Using intrinsic X-ray absorption spectral differences to identify and map peptides and proteins.
ABSTRACT The intrinsic variation in the near-edge X-ray absorption fine structure (NEXAFS) spectra of peptides and proteins provide an opportunity to identify and map them in various biological environments, without additional labeling. In principle, with sufficiently accurate spectra, peptides (<50 amino acids) or proteins with unusual sequences (e.g., cysteine- or methionine-rich) should be differentiable from other proteins, since the NEXAFS spectrum of each amino acid is distinct. To evaluate the potential for this approach, we have developed X-SpecSim, a tool for quantitatively predicting the C, N, and O 1s NEXAFS spectra of peptides and proteins from their sequences. Here we present the methodology for predicting such spectra, along with tests of its precision using comparisons to the spectra of various proteins and peptides. The C 1s, N 1s, and O 1s spectra of two novel antimicrobial peptides, Indolicidin (ILPWKWPWWPWRR-NH2) and Sub6 (RWWKIWVIRWWR-NH2), as well as human serum albumin and fibrinogen are reported and interpreted. The ability to identify, differentiate, and quantitatively map an antimicrobial peptide against a background of protein is demonstrated by a scanning transmission X-ray microscopy study of a mixture of albumin and sub6.
Using Intrinsic X-ray Absorption Spectral Differences To Identify and Map Peptides and
Jacob Stewart-Ornstein,†Adam P. Hitchcock,*,†Daniel Herna ´ndez Cruz,†Peter Henklein,‡
Joerg Overhage,§Kai Hilpert,§John D. Hale,§and Robert E. W. Hancock§
BIMR and Department of Chemistry, McMaster UniVersity, Hamilton, Ontario, Canada L8S 4M1, Institut fu ¨r
Biochemie, UniVersita ¨tsklinikum Charite ´, Humboldt-UniVersita ¨t, 10117 Berlin, Germany, and Department of
Microbiology, UniVersity of British Columbia, VancouVer, British Columbia, Canada V6T 1Z4
ReceiVed: March 15, 2007; In Final Form: April 29, 2007
The intrinsic variation in the near-edge X-ray absorption fine structure (NEXAFS) spectra of peptides and
proteins provide an opportunity to identify and map them in various biological environments, without additional
labeling. In principle, with sufficiently accurate spectra, peptides (<50 amino acids) or proteins with unusual
sequences (e.g., cysteine- or methionine-rich) should be differentiable from other proteins, since the NEXAFS
spectrum of each amino acid is distinct. To evaluate the potential for this approach, we have developed
X-SpecSim, a tool for quantitatively predicting the C, N, and O 1s NEXAFS spectra of peptides and proteins
from their sequences. Here we present the methodology for predicting such spectra, along with tests of its
precision using comparisons to the spectra of various proteins and peptides. The C 1s, N 1s, and O 1s spectra
of two novel antimicrobial peptides, Indolicidin (ILPWKWPWWPWRR-NH2) and Sub6 (RWWKIWVIR-
WWR-NH2), as well as human serum albumin and fibrinogen are reported and interpreted. The ability to
identify, differentiate, and quantitatively map an antimicrobial peptide against a background of protein is
demonstrated by a scanning transmission X-ray microscopy study of a mixture of albumin and sub6.
Although protein detection relative to other biological mac-
romolecules (DNA, polysaccharides, lipids, etc.) by spatially
resolved near-edge X-ray absorption fine structure (NEXAFS)
spectromicroscopy is well-developed,1-4to date there has been
only one example5in which two different proteins were
differentiated and mapped using the NEXAFS spectra of the
unlabeled species. In principle, with sufficiently accurate spectra,
peptides (usually defined as having <50 amino acids) or proteins
with unusual sequences (e.g., cysteine- or methionine-rich)
should be differentiable since the NEXAFS spectrum of each
amino acid is distinct.6,7We are interested in exploiting these
intrinsic spectral differences to map specific peptides or proteins
in various biological environments. In particular, the motivation
for the work described in this paper is to develop and use this
capability to locate cationic antimicrobial peptides (CAPs) in
biofilms, thereby assisting efforts to investigate the mechanisms
of their actions and the ability to penetrate bacterial cells.
