A solvatochromic model calibrates nitriles' vibrational frequencies to electrostatic fields

Article abstract of DOI:10.1021/ja303895k

Electrostatic interactions provide a primary connection between a protein's three-dimensional structure and its function. Infrared probes are useful because vibrational frequencies of certain chemical groups, such as nitriles, are linearly sensitive to local electrostatic field and can serve as a molecular electric field meter. IR spectroscopy has been used to study electrostatic changes or fluctuations in proteins, but measured peak frequencies have not been previously mapped to total electric fields, because of the absence of a field-frequency calibration and the complication of local chemical effects such as H-bonds. We report a solvatochromic model that provides a means to assess the H-bonding status of aromatic nitrile vibrational probes and calibrates their vibrational frequencies to electrostatic field. The analysis involves correlations between the nitrile's IR frequency and its ¹³C chemical shift, whose observation is facilitated by a robust method for introducing isotopes into aromatic nitriles. The method is tested on the model protein ribonuclease S (RNase S) containing a labeled p-CN-Phe near the active site. Comparison of the measurements in RNase S against solvatochromic data gives an estimate of the average total electrostatic field at this location. The value determined agrees quantitatively with molecular dynamics simulations, suggesting broader potential for the use of IR probes in the study of protein electrostatics.

Full text of DOI:10.1021/ja303895k

Communication
pubs.acs.org/JACS
A Solvatochromic Model Calibrates Nitriles’ Vibrational Frequencies
to Electrostatic Fields
Sayan Bagchi,‡ Stephen D. Fried,‡ and Steven G. Boxer*
Department of Chemistry, Stanford University, Stanford, California 94305-5080, United States
S
* Supporting Information
Moreover, a nitrile’s frequency, νCN, is impacted both by
̅
electrostatic fields and by hydrogen bonding.12,13 When a nitrile
ABSTRACT: Electrostatic interactions provide a primary
connection between a protein’s three-dimensional struc-
ture and its function. Infrared probes are useful because
vibrational frequencies of certain chemical groups, such as
nitriles, are linearly sensitive to local electrostatic field and
can serve as a molecular electric field meter. IR
spectroscopy has been used to study electrostatic changes
or fluctuations in proteins, but measured peak frequencies
have not been previously mapped to total electric fields,
because of the absence of a field-frequency calibration and
the complication of local chemical effects such as H-bonds.
We report a solvatochromic model that provides a means
to assess the H-bonding status of aromatic nitrile
vibrational probes and calibrates their vibrational frequen-
cies to electrostatic field. The analysis involves correlations
between the nitrile’s IR frequency and its 13C chemical
shift, whose observation is facilitated by a robust method
for introducing isotopes into aromatic nitriles. The method
is tested on the model protein ribonuclease S (RNase S)
containing a labeled p-CN-Phe near the active site.
Comparison of the measurements in RNase S against
solvatochromic data gives an estimate of the average total
electrostatic field at this location. The value determined
agrees quantitatively with molecular dynamics simulations,
suggesting broader potential for the use of IR probes in the
study of protein electrostatics.
probe is H-bonded, it is particularly difficult to determine from
IR measurements what electric field it is experiencing. Upon
accepting an H-bond, νCN shifts to the blue and broadens.12−16
̅
The shift has been shown theoretically to depend on the
distance and the angle between the nitrile and the proton on
the H-bond donor,14 in a manner that is not described by the
vibrational Stark effect.
Recently, we reported a technique17 to decompose IR shifts
of nitriles into H-bonding and electrostatic components based
on an analysis that requires both νCN and the 13C chemical shift,
̅
13
δ
CN, of the nitrile carbon, effectively using these two
observations to specify two unknowns. We initially applied
this approach to the cysteine thiocyanate probe, Cys-SCN,
which can be prepared by cyanylating cysteine with KCN.18
The synthetic procedure readily allowed for incorporation of
isotopically labeled 13CN, which is necessary for measuring
13CN
13CN
δ
in the protein. Knowledge of both νCN and δ
for the
̅
nitrile probe in situ was used to determine whether a series of
Cys-SCN probes installed at the active site of the enzyme
ketosteroid isomerase were H-bonded.7,17 Aromatic nitriles can
also be introduced into proteins, either as p-CN-Phe6,8,10 or as
a common functional group on inhibitors, including many
drugs, that bind to the active site of enzymes and signaling
proteins.4,5,11 For these reasons, we sought to generalize the
methods and analysis previously described for cysteine
thiocyanates17,18 to aromatic nitriles, beginning with a simple
synthetic route for incorporating 13CN into aromatic nitriles, in
particular p-13CN-Phe, and then testing whether similar
behavior is observed that would facilitate distinguishing H-
bonded from non-H-bonded aromatic nitriles.
