Interphylal product splicing: The first total syntheses of cephalostatin 1, the North hemisphere of ritterazine G, and the highly active hybrid analogue, ritterostatin G(N)1(N)1

Article abstract of DOI:10.1021/ja972160p

Convergent total syntheses of the extremely potent cell growth inhibitor cephalostatin 1 and two hybrid analogues, ritterostatins G(N)1(N) and G(N)1(s), have been achieved. Ritterostatin G(N)1(N) displays sub-nanomolar activity in the 60 cell line human t

Full text of DOI:10.1021/ja972160p

692
J. Am. Chem. Soc. 1998, 120, 692-707
Interphylal Product Splicing: The First Total Syntheses of
Cephalostatin 1, the North Hemisphere of Ritterazine G, and the
1
Highly Active Hybrid Analogue, Ritterostatin GN1N
Thomas G. LaCour,*,† Chuangxing Guo,† Sudhakar Bhandaru,† Michael R. Boyd,‡ and
P. L. Fuchs*,†
Contribution from the Department of Chemistry, Purdue UniVersity, West Lafayette, Indiana 47907, and
Laboratory of Drug DiscoVery, Research and DeVelopment, National Cancer Institute,
Frederick, Maryland 21702
ReceiVed June 30, 1997
Abstract: Convergent total syntheses of the extremely potent cell growth inhibitor cephalostatin 1 and two
hybrid analogues, ritterostatins GN1N and GN1S, have been achieved. Ritterostatin GN1N displays sub-nanomolar
actiVity in the 60 cell line human tumor panel of the National Cancer Institute. The North hemisphere of
ritterazine G was efficiently constructed from hecogenin acetate in 15% yield over 13 steps. Extension of a
key photolysis/Prins sequence to intermediates 19 and 32 proceeded in excellent yield, leading to installation
of the ∆14 moiety in the North G and South 1 steroidal subunits. Application of a method for directed
unsymmetrical coupling furnished the natural and analogue pyrazines in good yield from the cephalostatin
and ritterazine components.
Introduction
from the marine tube worm Cephalodiscus gilchristi in the
Indian Ocean,2,3 by Pettit at Arizona State University, and from
the tunicate Ritterella tokioka off the coast of Japan,4 by Fusetani
at the University of Tokyo (Figure 1). Surprisingly, they are
structurally related, featuring the union of two C27 steroids
(referred to as North and South) taken from an array of six basic
skeletal units which may be seen as substituted isomers of the
abundant plant-derived steroid hecogenin.5 Clinical trials of
cephalostatins 1 (1) and 7 (2) have stalled because of severe
difficulties in harvesting these rare materials (0.1 g of 1, <60
mg of 2, from 450 kg of worm) by SCUBA operations at 60-
80 m in the white shark infested waters off East Africa.6 This
scarcity will be nontrivial to alleviate via synthesis due to the
complexity of their steroid substructures, as evidenced by the
heroic preparations of cephalostatin 7 (2)7 and the 14′R,15′-
dihydro analogue of 1,1 wherein each of the 3-ketosteroid
precursors required 28-33 steps (1-3% overall yield) from
hecogenin acetate. Interestingly, several of the ritterazines
exhibit cytotoxicities approaching the same nanomolar range
as the most active of the cephalostatins, although the ritterazines
are far less oxygenated. Their relative simplicity promises
Cephalostatin 1 (1)2 is among the most powerful anticancer
agents ever tested by the National Cancer Institute. Cepha-
lostatin 7 (2)3a and ritterazine G (3)4 are other very potent (sub-
nanomolar antineoplastic activity) members of an expanding
group of trisdecacyclic pyrazines isolated from different phyla:
* To whom correspondence should be addressed. E-mail: pfuchs@
chem.purdue.edu or lacourtg@chem.purdue.edu. Fax: (765) 494-7679.
† Purdue University.
‡ National Cancer Institute.
(1) Cephalostatin synthesis. 11. Portions of this work have been
communicated in ref 12a and in paper 10 of this series: Guo, C.; Bhandaru,
S.; Fuchs, P. L.; Boyd, M. R. J. Am. Chem. Soc. 1996, 118, 10672; see
also references therein for syntheses of the cephalostatin subunit precursors
29 and 37 from hecogenin acetate.
(2) Pettit, G. R.; Inoue, M.; Kamano, Y.; Herald, D. L.; Arm, C.;
Dufresne, C.; Christie, N. D.; Schmidt, J. M.; Doubek, D. L.; Krupa, T. S.
J. Am. Chem. Soc. 1988, 110, 2006, reported ED50 < 10-4 ng/mL, corrected
) 0.01-0.001 ng/mL (P388 leukemia). Mean GI50 1.2 nM (ref b), 2.3 nM
(ref 1), 3.5 nM (8/4/97 test, this work).
(3) (a) Pettit, G. R.; Kamano, Y.; Inoue, M.; Dufresne, C.; Boyd, M. R.;
Herald, C. L.; Schmidt, J. M.; Doubek, D. L.; Christie, N. D. J. Org. Chem.
1992, 57, 429. Neither precise ED50 (P388 mouse leukemia) nor GI50 values
were ever reported. The paper states that “Cephalostatins 7-9 displayed
remarkable potency with TI50 (molar) values of 10-9 to <10-10 against a
number of (e.g. renal RXF-393 ... leukemia RPMI-8226) cell lines and values
of 10-8-10-9 for the breast MCF-7 cell line...” Our own testing results
(8/4/97) with synthetic cephalostatin 7 (2) revealed a mean GI50 of -6.8
(150 nM) across the entire 60-cell line panel, with certain cell lines showing
greater effect: GI50 -8.1 (7.9 × 10-9 molar ) 7.9 nM) renal RXF-393,
-8.0 (10 nM) leukemia RPMI-8226, but -6.6 (229 nM) for the breast
MCF-7 line, also see ref 2. Finally, Professor Pettit has indicated that these
latter results are in accord with his own and may be considered definitive.
(Personal communication, G. R. Pettit, December 1997.) (b) Pettit, G. R.;
Xu, J.; Schmidt, J. M.; Boyd, M. R. Bioorg. Med. Chem. Lett. 1995, 5,
2027 and references therein.
(4) (a) Fukuzawa, S.; Matsunaga, S.; Fusetani, N. Tetrahedron 1995,
51, 6707 and references therein; ritterazines B and G, IC50 0.15 and 0.73
ng/mL (P388 leukemia). (b) Fukuzawa, S.; Matsunaga, S.; Fusetani, N. J.
Org. Chem. 1997, 62, 4484. The southern hemisphere is acid sensitive and
isomerizes even at 25 °C to several compounds, including some apparently
related to North 5, the sixth and least active “basic”steroidal subunit: see
refs 5, 7, and 11.
(5) “North” has also been designated “right side” by Pettit2,3 and “East”
by Fusetani.4 Ranked by the bioactivities of the pyrazines containing
them: North and South 1, South 7 and North G, North A, North 5.
Cephalostatin basic units: North 1, South 1, South 7, and North 5.
Ritterazine basic units: North A and North G, with 7âOH-South 7 (“South
A”, not a “basic” unit) appearing commonly. In fact, South 7 and North G
may be viewed as deoxygenated spiroketal isomers of North 1, and 7â-
OH-South 7 as a spiroketal isomer of North 1 with translation of one
hydroxyl group. North G is considered a basic unit rather than North B
(the 14â dihydro, 22R spiroketal isomer of North G since the ∆14 function
permits access to all the D ring functions seen in the ritterazines; see ref 4.
(6) Pettit reports that ∼0.5 ton of the organisms yielded approximately
100 mg of cephalostatin 1 (1); ∼1 g of material is required for the initial
phases of the trials. Personal communication, Professor G. R. Pettit, 12/94.
(7) (a) Jeong, J. U.; Sutton, S. C.; Kim, S.; Fuchs, P. L. J. Am. Chem.
Soc. 1995, 117, 10157. (b) Kim, S.; Fuchs, P. L. Tetrahedron Lett. 1994,
35, 7163.
S0002-7863(97)02160-4 CCC: $15.00 © 1998 American Chemical Society
Published on Web 01/17/1998

Cephalostatin Synthesis. 11
J. Am. Chem. Soc., Vol. 120, No. 4, 1998 693
Figure 1.
greater synthetic accessibility with probable retention of sig-
nificant bioactivity.
Nature has provided a wealth of SAR data in the form of the
43 variations on this theme so far isolated by Pettit3 and
Fusetani,4 although a number of questions remain. Subtle
variations of stereochemistry and substitution on the six basic
steroid units display marked effects on bioactivity. We and
others have addressed some of these questions by the syntheses
of dihydrocephalostatin 1,1 derivatives of ritterazine B,4b and
pyrazines related more closely to hecogenin.10 Provocatively,
the 11 most potent (subnanomolar activity) pyrazines of the
natural series utilize only the four basic units North 1, South 1,
South 7, and North G (Figure 1),5 in four of the six possible
unsymmetric combinations, of which the South 1 / North 1
combination 1 is the most potent yet seen.
The unknown combinations are those of North G with North
1 and South 1, which would constitute the union of units
from disparate organisms, hence “interphylal natural product
splicing.” Indeed, North 1 and 7âOH-South 7 (the southern
hemisphere of many of the ritterazines, including ritterazine G
3) have the same level of oxygenation, differing only in the
location of one hydroxyl group (C23 vs C7) and in their isomeric
spiroketal arrangements. Since 1 is significantly more potent
than 2 or 3, which utilize South 7-type subunits, the question
of such chemical “cross-breeding” takes on significance.
Directed efforts to answer this and other remaining synthetic
and SAR challenges may lead not only to new chemical insights
but to practical and potent chemotherapeutic agents.
Figure 2.
Thus, in addition to the important goal of completing the first
total synthesis of the natural product 1, and as part of general
SAR explorations, we consider it useful to construct these
hybrids composed of ritterazine and cephalostatin steroidal
subunits (Figure 2). In the present study, we describe the first
synthesis of the North hemisphere of ritterazine G via a formal
isomerization of the spiroketal group of hecogenin acetate,
extension of an optimized photolysis/Prins protocol to install
the ∆14 functionality, use of the protocol to deliver the same
moiety into the South hemisphere of cephalostatin 1 from an
advanced saturated intermediate, and application of the unsym-
metrical pyrazine formation method developed in these labo-
ratories1 to provide the first total syntheses of cephalostatin 1
(1) and the hybrid analogues ritterostatin GN1N (4)8 and
ritterostatin GN1S (5).8
(8) The naming scheme employed for the unnatural ritterazine-cepha-
lostatin hybrids combines the lettering scheme of Fusetani4 with the number
designations of Pettit2,3 and is necessarily further extended by adopting a
“North-South” designator to indicate which hemisphere of each material
is present in the new analogue. Thus, ritterostatin GN1N (4) unambigiously
describes a material composed of the North steroidal unit of ritterazine G,
combined with the North steroidal segment of cephalostatin 1. In concept,
one could also name the same material cephalozine 1NGN. While this latter
designation is extremely attractive in recognition of Pettit’s seminal
contributions to the discovery and structural elucidation of the cephalostatins,
we ultimately opted for the ritterostatin appellation in deference to the
“statin” suffix which provides more immediate information vis-a`-vis the
compound’s biological activity.
(9) (a) Smith, S. C.; Heathcock, C. H. J. Org. Chem. 1992, 57, 6379.
(b) Heathcock, C. H.; Smith, S. C. J. Org. Chem. 1994, 59, 6828.
(10) Drogemuller, M.; Jautelat, R.; Winterfeldt, E. Angew. Chem., Int.
Ed. Engl. 1996, 35, 1572.

694 J. Am. Chem. Soc., Vol. 120, No. 4, 1998
Scheme 1. Guo, Bhandaru, and Fuchs (ref 1)
LaCour et al.
Scheme 2. Smith and Heathcock (ref 9)
Since we wish to be in a position to construct an array of
unsymmetrical pyrazines, the choice of method for coupling
different subunits is of prime concern. We recently reported
the first total synthesis of cephalostatin 7 (2) via a biomimetic
approach in which the central pyrazine ring was constructed
by a statistical coupling of the North and South R-amino
3-ketosteriods (produced in situ from the corresponding R-azido
ketosteroids);7 however, such a method is inappropriate for our
present purpose. Rather, the central pyrazine ring of cepha-
lostatin 1 (1) was envisioned as arising from the directed reaction
of the R-aminomethoxime “North 1” (6) with R-azidoketone
“South 1” (7) via the newer protocol that was successfully
employed to synthesize dihydrocephalostatin 1, the 14′R,15′-
dihydro analogue of 1 (Scheme 1).1 Likewise, the appropriate
derivatives of “North G” (8) would be coupled with 6 and 7 to
give the analogues. This method for the synthesis of cross-
coupled, “C2-pseudosymmetric” (i.e., trans vs cis) pyrazines is
much milder (80 °C, 3-6 h) and better yielding (60-90%) than
the seminal procedure reported by Heathcock and Smith, which
involved the reaction of an R-aminomethoxime with an R-acet-
oxy ketone at elevated temperatures (90-140 °C) for 2 days
with yields of 29-43% for two cases (Scheme 2).9 The
seemingly trivial substitution of an azido ketone for the acetoxy
ketone as the acceptor partner for imine formation has two
important consequences. The first of these simply relates to a
more efficient preparation of the acceptor (∼80%, two steps vs
∼40%, three steps). The more important difference pertains
to a change of mechanism, for the reaction produces N2 gas
and becomes basic (instead of acidic, as in Heathcock’s method)
through the loss of methoxylamine.
More recently, Winterfeldt and co-workers reported construc-
tion of nonsymmetrical steroid pyrazines via the reaction of a
steroidal azirine, formed in situ from a vinyl azide, with a
steroidal amino enone (Scheme 3).10 While the conditions and
yields for the highly creative coupling step (100 °C, dioxane,
PPTs, 51-67% for two cases) are more competitive with those
of our methodology, the azirine strategy suffers some draw-
backs: (i) a longer synthetic sequence to prepare the vinyl azide
from a 3-ketosteroid, with attendant loss of precious late-stage
material, and (ii) the use of homogeneous acid catalysis, which
causes isomerization of labile spiroketals such as that present
in the South 7 subunit.4b,11 The coupling protocol described
herein is milder, functions well with heterogeneous base
catalysis, and permits rapid derivation (two to four steps) of
the coupling partners from the parent 3-ketosteroids in high yield
(vide infra). Schemes 1-3 illustrate plausible mechanisms for
the three methods. For reasons of brevity, mildness and yield,
(11) (a) The Southern hemisphere of cephalostatin 7, “South 7”, contains
an acid-labile [5,6] spiroketal which undergoes extensive isomerization
(<50% remaining South 7 after brief reflux at 80 °C with catalytic PPTs)
when heated with acid and has exhibited a tendency to isomerize under
even mild provocation to a mixture of spiroketals (Jeong, J. U.; Guo, C.;
Fuchs, P. L. Synthesis of the South Hexacyclic Portion of Cephalostatin 7,
manuscript in preparation, and see ref 23). The 20R,22R [5,5] spiroketal
isomer, precisely “South C” (the southern hemisphere of ritterazine C, is
calculated to lie only 0.04 kcal above that of the natural 20R,22R [5,6]
spiroketal of South 7. (b) Calculations performed using CAChe 3.1; see
the Supporting Information for the energies of all isomers.

