Bare-minimum fluorous mixture synthesis of a stereoisomer library of 4,8,12-trimethylnonadecanols and predictions of NMR spectra of saturated oligoisoprenoid stereoisomers

Article abstract of DOI:10.1021/ja311606u

All four diastereomers of a typical saturated oligoisoprenoid, 4,8,12-trimethylnonadecanol, are made by an iterative three-step cycle with the aid of traceless thionocarbonate fluorous tags to encode configurations. The tags have a minimum number of total fluorine atoms, starting at zero and increasing in increments of one. With suitable acquisition and data processing, each diastereomer exhibits characteristic chemical shifts of methyl resonances in its ¹H and ¹³C NMR spectra. Together, these shifts provide a basis to predict the appearance of the methyl region of the spectrum of every stereoisomer of higher saturated oligoisoprenoids.

Full text of DOI:10.1021/ja311606u

Article
pubs.acs.org/JACS
Bare-Minimum Fluorous Mixture Synthesis of a Stereoisomer Library
of 4,8,12-Trimethylnonadecanols and Predictions of NMR Spectra of
Saturated Oligoisoprenoid Stereoisomers
Edmund A.-H. Yeh, Eveline Kumli, Krishnan Damodaran, and Dennis P. Curran*
Department of Chemistry, University of Pittsburgh, Pittsburgh, Pennsylvania 15260, United States
S
* Supporting Information
ABSTRACT: All four diastereomers of a typical saturated oligoisoprenoid, 4,8,12-
trimethylnonadecanol, are made by an iterative three-step cycle with the aid of traceless
thionocarbonate fluorous tags to encode configurations. The tags have a minimum
number of total fluorine atoms, starting at zero and increasing in increments of one.
With suitable acquisition and data processing, each diastereomer exhibits characteristic
1
chemical shifts of methyl resonances in its H and 13C NMR spectra. Together, these shifts provide a basis to predict the
appearance of the methyl region of the spectrum of every stereoisomer of higher saturated oligoisoprenoids.
INTRODUCTION
■
Fluorous mixture synthesis (FMS)1 is proving to be an
especially valuable tool to make natural product stereoisomer
libraries. Early work was focused on proving the principle that
stereoisomers could be tagged with fluorous tags, mixed, carried
through a multistep synthesis, and then reliably separated and
detagged to give individual, pure stereoisomers.2 Recent work
has had a dual focus of introducing more efficient tagging
strategies and solving structure problems by characterizing the
natural product stereoisomer library members.3 For example,
we introduced a new double-tagging method to make
macrosphelide stereoisomers that allows more compounds to
be tagged with fewer fluorine atoms compared to prior tagging
schemes.4 And by characterizing the library members, we
confirmed the structures of macrosphelides E and M and
corrected the structure of macrosphelide D.
Here we describe an exercise in synthesis of the four
diastereomers of 4,8,12-trimethylnonadecanol (1, Figure 1)
that was undertaken to answer three core questions. First, all
Figure 1. Saturated oligoisoprenoids have repeating stereocenters five
atoms apart.
prior work in fluorous mixture synthesis has been designed
have been characterized by Ingold and co-workers, who showed
based on the premise that protecting groups can do double
that each diastereomer has a unique 13C NMR spectrum.5
duty as fluorous tags. But there are no suitable functional
groups to protect in saturated oligoisoprenoids like 1. How can
fluorous mixture synthesis be applied to such compounds?
Second, reducing the fluorine content of fluorous tags is an
Likewise, other diastereoreomers with two different isoprenoid
units can be differentiated.6 But the differences in 13C NMR
spectra are very small (typically 0.05 ppm or less), so it is not
clear that these differences will translate to higher oligoisopre-
important goal for efficiency, but just how low can you go? Can
noids as more signals are packed into small areas. With its
you have tags with only 0, 1, 2, ... fluorine atoms? And third, are
1
smaller dispersion of resonances, H NMR spectroscopy has
the NMR spectra of the 4,8,12-trimethylnonadecanol isomers
not been useful in differentiating such stereoisomers. More
the same or different? It turned out that there were small but
typically, direct spectroscopic analysis is abandoned in favor of
reliable differences in the spectra of these isomers, and these
differences allowed us to pose and answer a fourth question:
can the NMR spectra of longer saturated oligoisoprenoid
stereoisomers be predicted?
analysis of derivatives bearing long-range chiral reporter
groups.7
With five stereocenters in their oligoisoprenoid chains, β-D-
mannosyl phosphomycoketides like 3a and 3b provide a
Saturated oligoisoprenoids have long alkyl chains with
repeating stereocenters bearing methyl groups every five
atoms on the chain. Vitamin E (2), for example, has a chain
Received: November 27, 2012
of three isoprene units and two stereocenters. Its diastereomers
© XXXX American Chemical Society
A
dx.doi.org/10.1021/ja311606u | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
difficult stereochemical assignment problem.8 To confirm the
two-dimensional structure of 3a, Crich intentionally synthe-
sized and characterized a stereorandom mixture of all 32
possible side-chain isomers.9 The NMR spectra of this mixture
corresponded well to those of the natural sample. Feringa and
Minnaard made the “all-S” isomer of 3b10 and showed that its
biological activity was similar to the natural product.11 If the
stereocenters of 3 are all biosynthesized in the same way, then
the natural product is presumably all-S. But could there be
other stereoisomers that have similar spectra and activity?
