Comparison of diffusion coefficients for matched pairs of macrocyclic and linear molecules over a drug-like molecular weight range

Article abstract of DOI:10.1039/c1ob05996c

The diffusion coefficients of a series of closely matched pairs of macrocyclic and linear molecules have been compared using NMR spectroscopy. The macrocyclic series was designed both to overlap with and extend beyond the molecular weight range typically employed for drug-like molecules. The linear molecules each represent a carbogenic fission of their macrocyclic counterparts, designed to minimize differences in functionality and physicochemical properties. Each series of molecules was prepared using copper catalyzed azide-alkyne cycloaddition (CuAAC) reactions conducted in a flow using a copper tube. The macrocyclic series exhibited consistently higher diffusion across the entire molecular weight range studied. The fold difference in diffusion coefficients between the macrocyclic and linear analogues appeared to be independent of either solvent viscosity or dielectric environment.

Full text of DOI:10.1039/c1ob05996c

View Article Online / Journal Homepage / Table of Contents for this issue
Organic &
Biomolecular
Chemistry
Dynamic Article Links
Cite this: Org. Biomol. Chem., 2011, 9, 7727
www.rsc.org/obc
PAPER
Comparison of diffusion coefficients for matched pairs of macrocyclic and
linear molecules over a drug-like molecular weight range†
Andrew R. Bogdan,a Nichola L. Daviesb and Keith James*a
Received 20th June 2011, Accepted 5th September 2011
DOI: 10.1039/c1ob05996c
The diffusion coefficients of a series of closely matched pairs of macrocyclic and linear molecules have
been compared using NMR spectroscopy. The macrocyclic series was designed both to overlap with
and extend beyond the molecular weight range typically employed for drug-like molecules. The linear
molecules each represent a carbogenic fission of their macrocyclic counterparts, designed to minimize
differences in functionality and physicochemical properties. Each series of molecules was prepared
using copper catalyzed azide-alkyne cycloaddition (CuAAC) reactions conducted in a flow using a
copper tube. The macrocyclic series exhibited consistently higher diffusion across the entire molecular
weight range studied. The fold difference in diffusion coefficients between the macrocyclic and linear
analogues appeared to be independent of either solvent viscosity or dielectric environment.
diffusion rate, which represents a sensitive measure of molecular
Introduction
shape,12 and is an important determinant of behavior in different
Cyclization of a linear molecule into a macrocyclic ring constitutes
a significant change in molecular shape. This transformation
physicochemical or biological environments. Previous studies
comparing the diffusion coefficients of macrocyclic versus linear
restricts the degrees of conformational freedom of the molecule,1
and imposes structural organization which was absent in the
linear precursor. The introduction of this conformational con-
straint comes at a synthetic cost, since macrocyclization reac-
tions are typically conducted under high dilution conditions
in order to avoid competing dimerization reactions.2,3 The pre-
organized shape of macrocycles has been used extensively in
the fields of molecular recognition,4 supramolecular chemistry,5
and molecular self-assembly.6 They are found widely in natural
products,7 and a number of natural product-derived macrocycles,
such as cyclosporin,8 are used as therapeutic agents, where they
appear to defy conventional wisdom regarding the relationship
between molecular size and drug-like behavior.9 Consequently,
there is growing interest in the design of macrocycle-based drugs,10
particularly as a strategy for addressing those challenging molec-
ular targets which involve an interaction between large protein
interfaces,11 where it might be necessary to deploy molecules well
outside the conventional drug-like molecular weight range.
molecules have compared either polymeric systems,13 or macrocy-
cles generated via chemical cyclization of a linear precursor.14 In
the former case, the molecules are comprised of simple, repeating
units, e.g. alkanes and siloxanes, which lack the range of functional
groups typically found in biologically active molecules. In the latter
case, the creation of the macrocycle entailed significant changes in
structure, such as the number of hydrogen-bonding groups present
and the overall charge of the molecule. Since these will affect the
interaction with the surrounding medium, they do not provide an
unambiguous assessment of the direction and magnitude of the
resultant change in diffusion coefficient for the system.