Cationic antimicrobial peptides are small (10-50 amino
acids), hydrophobic (>30% hydrophobic amino acids), posi-
tively charged (net charge between +2 and +9 at neutral pH)
peptides produced by a diverse range of organisms including
humans, insects, and plants.8Hundreds of natural CAPs having
been identified, and more have been designed through the
engineering of existing CAPs that aim to produce peptides with
increased potency.9,10The majority of CAPs work by interacting
and disrupting the integrity of the cytoplasmic membrane in a
susceptible bacterial cell.11However, many examples exist
where alternative targets have been identified.12The ability to
visualize CAPs in biofilms could help in understanding their
mode of action and thereby assist in refining tools that allow
prediction of peptide sequences with even greater activity. The
rate of advance in this field could be significantly increased if
it was possible to determine the location and local concentrations
of CAPs within individual cells or in complex microbial
We are using scanning transmission X-ray microscopy
(STXM)13-15to record spectra of pure peptides and proteins in
the solid state and also to investigate complex biological
samples.1-3,16,17The spectra of peptides and proteins on polymer
surfaces is being examined by photoemission electron micros-
copy (X-PEEM).4,18These techniques compliment more com-
mon methods for mapping proteins such as laser scanning
microscopy of fluorescent labeling proteins,19which has high
specificity but limited spatial resolution, and immunoelectron
microscopy, which has high spatial resolution but requires a
chemically fixed sample.20
The X-ray absorption spectra of proteins are quite distinct
from that of other biological materials, especially at the C 1s
edge. This is the basis for mapping locations and concentrations
of proteins relative to other biological materials or carbon-based
polymers.1-5Differentiation among proteins is, in contrast, quite
difficult and relies on spectral differences caused by differing
amino acid composition or, potentially, structure. To begin to
locate specific peptides or proteins in a biological sample, it is
necessary to understand and exploit these spectral differences,
which are often very small. It is impractical to record spectra
of all proteins or peptides of interest. In most cases it is likely
they would prove too similar to other proteins to provide
analytically useful spectral differentiation. Thus, it would be
useful to have a means of predicting the NEXAFS spectra of
peptides or proteins to evaluate the feasibility of a proposed
‡Universita ¨tsklinikum Charite ´.
§University of British Columbia.
J. Phys. Chem. B 2007, 111, 7691-7699
10.1021/jp0720993 CCC: $37.00© 2007 American Chemical Society
Published on Web 06/09/2007
detection and visualization scheme. Fortunately, the spectra of
large molecules often resemble a sum of the functional groups
presentsthe building block hypothesis.21Though this is typically
applied to small molecules (as in a recent comparison of the
spectra of phenylalanine to those of benzene and alanine22),
studies of small peptides (two to three monomer units) have
shown that their spectra are similar to the sum of the spectra of
their amino acid constituents.23,24The recent publication of high-
quality C 1s, N 1s, and O 1s spectra of the 20 common amino
acids7in their zwitterionic neutral form makes it possible to
predict the spectra of any peptide or protein using a weighted
sum of the spectra of its constituent amino acids. To evaluate
the potential for this approach, we have developed a software
tool for quantitatively predicting the C 1s, N 1s, and O 1s
NEXAFS spectra of peptides and proteins from their sequences,
within a modified building block approach. Specifically our code
generates a sum based on the sequence and then adds a distortion
which mimics the spectral modifications arising from the
structural changes associated with peptide bond formation.24,25
With the X-ray absorption spectral simulator (X-SpecSim) we
are able to explore the applicability of NEXAFS spectral
differentiation to any size of peptides and even to proteins made
up of hundreds of amino acids. Here we present the methodol-
ogy for predicting such spectra, explain how it deals with the
spectral changes associated with peptide bond formation, and
present tests of its precision by comparing its predictions to
the experimental spectra of two novel antimicrobial peptides
and two blood proteins, human serum albumin (HSA) and
fibrinogen. To demonstrate that intrinsic differences in their
NEXAFS spectra enable identification and mapping, we con-
clude by presenting a successful quantitative analysis of a
spatially heterogeneous peptide-protein mixture by STXM
using the intrinsic C 1s spectral differences.
Other recent work to simulate the NEXAFS spectra of
proteins includes a very elegant study by Liu et al.26who took
this concept in some sense one step further, by analyzing the
full 3D structure of a ribonuclease to predict the polarization
dependence of the N 1sf π*amideand S 1s f σ*S-Stransitions
in the NEXAFS spectra of an oriented sample. They were able
to show that polarization-dependent NEXAFS could be used
to differentiate oriented and unoriented ribonuclease molecules.
2. Experimental Section
2.1. Sample Sources and Preparation. Sub6 (RWWKI-
WVIRWWR-NH2),27an optimized variant of Bac2A, which in
turn is a linearized version of bactenecin, a naturally occurring
bovine peptide,28was synthesized by P. Henklein using 9-fluo-
renylmethyl carbamate (fMOC) solid-phase synthesis. Indoli-
cidin (ILPWKWPWWPWRR-NH2) was synthesized by P.
Owen at the Peptide Synthesis Facility, Biomedical Research
Centre, UBC, using tertiary butyloxycarbonyl (tBOC) solid-
phase synthesis.29,30Both peptides were purified by HPLC and
their respective masses confirmed by mass spectrometry.
Human serum albumin (HSA) was obtained from Behring-
werke AG, Marburg, Germany, and found to be homogeneous
as judged by sodium dodecyl sulfate polyacrylamide gel
electrophoresis (SDS-PAGE). Plasminogen-free human plasma
Figure 1. Illustration of the NEXAFS building block model with a tetrapeptide (IWRK-NH2): (a) C 1s spectra of the individual amino acids; (b)
ball and stick model of IWRK-NH2; (c) simple sum of C 1s spectra of the four amino acids, compared to the experimental C 1s spectrum of human
7692 J. Phys. Chem. B, Vol. 111, No. 26, 2007
Stewart-Ornstein et al.
fibrinogen (Calbiochem) was used. It is reported to be >95%
clottable by thrombin and pure as judged by SDS-PAGE.