We have commented before that, in cases where stereo-
isomers have identical or very similar spectra, it becomes
impossible to prove that a given stereoisomer is the natural
product.3b Instead, it is important to prove that its other
stereoisomers are not the natural product. Accordingly, there is
a need to make and carefully characterize more stereoisomers
of higher saturated oligoisoprenoids to see if their spectra reveal
stereochemical information.
produce a mixture of seven O-arylphenyl chlorothionoformates
6a−g. This was added to the alcohol 4 and pyridine in
dichloromethane, and the resulting product mixture 7 was
analyzed by HPLC on a FluoroFlash PF-8 column.
Although seven phenols 5a−g were added to the starting
mixture, only five peaks were observed on the chromatogram of
7. A preparative separation provided the five fractions, which
1
were identified by H and 19F NMR experiments. These were
(in order of elution) (1) phenylthionocarbonate 7a (19 min);
(2) o-fluorophenyl 7b (21 min); (3) a mixture of m- and p-
fluorophenyl 7c,d (23 min); (4) 3,4-difluorophenyl 7e (28
min); and (5) 3,4,5-trifluorophenyl 7f (37 min). The expected
product 7g from the pentafluorophenylchlorothionocarbonate
6g was not detected. Perhaps 6g is not stable under the
conditions of formation (NaOH and water present). Whatever
the case, we simply rejected this tag from further consideration.
Thus, provided that stereoisomers to be tagged all have the
same polarity, it should be possible to use tags that start with
zero fluorine atoms and differ by only one fluorine atom.
Further, it might be possible to use the o-fluorophenylth-
ionocarbonate as a kind of “fractional tag” (0.5) because it
comes out in between the unfluorinated thionocarbonate (0)
and the m- and p-monofluorinated thionocarbonates (1). This
small effect might be because the fluorine atom in the ortho
isomer is partially shielded from the fluorous silica gel by its
neighboring substituent.19
Iterative Cycle. The experiments in Scheme 1 and other
pilot studies18 were conducted by Brown−Ramachandran
crotylation; however, for the iterative cycles we opted for a
Roush crotylation20 because Roush reagents are convenient to
make and store, and their reactions are easy to conduct and
work up. The iterative cycle to make the isomers of 1 consists
of (1) Roush crotylation of an aldehyde, (2) tagging of the
resulting alcohol with a suitable chlorothionoformate to encode
configuration, and (3) hydroformylation of the terminal alkene
to give a chain-extended aldehyde ready for the next cycle.
In the first cycle, Scheme 2, heptanal 9 was reacted with the
Roush reagent SS-8 derived from the (−)-D-diisopropyl tartrate
RESULTS AND DISCUSSION
■
Model Tagging Experiments. We decided to use
asymmetric crotylation reactions to introduce the stereocenters
because there are several crotylation reagents that reliably
produce all stereoisomers with reagent control12 and because
the extra hydroxy group produced in the crotylation could serve
as a convenient tag location. After several iterations, we planned
to remove the tags by Barton−McCombie deoxygenation,13 so
we selected the aryl ring of the O-arylthionocarbonate group14
(OC(S)OAr) as the location for the fluorine atoms.
Fluorous tags have typically been fashioned by adding
spacers15 followed by perfluoroalkyl groups of various
lengths.16 Here we instead replaced hydrogen atoms on the
aryl ring of the tag with fluorine atoms (Ar = phenyl,
fluorophenyl, difluorophenyl, ...). In addition to evaluating tags
with 0, 1, 2, 3, and 5 fluorine atoms, we also evaluated
regioisomeric tags with one fluorine atom in the ortho, meta,
and para positions.
The performance of seven potential tags was evaluated on a
model compound 7 as shown in Scheme 1. Alcohol 4 was
synthesized by two Brown−Ramachandran crotylations17,18
and the standard O-phenylthionocarbonate group was added
after the first crotylation. After the second crotylation to give 4,
a mixture of the seven phenols 5a−g was reacted with
thiophosgene under standard conditions14 to presumably
Scheme 2. First Iteration by Traditional Synthesis of Single
Compounds Provides Aldehyde 13
Scheme 1. Evaluation of Fluorinated Tags with Model
Mixture 7
to give RS-10 in 88% yield. This reaction installs the C12
stereocenter21 of the target product. It is highly diastereose-
lective (>98% anti), but not completely enantioselective (see
below). Reaction of 10 with O-phenylchlorothionoformate 6a
and pyridine as above gave RS-11. Hydroformylation of 11 with
syngas (H2 and CO) and Rh(CO)2acac (7%) and 6-
(diphenylphospino)-2-pyridone 12 (30%) occurred smoothly
in THF at 60 °C/120 psi to give RS-13 in 80% yield.22
B
dx.doi.org/10.1021/ja311606u | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
Evidently, the thionocarbonate group does not disturb the
catalyst for this reaction. The overall isolated yield of 60% from
the three-step cycle was deemed suitable for three iterations.
To test whether the thionocarbonate survives the Roush
crotylation and to better understand the stereoselectivity of the
crotylation, we looked first at one of the products of the second
iterative cycle (Scheme 3). Reaction of 13 with Roush reagent
Scheme 4. Two Pilot Mixtures with Different Tagging
Patterns Form the Second and Third Iterations
Scheme 3. Synthesis of 14 as a Single Compound in 82%
Isomer Purity Shows the Stereoselectivity in Two Successive
Roush Crotylations
quasiisomers24 were then mixed back together for the
subsequent steps.