In order to explore the impact of macrocyclization on diffusion
behavior in molecules more representative of biologically active
agents, whilst avoiding the confounding effects of unintended
structural changes, we have examined diffusion rates in a carefully
matched set of macrocycles and linear analogues. We designed
a series of macrocycles which would overlap with, and extend
beyond, the molecular weight-range usually associated with drug
molecules, which would possess functional groups commonly
found in biologically active agents, and which provided an
approximately linear relationship between molecular weight and
ring size. We also designed a corresponding set of ‘acyclic controls’,
generated by carbogenic fission of each macrocycle,15 which
provided an almost identical match with regard to functionality
and molecular weight. Herein we report the results of studies
comparing the diffusion behavior of these two types of molecule in
three different solvent systems, representing a range of viscosities
and dielectric environments, using the bipolar pulse longitudinal
eddy current delay (BPPLED) NMR method.16
As part of a program exploring the use of macrocycle-based
systems to address biological problems, we have begun to charac-
terise the differences in physicochemical properties of macrocycles
in comparison to their linear counterparts. We were especially
interested in determining the impact of macrocyclization on
aDepartment of Chemistry, The Scripps Research Institute, 10550 N. Torrey
Pines Road, La Jolla, California, USA. E-mail: kjames@scripps.edu;
Fax: +1 858 784 7550; Tel: +1 858 784 2507
bPfizer Worldwide Research, Sandwich, Kent, CT13 9NJ, UK
† Electronic supplementary information (ESI) available: Full synthetic
procedures, compound characterization, 1H and 13C NMR spectra and
NMR diffusion data. See DOI: 10.1039/c1ob05996c
This journal is
The Royal Society of Chemistry 2011
Org. Biomol. Chem., 2011, 9, 7727–7733 | 7727
©

View Article Online
our design an approximately linear relationship between molecular
weight and ring size.
Results and discussion
Macrocycle and acyclic control design
We also designed the systems to feature drug-like functional
groups (amides, ethers, aryl/heteroaryl rings and stereocenters)
and physicochemical properties. Fig. 2b illustrates one important
measure of the drug-like properties of these systems, lipophilicity,
which reflects the affinity of a molecule for a lipid environ-
ment. The lipophilicity of the macrocyclic set, as reflected in
their logD7.4 values,18 lies within a typical range for drug-like
molecules.19
We designed a series of macrocycles and acyclic controls which
covered 12- to 29-membered rings and molecular weights from 300
to 730 respectively, and which could be synthesized efficiently using
copper catalyzed azide–alkyne cycloaddition (CuAAC) reactions
under flow, which we have reported recently (Fig. 1).17 This
range of ring sizes encompassed the most common size found
in a recent survey of natural products (14-membered),7 and
approached the size of cyclosporin itself (33-membered).8 The
molecular weight range extended well beyond that reflected in well-
established guidelines for orally bioavailable drugs (500 g mol-1).9
As illustrated in Fig. 2a, in order to maximize the chance of
detecting a small effect on diffusion behavior, we also built into
The acyclic controls were designed to represent a carbogenic
fission of their macrocyclic counterpart,15 which fully preserved
the functionality of the macrocycle and yielded an almost identical
MW. Thus, alkyl fragments were incorporated within the macrocy-
cle framework to allow at least one carbon–carbon bond-scission
Fig. 1 Matched pairs of macrocycles and acyclic controls. Ring size indicated within macrocyclic ring. Molecular weight indicated in parenthesis.
Fig. 2 (a) A plot of ring size versus MW for macrocycle series, R2 = 0.90 for linear fit. (b) A plot of logD7.4 versus MW for matched pairs of macrocycles
(circles) and acyclic controls (squares). Each matched pair shares approximately the same MW.
7728 | Org. Biomol. Chem., 2011, 9, 7727–7733
This journal is
The Royal Society of Chemistry 2011
©

View Article Online
Fig. 3 Diffusion data comparing macrocycles (circles) and acyclic controls (squares). The triangle denotes data point for the des-methyl analog of 2a.
Error bars are obscured by the data points. (a) A plot of the diffusion coefficient in CDCl3 (DCDCl ) versus MW. (b) A plot of the diffusion coefficient in
3
d6-DMSO (DDMSO) versus MW. (c) A plot of log(DCDCl ) versus log(MW). (d) A plot of log(DDMSO) versus log(MW); R2 for logarithmic plots combining all
3
data across both solvents = 0.99, p < 0.001.
which did not generate a significant functional group change. The
success of this approach is evident in the close correspondence
of logD7.4 values for the macrocyclic and acyclic control pairs,
indicating that there has been no significant change in lipophilicity
upon macrocyclization (Fig. 2b).