Samples of antimicrobial peptides and human serum albumin
were prepared by dissolving lyophilized material in deionized
water to a concentration of ∼1 mg/mL. A few microliters of
this solution was deposited onto a Si3N4window and allowed
to air-dry. The resulting nonuniform film was examined for
regions of appropriate thickness for STXM by optical micros-
copy. In the case of fibrinogen, initial attempts to dissolve and
solvent-cast gave unreliable results, probably due to limited
solubility and the presence of highly soluble lyophilization salts.
This problem was circumvented by placing small amounts of
the solid on a Si3N4window and then solubilizing and removing
most of the material by a micropipette, until areas thin enough
for STXM (∼100 nm) were left. A combined sample with Sub6
on HSA was prepared by first solvent-casting HSA onto a Si3N4
window, then sprinkling Sub6 powder on that surface, and
blowing off the excess.
2.2. Scanning Transmission X-ray Microscopy. STXM31
at the dedicated polymer STXM beamline 5.3.232of the
Advanced Light Source (ALS) was used to measure the spectra
of the pure peptides and proteins. STXM is a microscopy-based
version of NEXAFS spectroscopy, which combines good energy
resolution (<0.1 eV) with a spatial resolution (∼40 nm)
intermediate between optical techniques and electron or scanning
probe microscopies. The high spatial resolution was used to find
micrometer-sized regions of suitable thickness (∼100 nm) in
the solvent- cast samples. Typically spectra are recorded using
image sequences33because the samples are only homogeneous
over small areas, and the image sequence can be carefully
aligned postacquisition. Image sequence spectra also minimize
the radiation dose. Checks were made after each measurement
to ensure the radiation damage was negligible.
2.3. Method of Predicting Spectra. The key tenet of the
building block model21is that the spectral signature of a
functional group does not change from the reference compound
to the final chemical compound or that if there is a change, it
is uniform and well-quantified. The polymerization of amino
acids to form peptides results in the removal of two functional
groupssthe carboxylic acid and amine groupssand the creation
of a new groupsthe amide. There are modifications to the C
1s, N 1s, and O 1s NEXAFS spectra associated with this
change.24To produce reasonable simulations of the spectra of
peptides and proteins, these changes must be reproduced by the
software. The spectral modification associated with the polym-
erization of amino acids (the formation of the peptide bond) is
accomplished by addition of a “peptide bond spectrum” to the
spectrum of the amino acid constituents. The peptide bond
spectrum consists of addition of signal associated with the amide
group and subtraction of signal associated with the amine and
carboxylic acid groups.
2.3.1. X-ray Spectral Simulator (X-SpecSim). X-SpecSim is
a program written in Interactive Data Language.34Its input
is an amino acid sequence, which can be obtained from
the synthesis or analysis of peptides, or for proteins, from
standard databases such as the protein data bank (http://
Figure 2. Demonstration of correction for the peptide shift: (a) experimental C 1s spectrum of alanine (red) compared to the peptide-bond corrected
spectrum (green) (i.e., the spectrum expected of polyalanine) and the C 1s “peptide bond spectrum”, which is the correction applied; (b) experimental
N 1s spectrum of alanine compared to the peptide-bond corrected spectrum and the N 1s peptide bond spectrum; (c) experimental O 1s spectrum
of alanine compared to the peptide-bond corrected spectrum and the O 1s peptide bond spectrum; (d) the C 1s spectrum of human serum albumin,
predicted with and without the peptide bond correction, compared to the experimental spectrum.
Identifying and Mapping Peptides and Proteins
J. Phys. Chem. B, Vol. 111, No. 26, 2007 7693
www.pdb.org/).35X-SpecSim uses a library of C 1s, N 1s, and
O 1s spectra of the amino acids to compute the expected
NEXAFS spectra of that sequence. The default library is the
spectra reported by Zubavichus et al.,7which were measured
by total electron yield from powdered samples and thus avoided
a solvent contamination issue that plagued an earlier study of
the solid amino acids.6This library is stored in tab-delimited
text files. It can be modified easily as more accurate data
becomes available or if different edges (e.g., S 1s or S 2p) are
required. The output of the predicted spectrum is written to a
file in a format similar to the input. The user can choose the
energy scale and point spacing from the graphical interface, or
a stored spectrum can be used to set these parameters. The amino
acid spectra reported by Zubavichus et al.7are sampled at
approximately 0.1 eV in key spectral regions.