SS-8 produced RSRS-14 in 87% isolated yield. The sample
generally looked like a single compound by 13C NMR analysis,
except that the carbinol carbon exhibited two peaks at 74.83
and 74.57 ppm in a ratio of about 82/18. The major peak is
RSRS-14 (configuration shown in Scheme 3), and we assign the
minor peak as a mixture of SRRS-14 and its enantiomer RSSR-
14. The RSSR enantiomer arises from the minor path in the
first Roush crotylation and the major path in the second
crotylation, while its enantiomer results from the major path in
the first and minor in the second.
Hydroformylation in cycle 2 gave a new aldehyde sample 15
of predominately two quasiisomers that was again divided in
half. Each half was reacted with one enantiomer of the Roush
reagent, then the samples were tagged appropriately and
remixed. Another hydroformylation completed the third cycle,
giving a mixture of four aldehydes ready in principle for a fourth
cycle. However, for this study we reduced the aldehyde mixture
with DIBAL-H to give alcohols 16 ready for demixing.
The tables in the lower part of Scheme 4 show how the
mixture members are encoded and provide the total number of
fluorine atoms (F’s) in each member. All the pairs of adjacent
stereocenters in all members have anti configurations. Because
the hydroxy groups will be removed, we focus now on the
configurations of the methyl-bearing stereocenters at C4, C8,
and C12.
Both pilot mixtures have the (12R) configuration from the
starting aldehyde 13 (to the extent of 90%), and the C13-OH
group bears a standard phenylthionocarbonate (C6H5). In pilot
mixture 1, we used the regioisomeric mono-fluorophenyl tags.
After the first pair of Roush crotylations (cycle 2), we tagged
the (8R)-isomer with the phenyl group (C6H5) and the (8S)-
isomer with the “fractional” o-fluorophenyl group (o-FC6H4), as
indicated above. After the second pair of Roush crotylations
(cycle 3), we tagged the (4R)-isomers again with the phenyl
group (C6H5). Now the (4S)-isomers were tagged with the p-
fluorophenyl group (p-C6H4). Notice that only three tags are
used to encode four samples of 16 in pilot mixture 1, and that
the middle two samples have the same number of fluorines (1).
In pilot mixture 2, we stepped up the tags by one fluorine
atom each. After the first pair of Roush crotylations, we tagged
the (8R)-isomer with the phenyl group (C6H5) and the (8S)-
isomer with the p-fluorophenyl group (p-FC6H4). After the
second pair of Roush crotylation, we tagged the (4R)-isomer
again with the phenyl group, while the (4S)-isomer was tagged
with the 3,4-difluorophenyl group (3,4-F2C6H3). Again three
Assuming that the two crotylations give about the same
selectivity, this means that each gives a 90/10 ratio of isomers
(enantiomers after the first crotylation and diastereomers after
the second). The product of the minor path in both
crotylations is the enantiomer of the major product. Present
to the extent of about 1%, this product (SRSR-14, not shown)
can be neglected. The enrichment of the enantiomeric ratio of
the major diastereomer of 14 is similar to that observed in
sequential asymmetric transformations like dynamic kinetic
resolutions.23 Keep in mind, however, that the diastereomers of
1 cannot be directly separated at the end of this exercise
because the stereocenters are so far removed. Nonetheless, the
presence of minor isomers proved to be vitally important as
controls in the spectroscopic analysis of the final samples.
Second and Third Iterations in Mixture Mode. Because
of the ease of the cycle, we were able to conduct two pilot runs
in mixture mode, as summarized in Scheme 4. The cycles were
essentially the same; only the tags were changed. A sample of
aldehyde RS-13 was divided in two, and each half was treated
with one enantiomer of the Roush reagent. The products were
tagged with appropriate chlorothionocarbonates. For example,
in the first cycle of pilot mixture 1, the product from SS-8 was
tagged with the thionocarbonate-bearing phenyl (Ar1 = C6H5),
while that from RR-8 was tagged with the thionocarbonate-
bearing o-fluorophenyl (Ar1 = o-FC6H4). The tagged
C
dx.doi.org/10.1021/ja311606u | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
tags are used to encode four samples in pilot mixture 2, but
now there is no redundancy of total fluorine content.
Analytical HPLC traces of the two pilot mixtures are shown
in Figure 2. In mixture 2, the total fluorine content of six
mixture of four quasiisomers of 16 was conducted on a larger
FluoroFlash column, and a trace of a typical run is shown in
Figure S1 of the Supporting Information (SI). The recovery of
pure samples from the HPLC separation ranged from 68 to
74% for the four combined samples, and the combined weights
of each of the four quasiisomers 16 ranged from 16−20 mg.
The complete structures of the final purified quasiisomers
16a−d with both configurations and tags are shown in Scheme
5. Notice how the different aryl tags encode the configurations
at C4/5 and C8/9. The four quasiisomers 16 were
characterized as usual by NMR and MS experiments. There
are small differences in their NMR spectra, but these differences
may not be related to configuration and are simply because the
tags are chemically different. The purity of each sample was
assessed by analytical fluorous HPLC. The first eluting sample
16a was a single peak, but the subsequent samples 16b−d were
contaminated with 1−3% of the prior eluting quasiisomers.
These contaminants translate to stereoisomers after detagging.
Because the amounts are small relative to the already known
stereoisomer contaminants from the Roush crotylation, we did
not repurify the three slightly impure samples.
