(+)-Saxitoxin: A first and second generation stereoselective synthesis

Article abstract of DOI:10.1021/ja071501o

A stereoselective synthesis of the bis-guanidinium toxin (+)-saxitoxin (STX), the agent infamously associated with red tides and paralytic shellfish poisoning, is described. Our approach to this unique natural product advances through an unusual nine-membered ring guanidine intermediate 39 en route to the tricyclic skeleton that defines STX. The effectiveness of this strategy is notable, as only four steps are needed to transform 39 into the target molecule, including a four-electron alkene oxidation catalyzed by OsCl₃. Construction of the critical monocyclic guanidine has been achieved through two channels, the first of which makes use of Rh-catalyzed C-H amination and highlights a novel class of heterocyclic N,O-acetals as iminium ion equivalents for crafting functionalized amines. A second route to 39 relies on a stereoselective acetylide dianion addition to a serine-based nitrone, thereby facilitating the preparation of STX in just 14 linear steps from commercial material.

Full text of DOI:10.1021/ja071501o

Published on Web 07/21/2007
(+)-Saxitoxin: A First and Second Generation Stereoselective
Synthesis
James J. Fleming,1 Matthew D. McReynolds,2 and J. Du Bois*
Contribution from the Department of Chemistry, Stanford UniVersity,
Stanford, California 94305-0080
Received March 2, 2007; E-mail: jdubois@stanford.edu
Abstract: A stereoselective synthesis of the bis-guanidinium toxin (+)-saxitoxin (STX), the agent infamously
associated with red tides and paralytic shellfish poisoning, is described. Our approach to this unique natural
product advances through an unusual nine-membered ring guanidine intermediate 39 en route to the tricyclic
skeleton that defines STX. The effectiveness of this strategy is notable, as only four steps are needed to
transform 39 into the target molecule, including a four-electron alkene oxidation catalyzed by OsCl3.
Construction of the critical monocyclic guanidine has been achieved through two channels, the first of
which makes use of Rh-catalyzed C-H amination and highlights a novel class of heterocyclic N,O-acetals
as iminium ion equivalents for crafting functionalized amines. A second route to 39 relies on a stereoselective
acetylide dianion addition to a serine-based nitrone, thereby facilitating the preparation of STX in just 14
linear steps from commercial material.
Introduction
Neurotoxic agents can serve as important pharmacological
tools for understanding protein function associated with the
highly complex ionic mechanisms of electrical transmission in
cells. The guanidinium toxins, (+)-saxitoxin and (-)-tetrodot-
oxin (Figure 1), are exemplary in this regard and have been
instrumental for the identification, characterization, and study
of voltage-gated sodium channels (vg-Na+ channels). As
molecular targets, these natural products present incomparable
challenges for chemical methods research given the degree of
oxidation and the highly polar nature of the heteroatom
substituents. Although our initial interest in these compounds
drew principally from a desire to showcase new tactics for
oxidative C-N bond construction, it was apparent from the
outset of this work that efficient access to a molecule such as
saxitoxin (STX) would create opportunities to explore Na+
channel function through the systematic and rational design of
non-natural STX-based structures. Herein, we present a detailed
analysis of our synthetic strategies to this most unusual bis-
guanidinium natural product.3 The elegant and timeless works
of the Kishi and Jacobi labs, which constitute the only other
reported routes to the toxin, provided valuable guidance to our
own synthetic planning.4,5 Accordingly, our efforts have cul-
minated in both first and second generation asymmetric
Figure 1. Structurally unique guanidinium toxins act as site I selective
blockers of voltage-gated Na+ ion channels.
syntheses of (+)-STX, the shortest of which requires 14 linear
operations from commercial materials. The application of Rh-
catalyzed C-H amination is one of several unique transforma-
tions that highlight this work and make available the desired
target in a manner amenable to programmed modification.
Background
Ion channels are the molecular machinery that allow passive
diffusion of ions across the cell membrane and are found in all
animal, plant, and bacterial cells.6 Such proteins intimately
control the movement of Na+, K+, Ca2+, and Cl- ions to
produce small, transient changes in ion concentrations. Ion flux
in and out of a cell is, in turn, coupled to a diverse number of
biochemical processes, which include nerve and muscle excita-
tion, hormonal secretion, and sensory transduction. The ability
to modulate ion passage through specific channels could greatly
aid efforts to deconvolute the integrated circuitry of these
remarkable transmembrane conduits.6,7 One strategy to exact
such control would capitalize on small molecules that interact
(1) This work was taken, in part, from the Ph.D. thesis of J. J. Fleming, Stanford
University, 2006.
(2) Current address: Pfizer, Inc., St. Louis Laboratories, 700 Chesterfield
Parkway, Chesterfield, MO 63017.
(3) Fleming, J. J.; Du Bois, J. J. Am. Chem. Soc. 2006, 128, 3926-3927.
(4) (a) Tanino, H.; Nakata, T.; Kaneko, T.; Kishi, Y. J. Am. Chem. Soc. 1977,
99, 2818-2819. (b) Kishi, Y. Heterocycles 1980, 14, 1477-1495. (c) Hong,
C. Y.; Kishi, Y. J. Am. Chem. Soc. 1992, 114, 7001-7006.
(5) (a) Jacobi, P. A.; Martinelli, M. J.; Polanc, S. J. Am. Chem. Soc. 1984,
106, 5594-5598. (b) Martinelli, M. J.; Brownstein, A. D.; Jacobi, P. A.;
Polanc, S. Croat. Chem. Acta. 1986, 59, 267-295.
(6) (a) Ashcroft, F. M. Ion Channels and Disease; Academic Press: San Diego,
2000. (b) Hille, B. Ion Channels of Excitable Membranes, 3rd ed.; Sinauer
Associates: Sunderland, MA, 2001.
(7) (a) Cestele, S.; Catterall, W. A. Biochimie 2000, 82, 883-892. (b) Wang,
S. Y.; Wang, G. K. Cellular Signaling 2003, 15, 151-159. (c) Yamaoka,
K.; Vogel, S. M.; Seyama, I. Curr. Pharm. Des. 2006, 12, 429-442.
9
9964
J. AM. CHEM. SOC. 2007, 129, 9964-9975
10.1021/ja071501o CCC: $37.00 © 2007 American Chemical Society