The peptide shift correction signals, which take the loss of
amine and carboxylic acid groups and the gain of the amide
group into account, are computed by subtracting the spectra of
some known peptides and proteins from their amino acid
constituents, as described in more detail below. A further issue
requiring attention is the correct absolute scaling of the predicted
spectrum. The intensity scales of NEXAFS spectra recorded
by STXM are generally shown in units of optical density (OD)
or absorbance. Outside of the strongly structured near-edge
region (-10 to +30 eV relative to the ionization limit, typically),
the linear absorbance of a material (optical density per unit
thickness) is determined solely by its elemental composition
and density.36For quantitative analysis we find it useful to place
X-ray absorption reference spectra on an “OD1” scale, which
is the optical density of a nanometer thickness of a material at
its normal density. The final predicted spectrum is thus
constructed by (i) generating a sum of spectra of the amino acids,
weighted according to their frequency in the sequence; (ii)
modifying the spectrum to correct for spectral changes associ-
ated with peptide bond formation; and (iii) rescaling the summed
spectrum to match the elemental absorption of the final peptide
or protein predicted from the sequence using tabulated data.36
The final scaling step is necessary as peptide bond formations
specifically the loss of H2Oschanges the ratios of C, N, O, S,
and H in proteins relative to their amino acid constituents.
X-SpecSim is integrated into aXis2000,37a publicly available
program for analysis of spectromicroscopy data. Additionally,
the SF utility38is used to compute the expected OD1 elemental
response of the peptide or protein.
An important consideration is the ionic state of the peptide
or protein. The spectrum of glycine has been shown to undergo
Figure 3. Measured versus predicted C 1s, N 1s, and O 1s spectra of
indolicidin, a tryptophan-rich antimicrobial peptide (ILPWKWPWW-
PWRR-NH2). The elemental line represents the absorption of the
elemental composition of the peptide predicted from tabulated data.36
It is the expected spectrum without contributions from molecular
structure and is used to set the intensity scale of the spectrum.
Figure 4. Measured versus predicted C 1s, N 1s, and O 1s spectra of
sub6, a tryptophan-rich antimicrobial peptide (RWWKIWVIRWWR-
7694 J. Phys. Chem. B, Vol. 111, No. 26, 2007
Stewart-Ornstein et al.
significant changes in all edges as the pH of its environment
changes and the molecule transitions between cationic, zwitte-
rionic, and anionic states.6,39Less well-studied, but probably
equally significant, is the affect of ionization state on the spectra
of charged amino acid side chains such as arginine or lysine.
All spectra presented here are from molecules at neutral pH;
the amino acids are predominantly zwitterionic, lysine, argenine,
and the amino terminals of peptides positively charged, and
aspartic acid, glutamic acid, and the carboxyl terminal of
peptides negatively charged. The sub6 peptide has a net charge
of +5. Caution needs to be exercised when applying spectra
calculated with the standard amino acid library to samples in
low- or high-pH environments.
2.3.2. Illustrating X-SpecSim for Predicting Peptide Spectra.
Previous applications of the building block model to NEXAFS
spectra have been on a relatively small scale, typically involving
molecules consisting of a few dozen atoms at most, not the
hundreds of atoms which make up even relatively small
peptides. Thus, automation of the adding and scaling processes
was required. In addition to sheer scale, peptides offer two other
significant challenges. They are diverse polymers, with 20
common monomers, and as they polymerize they undergo
TABLE 1: Energies and Assignments of Spectral Features in the C 1s, N 1s, and O 1s Spectra of Indolicidin
(ILPWKWPWWPWRR-NH2) and Sub6 (RWWKIWVIRWWR-NH2)
indolicidinsub-6 assignment (final orbital) residue
π*CdC(aromatic) (V ) 0)
π*CdC(aromatic) (vC-H) 1)
C1s(C-R) f π*CdC
412(1) (sh, br)
413(1) (sh, br)
543 (br, sh)
543 (br, sh)
O 1s f π*COO
aC 1s calibration: -4.50(6) relative to C 1s f 3s (V ) 0) in CO2.45
1s calibration: -6.80(10) relative to O 1s f 3s in CO2.47
bN 1s calibration: -4.98(8) relative to N 1s f 3s (V ) 0) in N2.46
TABLE 2: Energies and Assignments of Spectral Features in the C 1s, N 1s, and O 1s Human Serum Albumin and Fibrinogen
albuminfibrinogen assignment (final orbital)
π*CdC(aromatic) (V ) 0)
π*CdC(aromatic) (vC-H) 1)
O 1s π*COO
291.2 (sh, br)
291.2 (sh, br)
Phe, Tyr, Trp, His
Phe, Tyr, Trp, His
Phe, Tyr, Trp, His
412(1) (sh, br)
413(1) (sh, br)
547 (br, sh)
545 (br, sh)
aC 1s calibration: -4.54(6) relative to C 1s f 3s (V ) 0) in CO2.45
1s calibration: -6.8(1) relative to O 1s f 3s in CO2.47
bN 1s calibration: -4.95(8) relative to N 1s f 3s (V ) 0) in N2.46
Identifying and Mapping Peptides and Proteins
J. Phys. Chem. B, Vol. 111, No. 26, 2007 7695
significant chemical change. Figure 1a plots the C 1s spectra
of four amino acids to illustrate the diversity of amino acid
spectra. This diversity is, however, nearly absent in protein
spectra, due to averaging over hundreds of residues. In smaller
peptides, or proteins with relatively unusual sequences, the
contribution of individual amino acids or classes of amino acids
can be observed. The amino acids with readily distinguishable
spectral features include the following: (i) aromatic amino acids,
which have a strong, relatively narrow peak at 285.2 eV
(phenylalanine, tyrosine, and tryptophan); (ii) S-containing
amino acids (methionine, cysteine) (relatively little work to date
has used S 2p or S 1s spectroscopy, a notable exception being
that of Liu et al.26); (iii) amino acids containing unsaturated
CdN bonds (histidine, arginine, and tryptophan).