ecent advances toward characterizing the complex
R
molecular environment present inside proteins and
biological macromolecules have focused on the use of infrared
probes.1,2 One of the important strengths of IR-based
methodologies is the linear dependence of vibrational
frequencies on an electrostatic field, a phenomenon known as
the vibrational Stark effect.3,4 A number of studies have
employed extrinsic IR probes as a molecular tool to measure
electrostatic field changes inside proteins.4−7 In particular, the
nitrile stretching mode has been proposed as an ideal IR probe
for investigations of protein structure4−10 and dynamics,11
because it is a local mode in an uncluttered region of the IR
spectrum that is particularly sensitive to its local electrostatic
field.3,5,7
Our analysis is based on solvatochromic trends of
benzonitrile (PhCN), a model compound used to represent
p-CN-Phe. As seen in Figure 1A, when PhCN is dissolved in
various non-H-bonding solvents, νCN varies from 2233.4 cm−1
̅
in the most nonpolar solvent (hexane) to 2227.6 cm−1 in the
most polar solvent (DMSO). This trend can largely be
attributed to the effective electric field created by the
polarization of the solvent. Using a simple analytic model
such as the Onsager reaction field approach,19,20 the magnitude
of the electric fields due to each solvent can be approximated
(the x-axis of Figure 1A).21 The Onsager equation estimates the
total electric field due to dielectric polarization, so it is unable
to describe chemical interactions such as H-bonds; however, it
While the vibrational Stark effect allows one to quantify the
relationship between changes in electrostatic field and an
accompanying IR frequency shift, e.g., in response to mutation
or pH change, it does not provide a calibration to associate a
particular total electrostatic field with any observed frequency.
Received: April 23, 2012
Published: June 13, 2012
© 2012 American Chemical Society
10373
dx.doi.org/10.1021/ja303895k | J. Am. Chem. Soc. 2012, 134, 10373−10376

Journal of the American Chemical Society
Communication
Figure 1. Solvatochromism calibrates the sensitivity of the C−N stretch and the 13C chemical shift of benzonitrile to electric field and H-bonding.
(A) νCN of PhCN in various non-H-bonding solvents (colored circles) and in water (black square) compared against the electric field each solvent
̅
2
13
exercises on PhCN (see footnote 20). νCN = 2235.2 + 0.597F; R = 0.86 (excluding water). (B) δ CN of PhCN in various non-H-bonding solvents
̅
(colored circles) and in water (black square) compared similarly. δ CN = 114.12 − 0.561F; R2 = 0.66. (C) Comparison between νCN and δ CN. νCN
=
13
13
̅
̅
2331.7 − 0.854δ CN; R2 = 0.60. All non-H-bonding solvents follow an approximate linear trend, but water is significantly removed from that region.
H-bonding is responsible for blue-shifting PhCN’s frequency in water away from what would be expected based on electrostatics. Three additional
points (brown circles) are compared against the solvatochromic model: amino acid p-CN-Phe (3), [p-CN-Phe]S-peptide (4), and [p-CN-
Phe]RNase S (5), all in 20 mM HEPES, pH 8.0. 3 and 4 fall far away from the electrostatic line, consistent with the nitrile being H-bonded, while 5
falls near the electrostatic line, suggesting an absence of H-bonding. Inset shows 13C NMR spectrum of 5.