Stereoselective Synthesis of (+)-Saxitoxin
A R T I C L E S
with specific channel isoforms. Because of the general absence
of detailed structural information on most ion channel
proteins, however, the de noVo design of such chemical
agents is exceedingly difficult.8 Fortunately, Nature has provided
a number of topologically unique, low molecular weight
compounds that act as modulators of ion channel proteins;
such compounds can serve as scaffolds from which to
craft new materials with unprecedented pharmacological
properties.
The paralytic shellfish poison, (+)-saxitoxin, most notably
associated with outbreaks of red tide, was first isolated in pure
form by Schantz in 1957 and structurally characterized by X-ray
analysis by the groups of Schantz/Clardy and Rapoport in
1975.9,10 This small molecule is among the most lethal non-
proteinaceous substances known, its acute toxicity resulting from
its ability to disable ionic conductance through the voltage-gated
sodium channel, a characteristic shared by the fugu poison, (-)-
tetrodotoxin (TTX).11,12 At a molecular level, STX and TTX
act by occluding the channel pore, lodging with nanomolar
affinity in the extracellular mouth of certain voltage-gated Na+
channel isoforms.11,13 Physiological studies of the Na+ channel
protein have been empowered with the availability of STX,
TTX, and a small number of naturally occurring isolates having
related guanidinium structures.14 More information pertinent to
understanding both Na+ channel structure and function would
be gained if specific chemical tailoring could be done to either
toxin.15 Semisynthetic protocols, however, have proven largely
ineffective for modifying these heteroatom-rich, charged, hy-
drophilic natural products.11a,16 The development of a concise,
easily modified route to STX is therefore warranted, as it is
only through total synthesis that the potential of STX to serve
as a blueprint for new Na+ channel blockers having designed
function may be realized.
Figure 2. Retrosynthetic scheme for STX assembly: two strategies for
accessing the nine-membered ring guanidine 2.
Results and Discussion
Synthetic Plan. The challenges associated with STX as a
target for chemical synthesis arise, in part, from the remarkably
dense configuration of heteroatoms about its tricyclic core. As
indicated by the molecular formula, C10H19N7O4, the total carbon
count is in fact less than the combined number of oxygen and
nitrogen atoms that comprise the natural product. Furthermore,
the dicationic nature of STX adds complications to the handling
and purification of this target. Nevertheless, shortly after the
molecular structure of STX was revealed by X-ray analysis,
two highly creative synthetic plans for its preparation were
described.4,5 In both instances the tricyclic core of the natural
product is first assembled and the desired toxin is revealed
following a common end-game strategy where pseudourea
moieties are converted to their guanidinium equivalents. Ac-
cordingly, a more expeditious synthesis of STX might avoid
these types of late-stage functional group exchange reactions
and put directly in place the intact guanidinium groups.
Following this line of analysis, unraveling the tetrasubstituted
C4 aminal to its component parts, a C4 ketone and two
guanidine units, thus became a defining disconnection in our
approach (Figure 2). Such a plan posits that all three rings of
the STX core can be formed from a nine-membered ring bis-
guanidine 1. Importantly, the stereochemical configuration of
the tetrasubstituted C4 aminal position would be controlled by
the chirality of the attendant groups at C5 and C6. Although
the decision to prepare a medium-sized ring intermediate such
as 1 would not appear simplifying at first glance, such a move
transforms a highly complex problem in stereocontrolled cyclic
synthesis to the seemingly more manageable problem of
preparing an acyclic polyamine (e.g., 3 or 4) en route to 1. The
assembly of such an acyclic material could capitalize on the
utility of our directed C-H amination methods.
(8) Potassium and chloride ion channels have been characterized crystallo-
graphically. For recent reviews of this work, see: (a) Tombola, F.; Pathak,
M. M.; Isacoff, E. Y. Annu. ReV. Cell DeV. Biol. 2006, 22, 23-52. (b)
Dutzler, R. Curr. Opin. Struct. Biol. 2006, 16, 439-446. (c) Gouaux, E.;
MacKinnon, R. Science 2005, 310, 1461-1465.
(9) Schantz, E. J.; Mold, J. D.; Stanger, D. W.; Shavel, J.; Riel, F. J.; Bowden,
J. P.; Lynch, J. M.; Wyler, R. S.; Riegel, B.; Sommer, H. J. Am. Chem.
Soc. 1957, 79, 5230-5235.
(10) (a) Schantz, E. J.; Ghazarossian, V. E.; Schnoes, H. K.; Strong, F. M.;
Springer, J. P.; Pezzanite, J. O.; Clardy, J. J. Am. Chem. Soc. 1975, 97,
1238-1239. (b) Bordner, J.; Thiessen, W. E.; Bates, H. A.; Rapoport, H.
J. Am. Chem. Soc. 1975, 97, 6008-6012.
(11) (a) Tetrodotoxin, Saxitoxin, and the Molecular Biology of the Sodium
Channel; Kao, C. Y., Levinson, S. R., Eds.; Annals of the New York
Academy Science: New York, 1986; Vol. 479. (b) Narahashi, T. J. Toxicol.,
Toxin ReV. 2001, 20, 67-84. (c) Novakovic, S. D.; Eglen, R. M.; Hunter,
J. C. Trends Neurosci. 2001, 24, 473-478.
(12) (a) Evans, M. H. Br. J. Pharmacol. Chemother. 1964, 22, 478-479. (b)
Kao, C. Y.; Nishiyama, A. J. Physiol. (London, U.K.) 1965, 1, 50-51. (c)
Kao, C. Y. Pharmacol. ReV. 1966, 2, 997-1049.
(13) For a general discussion of Na+ channel isoforms, see: (a) Catterall, W.
A.; Goldin, A. L.; Waxman, S. G. Pharmacol. ReV. 2003, 55, 575-578.
(b) Catterall, W. A.; Goldin, A. L.; Waxman, S. G. Pharmacol. ReV. 2005,
57, 397-409.
The preparation of medium-sized rings, particularly hetero-
cycles of eight and nine units in size, is challenging, and the
construction of a nine-membered cyclic guanidine is without
precedent.17,18 In considering approaches to compounds such
as 1, two strategies were most appealing: ring closing-metathesis
(14) (a) Hall, S.; Strichartz, G.; Moczydlowski, E.; Ravindran, A.; Reichardt,
P. B. ACS Symp. Ser. 1990, 418, 29-65. (b) Yotsu-Yamashita, M. J.
Toxicol., Toxin ReV. 2001, 20, 51-66. (c) Llewellyn, L. E. Nat. Prod. Rep.
2006, 23, 200-222. (d) Llewellyn, L.; Negri, A.; Robertson, A. J. Toxicol.,
Toxin ReV. 2006, 25, 159-196.
(15) Kishi has prepared a small number of STX derivatives through de novo
synthesis, see: Strichartz, G. R.; Hall, S.; Magnani, B.; Hong, C. Y.; Kishi,
Y.; Debin, J. A. Toxicon 1995, 33, 723-737.
(16) (a) Koehn, F. E.; Ghazarossian, V. E.; Schantz, E. J.; Schnoes, H. K.; Strong,
F. M. Bioorg. Chem. 1981, 10, 412-428. (b) Hu, S. L.; Kao, C. Y.; Koehn,
F. E.; Schnoes, H. K. Toxicon 1987, 25, 159-165. (c) Sato, S.; Sakai, R.;
Kodama, M. Bioorg. Med. Chem. Lett. 2000, 10, 1787-1789. (c) Watanabe,
R.; Samusawa-Saito, R.; Oshima, Y. Bioconjugate Chem. 2006, 17, 459-
465.
(17) For general reviews on the synthesis of medium-sized rings, see: (a) Evans,
P. A.; Holmes, A. B. Tetrahedron 1991, 47, 9131-9166. (b) Yet, L. Chem.
ReV. 2000, 100, 2963-3007.
(18) The assembly of an eight-membered ring guanidine has been reported;
however, the formation of a nine-membered ring was not possible under
the same conditions, see: Fukada, N.; Takano, M.; Nakai, K.; Kuboike,
S.; Takeda, Y. Bull. Chem. Soc. Jpn. 1993, 66, 148-152.
9
J. AM. CHEM. SOC. VOL. 129, NO. 32, 2007 9965