Diffusion studies
Diffusion coefficients for the entire set of molecules were mea-
sured in 10 mM CDCl3 and d6-DMSO solutions, using the
bipolar pulse longitudinal eddy current delay (BPPLED) method
(see Supporting Information for full experimental details†).16
These solvents were chosen as models for the low (CDCl3, 4.8)
and high (d6-DMSO, 46.7) dielectric environments encountered in
physiological systems, whilst enabling complete dissolution of all
the compounds examined at the necessary concentrations, which
was not possible in pure D2O. These solvents also afforded two
different viscosities for the diffusion medium, in order to confirm
that any effects on the diffusion rate were independent of this
parameter.
The observed diffusion coefficients were highly reproducible
(see Supporting Information for full set of diffusion data†). Fig. 3
depicts the observed diffusion coefficients plotted versus molecular
weight, for CDCl3 and d6-DMSO. The results demonstrate an
expected inverse relationship between MW and diffusion co-
efficient in both solvent systems. Most importantly, they also
Macrocycle and acyclic control synthesis
The macrocycles were synthesized via a copper catalyzed azide–
alkyne cycloaddition (CuAAC), or ‘click’ reaction, conducted in
flow,17 typically affording >100 mg quantities of product under
modest dilution (0.017 M) conditions. The azido–alkyne precursor
for each macrocycle was synthesized in a short, modular synthetic
sequence from commercially available, homochiral fragments. The
acyclic controls were also synthesized in flow via in situ generation
of each alkyl–azide reactant,20 utilizing common intermediates
from the corresponding macrocycle synthesis wherever possible
(see Supporting Information for synthetic sequences and full
experimental details†).
This journal is
The Royal Society of Chemistry 2011
Org. Biomol. Chem., 2011, 9, 7727–7733 | 7729
©

View Article Online
illustrate a clear increase in diffusion rate for the macrocyclic series
versus the acyclic controls, which is highly statistically significant
(5% increase for macrocycle series, R2 > 0.99, p < 0.001 for the
combined data across both solvents; see experimental section for
details of statistical treatment). In fact, every macrocycle examined
exhibits a higher diffusion rate than its acyclic counterpart. The
diffusion rates are some 3.5-fold higher in CDCl3, compared to
d6-DMSO, reflecting the higher viscosity term h for d6-DMSO
(2.0 cP) versus CDCl3 (0.54 cP) in the Stokes–Einstein equation:12,21
diffusion in CDCl3.25 The consistency of slope, regardless of
solvent or molecular series, is noteworthy.
Solvent effects
It appears in Fig. 3b and 3d that the components of matched
pair 2 (MW 340 and 342), although fitting the general trend of
higher diffusion rate for the macrocycle, are outliers from the line
in d6-DMSO. This is the only pair which does not possess an
amide N–H group, the N-methyl analog having been chosen to
provide a larger MW increment over pair 1. Since pair 2 is not
an outlier in CDCl3 (Fig. 3a), we hypothesized that in d6-DMSO,
the absence of an amide N–H group meant less interaction with
solvent and therefore a resultant boost in diffusion rate compared
to closely related molecules which do possess an amide N–H
group. The corresponding des-methyl analog,26 of macrocycle 2a
was therefore examined in both CDCl3 and d6-DMSO. Pleasingly,
the measured diffusion coefficient for this molecule in d6-DMSO
was lower (Fig. 3b and d) and more in line with the other members
of the macrocyclic series, thereby yielding an improved fit for the
curve (R2 = 0.99 versus 0.97, when the des-methyl analog was
substituted for 2a; data included in Supporting Information†).
This result is consistent with reports of lower diffusion rates in
d6-DMSO versus CDCl3 for molecules which can form hydrogen-
bonds to DMSO.27 It also highlights the challenges of designing a
compound test set which provides smooth structural variation over
a wide MW range, as well as access to relevant controls and which
can be synthesized efficiently. It also indicates that macrocycles
of much different structure or polarity compared to those studied
here might exhibit a different y-intercept for the graphs of diffusion
coefficient versus MW.
