As part of our synthetic work directed toward glycopeptide mimetics, we required a
suitably protected (2S ,3S )-β-vinylserine (β-VSer) for use as a synthetic building block. Many noncanonical
amino acids have been incorporated into protein and peptide structures to interrogate
various cellular functions.[1 ] In particular, alkenyl amino acids incorporated into peptides have proven to be
useful for peptide stapling by a cross-metathesis reaction to afford conformationally
restricted peptidomimetics.[2 ] In addition, Zhang and van der Donk have examined the effect of direct alkenyl amino
acid incorporation.[3 ] They incorporated a diastereomer of our desired β-VSer (referred to as a threonine
analogue) into a peptide sequence of lacticin synthetase to examine substrate selectivity
toward dehydration reactions. The pentenoic backbone of β-VSer itself is also a common
scaffold for dipeptide isosteres,[4 ] which have been investigated as enzyme inhibitors and as receptor antagonists.[5 ] This platform has also been a versatile synthetic intermediate for preparing sphingomyelin
analogues[6 ] and glycosidase inhibitors such as the deoxynojirimycins.[7 ] It has also served as a building block for antitumor agents such as 2-epi -pachastrissamine[8 ] or for glycopeptide[9 ] and β-lactam antibiotics.[10 ] For our purposes, we sought to elaborate the β-VSer alkene through cross-metathesis
and/or Trost–Tsuji π-allylic alkylation chemistry for the development of novel glycopeptides.
Given the versatility and interest in this simple building block, we elected to exploit
an oxazolidinone scaffold 1 as a β-VSer synthetic equivalent in which both the amine and the hydroxy functions
are simultaneously protected (Scheme [1 ]). Although there are excellent reports on carbamate cyclizations[11 ] and an allylic C–H amination[12 ] that yield trans -4,5-disubstituted oxazolidinones stereospecifically, our studies required a cis -oxazolidinone. cis -4,5-Disubstituted oxazolidinones of this sort are known and are commonly derived
from anti -2-aminopent-4-en-1,3-diols such as 2 .
Scheme 1 Target β-vinylserine (β-VSer) synthetic equivalent 1 and precursor
Both vinyl oxazolidinones and functionalized 2-aminopent-4-en-1,3-diols are valuable
synthetic intermediates that have been used to prepare numerous natural products and
medicinal targets, as discussed above. Although synthetic approaches from carbohydrates,[13 ] azide epoxide openings,[6a ] and chiral glycine enolate aldols[14 ] are available, the more common synthetic approaches entailing nucleophilic additions
to α-amino-β-hydroxy aldehydes or ketones provide varying degrees of control of stereochemistry
(Scheme [2 ]).
Scheme 2 Approaches to cis -oxazolidinones
A survey of the literature indicated one could proceed by a vinyl Grignard addition
onto the well-known d -serine-derived Boc-protected Garner’s aldehyde[15 ] or the OTBS-Boc-serinal 4 ,[7a ]
[10 ] followed by an intramolecular cyclization onto the Boc group to form an oxazolidinone.
The Grignard approach has been widely used,[7b,9,16 ] but is limited due to the selectivity of the Grignard addition; this led Herold
to develop a three-step approach employing trimethylsilyl acetylide additions for
improved anti -stereoselectivity.[17 ] Although the tert -butyl(dimethyl)silyl ether substrate 4 gives 5 directly, it results in an undesirable 1:2 anti /syn diastereomeric ratio.[7a ] The typical anti -selectivity for vinyl addition to Garner’s aldehyde is reported to range from 3:1[16a ] to 6:1 anti /syn , and experimental details indicate that additional purification by chromatography
is necessary. From the Grignard product of Garner’s aldehyde, hydrolysis of the N ,O -acetal and selective protection of the primary hydroxy groups is needed, followed
by formation of the oxazolidinone by a base-induced intramolecular cyclization onto
the tert -butyl carbamate to afford 6 .[18 ] In an improvement to these early approaches, the Weinreb amide 7 of a protected d -serine, available in four steps, has been employed to form an enone upon addition
of vinylmagnesium bromide; this enone can be stereoselectively reduced with Li(t -BuO)3 AlH in ethanol giving 5 with a 10:1 preference toward the anti -diastereomer.[19 ]
Here, we report a highly selective alternative approach in which the N -tosylamide 8 is used as a stereodirecting orthogonal protecting group; this approach is complementary
to the approaches discussed above.