A R T I C L E S
Fleming et al.
Figure 3. Highlighting the versatility of oxathiazinane dioxide heterocycles
for amine synthesis.
Figure 5. Exploring reaction chemoselectivity with sulfamate ester model
systems.
attracted to such a strategy, at the time these studies were
initiated, we had little experience with amination reactions of
unsaturated sulfamates and were unsure as to whether olefin
aziridination would compete with the desired allylic insertion
pathway.22,23 To examine questions pertaining to functional
group chemoselectivity, model sulfamates 7 and 8 were
subjected to standard reaction conditions that employed 2 mol
% Rh2(O2CR)4, PhI(OAc)2, and MgO (Figure 5). In each case,
the product of alkene aziridination formed in amounts ap-
proximately equal to or greater than the allylic amine, despite
the larger-size ring formation. Changing catalysts from Rh2-
(OAc)4 to Rh2(O2CCPh3)4 or Rh2(esp)2 with the hope of favoring
the oxathiazinane product did not have the intended effect.24
Although the C-H insertion product from 8 would provide an
ideal starting material for evolving a metathesis-based route to
1, our inability to generate this compound in yields greater than
20% necessitated a new approach to such targets.
Confident in the “oxathiazinane-approach” for preparing STX,
we considered a new tactic by which C-H amination could be
employed to create highly functionalized, unsaturated hetero-
cycles of this type. Knowledge accrued through mechanistic
studies on the amination process had revealed that C-H bond
reactivity follows the general order: 3° > R-ethereal ∼ benzylic
> 2° . 1°.22 The high intrinsic activity of R-ethereal centers
relative to most other C-H bonds suggested to us that such
groups could be used to direct the insertion event in sulfamate
esters bearing a large degree of attendant functionality.25
Importantly, amination of an ethereal R-C-H bond would afford
a novel N,O-acetal product that could serve as an imine or
iminium ion surrogate for subsequent modification (Figure 6).
Implementing this strategy would afford access to oxathiazinane
materials not easily prepared through direct means; in addition,
the utility of such a process might ultimately transcend the
synthesis of STX itself.
Figure 4. Possible strategy for employing C-H amination.
(RCM) and intramolecular isothiourea-amine condensation
(Figure 2). Both of these plans require the design of function-
alized acyclic substrates that contain a contiguous hydroxy
diamine triad (i.e., C5-C6-C13, STX numbering).19 The
preparation of nitrogen-rich molecules by way of catalytic C-H
amination is of special interest in our laboratory; thus, in
considering paths to an anti-diamino alcohol intermediate, we
recognized the unique potential of the [1,2,3]-oxathiazinane-
2,2-dioxide heterocycle for such a purpose (Figure 3).20
Structures of this type are valuable synthons for 1,3-substituted
amine derivatives and, in the context of the STX work, appeared
ideally suited for masking the C5,C13 amino alcohol and
crafting an appropriate precursor to the nine-membered ring
guanidine. As our lab has shown, oxidative C-H insertion of
sulfamate esters enables construction of substituted oxathiazi-
nanes;20 however, the application of this method in the context
of the STX assembly would provide a formidable test of reaction
chemoselectivity and efficiency (Vide infra).
C-H Insertion for Preparing Amine Derivatives. In
principle, there exist a number of potentially viable routes using
C-H amination to access cyclic guanidine 1 and related
structures (Figure 4). One possibility is to establish the C5 amino
group through oxidative cyclization of a C13 sulfamate ester
5. Such a transformation would require that C-H insertion occur
at a methylene center adjacent to a ketone moiety (or some
protected form). In general, amination reactions at carbon centers
proximal to electron-withdrawing groups are difficult and often
result in depressed product yields.21 As an alternative plan,
masking the requisite C4,C12 R-ketol as an alkene would
provide a more easily oxidized substrate, which could then be
manipulated to an intermediate resembling 1. While we were
The utilization of ethereal groups to bias the positional
selectivity of the C-H amination reaction has been reduced to
(19) For the synthesis of anti-1,2-diamines, see: (a) Davis, F. A.; Deng, J. H.
Org. Lett. 2004, 16, 2789-2792. (b) Luo, Y.; Blaskovich, M. A.; Lajoie,
G. A. J. Org. Chem. 1999, 16, 6106-6111. (c) Rossi, F. M.; Powers, E.
T.; Yoon, R.; Rosenberg, L.; Meinwald, J. Tetrahedron 1996, 31, 10279-
10286.
(20) For the initial report on the oxidative cyclization of sulfamate esters, see:
Espino, C. G.; Wehn, P. M.; Chow, J.; Du Bois, J. J. Am. Chem. Soc.
2001, 28, 6935-6936. Also, see: (a) Wehn, P. M.; Lee, J. H.; Du, Bois,
J. Org. Lett. 2003, 25, 4823-4826. (b) Wehn, P. M.; Du, Bois, J. Org.
Lett. 2005, 7, 4685-4688.
(21) For general reviews on C-H amination, see: (a) Espino, C. G.; Du Bois,
J. In Modern Rhodium-Catalyzed Organic Reactions; Evans, P. A., Ed.;
Wiley-VCH: Weinheim, 2005; pp 379-416. (b) Davies, H. M. L. Angew.
Chem., Int. Ed. 2006, 45, 6422-6425. (c) Lebel, H.; Leogane, O.; Huard,
K.; Lectard, S. Pure Appl. Chem. 2006, 78, 363-375. (d) Halfen, J. A.
Curr. Org. Chem. 2005, 9, 657-669. (e) Mu¨ller, P.; Fruit, C. Chem. ReV.
2003, 103, 2905-2920. (f) Dauban, P.; Dodd, R. H. Synlett 2003, 1571-
1586.
(22) Du Bois, J. CHEMTRACTSsOrg. Chem. 2005, 18, 1-13.
(23) Alkene aziridination using sulfamate esters has been described, see: (a)
Guthikonda, K.; Du Bois, J. J. Am. Chem. Soc. 2002, 46, 13672-13673.
(b) Di Chenna, P. H.; Robert-Peillard, F.; Dauban, P.; Dodd, R. H. Org.
Lett. 2004, 6, 4503-4505. (c) Guthikonda, K.; Wehn, P. M.; Caliando, B.
J.; Du Bois, J. Tetrahedron 2006, 62, 11331-11342.
(24) Oxidation of 8 using Rh2(OAc)4 gave a ∼1.5:1 mixture favoring the six-
membered ring product. Results using catalytic Rh2(O2CCPh3)4 were similar.
The stereochemistry of the bicyclic aziridine was not determined.
(25) Nitrene insertion into an ethereal R-C-H bond parallels carbene reactivity,
see: (a) Doyle, M. P.; McKervey, M. A.; Ye, T. Modern Catalytic Methods
for Organic Synthesis with Diazo Compounds: From Cyclopropanes to
Ylides; Wiley and Sons, Ltd.: New York, 1997. (b) Davies, H. M. L.;
Beckwith, E. J. Chem. ReV. 2003, 103, 2861-2904. (c) Wang, P.; Adams,
J. J. Am. Chem. Soc. 1994, 116, 3296-3305.
9
9966 J. AM. CHEM. SOC. VOL. 129, NO. 32, 2007

Stereoselective Synthesis of (+)-Saxitoxin
A R T I C L E S
to consider this process as a means for fabricating the desired
nine-membered guanidine.31 Myriad displays of medium-sized
ring assembly by RCM notwithstanding, only a handful of
examples of nine-membered ring formation dot the literature,
the preparation of a nine-membered cyclic urea serving as
perhaps the closest precedent to our desired goal.32,33 Starting
from N,O-acetal 10, an appropriate RCM diene substrate was
assembled through a sequence highlighted by the stereoselective
addition of divinyl zinc to 10 (Scheme 1). This reaction
represents our first example of a vinyl metal coupling to an
N,O-acetal derivative and affords a particularly useful unsatur-
ated oxathiazinane 12. Bis-guanidine 16 was accessible from
12 through a series of straightforward functional group trans-
formations. For reasons not entirely obvious to us, installation
of the azido moiety proved quite difficult and was poor yielding.
Still more vexing, diene 16 was an unwilling participant for
RCM under conditions employing either first or second genera-
tion Grubbs or Hoveyda-Grubbs catalysts. Modification of the
reaction conditions, including the use of Ti(OiPr)4 as a Lewis
acid addend, also failed to provide any of the desired nine-
membered ring product.34 In all cases, starting material 16 and
unidentifiable (assumed-to-be polymeric) products were ob-
tained. The inability to execute RCM on 16 may be due to the
polar nature of the two guanidine units and/or aggregation effects
caused by the large number of hydrogen bond donor and
acceptor groups. Different solvents, which included THF, did
not influence the reaction outcome. As we would determine
subsequently, the same protocols tried on an analogous urea
substrate 18 were unable to induce nine-membered ring forma-
tion (Figure 8).35
Figure 6. N,O-Acetal strategy for accessing substituted oxathiazinane
heterocycles.
Figure 7. Organozinc addition to N,O-acetal 10 affords alkyne-derived
oxathiazinane products.
practice, thus enabling facile preparation of unique oxathiazinane
N,O-acetals.26 These compounds, when treated with a Lewis
acid (e.g., BF3‚OEt2, Sc(OTf)3), undergo smooth coupling with
allylsilane, silyl enol ether, and silyl ketene acetal nucleophiles
to give value-added oxathiazinane products. In the addition
reaction, nucleophilic attack on the putative iminium ion
intermediate is found to be highly diastereoselective, and the
sense of induction is decidedly predictable. The development
of this chemistry has been chronicled, and the reader is referred
to previously published reports from this lab for a more
extensive discussion on this topic.26
For the expressed purpose of the STX synthesis, a glycerol-
derived sulfamate ester 9 was deemed an optimal starting
material due to its ease of assembly in enantiomerically pure
form on multigram scale and the facility by which it submitted
to oxidative cyclization (Figure 7).27,28 In fact, using our newest
catalyst, Rh2(esp)2, sulfamate 9 oxidation is easily effected at
0.3 mol % catalyst loading.29,30 The product N,O-acetal 10 is
stable to isolation and storage and is formed in sufficiently high
purity that it may be employed without purification. Finally, in
the presence of alkynyl zinc reagents and BF3·OEt2, addition
reactions proceed efficiently to give the coupled product (i.e.,
11). Such alkynyl oxathiazinane heterocycles represent the
cornerstone of our first generation STX construction.
Undeterred by the inability to prepare 17 or 19 through RCM,
a second plan was adopted that would rely on a condensation
reaction to drive ring-closure. Arguably, the most successful
methods in the literature for the generation of medium-sized
and macrocyclic rings involve the addition of an alcohol to an
activated carbonyl (i.e., lactonization).36 By analogy, a strategy
for cyclic guanidine synthesis can be envisioned in which an
amine is added to an isothiourea. This type of condensation is
commonly employed for the preparation of substituted acyclic
guanidines; in most cases, the reaction is presumed to operate
through a reactive carbodiimide intermediate.37 Accordingly,
such a process could be staged in an intramolecular fashion to
give 2 (Figure 9). The ability to access isothiourea-derived amine
4 from N,O-acetal 10 is an attractive feature of this design.
Ready access to differentially substituted alkynyl oxathiazi-
nanes, as exemplified by 11, has allowed us to examine
numerous approaches for forming the nine-membered ring
guanidine, and in this context the value of the N,O-acetal method
cannot be overstated. With the absence of any prior art to guide
our attempts to fashion a nine-membered ring guanidine, two
disparate approaches were explored; each, however, would share
a common origin in 10.
(31) (a) Handbook of Metathesis; Grubbs, R. H., Ed.; Wiley-VCH: Weinheim,
2003; Vol. 2. (b) Deiters, A.; Martin, S. F. Chem. ReV. 2004, 104, 2199-
2238. (c) Conrad, J. C.; Fogg, D. E. Curr. Org. Chem. 2006, 10, 185-
202.
Formation of the Nine-Membered Ring Guanidine. The
power of ring-closing metathesis (RCM) for the construction
of polar, nitrogen-containing heterocycles gave us ample cause
(32) For some recent examples, see: (a) Hirama, M.; Oishi, T.; Uehara, H.;
Inoue, M.; Maruyama, M.; Guri, H.; Satake, M. Science 2001, 294, 1904-
1907. (b) Clark, J. S.; Marlin, F.; Nay, B.; Wilson, C. Org. Lett. 2003, 5,
89-92. (c) Crimmins, M. T.; Powell, M. T. J. Am. Chem. Soc. 2003, 125,
7592-7595. (d) Gaich, T.; Mulzer, J. Org. Lett. 2005, 7, 1311-1313. (e)
Sato, K.; Sasaki, M. Org. Lett. 2005, 7, 2441-2444. (f) Pansare, S. V.;
Adsool, V. A. Org. Lett. 2006, 8, 5897-5899. (g) Li, Y.; Hale, K. J. Org.
Lett. 2007, 9, 1267-1270.
(33) McReynolds, M. D., Ph.D. Thesis, University of Kansas, Lawrence, 2004.
(34) Fu¨rstner, A.; Langemann, K. J. Am. Chem. Soc. 1997, 119, 9130-9136.
(35) Urea 18 was synthesized from 12 following a sequence of steps analogous
to those shown in Scheme 1.
(36) (a) Meng, Q. C.; Hesse, M. Top. Curr. Chem. 1992, 161, 107-176. (b)
Parenty, A.; Moreau, X.; Campagne, J.-M. Chem. ReV. 2006, 106, 911-
939.
(37) For a recent comprehensive review on guanidine synthesis, see: Katritzky,
A.; Rogovoy, B. V. ArkiVoc 2005, 49-87. Also, see: Ma, D. W.; Xia, C.
F.; Jiang, J. Q.; Zhang, J. H.; Tang, W. J. J. Org. Chem. 2003, 68, 442-
451.
(26) (a) Fleming, J. J.; Fiori, K. W.; Du Bois, J. J. Am. Chem. Soc. 2003, 125,
2028-2029. (b) Fiori, K. W.; Fleming, J. J.; Du Bois, J. Angew. Chem.,
Int. Ed. 2004, 43, 4349-4352.
(27) (R)-(-)-2,2-Dimethyl-1,3-dioxolane-4-methanol (i.e., (R)-glycerol acetonide)
was purchased from Oakwood Products Inc., West Columbia, SC 29172.
(28) (R)-Glycerol acetonide can also be obtained by resolution of (()-glycerol
acetonide or from (S)-glyceraldehyde acetonide, see: (a) Pallavicini, M.;
Valoti, E.; Villa, L.; Piccolo, O. Tetrahedron: Asymmetry 1994, 5, 5-8.
(b) Pallavicini, M.; Valoti, E.; Villa, L.; Piccolo, O. Tetrahedron:
Asymmetry 1996, 7, 1117-1122. (c) Hubschwerlen, C.; Specklin, J. L.;
Higelin, J. Org. Synth. 1995, 72, 1-5.
(29) An earlier, optimized preparation of 9 required 3 mol% Rh2(O2CCPh3)4,
see ref 26a.
(30) Rh2(esp)2 ) Rh2(R,R,R′,R′-tetramethyl-1,3-benzenedipropionate)2, see: Es-
pino, C. G.; Fiori, K. W.; Kim, M.; Du Bois, J. J. Am. Chem. Soc. 2004,
126, 15378-15379.
9
J. AM. CHEM. SOC. VOL. 129, NO. 32, 2007 9967