In order to confirm that these shape-driven changes in diffusion
coefficient would also be observed under more physiological
conditions, we measured the diffusion coefficients of selected pairs
in an aqueous medium, adding the minimum possible quantity of
DMSO in order to maintain solubility. Thus, Fig. 4 shows the
diffusion coefficients for pairs 1, 3, 5 and 8 in a 2.5 mM D2O : d6-
DMSO (9 : 1) solution. The macrocyclic examples show the same
statistically significant increase in diffusion rate compared to their
matched acyclic controls as we had observed in the other solvent
systems.
Comparing the diffusion coefficients for this set of four
macrocycles (1a, 3a, 5a and 8a) across the three different solvent
systems examined, it is apparent that the diffusion coefficients
under aqueous conditions fall between those measured in CDCl3
and d6-DMSO, as would be expected based upon their viscosities
(Fig. 5). Thus, we feel confident that the differences in diffusion
rates we have observed for the macrocyclic versus acyclic control
series would be observed in any media which possess a viscosity
in this range.
kT
D =
6phrH
The data also demonstrate that, for these examples, the effect is
present at both low MW (300) and high MW (730), well beyond
the Lipinski Rule-of-Five range.9 The effect is modest in absolute
terms, averaging a change of +5% in CDCl3, and translates to
a macrocycle of MW 550 exhibiting the diffusion behavior of a
linear molecule of MW 500.
Our findings differ from those of earlier studies in terms
of the magnitude of observed increase in diffusion rates upon
macrocyclization. Thus, a comparison of the diffusion rates for
macrocyclic versus linear siloxane polymers in toluene found a
ratio of 0.85 0.01 for the diffusion coefficients Dl/Dr of the
linear (l) and ring (r) molecules respectively.13b This is within
experimental error of the theoretical ratio of 8/3p (0.85) predicted
for cyclic versus linear molecules in the absence of free-draining
and excluded volume effects.13d In comparison, the average ratio
Dl/Dr in CDCl3 for the macrocycles and acyclic controls in our
study is 0.95 0.01, suggesting that excluded volume effects are
more significant in these more drug-like molecules compared
to simple polymers. Interestingly, the ratio of sedimentation
coefficients sl/sr for linear and nicked DNA for polyoma virus,22
and lDNA,23 was found to be 0.91 and 0.88 respectively. Since
sedimentation coefficient s is directly proportional to diffusion
coefficient D,13b this ratio is directly comparable to Dl/Dr. These
ratios for cyclic versus linear DNA molecules therefore fall between
the value that we have observed and the value reported for cyclic
versus linear polydimethylsiloxane.
The closeness of the ratio Dl/Dr to unity for our experiments
suggests that the acyclic molecules probably do not exist in a fully
extended, ‘rod-like’, conformation, and instead undergo some
form of hydrophobic collapse to a more compact structure.24
Furthermore, the macrocyclic systems, despite the conformational
constraint imposed by the ring, probably still retain a considerable
degree of flexibility. Nevertheless, the consistency of the effect is
striking and whilst the differences in diffusion coefficients between
macrocycles and acyclic controls are small, they are statistically
significant.
As illustrated in Fig. 3a and 3b, a logarithmic plot of the data
from both series of molecules and in both solvents is a straight
line, which fits the equation:
Comparing the slopes of the lines in Fig. 5, whilst the slopes
in CDCl3 and d6-DMSO are the same, the slope of the line in
D2O : d6-DMSO (9 : 1) is more shallow (statistically significant
difference), indicating that in this medium there is less reduction
in diffusion coefficient as molecular weight increases. This could
be a consequence of the higher dielectric nature of this aqueous
system, resulting in more hydrophobic collapse of the lipophilic
substituents in the larger, more flexible molecules,24 and therefore
D = C(MW) - 0.56 ( 0.079)
(where D = diffusion coefficient and C is a constant) extremely
well (R2 = 0.99, p < 0.001). This quantitative relationship between
the logarithm of diffusion coefficient and the logarithm of MW
is consistent with previous studies examining small molecule
7730 | Org. Biomol. Chem., 2011, 9, 7727–7733
This journal is
The Royal Society of Chemistry 2011
©

View Article Online
Fig. 4 Diffusion data comparing macrocycles (circles) and acyclic controls (squares). a) A plot of the diffusion coefficient in D2O : d6-DMSO (9 : 1)
(DH2O) versus MW. b) A plot of log(DH2O) versus log(MW); R2 for logarithmic plots = 0.99, p < 0.001.
macrocycles are very similar. There appears to be no support for
a model in which the acyclic systems exhibit an extended, rod-
like structure and the macrocycles possess a compact, spheroidal
structure.