For our purpose, we had concerns about the N -Boc protecting group due to its potential for neighboring-group participation in
our planned synthetic manipulations; we therefore initially desired an N -tosyl protected nitrogen on the oxazolidinone 9 . Although one could simply tosylate the known oxazolidinone 6 to give 9 , we considered initiating our synthesis with the acyclic silyl-protected N -tosyl-d -Ser[20 ] or the N -tosyl equivalent of Garner’s aldehyde.[21 ] Vinyl Grignard additions to N -sulfonyl-protected acyclic amino acids are not usually selective. Literature reports
suggest that additions to the aldehydes of TsNH-Ala[22 ] and TsNH-Phe[23 ] give poor diastereoselectivities (2:3 anti /syn and 2:1 with the major isomer not identified, respectively). Given the poor selectivity
of additions to acyclic amino aldehydes, we opted to pursue the use of a toluenesulfonamide
derivative of Garner’s aldehyde 8 . Surprisingly, no Grignard chemistry has been reported on this aldehyde. We found
that vinylmagnesium bromide added cleanly to give a >95% yield[24 ] (Scheme [3 ]) and was more selective than the N -Boc-protected Garner’s aldehyde, giving the anti -allylic alcohol 10 with an 8.5:1 dr before chromatography. The use of LiCl as an additive in the vinylmagnesium
bromide reaction did not alter the results. Although some trial runs using vinylmagnesium
chloride directly did show >10:1 diastereoselectivity, these seemed highly dependent
on the commercial source and age of the reagent. Conveniently, no rotamers are observed
in the NMR spectra of the tosylamides, unlike the Boc-derivatives, making their interpretation
more straightforward; moreover, TLC visualization and chromatographic detection is
aided by the UV activity of the aromatic sulfonamide.
Scheme 3 Synthesis of β-VSer derivatives 12 and 14 ; pNs = 4-O2 NC6 H4 SO2 .
The improved diastereoselectivity can be partially explained by examining the LUMO
energies of the reactive Felkin–Anh conformations (Scheme [4 ]). With the N -sulfonamide there is a strong preference for the C–NTs bond of 9b to lie perpendicular to the plane defined by the aldehyde carbonyl as opposed to
the C–CH2 O bond in 9a . The LUMO of 9a is 3.46 kcal mol–1 higher in energy than that of 9b , as determined by ground-state gas-phase DFT calculations using an ω-897XD hybrid
GGA functional. This predicts that nucleophilic approach should favor attack on 9b , leading to the 2,3-anti -product. In contrast, the N -Boc derivative has a smaller LUMO energy difference (2.77 kcal mol–1 ) between the two Felkin–Anh conformations, so it would not be expected to be as stereoselectively
based on this analysis.
Scheme 4 Felkin–Anh depiction of nucleophilic attacks
The trend favoring the 2,3-anti -diastereomer is also observed for aryl and methyl Grignards, with >7:1 ratios being
observed (Table [1 ]). Interestingly, ethyl Grignard also afforded an 8:1 selectivity toward the anti -product, which is a near reversal of the syn -preference observed by Joullié and others.[25 ] The 2,3-syn -selectivity has been suggested to arise from chelation to the Boc carbonyl oxygen,[26 ] which might contribute to our observed anti -preference with the less chelation-prone tosylamide. Finally, the allyl Grignard
gave poor selectivity in this reaction.
Table 1 Comparison of Grignard Additions to 8 and to Garner’s Aldehyde
Entry
R
Pg = Tsanti /syn
a
Yieldb
(%)
Pg = Bocanti /syn
Ref.