A R T I C L E S
Scheme 1a
Fleming et al.
a Reagents and conditions: (a) PMBCl, K2CO3, nBu4NI, 71%; (b) Tf2O, C5H5N; (c) NaN3, DMF, -15 °C, 8% (two steps); (d) (NH4)2Ce(NO3)6, tBuOH,
55°C,92%;(e)NaOtBu,MbsNdCCl2,CH2Cl2;then(Me3Si)2NH,63%;(f)H2O,CH3CN,65°C,70%;(g)Me3P,H2O/THF;(h)MbsNdC(SMe)NHCH2CH2CHdCH2,
AgNO3, 80% (two steps).
Scheme 2 a
Figure 8. Urea analog also fails to provide nine-membered ring through
metathesis reaction.
a Reagents and conditions: (a) Pd/CaCO3/Pb, quinoline, THF, H2; (b)
NaN3, nBu4NI, DMF, 80% (two steps); (c) PMBCl, K2CO3, nBu4NI, 80%;
(d) Me3P, H2O/THF; (e) MbsNdC(Cl)SMe 24, iPr2NEt, CH3CN, 70% (two
steps).
alkylation of 22 under mild base treatment (K2CO3, 80%).
Reduction of azide 23 using Me3P followed by treatment of
the intermediate amine with an imidoyl chloride reagent,
MbsN)C(Cl)SMe 24, cleanly furnished isothiourea 25 (70%,
Figure 9. Cyclic guanidine formation via carbodiimide addition.
two steps). Rapoport has explored the general use of imidoyl
chloride electrophiles such as 24 for guanidine synthesis,
although this particular derivative had not been previously
prepared.38 Despite the sparing use of such compounds in the
literature, 24 and a related dichloro-form (Vide infra) have been
of exceptional service in our STX effort. In addition, and as
will be highlighted in a later section, the p-methoxybenzene-
sulfonyl (Mbs) group proved ideal for protecting the polar
guanidine residues.
Figure 10. Stereoselective alkynylzinc addition furnishes a highly versatile
intermediate.
Access to isothiourea 25 now possible, replacement of the
C6 alcohol with the requisite amine unit remained the final task
prior to attempting nine-membered ring closure. The added need
to invert the stereochemistry at this C6 position suggested a
sequence involving alcohol activation (Tf2O or CF3CH2SO2Cl,
pyridine) and displacement with NaN3. Attempts to employ other
activating groups (i.e., mesylate, tosylate) failed, as such
substrates were found to be unacceptable for the SN2 reaction.
Although a solution involving triflate formation ultimately
proved successful, our initial attempts to install the C6 azide
revealed an unexpected roadblock. Through earlier routes that
lacked PMB-protection of the oxathiazinane nitrogen, we learned
that activation of the C6 alcohol and treatment with azide (or
other mild bases) resulted in intramolecular displacement to give
the unusual bicyclic aziridine 28 (Scheme 3). This material has
been isolated and characterized; however, under most reaction
conditions, nucleophilic opening at the allylic center ensued to
afford the five-membered sulfamidate product 29. Our desire
to avoid the introduction of an oxathiazinane NH protecting
group caused us to examine many different reaction conditions
for introducing the C6 amine; unfortunately, all attempts were
complicated by the rather remarkable facility by which the
aziridine forms. In the final analysis, blocking the NH position
allowed this problem to be circumvented (Scheme 4). Notably,
Capitalizing on earlier work from our lab, oxathiazinane N,O-
acetal 10 was transformed to the corresponding alkyne derivative
in a highly stereoselective addition process (>20:1 ds, Figure
10).26a Following an optimized protocol, more than 10 g of 21
was synthesized in a single pass without recourse to chroma-
tography. X-ray analysis of this crystalline material confirmed
the desired cis-stereochemistry between C5 and C6 (STX
numbering). The sense of induction in this reaction is consistent
with directed attack of the alkynyl nucleophile by the C6
heteroatom. Optically pure 21, obtained in just three steps from
commercial material, affords an extremely versatile building
block from which to elaborate 4.
To forward our synthetic plan, transforming the C10 tosylate
to an isothiourea was considered the most direct and practicable
means to enact nine-membered ring closure. Azide displacement
of the C6 alcohol could provide the necessary amine nucleophile
for this intramolecular condensation. Accordingly, partial
hydrogenation of oxathiazinane 21 was accomplished using
Lindlar’s catalyst (Scheme 2). The efficiency of this process
was sufficiently high that subsequent azide displacement of the
1° tosylate was performed on the unpurified material (80%, two
steps). Our exploratory studies aimed at trying to manipulate
22, and in particular the C6 hydroxyl center, made evident the
need to first block the NH moiety of the oxathiazinane nucleus
(Vide infra). For this purpose, the p-methoxybenzyl (PMB)
protecting group was chosen and could be easily installed by
(38) Bosin, T. R.; Hanson, R. N.; Rodricks, J. V.; Simpson, R. A.; Rapoport,
H. J. Org. Chem. 1973, 38, 1591-1600, and references therein.
9
9968 J. AM. CHEM. SOC. VOL. 129, NO. 32, 2007