Although the systems examined here point the way towards
a general strategy for increasing the diffusion rate of drug-like
structures via macrocyclization, they are still relatively simple
models of the therapeutically important macrocyclic natural
products such as cyclosporin. Thus, whilst they are macrocyclic
and possess a similar lipophilicity, they are still lower in molecular
weight, and perhaps most importantly, they possess neither the
superb network of trans-annular intramolecular hydrogen bonds
observed in this natural product, nor its array of N-methylated
amide bonds.8 We suspect therefore that a macrocyclic structure,
whilst important to the success of such agents, constitutes just one
component in a set of structural features which operate in concert
to confer drug-like behavior.
Future studies will therefore examine still higher MW systems,
as well as the impact of introducing intramolecular hydrogen-
bonds, in an effort both to constrain these systems further and
‘conceal’ polar functional groups within the molecule. We will also
expand our studies beyond the measurement of diffusion rates to
Fig. 5 A plot of the logarithms of the diffusion coefficients in CDCl3
(circles), D2O : d6-DMSO (9 : 1; triangles) and d6-DMSO (squares) for
macrocycles 1a, 3a, 5a and 7a versus the logarithm of MW.
include the impact of macrocyclization on passage through both
synthetic and biological membranes.
(in comparison to the other solvents) more compact structures
with lower hydrodynamic radius and higher diffusion coefficient.
Experimental procedures
Conclusions
General considerations
We have shown that macrocyclization of polyfunctional, non-
polymeric molecules in the molecular weight range 300–730,
generating systems with 12- to 29-membered rings, results in a
statistically significant increase in diffusion coefficient. This effect
is evident across this entire range of molecular weight and ring size
examined. It also appears to be independent of solvent viscosity
and solvent dielectric over the range studied. The magnitude of
increase in diffusion coefficient for the macrocyclic series is smaller
than observed in simpler, more uniform oligomers such as alkanes,
poly(dimethoxysiloxanes) or even DNA. This suggests that for
this set of matched pairs, the molecular volumes occupied by
the ensemble of conformations for the linear systems and the
All reagents and solvents were used as received. THF was distilled
from sodium. NMR spectra were recorded on Bruker DRX-600,
Bruker DRX-500, or Bruker AMX-400 instruments using residual
solvent peak as a reference. Data are reported as s = singlet,
d = doublet, t = triplet, q = quartet, m = multiplet, br = broad.
LC/MS analyses were carried out using an Agilent Technologies
HPLC (Agilent Technologies 1100 Series diode array detector,
Agilent Technologies 1100 Series column heater, Agilent 1100
Series pump, and Agilent 1100 Series degasser) interfaced with an
Agilent Technologies 6110 Quadrupole LC/MS. Column chro-
matography was performed using a Biotage Horizon automated
This journal is
The Royal Society of Chemistry 2011
Org. Biomol. Chem., 2011, 9, 7727–7733 | 7731
©

View Article Online
flash chromatography system equipped with a Biotage Horizon
detector, fraction collector and pump where noted.
been averaged for each run in order to derive the diffusion
coefficient and a standard deviation. Statistical analysis confirms
that the variation in diffusion coefficient within each run is the
same as the comparison between runs. Confidence intervals were
determined based upon the observation that measurements on all
the compounds were of equal precision, therefore a single estimate
of variance was derived based upon the between-run differences
for all compounds. The analysis demonstrates that a difference in
diffusion coefficients between macrocycles and linear controls of
1% would be statistically significant. The observed difference of
5% is highly statistically significant.
TMS was included in the CDCl3 and DMSO datasets as a
control to check temperature was consistent in all experiments.
As can be seen in the table of diffusion data, this is very
reproducible and allows confidence in the comparison of data
between macrocycles and linear analogues.