13
qualitatively describes the electric field caused by non-H-bond-
donating solvents. Electric fields so calculated display a good
correlation to observed frequencies (R2 = 0.86); moreover, the
slope of the best-fitting line (0.60 cm−1/(MV/cm)) shows
excellent agreement with the independently measured Stark
tuning rate of PhCN (0.61 cm−1/(MV/cm)),9,22 which is the
frequency shift per unit field found when an external electric
field is applied to the compound. This agreement reinforces the
view that the solvatochromic trend is reporting largely on
electric fields. On the other hand, the frequency of PhCN in
water (the black square in Figure 1A) does not follow the
trend, consistent with an additional non-electrostatic effect
associated with the nitrile being H-bonded.12,14,17 A parallel
To selectively label the carbon in the nitrile of p-13CN-Phe,
we adopted the Rosenmund−von Braun reaction,23 which
exchanges the iodine in aromatic iodides for a nitrile using a
cuprous reagent. As shown in Scheme 1 (synthetic methods are
Scheme 1. Incorporation of Isotopically Labeled Nitriles into
a
Amino Acids, Peptides, and Proteins
13CN
study on δ
of PhCN in the same solvents shows a similar
correlation (albeit weaker) with the calculated electric field
13
(Figure 1B). In this case, unlike νCN in water, δ
in water is
CN
̅
relatively similar to its value in another very polar solvent
(DMSO). In the context of the Onsager model, which ascribes
water and DMSO very similar reaction fields, this observation
13CN
suggests that δ
sensitive to H-bonding as νCN
is sensitive to electrostatic field but not as
17
.
̅
As described in earlier work on thiocyanate probes, a plot of
13
CN
νCN versus δ
̅
(Figure 1C) recapitulates the linear sensitivity
of both observables to electrostatic field but is independent of
any specific model to calculate it.17 The best-fit line in Figure
1C, drawn for the non-H-bonding solvents’ points, describes
in environments free of H-
bonds. Therefore, deviations from this line represent effects of
specific interactions, which alter νCN and δ differently and
13CN
the covariance of νCN and δ
̅
a
Conditions: (i) 1 h, 46 °C; (ii) 1 h, 150 °C, microwave; (iii) 2 h, RT;
13CN
(iv) RT, 20 mM HEPES, pH 8.0.
̅
13CN
confound the mutual linear dependence of νCN and δ
on
̅
field. Points that occur in the region near the line given by
electrostatics (shown in gray in Figure 1C) correspond to
environments free of H-bonds, while points that fall
significantly off this line appear to experience a non-
electrostatic contribution, assigned to H-bond effects. Noting
this observation, IR and NMR data can be used to assess
whether a nitrile is H-bonded in complicated environments
where it is not known a priori.
given in the Supporting Information), Cu13CN was prepared
from commercial K13CN with aqueous chemistry24 (i). N-
Fmoc-p-13CN-Phe (2) was prepared in moderate (40%) yield
using a microwave-mediated reaction (ii) between the
corresponding N-Fmoc-p-I-Phe (1) and Cu13CN. With 2 in
hand, Fmoc can be easily removed to give the isotopically
labeled amino acid (3).25 This same approach has been used to
10374
dx.doi.org/10.1021/ja303895k | J. Am. Chem. Soc. 2012, 134, 10373−10376

Journal of the American Chemical Society
Communication
label other aromatic nitriles26 and in more complex nitrile-
bearing enzyme inhibitors (unpublished results).
an unstructured peptide into a complete protein) can result in a
fairly small shift.
We recently proposed a semiempirical strategy to determine
total electrostatic fields in proteins:26 the vibrational
frequencies of a model compound in a series of solvents paired
with their calculated electric fields (such as from Onsager’s
theory) might be used as a reference to assign a vibrational
frequency recorded in a protein to an electrostatic field. This
concept rests on the assumptions that (1) the model being used
to calculate solvent fields is accurate and (2) the solvatochromic
frequency shifts are principally due to electrostatics.
