite
phosphonium iodide 48 (Scheme 10) began with selective
removal of the PMB group. As anticipated from our earlier
model study (Scheme 5), exposure of AB to 1 equiv of DDQ
(CH2Cl2, H2O) effected selective removal of the PMB ether in
the presence of the PMP acetal, furnishing 49 in 91% yield.
The relative stereochemistry of 49 was confirmed by single-
crystal X-ray analysis.21 Further elaboration of 49, however,
presented unforeseen difficulties. Quite surprisingly, attempts
to convert 49 to iodide 51 via tosylate 50 proved unsuccessful.
While some tosylate appeared to be present in the reaction
mixture (1H NMR), attempts to chromatograph the tosylate
failed. Attempts to convert the tosylate directly to iodide 51
1
also proved unrewarding. Detailed H NMR analysis of the
of the Garegg conditions (PPh3, I2, imidazole, 3:1 Et2O/CH3-
CN)48 did permit conversion of 52 to iodide 54 (74%); minor
amounts of 56 and 57 were also isolated and characterized (8%
each). The observation of cyclopentanes 56 and 57 provides
clear evidence for the competitive solvolysis process.49
Analysis of the competing transition states (e.g., intra-
molecular vs bimolecular displacement) suggested that solvent
product mixture revealed the loss of the trisubstituted olefin,
presumably the result of olefin-assisted displacement at C(9).
Analogous olefin-assisted solvolyses of 5-hexenylsulfonates are
well-known.47 However, such reactions usually require more
vigorous conditions (e.g., formic acid, 75 °C).47
(46) In the Marshall synthesis of 1, a Suzuki coupling strategy was
employed requiring 2 equiv of the alkyl iodide; see ref 11d.
(47) (a) Johnson, W. S.; Bailey, D. M.; Owyang, R.; Bell, R. A.; Jaques,
B.; Crandall, J. K. J. Am. Chem. Soc. 1964, 86, 1959. (b) Johnson, W. S.;
Owyang, R. J. Am. Chem. Soc. 1964, 86, 5593.
(48) (a) Corey, E. J.; Pyne, S. G.; Su, W.-G. Tetrahedron Lett. 1983,
24, 4883. (b) Garegg, P. J.; Samuelsson, B. J. Chem. Soc., Perk. Trans. 1
1980, 2866.

8660 J. Am. Chem. Soc., Vol. 122, No. 36, 2000
Smith et al.
polarity should influence the product distribution.50 Indeed, the
use of CH3CN or CH3CN/Et2O (3:1) as solvent significantly
increased cyclization vs displacement (e.g., 54:56:57 ) 9:1:1;
90% yield) whereas the less polar medium, Et2O/PhH (2:1),
provided a greater proportion of iodide 54 (18:1:1; 97% yield).
To our delight similar treatment of alcohol 49 [PPh3, I2, Et2O/
PhH (2:1)] furnished the requisite iodide 51 in near quantitative
yield (Scheme 12). The sensitive iodide could be either purified
Scheme 13
Scheme 12
A First-Generation Endgame. Deprotonation of phospho-
nium salt 48 with sodium bis(trimethylsilyl)amide in THF gave
rise to an ylide which underwent high Z-selective coupling
(>49:1 Z/E) with aldehyde 40, to furnish 62 in 76% yield
(Scheme 14). DIBAL reduction52 then led to a primary alcohol
Scheme 14
by rapid silica gel chromatography (pretreated, Et3N) or used
without purification.
Turning next to the preparation of the requisite Wittig reagent
(Scheme 12), initial efforts involved treatment of 51 with PPh3
in solvent (not shown); again extensive cyclization and some
decomposition occurred. To circumvent intramolecular cycliza-
tion, we turned to molten PPh3. In the event, treatment of 51
with excess triphenylphosphine (15 equiv) and i-Pr2NEt (3
equiv) at 80 °C in the absence of additional solvent generated
48 in 37% yield from 49. The major byproduct was again
cyclopentane 60 (35% yield). Critical to the success of this
reaction was the use of an amine base to prevent the HI,
produced during cyclization, from promoting decomposition.