Stereoselective Synthesis of (+)-Saxitoxin
A R T I C L E S
a
Scheme 3. An Unexpected Ring Contraction
Figure 11. Oxathiazinane N-guanidinylation facilitates subsequent ring
opening with H2O.
a Reagents and conditions: (a) CF3CH2SO2Cl, C5H5N, DMAP, CH2Cl2,
0 °C, 89%; (b) NaN3, DMF, 0 °C, 70%; R ) MbsNdC(SMe).
Scheme 4 a
Figure 12. Efficient nine-membered ring formation from an acyclic starting
material.
with (Me3Si)2NH afforded the desired product in 70% yield
(+20% of recovered 31). By activating the oxathiazinane
heterocycle in this way, subsequent hydrolysis is greatly
facilitated and acyclic alcohol 37 can be obtained. It is
interesting to note that this rather uncomplicated structure, 37,
contains all of the carbons found in the tricyclic core of STX.
a Reagents and conditions: (a) Tf2O, C5H5N, DMAP, CH2Cl2, 0 °C; (b)
The decision to prepare alcohol 37 proved advantageous, as
cyclization of this intermediate to the desired nine-membered
guanidine was considerably more effective than prior attempts
to close 32 (Figure 12). In this instance, azide reduction with
Me3P was optimal over other methods tested (e.g., SnCl2).
Treatment of the unpurified amine with AgNO3 triggered
carbodiimide formation and ring closure in 65% yield for the
two-step process. Unlike its predecessor 34, this monocyclic
compound 39 was stable to purification. With the improved
performance of this cyclization reaction, and the ability to
prepare gram quantities of the nine-membered guanidine 39,
the final elements of the synthetic plan could now be tried.
t
NaN3, DMF, -15 °C, (70%, two steps); (c) (NH4)2Ce(NO3)6, BuOH,
CH2Cl2, 74%; (d) SnCl2, THF, MeOH; (e) AgNO3, iPr2NEt, CH3CN, 40%
(two steps).
our completed synthesis of STX relies exclusively on this lone
p-methoxybenzyl and just two p-methoxybenzenesulfonyl pro-
tecting groups.
Successful introduction of the C6 azide rendered the PMB
group unnecessary, and therefore its removal was effected under
oxidative conditions (ceric ammonium nitrate, 74%, Scheme
4).39 Closure of the nine-membered guanidine ring could now
be attempted from the oxathiazinane intermediate 31. Initial
reduction of the azide using SnCl2 was effective for preparing
the C6 amine without recourse to purification. Under highly
optimized conditions, treatment of this compound with AgNO3
and iPr2NEt provided the sought-after bicycle 34 in 40% yield.
This material was unstable to chromatography and storage,
presumably due to strain-promoted ring opening of the fused
oxathiazinane. The strain induced by the trans-substituted
oxathiazinane ring may also have a deleterious influence on
the cyclization reaction itself. Given our inability to optimize
this process, we decided to investigate nine-membered ring
formation from an acyclic starting material (e.g., 4, see Figure
9). Such an intermediate could be prepared by first converting
oxathiazinane 31 to the Mbs-guanidine 36 (Figure 11). This
unusual guanidinylation reaction was made possible with the
use of a second imidoyl chloride, MbsNdCCl2 35.40,41 Initial
condensation of 31 with MbsNdCCl2 followed by quenching
Transforming the Monocyclic Guanidine to STX. Conver-
sion of a nine-membered ring guanidine such as 39 to the
tricyclic core of STX is a problem laden with potential
complications (Figure 13). In the natural product, both C4 and
C12 centers reside at the ketone oxidation level. The direct six-
electron oxidation of an alkene to a diketone is, however, a
reaction with very little precedent.42,43 Moreover, it is likely
that any method for transforming an alkene to an R-dione would
proceed through intermediate species of varying oxidation states
(i.e., 1,2-diol, R-hydroxy ketone). Selectivity in the formation
of a C4,C12-hydroxy ketone 40 is essential, as the regioisomeric
structure 42 has the potential for transannular collapse to give
an undesired bicyclic product 43. In principle, the isomeric
bicycle could be made to equilibrate to the proper form, but
such a reaction would require ring opening back to the nine-
(41) Compound 35 was prepared following a published protocol, see: Tanga,
M. J.; Bradford, W. W.; Bupp, J. E.; Kozocas, J. A. J. Heterocycl. Chem.
2003, 40, 569-573.
(39) To our knowledge, p-methoxybenzyl cleavage of an N-substituted oxathi-
azinane heterocycle or sulfamate ester has not been reported. For depro-
tection of an N-PMB amide, carbamate, and thiazolidinone, see respec-
tively: (a) Williams, R. M.; Armstrong, R. W.; Dung, J. S. J. Am. Chem.
Soc. 1984, 106, 5748-5750. (b) Alcon, M.; Moyano, A.; Pericas, M. A.;
Riera, A. Tetrahedron: Asymmetry 1999, 10, 4639-4651. (c) Smith, A.
B.; Leahy, J. W.; Noda, I.; Remiszewski, S. W.; Liverton, N. J.; Zibuck,
R. J. Am. Chem. Soc. 1992, 114, 2995-3007.
(40) For the general preparation of reagents such as 35, see: (a) Neidlein, R.;
Haussman, W. Tetrahedron Lett. 1965, 1753-1755. (b) Gompper, R.; Kunz,
R. Chem. Ber. Recl. 1966, 99, 2900-2904. (c) Merchan, F. L.; Garin, J.;
Tejero, T. Synthesis 1982, 984-986.
(42) For examples, see: (a) Sharpless, K. B.; Lauer, R. F.; Repic, O.; Teranishi,
A. Y.; Williams, D. R. J. Am. Chem. Soc. 1971, 93, 3303-3304. (b)
Yusubov, M. S.; Filimonov. V. D.; Vasilyeva, V. P.; Chi, K.-W. Synthesis
1995, 1234-1236. (c) Yusubov, M. S.; Krasnokutskaya, E. A.; Vasilyeva,
V. P.; Filimonov, V. D.; Chi, K.-W. Bull. Korean Chem. Soc. 1995, 16,
86-88. (d) Clayton, M. D.; Marcinow, Z.; Rabideau, P. W. Tetrahedron
Lett. 1998, 39, 9127-9130.
(43) Methods for alkene ketohydroxylation have been recently reviewed, see:
Plietker, B. Tetrahedron: Asymmetry 2005, 16, 3453-3459. Also, see:
Plietker, B. Synthesis 2005, 2453-2472.
9
J. AM. CHEM. SOC. VOL. 129, NO. 32, 2007 9969

A R T I C L E S
Fleming et al.
Figure 13. Potential complications associated with C4,C12-alkene oxidation. DFT analysis (B3LYP/6-31G*) suggests that the desired isomer 41 is
thermodynamically preferred. Calculations were performed on all four diastereomers of each configurational isomer (i.e., 41 and 43). The three lowest
energy structures are shown (R ) C(O)NH2), see Supporting Information for details.
Figure 14. Regioselective alkene ketohydroxylation gives the desired
Figure 15. Ru-catalyzed oxidation of 45 affords a mixture of hemiaminal
hemiaminal 44.
products.
As our initial experiments would indicate, the absence of the
membered guanidine. Also, relying on thermodynamic equili-
oxathiazinane ring in guanidine 39 jeopardizes the ability to
bration between structures 41 and 43 presupposes that an
control isomer formation in the transannular ring-closing event.
energetic difference between these two compounds exists and
Treatment of 45 with catalytic RuCl3 and Oxone furnished both
the desired hemiaminal 46 and a second product tentatively
is biased toward the requisite product 41. In this regard, density
functional theory (DFT) calculations (B3LYP/6-31G*) do, in
fact, confirm that the desired bicyclic isomer 41 is strongly
preferred, irrespective of the stereochemical consequences of
assigned as the undesired isomer 47 (∼3:1 selectivity, Figure
15).47,48 More troubling than the low product selectivity,
however, was the poor overall yield of this reaction (30%). In
the ketohydroxylation reaction (Figure 13).44
spite of these results, a sufficient quantity of the bicyclic
It bears mentioning that the potential problems surrounding
compound 46 was obtained to enable forward progress. To
the oxidation of the monocyclic guanidine 39 guided our earliest
complete the synthesis of the STX core, 46 was exposed to
approach to STX. With bicyclic oxathiazinane 34, we surmised
that generation of the undesired hemiaminal product following
alkene oxidation would be disfavored due to the strain energy
p-TsOH, conditions that we anticipated would induce iminium
ion formation and ring-closure of the five-membered guanidine.
This reaction did indeed furnish a new product (Figure 16), but
associated with forming a fused trans-5,6 ring system (Figure
a definitive structural assignment of this material was not
14). This expectation was indeed realized, as oxidation of 34
possible in the absence of unambiguous coupling constant data.
under conditions described by Plietker (2 mol % RuCl3, NaIO4,
Thus, we advanced under the assumption that formation of the
H2SO4) furnished only the tricyclic compound 44, albeit in a
desired N,N-aminal at C4 had occurred in the expected fashion,
rather modest 35% yield.45 Compound 44 is produced as a single
conducting the subsequent oxidation of the C12 alcohol with
diastereomer, the structure of which was confirmed by X-ray
C5H5N·SO3 in DMSO.49 This transformation proved quite
analysis.46 Our excitement with having prepared 44 was,
revealing, as it became clear that the bond connectivity in the
unfortunately, tempered by the suboptimal yields for both the
isolated material 48 was different from that of the natural
nine-membered guanidine assembly (Vide supra) and this Ru-
product. Perhaps the most telling discrepancy between the two
catalyzed four-electron oxidation. Accordingly, focus turned
structures was the C5,C6 coupling constant, measured at 3.4
toward ketohydroxylation of the more readily accessible mono-
Hz for 48 and ∼1 Hz for decarbamoylsaxitoxin (dcSTX).16a
cyclic compound, 39.
(47) Oxone has been shown to be a more effective oxidant than NaIO4 for alkene
(44) (a) Calculations performed using Gaussian 03, Revision C.02, Frisch, M.
J. et al. Gaussian, Inc., Wallingford, CT, 2004. (b) Becke, A. D. J. Chem.
Phys. 1993, 98, 5648-5652.
ketohydroxylation, see: (a) Plietker, B. J. Org. Chem. 2003, 68, 7123-
7125. (b) Plietker, B. J. Org. Chem. 2004, 69, 8287-8296. (c) Plietker, B.
Eur. J. Org. Chem. 2005, 1919-1929.
(45) The conditions employed were analogous to those described for olefin
dihydroxylation, see: (a) Plietker, B.; Niggemann, M. Org. Lett. 2003, 5,
3353-3356. (b) Plietker, B.; Niggemann, M.; Pollrich, A. Org. Biomol.
Chem. 2004, 2, 1116-1124.
(48) The broadened signals in the 1H NMR spectrum of this product made
difficult an accurate determination of the relative amount of the minor
isomer.
(49) The ketone was isolated in the hydrated form as determined by IR and 13
C
(46) X-ray analysis was performed on the p-toluenesulfonyl-protected form of
44. The tosyl-protecting group was employed in our earliest studies to
prepare STX and later replaced with p-methoxybenzenesulfonyl (Mbs).
NMR. For examples of guanidine-derived structures possessing a hydrated
ketone, see: Houghten, R. A.; Simpson, R. A.; Hanson, R. N.; Rapoport,
H. J. Org. Chem. 1979, 44, 4536-4543.
9
9970 J. AM. CHEM. SOC. VOL. 129, NO. 32, 2007

Stereoselective Synthesis of (+)-Saxitoxin
A R T I C L E S
Scheme 5 a
Figure 16. An attempt to form the tricyclic core of STX gives an undesired
isomeric structure.
i
a Reagents and conditions: (a) Me3P, H2O/THF; (b) AgNO3, Pr2NEt,
CH3CN, 65% (two steps); (c) 20 mol % OsO4, NMO, DABCO, H2O,
tBuOH/acetone, 84%; (d) DMP, CH2Cl2, 85% (50% 53, 35% 54).
Figure 17. Acid treatment of 48 does not afford decarbamoyl STX.
Figure 18. Acid-promoted ring closure gives an unstable tricyclic product
55, which reverts to starting material 53 upon standing.
Although mass spectral data could not distinguish between the
two potential isomeric forms, additional NMR analysis, includ-
1
ing H-13C heteronuclear multiple bond correlation (HMBC),
conditions (TsOH), and thus we speculate that the selectivity
observed for 53 (albeit small) is kinetic in origin (Figure 18).53
While a method for selective diol oxidation would be needed
to improve material throughput in this step, we were quite
pleased to find that treatment of the correct hemiaminal 53 with
TsOH resulted in ring closure to the tricyclic frame of STX 55.
Thus, our strategy for preparing STX through late-stage dehy-
drative aminal formation appeared viable. Rather surprisingly,
however, it was not possible to further manipulate tricycle 55,
as it reverted back to starting material 53 upon attempted
purification. Given the known stability of â-STXol (a dication),
the Mbs guanidine protecting groups in 55 were presumed to
be responsible for disfavoring the equilibrium between 53 and
55. Our next move, it therefore seemed, was to identify reaction
conditions that would induce ring closure and simultaneous
guanidine deprotection without cleaving the C13 benzyl ether.
Prior success at deblocking the two Mbs-guanidines in 48
with CH3SO3H (see Figure 17) prompted our decision to test
this acid for the single-step conversion of 53 to decarbamoyl
â-saxitoxinol (â-dcSTXol). The strength of CH3SO3H in CH2-
Cl2 is rather remarkable and, while some of the desired product
was indeed obtained, an equivalent amount of N,O-acetal 56
was generated (Figure 19). Assuming that 56 forms irreversibly,
the observed product ratio suggests that the relative rate of
benzyl ether acidolysis and N,O-acetal formation is comparable
to that of ring closure and guanidine deprotection. Thus, in an
attempt to alter favorably the course of this reaction, Lewis
acidic conditions were explored. Prior work by Pless, Bauer,
and Nishimura suggested to us that boron tris(trifluoroacetate)
(B(O2CCF3)3) might be a uniquely effective reagent for
revealed that the C13-hydrogens were within three contiguous
bonds of the C4 center, suggesting that N,O-acetal 48 had
formed instead of the desired N,N-aminal.50 We speculated that
the Mbs-protecting groups on both guanidine units might be
responsible for the unexpected preference of the N,O-acetal
linkage; accordingly, their removal was smoothly effected using
CH3SO3H and thioanisole.51 This transformation, however,
failed to promote the anticipated isomerization reaction of the
recalcitrant N,O-acetal 48 (Figure 17); thus, a new tactic was
needed to facilitate the desired cyclization pathway.
Blocking the C13-hydroxyl moiety with an acid-stable
protecting group presented the most straightforward means for
circumventing the formation of N,O-acetal 48. Fortunately, our
synthetic plan to the nine-membered ring guanidine was
sufficiently flexible to allow for facile introduction of an ether
or ester group at C13. Benzyl ether 50 could be prepared by
direct displacement of oxathiazinane 36 with BnOH (Scheme
5), and by following our earlier protocol (Me3P, then AgNO3),
the nine-membered ring olefin 51 was readily assembled.
Ketohydroxylation of 51, however, using catalytic RuCl3 and
Oxone lead only to intractable material, possibly owing to
competitive oxidation of the benzylic unit. Recourse to a two-
step process of dihydroxylation and monooxidation with Dess-
Martin periodinane (DMP) furnished hemiaminal 53 in 50%
yield along with 35% of its structural isomer 54.52 Having
isolated hemiaminal 54 in pure form, the prospect of equilibrat-
ing this compound to the requisite bicycle 53 could now be
discerned. No such reaction was observed under neutral or acidic
(50) Complete HMBC data for compound 48 can be found in the Supporting
Information.
(53) It is not possible to discount a mechanistic scenario in which rapid
tautomerization of the R-ketol isomers precedes transannular ring closure.
We believe, however, that the collective results of our studies to oxidize
substrates such as 51 are more consistent with an oxidation reaction in
which R-ketol selectivity is kinetically controlled.
(51) Structural assignment of intermediate 49, as determined by 1H NMR and
mass spectral analysis, is tentative.
(52) Although the undesired hemiaminal 54 formed as a single diastereomer,
its stereochemistry was never established conclusively.
9
J. AM. CHEM. SOC. VOL. 129, NO. 32, 2007 9971