The challenge of spectral changes associated with peptide
bond formation is also shown in Figure 1c, which compares
the spectrum of the hypothetical peptide IWRK-NH2, simulated
according to the “classic” building block approach (i.e., without
correction for peptide bond formation), to that of human serum
albumin, a protein with a relatively typical amino acid content
in terms of the ratio of aromatic to nonaromatic amino
acids. There is a clear difference in the 288 eV region. Albumin
has a C 1s f π*CdO peak at 288.2 eV, a position which
is consistent with an amide bond. It is red-shifted relative
to the 288.5 eV C 1s f π*CdOpeak of the carboxylic acid of
an amino acid. This shift is well-documented24and is the
principal change in the C 1s spectra of peptides compared to
amino acids. The change at the N 1s edge is much more
dramatic, with introduction of a low-lying N 1s f π*amidepeak
at 401.2 eV,24a feature entirely absent in the N 1s spectrum of
amino acids. The change at the O 1s edge includes a ∼0.3 eV
shift from a O 1s(CdO) f π*CdOpeak to the O 1s fπ*amide
peak, and loss of the O 1s(C-O) f π*CdOpeak at 535.5 eV,
as the carboxylic acid functional group is transformed to an
To correctly estimate the NEXAFS spectra of peptides,
it is critical to correct for the spectral changes associated with
peptide bond formation. X-SpecSim achieves this by adding a
“peptide bond spectrum” to the simple summation of amino
acids. The peptide bond spectrum is a differential correction
which incorporates changes associated with the loss of the amine
and carboxylic groups and the addition of the amide group. This
correction and its influence on the resulting spectrum is
illustrated in Figure 2. The peptide bond spectrum that is added
to the amino acid spectra to account for polymerization is
shown in Figure 2a-c, for the case of converting the spectrum
of alanine into that of polyalanine. The major features of the C
1s peptide bond spectrum (Figure 2a) are the negative peak at
288.5 eV, which arises from loss of the carboxylic acid group,
and the positive peak at 288.2 eV that arises from formation of
the amide group. The principal effects of this correction, a shift
in the location, mild loss in intensity, and slight broadening
of the 288.2 eV peak, can be observed in Figure 2a, where the
C 1s spectrum of polyalanine is shown before and after the
peptide bond correction. The N 1s peptide bond spectrum
(Figure 2b) has three distinct spectral features: the N 1s f
π*amidepeak at 401.2 eV; a mild loss of intensity from the broad
N 1s f σ*N-Cpeak at 406 eV; and the addition of a broad N
1s f σ*C-N peak at 412 eV, due to the formation of an
additional N-C bond. The effect of these changes is illustrated
in Figure 2b, which compares the N 1s spectrum of polyalanine
before and after the peptide bond correction. The O 1s peptide
bond spectrum (Figure 2c) is a differential shape consisting
of the shift in the location of the O 1s f π*CdOpeak, which
is at 532.3 eV in amino acids and at 532.0 eV in peptides.
This shift can be seen in Figure 2c, where the peptide bond
spectrum has a negative peak at 532.2, showing the loss of
the carboxylic acid, and a positive peak at 532.0 eV, corre-
sponding to the formation of the amide group. The effect of
these changes is illustrated in Figure 2c, which compares the O
1s spectrum of polyalanine before and after the peptide bond
Figure 2d presents the X-SpecSim spectrum of albumin,
predicted with and without the correction, in comparison to the
experimental C 1s spectrum of albumin. Away from the 288
eV region, both spectra show good agreement; for example,
the height of the 285 eV aromatic peak is nearly identical
in the predicted and experimental spectra. This is comforting,
as this region should be unaffected by peptide bond for-
mation according to the building block model. The corrected
spectrum deviates somewhat from the experimental spectrum
in several slowly varying regions, but it fits well at both major
3. Results and Discussion
3.1. Antimicrobial Peptides. The antimicrobial peptides we
have studied have unusual amino acid compositions. Both
Figure 5. Measured versus predicted C 1s, N 1s, and O 1s spectra of
human serum albumin and fibrinogen.
7696 J. Phys. Chem. B, Vol. 111, No. 26, 2007
Stewart-Ornstein et al.
indolicidin and sub6 are rich in tryptophan and arginine residues.
Sub6 is 35% arginine and 42% tryptophan, while indolicidin
is 38% tryptophan and 15% arginine. The tryptophan aro-
matic residue affects the strength of the 285.2 eV peak, and
the guanidine group of arginine produces a peak at 289.3 eV.