Similar treatment of model iodide 54 furnished phosphonium
salt 61 in 38% overall yield from 52. In marked contrast, the
fully saturated congener obtained by hydrogenation of olefin
52 afforded the corresponding saturated phosphonium salt (via
the iodide) in almost quantitative yield (94%).
Phosphonium Salts via Ultrahigh Pressure. Although the
use of molten triphenylphosphine as solvent led to displacement
of the iodide, the 37% yield was not acceptable for large-scale
material advancement. We therefore turned to high pressure to
facilitate displacement. In 1984 Dauben and co-workers reported
that alkylation of phosphines can be dramatically accelerated
by ultrahigh pressure.51 In the event, reaction of a mixture of
iodide 54, triphenylphosphine (3 equiv), and i-Pr2NEt (0.5 equiv)
in benzene/toluene (7:3) at 12.8 kbar (185 000 psi) for 72 h
furnished an 86% yield of phosphonium salt 61 (Scheme 13).
Gratifyingly, iodide 51 was similarly transformed into 48 via
these conditions (6 d; 70% overall yield from alcohol 49).
(49) Interestingly a Monte Carlo conformational analysis of iodide 54
using the MM2* force field (Macromodel, v. 5.0) revealed that the
substituents on 54 enforce a half-chair arrangement in a large percentage
of the lower energy conformers. In particular, 12% of the lowest energy
conformers (Erel ) 0-5.48 kJ/mol) were judged suitably disposed to
facilitate displacement of the iodide by the olefin. Of these conformers,
the average distance between C(1) and C(5) was only 3.05 Å (SD ) 0.02
Å) while the average I-C(1)-C(5) angle was 161.2° (SD ) 0.4°).
(50) Reichardt, C. SolVents and SolVent Effects in Organic Chemistry;
VCH: Weinheim, 1990.
(88% yield), which in turn was oxidized to aldehyde 64 (96%
yield). The terminal (Z)-diene 65 was next installed via the
Yamamoto protocol53 in 70% yield, completing assembly of
the discodermolide backbone; the selectivity was 16:1 Z/E.54

Gram-Scale Synthesis of (+)-Discodermolide
J. Am. Chem. Soc., Vol. 122, No. 36, 2000 8661
Hydrolysis of the hemithio acetal followed by Kuzuhara-
Fletcher oxidation55 provided lactone 66 in 82% yield for the
two steps. Removal of the PMB group (DDQ, CH2Cl2, H2O,
95%) and introduction of the carbamate via the Kocovsky
protocol (Cl3CCONCO, CH2Cl2; K2CO3, MeOH, 83%)56 then
led to 68. Final deprotection with HF/CH3CN (48%, 1:9)
furnished (-)-discodermolide (60% yield), identical in all
respects, with an authentic sample provided by Schreiber (e.g.,
We selected as our initial nucleophile the silyl enol ether
derived from mesityl oxide (72) (Scheme 16).60 There was of
Scheme 16
1
500 MHz H NMR, 125 MHz 13C NMR, IR, HRMS, optical
rotation, and TLC in four solvent systems).57
A Second-Generation Endgame. Although the endgame
shown in Scheme 14 permitted completion of (-)-discoder-
molide in 1995 (40 total steps; 28 in the longest linear
sequence),11 the synthesis was far from optimal. Having
subsequently learned of the potential of (+)-discodermolide as
an anticancer agent, we felt compelled both to shorten the
existing synthetic route and to improve the overall efficiency
of the synthesis to permit preparation of significant quantities
of this natural product for further clinical development. We set
as our goal the synthesis of 1 g.