Evidence supporting these two assumptions for the non-H-
bonding case was presented,26 and the solvatochromic study on
PhCN presented here (Figure 1) further supports the previous
work. The NMR/IR data for RNase S (point 5 in Figure 1C)
confirm that the nitrile probe is not H-bonded, indicating that
the solvatochromic data might be a useful reference to assign a
When ribonuclease A is digested by subtilisin, a 20-residue
peptide called the S-peptide is liberated, forming a truncated
ribonuclease called the S-protein. Subsequent combination of
S-peptide with S-protein reconstitutes a functional semi-
synthetic protein called RNase S.6 We chose to use this system
as a test case because the S-peptide can be exploited to deploy
an IR/NMR probe into RNase S by replacing Phe8 with p-CN-
Phe. This nitrile-bearing RNase S has been characterized, and
its crystal structure has been solved (PDB: 3OQY).6 We used 2
as a reagent to generate an S-peptide bearing an isotopically
labeled nitrile ([p-13CN-Phe]S-peptide, 4), which in turn can
be used to form [p-13CN-Phe]RNase S (5) as shown in Scheme
1. 5 has been shown to have nearly the same catalytic properties
as the native ribonuclease.6 The nitrile in 5 is buried in a
hydrophobic region near the active site, with no access to the
solvent and no H-bonding partners within a reasonable
distance.
total electrostatic field to νCN observed in the protein complex.
̅
Applying the field−frequency correlation from Figure 1A, the
electric field corresponding to the nitrile stretching frequency in
5 is −7 MV/cm. Molecular dynamics (MD) simulations using
the Amber-99 force field were carried out on the [p-CN-
Phe]RNase S construct as previously described (details found
in Supporting Information).6,27 The electric field at the
midpoint of the nitrile bond was calculated every 2 fs steps
during the 20 ns trajectory, and the average total electric field
experienced by the nitrile was found to be −7.8 MV/cm
(Figure 2), which agrees well with the value determined by the
proposed semiempirical method. This agreement supports the
use of a solvatochromic scheme to calibrate a mapping between
vibrational frequency and electrostatic field; it also provides an
uncommonly convincing testament to the accuracy of the
electrostatic parameters in a standard MD force field.
13C NMR experiments were carried out on the labeled amino
acid (3), the labeled peptide fragment (4), and the labeled split
protein (5). With quantitative isotopic enrichment of a single
atom, high-quality NMR spectra (see inset of Figure 1C) could
be obtained on protein samples with fewer than 100 scans.
These measurements, in combination with νCN’s of these three
̅
species, allow us to apply the analysis described in Figure 1.
13
The three (δ CN, νCN) ordered pairs for the free amino acid
̅
(3), S-peptide (4), and RNase S (5) can be compared against
the solvatochromic model (brown circles) and are shown
alongside it in Figure 1C. It is apparent that the points for 3
and 4 lie well off the line described by a purely electrostatic
model, and in fact are very close to the point representing
PhCN in water (the black square). These data suggest that the
nitriles of p-CN-Phe in the free amino acid and in the S-peptide
are H-bonded (most likely to water), and that their local
electrostatic environments are largely determined by the
surrounding water molecules.
When the nitrile is embedded into a protein environment, by
13
incorporating the S-peptide into RNase S (5), (δ CN, νCN
)
̅
moves into a region very close to the line described by the
electrostatic model, within the range found for other non-H-
bonding solvents. This result indicates that the nitrile in RNase
S is not H-bonded, consistent with expectations about the H-
bonding status of this nitrile from the crystal structure6 and
from 2D IR experiments.27 This conclusion could not have
been reached if one only knew νCN in 4 and 5, because the red-
̅
shift upon complexation of the S-peptide to the truncated
protein is only 4.0 cm−1, which is too small to assign to the
breaking of an H-bond. Figure 1C reveals that the modest red-
shift occurs because there are two partially counteracting
contributions: when the nitrile is embedded into a protein, it
loses an H-bond but also gets placed in a comparatively weaker
electric field. This weaker electric field can be explained
qualitatively by the fact that the local hydrophobic environment
inside RNase S is less polar than the environment in aqueous
solution. The superposition of these two effects results in a shift
that is difficult to interpret from the IR measurements alone.
However, in combination with NMR measurements, we can
deconvolute the two effects. This discussion highlights how the
dual IR/NMR technique can be used to dissect an IR frequency
shift and resolve a puzzle such as understanding how a large
alteration in local environment (associated with reconstituting
Figure 2. (A) Structure of [p-CN-Phe]RNase S (5). The probe,
located at position 8 on the S-peptide, is shown in sticks. It is buried in
a hydrophobic pocket and is close to the catalytic proton shuttle
His12, shown in sticks. (B) Histogram of the electric fields
experienced by the nitrile in 5 during a 20 ns MD trajectory. The
mean electric field is −7.8 MV/cm; the standard deviation is 10.0 MV/
cm.