NMR diffusion coefficient determination
All the NMR experiments were acquired on a Bruker Avance
600 MHz spectrometer, equipped with a 5 mm TCI cryoprobe
with z-gradients capable of generating 54 G/cm field strengths.
The temperature controller was set to 298 K with an air flow of
535 l h-1 in order to avoid any temperature fluctuations due to
sample heating during acquisition and to avoid sample vibrations
from a high air-flow. Samples were made up to 10 mM solutions in
DMSO-d6 or CDCl3 with some TMS vapour, and 180 uL of this
solution was added to a 3 mm NMR tube to avoid problems of
convection. In the case of samples in 10% DMSO in D2O, samples
were made up at 2.5 mM and 600 mL of this solution was added
to a 5 mm NMR tube. The lower concentration was required due
to the limited solubility in this solvent. To maintain good signal
to noise, the use of 5 mm NMR tubes is possible for more viscous
solvents, as the onset of Rayleigh–Benard convection is ablated at
the temperatures used during this investigation.
Diffusion coefficients were determined with a high degree of
reproducibility on the NMR system used. Using robust statistical
analysis, we have determined that the relative differences in diffu-
sion between the linear and macrocyclic analogues are significant.
However, the absolute accuracy of the diffusion coefficient as
measured by NMR is limited by the accuracy of the calibration of
the gradient and temperature, and often prone to errors introduced
during the calculation method. Steps were taken to minimise these
errors as detailed, however, it would be difficult to accurately
compare diffusion coefficients between different NMR systems.
The gradient strength was calibrated using the diffusion coefficient
of water in a standard solution of 0.1 mg ml-1 GdCl3, 0.1% DSS,
1% H2O in D2O. The values of the measured diffusion coefficient
(D) of water, the known diffusion coefficient of water and the
current gradient calibration value (gc(old)) were used to obtain
the new gradient calibration value (gc(new)) using the following
equation:28
General procedure for preparative scale flow macrocyclization
Azido–alkyne (0.10 M in EtOH, 100 mL, 0.010 mmol, 1.0 eq),
TTTA (0.01 M in EtOH, 100 mL, 0.001 mmol, 0.10 eq), DIPEA
(0.1 M in EtOH, 200 mL, 0.020 mmol, 2.0 eq) and EtOH (200 mL)
were aspirated from their respective source vials, mixed through
a PFA mixing tube (0.2 mm inner diameter), and loaded into an
injection loop. The reaction segment was injected into the flow
reactor set at 150 ◦C, passed through the reactor at 300 mL min-1
(5 min residence time). A total of 40 reaction segments prepared
in this manner were collected in a round bottom flask. Upon
completion, the reaction mixture was concentrated and dried in
vacuo. The crude reaction mixture was purified using a Biotage
Horizon automated flash column chromatography system (silica
gel, EtOAc, Rf 0.17) to yield 1a as a white solid (103.8 mg, 87%
yield): 1H NMR (600 MHz, CD3OD): d 8.06 (br. s., 1 H), 7.64 (br.
s., 1 H), 7.26–7.39 (m, 4 H), 7.19–7.25 (m, 1 H), 4.88 (d, J = 13.2
Hz, 1 H), 4.73 (br. s., 1 H), 4.39 (br. s., 1 H), 4.21 (d, J = 13.2 Hz,
2 H), 3.53 (d, J = 6.6 Hz, 1 H), 2.20–2.58 (m, 3 H), 1.97 (br. s.,
1 H), 0.79 (d, J = 6.6 Hz, 3 H); 13C NMR (150 MHz, CD3OD):
d 173.3, 147.3, 141.0, 129.5, 128.4, 128.3, 127.4, 77.5, 63.3, 52.8,
52.2, 34.2, 26.1, 11.6; HRMS (ESI-TOF): C16H20N4O2: [M + H]+:
calculated 301.1659, found 301.1664.
gc(new) = gc(old) ¥ sqrt(D(measured)/D(known))
The temperature was calibrated with a sample of methanol-
d4 (99.8 at%), sealed under atmospheric pressure, using details
described elsewhere.29
General procedure for preparative scale flow intermolecular
CuAAC reaction
All DOSY experiments used the ledbpgp2s sequence (available
in standard Bruker pulse sequence library). A gradient duration
(d) of 2 ms and an eddy current delay of 5 ms was used in all
cases. The diffusion time (D) was 100 ms in the case of CDCl3,
and 200 ms in the case of d6-DMSO and 10% d6-DMSO in D2O.