Fragment C: A Second-Generation Strategy. In our initial
synthesis of (-)-discodermolide, elaboration of subunit C (40)
from the common precursor CP was clearly the least efficient
aspect of the synthetic venture, requiring 14 steps (CP f 40;
Scheme 6). The length of this sequence was due, at least in
part, to our desire to exploit the coupling of dithiane 31 with
an epoxide (31 f 33; Scheme 6), a tactic utilized to great
advantage in earlier syntheses in our laboratory. While the above
protocol proved effective, a direct three-carbon extension of
aldehyde 28 appeared more expedient (Scheme 15). The
course no guarantee that we would be able to control the facial
selectivity of aldehyde 28. Two options were however available.
One would be to replace the TBS ether of 28 with a more Lewis
basic protecting group to direct nucleophilic addition to the
aldehyde via metal ion chelation.10k,l The alternative would be
to use an external asymmetric promoter. In the interest of
synthetic economy and the demonstrated effectiveness of the
C(14) TBS ether in the first-generation endgame (vide supra),
we elected to explore initially the addition of enol ether 72 to
28 (Scheme 16). To our delight treatment of a solution of 28
and TiCl4 (1 equiv) with enol ether 72 at -78 °C proceeded in
77% yield with near complete anti-Felkin selectivity (ds > 20:
1); lactone 74 was then available after acid-catalyzed cyclization
of the intermediate hydroxy amide 73.61
Scheme 15
Anti-Felkin addition was also observed in the TiCl4-promoted
reactions of allyltrimethylsilane and allyltributylstannane with
28 (ca. 49:1 and 19:1, respectively; Scheme 17). In contrast,
Scheme 17
challenge, however, would entail stereocontrol in the addition
of an appropriate nucleophile to 28, to generate the required
anti relationship between C(4) and C(5) in 70.58 This strategy
also held the promise of allowing the unprotected C(1) lactone
to be carried throughout the endgame sequence. We thus set
lactone-aldehyde 71 as our second-generation fragment C.59
BF3‚OEt2-promoted the addition of allyltributylstannane with
no selectivity.62 Taken together, these results suggest that
bidentate complexation of the TiCl4 with the carbonyl groups
of the Weinreb amide and the aldehyde, via an eight-membered
chelate, is responsible for the observed high anti-Felkin selectiv-
ity.63,64
(51) Dauben, W. G.; Gerdes, J. M.; Bunce, R. A. J. Org. Chem. 1984,
49, 4293. For reviews on the use of high pressure in organic synthesis, see:
(a) Organic Synthesis at High Pressure; Matsumoto, K., Morrin Acheson,
R., Eds.; John Wiley: New York, 1991. (b) Matsumoto, K.; Sera, A.;
Uchida, T. Synthesis 1985, 1. (c) Matsumoto, K.; Sera, A. Synthesis 1985,
999.
(52) Takano, S.; Akiyama, M.; Sato, S.; Ogasawara, K. Chem. Lett. 1983,
1593.
(53) Ikeda, Y.; Ukai, J.; Ikeda, N.; Yamamoto, H. Tetrahedron 1987,
43, 723.
(59) Paterson reported the synthesis of the enantiomer of 71; see ref 10a.
However, use as a substrate for Wittig olefination was not reported.
(60) Paterson, I. Tetrahedron Lett. 1979, 1519.
(61) The relative stereochemistries of 74 and the C(5) epimer 75 were
established by NMR: 74, J4,5 ) 10.0 Hz; 75, J4,5 ) 2.4 Hz, J2,3 ) 9.8 Hz.
(62) The BPS ether aldehyde 79, a substrate lacking the Weinreb amide
moiety furnished the Felkin addition product 80 over the anti-Felkin adduct
81 (3:1):
(54) Heathcock has also utilized the Yamamoto protocol; see ref 10b.
(55) Kuzuhara, H.; Fletcher, H. G., Jr. J. Org. Chem. 1967, 32, 2531.
(56) Kocovsky, P. Tetrahedron Lett. 1986, 27, 5521.