These features provide spectral handles which may enable
detection of these peptides against a background of signal from
proteins in a biological environment such as a cell. Thus,
they provide a good test of our prediction method. Figure 3
compares measured versus predicted C 1s, N 1s, and O 1s
spectra for indolicidin, a 13 residue, tryptophan-rich peptide,
which has antimicrobial properties.9,10Note that both the
energy and intensity scales are absolute. There is very good
agreement between the experimental and predicted spectra,
indicating that, with inclusion of the peptide shift correction,
the spectral simulation works. Table 1 reports the energies
and assignments of the spectral features. As parts a-c of Figure
3 show, there is good general agreement between the measured
and predicted NEXAFS spectrum of indolicidin at the C 1s,
N 1s, and O 1s edges. At the C 1s edge the aromatic and
guanidine peaks agree nearly perfectly with the expected values;
the π*CdOpeak and the σ*C-Cregion of the predicted spec-
trum agree well in shape and location, but there are minor
intensity differences as compared to the measured spectrum.
The better definition of the near-edge fine structure (especially
on the high-energy side of the 285 eV peak) in the experi-
mental spectrum of indolicidin is probably due to the
higher experimental resolution in our STXM measurement
(∼150 meV full width at half-maximum (fwhm)) as compared
to that of 300 meV used by Zubavachus et al.7in
their measurements of the constituent amino acids. The
N 1s and O 1s spectra are an excellent match. The pre-
dicted and measured C 1s, N 1s, and O 1s spectra of sub6
are shown in Figure 4, in comparison to the spectra pr-
edicted by X-SpecSim. As with indolicidin, there is good
3.2. Proteins. The same approach for predicting peptide
spectra from their constituent amino acids can be applied to
proteins of hundreds or thousands of amino acids. Figure 5 plots
the C 1s, N 1s, and O 1s spectra of HSA and fibrinogen in
comparison to the predicted spectra. These spectra are newly
recorded and differ slightly from previously reported spectra
of fibrinogen18and albumin.40Spectral features and assignments
are given in Table 2. Figure 5 also displays the predicted spectra,
based on the known sequences of HSA41and fibrinogen.42,43
There is excellent agreement between the measured and
predicted spectra. Figure 5 illustrates one of the challenges of
using natural spectral contrast for protein identification. Because
of the large numbers of amino acids, there is a high degree of
spectral averaging, making proteins with “typical” ratios of
aromatic to nonaromatic amino acids almost indistinguishable.
One exception to this “indistinguishability by averaging” is
proteins composed of a repeating motif, such as many structural
proteins. Proteins with a relatively short repeated sequence of
amino acids are potentially very different spectrally from
proteins with more typical amino acid compositions. Finally,
in cases where there is a high degree of regularity (such as
proteins with large fractions of ?-sheets) and these regions are
aligned across different proteins, spectral differences can arise
from linear dichroic effects.5,44
Figure 6. Demonstration of the ability to differentiate sub6 and human serum albumin in a heterogeneous system: (a) component map of albumin
(the gray scale is thickness in nanometers); (b) component map of sub6, derived from fit to a C 1s image sequence (280-320 eV); (c) color-coded
composite map (red ) albumin, blue ) Sub6). Extracted C 1s spectra of (d) HSA, and (e) sub6, based on threshold masking the high-intensity
regions of the component maps, compared with the spectra of the pure materials.
Identifying and Mapping Peptides and Proteins
J. Phys. Chem. B, Vol. 111, No. 26, 2007 7697
3.3. Experimental Test of Distinguishing Peptide and
Protein by NEXAFS. As a test of the ability of the intrinsic
differences in the NEXAFS spectra to provide a contrast
mechanism and thus provide a means to quantitatively map
peptides against a protein background, STXM was used to
measure a C 1s image sequence of a sample consisting of a
Si3N4window coated with a uniform layer of HSA with a small
amount of sub6 deposited on top. The results of a fit of this C
1s image sequence to the reference spectra for albumin and sub6
are shown in Figure 6. Figure 6a is the component map for the
albumin indicating the film is relatively uniform, and ∼40 nm
thick. Figure 6b is the component map for the sub6 peptide,
indicating much thicker, but irregular, deposits, up to 400 nm
thick. Figure 6 demonstrates clearly that it is possible to
discriminate between the uniform background of HSA and the
nonuniform sub6 deposits using the differences in their C 1s
spectra and to derive quantitative maps. Parts d and e of Figure
6 plot the spectra extracted from two regions selected by
threshold masks of the high-intensity pixels of the component
maps. These extracted spectra are found to be very good matches
to the spectra of pure HSA and sub6. Although this is an
artificial two-component mixture, it indicates that differentiation
in a more complex sample should be possible, if sufficient
concentrations of the peptide or protein of interest are present
within the spatially resolved analytical zone, which can be as
small as a 30 nm diameter circle in state-of-the-art STXMs.