Cephalostatin Synthesis. 11
J. Am. Chem. Soc., Vol. 120, No. 4, 1998 695
Scheme 3. Drogemuller, Jautelat, and Winterfeldt (ref 10)
we therefore chose the same unsymmetrical coupling utilized
in the synthesis of the dihydrocephalostatin 1 analogue.1 In
this way, the union of “North 1” aminomethoxime 6 with “South
1” azido ketone 7 leads to cephalostatin 1 (1), that of 6 with
“North G” azido ketone 9 leads to ritterostatin GN1N (4), and
that of 7 with “North G” aminomethoxime 10 leads to
ritterostatin GN1S (5).
11 affords lumihecogenin acetate 13 as the main product (∼80%
on 250 g scale), which on further irradiation undergoes Paterno-
Bu¨chi cyclization to photohecogenin acetate 14 (Scheme 4).
Chinn15b demonstrated that oxetane 14 participated in the same
acid-catalyzed ene and Prins reactions by prior conversion back
to 13. Welzel more fully explored these reactions, identifying
both of the Prins products and correcting the assignment of the
14-hydroxy product stereochemistry. Additional pertinent
transformations of the photoproducts reported by Bladon,15a
Chinn,15b Welzel,13a and Winterfeldt14b are summarized in the
scheme. Under Welzel’s protocol, a mixture13b of 13 and 14
underwent the intramolecular Prins reaction16 in modest yield
to give a diastereomeric mixture of diols 15 (6.5:1 R/â, 52%),
which on oxidation17 afforded the keto alcohol 17 (67%; also
accessible in one pot by oxidation of 13 in AcOH). Elimination
of the 14â-hydroxyl gave the ∆14 ketone 18 (69%) contaminated
by an alkyl chloride.18 The calculated conversion from ketone
11 to 17 of 27%, 19% overall to impure 18, was somewhat
daunting. After our investigations of the 11 to 18 sequence,
Winterfeldt has published an improved BF3‚OEt2-mediated ene
reaction of 13, achieving an 80% yield of ∆14 alcohol 16R from
ketone 11 in only two steps,14b now making 16R extremely
attractive as an alternate precursor to 18.37
In light of studies on fundamental carbonyl photochemistry,
insights of practical significance may be gleaned by examination
of the factors that account for the formation of Norrish type I
(R-cleavage) product 13 in such preparatively useful yield. It
appears that structural features of 11, both gross and subtle,
play a most significant role, but the solvent must also be
considered. The acyl and tertiary alkyl radicals produced
(reversibly) by R-cleavage are relatively stable, and intramo-
lecular 14R-H abstraction to give the secoaldehyde 13 may
proceed without undue strain, so that 11-ketene is not formed;15b
however, the stability of such radicals alone is insufficient to
ensure exclusive type I cleavage. Type II reactions of excited
ketones often proceed at a higher rate, and the well-known
As part of our ongoing program to synthesize the most
bioactive members of the ritterazines and cephalostatins, we
desired a more direct route to functionalization of ring D than
that employed in previous syntheses which also delivered C-17
oxygenation.1,7 The ∆14 moiety is present in 55 of the 86
steroidal subunits of the known trisdecacyclic pyrazines, includ-
ing the North half of ritterazine G (3) and the South half of
cephalostatin 1 (1), neither of which require a 17-hydroxyl. To
efficiently access this unsaturation from existing synthetic
materials,12 we inquired whether the known photolysis/Prins/
elimination strategy of Welzel13a (recently employed on heco-
genin acetate to generate pyrazine steroidal subunits by
Winterfeldt)10,14a could be utilized for our substrates. The
importance of at least one such unsaturation to bioactivity is
indicated by the low activities of Winterfeldt’s saturated
analogues,10 our own 14′â,15′-dihydrocephalostatin 1 analogue
(still highly active with one ∆14 in the North and a trans-fused
C/D junction in the South, but less active than 1 in which both
hemispheres are unsaturated),1 and the inferior activity of the
natural fully saturated ritterazines.4b This moiety is also essential
for synthetic access to the type of cis-fusion of the C/D rings
found in the most potent ritterazines such as B and F, and other
14-hydroxylated or epoxidized subunits.3,4
Welzel’s protocol drew from the pioneering work of Bladon.15a
Specifically, Bladon found that irradiation of hecogenin acetate
(12) (a) Bhandaru, S.; Fuchs, P. L. Tetrahedron Lett. 1995, 36, 8351.
(b) Bhandaru, S.; Fuchs, P. L. Tetrahedron Lett. 1995, 36, 8347.
(13) (a) Welzel, P.; Janssen, B.; Duddeck, H. Liebigs. Ann. Chem. 1981,
546 and references therein; (b) In response to a reviewer’s question, Welzel
reports reactions on what he terms “Gemisches von 4 und 5” [here 13 and
14] obtained directly from photolysis of 11; Winterfeldt (ref 14a) makes
no mention of the composition of his photolysate, but simply references
Bladon and Welzel. However, the crude product from photolysis demon-
strably contains multiple products and is not merely a mixture of 13 and
14. The amorphous mass in the mother liquor after crystallization of 13
contained little or no 14 by NMR, and its quantitative conversion to diols
15, as well as the quantitative conversion of the crude photolysate, is in
our opinion a repetition of neither Winterfeldt’s nor Welzel’s efforts but a
useful improvement; (c) Welzel has also extended this protocol to the
synthesis of cardenolides and bufadienolides from less-strained, tetracyclic
steroids, although the yields in this series suffer by comparison to hecogenin
acetate 11; see ref 30.
(14) (a) Kramer, A.; Ullmann, U.; Winterfeldt, E. J. Chem. Soc., Perkin
Trans. 1 1993, 2865. (b) Jautelat, R.; Winterfeldt, E.; Muller-Fahrnow, A.
J. Prakt. Chem. 1996, 338, 695.
(15) (a) Bladon, P.; McMeekin, W.; Williams, I. A. J. Chem. Soc. 1963,
5727 and references therein. (b) Chinn, L. J. J. Org. Chem. 1967, 32, 687,
reported only 15R. Chinn and Bladon (ref. 15a) both assigned the
stereochemistry of the 14-OH group in 15 and 17 as “R” but were
subsequently corrected by Welzel: the correct assignment is “â” as shown
(ref 13).
(16) (a) Prins, H. J. Chem. Weekbald 1919, 16, 1510 (b) Arundale, F.;
Mikeska, L. A. Chem. ReV. (Washington, D.C.) 1952, 51, 505 (c) Smissman,
E. E.; Schnettler, R. A.; Portoghese, P. S. J. Org. Chem. 1965, 30, 797.
(17) “Jones” reagent, originally known as “Kiliani reagent”, prepared
as per Bladon et al. J. Chem. Soc. 1951, 2402.
(18) Welzel reports that the product isolated after chromatography (69%)
gave a peak in its mass spectrum which corresponds to formal HCl addition
to 18; the amount of this impurity was estimated to be ∼10% of the isolated
material. Under our modified conditions (-10 to 0 °C instead of room
temperature, much less SOCl2, early quench) the reaction was found to
give a quantitative mass balance of material containing ∼15-20% of such
side product (varying by run, amount estimated by NMR), tentatively
identified as the 14â chloride, which almost completely coeluted with desired
18 (TLC in several systems show a single spot; crystallization has not yet
permitted separation). Careful chromatography furnished an 83% yield of
material which was nearly pure (by sacrifice of later fractions, which
contained more substantial amounts of the chloride mixed with the desired
olefin) but which still contained ∼6-9% of the chloride. While we have
not yet been able to isolate this chloride wholly pure, MS and NMR
comparisons to 14â chloride 25d make its identification as the 14â chloride
persuasive.

696 J. Am. Chem. Soc., Vol. 120, No. 4, 1998
LaCour et al.
Scheme 4a
a (a) hν, dioxane; (b) 75% AcOH in H2O, 25 °C; (c) BF3‚OEt2, benzene, 25 °C; (d) CrO3, aqueous H2SO4, acetone, 25 °C, 5 min; (e) PCC,
CH2Cl2, 25 °C; (f) SOCl2, Pyr, 0-25 °C, 1 h; (g) CrO3 in aqueous AcOH, 25 °C; (h) LiAlH4, THF, 25 °C; (i) NaBH4, MeOH, 25 °C; (j) excess
BF3‚OEt2, toluene, 0 °C.
Scheme 5
or net type II product as with camphor22 (Scheme 5). The
importance of ring-strain relief to successful dominance of
R-cleavage is further demonstrated by the behavior of 17-oxo
androstanes, which often display only epimerization (by re-
combination of type I acyl/tertiaryalkyl radicals) and reduction.23
Excessive ring strain, while useful for maximizing type I
cleavage even when the ketone is only R-disubstituted, as in
11-oxo-nor-C-steroids vs 11-oxosteroids of otherwise similar
structure, may prevent subsequent recyclization (Scheme 5).24
Ketone 11 apparently possesses the requisite balance of ring
strain for both photolysis and cyclization to proceed. Its rigid
structure might preclude any â-cleavage resulting from inter-
molecular, solvent assisted production of a type II 1,4-biradical
(via net γ-hydrogen abstraction at C16, C20, or C8) since
considerable reorganization would be required for sufficient
overlap of the developing p-orbitals.21d Thus, aside from
reduction, we may surmise that only products deriVed from the
type I process might be expected from photolysis of 11.
Results and Discussion
Synthesis of South 1 (7) and North G (8). Our initial
approach to North G (8) from 11 relied upon this method to
introduce the ∆14 moiety in 18, with intended subsequent
propensity of R-trialkyl ketones to suffer R-scission was shown
by Yang, Lewis, Wagner, and others to be subject to effective
competition from type II pathways when γ-H (or even δ-H in
suitable substrates)19 abstraction is conformationally available.20
In 11, no such hydrogen atom is within reach of the excited
carbonyl oxygen, and type I should predominate.21 Yet even
when intramolecular abstraction is structurally precluded,
intermolecular participation by solvent can lead to reduction
(21) (a) Lewis, F. D.; Johnson, R. W.; Johnson, D. E. J. Am. Chem.
Soc. 1974, 96, 6090. For discussion of transition-state geometry for type II
reaction and subsequent reactions of the 1,4 biradical, see: (b) Lewis, F.
D.; Johnson, R. W.; Kory, D. R. J. Am. Chem. Soc. 1974, 96, 6100. (c)
Wagner, P. J.; Kelso, P. A.; Kemppainen, A. E.; Zepp, R. G. J. Am. Chem.
Soc. 1972, 94, 7500. (d) Wagner, P. J.; Kelso, P. A.; Kemppainen, A. E.;
McGrath, J. M.; Schott, H. N.; Zepp, R. G. J. Am. Chem. Soc. 1972, 94,
7506. (e) Dawes, K.; Dalton, J. C.; Turro, N. J. Mol. Photochem. 1971, 3,
71. (f) Wagner, P. J. Acc. Chem. Res. 1971, 4, 168.
(19) For a leading review, see: Wagner, P. J. Acc. Chem. Res 1989, 22,
83.
(20) (a) Yang, N. C.; Feit, E. D. J. Am. Chem. Soc. 1968, 90, 504. (b)
Wagner, P. J.; McGrath, J. M. J. Am. Chem. Soc. 1972, 94, 3849. (c) Lewis,
F. D.; Hilliard, T. A. J. Am. Chem. Soc. 1972, 94, 3852. For reviews,
including discussion of the relative rates for Type II reaction, see: (d) Turro,
N. J.; Dalton, J. C.; Dawes, K.; Farrington, G.; Hautala, R.; Morton, D.;
Niemczyk, M.; Schore, N. Acc. Chem. Res 1972, 5, 92. (e) Wagner, P. J.
Acc. Chem. Res 1983, 16, 461.
(22) (a) Yates, P. Pure Appl. Chem. 1968, 93. (b) Srinivasan, R. J. Am.
Chem. Soc. 1959, 81, 2604.
(23) Wehrli, H.; Schaffner, K. Ber. 1962, 45, 385 and references therein.
(b) Butenandt, A.; Poschmann, L. Ber. Deutsch. Chem. Ges. 1944, 77, 392
and 394.
(24) Iriarte, J.; Schaffner, K.; Jeger, O. Ber. 1964, 47, 1255 and references
therein.