(57) We thank Professor Schreiber (Harvard University) for providing a
sample of synthetic (-)-discodermolide.
(58) For a general discussion of nucleophilic additions to R-chiral
carbonyl compounds, see: Devant, R. M.; Radunz, H.-E. In Methods of
Organic Chemistry; Helmchen, G., Hoffman, R. W., Mulzer, J., Schaumann,
Eds.; Thieme Verlag: New York, 1995; Vol. E21b, Chapter 1.3.

8662 J. Am. Chem. Soc., Vol. 122, No. 36, 2000
Smith et al.
With a method available to access 73, 74, and 77 with
excellent stereocontrol at C(5), attention turned to introduction
of the requisite R-hydroxy aldehyde functionality. Initial studies
entailed asymmetric dihydroxylation of olefin 77.65 Attempts
to effect this transformation led, at best, only to a 3:1 mixture
of diols. Attention thus shifted to the acyclic Mukaiyama aldol
adduct 73. We reasoned that the stereogenicity of the C(5)
hydroxyl could be employed to direct reduction of the enone
carbonyl to the anti-diol similiar to a reduction reported by
Evans.37,66
subsequently secured via single-crystal X-ray analysis (Scheme
19).
Scheme 19
Having failed to observe high selectivity in the directed
reduction of the C(7) enone in 73, we explored the reduction
of the enone in lactone 74 (Scheme 18), readily available (77%
Scheme 18
With alcohol 89 in hand, conversion to our second-generation
fragment C (71) proved straightforward. Protection as the TBS
ether furnished 93 (87% yield), which was then subjected to
ozonolysis to give 71 (95% yield). Thus, our second-generation
subunit C was available in seven steps (47% overall yield, 89%
average) from CP. While this synthesis was considerably shorter
and more efficient than that of 40 (e.g.,14 steps, 24% overall
yield), the vitality of 71 as a Wittig coupling partner was yet to
be established.
Chemoselectivity in the Wittig Coupling. Treatment of 71
(Scheme 20) with the ylide derived from 48 (KHMDS, toluene,
0 °C) furnished olefin 94 in 57% yield with excellent Z/E
selectivity (>20:1). The choice of both base and solvent proved
yield) from 73 by acidic treatment (Scheme 16). A variety of
reducing conditions were examined. Only the Selectride reagents
furnished the desired allylic alcohol 89 as the predominant
product.67 Higher selectivities were obtained in toluene com-
pared to THF or ether. The selectivies also proved sensitive to
the counterion; K-Selectride (89:90 ) 9:1) was notably more
selective than both N-Selectride (89:90 ) 3:1) and L-Selectride
(89:90 ) 1.2:1). The stereochemistry of alcohol 89 was
Scheme 20
(63) For a discussion of chelation-controlled carbonyl additions see:
Reetz, M. T. Acc. Chem. Res. 1993, 26, 462.
(64) Interestingly the structurally similar aldehyde ester 82 has also been
postulated to participate in bidentate coordination with both TiCl4 and SnCl4
to afford anti-Felkin products upon reaction with crotylsilanes and crotyl-
stannanes. 13C NMR data strongly supports coordination of Sn(IV) with
the carbonyls of 82. (a) Santelli-Rouvier, C. Tetrahedron Lett. 1984, 25,
4371. (b) Yamamoto, Y.; Yatagai, H.; Ishihara, Y.; Maeda, N.; Maruyama,
K. Tetrahedron 1984, 40, 2239. (c) Yamamoto, Y.; Nemoto, H.; Kikuchi,
R.; Komatsu, H.; Suzuki, I. J. Am. Chem. Soc. 1990, 112, 8598.
(65) Kolb, H. C.; VanNieuwenhze, M. S.; Sharpless, K. B. Chem. ReV.
1994, 94, 2483.
(66) Surprisingly, treatment of 73 with Me4NB(OAc)3H again afforded
only a 3: 1 mixture of diols.
(67) Minor amounts of elimination products 91 and 92 were also
observed.

Gram-Scale Synthesis of (+)-Discodermolide
J. Am. Chem. Soc., Vol. 122, No. 36, 2000 8663
critical for this coupling reaction. Subsequent exposure of 94
to excess DIBAL (10 equiv, -78 °C f 0 °C) effected reductive
cleavage of the PMP acetal, with concomitant reduction of the
lactone carbonyl, to furnish 95. Reduction of the lactone proved
of little consequence, since pyridinium dichromate (PDC) readily
converted the primary hydroxyl and the lactol to lactone
aldehyde 96 (62% yield, two steps).
The stage was thus set for incorporation of the terminal diene.
Although our first synthesis of discodermolide (cf. 64, Scheme
14) made efficient use of the Yamamoto protocol for this
transformation (70% yield, 16:1 Z/E), in this case only a modest
yield of 66 with significantly lower Z selectivity (40-58%, 4:1
Z/E) was observed. Apparently the lactone moiety is sensitive
to the reaction conditions, although it is difficult to account for
the low Z selectivity (e.g., 65).68 The low efficiency of the diene
installation stimulated us to explore an alternative route in which
the diene would be incorporated at an earlier stage.
in 75-82% yield on a multigram scale (Scheme 22). Chemo-
selective addition of the derived ylide also proceeded in good
yield and with excellent selectivity (Z/E 15-24:1). Again the
solvent of choice (vide supra) proved to be THF, with aldehyde
71 added to the ylide at -20 °C. Addition at lower temperature
resulted in considerable decomposition. The phosphonium salt
99 proved to be highly hygroscopic; thus, great care was
necessary to remove residual water, which if present led to the
formation of the corresponding diphenylphosphine oxide upon
addition of base.72 Best results were obtained by azeotropic
removal of water from the salt with toluene (three times),
followed by drying in vacuo at 50 °C for 12 h (ca. 0.1 Torr).
Attention to these details led to 66 reproducibly in 59-69%
yield up to the 2.0 g scale; importantly the phosphonium salt
99 could be recovered. The yield of 66 based on recovered 99
was 85-95%.
Having achieved assembly of the discodermolide carbon
skeleton, we were poised to complete the large-scale synthesis.
However, the Z/E selectivity (8-12:1) obtained in the diene
synthesis demanded attention. Previously the minor (E)-diene
was removed via AgNO3-impregnated silica gel chromatogra-
phy; however, for large-scale work this was less than optimal.
Fortunately we discovered that when 66, an 8:1 Z/E mixture at
C(22,23), was subjected to DDQ oxidation to remove the C(19)
PMB group, 67 was obtained as a single diastereomer along
with an easily separable byproduct (not shown). The byproduct
proved to be the result of a fortuitous Diels-Alder reaction
between the (E)-diene and DDQ. The ability of the (E)-diene
to adopt the s-cis-conformation required for cycloaddition to
DDQ leads to the observed chemoselectivity, whereas the
desired (Z)-isomer is prevented from adopting a planar confor-
mation due to significant steric interaction between the C(24)
vinyl and C(20) methine moieties.
Early Installation of the Terminal Diene: Completion of
a Gram-Scale Synthesis of (+)-Discodermolide. We began
with the previously prepared C(9-22) AB fragment (Scheme
21). To maintain eventual discrimination of the C(19) PMB
Scheme 21
With pure 67 in hand, completion of the 1 g synthesis of
(+)-discodermolide now only required installation of the
carbamate and global desilylation. Importantly, carbamate
formation (Cl3CCONCO; Al2O3) proceeded in 90-95% yield.
For global deprotection we had employed HF/CH3CN in our
1995 synthesis. However, on scale-up we found these conditions
to be capricious (e.g., yields ranging from 20% to 60%).