We have developed X-SpecSim, a tool that allows prediction
of the NEXAFS spectra of peptides and proteins of arbitrary
sequence from a data bank of the spectra of the constituent
amino acids.7This methodology can be readily extended
to predictions of the NEXAFS spectra of any substance
consisting of repeat units, such as block copolymers, polysac-
charides, and nucleic acids, etc. In addition we have demon-
strated the effectiveness of X-SpecSim by comparing its
predictions to the experimental spectra of previously unmeasured
peptides and several proteins. Finally we have demonstrated
that the intrinsic spectral differences are sufficiently large, at
least in some cases, to enable mapping of a peptide against a
background of proteins. Efforts are presently underway to use
this capability to track these and other antimicrobial peptides
dosed at sublethal levels in bacterial biofilms. If this works,
we hope to gain insights into mechanisms of antimicrobial
Acknowledgment. We especially thank Drs. Zubavichus,
Zharnikov, and Grunze for permission to use their amino acid
spectra in the spectral simulator program. Research was sup-
ported by Advanced Food and Materials Network (AFMnet),
NSERC, Canada Foundation for Innovation, and the Canada
Research Chair program. STXM was performed at beamline
5.3.2 at the Advanced Light Source, Berkeley, CA, which is
supported by the U.S. Department of Energy under Contract
DE-AC03-76SF00098. We thank David Kilcoyne and Tolek
Tyliszczak for their excellent support of the STXM instrumenta-
tion at the ALS.
References and Notes
(1) Zhang, X.; Balhorn, R.; Mazrimas, J.; Kirz, J. J. Struct. Biol. 1996,
(2) Hitchcock, A. P.; Morin, C.; Heng, Y. M.; Cornelius, R. M.; Brash,
J. L. J. Biomater. Sci., Polym. Ed. 2002, 13, 919-938.
(3) Lawrence, J. R.; Swerhone, G. D. W.; Leppard, G. G.; Araki, T.;
Zhang, X.; West, M. M.; Hitchcock, A. P. Appl. EnViron. Microbiol. 2003,
(4) Li, L.; Hitchcock, A. P.; Robar, N.; Cornelius, R.; Brash, J. L.;
Scholl, A.; Doran, A. J. Phys. Chem. B 2006, 110, 16763-16773.
(5) Herna ´ndez Cruz, D.; Hitchcock, A. P.; West, M. M.; Rousseau,
M.-E.; Pe ´zolet, M. Biomacromolecules 2006, 7, 836-843.
(6) Kaznacheyev, K.; Osanna, A.; Jacobsen, C.; Plashkevych, O.;
Vahtras, O.; Ågren, H.; Carravetta, V.; Hitchcock, A. P. J. Phys. Chem. A
2002, 106, 3153-3168.
(7) Zubavichus, Y.; Shaporenko, A.; Grunze, M.; Zharnikov, M. J.
Phys. Chem. A 2005, 109, 6998-7000.
(8) Hancock, R. E.; Sahl, H. G. Nat. Biotechnol. 2006, 24, 1551-
(9) Hilpert, K.; Elliott, M. R.; Volkmer-Engert, R.; Henklein, P.; Donini,
O.; Zhou, Q.; Winkler, D. F.; Hancock, R. E. Chem. Biol. 2006, 13, 1101-
(10) Zhang, L.; Scott, M. G.; Yan, H.; Mayer, L. D.; Hancock, R. E.
Biochemistry 2000, 39, 14504-14514.
(11) Brogden, K. A. Nat. ReV. Microbiol. 2005, 3, 238-250.
(12) Jenssen, H.; Hamill, P.; Hancock, R. E. Clin. Microbiol. ReV. 2006
(13) Kirz, J.; Jacobsen, C.; Howells, M. Q. ReV. Biophys. 1995, 28, 33-
(14) Ade, H. X-Ray Spectromicroscopy. In Experimental Methods in
the Physical Sciences, Vol. 32; Samson, J. A. R., Ederer, D. L., Eds.;
Academic Press: New York, 1998; pp 225-261.
(15) Ade, H.; Urquhart, S. G. NEXAFS Spectroscopy and Microscopy
of Natural and Synthetic Polymers. In Chemical Applications of Synchrotron
Radiation; Sham, T. K., Ed.; World Scientific: Singapore, 2002; pp 285-
(16) Hitchcock, A. P.; Morin, C.; Zhang, X.; Araki, T.; Dynes, J. J.;
Sto ¨ver, H. D. H.; Brash, J. L.; Lawrence, J. R.; Leppard, G. G. J. Electron
Spectrosc. Relat. Phenom. 2005, 144-147, 259-269
(17) Hitchcock, A. P.; Sto ¨ver, H. D. H.; Croll, L. M.; Childs, R. F. Aust.
J. Chem. 2005, 58, 423-432.
(18) Morin, C.; Hitchcock, A. P.; Cornelius, R. M.; Brash, J. L.;
Urquhart, S. G.; Scholl, A.; Doran, A. J. Electron Spectrosc. Relat. Phenom.
2004, 137-140, 785-794.
(19) Lippincott-Schwartz, J.; Patterson, G. H. Science 2003, 300, 87-
(20) Roth, J.; Heitz, P. U. Ultrastruct. Pathol. 1989, 13, 467-484.