A R T I C L E S
Fleming et al.
lation of 57 with OsO4/N-methylmorpholine N-oxide (NMO)
proceeded smoothly and furnished a single diol product. The
subsequent reaction with DMP, however, was marred by an
unfavorable isomer product ratio (∼1:2 desired/undesired). The
discernible difference between this reaction and oxidation of
52 (see Scheme 5), while small, is quite surprising in light of
the number of carbon centers separating the C13-O protecting
group and the C4/C12 alkene. Fortunately, this selectivity was
reversed in the one-step ketohydroxylation of 57 with catalytic
RuCl3 and Oxone, conditions that afforded a 2.3:1 product
mixture favoring 58 (entry 1, Table 1). The combined yield for
this reaction, however, could never be made to exceed 40%
(entry 2). Determined to solve this problem in a direct fashion,
efforts focused on improving the one-step, four-electron conver-
sion of 57 to the desired hemiaminal 58. The low yield and
mass recovery for the Ru-catalyzed oxidation was likely due to
competitive C-C bond cleavage of the intermediate Ru-
glycolate. Changing the oxidant and pH of the solution did little
to improve the reaction performance. The switch to an Os
catalyst in place of Ru, thus, followed as a logical progression.56
Direct conversion of an olefin to an R-ketol using OsO4 can
be traced as far back as 1942 in seminal work from Prins and
Richstein.57 Later studies at Upjohn would demonstrate more
fully the power of this transformation in the context of
modifying unsaturated steroidal systems.58 This same group and
others have noted that Os-catalyzed dihydroxylation using H2O2
Figure 19. Guanidine deprotection affords decarbamoyl â-saxitoxinol.
transforming 53 to â-dcSTXol.54 Although milder than BBr3,
B(O2CCF3)3 has been used to deprotect Mbs-guanidines.54b Such
a Lewis acid would undoubtedly cleave the C13 benzyl ether
as well, but here it was assumed that the intermediate alkoxy
boron species would be incapable of closing to form the
unwanted N,O-acetal 56. When put to practice, B(O2CCF3)3
exceeded all expectations, affording â-dcSTXol in >80% yield
and giving no trace of the isomeric structure 56 in the 1H NMR
spectrum of the unpurified reaction mixture. Now only two steps
remained to complete the natural product. Selective carbamoy-
lation, however, of the C13 alcohol with the 2° alcohol free at
C12 would be necessary to accomplish this feat. The highly
polar nature and poor solubility of the bis-guanidinium salt in
organic solvents further complicated this problem.
As a consequence of our inability to manipulate â-dcSTXol,
an alternative end-game strategy was devised that would attempt
the critical Mbs-deprotection/guanidine cyclization step with the
C13-carbamate in place. The reported functional group compat-
ibility of B(O2CCF3)3 toward carboxylic esters encouraged us
to examine such a strategy.54 Moreover, such a plan, if
successful, would afford a most expeditious path from 39 to
STX.
Alkene Ketohydroxylation: An Optimal Solution. Conver-
sion of the C13-alcohol in 39 to the requisite 1° carbamate was
readily accomplished using trichloroacetyl isocyanate (eq 1).55
It should be noted that in both the Kishi and Jacobi routes to
STX, carbamate installation was conducted as a modest-yielding,
final step on dcSTX, a bis-guanidiniun salt that is insoluble in
most solvents.4,5 With the guanidines suitably masked, the ability
to synthesize, isolate, and purify carbamate 57 is facile and, as
an added benefit, straightforward access to carbamate derivatives
of STX also becomes possible.
t
or BuOOH gives varying amounts of the vicinal keto-alcohol
as a side product.59 To our knowledge, however, no systematic
investigation of OsO4 for ketohydroxylation of nonsteroidal
olefins has been reported. By employing a combination of 8
mol % OsO4 and tBuOOH, efficient four-electron oxidation of
alkene 57 was achieved (70%); regrettably, it soon became clear
that the product had formed exclusively as the undesired
hemiaminal 59 (entry 3). Buoyed by earlier data from our studies
with Ru-catalysts, which had shown that oxidant and base
additives could alter selectivity in the ketohydroxylation event,
we continued to examine this unique reaction. Further guided
by a report by Murahashi describing alkene ketohydroxylation,
we turned to OsCl3‚H2O as a catalyst in combination with
different terminal oxidants including Oxone.60 Rather astonish-
ingly, treatment of 57 with a mixture of 10 mol % OsCl3‚H2O,
Oxone, and NaHCO3 reversed the reaction outcome and favored
the wanted bicycle 58 as a 5:1 isomeric mixture along with a
substantial fraction of diol 60 (entry 4). The amount of this latter
product was sensitive to the base employed, and the addition
of excess Na2CO3 in lieu of NaHCO3 almost completely
suppressed its formation. Subsequent optimization studies
revealed conditions in which hemiaminal 58 could be obtained
in good yield (57%) and with high regio- (12:1) and stereo-
control (>20:1) using 10 mol % OsCl3‚H2O, 10 equiv of Na2-
(56) For a discussion of C-C bond cleavage by Ru and Os oxidants, see:
Frunzke, J.; Loschen, C.; Frenking, G. J. Am. Chem. Soc. 2004, 126, 3642-
3652.
(57) (a) Prins, D. A; Reichstein, T. HelV. Chim. Acta 1942, 25, 300-322. (b)
Miescher, K. HelV. Chim. Acta 1950, 6, 1840-1847. (c) Miescher, K.;
Schmidlin, R.; Schmidlin, J. U.S. Patent 2,668,816, 1954.
(58) Harkema, J. U.S. Patent 3,582,270, June 1, 1971, and references cited
therein.
(59) (a) VanRheenen, V.; Kelly, R. C.; Cha, D. Y. Tetrahedron Lett. 1976, 23,
1973-1976. (b) Lohray, B. B.; Bhushan, V.; Kumar, R. K. J. Org. Chem.
1994, 6, 1375-1380.
(60) Murahashi, S. I.; Naota, T.; Hanaoka, H. Chem. Lett. 1993, 10, 1767-
1770. Murahashi reports the use of OsCl3 with peracetic acid in EtOAc
for R-ketol formation. We did not investigate these specific conditions due
to the inability to obtain a commercial supply of peracetic acid in EtOAc.
Initial attempts to execute the critical ketohydroxylation of
alkene 57 would capitalize on a two-step sequence developed
in our earlier studies (see Scheme 5). Accordingly, dihydroxy-
(54) (a) Pless, J.; Bauer, W. Angew. Chem., Int. Ed. 1973, 12, 147-148. (b)
Nishimura, O.; Fujino, M. Chem. Pharm. Bull. 1976, 24, 1568-1575.
(55) Kocovsky, P. Tetrahedron Lett. 1986, 27, 5521-5524.
9
9972 J. AM. CHEM. SOC. VOL. 129, NO. 32, 2007

Stereoselective Synthesis of (+)-Saxitoxin
A R T I C L E S
Table 1. Reaction Conditions Influence Alkene Ketohydroxylation
none of the undesired N,O-acetal product 56 that had plagued
our earlier routes. Facile isolation of the pure bis-hydrochloride
salt of â-STXol was easily accomplished using ion-exchange
chromatography.
The final transformation in the path to (+)-STX relied on a me-
thod reported by Schantz et al. for oxidation of R-STXol‚2Cl-
under Pfitzner-Moffat conditions (DCC, DMSO, C5H5N‚HO2-
CCF3).16a,62 We were optimistic that such a reaction would
perform successfully on â-STXol, as the three-dimensional
arrangement of the STX skeleton leaves the â-OH group more
accessible than its R-counterpart. By employing this protocol,
1
high conversion (>90% as determined by H NMR analysis)
of the starting material to STX was successfully accomplished,
and the natural product was isolated in 70% yield. Initially,
purification of STX‚2Cl- by reverse phase HPLC was quite
difficult to reproduce as a result of its extremely short retention
time (<1 min) on a C18 column. We later discovered that an
eluent system composed of CH3CN, H2O, and 10 mM hep-
tafluorobutyric acid dramatically extended HPLC tR (13.0 min
@ 6 mL/min, C18, 10 × 250 mm), thereby providing highly
pure samples of the target as the bis-perfluorobutyrate salt
(C3F7CO2-). The synthetic material matched the physical data
reported for natural (+)-STX‚2Cl- in all respects (1H and 13C
NMR, HRMS). In addition, an IC50 of 2.3 nM was measured
against heterologously expressed NaV1.4 channels (CHO cells),
a value consistent with literature reports.63
a Reactions performed with 2-10 mol % catalyst at 25 °C. b Product
ratio determined by HPLC analysis. c %Yield is combined for 58 and 59.
d Reaction performed at 10 °C.
Enantioselective preparation of the paralytic shellfish poison,
(+)-STX, has been accomplished through a 19-step sequence
and in 1.3% overall yield from commercially available (R)-
glycerol acetonide. Critical to the success of this program was
the discovery of a new class of heterocyclic iminium ion
surrogates for the rapid assembly of highly functionalized
propargylic amine derivatives. The intermediate alkynyl ox-
athiazinane 21 (Figure 10) proved an exceptionally versatile
substructure, allowing us to explore and evaluate disparate
strategies for assembling an unprecedented nine-membered ring
guanidine. These findings add to the general methods that we
have described for the synthesis and selective manipulation of
oxathiazinane N,O-acetals.26 When combined with the demon-
strated utility of oxathiazinanes for crafting substituted amine
products, it is expected that the value of such chemistries will
extend beyond the work presented herein. Nevertheless, with
hindsight, we recognized that a preparative route to an acyclic
target such as 38 (Figure 12) could be conducted with greater
step-efficiency by employing alternative protocols for vicinal
diamine synthesis. Propelled by an overarching interest to
develop STX-like molecules for investigating the structure and
integrated function of Na+ channels in neuronal circuits, a
second-generation approach to the nine-membered ring guani-
dine 39 was conceived. Successful implementation of this new
strategy thus reduces the assembly of (+)-STX to only 14 linear
steps. The details of this work are outlined below.
Figure 20. Completed first-generation synthesis of the desired target.
CO3, and 7 equiv of Oxone (entry 5). The fact that this method
delivered the oxidized product 58 out of a mixture of eight
possible isomers with selectivity completely opposite that of
the OsO4/tBuOOH combination deserves special comment. As
a final note, OsCl3 can be substituted with RuCl3, but not without
a cost to both product yield and selectivity (entry 6).
The four-electron oxidation of alkene 57 would ultimately
distinguish itself as one of the most intriguing and critical
transformations in our completed synthesis of STX. The side-
chain element at C13, catalyst, oxidant, and base all transpire
to influence the reaction outcome.61 Optimization of this unique
oxidation process at such a late stage underscores the efficiency
of the synthetic route to this point.
The Completed Synthesis. With access to hundreds of
milligrams of hemiaminal 58 as a single structural and stere-
oisomer, completion of the STX synthesis now seemed at hand.
Conditions previously developed for the deprotection and
cyclization of benzyl ether 53 (Figure 19) were anticipated to
function with 58, although the stability of the carbamate side
chain to B(O2CCF3)3 was uncertain. These concerns were
quickly assuaged when exposure of the bicyclic starting ma-
terial to 30 equiv of B(O2CCF3)3 in trifluoroacetic acid pro-
vided the known compound, â-STXol, in 82% yield (Figure
20). For this step, the unpurified reaction mixture was excep-
A Second Generation Route to (+)-Saxitoxin. A retrospec-
tive analysis of our initial route to STX underscores the power
of the cyclodehydrative strategy for quickly (four steps) ac-
cessing the natural product from a monocyclic guanidine
precursor 39 (Figure 21). The preparation of 39 itself, however,
requires 15 unique transformations from commercial material,
1
tionally clean, as judged by H NMR analysis, and revealed
(61) We speculate that the combination of OsCl3 and Oxone leads to the
formation of an Os(V)dO species, which serves as the operative oxidant.
The pendant guanidine at C5 may have a critical role as a base to promote
selective elimination from the C4 center. Studies are on-going to explore
the mechanistic details of this process and will be reported in due course.
(62) Pfitzner, K. E.; Moffatt, J. G. J. Am. Chem. Soc. 1965, 87, 5661-5670.
(63) Moran, O.; Picollo, A.; Conti, F. Biophys. J. 2003, 5, 2999-3006.
9
J. AM. CHEM. SOC. VOL. 129, NO. 32, 2007 9973