Each of the O-arylthionocarbonates was reduced by exposure
to 1,3-dimethylimidazol-2-ylidene borane (di-Imd-BH3, 5
equiv) and AIBN in refluxing benzene.25 Simple solvent
evaporation and flash chromatography provided the final
4,8,12-trimethylnonadecanol stereoisomer samples 1 in 52−
67% yields. This reduction with the carbene-borane reagent is
superior to the usual tin hydride reduction because complete
separation of tin byproducts from nonpolar samples like 3 can
be tedious.
Notice the CIP priority switch that occurs at all the
remaining stereocenters on deoxygenation of the quasiisomers
16 to give true isomers 1. This is of little importance because
we are concerned with relative stereochemistry, and going
forward we indicate this in the usual way as anti/syn. For
example, (4S,8S,12S)-1 has the syn relative configuration
between both C4/C8 and C8/C12 so is called 1-syn,syn.
Each sample of 1 can contain up to eight stereoisomers. In
round numbers, each sample should consist of about 70% of
the stereoisomer from the favored product of each of the three
Roush crotylations. This is the structure shown in Scheme 5.
Each major product should contain about 9% each of three
significant minor isomers (one from the minor product being
formed in each of the three Roush crotylations). There are also
three double-minor diastereomers, which are enantiomers of
Figure 2. Analytical HPLC traces of pilot mixtures 1 (left, 65/35
MeCN/H2O) and 2 (right, 70/30 MeCN/H2O); FluoroFlash PF-8
column.
fluorine atoms spread over the four products is 0, 1, 2, and 3. In
pilot mixture 1, the total fluorine content is four fluorine atoms,
spread 0, 1, 1, and 2. Both provided four peaks, as expected. But
the separation of pilot mixture 2 (right trace) was clearly
superior, with baseline resolution being observed between every
pair of peaks.
Interestingly, in the inferior pilot mixture 1 (left trace), the
two compounds with one fluorine atom and regioisomeric tags
(the two middle peaks of the chromatogram) separate rather
well, with the o-fluoro-tagged compound emerging before the
p-fluoro compoud. The problem is that the o-tag compound
(peak 2) is shifted toward the zero-tag compound (peak 1) and
the p-tag compound (peak 3) is shifted toward the difluoro-tag
compound (peak 4). In essence, the two samples with
regioisomeric tags can be separated from each other, but it is
difficult to “fit” two tags with the same fluorine content in the
small space between the higher and lower tags.
As a control, pilot mixture 2 was injected onto a standard
reverse-phase HPLC column (RP-C18, comparable retention
times observed with 80/20 MeCN/H2O). However, only a
single, very broad peak was observed. Thus, despite the low
fluorine content of the tags, the use of a fluorous HPLC column
is crucial for demixing.
Based on these results, we scaled up the mixture synthesis by
using the tagging pattern in pilot mixture 2. Yields of the cycles
on preparative scale were comparable: 60% for cycle 1 (Scheme
1), 59% for cycle 2, and 56% for cycle 3 (including the extra
step of DIBAL-H reduction). Preparative separation of the final
Scheme 5. Structures of Final Tagged Products 16a−d and Reduced Nonadecanol Stereoisomers 1, All in Individual Form
D
dx.doi.org/10.1021/ja311606u | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
the major diastereomer impurities, in ∼1% each. Adding these
to their major enantiomers gives a total diastereomer ratio of
about 70/10/10/10. The eighth isomer in each sample is the
triple-minor product. This enantiomer of the major product can
be neglected because the projected er of the major product is
>100/1.
mutually consistent set of assignments27 that is shown under
the structures in Scheme 5 and collected in Table 1.
Table 1. The Seven Resonances of the Branched Methyl
1
Groups in 13C and H NMR Spectra of 1
To summarize, the mixture synthesis provided ample
quantities (5−6 mg) of the four target isomers of 1 for
complete characterization. Every major isomer is projected be
present in about 70% along with 10% each of three minor
diastereomers. This defect from the standpoint of asymmetric
synthesis proved to be an asset in the analysis.
Analysis of NMR Spectra. NMR spectra were initially
recorded and processed in the usual way for each of the four
samples: 1H NMR at 700 MHz and 13C NMR at 175 MHz, all
in CDCl3. We could see that the spectra, though very similar,
were not identical in places. But we could not assign any
specific resonances or determine isomer ratios because the
resonances were too close together.
assignment
Me-20, syn
13C NMR ppm
1H NMR ppm
19.67
19.61
19.79
19.73
19.66
19.77
19.70
0.874
0.872
0.844
0.842
0.841
0.840
0.839
Me-20, anti
Me-21, syn,syn
Me-21, syn/anti
a
Me-21, anti/anti
Me-22, syn
Me-22, anti
a
Or anti/syn.
Additional processing of the data by the Traficante algorithm
followed by forward linear prediction26 gave a significant
improvement in resolution while maintaining needed sensi-
tivity. Expansions of the region of the 13C NMR spectrum with
the resonances of the branched methyl groups (Me-21, 22, 23)
are shown in Figure 3 along with simulated spectra whose
production will be discussed presently.
The chemical shift of each kind of methyl group (Me-20, 21,
22) is influenced by the configuration of the nearest
stereocenter(s). Thus, there are only two resonances for Me-
20 and two for Me-22 because these methyl groups can be syn
or anti to Me-21. In turn, Me-21 has two neighbors, so it can
have three relationships, syn/syn, syn/anti (or anti/syn), and
anti/anti. It does not matter whether Me-21 is anti to Me-20
and syn to Me-22 or the reverse, so this is why it has three
resonances rather than four.
The middle unit is comparable to the repeating monomer in
assorted saturated polyisoprenoids [CH2CH2CH(CH3)CH2]n.