(21) Sto ¨hr, J. NEXAFS Spectroscopy; Springer Tracts in Surface Science;
Springer: Berlin, 1992; Vol. 25.
(22) Cooper, G.; Gordon, M.; Tulumello, D.; Turci, C. C.; Kaznatcheev,
K.; Hitchcock, A. P. J. Electron Spectrosc. Relat. Phenom. 2004, 137-
(23) Boese, J.; Osanna, A.; Jacobsen, C.; Kirz, J. J. Electron Spectrosc.
Relat. Phenom. 1997, 85, 9-15.
(24) Gordon, M. L.; Cooper, G.; Morin, C.; Araki, T.; Turci, C. C.;
Kaznatcheev, K.; Hitchcock, A. P. J. Phys. Chem A 2003, 107, 6144-
(25) Zubavichus, Y.; Zharnikov1, M.; Schaporenko, A.; Grunze, M. J.
Electron Spectrosc. Relat. Phenom. 2004, 134, 25-33.
(26) Liu, X.; Jang, C.-H.; Zheng, F.; Jurgensen, A.; Denlinger, J. D.;
Dickson, K. A.; Raines, R. T.; Abbott, N. L.; Himpsel, F. J. Langmuir 2006,
(27) Hilpert, K.; Volkmer-Engert, R.; Walter, T.; Hancock, R. E. Nat.
Biotechnol. 2005, 23, 1008-1012.
(28) Romeo, D.; Skerlavaj, B.; Bolognesi, M.; Gennaro, R. J. Biol. Chem.
1988, 263, 9573-9575.
(29) Hilpert, K.; Hancock, R. E. W. Nat. Protoc., in press.
(30) Hilpert, K.; Winkler, F. H. D.; Hancock, R. E. W. Nat. Protoc.
2007, 2, 1333-1349.
(31) Kilcoyne, A. L. D.; Tylisczak, T.; Steele, W. F.; Fakra, S.;
Hitchcock, P.; Franck, K.; Anderson, E.; Harteneck, B.; Rightor, E. G.;
Mitchell, G. E.; Hitchcock, A. P.; Yang, L.; Warwick, T.; Ade, H. J.
Synchrotron Radiat. 2003, 10, 125-136.
(32) Warwick, T.; Ade, H.; Kilcoyne, A. L. D.; Kritscher, M.; Tylisc-
zcak, T.; Fakra, S.; Hitchcock, A. P.; Hitchcock, P.; Padmore, H. A. J.
Synchrotron Radiat. 2002, 9, 254-257.
(33) Jacobsen, C.; Wirick, S.; Flynn, G.; Zimba, C. J. Microsc. 2000,
(34) IDL is a scientific computing language developed by Research
Systems Inc, currently part of ITT Visual Imaging Solutions (http://
(35) Berman, H. M.; Henrick, K.; Nakamura, H. Nat. Struct. Biol. 2003,
(36) Henke, B. L.; Gullikson, E. M.; Davis. J. C. At. Data Nucl. Data
Tables 1993, 54, 181-342.
(37) aXis2000 is written in Interactive Data Language (IDL). It is
available free for noncommercial use from http://unicorn.mcmaster.ca/
(38) SF is a precursor to the program on the CXRO web site (http://
www.cxro.lbl.gov/optical_constants/) which computes elemental X-ray
absorption. The version implemented in aXis2000 was programmed for IDL
by Billy Loo.
7698 J. Phys. Chem. B, Vol. 111, No. 26, 2007
Stewart-Ornstein et al.
(39) Zubavichus, Y.; Shaporenko, A.; Grunze, M.; Zharnikov, M. J.
Phys. Chem. B 2006, 110, 3420-3427.
(40) Loo, B. W., Jr.; Sauerwald, I. M.; Hitchcock, A. P.; Rothman, S.
S. J. Microsc. 2001, 204, 69-86.
(41) Minghetti, P. P.; Ruffner, D. E.; Kuang, W. J.; Dennison, O. E.;
Hawkins, J. W.; Beattie, W. G.; Dugaiczyk, A. J. Biol. Chem. 1986, 261,
(42) Rixon, M. W.; Chan, W. Y.; Davie, E. W.; Chung, D. W.
Biochemistry 1983, 22, 3237-3244.
(43) Rixon, M. W.; Chung, D. W.; Davie, E. W. Biochemistry 1985,
(44) Rousseau, M. E.; Herna ´ndez-Cruz, D.; West, M. M.; Hitchcock,
A. P.; Pe ´zolet, M. J. Am. Chem. Soc. 2007, 129, 3897-3905.
(45) Ma, Y.; Chen, C. T.; Meigs, G.; Randall, K.; Sette, F. Phys. ReV.
A 1991, 44, 1848-1858
(46) Chen, C. T.; Ma, Y.; Sette, F. Phys. ReV. A 1989, 40, 6737-6740.
(47) Ishii, I.; Hitchcock, A. P. J. Electron Spectrosc. Relat. Phen. 1987,
Identifying and Mapping Peptides and Proteins
J. Phys. Chem. B, Vol. 111, No. 26, 2007 7699