1
vinyl
8.5:1
95
3–6:1
[7b ]
[9 ]
[16 ]
2
Ph
12:1
70
1.5–5:1
[27 ]
[25b ]
3
4-MeOC6 H4
14:1
n.d.c
5:1d
[27 ]
4
Me
7:1
93
2:1
[25a ]
5
Et
8:1
87e
1:9
[25a ]
6
All
1:1.6
94
1.5:1
[28 ]
a Determined by 1 H NMR integration on the crude sample or after hydrolysis to the diol.
b The crude product contained 1–4% of starting aldehyde.
c Not determined due to contamination by anisole. Hydrolysis gave the diol in 59% yield
over two steps.
d Aryllithium rather than Grignard.
e
anti -Configuration confirmed by comparison with hydrogenated 10 .
For most of the N -tosyl Grignard products, we observed significant decomposition to the diol or rearrangement
to dioxolanes on silica gel chromatography, so for 10 , the crude product was always carried forward. Acidic hydrolysis of the N ,O -acetal by using 4-toluenesulfonic acid in an ethanol/methanol mixture gave chromatographically
pure diol 3 ,[24 ] which could be selectively protected at the primary hydroxy group with tert -butyl(dimethyl)silyl chloride to supply 11 in 80% over three steps from 8 . Note that this silylation is much more easily achieved than that of the similar
Boc-amine diol 2 derived from Garner’s aldehyde, which tends to give disilylation products if great
care is not taken.
To confirm our stereochemical assignment of the vinyl addition, the known oxazolidinone[29 ]
9 was formed in 75% yield from 11 by using triphosgene and pyridine. Unfortunately, the 1 H NMR spectrum reported in the literature was not sufficiently resolved to permit
comparison of coupling constants, but, in general, the H-4 to H-5 coupling (oxazolidinone
numbering) can be easily used to distinguish between the cis - and trans -diastereomers, with cis J
4,5 ≈ 7 Hz and the trans J
4,5 ≈ 4 Hz.[30 ] Oxazolidinone 9 has J
4,5 of 7.6 Hz, indicative of a cis -relationship. In addition, removal of the toluenesulfonyl protecting group could
be accomplished in good yield (83%) by using Na/naphthalene in 1,2-dimethoxyethane,
and the cis -coupling constant between H5 at δ = 5.04 ppm and H4 at δ = 3.83 ppm of oxazolidinone
6 was revealed to be 8.1 Hz, matching that reported by Ibuka,[18 ] and thereby confirming our assignment of the anti -diastereomer 10 from the Grignard chemistry. Note that this synthetic route to 6 via N -tosyl serinal 8 is a significant improvement compared with previously reported Grignard chemistry.
In our case, we had no desire to remove the N -tosyl protection; instead, we sought to deprotect the primary hydroxy and to oxidize
it to a carboxylic acid to form our β-VSer synthetic equivalent. Although there are
reports of both steps being achieved in one pot with KF, Jones reagent, or similar
compounds[31 ] we found it better to do this in a stepwise manner by using HCl and MeOH to remove
the silyl protection in 92% yield, and subsequent Jones oxidation to supply methyl
ester 12 in 82% yield after diazomethane treatment. Unfortunately, attempts at oxidation with
TEMPO-type reagents did not give a complete reaction, giving yields of around 50%
in our hands.
Although we desired the N -tosyl protection, we recognize its versatility is limited for some cases, so we demonstrated
that the final steps can also be carried out with a p -nosyl-protected nitrogen. From 6 , the para -nosyl group can be introduced using sodium hydride in THF to give 13 in 90% yield. Similar reactions have been reported to run in DMF and to give concomitant
silyl ether cleavage,[29 ] but in our case a mixture was always observed. Therefore, we removed the silyl ether
under acidic conditions and employed a Jones oxidation, as described earlier for 12 , to give 14 in similar yields.
In summary, an efficient synthesis of a β-vinyl serine (β-VSer) synthetic equivalent
is reported that exploits the stereodirecting effect of the N -toluenesulfonamide group in a highly diastereoselective vinyl Grignard addition.