The absolute chemical shifts of Me-21 in the stereoisomers of 2
cannot be compared to the polymers because the polymer 13C
NMR spectra are acquired at high temperatures.28 But the
various polymer isomer units exhibit the same trends. The
polymer terms for the adjacent stereochemical relationships are
meso and racemic, and polymer resonances move slightly upfield
in the order: meso/meso (syn/syn), meso/rac or rac/meso (syn/
anti or anti/syn), rac/rac (anti/anti).
The isomer purities of the samples were assessed by 13C
NMR analysis with the aid of the simulated spectra shown in
Figure 3. These simulations were produced by inputting the
measured chemical shifts and the predicted isomer ratios
(always 70/10/10/10, but with a different major isomer in each
case). The good correspondence of the relative peak heights in
the experimental and simulated spectra shows that the actual
purities of the samples are similar to the projected purities. This
in turn shows that the Roush crotylations on mixture samples
(cycle 3) worked as well as those on individual samples (cycles
1, 2).
Figure 3. Expansions of the methyl region of the 13C NMR spectra of
the four samples of 1: left, actual spectra with Traficante processing;
right, simulated spectra based on predicted isomer ratios. From top to
bottom, syn/syn (S,S,S), anti/anti (S,R,S), anti/syn (R,S,S), and syn/anti
(R,R,S).
There are four isomers of 1 with three branched methyl
groups each, so twelve signals might be expected in total.
However, notice that each experimental spectrum contains the
same seven peaksthree larger ones and four smaller ones.
The differences between the spectra are which peaks are large
and which are small.
With this limited information, it is already possible to assign
all of the resonances by a process of elimination.18 For example,
isomers 1-syn,syn and 1-syn,anti share one large resonance at
19.67 ppm while the other large ones are different. The shared
resonance has to belong to Me-20 because this is the only
methyl group that has the same stereochemical relationship
(syn to Me-21) in both compounds. Continuing to compare
shared large peaks and stereochemical relationships provides a
Though more complicated, the set of 1H NMR spectra yield
to the same analysis. There is extensive overlapping in the
methyl region because of chemical shift compression, J
coupling, and addition of a new resonance (the terminal
methyl group, C19). Again Traficante processing is crucial, and
again spectra were simulated with 70/10/10/10 isomer purities.
These pairs of spectra are shown in Figure 4. The simulated
spectra are easier to interpret; each sample exhibits three
doublets and a triplet in the Me region of the spectrum.
The triplet for the terminal Me group resonates at 0.882 ppm
in all of the isomers. The seven branched Me resonances are
again listed with full assignments in Scheme 5 and Table 1. Five
doublets (three for Me-21 and two for Me-22) are packed
between 0.839 and 0.844 ppm, while the other two doublets
E
dx.doi.org/10.1021/ja311606u | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
resonances in a 2/1/1 ratio (present in four isomers) or four
resonances in 1/1/1/1 ratio (present in the other four
isomers). Although there are only two patterns of peak ratios,
the chemical shifts vary in each. So every one of the eight
1
diastereomers of 17 has a unique predicted H and 13C NMR
spectrum.
Compound 18 has the same seven chemical shifts as 17 and
1, but now the patterns of peak heights vary more because the
three signals of the middle carbons can be present with relative
intensities of 0, 1, 2, or 3 in a given spectrum. Globally now,
there are only 14 (not 16) different predicted spectra with four
spectra having three resonances (3/1/1), eight spectra having
four resonances (2/1/1/1) and two spectra having the
maximum five resonances (1/1/1/1/1). Twelve of the isomers
have a unique spectrum, while two pairs of isomers have
identical predicted spectra (both in the “four peak” class). The
redundancies arise because the spectra do not differentiate
middle carbons with syn/anti relationships from those with
anti/syn relationships.29
Predicted spectra of higher oligomers are easily generated as
well by the same method. They will never have more than five
different peaks (for a pure isomer) because each compound can
only have one kind of carbon on each end (syn or anti) and up
to three kinds of carbons in the middle (syn/syn, anti/anti, syn/
anti same as anti/syn). But the intensities of the peaks continue
to vary with the number and configuration of the middle units.
Simulated pairs of 1H and 13C spectra of three variants of 18
are shown in Figure 6. These illustrate how the predictions are
1
Figure 4. Expansions of the methyl region of the H NMR spectra of
the four samples of 1: left, actual spectra with Traficante processing;
right, simulated spectra based on predicted isomer ratios. From top to
bottom, syn/syn (S,S,S), anti/anti (S,R,S), anti/syn (R,S,S), and syn/anti
(R,R,S).
(Me-20) at 0.874 and 0.872 ppm overlap with each other and
with the triplet of the terminal methyl group. Each peak is clear
even though the entire range of peaks is less than 35 ppb (parts
per billion).
The deduced assignments of the methyl resonances in the 1H
NMR spectra were confirmed by 1D TOCSY experiments that
are shown in the SI. Finally, 2D COSY experiments correlated
each 1H methyl resonance with its 13C counterpart to complete
the assignments. These experiments prove that the deduced
assignments are correct.
Predictions of NMR Spectra of Higher Oligomers.
Nonadecanol 1 is a model for higher saturated oligoisoprenoid
side chains like 17 and 18 (Figure 5). All these oligomers have
Figure 5. The basis for prediction of spectra: oligomers 17 and 18
have the same left and right ends as 1, and the middle unit simply
repeats.