In each PFG NMR experiment, a series of 16–32 spectra on 16
K data points were collected, using a linear gradient ramp from
5–95% of the maximum gradient strength.
After acquisition, the data was zero filled to 32k, Fourier
transformed and baseline corrected in f2. The diffusion coefficients
were calculated with the T1/T2 relaxation module using mono-
exponential fitting, rather than the 2D processing protocol. This
is available in Bruker Topspin v.2. For each sample, several
well resolved signals were used to extract individual diffusion
coefficients (an example of the raw data and the fitting report
is included in the Supporting Information†). These signals have
Alkyne (0.25 M in DMF, 150 mL, 0.038 mmol, 1.0 eq), iodoethane
(0.50 M in DMF, 150 mL, 0.075 mmol, 2.0 eq), NaN3 (0.30 M in
DMF, 187.5 mL, 0.056 mmol, 1.5 eq) and DIPEA (0.5 M in DMF,
75 mL, 0.038 mmol, 1.0 eq) were aspirated from their respective
source vials, mixed through a PFA mixing tube (0.2 mm inner
diameter), and loaded into an injection loop. T◦he reaction segment
was injected into the flow reactor set at 150 C, passed through
the reactor at 150 mL min-1 (10 min residence time). A total of
10 reaction segments prepared in this manner were collected in
a round bottom flask. Upon completion, the reaction mixture
was concentrated and dried in vacuo. The crude reaction mixture
was purified using a Biotage Horizon automated flash column
chromatography system (silica gel, 5% MeOH in CH2Cl2, Rf 0.17)
followed by PTLC (silica gel, 500 mm plates, 5% MeOH in CH2Cl2)
1
to yield 1b as an off-white solid (78.5 mg, 69% yield): H NMR
7732 | Org. Biomol. Chem., 2011, 9, 7727–7733
This journal is
The Royal Society of Chemistry 2011
©

View Article Online
(600 MHz, CDCl3): d 7.44 (s, 1 H), 7.16–7.35 (m, 5 H), 6.56 (d,
J = 7.2 Hz, 1 H), 4.73 (d, J = 13.8 Hz, 1 H), 4.39 (d, J = 2.4 Hz, 1
H), 4.33–4.43 (m, 3 H), 4.15 (m, 1 H), 1.91 (s, 3 H), 1.52 (t, J = 7.2
Hz, 3 H), 0.91 (d, J = 6.6 Hz, 3 H); 13C NMR (150 MHz, CDCl3):
d 169.7, 144.9, 138.7, 128.6, 127.8, 126.7, 121.9, 82.8, 62.7, 50.2,
45.5, 23.5, 15.6, 13.5; HRMS (ESI-TOF): C16H22N4O2: [M + H]+:
calculated 303.1815, found 303.1823.
9 C. A. Lipinski, F. I. Lombardo, B. W. Dominy and P. J. Feeney, Adv.
Drug Delivery Rev., 1997, 23, 3.
10 (a) E. M. Driggers, S. P. Hale, J. Jinbo Lee and N. K. Terrett, Nat. Rev.
Drug Discovery, 2007, 7, 608; (b) E. Marsault and M. L. Peterson, J.
Med. Chem., 2011, 54, 1961.
11 J. A. Wells and L. McClendon, Nature, 2007, 450, 1001.
12 A. Macchioni, G. Ciancaleoni, C. Zuccaccia and D. Zuccaccia, Chem.
Soc. Rev., 2008, 37, 479.
13 (a) R. Ozisik, E. D. von Meerwall and W. L. Mattice, Polymer, 2002, 43,
629; (b) C. J. C. Edwards, R. F. T. Stepto and J. A. Semlyen, Polymer,
1980, 21, 781; (c) C. J. C. Edwards, R. F. T. Stepto and J. A. Semlyen,
Polymer, 1982, 23, 865; (d) M. Fukatsu and M. Kurata, J. Chem. Phys.,
1966, 44, 4539.