Screening a series of conditions revealed that slow addition of
3 N HCl to a methanol solution of 68 over a 12 h period
produced the natural product cleanly and reproducibly up to
the 200 mg scale. Utilizing this route, we were able to prepare
1.06 g of (+)-discodermolide; crystallization from acetonitrile73
furnished 1.043 g of the natural product (mp 117-120 °C).
Importantly, in contrast to the original report on the isolation,
we found (+)-discodermolide to be stable as a crystalline solid.
However, when (+)-discodermolide was concentrated from
CDCl3, decomposition occurred, due to residual acid derived
from the deuteriochloroform. Thus, the reported instability of
natural discodermolide is most likely an artifact.
Summary. A highly convergent total synthesis of discoder-
molide (1) has been achieved, exploiting a common precursor
(CP) to permit rapid and efficient construction of three advanced
subtargets. The subtargets, each of which embody the common
triad of contiguous stereogenic centers, were then coupled via
highly stereoselective σ-bond (palladium-catalyzed cross-
coupling) and π-bond (Wittig) constructions. Importantly, the
synthesis is highly efficient, proceeding in 6% overall yield
(longest linear sequence, 24 steps), stereocontrolled, and
ether, a protecting group interchange was required. Initially we
selected the TBS moiety to protect the primary hydroxyl.
However, during reductive-opening of the acetal, we experienced
competitive removal of the silyl group.69 Accordingly, we
exploited the trityl group. Reduction of 97 with DIBAL then
gave the primary alcohol (Scheme 21) which in turn was
subjected to oxidation with buffered Dess-Martin periodinane,70
installation of the diene via the Yamamoto diene protocol,53
and removal of the trityl group71 to furnish 98 in yields ranging
from 50% to 74% (four steps) with good diastereocontrol (8-
12:1 Z/E, 10 g scale). As will be described, the undesired (E)-
congener eventually proved inconsequential since it could be
readily removed at a later stage in the synthesis (vide infra).
Conversion of alcohol 98 to the corresponding phosphonium
salt via our previously optimized high-pressure conditions
proceeded without difficulty, providing phosphonium salt 99
(68) Both Myles and Paterson employed a two-step diene synthesis in
their syntheses of discodermolide. In our hands this route did successfully
deliver discodermolide in 31 total steps (22 in the longest linear sequence);
however, due to the capricious yields observed, this approach was not
conducive to large-scale synthesis.
(69) All attempts to remove selectively the primary PMB ether in the
presence of the secondary PMB ether proved unsuccessful.
(70) Dess, D. B.; Martin, J. C. J. Am. Chem. Soc. 1991, 113, 7277.
(71) Boeckman, R. K, Jr.; Potenza, J. C.; Tetrahedron Lett. 1985, 26,
11, 1411.
(72) For a discussion on this side reaction see: Schnell, A.; Tebby, J.
C. J. Chem. Soc., Perkin Trans. 1 1977, 1883.
(73) We thank Novartis for informing us that discodermolide is stable
in acetonitrile.

8664 J. Am. Chem. Soc., Vol. 122, No. 36, 2000
Smith et al.
Scheme 22
amenable to gram-scale production of this potentially important
natural product.
assistance in securing and interpreting high-field NMR spectra,
X-ray crystal structures, and mass spectra, respectively.
Acknowledgment. Financial support was provided by the
National Institutes of Health (Institute of General Medical
Sciences) through Grant GM-29028, and NIH Postdoctoral
Fellowship to T.J.B and M.D.K. We also thank Novartis for
financial support and for supplying spectral data for (+)-
discodermolide in CD3CN. Finally, we thank Drs. George T.
Furst and Patrick J. Carroll and Mr. John Dykins of the
University of Pennsylvania Spectroscopic Service Center for
Supporting Information Available: Experimental proce-
dures and characterization data for all synthetic intermediates
(PDF). This material is available free of charge via the Internet
at http://pubs.acs.org.
JA0015287