Figure 6. Predicted 1H (left) and 13C (right) NMR spectra for
representative samples of 18: top, an equimolar mixture of all 16
diastereomers; middle, the “all-S” (or “all-R”) isomer with all four
relationships syn; and bottom, the isomer with the relationship syn/
anti/anti/syn.
the same left and right chain ends. The repeating middle unit
appears once in nonadecanol 1, twice in 17, and three times in
18 (which is the side chain of phosphomycoketide 3b in Figure
1). The methyl regions of the spectra for all the isomers of 17,
18, and higher oligomers are readily predicted simply by
assigning each carbon in an oligomer to one of seven chemical
shifts of 1 and then plotting the results.
made and what some typical differences are. The top pair of
spectra are simulations of an equimolar mixture of all sixteen
diastereomers of 18. This is what the spectra of Crich’s side-
chain sample9 might look like under our recording and
processing conditions (though he made 3a with a C4H9 tail,
see below). There are only seven peaks in the model, so there
are seven peaks in the simulated spectra of the all-isomer
mixture. This is most easily seen in the 13C NMR spectrum
(upper right). The relative intensities of the peaks total to 3 for
each of the three middle Me resonances and 1 for each of the
four end Me resonances.
We simulated the 1H and 13C NMR spectra of the diagnostic
methyl regions of the 8 diastereomers of 17 and the 16
diastereomers of 18. These spectra (48 in total) are displayed in
the SI.
The number of branched methyl resonances in the spectrum
of each isomer of 17 depends on whether the relative
configurations of the pair of middle fragments are the same
(one peak, twice as large) or different (two peaks). Add these
to the two different ends, and the spectra can have three
The spectra of the “all-S” isomer of 18 made by Feringa and
Minnaard10 might look the middle spectrum of Figure 6. In the
F
dx.doi.org/10.1021/ja311606u | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
13C spectrum, the smaller peaks at 19.67 and 19.77 ppm are
resonances for the two end carbons, both with adjacent syn
relationships. All three middle carbons have syn/syn relation-
ships, so they are predicted to come as one resonance at 19.79
ppm with a relative intensity of three.
related but not identical to 1. By comparing the model chemical
shifts to those of an actual mixture according to the steps above,
the chemical shifts of all the components can be extracted and
assigned to their relevant isomers without ever recording
spectra of the pure isomers.
Finally, the process can of course be conducted in reverse;
that is, as an assignment rather than a prediction. For example,
what isomer of 18 has the spectra exhibited in the bottom of
Figure 6? Again looking at the simpler 13C NMR spectrum, the
peaks at 19.67 and 19.77 show that the right and left end
methyl group both have syn relationships to their neighbors.
The peak at 19.73, relative intensity two, shows that two middle
carbons have a syn/anti relationship while the peak at 19.66
shows that the third is anti/anti.
The result is five piecessyn, two times; syn/anti, two times;
and anti/anti, one timethat fit together like a little puzzle.
The two end pieces, syn and syn, have the three middle pieces in
between. In the puzzle, each relationship has to match its
adjacent relationship, and every syn/anti middle piece can be
inserted either way (syn/anti or anti/syn). In this puzzle, for
example, you have to put one of the syn/anti pieces after the
first syn piece; you cannot put the anti/anti piece there. So
these five pieces fit together in only one way (syn-syn/anti-anti/
anti-anti/syn-syn), and this is the (4S,8S,12R,16S,20R) isomer of
18.
Users of the predicted spectra sets need to keep in mind the
assumption. We already know from the data set that the
chemical shifts of the two end stereocenters are only affected by
the configuration of the middle stereocenter, not by each other.
This is the basis of the assumption that the chemical shifts can
be predicted based only on nearest neighbor configurations.
This assumption can only be tested by making higher
oligomers. Even if it ultimately proves to be wrong, secondary
effects of more remote stereocenters are likely to be smaller
than the observed differences in 1, so the predicted spectra still
have value.
Notice also in Table 1 that the resonances of both syn ends
are downfield from their anti counterparts, in both 1H and 13C
spectra. And in the middle the chemical shifts have the same
trend: syn/syn, most downfield; syn/anti, in the middle; anti/
anti, most upfield. Converting this to an assumption (syn always
downfield from anti), the spectra of other saturated
oligoisopreniods having different right/left ends from 1 can
be analyzed, even as mixtures.
CONCLUSIONS
■
Each of the four questions posed at the outset has been
answered satisfactorily. First the combination of asymmetric
crotylation reactions, thionocarbonate encoding, hydroformy-
lation and Barton−McCombie deoxygenation allows for the
reliable synthesis of saturated oligoisoprenoids by mixture
methods. The strategy introduces extra hydroxy-bearing
stereocenters in each cycle. These provide sites for tags and
allow the use of asymmetric crotylation, one of a few reactions
that can provide reliable access to all stereoisomers with
complete reagent control.
Second, the bare minimum increment in fluorous tags has
been reached with the quasiisomers of 16 having 0, 1, 2, or 3
total fluorine atoms. Indeed, these compounds do not even fit
the accepted definition of a “fluorous compound”.30 More
precisely, the technique here should be called “fluorinated
mixture synthesis” rather than “fluorous mixture synthesis”. But
it is still “FMS”, not just because the abbreviations coincide but
because the unique interaction between fluorine atoms and the
fluorous stationary phase is crucial to the separation. We even
showed in principle that it is possible to go below the bare
minimum with a “fractional tag” (o-fluorophenyl). However, in
practice, the observed separation was tight and the incremental
tags were better. Nonetheless, with improved separation, such
fractional tags might be useful.