A R T I C L E S
Fleming et al.
Scheme 6. (+)-Saxitoxin in 14 Steps from N-Boc-L-Serine Methyl Estera
a Reagents and conditions: (a) tBuPh2SiCl, imidazole, DMF, 95%; (b) iBu2AlH, CH2Cl2, 71%; (c) PMBNHOH, MgSO4, CH2Cl2, 76%; (d)
i
MbsNdC(SMe)NHCH2CH2CtCH 63, PrMgCl, THF, -78 °C, 78%; (e) p-TsNHNH2, NaOAc, THF, H2O, 100 °C, 78%; (f) Zn, Cu(OAc)2, HOAc, H2O,
70 °C, 81%; (g) MbsNdC(SMe)NHBoc 65, HgCl2, Et3N, CH2Cl2, 74%; (h) HCl, MeOH, 52%; (i) AgNO3, Et3N, CH3CN, 73%; (j) CF3CO2H, 60 °C, 91%.
from 65 was effected using HgCl2 and Et3N, which reacts
efficiently with the PMB-amine to afford the N7-guanidine
t
(74%, STX numbering). Removal of both N-Boc and BuPh2-
Si-protecting groups using methanolic HCl gave an acyclic
structure that was suitably configured for closure to the nine-
membered ring. Although attempts were made to cleave the
p-methoxybenzyl moiety in addition to the other three protecting
groups by employing reagents such as cerium ammonium nitrate
Figure 21. Step-count analysis of the finished route to saxitoxin.
or CF3CO2H/∆, yields for this process were generally poor.
Accordingly, cyclization of the intermediate C6-amine to
guanidine 66 was first executed (AgNO3, Et3N, 73%) followed
by removal of the PMB group with CF3CO2H. The isolated
compound 39 intercepts our initial STX preparation in just 10
linear operations from N-Boc-L-serine methyl ester and is easily
processed to the optically pure natural product in four steps.
This new route has enabled multigram production of the nine-
a somewhat lengthy operation for a structure containing only
two stereogenic centers. It is worth recalling that the initial role
of the oxathiazinane, which ultimately allowed for the assembly
of 39, was not only as a synthetic intermediate but as a structural
restraint to bias formation of the correct hemiaminal isomer in
the ketohydroxylation/transannular cyclization event. Having
established the viability of oxidizing regioselectively the mono-
cylic guanidine 39 without imposed constraints, the plan for
membered guanidine and has greatly facilitated access to
modified forms of the toxin.
synthesizing this material bears reconsidering. To this end, our
attention focused on isothiourea 38 (or a closely related
structure) given the success of the subsequent carbodiimide ring-
Conclusion
closure to fashion the nine-membered ring 39. Borrowing from
earlier work of Merino, addition of an organometallic nucleo-
phile to the nitrone derived from L-serine 62 appeared aptly
suited for achieving a more direct means to the desired
intermediate (Scheme 6).64
The completed asymmetric synthesis of the guanidinium
poison, (+)-STX, marks a significant first step in our effort to
understand the complex structure and function of vg-Na+
channels using novel pharmacological tools made available
through de novo synthesis. A short, modular route to STX was
deemed essential to accomplish these longer-term goals and,
through the work described herein, such a recipe for assembling
the toxin is now in place. Through these studies, new chemical
tools and strategies for crafting highly polar, nitrogen-rich small
molecules were delineated, some of which should enjoy
application beyond the STX assembly. In particular, the
development of unique heterocyclic N,O-acetals as iminium ion
equivalents expands greatly the ability to employ C-H ami-
nation for accessing highly functionalized amine derivatives.
Other highlights of this work include the first reported pre-
paration of a nine-membered ring guanidine, catalytic ketohy-
droxylation methods for regioselective alkene oxidation, the
application of B(O2CCF3)3 to deblock Mbs-guanidine
groups, stereoselective aminal formation through cyclodehy-
dration, and acetylide dianion-nitrone addition to generate
polyfunctionalized vicinal diamines. The combination of
these tactics has reduced a complicated problem in cyclic
stereocontrolled synthesis to one of assembling a heteroatom-
Serine-derived aldonitrones analogous to 62 undergo smooth
attack by a variety of simple organomagnesium and organo-
lithium compounds to furnish hydroxylamine products with good
to excellent levels of anti selectivity. In our minds, an ideal
path to an intermediate such as 38 could make use of 62 to
assemble 64 in a single step that would import the entire
isothiourea substructure via addition of the corresponding metal
acetylide. Importantly, preparation of nitrone 62 could be
accomplished from commercially available N-Boc-L-serine
methyl ester in just three transformations.65 Although additions
of dianionic nucleophiles to this starting material are unprec-
edented, the reaction of the Mg salt of 63 with 62 proceeded
nearly quantitatively to afford the product as a 4-5:1 mixture
of diastereomers favoring the anti-stereoisomer. Following
purification, the desired hydroxylamine 64 was isolated in 78%
yield. This dianion coupling step highlights our second genera-
tion route of STX and eliminates all functional group manipula-
tions associated with assembling the isothiourea moiety from
an alkyl tosylate (Vide supra).66
Partial reduction of alkyne 64 using diimide and subsequent
Zn-induced N-O bond cleavage proceeded without event and
gave an intermediate 2° PMB-amine onto which the first
guanidine unit could be installed. In this instance, a differentially
protected isothiourea reagent 65 was prepared to facilitate the
coupling reaction. In situ generation of the carbodiimide derived
(64) (a) Merino, P.; Franco, S.; Merchan, F. L.; Tejero, T. Synlett 2000, 442-
454, and references cited therein. (b) Merino, P.; Franco, S.; Merchan, F.
L.; Tejero, T. J. Org. Chem. 1998, 16, 5627-5630.
(65) Merino, P.; Lanaspa, A.; Merchan, F. L.; Tejero, T. Tetrahedron:
Asymmetry 1998, 4, 629-646.
(66) All attempts to add the zinc acetylide of 63 to oxathiazinane 10 (see Figure
10) were unsuccessful.
9
9974 J. AM. CHEM. SOC. VOL. 129, NO. 32, 2007

Stereoselective Synthesis of (+)-Saxitoxin
A R T I C L E S
substituted cis-heptene derivative. As such, the finished route
to STX requires only 14 linear transformations, affording the
natural product with unprecedented efficiency and em-
powering efforts to construct unique, non-natural toxin designs.
support of this work. J.J.F. and M.D.M. were the recipients of
an Amgen graduate student fellowship (2002 and 2004) and an
NIH postdoctoral fellowship, respectively.
Supporting Information Available: Complete ref 44a (page
S24, ref 12); experimental procedures and characterization data.
This material is available free of charge via the Internet at
http://pubs.acs.org.
Acknowledgment. Benjamin Brodsky and Brian Andresen
are graciously acknowledged for their assistance with compu-
tational studies and electrophysiology experiments. We are
grateful to the National Institutes of Health, Amgen, Boehringer
Ingelheim, GlaxoSmithKline, Merck, and Pfizer for generous
JA071501O
9
J. AM. CHEM. SOC. VOL. 129, NO. 32, 2007 9975