Cephalostatin Synthesis. 11
J. Am. Chem. Soc., Vol. 120, No. 4, 1998 697
Scheme 6
Scheme 7a
a (a) (Cl2CHCO)2O, 150-160 °C; (b) KOH/aqueous MeOH; (c) TsCl/pyr; (d) NaI; (e) DBU; (f) (i) 75% HOAc, (ii) Ac2O/pyr.
opening of the [5,6] spiroketal and manipulation to the [5,5]
system. This entry proved quite rewarding in its first stage, as
several improvements in the protocol were developed (Scheme
6). We noted that the crude photolysate from 11 was composed
of not only 13 and 14 as intimated by Welzel but rather many
products (a fact alluded to by Bladon). Examination by NMR
and MS revealed neither undesired vinyl (via net γ-hydrogen
abstraction at C16, C20, or C8) nor 12-OH (via reduction)
compounds derived from solvent participation, indicating that
dioxane remains effectively inert to this compound. Suspecting
that the normal plethora of unidentified photo side products
might, like oxetane 14, be convertible to secoaldehyde 13 and
thence to desired diols 15, we first attempted such conversion
on the complex mixture found in the mother liquor after
crystallization of 13. Several conditions (75% AcOH as per
Welzel, dilute H2SO4, aqueous HBr/HOAc, aqueous TFA/CH2-
Cl2) gave inferior results, but extended exposure to 75% AcOH
in a much more dilute reaction than indicated in the literature
proved excellent, delivering pure diols 15 from the complex
mixture. We further noted that 15R (but not 15â) is easily air-
oxidized and that silica gel causes some degradation of both
15 and 17, so care was taken to keep the oxidation cold and
brief to achieve quantitative, clean conversion to 17 (the
“milder” Brown-Jones method25 gave inferior yields and
purity). Subsequently applying these observations, and avoiding
chromatography, we were pleased to find that irradiation of a
dioxane solution of 11 followed by slurry of the crude photoly-
sate with 75% acetic acid (dilute, 25 °C, 24 h) smoothly and
quantitatiVely effects the intramolecular Prins reaction to yield
a ∼5:1 mixture of diols 15R/15â.26 This crude mixture gaVe,
upon Jones oxidation,17 keto alcohol 17 in a remarkable 94%
oVerall yield (99% when the 5% tigogenin (12-deoxyhecogenin)
acetate present in commercial 11 is considered) for three steps
from 11. Dehydration of alcohol 17 to keto olefin 18 was
accomplished in 83% yield18 by modifying (cold, brief) Welzel’s
conditions. Other attempts to achieve improved conditions for
elimination were totally unrewarding (POCl3,31a Swern,27
BF3‚Et2O,28 sulfide or bromide installation during the Prins
reaction, mesylation, xanthate formation, etc.). The chloride
side product18 always present in this alkene proved surprisingly
resistant to elimination (DBU, hot pyridine, etc.). Further work
on the transformation was suspended because, unfortunately,
neither the 14â-OH (which resisted vigorous attempts at
protection) nor the ∆14 moieties37 were stable to the conditions
necessary (>155 °C, acid anhydride) for spiroketal opening.29
Before commencing a potentially lengthy search for unprec-
edented spiroketal-opening conditions which would allow us
to utilize D-ring functionalized materials 16-18, we considered
reversing the order of events. Our experience with spirostan-
12-one 11 under the Welzel protocol encouraged attempted
application to the related furostan-12-one 19, which would arise
from 11 by a prior spiroketal adjustment sequence (Scheme 7).
The method had been extended by Welzel to 12-ketosteroids
appropriate to the syntheses of cardioactive digitalins, bufadi-
enolides, and cardenolides, but these tetracyclic precursors
invariably gave poorer yields in the photolysis step,30 possibly
due to insufficient ring strain and type II competition. Indeed,
a change of solvent from dioxane to CH2Cl2 was warranted.
Further, there are many cases of substantial variance in the
reactivities of distally differentiated steroids,13,29,31 wherein what
appears to be a minor perturbation in the steroid structure
(27) Corey, E. J.; Gin, D. Y.; Kania, R. S. J. Am. Chem. Soc. 1996, 118,
9202.
(28) Posner, G. H.; Shulman-Roskes, E. M.; Oh, C. H.; Carry, J.-C.;
Green, J. V.; Clark, A. B.; Dai, H.; Anjeh, T. E. N. Tetrahedron Lett. 1991,
32, 6489.
(29) Jeong, J. U.; Fuchs, P. L. J. Am. Chem. Soc. 1994, 116, 773-774.
See also: Cameron, A. F. B.; Evans, R. M.; Hamlet, J. C.; Hunt, J. S.;
Jones, P. G.; Long, A. G. J. Chem. Soc. 1955, 2807 for high-yield openings
using fatty acid solvents (T > 150 °C) but which also fail on unsaturated
steroids.
(30) (a) Digipurpurogenins I and II (∼30% in the photostep): Welzel,
P.; Moschner, R.; Ponty, A.; Pommerenk, U.; Sengewein, H. Liebigs. Ann.
Chem. 1982, 564. (b) Digitoxigenin (from hecogenin acetate): Milkova,
T.; Stein, H.; Ponty, A.; Bo¨ttger, D.; Welzel, P. Tetrahedron Lett. 1982,
23, 413. (c) Milkova, T.; Stein, H.; Welzel, P. Liebigs. Ann. Chem. 1982,
2119. (d) Bufalin (36% in the photostep): Hoppe, H.-W.; Welzel, P.
Tetrahedron Lett. 1986, 27, 2459. (e) Digoxigenin (72% in the photostep):
Stein, H.; Welzel, P. Tetrahedron Lett. 1981, 22, 3385 and ref 30c.
(31) For leading examples, see: (a) Fieser, L. F.; Fieser, M. Steroids;
Rheinhold: New York, 1959, and references therein. (b) Djerassi, C.
Reactions of Steroids; Holden-Day: San Francisco, 1963, and references
therein.
(25) Brown, H. C.; Garg, C. P.; Liu, K.-T. J. Org. Chem. 1971, 36, 387.
(26) Interestingly, the Prins reaction produces a higher proportion of 15â
if allowed to stir for extended periods of time. The implied equilibrium
between the diols was explored by resubjecting pure 15R (the apparent
kinetic product) to the reaction conditions (25 °C) for 15 h, whereupon a
6:1 ratio of 15R/15â was obtained, presumably via a retro-Prins/Prins
sequence. A 4:1 mixture of 15R/15â was heated at 75 °C in 75% acetic
acid for 15 h, at which time a 1:6 ratio of 15R/15â was produced, the
thermodynamic product 15â greatly predominating; neither decomposition
nor dehydration was observed. For similar evidence of such equilibrium in
a related 1,3-diol, see ref 30a.

698 J. Am. Chem. Soc., Vol. 120, No. 4, 1998
LaCour et al.
Scheme 8a
a (a) hν, dioxane, 4 h; (b) 75% AcOH, 35 h; (c) CrO3, H2SO4, acetone, 5 min; (d) 2 equiv of SOCl2, 10 equiv of pyr, toluene, 5 min; (e) pTsOH,
CH3CN, 15 min, 85%; (f) DBU, CH3CN, reflux, 30 min, 80%; (g) DBU, LiF, 100 °C, 1 h, 10%.
exercises profound changes in chemical behavior. The E/F ring
strain difference expected between 11 and 19 which might be
communicated to the C/D rings, and the possible increase in
steric compression near the D ring in 19 were concerns, albeit
ones which fortunately proved surmountable with respect to
photolysis and Prins reaction (although all too well-founded with
respect to ene or elimination reactions, vide infra).
The spiroketal of 11 was therefore opened and tosylate
derivative 20c obtained utilizing standard procedures29 and then
eliminated via the iodide 20d with DBU to afford enol ether-
olefin 21 in excellent yield. Several attempts to add water in
Markownikoff fashion and form the [5,5] spiroketal were only
marginally successful (Hg(OAc)2/NaBH4, aqueous CH3CN with
TFA or TsOH), but 21 hydratively cyclized through the action
of hot aqueous acetic acid (with some loss of the 3-acetate,
necessitating reacetylation) to furnish furostanone 1932 in 73%
yield (98% based on recovered 21).
Differences, sometimes dramatic, were noted in the reactivity
of 19 compared to that of 11 throughout the remainder of the
sequence. Photolysis of the furostanone to secoaldehyde 22
proceeded more quickly but less cleanly, and the Prins reaction
yielded diols 23 more slowly (even as a solution rather than a
slurry as for 13, some unconverted photolysate remained after
36 h, but triol and ∆14 olefin were becoming evident).33 These
results bespeak the effect of increased C ring strain in 19 during
the protocol compared to the corresponding steps from
spirostanone 11, although the yields remained high by faithfully
carrying on the crude products to convergence (Scheme 8). By
contrast, a tempting shortcut to diols 23 via photolysis of alkene
21 was unrewarding. It was hoped that the R-cleavage product
would undergo one-pot successive Prins and hydrative spiro-
cyclization reactions by temperature-controlled exposure to
aqueous AcOH. Unfortunately, the desired cleavage was
excruciatingly sluggish. After a 3 h irradiation of 21 in dioxane,
only ∼7% aldehyde was produced, and after 20 h only a 1:2
ratio of aldehyde/starting ketone was obtained, with sizable
buildup of low Rf byproducts already evident. The reaction did
not approach completion even after 60 h. Dioxane participation
as an H-atom donor was ruled out by conducting the reaction
in acetonitrile,34 but even less aldehyde and more byproducts
were formed. Whether reduced C ring strain (from removal of
the 22-spiro connection), alkene quenching of the excited ketone
(by through-bond energy transfer35 to the C20,22 bond), type II
â-abstraction20 of the activated allylic 17R hydrogen (2.61 Å
away), or a combination of these factors could account for the
observations, it is clear that the method is no panacea applicable
to all 12-ketosteroids.
Oxidation of diols 23 also proceeded less smoothly than had
the analogous conversion of 15 to 17 which evidenced no
byproducts. Jones oxidation of 23 gave 14â-hydroxyfuro-
stanone 24, the North segment of ritterazine I, containing
5-10% of an inseparable side product, possibly the 22R
epimer.36 However, dehydration to keto olefin 25a was
significantly more involved. Using even our modified condi-
tions (vide supra), the reaction of 24 (stereochemistry confirmed
by X-ray)38 was lower yielding (42%) and much more complex
than reaction of the 14-hydroxyspirostanone 17. Revised
conditions (dilute in toluene, 10 equiv of pyridine, 2 equiv of
SOCl2 required) and repeated fractionation by chromatography
gave pure desired keto olefin 25a (63%) and four side products
25b-e. Repeated attempts to access the ∆14 moiety by ene
reaction of 22 under Winterfeldt’s conditions14b afforded only
complex mixtures from which the desired homoallylic alcohol
was obtained in poor yield.37
(32) The stereochemistry of 19 (20S,22S) was determined by Luche
reduction at C-12 (13:1 â/R) followed by acetylation, and comparison of
the resulting furostandiol diacetate to material previously synthesized by
an independent route (S. Ma, unpublished results) for which single-crystal
X-ray confirmation had been obtained.
(33) For production of a related homoallylic alcohol as a minor product
during the Prins, see ref 30a. Attempted equilibration of 23R by resubjection
to the reaction conditions for 18 h gave unchanged 23R (89%), 23â (6%),
and further 12R-hydroxy-14-alkene (5%). This confirmed both the acid-
catalyzed equilibration of the diols (noted for the spirostanols 15R/15â,
vide supra) and the willingness of this 12R-furostanol to produce the formal
ene adduct (also probably via a retro-Prins, but conceivably via direct
dehydration). An equilibration attempt on 23R at 95 °C resulted in extensive
decomposition.
(34) Gong, J.; Fuchs, P. L. Tetrahedron Lett. 1997, 38, 787 and references
therein.
(35) Agyin, J. K.; Timberlake, L. D.; Morrison, H. J. Am. Chem. Soc.
1997, 119, 7945. Since the excited endocyclic C20,22 π bond resulting
from intraTTET cannot relax via rotation and has no H atom within
abstractable range, we consider it possible that subsequent transfer, probably
through space, to the terminal 26-ene (a “free rotor”, see: Zimmerman, H.
E.; Epling, G. A. J. Am. Chem. Soc. 1972, 94, 8749) completes the
dissipation of the excitation energy. We cannot, at this point, exclude attack
of the excited E ring olefin/biradical onto the 26-ene moiety, but we have
no evidence for cyclic products or polymers expected from such addition,
nor those from 17H abstraction. See Scheme 8a in the Supporting
Information.

Cephalostatin Synthesis. 11
J. Am. Chem. Soc., Vol. 120, No. 4, 1998 699
Scheme 9
The 22R epimer 25b, for which we obtained single-crystal
X-ray confirmation,38 may arise from the side product in 24, or
directly from 25a by Lewis acid catalysis in toluene.36 Indeed,
resubjection of 25a to the reaction conditions gave a 10:1 ratio
of 25a/b. The epimer is rapidly (and completely!) consumed
to give mainly 25a with catalytic TsOH in acetonitrile (15 min,
25 °C), a solvent which we have found is superbly suited to
production of the thermodynamically most stable 20,22 arrange-
ment of the spiroketal moiety in several compounds. The
presence of 14â-chloride 25d was anticipated, based on earlier
results with 17.18 Formation of the 14R-chloride 25c (also
confirmed by X-ray)38 was not expected, as no evidence for
such a product was found in the dehydration of 14â-hydroxy-
spirostanone 17,18 and implies a stronger carbocation-like
character (E1 rather than E2) in the dehydration intermediate
from 24 than is observed with 17. Indeed, analysis of the
dehydration of 24 run in a solvent expected to better stabilize
such a carbenium ion (pyridine, as for 17) reveals formation of
much more (10%) of this chloride along with 25d (11%).
Elimination of 14R-chloride 25c proceeded relatively well
(DBU, CH3CN, reflux 3h, 80%), but the epimer 25d proved as
recalcitrant as the chloride side product accompanying 18.
Finally, LiF in DBU solvent39 (100 °C, 1 h) consumed 25d but
gave the desired elimination product in only poor yield,
delivering mainly baseline material. The final side product 25e
resisted crystallization and attempts to convert it to known
compounds; besides an apparent isomeric relationship to 25a/b
as a keto olefin, it has so far remained unidentified.
reaction of 12-ketosteroids which retain a trans (saturated) C/D
ring fusion or a [5,6] spiroketal: similar reductions of 11 (17:1
â/R) and 19 (13:1 â/R) were substantially more selective, as
was Winterfeldt’s uncatalyzed borohydride reduction of the 12-
keto function in the pyrazine dimer of 18 (∆14-spirostenone).14a
Reported reductions of cis-fused steroid 17 (Scheme 4) with
LAH or borohydride (1:1.6 â/R) showed the opposite selectiv-
ity.13,15 A greater divergence from previous experience occurred
in the selective saponification of the 3â-acetate group of
intermediate 27, normally facile with KHCO3 in aqueous
methanol at reflux.23 However, the 12â acetate of 27 proved
surprisingly labile to these and several other conditions (LiOH,
KOH, or KHCO3 in numerous solvents, concentrations, and
temperatures). Finally, potassium tert-butoxide (t-BuOK/i-
PrOH, 5 °C) provided a useful 68% yield of the 3-hydroxy-
12-acetate 28; two recycles (reacetylation of monool 26 and
the diol [not pictured], and resubjection to base) raised the
cumulative yield to 96%. In addition, Jones oxidation of 28 to
give 8 introduced 5% of an inseparable compound which
appears (again!) to be its 22â epimer (not shown), a recurrent
issue with this ∆14 [5,5] spiroketal system.36b A similar epimer
of 27 was also evidenced following one reacetylation step
(Ac2O/pyridine, allowed to stir at 25 °C overnight) of the
recycles producing additional 28 (vide supra), whereas acety-
lations conducted at 0 °C and quenched within 5 h caused no
formation of the side product. By contrast, acetylation, selective
hydrolysis (3,12-diacetate to 12-monoacetate), and oxidation
reactions of the analogous homoallylic alcohol 16â retaining a
[5,6] spiroketal evidenced no such side products.
Notwithstanding these sometimes surprising side paths,
conversion of hecogenin acetate 11 to the ∆14-furostene North
G 8 was accomplished in 15% yield over 13 steps, an
encouraging gain over syntheses of the highly oxygenated South
7 and North 1 subunits (30-33 steps, 1-3%, vide supra) as
well as that of South 1 (35 steps, 1%, vide infra), and
optimizations are likely to further improve the process. Sig-
nificantly, the key three-step Welzel sequence furnishing the
14â-OH of keto alcohol 24 in 86% overall yield from furo-
stanone 19 had proceeded comparably to that from spirostanone
11 to keto alcohol 18 (94%).
A more adventurous application of this sequence was
envisioned to introduce the ∆14 unsaturation present in the south
half of cephalostatin 1 (1). The saturated ketone 32 (prepared
rapidly and in good yield from the known advanced intermediate
29,12 Scheme 10) differs dramatically from ketones 11 and 19
in close proximity to the reacting center. The effects of the
altered ring strain and steric repulsions on its reactivity during
the photolytic opening and acid-catalyzed recyclization steps
were nonobvious. The â-alkoxy moiety should destabilize
formation of the tertiary alkyl radical,41 and the E/D ring fusion
Further differences in this series from experiences with related
steroids also deserve brief mention. Luche reduction40 (Scheme
9) of ∆14-furostenone 25 (6.5:1 â/R) suffered by comparison to
(36) (a) Only after we obtained the X-ray of 25b, which we expected to
be the 20â epimer of 25a, did we suspect this possible identity. The NMR
changes seen in going from 25a to 25b are similar to those visible for the
side product in 24. However, while 25a is calculated to lie only 0.83 kcal/
mol lower than 25b, 24 is calculated to lie 3.81 kcal/mol lower than (22R)-
epi-24, the Prins products 23R 3.96 kcal/mol and 23â 3.90 kcal/mol lower
than their respective 22R isomers, and 19 4.40 kcal/mol lower than (22R)-
epi-19. In the face of such large energy differences, it seems rather more
likely that any 22R epimer 25b is produced from 25a directly via Lewis
acid catalysis after elimination. It remains difficult at this point to assign
the identity of the side product in 24 with any certainty, and the possibility
that it is the source of the unidentified side product 25e remains viable. (b)
For compounds 26, 27, 28, 3,12-diol 8 and deacetylated 8, each 22S isomer
is calculated to lie 1.99-2.17 kcal/mol lower than its respective 22R isomer.
(c) All calculations performed using CAChe version 3.5. (d) Brown-Jones
oxidation produced even more of the impurity in 24, see ref 25. Swern
conditions gave less impure product but in lower yield.
(37) One referee questioned why we explored the Prins and elimination
reactions to give 18 rather than access it more directly via 16: to reiterate,
we had already explored the chemistry proceeding from 11 and had moved
on to an alternate approach via 19 (because substrates 17 and 18 proved
unstable to spiroketal opening) several months prior to the appearance of
ref 14b. The ene reaction of 22 was, however, attempted. These reactions
were performed on crude photolysate 22 by Mr. Zhiwei Tong. Some
adjustment of conditions was attempted in order to obtain yields more like
those available from crude 13 without success, but more fine-tuning of the
conditions may yet provide a synthetically useful reaction for this compound.
(38) X-ray data for compounds 24, 25b, and 25c have been submitted
to the Cambridge Crystallographic database.
(39) Kim, S. Ph.D. Thesis, Purdue University, 1994.
(40) Gemal, A. L.; Luche, J.-L. J. Am. Chem. Soc. 1981, 103, 5454.
(41) (a) Wagner, P. J.; Kemppainen, A. E. J. Am. Chem. Soc. 1972, 94,
7495. (b) Reference 21c.

700 J. Am. Chem. Soc., Vol. 120, No. 4, 1998
LaCour et al.
Scheme 10a
a (a) H2CrO4, Et2O, 25 °C, 15 min, 96%; (b) CSA, 1,2-DCE, 83 °C, 13 h, 87%; (c) 10% Pd/C (0.2 equiv), H2 (1 atm), AcOH, 25% EtOH/EtOAc,
24 h, 88%; (d) Ac2O, NEt3, DMAP CH2Cl2, 25 °C, 1 h, quantitative; (e) TBAF, THF, 80 °C, 2.5 h, 87%; (f) 300 nm, hν, dioxane, 1.5 h; (g) 75%
AcOH in H2O, 25 °C, 2.5 d; (h) H2CrO4, Et2O, 25 °C, 15 min.
of 32 could either hamper or aid collapse to desired secoalde-
hyde. Additionally, increased C ring strain would retard
recyclization as seen for 19 and other steroids.24 The greater
concern, however, derived from the possibility of a type II
abstraction of the 20â hydrogen, which lies, unavoidably, rigidly
held well within reach of the excited carbonyl oxygen (only
2.60 Å away) via a nearly optimal six-membered transition
state.21,42 Of the three paths the resulting 1,4-biradical might
take, reversion to ketone would be most facile due to the same
rigid nature of the steroid and would afford renewed opportunity
for the desired type I cleavage. Cyclobutanol formation should
likewise be excluded structurally. However, reorganization to
permit â-cleavage might not be so very difficult, as a 15° smaller
change in dihedral angle to permit overlap of developing
p-orbitals is needed in 32 than that required for 11 or 19, with
a calculated release of >9 kcal/mol by relief of the strain of
D/E fusion and formation of the enol/nonene array (the 13R
ketone lies a further 6 kcal/mol lower in energy).11b,21
In the event, the Welzel sequence was applied as before. A
solution of 32 in dioxane was irradiated at 300 nm for a mere
1.5 h, at which time TLC revealed complete consumption of
starting material. The crude photolysate was a complex mixture
which evidenced a major spot by TLC, a minor, less polar spot,
and a host of minor tailings. We note that oxetane 14 is more
polar than 13. NMR showed a smaller than expected aldehyde
peak (in one run, none at all in another run) as well as minor
(<10%) signals in the vinyl region, suggestive of the type II
product. Unfortunately, isolation of the putative ∆17(20) nonene
proved elusive both at this stage and at the completion of the
protocol, possibly because of enhanced spiroketal lability to
silica gel (the 9-ring oxonium ion would be conjugated to the
C17(20) double bond).42 The mixture required much longer
reaction times than crude 13 or 22 for the Prins reaction (at
least 60h stirring with 75% acetic acid, perhaps much longer
for consumption of all tailings) and afforded not two but a
mixture of several products. Subsequent oxidation gave, much
to our surprise and delight, an easily separable mixture of the
diketo olefin 34 and diketo alcohol 35 in overall yields (for
three steps from ketone 32) of 58% and 22%, respectively.
Chemical conversion therefore shows that type I cleavage had
predominated in the unprecedented photolysis of 32. That the
majority of the “Prins” product contained the desired unsatura-
tion (whether from spontaneous elimination of the kinetic Prins
product or its intermediate carbocation, as was implied in the
reaction of 23, or from Bronsted acid-catalyzed ene reaction)
is also unprecedented as well as serendipitous.30a,33 It is
instructive to note, as a consequence of substrate structure and
ring strain, the striking divergence over the course of this three-
step sequence of the proximally differentiated ketone 32 from
spirostanone 11 and furostanone 19, especially in terms of their
rates of photolysis (much faster for 32, some apparent type II
reaction) and their product distributions (formal ene reaction
versus Prins-reaction products) in the acid-catalyzed cyclization
of their secoaldehyde photolysates.
Azido Ketone and Aminomethoxime Coupling Partners.
Smooth elaboration of 34 to “South 1” azido ketone 7 occurred
by application of the bromination (PTAB) / azide substitution
(tetramethylguanidinium tribromide, TMGA) procedure estab-
lished in these laboratories to give bromide 36 and azido ketone
7 successively in high yield (Scheme 11).1 Azide substitution
worked better with freshly distilled nitromethane: older solvent
displayed an annoying tendency to generate â-nitro alcohol
(Henry reaction at C3)43 as a side product. Equatorial azide
was, as usual, the main product, presumably arising from
enolization following substitution. The advanced North 1
intermediate azidoketone 37 (78 × 95 ) 72% from the ketone)7
gave azidomethoxime 38 (99%), which was subjected to
Staudinger reduction44 to afford aminomethoxime 6 (80%).1
Some improvements in conversion were achieved with North
G ketone 8, which was brominated using a modified procedure
to furnish 39 (accompanied by ∼5% axial bromide) in quantita-
tive yield. Azide substitution operated very satisfactorily in
acetonitrile,45 giving azido ketone 9 (without Henry reaction
complications) in 90% yield over two steps. Azidomethoxime
40 and aminomethoxime 10 were obtained from 9 via the
protocol described above for the conversion of 37 to 6 (75%
for two steps). Thus, all four coupling partners derived quickly
and in good yield from their 3-ketosteroids although they contain
substantial dissimilarities in rings D-F.
Unsymmetric Coupling and Deprotection. For the key
coupling reaction leading to the first total synthesis of cepha-
(43) For recent advances in the addition of nitroalkanes to carbonyl
compounds, see: Simoni, D.; Invidiata, F. P.; Manfredini, S.; Ferroni, R.;
Lampronti, I.; Roberti, M.; Pollini, G. P. Tetrahedron Lett. 1997, 38, 2749
and references therein.
(44) Procedure taken from the following: Nagarajan, S.; Ganem, B. J.
Org. Chem. 1987, 52, 5044. For a review, see: Gololobov, Y. G.; Kasukhin,
L. F. Tetrahedron 1992, 48, 1353.
(45) The mixing order is critical: less than 3% of R-amino enone (not
shown) and 5-10% of the axial azide were produced when a cold solution
of bromide 39 was added to a cold solution of TMGA prior to warming,
but the reverse order produced ∼15% R-amino enone and >10% axial azide.
(42) See the expanded Scheme 10a and CAChe 3D drawings in the
Supporting Information.

Cephalostatin Synthesis. 11
J. Am. Chem. Soc., Vol. 120, No. 4, 1998 701
Scheme 11a
a (a) PTAB, THF, 25 °C, 6 min; (b) TMGA, CH3NO2, 25 °C, 2.5 h; (c) NH2OMe‚HCl, 10% pyr in CH2Cl2, 0 °C to 25 °C, 4 h; (d) PPh3, 3%
aqueous THF, 25 °C, 24 h; (e) PTAB, THF, 0 °C, 10 min; (f) TMGA, CH3CN, 0 °C to 25 °C, 6 h; (g) as (d), 48 h.
Scheme 12a
a (a) Bu2SnCl2 (10 mol %), PVP (100 wt %), benzene, 80-85 °C, 3 h (b) i. TBAF, THF, 83 °C, 2 h; ii. K2CO3, MeOH/H2O (8:1), 0.5 h.
lostatin 1 (1), the North 1 R-aminomethoxime 6 (1 equiv) was
heated with the south R-azido ketone 7 (1 equiv) in benzene in
the presence of 10 mol % dibutyltin dichloride (a catalyst for
imine formation)46 and 100 wt % of polyvinylpyridine (PVP),
with azeotropic removal of water for 3 h (Scheme 12).
Chromatography gave protected cephalostatin 1 (41) in 59%
yield (76% based on recovered 6). Cleavage of the C23 and
C26 silyl protecting groups (TBAF), followed by removal of
the C12 and C23′ acetate groups (K2CO3/aqueous MeOH), gave
the first synthetic sample of (+)-cephalostatin 1 (1) in 80% yield
yields observed with model couplings1 and in retrospect may
explain the lower yields obtained in the cephalostatin 1 and
dihydrocephalostatin 1 cases, for they also had evidenced similar
TLC profiles. Regarding the origin of the more polar pyrazine
product, suspicion fell on possible loss of the C12 acetate, which
had shown some lability in the synthesis of North G ketone 8
(vide supra, Scheme 9), or possibly the C12′ acetate; unprec-
edented formation of the cis-cross-coupled pyrazine seemed
unlikely. Deprotection of 42 (TBAF then KOH in a single pot)
gave ritterostatin GN1N (4) in 94% yield. Lacking unambiguous
NMR confirmation of the more polar pyrazine’s identity, the
material was separately deprotected (TBAF, then KOH) and
shown to give the desired ritterostatin. The combined yield of
ritterostatin GN1N (4) was thus 71% overall (82% based on
recovered 9) for the complete coupling/deprotection sequence.
Examination of the NMR spectra of the cephalostatins and
ritterazines shows that the chemical shifts and coupling patterns
of the steroidal subunits are essentially invariant, regardless of
1
from 41. The TLC, H and 13C NMR, and HPLC profiles of
synthetic and natural cephalostatin 1 (1) were found to be
identical.47
Similar conditions were employed in the coupling of North
G azido ketone 9 with North 1 aminomethoxime 6 to give
protected ritterostatin GN1N 42 as the main product in 49% yield
after chromatography. In addition, attention was paid to a more
polar pyrazine (31%) closely resembling 42; 23% of ami-
nomethoxime 6 was recovered unchanged (Scheme 13). The
total yield of isolated coupled pyrazines was thus 80% (92%
based on recovered 6). This is more in line with the type of
1
the partner to which they are coupled. The H NMR and 13C
NMR spectra of ritterostatin GN1N (4) matched those of the
North units of cephalostatin 1 (1) and ritterazine G (3),
confirming the identity and trans orientation of the pyrazine-
fused steroids.48
(46) Stein, C.; de Jeso, B.; Pommier, J. C. Synth. Commun. 1982, 12,
495.
(47) We thank Professor G. R. Pettit of Arizona State University for
kindly providing us with a sample of natural cephalostatin 1 (1) and Dr.
Douglas Lantrip for performing the HPLC comparisons of synthetic and
natural cephalostatin 1 (1).
In the reaction of North G aminomethoxime 10 with South
1 azido ketone 7 to give protected ritterostatin GN1S 43, Nafion
H was employed in lieu of PVP in order to test the stability of

702 J. Am. Chem. Soc., Vol. 120, No. 4, 1998
LaCour et al.
Scheme 13a
a (a) 10% Bu2SnCl2, 100 wt % PVP, 100 wt % 4A sieves, benzene, reflux; (b) (i) TBAF, THF, reflux; (ii) KOH, MeOH added, reflux.
Scheme 14a
a (a) 10% Bu2SnCl2, 100 wt % Nafion H, benzene, reflux; (b) K2CO3, aqueous MeOH, reflux.
these partners’ spiroketal moieties to heterogeneous Bronsted
acid catalysis (Scheme 14). Couplings of model steroids have
proceeded in somewhat higher yield for each partner set when
Nafion H was utilized instead of PVP.1 Such results might be
understood in terms of the release during the course of the
reaction of methoxylamine (see Scheme 1), which could
diminish the effectiveness of the tin Lewis acid when left
unneutralized, or capture azido ketone coupling partner as an
azidomethoxime. The chemically resistant, hindered ∆14 func-
tions were expected to survive unchanged. The natural 20R22R
configuration of North G (as the 3-ketone) is calculated to lie
2.0 kcal/mol below its energetically nearest (20R22â) epimer,
while the calculated energy of the natural 20R22R configuration
of South 1 (as the 23-acetate) lies 3.6 kcal/mol below its closest
(20R22â) epimer. For steroids which are thus expected to be
stable to heterogeneous Bronsted acid (NOT South 7, for
example, whose [5,6] spiroketal lies only 0.04 kcal below a
[5,5] isomer: its 7âOH relative South B has been shown to
isomerize extensively at 25 °C with HCl),4b,11 such a protocol
might prove advantageous in terms of yield in the coupling step.
In the event, North G aminomethoxime 10 (1.0 equiv) and
South 1 azido ketone 7 (1.1 equiv) reacted as expected, and no
evidence for further epimerization was found. Chromatography
afforded a 52% yield of 43; lower Rf pyrazines arising from
loss of acetate at C23′ (which had shown some lability, Scheme
10) and/or C12 amounted to another 21%. Unchanged ami-
nomethoxime 10 (26%) was recovered, for a total yield of 73%
(99% based on recovered 10) in the coupling step. The actual
yield of pyrazine was thus not substantially superior to that
obtained in comparable PVP-mediated reactions, although a
direct comparison for this case (keeping partners and equivalents
of each constant) between PVP and Nafion H was not made.
Saponification of the protected pyrazines afforded the hybrid
analogue ritterostatin GN1S (5) in 94% combined yield for the
deprotection, overall 69% (93% based on recovered 10) for the
1
two-step coupling/deprotection sequence. Again, the H and
13C NMR spectra of ritterostatin GN1S (5) closely matched those
of its constituent subunits, the North hemisphere of ritterazine
G and the South hemisphere of cephalostatin 1 (1).
Biological Activity. Testing of the analogues against natural
cephalostatin 1 (1) in the National Cancer Institute’s (NCI) in
vitro, human cancer cell panel revealed that ritterostatin GN1N
(4) displays exceptionally high potency. It is more potent than
1 for the leukemia K-562 line, equipotent for ovarian OVCAR-
8, renal SN12C, and breast MCF7/ADR-RES cell lines, and
nearly so for several others, including renal RXF-353 (1, GI50
0.1 nM; 4, 0.3 nM), with a mean GI50 < -7.4 (42 ( 7 nM, 60
of 60 cell lines affected) and an activity profile similar to that
of 1 (mean GI50 < -8.5; i.e., 3.5 ( 0.7 nM, 60 of 60 cell lines
affected).49 Ritterostatin GN1N (4) possesses mean tumor
inhibiting activity approaching that of taxol (mean GI50 -7.9;
i.e., 12.6 nM), superior to that of cephalostatin 7 (2) in all
categories of cancer cell lines tested, and superior to that of all
standard chemotherapeutics, including adriamycin (mean GI50
-6.9), cisplatin (mean GI50 -5.7), 5-fluorouracil (mean GI50
-4.7), and cyclophosphamide (mean GI50 -3.7).50
That ritterostatin GN1N (4) retains most of the activity of
cephalostatin 1 (1) represents a significant new advance.
Synthesis of its southern subunit 8 (North G) was accomplished
in only a third of the number of steps needed to construct the
south hemisphere of 1 (South 1), with a 1500% increase in yield
(49) Originally, cephalostatin 1 (1) was reported to display a mean GI50
< -8.9 (i.e., 1.2 nM, J. Nat. Prod. 1994, 57, 52), but in head-to-head testing
against dihydrocephalostatin a mean GI50 of -8.6 (i.e. 2.3 nM) was observed
(ref 1). During head-to-head testing against the ritterostatin GN1N (GI50
<
(48) See the Supporting Information for a comparison of the NMR data
of each ritterostatin with its constituent subunits as observed in the relevant
ritterazine or cephalostatin.
-7.4 [42 ( 7 nM] NSC D-699012) and GN1S (GI50 > -6.1 [900 nM]
NSC D-699013) analogues, a mean GI50 of -8.45 (3.5 ( 0.7 nM) was
observed.

Cephalostatin Synthesis. 11
J. Am. Chem. Soc., Vol. 120, No. 4, 1998 703
from hecogenin acetate 11. Ritterostatin GN1N (4) fulfills the
desired increase in synthetic accessibility with retention of high
potency which we hoped to find in such interphylal hybrids.
The North G unit is thus shown to be an adequate analogue of
South 1, and it is noteworthy that almost all of the most active
cephalostatins include South 1 or a close relative thereof. The
presence of a 12â-hydroxyl in North G probably contributes to
its activity relative to South 1, since compounds with 12-keto
or 12-acetoxy moieties are invariably less active than their direct
12â-hydroxyl counterparts.51 The 10-fold decrease in activity
in going from 1 to 4 may be more a reflection of the loss of the
23-hydroxyl (polarity loss)4b than either the change from an 18,-
22-epoxy linkage to the 16,22-epoxide (a major spatial transla-
tion of the E/F rings) or the change from a [6,5] to [5,5]
spiroketal arrangement. A short route to such a 23-hydroxy
North G relative is currently being explored in order to test the
theory. It will also be interesting to discover the effect of the
[5,6] spiroketal arrangement present in hecogenin itself and its
14,15-dehydro relative, whose C rings and spiro connections
have less strain energy to spend. Shorter routes to North 1 and
analogues thereof to function as the North partner, especially
those featuring a 17R-hydroxyl group, are also of prime interest
and under current inquiry.
Ritterostatin GN1S (5), by contrast, was not expected to show
high activity because it lacks a 17R-hydroxyl group, a feature
present in at least one hemisphere of all of the most active
ritterazines and cephalostatins. Ritterostatin GN1S (5) was
significantly weaker than 4, with a mean GI50 > -6.1, affecting
only 10 of 60 cell lines at 900 nM concentration. Compounds
which lack this moiety are dramatically less active than their
direct 17-hydroxylated counterparts,52 and the importance of this
group to tumor inhibition has also been documented in angio-
static steroids such as 17R-hydroxyprogesterone and ana-
logues.53 Furthermore, the spiroketals of both South 1 and North
G are calculated to lie at the bottom of their respective isomeric
energy wells by at least 2 kcal/mol. The most actiVe pyrazines
all contain at least one spiroketal which calculations (and in
some cases, experimental eVidence)1,7,11,29 show is a less stable
isomer. Since there are indications that related steroids such
as OSW-1 (which may be considered structurally analogous to
a spiroketal, and the activity of which correlates well with that
of the cephalostatins)54 and solasodine act biologically in part
by reaction at the spiro center,55 the concept of a “cocked gun”
may appropriately aid our understanding. North 1 (20R,22â),
for example, is calculated to lie ∼1 kcal/mol above the most
stable (20â,22â) [5,5] spiroketal and more than 6 kcal/mol aboVe
the most stable isomer, the (20R,22R) [5,6] spiroketal. Interest-
ingly, the latter happens to be the same isomeric arrangement
(with an axial 23-hydroxyl instead of hydrogen) found in the
South 7 steroid.11
On the other hand, if North G actually functioned as a very
close mimic of South 1, not only would 4 be more active but
the union 5 of these subunits would be expected to show low
activity as do “symmetric” trisdecacyclic pyrazines (ritterazines
K, N, and R, IC50’s of 10, 460, and 2100 ng/mL, cephalostatin
12 and Winterfeldt’s symmetric analogues, all with mean GI50’s
> -4.3). That is, it should display greatly diminished scope
and level of activity, perhaps due to the lack of a polar/apolar
pair of subunits as proposed by Fusetani.4b However, its activity
level remains higher than expected (although its scope is
diminished), likely due to the contribution of the 12âOH/∆14
array in North G, and perhaps some polar/apolar pair character
as a result of the difference between the steroid subunits
imparted by the 23-hydroxyl present in South 1 and absent in
North G.
Conclusion
We have successfully completed the first total synthesis of
cephalostatin 1 (1), which remains among the most potent
antineoplastic agents ever tested by the NCI. The lengthy
sequence required to produce 1 has also been somewhat obviated
by the synthesis of ritterostatin GN1N (4), which retains very
high activity in the tumor panel but which is much more
synthetically accessible. North G is shown to be a useful
analogue of South 1, an important step toward the practical
synthesis of extremely potent anticancer agents. These syn-
theses were made possible by logical modification and extension
of a key photolysis/Prins sequence which was not known to be
as general nor nearly as high yielding as has now been shown.
New chemical insights and important new SAR data have been
provided by these syntheses. The concept of interphylal product
splicing is thus justified. Fruitful directions for future inquiry
are indicated and are actively being pursued.
Experimental Section
General Methods. Unless otherwise stated, reactions were carried
out under a positive argon atmosphere in flame-dried glassware using
magnetic stirring. Diethyl ether (Et2O) and tetrahydrofuran (THF) were
distilled from sodium benzophenone ketyl. Dichloromethane (CH2-
Cl2), benzene, and toluene were distilled from calcium hydride.
Deuterated NMR solvents (CDCl3, C6D6, and 99.5% C5D5N) were
stored over 4 Å molecular sieves for several days prior to use, with
the exception of 99.99% C5D5N (used for <3 mg samples), which was
used directly from ampules. All other chemicals were used as supplied
from commercial sources unless otherwise specified. TLC plates were
developed with anisaldehyde solution. Flash chromatography (hereafter
sgc) was carried out as described by Still56 (230-400 mesh silica gel).
Melting points are uncorrected. 1H and 13C NMR spectra were obtained
at 300 or 500 MHz and 75 or 125 MHz, respectively. 1H NMR
chemical shifts are reported in ppm relative to the residual protonated
solvent resonance: CHCl3, δ 7.26; C6D5H, δ 7.15; C5HD4N, δ 8.71.
Splitting patterns are designated as s, singlet; d, doublet; t, triplet; q,
quartet; p, pentet; m, multiplet; br, broadened; ap, apparent. Coupling
constants (J) are reported in hertz. 13C NMR chemical shifts are
reported in ppm relative to solvent resonance: CDCl3, δ 77.00; C6D6,
δ 128.00; C5D5N, δ 135.50. Mass spectral data include the molecular
ion designated as M.
(50) Detailed testing information for all standard therapeutic agents
(∼150) is available on the World Wide Web: http://epnws1.ncicrf.gov:
2345/dis3d/itb/stdagnt/tab.html. Useful NSC numbers: cyclophosphamide
(NSC 26271), 5-fluorouracil (NSC 19893), cisplatin (NSC 119875),
adriamycin (NSC 123127), tamoxifen (NSC 180973), and paclitaxel (NSC
125973). Our compliments to Professor Winterfeldt for disseminating this
site in his report (ref 10) as well.
(51) Compare, for example, ritterazines B (12âOH, ED50 0.15 ng/mL
against P388 leukemia), H (12-keto, ED50 16 ng/mL, a 100-fold decrease
in activity), and B-12-acetate (ED50 3.5 ng/mL, a 23-fold decrease in
activity). See ref 4b.
(52) Compare, for example, ritterazines A (17′-OH, IC50 3.5 ng/mL) and
T (17′-H, IC50 460 ng/mL, a 330-fold decrease), or B (0.15 ng/mL) and Y
(3.5 ng/mL).
(53) Schweiger, E. J.; Joullie, M. M.; Weisz, P. B. Tetrahedron Lett.
1997, 38, 6127 and references therein.
Furostanone 19. Alkene 21 from combined runs (0.890 g, 1.86
mmol, contained 5% hecogenin acetate) was suspended in 75% acetic
acid and heated at 90-95 °C during 40 h of stirring, when TLC analysis
(3:1 hexane/EtOAc) of a worked up aliquot showed no remaining 21.
(In fact, some starting material did remain as the desacetate, but its
high polarity obscured its identity, being mistaken for expected
(54) Mimaki, Y.; Kuroda, M.; Kameyama, A.; Sashida, Y.; Hirano, T.;
Oka, K.; Maekawa, R.; Wada, T.; Sugita, K.; Beutler, J. A. Bioorg. Med.
Chem. Lett. 1997, 7, 633.
(55) Kim, Y. C.; Che, Q.-M.; Gunatilaka, A. A. L.; Kingston, D. G. I.
J. Nat. Prod. 1996, 59, 283.
(56) Still, W. C.; Kahn, M.; Mitra, A. J. Org. Chem. 1978, 43, 923.

704 J. Am. Chem. Soc., Vol. 120, No. 4, 1998
LaCour et al.
deacetylated spiroketal 19.) The resulting solution was partitioned
between EtOAc and water, and the organic layer was washed with
saturated bicarbonate solution. The original aqueous layer (acidic) was
back-extracted with EtOAc, the organic layer was washed with saturated
bicarbonate, and the combined organic layers were dried over Na2SO4
and concentrated to give 0.90 g of pale yellow solids. The solids were
acetylated in the usual way (CH2Cl2, TEA, Ac2O, catalyst DMAP, 0
°C, 2 h) and worked up to give 0.91 g of pale orange solids. Sgc
(40:1 CH2Cl2/THF) afforded 0.212 g of recovered alkene 21 (25%)
and 0.646 g of furastanone 19 (73%, 98% borsm) as white solids: mp
230-232 °C (MeOH); 1H NMR (300 MHz, CDCl3) δ 4.73-4.60 (1H,
m, H3R), 4.43-4.35 (1H, m, H16R), 2.51 (1H, dd, J ) 8.7, 6.9 Hz, H17R),
2.39 (1H, apt, J ) 13.9 Hz, H11â), 2.21 (1H, dd, J ) 14.2, 5.1 Hz,
) -12.8° (c ) 1.0, CHCl3); MS (EI) m/z 472 (M - H2O, 56); (CI)
473 (M + H - H2O, 84), 455 (100); HRMS (CI) calcd for C29H45O5
(M + H - H2O) 473.3267, found 473.3252.
Diol 23â: 1H NMR (300 MHz, CDCl3) δ 4.70-4.59 (1H, m), 4.55
(1H, apt, J ) 6.8 Hz), 3.09 (1H, dd, J ) 11.8, 4.1 Hz, H12R), 2.48 (1H,
apt, J ) 7.7 Hz, H17R), 2.37-2.27 (2H, m), 2.00 (3H, s), 1.33 (3H, s),
1.16 (3H, s), 0.97 (1H, d, J ) 6.4 Hz), 0.97 (3H, s, H18), 0.80 (3H, s);
13C NMR (75 MHz, CDCl3): δ 170.8 (s), 116.8 (s), 86.7 (s), 82.5 (s),
79.5 (d), 74.6 (d), 73.5 (d), 57.0 (d), 53.2 (s), 46.3 (d), 44.6 (d), 42.8
(d), 39.3 (d), 38.8 (t), 37.2 (t), 36.9 (t), 35.8 (s), 33.9 (t), 33.6 (t), 30.2
(q), 29.4 (t), 28.5 (t), 28.5 (q), 27.7 (t), 27.4 (t), 21.5 (q), 14.8 (q), 12.2
(q), 7.6 (q); MS (EI) m/z 490 (M, 36); (CI) 491 (M + H, 84).
Keto Alcohol 24 (North I as 3-Acetate). Jones oxidation (see
Supporting Information) of 23R (0.245 g, 0.499 mmol) afforded keto
alcohol 24 (0.237 g, 97%) as white solids, mp 231.5-232 °C (CH2-
Cl2), which by NMR contained ∼8% of an impurity (tentatively
identified as the 22R epimer). Oxidation of crude diols 23 proceeded
in similar yield: 1H NMR (300 MHz, CDCl3) δ 4.71-4.60 (1H, m),
4.46 (1H, apt, J ) 6.6 Hz, H16R), 3.32 (1H, apt, J ) 7.8 Hz, H17R),
2.78 (1H, brs, HOH), 2.45 (1H, apt, J ) 13.5 Hz, H11â), 2.31-2.11
(3H, m), 2.00 (3H, s), 1.33 (3H, s), 1.23 (3H, s, H18), 1.16 (3H, s),
H11R), 2.09 (1H, m), 2.00 (3H, s, HAc), 1.32 (3H, s, H27), 1.16 (3H, s,
H26), 1.05 (1H, d, J ) 6.4 Hz, H21), 1.03 (3H, s, H18), 0.91 (3H, s,
H19); 1H NMR (300 MHz, C6D6) and 1H NMR (300 MHz, pyr-d5) see
Supporting Information; 13C NMR (75 MHz, CDCl3) δ 214.0 (s), 171.0
(s), 120.1 (s), 82.6 (s), 79.1 (d), 73.4 (d), 55.6 (d), 55.4 (d), 55.3 (s),
53.3 (d), 44.5 (d), 39.0 (d), 37.8 (t), 37.1 (t), 36.3 (t), 36.2 (s), 34.4
(d), 33.8 (t), 31.5 (t), 31.2 (t), 30.2 (q), 28.5 (q), 28.2 (t), 27.3 (t), 21.5
(q), 16.0 (q), 13.6 (q), 11.9 (q); 13C NMR (75 MHz, C6D6) see
Supporting Information; MS m/z 472 (M+, 3), 139 (100); HRMS calcd
for C29H44O5 472.3189, found 472.3175.
1
0.96 (1H, d, J ) 6.7 Hz), 0.88 (3H, s); H NMR (300 MHz, pyr-d5)4
δ 5.46 (1H, brs, HOH), 4.86-4.74 (1H, m, H3R), 3.74 (1H, apt, J ) 6.6
Hz, H16R), 2.95-2.86 (1H, m, H17R), 2.55 (1H, apt, J ) 13.4 Hz, H11â),
2.38-2.30 (2H, m), 2.05 (3H, s, HAc), 1.48 (3H, s, H27), 1.44 (3H, s,
H18), 1.20 (3H, s, H26), 1.10 (1H, d, J ) 6.7 Hz, H21), 0.76 (3H, s,
H19); 13C NMR (75 MHz, CDCl3) δ 213.6 (s), 170.7 (s), 116.7 (s),
87.6 (s), 82.7 (s), 78.8 (d), 73.2 (d), 62.4 (s), 51.3 (d), 47.0 (d), 44.6
(d), 42.5 (d), 39.0 (d), 37.6 (t), 37.2 (t), 36.6 (t), 36.3 (s), 33.9 (t), 33.8
(t), 30.2 (q), 28.5 (q), 28.3 (t), 27.9 (t), 27.3 (t), 21.5 (q), 14.9 (q), 14.8
(q), 12.0 (q); [R] ) -25.9° (c ) 0.99, CHCl3); MS (CI) m/z 489 (M
+ H, 91), 471 (100); HRMS (CI) calcd for C29H45O6 (M + H) 489.3216,
found 489.3206; single-crystal X-ray.
Keto olefin 25a. Keto alcohol 24 (0.2164 g, 0.443 mmol) was
dissolved in toluene (3.4 mL). Pyridine (0.036 mL, 0.487 mmol, 10
equiv) was added, followed by dropwise addition of a solution of thionyl
chloride (0.030 mL, 0.487 mmol, 1.1 equiv) in toluene (1.0 mL) over
1 min. After 5 min, and after 2 h, TLC showed ∼40% 24 remaining.
Additional SOCl2 (0.025 mL, 0.399 mmol, 0.9 equiv) was added neat;
after 5 min, TLC showed no 24 remaining. The reaction was quenched
by the addition of ice (5 g) and partitioned between toluene (15 mL)
and 0.003 M HCl (27 mL), and the aqueous layer was extracted with
toluene. The organic layers were combined, washed with brine, and
dried over sodium sulfate, and the solvent was removed in vacuo. The
residue was fractionated repeatedly by sgc (gradient 50:1 to 10:1 CH2-
Cl2/THF) to afford (in order of elution) 14R-chloride 25c (6.7 mg, 3%),
(22R)-keto olefin 25b (16.5 mg, 8%), desired keto olefin 25a (130 mg,
63%), 14â-chloride 25d (29 mg, 13%), and 23 mg (11%) of an
unidentified product 25e.
Secoaldehyde 22: Ketone 19 (0.630 g, 1.33 mmol) in dioxane (70
mL) was deoxygenated with argon, and the solution was irradiated with
300 nm light (Rayonet apparatus) for 4 h, at which time TLC analysis
(3:1 hexane/EtOAc) showed only a trace of 19 and the appearance of
a major new spot of lower Rf accompanied by many faint tailing spots.
The solvent was removed in vacuo to afford 0.75 g (119%, some solvent
left) of photolysate as a white foam, 36 mg of which was purified by
sgc to give an analytical sample: 1H NMR (300 MHz, CDCl3) δ 9.47
(<1H, s, H12), 4.73-4.60 (1H, m, H3R), 4.55-4.51 (1H, m, H16R), 2.70-
2.64 (1H, apt, J ) 7.3 Hz, H17R), 2.45-2.05 (4H, m), 2.02 (3H, s,
H
Ac), 1.61 (3H, s, H18), 1.35 (3H, s, H27), 1.16 (3H, s, H26), 1.06 (1H,
d, J ) 6.9 Hz, H21), 0.84 (3H, s, H19); 13C NMR (75 MHz, CDCl3): δ
201.4 (s), 170.7 (s), 135.5 (s), 135.3 (s), 117.0 (s), 81.9 (s), 77.4 (d),
73.2 (d), 61.3 (d), 46.3 (d), 45.3 (d), 43.8 (d), 43.4 (t), 37.9 (t), 37.8
(d), 37.4 (t), 37.3 (t), 36.5 (s), 34.0 (t), 33.6 (t), 30.9 (t), 30.2 (q), 28.6
(q), 28.3 (t), 21.5 (q), 14.6 (q), 12.8 (q), 12.2 (q); MS (CI, isobutane)
m/z (%) 473 (100, M + H), 471 (61); HRMS calcd for C29H45O5 (M
+ H) 473.3267, found 473.3257.
Diols 23: Crude photolysate containing mainly secoaldehyde 22
(0.714 g, 1.27 mmol maximum) was suspended in 75% AcOH (6 mL)
and stirred for 35 h. The resulting solution was partitioned between
EtOAc (30 mL) and water (90 mL), the aqueous layer was extracted
with EtOAc (30 mL), and the combined organic layers were washed
with water and then saturated bicarbonate and dried over sodium sulfate.
Removal of solvent in vacuo afforded diols 23 (0.628 g, 101%) as
white solids. 1H NMR analysis indicated a 4.5:1 ratio of 23R/23â
(∼90%) accompanied by traces of ∆14-containing material, keto alcohol
24 (from dissolved O2; degassed aqueous AcOH produced none of this
material) and unphotolyzed furastanone 19. Normally, this crude
material was carried on without purification; in this instance, for
characterization, sgc (gradient from 3:1 to 1:3 hexane/EtOAc) gave
diols 23R (0.400 g, 64%) and 23â (0.077 g, 13%). In addition, there
was obtained 24 mg (4%) of ketone 19; 15 mg (2%) of keto alcohol
24; 72 mg of material containing 23R (59%, 7% of theory), 23â (27%,
3% of theory), and what appeared to be the formal ene adduct (12ROH-
∆14, 14%, 2% of theory); 19 mg (3%) of presumed unreacted
photolysate side products, and 18 mg (3%) of mixed triols (3â,12R/
â,14â-trihydroxy epimers).
Diol 23r: 1H NMR (300 MHz, CDCl3) δ 4.71-4.61 (1H, m), 4.55
(1H, apt, J ) 6.8 Hz, H16R), 3.62 (1H, brs, H12â), 2.44-2.29 (3H, m),
2.00 (3H, s), 1.35 (3H, s, H27), 1.16 (3H, s, H26), 1.02 (3H, s, H18),
0.95 (1H, d, J ) 6.4 Hz, H21), 0.80 (3H, s, H19); 13C NMR (75 MHz,
CDCl3) δ 171.1 (s), 116.8 (s), 86.2 (s), 82.3 (s), 81.2 (d), 76.1 (d),
73.7 (d), 57.9 (d), 50.9 (s), 44.4 (d), 43.0 (d), 42.8 (d), 40.8 (t), 40.1
(d), 37.3 (t), 36.8 (t), 35.4 (s), 34.0 (t), 33.6 (t), 30.2 (q), 28.7 (q), 28.6
(t), 28.5 (t), 27.4 (t), 27.3 (t), 21.6 (q), 15.4 (q), 15.1 (q), 12.2 (q); [R]
25a: mp 189-191 °C (CH2Cl2/hexane); 1H NMR (300 MHz, CDCl3)
δ 5.40 (1H, brs, H15), 4.81 (1H, apd, J ) 8.0 Hz, H16R), 4.70-4.61
(1H, m), 3.32 (1H, apt, J ) 8.6 Hz), 2.54 (1H, apt, J ) 14.2 Hz, H11â),
2.49-2.38 (1H, m), 2.33 (1H, dd, J ) 14.5, 4.4 Hz, H11R), 2.01 (3H,
s, HAc), 1.99 (2H, aps), 1.35 (3H, s), 1.28 (3H, s, H18), 1.18 (3H, s),
1.02 (1H, d, J ) 6.6 Hz), 0.93 (3H, s); 13C NMR (75 MHz, CDCl3):
δ 211.3 (s), 170.7 (s), 154.6 (s), 121.3 (d), 117.8 (s), 83.7 (d), 82.1 (s),
73.1 (d), 62.4 (s), 53.5 (d), 49.9 (d), 44.1 (d), 41.0 (d), 37.4 (t), 37.3
(t), 36.4 (s), 36.3 (t), 34.2 (d), 33.8 (t), 33.5 (t), 30.1 (q), 29.5 (t), 28.6
(q), 28.0 (t), 27.3 (t), 21.5 (q), 21.2 (q), 14.1 (q), 11.8 (q); COSY NMR
1
(300 MHz, CDCl3), H NMR (300 MHz, pyr-d5), COSY and CSCM
NMR (300 MHz, pyr-d5), 13C NMR (75 MHz, pyr-d5), see Supporting
Information; [R] ) +71.1° (c ) 1.02, CHCl3); MS (CI) m/z 471 (M +
H, 100); HRMS (CI) calcd for C29H43O5 (M + H) 471.3111, found
471.3117.
25b: 1H NMR (300 MHz, CDCl3) δ 5.39 (1H, brs), 4.72-4.61 (1H,
m), 4.69 (1H, ap brd, J ) 7.0 Hz, H16R), 3.18 (1H, apt, J ) 7.6 Hz,
H17R), 2.56 (1H, apt, J ) 13.9 Hz), 2.52-2.42 (1H, m), 2.31 (1H, dd,
J ) 14.5, 4.7 Hz), 2.11 (1H, dd, J ) 13, 6.2 Hz, H20â), 2.02 (3H, s),
1.38 (3H, s), 1.32 (3H, s, H18), 1.15 (3H, s), 1.04 (1H, d, J ) 7.0 Hz),
0.94 (3H, s); COSY NMR (300 MHz, CDCl3) see Supporting
Information; MS (CI) m/z 471 (M + H, 100); single-crystal X-ray.

Cephalostatin Synthesis. 11
J. Am. Chem. Soc., Vol. 120, No. 4, 1998 705
1
25c: H NMR (300 MHz, CDCl3) δ 4.75-4.65 (2H, m, H3R &16R),
aqueous acetic acid (3 mL) and stirred at 25 °C for 2.5 days. The
reaction mixture was diluted with EtOAc, washed with water (4-5
times) and saturated NaHCO3, and dried over sodium sulfate. The
solvent was evaporated and the residue dissolved in ether. Brown-
Jones oxidation (see Supporting Information) followed by sgc (25%
EtOAc in hexane) gave 75 mg (58% for three steps from 32) of diketo
olefin 34 and 28 mg (22%) of the diketo alcohol 35 as white foams.
Diketo Olefin 34. 1H NMR (300 MHz, CDCl3): δ 5.43 (1H, br s,
H15), 5.14 (1H, dd, J ) 6.6, 4.5 Hz, H23), 3.83 (1H, d, J ) 11.7 Hz,
H18), 3.76 (1H, d, J ) 12 Hz, H18), 2.07 (3H, s), 1.28 (3H, s), 1.26
(3H, s), 1.05 (3H, d), 1.04 (3H, s); 13C NMR (APT, 75 MHz, CDCl3)
[EVEN (17)] δ 211.7, 210.9, 169.9, 148.1, 109.3, 82.3, 63.6, 61.4, 44.5,
44.0, 38.8, 37.9, 36.2, 31.9, 29.3, 28.2; [ODD (12)] δ 123.4, 82.8, 51.9,
45.7, 43.4, 35.7, 32.3, 29.6, 29.5, 21.4, 14.9, 10.9; [R] ) +23.4° (c )
1.2, CHCl3); MS (CI) 485 (M + H, 100); HRMS (CI) calcd for
C29H40O6 485.2903, found 485.2912.
3.15 (1H, dd, J ) 8.0, 7.4 Hz, H17R), 2.54 (1H, dd, J ) 13.7, 7.1 Hz),
2.38-2.30 (3H, m), 2.01 (3H, s), 1.35 (3H, s), 1.19 (3H, s, H18), 1.17
(3H, s), 1.09 (1H, d, J ) 6.9 Hz), 0.98 (3H, s); COSY NMR (300
MHz, CDCl3) see Supporting Information; 13C NMR (75 MHz, CDCl3)
δ 209.4, 170.5, 120.0, 96.3, 82.5, 79.4, 73.0, 61.1, 52.6, 49.1, 44.2,
41.6, 39.6, 39.2, 37.0, 36.5, 36.1, 33.9, 33.7, 30.2, 29.7, 28.4, 27.8,
27.5, 27.1, 20.6, 13.5, 12.0; MS (CI) m/z 507/509 (M + H, 3.4/2.2);
HRMS (CI) calcd for C29H44ClO5 (M + H) 507.2877, found 507.2856;
single-crystal X-ray.
25d: 1H NMR (300 MHz, CDCl3) δ 4.72-4.61 (1H, m), 4.47 (1H,
apt, J ) 7.5 Hz), 3.46 (1H, apt, J ) 8.7 Hz, H17R), 2.66 (1H, apdq, J
) 9.2, 6.8 Hz), 2.49 (1H, apt, J ) 13.3 Hz, H11â), 2.41 (1H, apdq, J )
13, 3.1 Hz), 2.32 (1H, dd, J ) 13.1, 3.9 Hz, H11R), 2.23 (1H, d, J )
16.3 Hz), 2.20 (1H, dt, J ) 11.9, 3.3 Hz), 2.01 (3H, s), 1.99 (2H, aps),
1.37 (3H, s, H18), 1.35 (3H, s), 1.17 (3H, s), 0.96 (1H, d, J ) 6.8 Hz),
0.89 (3H, s); 13C NMR (75 MHz, CDCl3) δ 211.1 (s), 170.6 (s), 117.4
(s), 88.5 (s), 82.0 (s), 77.8 (d), 73.0 (d), 63.6 (s), 52.9 (d), 48.4 (d),
44.2 (d), 42.1 (d), 37.3 (t), 37.2 (t), 36.5 (t), 36.5 (s), 33.8 (t), 33.5 (t),
30.3 (q), 28.6 (q), 28.6 (t), 28.2 (t), 27.2 (t), 21.5 (q), 18.0 (q), 14.9
(q), 11.9 (q); MS (CI) m/z 507/509 (M + H, 16/6); HRMS (CI) calcd
for C29H44ClO5 (M + H) 507.2877, found 507.2867.
Mono-ol 28. To a solution of diacetate 27 (101.6 mg, 0.197 mmol)
in i-PrOH (4 mL) cooled to -15 °C was added t-BuOK (217 µL of a
1.0 M solution in t-BuOH, 0.217 mmol, 1.1 equiv). The solution was
warmed to 0 °C, stirred for 8 h, and then placed in the freezer (-15
°C) for 14 h. TLC analysis (1:1 hexane/EtOAc) showed substantial
27 (Rf 0.60) remaining, with the appearance of a major new spot (desired
mono-ol 28, Rf 0.33), a minor spot (diol, Rf 0.16), and a trace spot
(mono-ol 26, Rf 0.43). The reaction was warmed to 0 °C and stirring
continued for 10 h, at which time TLC showed only a trace of 27
remaining. The solution was partitioned between ether and 0.5% HCl,
and the aqueous layer was extracted with ether. The combined organics
were washed with bicarbonate, dried (Na2SO4), and concentrated, and
sgc (1:1 hexane/EtOAc) afforded desired mono-ol 28 (63.0 mg, 68%)
and diol (17.8 mg, 21%). The earlier fractions containing 27 and mono-
ol 26 were combined with the diol and reacetylated to give 32.2 mg of
diacetate 27, which was resubjected to the above hydrolysis procedure
(1.3 mL of i-PrOH, 69 µL of t-BuOK solution) to afford additional 28
(19.5 mg, 66% for this recycle). Reacetylation and hydrolysis of the
remaining material afforded another 6.8 mg of desired mono-ol, bringing
the total production of 28 to 89.3 mg (96%): 1H NMR (300 MHz,
CDCl3) δ 5.40 (1H, apt, J ) 1.8 Hz), 4.92 (1H, dd, J ) 8.2, 1.8 Hz),
4.35 (1H, dd, J ) 11.3, 4.7 Hz, H12R), 3.64-3.53 (1H, m, H3R), 2.34
(1H, apt, J ) 8.8 Hz), 2.04 (3H, s), 1.99 (2H, aps), 1.36 (3H, s), 1.18
(3H, s), 1.08 (3H, s, H18), 0.98 (1H, d, J ) 6.7 Hz), 0.86 (3H, s);
COSY and CSCM NMR (300 MHz, CDCl3) see Supporting Informa-
tion; 13C NMR (75 MHz, CDCl3): δ 170.6 (s), 156.6 (s), 120.2 (d),
117.4 (s), 84.3 (d), 82.1 (s), 81.1 (d), 71.0 (d), 56.1 (d), 52.3 (d), 51.3
(s), 44.5 (d), 41.2 (d), 37.9 (t), 37.3 (t), 36.8 (t), 36.0 (s), 34.1 (d), 33.3
(t), 31.3 (t), 30.1 (q), 29.6 (t), 28.5 (q), 28.3 (t), 26.5 (t), 21.3 (q), 15.0
(q), 14.0 (q), 12.2 (q); MS (CI, isobutane) m/z 473 (M + H, 45), 455
(64), 413 (100); HRMS (CI, isobutane) calcd for C29H45O5 (M + H)
473.3267, found 473.3281.
Ketone 8. Jones oxidation (see Supporting Information) of alcohol
28 (50.3 mg, 0.106 mmol) afforded ketone 8 (45.1 mg, 90%) as waxy
white solids: 1H NMR (300 MHz, CDCl3) δ 5.43 (1H, apt, J ) 1.8
Hz), 4.92 (1H, dd, J ) 8.3, 1.8 Hz), 4.37 (1H, dd, J ) 11.3, 4.7 Hz),
2.39-2.22 (4H, m), 2.16-2.05 (3H, m), 2.05 (3H, s), 2.00 (2H, s),
1.37 (3H, s), 1.18 (3H, s), 1.11 (3H, s), 1.06 (3H, s, H19), 0.99 (1H, d,
J ) 6.7 Hz); CSCM NMR (300 MHz, CDCl3) see Supporting
Information; 13C NMR (75 MHz, CDCl3) δ 211.2 (s), 170.6 (s), 155.9
(s), 120.8 (d), 117.5 (s), 84.2 (d), 82.2 (s), 80.8 (d), 56.1 (d), 51.8 (d),
51.4 (s), 46.2 (d), 44.5 (t), 41.2 (d), 38.2 (t), 38.0 (t), 37.3 (t), 36.1 (s),
34.0 (d), 33.3 (t), 30.1 (q), 29.3 (t), 28.6 (q), 28.5 (t), 26.7 (t), 21.3 (q),
15.0 (q), 14.0 (q), 11.4 (q); MS (CI) m/z 471 (M + H, 100); HRMS
(CI) calcd for C29H43O5 (M + H) 471.3111, found 471.3134.
Diketo Olefin 34 and Diketo Alcohol 35. A solution of 32 (130
mg, 0.27 mmol) in dioxane (12 mL) was irradiated as for 19 for 1.5 h
at which time TLC (40% EtOAc/hexane, 2×) showed no 32. The
dioxane was evaporated and the crude photolysate dissolved in 75%
1
Diketo Alcohol 35: H NMR (300 MHz, CDCl3) δ 5.15 (1H, apd,
J ) 6 Hz), 4.77 (1H, br s, OH), 4.04 (1H, d, J ) 11.7 Hz, H18), 3.38
(1H, d, J ) 12 Hz, H18), 2.06 (3H, s), 1.31 (3H, s), 1.27 (3H, s), 1.10
(3H, d, J ) 6.6 Hz), 0.99 (3H, s); 13C NMR (APT, 75 MHz, CDCl3)
[EVEN (18)] δ 211.9, 210.9, 170.1, 110.7, 87.5, 83.9, 64.0, 61.9, 44.5,
42.6, 41.0, 38.2, 37.8, 36.3, 34.6, 28.5, 27.7, 25.0; [ODD (11)] δ 81.0,
45.9, 44.8, 43.8, 43.0, 31.5, 30.9, 29.3, 21.4, 14.5, 11.2; MS (CI) 503
(M + H, 100); HRMS (CI) calcd for C29H42O7 503.3009, found
503.3024.
South 1 Azido Ketone 7. To a solution of 36 (29 mg, 0.05 mmol)
in freshly distilled nitromethane (5.2 mL) was added tetramethylguani-
dinium azide (TMGA, 33 mg, 0.21 mmol). The mixture was stirred
1
at 25 °C for 2.5 h, when a H NMR of an aliquot of showed no 36.
After removal of solvent, sgc (15% EtOAc/hexane) gave 24 mg (87%)
of 7 as a foam: 1H NMR (300 MHz, CDCl3) δ 5.44 (1H, s), 5.16 (1H,
dd, J ) 6.3, 4.8 Hz), 3.98 (1H, dd, J ) 12.9, 6.9 Hz, H2â), 3.83 (1H,
d, J ) 12 Hz), 3.75 (1H, d, J ) 12 Hz), 2.08 (3H, s), 1.29 (3H, s),
1.27 (3H, s), 1.12 (3H, s), 1.06 (3H, d, J ) 6 Hz); 13C NMR (APT, 75
MHz, CDCl3) [EVEN (16)] δ 210.9, 204.4, 169.9, 147.4, 109.3, 82.3,
63.6, 61.4, 44.9, 44.0, 43.5, 38.7, 37.3, 32.0, 29.2, 27.7; [ODD (13)] δ
123.8, 82.8, 63.5, 51.5, 46.6, 44.4, 35.2, 32.2, 29.6, 29.5, 21.4, 14.8,
11.9; MS (FAB, DTT/DTE) 526 (M + H, weak), 500 (M - N2 + H).
Bromo Ketone 39. A solution of ketone 8 (34.8 mg, 73.9 mmol)
in THF (0.75 mL) was cooled to 0 °C; PTAB (30.6 mg, 81.3 mmol,
1.1 equiv) in THF (1.5 mL) was cooled and then added to the solution
of 8 rapidly via cannula. The resulting orange solution deposited
copious precipitate and faded to a beige color within 2 min. After 10
min of stirring at 0 °C, the reaction was quenched with brine (10 mL)
and extracted twice with ether. The combined organics were washed
with brine, dried (Na2SO4), concentrated, and filtered through a plug
of silica gel (2:1 hexane/EtOAc) to afford bromo ketone 39 (41 mg,
quantitative) as off-white solids which by NMR contained ∼5% of the
3S (axial) epimer: 1H NMR (300 MHz, CDCl3) δ 5.44 (1H, apt, J )
1.9 Hz), 4.92 (1H, dd, J ) 8.3, 1.9 Hz), 4.71 (1H, dd, J ) 13.4, 6.3
Hz, H2â), 4.37 (1H, dd, J ) 11.2, 4.7 Hz), 2.55 (1H, dd, J ) 13.0, 6.3
Hz, H1â), 2.48-2.42 (2H, m), 2.37 (1H, apt, J ) 8.8 Hz), 2.06 (3H, s),
2.00 (2H, s), 1.36 (3H, s), 1.18 (3H, s), 1.14 (3H, s, H19), 1.10 (3H, s),
0.99 (1H, d, J ) 6.7 Hz); 13C NMR (75 MHz, CDCl3) δ 200.6 (s),
170.6 (s), 155.2 (s), 121.2 (d), 117.5 (s), 84.1 (d), 82.2 (s), 80.4 (d),
56.1 (d), 53.7 (d), 51.4 (d), 51.4 (s), 51.0 (t), 47.0 (d), 43.7 (t), 41.2
(d), 39.3 (s), 37.3 (t), 33.5 (d), 33.3 (t), 30.1 (q), 29.1 (t), 28.6 (q),
28.0 (t), 26.7 (t), 21.3 (q), 15.1 (q), 14.0 (q), 12.0 (q); [R] ) +33.7°
(c ) 1.00, CHCl3); MS (CI) m/z 551/549 (M + H, 10/12), 393/391
(100/35); HRMS (CI) calcd for C29H42BrO5 (M + H) 549.2216, found
549.2226.
Azido Ketone 9: A solution of bromo ketone 39 (31.4 mg, 57 mmol)
in CH3CN (3.2 mL) was cooled to 0 °C and then rapidly added, via
cannula, to a cold (0 °C) solution of TMGA (45.8 mg, 290 mmol, 5
equiv) in CH3CN (2.6 mL). The reaction was allowed to warm slowly
to 25 °C during 18 h of stirring and partitioned between EtOAc and
brine, and the organic layer was dried (Na2SO4) and concentrated to
give crude azido ketone 9 (30 mg, quantitative). 1H NMR analysis
indicated that the material contained 5% R-amino enone side product
and ∼10% 2-epi-9 (axial azide). Crude 9 coupled admirably in the
next step, but in this run sgc (2:1 hexane/EtOAc) furnished pure 9 (24.7

706 J. Am. Chem. Soc., Vol. 120, No. 4, 1998
LaCour et al.
mg, 85%) as white solids for analytical purposes: 1H NMR (300 MHz,
CDCl3) δ 5.43 (1H, apt, J ) 1.9 Hz), 4.92 (1H, dd, J ) 8.2, 1.9 Hz),
4.37 (1H, dd, J ) 11.2, 4.7 Hz), 3.95 (1H, dd, J ) 13.0, 6.7 Hz, H2â),
2.42-2.32 (1H, m), 2.33 (1H, apt, J ) 8.5 Hz), 2.29 (1H, dd, J )
14.4, 4.2 Hz, H4R), 2.22 (1H, dd, J ) 12.6, 6.3 Hz, H1â), 2.06 (3H, s),
1.99 (2H, aps), 1.36 (3H, s), 1.18 (3H, s), 1.13 (3H, s, H19), 1.10 (3H,
s), 0.99 (1H, d, J ) 6.7 Hz); COSY NMR (300 MHz, CDCl3) see
Supporting Information; 13C NMR (75 MHz, CDCl3) δ 204.6 (s), 170.6
(s), 155.3 (s), 121.2 (d), 117.5 (s), 84.1 (d), 82.2 (s), 80.5 (d), 63.7 (d),
56.1 (d), 51.5 (d), 51.4 (s), 47.1 (d), 45.1 (t), 43.6 (t), 41.2 (d), 37.3
(s), 37.3 (t), 33.4 (d), 33.3 (t), 30.1 (q), 29.1 (t), 28.6 (q), 28.0 (t), 26.8
(t), 21.3 (q), 15.0 (q), 14.0 (q), 12.4 (q); MS (FAB, DTT/DTE) 512
(M + H); HRMS (FAB, DTT/DTE) calcd for C29H42N3O5 (M + H)
512.3124, found 512.3134.
Aminomethoxime 10: Azido ketone 9 (9.9 mg, 19 µmol) was
converted to azidomethoxime 40 (10.4 mg, 19 µmol) followed by
Staudinger reduction to give aminomethoxime 10 (7.4 mg, 75% from
9) as a white foam, which resisted crystallization (see Supporting Infor-
mation): 1H NMR (300 MHz, CDCl3) δ 5.40 (1H, apt, J ) 1.8 Hz),
4.91 (1H, dd, J ) 8.3, 1.8 Hz), 4.36 (1H, dd, J ) 11.2, 4.7 Hz), 3.84
(3H, s), 3.49 (1H, dd, J ) 12.2, 5 Hz, H2â), 3.01 (1H, dd, J ) 14.3, 2.6
Hz, H4R), 2.34 (1H, apt, J ) 8.8 Hz), 2.05 (3H, s), 1.99 (2H, aps), 1.36
(3H, s), 1.18 (3H, s), 1.08 (3H, s), 0.98 (1H, d, J ) 6.7 Hz), 0.97 (3H,
s, H19); 13C NMR (75 MHz, CDCl3) δ 170.5 (s), 156.1 (s), 120.6 (d),
117.5 (s), 84.2 (d), 82.1 (s), 80.9 (d), 61.6 (q), 56.1 (d), 51.9 (d), 51.4
(s), 49.7 (d), 48.8 (t), 45.5 (t), 41.2 (d), 37.3 (t), 37.0 (s), 33.6 (d), 33.3
(t), 30.1 (q), 29.3 (t), 28.6 (q), 27.9 (t), 27.3 (t), 26.5 (t), 21.3 (q), 15.1
(q), 14.0 (q), 12.4 (q); MS (FAB, DTT/DTE) 515 (M + H); HRMS
(FAB, DTT/DTE) calcd for C30H47NO5 (M + H) 515.3485, found
515.3468.
Protected Cephalostatin 1 (41). To a solution of R-azido ketone
7 (5 mg, 0.01 mmol) and R-aminomethoxime 6 (9.6 mg, 0.01 mmol)
in benzene (3 mL) was added dichlorodibutylstannane (0.3 mg, 10 mol
%) and polyvinylpyridine (15 mg). The reaction flask was equipped
with a Dean-Stark trap, and the mixture was heated at reflux for 3 h
(2-4 mL of fresh benzene was added twice to maintain the solvent
level in the reaction vessel), at which time TLC (25% EtOAc/hexane)
indicated no remaining 7. The reaction mixture was cooled and filtered,
and the solids were washed with CH2Cl2. Evaporation of the filtrate
and sgc (15-20% EtOAc/hexane) of the residue gave 7.6 mg (59%)
of pure 41 as a white solid and 2.8 mg of recovered R-aminomethoxime
6: 1H NMR (600 MHz, CDCl3) δ 7.86 (2H, d, J ) 6.9 Hz), 7.75 (2H,
d, J ) 7.2 Hz), 7.48-7.38 (6H, m), 5.57 (1H, aps), 5.48 (1H, aps),
5.17 (1H, dd, J ) 6.6, 4.2 Hz), 5.07 (1H, dd, J ) 11.4, 5.4 Hz), 4.95
(1H, aps), 3.97 (1H, s), 3.90-3.79 (3H, m), 3.10 (1H, apd, J ) 9.6
Hz), 2.98 (1H, apd, J ) 10.2 Hz), 2.91-2.81 (4H, m), 2.09 (3H, s),
2.00 (3H, s), 1.31 (3H, s), 1.29 (3H, s), 1.26 (3H, s), 1.25 (3H, s), 1.12
(3H, d), 1.08 (3H, d, J ) 6 Hz), 1.01 (9H, s), 0.86 (3H, s), 0.85 (3H,
s), 0.76 (9H, s), -0.13 (3H, s), -0.14 (3H, s).
(+)-Cephalostatin 1 (1). To a solution of 41 (7 mg, 0.005 mmol)
in THF (2 mL) was added a 1.0 M solution of tetrabutylammonium
fluoride (TBAF) in THF (16 mL, 0.016 mmol), and the mixture was
heated at reflux for 2 h and cooled and the solvent evaporated. The
residue was dissolved in an 8:1 mixture of MeOH/H2O (2 mL), and
K2CO3 (7.5 mg, 0.054 mmol) was added. The resulting suspension
was heated at reflux for 0.5 h, cooled, and concentrated. The residue
was dissolved in EtOAc, washed with water (two to three times) and
dried (Na2SO4). Evaporation of the solvent followed by sgc (3-5%
MeOH in chloroform) of the residue gave 3.8 mg (80%) of pure
cephalostatin 1 (1) as a white solid. The 1H and 13C NMR, TLC, and
HPLC profiles of synthetic 1 and natural 1 were found to be identical.2
For a comparative summary of 1H and 13C NMR of synthetic and natural
cephalostatin 1 (1), see Supporting Information: 1H NMR (500 MHz,
C5D5N) δ 8.16 (1H, d, J ) 7.5 Hz), 7.27 (1H, d, J ) 4 Hz), 6.63 (1H,
t, J ) 5.5 Hz), 6.26 (1H, s), 5.64 (1H, s), 5.44 (1H, s), 5.25 (1H, s),
4.81 (2H, m), 4.71 (1H, d, J ) 1 Hz), 4.08 (1H, d, J ) 12 Hz), 4.06
(1H, m), 4.03 (1H, d, J ) 12 Hz), 3.82 (1H, dd, J ) 11, 5.5 Hz), 3.72
(1H, dd, J ) 11.5, 5 Hz), 3.18 (1H, dq, J ) 7, 6 Hz), 3.08 (1H, d, J
) 16 Hz), 3.05 (1H, d, J ) 17 Hz), 1.65 (3H, s), 1.47 (3H, s), 1.47
(3H, d, J ) 6.5 Hz), 1.39 (3H, s), 1.35 (3H, d, J ) 7 Hz), 1.33 (3H,
s), 0.75 (3H, s), 0.72 (3H, s); 13C NMR (150 MHz, C5D5N) δ 211.75,
152.72, 149.46, 148.99, 148.65, 148.43, 148.37, 123.17, 122.28, 117.15,
110.91, 93.14, 91.66, 82.79, 81.53, 81.13, 75.58, 71.51, 69.29, 64.19,
61.83, 55.40, 53.23, 52.22, 47.33, 45.98, 45.82, 44.50, 44.23, 41.81,
41.24, 39.52, 38.82, 36.33, 36.29, 35.79, 35.73, 35.61, 33.81, 32.87,
32.36, 29.75, 29.51, 29.46, 28.96, 28.69, 28.23, 27.95, 26.40, 15.47,
12.57, 11.73, 11.32, 8.99; [R] ) +95° (c ) 0.04, MeOH) [lit.2 [R] )
+102° (c ) 0.04, MeOH)].
Protected Ritterostatin GN1N 42. North G azido ketone 9 (4.7 mg,
0.0092 mmol), North 1 aminomethoxime 6 (9.6 mg, 0.0105 mmol),
polyvinylpyridine (14 mg), freshly crushed 4A molecular sieves (14
mg), dichlorodibutylstannane (catalyst), and 10 mL of benzene,
procedure as for 41 except 7 mL of distillate was collected over 2 h
without addition of fresh solvent. Sgc (gradient from 2:1 to 1.5:1
hexane/EtOAc, then 100:10:1 EtOAc/MeOH/TEA) afforded 5.9 mg
(49%) of 42 as a white foam, 3.7 mg (31%) of material containing
mainly what appeared by NMR to be partially deacetylated 42, and
2.2 mg of aminomethoxime 6 (23%). The combined yield of pyrazines
was thus 80% based on 9, 92% based on recovered 6: 1H NMR (300
MHz, CDCl3) δ 7.86 (2H, m), 7.74 (2H, dd, J ) 7.9, 1.6 Hz), 7.46-
7.28 (6H, m), 5.56 (1H, aps), 5.45 (1H, aps), 5.06 (1H, dd, J ) 11.2,
5.1 Hz), 4.95 (1H, aps), 4.93 (1H, dd, J ) 8.6, 1.9 Hz), 4.40 (1H, dd,
J ) 11.1, 4.5 Hz), 4.30 (1H, dd, J ) 10.5, 7.9 Hz), 3.97 (1H, s), 3.10
(1H, d, J ) 10.1 Hz), 2.97 (1H, d, J ) 10.1 Hz), 2.87-2.82 (4H, m),
2.78-2.44 (5H, m), 2.38 (1H, dd, J ) 8.8, 8.7 Hz), 2.06 (1H, s), 2.04
(1H, s), 1.37 (3H, s), 1.24 (3H, s), 1.19 (3H, s), 1.12 (3H, d, J ) 6.0
Hz), 1.11 (6H, s), 1.00 (9H, s), 0.99 (1H, d, J ) 7 Hz), 0.86 (3H, s),
0.85 (3H, s), 0.75 (9H, s), -0.14 (3H, s), -0.15 (3H, s); COSY NMR
(300 MHz, CDCl3) see Supporting Information; MS (FAB, DTT/DTE)
1134 (M + H).
Ritterostatin GN1N 4. To 42 (4.7 mg, 0.0035 mmol) in THF (1.5
mL) was added TBAF (0.0105 mL of 1.0 M THF solution, 3 equiv);
the resulting yellow solution was heated at reflux for 2 h and then
cooled. Methanol (1 mL) and 10% KOH (0.1 mL) were added, and
the reaction mixture was heated at reflux for 45 min. The mixture
was cooled and partitioned between EtOAc (10 mL) and brine (10 mL),
and the organic layer was washed with 0.01% HCl. The combined
aqueous layers were extracted with EtOAc (10 mL). The combined
organic layers were washed with saturated NaHCO3, dried over sodium
sulfate, concentrated, and chromatographed (gradient 3-7% MeOH in
CH2Cl2) to afford 2.9 mg of 4 (94%) as white solids.
Partially deacetylated 42 (3.7 mg, the second fraction from the
coupling reaction) was treated as above to afford 2.0 mg of 4 as white
solids (89% combined yield; 62% overall from 9, 69% based on
recovered 6). See the Supporting Information for comparison of the
1H and 13C NMR spectra of 4 with signals of its constituent subunits
in natural 1 and 3: 1H NMR (500 MHz, pyr-d5) δ 8.12 (1H, apd, J )
7.2 Hz), 6.59 (1H, apt, J ) 5.3 Hz), 6.29 (1H, d, J ) 4.8 Hz), 6.24
(1H, s), 5.63 (1H, aps), 5.54 (1H, aps), 5.27 (1H, dd, J ) 8.4, 1.7 Hz),
5.24 (1H, aps), 4.80 (1H, m), 4.70 (1H, d, J ) 1.5 Hz), 4.04 (1H, dd,
J ) 10.6, 4.7 Hz), 3.80 (1H, dd, J ) 11, 4.8 Hz), 3.71 (1H, dd, J )
11, 4.8 Hz), 3.48 (1H, m), 3.11-3.05 (3H, m), 2.92-2.83 (3H, m),
2.72 (1H, dd, J ) 11.4, 8.0 Hz), 2.67-2.61 (4H, m), 2.32 (1H, apt, J
) 11.1 Hz), 2.22 (1H, dq, J ) 8, 7 Hz), 2.16-2.00 (7H, m), 1.92-
1.72 (4H, m), 1.64 (3H, s), 1.46 (3H, s), 1.35 (3H, d, J ) 7 Hz), 1.32
(3H, s), 1.31 (3H, s), 1.24 (3H, d, J ) 6.7 Hz), 1.19 (3H, s), 0.94-
0.83 (2H, m), 0.75 (3H, s), 0.73 (3H, s); 13C NMR (125 MHz, pyr-d5)
δ 157.1, 152.7 (C14′), 150.2, 148.9, 148.6, 148.5, 122.3, 120.4, 117.8,
117.2, 93.2, 91.6, 85.0, 82.8, 81.5, 78.5, 75.6, 71.5, 69.3, 56.3, 55.4,
53.2, 52.6, 46.1, 46.0, 44.5, 42.1, 41.8, 41.8, 39.5, 37.8, 36.3, 36.2,
35.8, 35.8, 34.0, 33.8, 33.6, 30.9, 30.3, 29.7, 28.8, 28.9, 28.7, 28.4,
27.9, 26.4, 14.5, 13.9, 12.6, 11.8, 11.7, 9.0; [R] ) +166° (c ) 0.01,
CHCl3); MS (FAB, DTT/DTE) 879 (M + H); HRMS (FAB, DTT/
DTE) calcd for C54H75N2O5Si (M + H) 879.5523, found 879.5449.
Protected Ritterostatin GN1S 43. To a 25 mL flask equipped with
a Dean-Stark trap and condenser was charged North G ami-
nomethoxime 10 (5.4 mg, 0.0105 mmol), South 1 azido ketone 7 (6.0
mg, 0.011 mmol), Nafion H (11.5 mg), dichlorodibutylstannane
(catalyst), and 10 mL of benzene. The apparatus was heated to reflux,
with removal of the contents of the Dean-Stark trap (3.5 mL) after 1
h and again after 3 h heating; a further 1.5 mL was removed after 4 h,
leaving the final reaction volume at 1.5 mL, TLC analysis still showed

Cephalostatin Synthesis. 11
J. Am. Chem. Soc., Vol. 120, No. 4, 1998 707
remaining 7 (3:1 hexane/EtOAc) and 10 (100:10:1 EtOAc/MeOH/TEA)
with one main new spot and another more polar spot (1:1 hexane/
EtOAc). The mixture was cooled, filtered, and chromatographed
(gradient from 1:1 to 1:2 hexane/EtOAc, then 100:10:1 EtOAc/MeOH/
TEA) to afford 1.3 mg (21%) recovered impure 7, 5.2 mg (52%)
protected ritterostatin GN1S 43 as white solids, 2.0 mg (21%) of material
containing mainly what appeared by NMR to be partially deacetylated
pyrazines 43, and 1.5 mg (26%) of clean 10. The combined yield of
the pyrazines was thus 73% based on azido ketone 7, 99% based on
recovered aminomethoxime 10: 1H NMR (500 MHz, CDCl3) δ 5.48
(1H, aps), 5.44 (1H, aps), 5.16 (1H, dd, J ) 6.7, 4.6 Hz,), 4.93 (1H,
dd, J ) 8.2, 1.8 Hz), 4.40 (1H, dd, J ) 11.2, 4.7 Hz), 3.85 (1H, d, J
) 8.0 Hz), 3.85 (1H, d, J ) 8.0 Hz), 2.91-2.78 (4H, m), 2.71-2.28
(6H, m), 2.08 (3H, s), 2.05 (3H, s), 1.37 (3H, s), 1.30 (3H, s), 1.28
(3H, s), 1.18 (3H, s), 1.11 (3H, s), 1.07 (3H, dd, J ) 6.2 Hz), 1.00
(3H, dd, J ) 6.7 Hz), 0.85 (3H, s), 0.84 (3H, s); MS (FAB, DTT/
DTE) 947 (M + H); HRMS (FAB, DTT/DTE) calcd for C58H79N2O9
(M + H) 947.5786, found 947.5737.
(1H, dt, J ) 13.4, 6.9 Hz), 2.17-1.80 (10H, m), 1.70-1.53 (6H, m),
1.47 (3H, s), 1.47 (3H, d, J ) 6.8 Hz), 1.46 (3H, s), 1.38 (3H, s), 1.32
(3H, s), 1.25 (3H, d, J ) 6.5 Hz), 1.19 (3H, s), 0.77 (3H, s), 0.73 (3H,
s); 13C NMR (125 MHz, pyr-d5) δ 157.1, 149.4, 149.0, 148.7, 148.4,
148.3, 123.0, 120.4, 117.8, 110.9, 85.0, 81.5, 81.5, 81.1, 78.6, 64.2,
61.8, 56.4, 53.5, 52.6, 52.2, 47.3, 46.1, 45.8, 44.2, 42.1, 41.7, 41.2,
38.8, 37.8, 36.3, 36.2, 35.8, 35.8, 35.6, 34.0, 33.6, 32.9, 32.4, 30.9,
30.3, 30.0, 29.7, 29.7, 29.4, 28.7, 28.4, 28.0, 15.5, 14.5, 13.9, 11.8,
11.3; IR (CHCl3) 1705; [R] ) +66° (c ) 0.05 MeOH); MS (FAB,
DTT/DTE) 863 (M + H); HRMS (FAB, DTT/DTE) calcd for
C54H75N2O7 (M + H) 863.5574, found 863.5548.
Acknowledgment. We thank the National Institute of Health
(CA 60548) and the Indiana Elks/Purdue Cancer Center (T.G.L.
and C.G.) for financial support of this work. We are grateful to
Arlene Rothwell for MS data and to Drs. Tom Crawford and
Tamim Braish of Pfizer Pharmaceutical Company for providing
us a gift of hecogenin acetate. We thank Mr. Sunghoon Ma for
his work on material related to the ritterazines, and for providing
a sample and X-ray data which allowed us to confirm the
stereochemistry of an intermediate. We also thank Mr. Zhiwei
Tong for his efforts to achieve an improved route to North G
via ene reaction of 22 and for preparation of some intermediates.
We are indebted to Professor Harry Morrison and his research
group, most especially Dr. Taj Mohammad, for discussions,
considerable and patient help, and use of equipment for
performing photolyses.
Ritterostatin GN1S 5. Diacetate 43 (5.1 mg, 0.0054 mmol) was
dissolved in 88% aqueous MeOH (2 mL), and K2CO3 (4 mg) was added.
The mixture was heated at reflux for 30 min and then cooled. The
mixture was partitioned between CH2Cl2 (10 mL) and water (10 mL).
The organic layer was dried over sodium sulfate and then concentrated
to afford 5 mg of 5 (quantitative) as white solids.
Partially deacetylated 43 (2.0 mg, the second fraction from the
coupling reaction) was treated as above to afford 2 mg of impure 5 as
white solids. The two lots of 5 were combined and chromatographed
(gradient 2-4% MeOH in CH2Cl2) to afford 6.0 mg of 5 as white solids
(94% combined yield; 68% overall from 10, 93% overall based on
1
recovered 10). See Supporting Information for comparison of the H
Supporting Information Available: Experimentals for
compounds 6, 13-18, 20a-d, 21, 27-32, 36-38, and 40; NMR
spectra and extensive signal assignments for all compounds;
Schemes 8a and 10a; CAChe drawing of 32 (ground state);
calculated energies of spiroketal isomers; and tables comparing
1H and 13C NMR data for natural versus synthetic pyrazines
(155 pages). See any current masthead page for ordering and
Internet access instructions.
and 13C NMR spectra of 5 with signals of its constituent subunits in
natural 1 and 3: 1H NMR: (300 MHz, pyr-d5) δ 7.19 (1H, brs), 5.55
(1H, aps), 5.54 (1H, aps), 5.28 (1H, brd, J ) 8.0 Hz), 4.80 (1H, m),
4.07 (1H, d, J ) 11.9 Hz), 4.04 (1H, d, J ) 11.9 Hz), 3.49 (1H, m),
3.18 (1H, dq, J ) 6.8, 6.8 Hz), 3.10 (1H, d, J ) 17.5 Hz), 3.09 (1H,
apq, J ) 8.0 Hz), 3.08 (1H, d, J ) 16.9 Hz), 2.93 (1H, dd, J ) 18.1,
5.0 Hz), 2.91 (1H, d, J ) 18.0, 5.0 Hz), 2.85 (1H, apdt, J ) 12, 4 Hz),
2.78 (1H, apt, J ) 13.7 Hz), 2.81-2.76 (1H, m), 2.71-2.61 (3H, m),
2.64 (1H, d, J ) 16.1 Hz), 2.62 (1H, dd, J ) 13.5, 3.2 Hz), 2.57 (1H,
d, J ) 17.5 Hz), 2.35 (1H, dd, J ) 12.3, 7.1 Hz), 2.32 (1H, m), 2.23
JA972160P