Third, we made the four enriched samples of the
stereoisomers of 4,8,12-trimethylnonadecanol 1 and discovered
1
that their H and 13C NMR spectra were very similar but not
identical. With processing to enhance resolution, characteristic
differences were identified in the methyl-group regions of the
spectra. The 90/10 selectivity of the Roush crotylation left
minor isomer impurities that conveniently functioned as
controls in sorting out the NMR spectra of isomers 1 and
proving that the Roush crotylation worked on mixtures as well
as on single isomers. However, better stereoselectivites are
desirable going forward to make higher oligomers like 17 and
18.
Fourth, the simple correlation of the seven types of
stereogenic carbon atoms to the seven methyl resonances in
For example, the right chain end of 3a has a C4H9 group, not
C6H13. And assorted compounds with other left end groups
(other functional groups or oxidation states, shorter/longer
chain lengths) are likewise of interest. A stereoisomer library of
compounds with, say, the left end different from 1 will have five
of the same resonances (middle and right end) and two new
resonances. The two new resonances have to be from the
stereocenter nearest to the left end, with the downfield one
being syn, the upfield, anti. A library of isomers with both ends
different will have four new resonances (two from each new
end). Now one end pair has to be assigned spectroscopically
(as we did with the 1D TOCSY experiments to prove the
assignment of Me-20). Process of elimination gives the other
end pair, then the “syn downfield” guideline completes the
assignment.
1
both the H and 13C NMR spectra of 1 can be extrapolated to
predict the NMR spectra of oligomers, and the SI contains the
48 predicted spectra for all of the isomers of 17 and 18. Spectra
for higher oligomers up to polymers are readily generated.
This work sets the stage for synthesis of higher oligomer
stereoisomer libraries like those of 17 and 18. Such work could
validate the predicted spectra or perhaps refine the predictions.
This is especially important for mycoketide side chains like 18
because it now appears certain that careful analysis of NMR
spectra of such compounds will yield valuable information
about relative configuration and isomer purity.
ASSOCIATED CONTENT
■
S
* Supporting Information
To determine these chemical shifts and guidelines, we
needed to make pure stereoisomers of 1 to assign the peaks.
However, now that the model is complete, it becomes possible
to make true mixtures of isomers in future work on compounds
Complete experimental details, additional figures, predicted
spectra of the isomers of 17 and 18, and H and 13C NMR
1
spectra of all final products and key intermediate products. This
G
dx.doi.org/10.1021/ja311606u | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
Gladysz, J. A., Curran, D. P., Horvath, I. T., Eds.; Wiley-VCH:
Weinheim, 2004; pp 128−156.
(17) Ramachandran, P. V. Aldrichimica Acta 2002, 35, 23−35.
(18) For extensive additional information about the synthesis,
characterization, and spectral predictions, see: Yeh, E. A.-H. Ph.D.
Thesis, University of Pittsburgh, 2012.
(19) Curran, D. P. In The Handbook of Fluorous Chemistry; Gladysz, J.
A., Curran, D. P., Horvath, I. T., Eds.; Wiley-VCH: Weinheim, 2004;
pp 101−127.
(20) Roush, W. R.; Ando, K.; Powers, D. B.; Palkowitz, A. D.;
Halterman, R. L. J. Am. Chem. Soc. 1990, 112, 6339−6348.
(21) The carbon numbers of all the intermediates are based on the
numbering of the final product 1.
material is available free of charge via the Internet at http://
pubs.acs.org.
AUTHOR INFORMATION
Corresponding Author
curran@pitt.edu
■
Notes
The authors declare no competing financial interest.
ACKNOWLEDGMENTS
■
We thank the National Institutes of Health for funding this
work and for funds to help purchase a 700 MHz NMR
spectrometer. We thank Prof. Tara Y. Meyer, University of
Pittsburgh, for helpful discussions on the polymer comparisons.
(22) Breit, B.; Seiche, W. J. Am. Chem. Soc. 2003, 125, 6608−6609.
(23) (a) Pellissier, H. Tetrahedron 2003, 59, 8291−8327. (b) Ward,
R. S. Tetrahedron: Asymmetry 1995, 6, 1475−1490.
(24) Zhang, Q. S.; Curran, D. P. Chem.Eur. J. 2005, 11, 4866−
4880.
REFERENCES
■
́
(25) (a) Ueng, S.-H.; Makhlouf Brahmi, M.; Derat, E.; Fensterbank,
L.; Lacote, E.; Malacria, M.; Curran, D. P. J. Am. Chem. Soc. 2008, 130,
10082−10083. (b) Ueng, S.-H.; Fensterbank, L.; Lacote, E.; Malacria,
M.; Curran, D. P. Org. Lett. 2010, 12, 3002−3005.
(26) Traficante, D. D.; Nemeth, G. A. J. Magn. Reson. 1986, 71, 237−
245.
(1) (a) Luo, Z. Y.; Zhang, Q. S.; Oderaotoshi, Y.; Curran, D. P.
Science 2001, 291, 1766−1769. (b) Zhang, Q. S.; Rivkin, A.; Curran,
D. P. J. Am. Chem. Soc. 2002, 124, 5774−5781.
̂
̂
(2) (a) Dandapani, S.; Jeske, M.; Curran, D. P. Proc. Natl. Acad. Sci.