14 (a) F. W. Okumu, G. M. Pauletti, D. G. Vander Velde, T. J. Siahaan and
R. T. Borchardt, Pharm. Res., 1997, 14, 169; (b) B. Wang et al., J. Pept.
Res., 1999, 53; (c) H. T. He, C. R. Xu, X. Song and T. J. Siahaan, J.
Pept. Res., 2003, 61, 331.
15 For recent examples of generating acyclic controls of macrocycles
using this strategy see: (a) D. G. Udugamasooriya and M. R. Spaller,
Biopolymers, 2008, 89, 653; (b) J. E. DeLorbe, J. H. Clements, B. B.
Whiddon and S. F. Martin, ACS Med. Chem. Lett., 2010, 1, 448.
16 D. Wu, A. Chen and C. S. Johnson, J. Magn. Reson., Ser. A, 1995, 115,
260.
17 A. R. Bogdan and K. James, Chem.–Eur. J., 2010, 16, 14506.
18 logD7.4 is defined as the logarithm of the partition coefficient between
1-octanol and an aqueous buffer at pH 7.4.
19 P. D. Leeson and J. R. Empfield, Annu. Rep. Med. Chem., 2010, 45, 393.
20 A. R. Bogdan and N. Sach, Adv. Synth. Catal., 2009, 351, 849.
21 D = diffusion coefficient, T = temperature, k = Boltzmann constant, h =
solvent viscosity and rH = hydrodynamic radius.
22 J. Vinograd, J. Liebowitz, R. Radloff and P. Laipis, Proc. Natl. Acad.
Sci. U. S. A., 1965, 53, 1104.
23 A. D. Hershey, E. Burgi and L. Ingram, Proc. Natl. Acad. Sci. U. S. A.,
1963, 49, 748.
24 D. H. Rich, Perspect. Med. Chem., 1993, 15.
25 S. Auge´ et al., J. Phys. Chem. B, 2009, 113, 1914.
26 See ref. 17 for details of the preparation of this compound.
27 G. S. Kapur, E. J. Cabrita and S. Berger, Tetrahedron Lett., 2000, 41,
7181.
Acknowledgements
This work was supported by a grant from Pfizer Pharmather-
apeutics Research. The authors would like to thank Phil Brain
(Pfizer) for statistics support and Lee Roberts (Pfizer) for advice
and valuable discussions.
Notes and references
1 N. Go and H. Scheraga, H. A., Macromolecules, 1970, 3,
178.
2 L. A. Wessjohann and E. Ruijter, Top. Curr. Chem., 2005, 243,
137.
3 G. Illuminati and L. Mandolini, Acc. Chem. Res., 1981, 14, 95.
4 For selected examples see: (a) M. Krishnamurthy, K. Simon, A. M.
Orendt and P. A. Beal, Angew. Chem., Int. Ed., 2007, 46, 7044; (b) A.
Kumara, S.-S. Suna and A. J. Lees, Coord. Chem. Rev., 2008, 252, 922;
(c) T. Ema, D. Tanida, K. Hamada and T. Sakai, J. Org. Chem., 2008,
73, 9129.
5 For selected examples see: (a) A. Moretto, M. Crisma, F. Formaggio
and T. Claudio, Biopolymers, 2010, 94, 721; (b) M.-X. Wang, Chem.
Commun., 2008, 4541; (c) M. J. MacLachlan, Pure Appl. Chem., 2006,
78, 873.
6 For selected examples see: (a) P. Ballester et al., J. Org. Chem., 2005, 70,
6616; (b) C. A. Schalley, H. T. Baytekin and B. Baytekin, Macrocyclic
Chemistry, 2005, 37; (c) K. Balakrishnan et al., J. Am. Chem. Soc.,
2006, 128, 6576.
7 W. Brandt, V. J. Haupt and L. A. Wessjohann, Curr. Top. Med. Chem.,
2010, 10, 1361.
28 Bruker, DOSY and Diffusion by NMR, A Tutorial for Topspin 2.0,
Version 2.0.0, Rheinstetten, Germany, 2006.
29 M. Findeisen, T. Brand and S. Berger, Magn. Reson. Chem., 2007, 45,
175.
8 R. M. Wenger, Progress in the Chemistry of Organic Natural Products,
1986, 50, 123.
This journal is
The Royal Society of Chemistry 2011
Org. Biomol. Chem., 2011, 9, 7727–7733 | 7733
©