10375
dx.doi.org/10.1021/ja303895k | J. Am. Chem. Soc. 2012, 134, 10373−10376

Journal of the American Chemical Society
Communication
(7) Fafarman, A. T.; Sigala, P. A.; Schwans, J. P.; Fenn, T. D.;
Herschlag, D.; Boxer, S. G. Proc. Natl. Acad. Sci. U.S.A. 2012, 109,
E299−E308.
To the best of our knowledge, this work constitutes the first
instance of a meaningful comparison between computation and
experiment of a single-state total electric field. Previous studies
have focused on electric field differences between two
states2,4−7,28,29 or electric field fluctuations,27,30 which do not
require an external reference and depend only on the Stark
tuning rate. The total electric field that a protein exercises on a
target biomacromolecule, ligand, or transition state defines the
energetics that underlie molecular recognition, binding, and
catalysis, respectively. We therefore expect that translating IR
spectroscopic data into semiempirical absolute electric field
maps in proteins will lead to a deeper physical understanding of
protein function. Two aspects of our system likely made this
agreement between experiment and theory possible: (1) the
nitrile probe is not H-bonded at any point during the MD
trajectory, which would introduce non-electrostatic contribu-
tions to the vibrational frequency, and (2) the absence of slow
dynamics, as evidenced by 2D IR studies on this construct,27
allowed the relatively short simulation to adequately sample
configurations of the protein present in the equilibrium
ensemble.
In summary, we have reported the synthesis and application
of a dual NMR/IR probe (p-13CN-Phe), which was able to
determine that a nitrile probe installed in RNase S is not H-
bonded. Using experimental solvatochromic data as a
calibration tool, we then translated the nitrile stretching
frequency to a total electrostatic field and found excellent
agreement with simulation. Although this single example’s
agreement is encouraging, current work is underway to further
benchmark this method and examine its range of validity.
(8) Lindquist, B. A.; Furse, K. E.; Corcelli, S. A. Phys. Chem. Chem.
Phys. 2009, 11, 8119−8132.
(9) Dalosto, S. D.; Vanderkooi, J. M.; Sharp, K. A. J. Phys. Chem. B
2004, 108, 6450−6457. Dalosto et al. calculated from DFT that the
linear dependence of νCN on electrostatic field breaks down at very
̅
large fields, ca. −50 MV/cm. This effect would complicate the simple
linear solvatochromic model presented herein; however, our measure-
ments suggest that average total electrostatic fields in solvents and
proteins are low enough to remain in a linear regime.
(10) Getahun, Z.; Huang, C.-Y.; Wang, T.; De Leon
́
, B.; DeGrado,
W. F.; Gai, F. J. Am. Chem. Soc. 2002, 125, 405−411.
(11) Fang, C.; Bauman, J. D.; Das, K.; Remorino, A.; Arnold, E.;
Hochstrasser, R. M. Proc. Natl. Acad. Sci. U.S.A. 2008, 105, 1472−
1477.
(12) Aschaffenburg, D. J.; Moog, R. S. J. Phys. Chem. B 2009, 113,
12736−12743.
(13) Maienschein-Cline, M. G.; Londergan, C. H. J. Phys. Chem. A
2007, 111, 10020−10025.
(14) Choi, J.-H.; Oh, K.-I.; Lee, H.; Lee, C.; Cho, M. J. Chem. Phys.
2008, 128, 134506.
(15) Choi, J.-H.; Cho, M. J. Chem. Phys. 2011, 134, 154513.
(16) Lindquist, B. A.; Corcelli, S. A. J. Phys. Chem. B 2008, 112,
6301−6303.
(17) Fafarman, A. T.; Sigala, P. A.; Herschlag, D.; Boxer, S. G. J. Am.
Chem. Soc. 2010, 132, 12811−12813.
(18) Fafarman, A. T.; Webb, L. J.; Chuang, J. I.; Boxer, S. G. J. Am.
Chem. Soc. 2006, 128, 13356−13357.
(19) Onsager, L. J. Am. Chem. Soc. 1936, 58, 1486−1493.
(20) The Onsager field, FOnsager, is given by the expression FOnsager
=
(μ0/a3)[2(ε − 1)(n2 + 2)/3(2ε + n2)]. It is a function of the solvent’s
static dielectric constant, ε, the solute’s gas-phase dipole moment, μ0,
and the solute’s refractive index, n. The term a is the Onsager cavity
radius and is related to the molecular volume of the solute.
(21) The dipole of PhCN was taken to be 4.48 D ( Borst, D. R.;
Korter, T. M.; Pratt, D. W. Chem. Phys. Lett. 2001, 350, 485−490 ). Its
index of refraction is 1.528 (CRC Handbook), and its volume factor
was taken to be 171 Å3, as determined from its formula weight 103 g/
mol and density 1.0 g/mL. Static dielectric constants for all the
solvents were taken from the CRC Handbook.
ASSOCIATED CONTENT
■
S
* Supporting Information
Synthetic, spectroscopic, and simulation methods. This material
is available free of charge via the Internet at http://pubs.acs.org.
AUTHOR INFORMATION
■
(22) Andrews, S. S.; Boxer, S. G. J. Phys. Chem. A 2000, 104, 11853−
11863.
Corresponding Author
sboxer@stanford.edu
(23) Ellis, G. P.; Romney-Alexander, T. M. Chem. Rev. 1987, 87,
779−794.
Author Contributions
‡These authors contributed equally.
(24) Matloubi, H.; Shafiee, A.; Saemian, N.; Shirvani, G.; Daha, F. J. J.
Labelled Compd. Radiopharm. 2004, 47, 31−36.
Notes
(25) Wang, L.; Zhang, Z. W.; Brock, A.; Schultz, P. G. Proc. Natl.
Acad. Sci. U.S.A. 2003, 100, 56−61.
The authors declare no competing financial interest.
(26) Levinson, N. M.; Fried, S. D.; Boxer, S. G. J. Phys. Chem. B 2012,
DOI: 10.1021/jp301054e.
ACKNOWLEDGMENTS
■
(27) Bagchi, S.; Boxer, S. G.; Fayer, M. D. J. Phys. Chem. B 2012, 116,
4034−4042.
We thank the Trost laboratory for use of their microwave
reactor. S.D.F. thanks the NSF for a predoctoral fellowship.
This work is supported in part by a grant from the NIH
(GM27738).
(28) Park, E. S.; Andrews, S. S.; Hu, R. B.; Boxer, S. G. J. Phys. Chem.
B 1999, 103, 9813−9817.
(29) Laberge, M.; Vanderkooi, J. M.; Sharp, K. A. J. Phys. Chem.
1996, 100, 10793−10801.
(30) Merchant, K. A.; Noid, W. G.; Akiyama, R.; Finkelstein, I. J.;
Goun, A.; McClain, B. L.; Loring, R. F.; Fayer, M. D. J. Am. Chem. Soc.
2003, 125, 13804−13818.
REFERENCES
■
(1) Taskent-Sezgin, H.; Chung, J.; Banerjee, P. S.; Nagarajan, S.;
Dyer, R. B.; Carrico, I.; Raleigh, D. P. Angew. Chem., Int. Ed. 2010, 49,
7473−7475.
(2) Hu, W.; Webb, L. J. J. Phys. Chem. Lett. 2011, 2, 1925−1930.
(3) Suydam, I. T.; Boxer, S. G. Biochemistry 2003, 42, 12050−12055.
(4) Suydam, I. T.; Snow, C. D.; Pande, V. S.; Boxer, S. G. Science
2006, 313, 200−204.
(5) Webb, L. J.; Boxer, S. G. Biochemistry 2008, 47, 1588−1598.
(6) Fafarman, A. T.; Boxer, S. G. J. Phys. Chem. B 2010, 114, 13536−
13544.
10376
dx.doi.org/10.1021/ja303895k | J. Am. Chem. Soc. 2012, 134, 10373−10376