U.S.A. 2004, 101, 12008−12012. (b) Zhang, Q. S.; Lu, H. J.; Richard,
C.; Curran, D. P. J. Am. Chem. Soc. 2004, 126, 36−37. (c) Dandapani,
S.; Jeske, M.; Curran, D. P. J. Org. Chem. 2005, 70, 9447−9462.
(3) (a) Yang, F.; Newsome, J. J.; Curran, D. P. J. Am. Chem. Soc.
2006, 128, 14200−14205. (b) Curran, D. P.; Zhang, Q. S.; Lu, H. J.;
Gudipati, V. J. Am. Chem. Soc. 2006, 128, 9943−9956. (c) Curran, D.
P.; Zhang, Q. S.; Richard, C.; Lu, H. J.; Gudipati, V.; Wilcox, C. S. J.
Am. Chem. Soc. 2006, 128, 9561−9573. (d) Jung, W.-H.; Guyenne, S.;
Riesco-Fagundo, C.; Mancuso, J.; Nakamura, S.; Curran, D. P. Angew.
Chem., Int. Ed. 2008, 47, 1130−1133. (e) Curran, D. P.; Sui, B. J. Am.
Chem. Soc. 2009, 131, 5411−5413. (f) Moretti, J. D.; Wang, X.;
Curran, D. P. J. Am. Chem. Soc. 2012, 134, 7963−7970.
(27) “Mutually consistent” means that the same assignments emerge
independent of how you conduct the process of elimination.
(28) Depending on how they are made, comparable polymers are
called poly(1-methyltetramethylene) or alternating copolymers of
ethylene and propylene. (a) Zetta, L.; Gatti, G.; Audisio, G.
Macromolecules 1978, 11, 763−766. (b) Fan, W.; Leclerc, M. K.;
Waymouth, R. M. J. Am. Chem. Soc. 2001, 123, 9555−9563.
(29) For example, the isomer with syn/syn/anti/syn relative
configuration (two syn ends and the middle carbons in the order
syn/syn next to syn/anti next to anti/syn) has the same four model
carbons in the same ratio (2/1/1/1) as the isomer syn/anti/syn/syn
(two syn ends and the middle carbons in the order syn/anti next to
anti/syn next to syn/syn).
(4) Curran, D. P.; Sinha, M. K.; Zhang, K.; Sabatini, J. J.; Cho, D.-H.
Nat. Chem. 2012, 4, 124−129.
(5) (a) Baker, J. W.; Myers, C. W. Pharm. Res. 1991, 8, 763−770.
(b) Brownstein, S.; Burton, G. W.; Hughes, L.; Ingold, K. U. J. Org.
Chem. 1989, 54, 560−569.
(30) Gladysz, J. A.; Curran, D. P. Tetrahedron 2002, 58, 3823−3825.
(6) Heathcock, C. H.; Finkelstein, B. L.; Jarvi, E. T.; Radel, P. A.;
Hadley, C. R. J. Org. Chem. 1988, 53, 1922−1942.
(7) Ohrui, H. Proc. Jpn. Acad., Ser. B 2007, 83.
(8) Moody, D. B.; Ulrichs, T.; Muhlecker, W.; Young, D. C.; Gurcha,
S. S.; Grant, E.; Rosat, J.-P.; Brenner, M. B.; Costello, C. E.; Besra, G.
S.; Porcelli, S. A. Nature 2000, 404, 884−888.
(9) Crich, D.; Dudkin, V. J. Am. Chem. Soc. 2002, 124, 2263−2266.
(10) van Summeren, R. P.; Moody, D. B.; Feringa, B. L.; Minnaard,
A. J. J. Am. Chem. Soc. 2006, 128, 4546−4547.
(11) de Jong, A.; Arce, E. C.; Cheng, T.-Y.; van Summeren, R. P.;
Feringa, B. L.; Dudkin, V.; Crich, D.; Matsunaga, I.; Minnaard, A. J.;
Moody, D. B. Chem. Biol. 2007, 14, 1232−1242.
(12) (a) Buitrago Santanilla, A.; Leighton, J. L. In Stereoselective
Synthesis 2: Stereoselective Reactions of Carbonyl and Imino Groups;
Molander, G. A., Ed.; Georg Thieme Verlag: Stuttgart-New York,
2010; Vol. 2, pp 401−447. (b) Hoffmann, R. W. In Asymmetric
SynthesisThe Essentials; Christmann, M., Brase, S., Eds.; Wiley-VCH:
̈
Weinheim, 2007; pp 29−33.
(13) (a) Crich, D.; Quintero, L. Chem. Rev. 1989, 89, 1413−1432.
(b) Saraiva, M. F.; Couri, M. R. C.; Le Hyaric, M.; de Almeida, M. V.
Tetrahedron 2009, 65, 3563−3572. (c) Barton, D. H. R.; McCombie,
S. W. J. Chem. Soc., Perkin Trans. 1 1975, 1574−1585.
(14) Robins, M. J.; Wilson, J. S. J. Am. Chem. Soc. 1981, 103, 932−
933.
(15) Gladysz, J. A. In The Handbook of Fluorous Chemistry; Gladysz, J.
A., Curran, D. P., Horvath, I. T., Eds.; Wiley-VCH: Weinheim, 2004;
pp 41−55.
(16) (a) Zhang, W. Handbook of Fluorous Chemistry; Gladysz, J. A.,
Curran, D. P., Horvath, I. T., Eds.; Wiley-VCH: Weinheim, 2004; pp
222−236. (b) Curran, D. P. In The Handbook of Fluorous Chemistry;
H
dx.doi.org/10.1021/ja311606u | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX