Qualification of FTIR Spectroscopic Method for Protein

Secondary Structural Analysis

YIJIA JIANG,1 CYNTHIA LI,1 XICHDAO NGUYEN,1 SALMAN MUZAMMIL,2 ED TOWERS,3
JOHN GABRIELSON,3 LINDA NARHI1
1
Formulation and Analytical Resources Department, Product and Process Development, Amgen Inc.,
Thousand Oaks, California 91320
2
Biophysics and Developability, Antibody Drug Discovery, Centocor Inc., Radnor, Pennsylvania 19406
3
Analytical Sciences Department, Product and Process Engineering, Amgen Inc., Longmont, Colorado 80503

Received 15 November 2010; revised 16 March 2011; accepted 9 June 2011

Published online 28 June 2011 in Wiley Online Library (wileyonlinelibrary.com). DOI 10.1002/jps.22686

ABSTRACT: Fourier transform infrared (FTIR) spectroscopy is widely used to study protein
secondary structure both in solution and in the solid state. The FTIR spectroscopic method
has also been employed as a characterization method by the biopharmaceutical industry to
determine the higher order structure of protein therapeutics, and to determine if any changes in
protein conformation have occurred as a result of changes to process, formulation, manufacture,
and storage conditions. The results of these studies are often included in regulatory filings;
when comparability is assessed, the comparison is often qualitative. To demonstrate that the
method can be quantitative, and is suitable for these intended purposes, the precision and
sensitivity of the FTIR method were evaluated. The results show that FTIR spectroscopic
analysis is reproducible with suitable method precision, that is, spectral similarity of replicate
measurements is greater than 90%. The method can detect secondary structural changes caused
by pH and denaturant. The sensitivity of the method in detecting structural changes depends on
the extent of the changes and their impact on the resulting spectral similarity and characteristic
FTIR bands. The results of these assessments are described in this paper. © 2011 Wiley-Liss,
Inc. and the American Pharmacists Association J Pharm Sci 100:4631–4641, 2011
Keywords: algorithm; Infrared spectroscopy; method qualification; proteins; structure; FTIR;
protein secondary structure; spectral similarity

INTRODUCTION gion of 1600–1700 cm−1 , is primarily due to the amide

C O stretching vibrations. Different secondary struc-
Fourier transform infrared (FTIR) spectroscopy is an
tures, such as alpha-helix, beta-sheet, and beta-turn,
absorption spectroscopy that can be used to obtain in-
exhibit characteristic frequencies and intensities in
formation about the vibrational states of molecules.
the amide I band region due to differences in the
The FTIR spectrum of a protein is composed of many
hydrogen bonds in these structures.1,2 Therefore the
vibrational bands arising from different functional
FTIR spectrum of a protein is routinely used to deter-
groups such as N H, C O, and so on. The protein
mine and characterize protein secondary structures
backbone amide groups generate a number of charac-
in solution as well as in the solid state.3–5
teristic IR bands that can be used to determine pro-
Many attempts have been made to quantify the
tein backbone conformation and secondary structure.
percentage of different secondary structural compo-
In particular, the amide I band, which is in the re-
nents for proteins in solution by FTIR and CD (cir-
cular dichroism); these results are compared with
Abbreviations used: FTIR, Fourier transform infrared; Gdn,
guanidine hydrogen chloride; C3N, 20 mM sodium citrate with those obtained by X-ray crystallography in order to
140 mM sodium chloride buffer at pH 3.0.DTGD, Deuterated assess their accuracy.1,6–8 For FTIR, there are two pri-
Triglycine Sulfate mary ways of doing this. The first uses curve fitting
Correspondence to: Yijia Jiang (Telephone: +805-447-1116;
Fax: +805-499-3654; E-mail: yjiang@amgen.com)
of the second derivative or self-deconvoluted spec-
Journal of Pharmaceutical Sciences, Vol. 100, 4631–4641 (2011)
tra to obtain the relative amounts of different types
© 2011 Wiley-Liss, Inc. and the American Pharmacists Association of secondary structure based on the band areas.1,9

JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 100, NO. 11, NOVEMBER 2011 4631
4632 JIANG ET AL.
The second involves peak fitting of the nondeconvo-
luted and baseline corrected amide I bands, and then
obtains the percentage of secondary structures by cor-
relating with the shape and intensity using an in-
terval partial least squares algorithm.6 These meth-
ods are helpful in determining the primary secondary
structure components and estimating their percent-
ages in a protein. However, because the quantitation
of the secondary structure composition can vary de-
pending on the methods and parameters used and
the process typically involves considerable mathe-
matical manipulations, they are not routinely carried
out for protein secondary structure assessment by
the biopharmaceutical industry. Furthermore, these
methods are focused on quantitation of the secondary
structure composition of each individual protein, but
are not optimized for assessing overall similarity of
protein structure between samples from different pro-
cesses, or formulations. In contrast, the focus of this
work is to evaluate the precision and sensitivity of
the FTIR method for biopharmaceutical applications
in which a degree of similarity between the spectra
obtained from two or more samples is often the de-
sired result. The higher order structure of protein
therapeutics is complex in nature; the correct three-
dimensional structure is required for proteins to be
able to carry out their function and remain stable. It
can be affected by manufacturing processes includ-
ing refolding from inclusion bodies,10 low pH elu-
tion during chromatographic and viral inactivation
steps,11 and so on. To ensure the product quality
of a protein therapeutic, its higher order structure
needs to be assessed.12–14 FTIR spectroscopy is one
of the spectroscopic techniques that has been used
routinely to assess the secondary structure of pro-
tein therapeutics with the data included in regula-
tory filings. However, the qualification of the FTIR
method has mostly remained elusive, mainly due to
the lack of a quantitative way to directly compare
the FTIR spectra of different samples without curve
fitting or self-deconvolution. For this type of determi-
nation, the assignment of particular secondary struc-
ture content to the protein is not important. The fo-
cus is instead on the ability to detect small changes
in structure, regardless of the source of the change.
At present, there is no consistent method applied to
quantitatively determine the comparability of differ-
ent proteins, or of different lots of the same protein,
so visual comparisons of a sample spectrum to a ref-
erence have typically been used as a way to verify
proper folding of the protein. It is not straightforward
to apply principles from International Conference on
Harmonization guidelines Q2 (R1) “Validation of An-
alytical Procedures” to FTIR analyses because the
spectra are not quantitative without subsequent
mathematical treatment. A few approaches (mathe-
matical algorithms) have been explored to compare
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 100, NO. 11, NOVEMBER 2011
the FTIR spectra quantitatively,15,16 including the
correlation coefficient and area of overlap methods by
Kendrick et al.15 and Prestrelski et al.16 Each one has
strengths and weaknesses, which will be detailed in
a separate paper. In this paper, we applied a quanti-
tative function—OMNIC QC Compare17 —to directly
compare FTIR spectra and identified important per-
formance characteristics for the qualification of the
FTIR method.
To demonstrate that the FTIR method is suitable
for its intended applications, that is, to analyze a
protein therapeutic conformation and changes under
various process-related conditions, and to ensure the
proper folding of the protein during storage and deliv-
ery, the precision and sensitivity of the method were
assessed.
The precision of the method was evaluated through
a multisite/instrument and multi-analyst study,
wherein the same data collection and analysis method
were used. Proteins containing different types of sec-
ondary structure such as alpha-helical and beta-sheet
were included to ensure that the precision assess-
ment would apply to secondary structure analysis
of all proteins, regardless of the specific structural
type. However, most of the data were acquired using
proteins with beta-sheet secondary structure because
antibody and antibody fragment-based therapies are
prevalent in the biopharmaceutical industry. In addi-
tion, interday repeatability and effect of protein con-
centration (±10% of the targeted concentration) dif-
ferences on the precision of the method was also eval-
uated. Characteristic FTIR bands and overall spectral
similarity of spectra collected on the same proteins
were compared. The precision of FTIR analysis from
this study can be used to define method performance
parameters.
The sensitivity of the method was evaluated by
comparing the spectrum of a native or control pro-
tein to that of the partially or fully unfolded protein
after low pH or denaturant treatment and also by
blending studies, wherein a native protein spectrum
was mixed with that of a denatured protein and the
resulting spectra were compared with that of the na-
tive protein. Standard curves were generated where
the spectral similarity was plotted as a function of
the percentage of the unfolded protein. Spectral simi-
larity determined using the OMNIC QC compare tool
was used to quantitatively assess the sensitivity of
the FTIR method.
MATERIALS AND METHODS
Materials
Proteins 1–11 in Table 1 were produced at Amgen Inc.
and were at least 98% pure by size-exclusion chro-
matography. They were kept in their stable storage
DOI 10.1002/jps
 FTIR SPECTROSCOPIC METHOD FOR PROTEIN STRUCTURAL ANALYSIS
Table 1. Proteins Used in the Studies
Protein Characteristics
Protein 1 IgG2 mAb, beta-sheet protein
Protein 2 Fc conjugate, beta-sheet protein
Protein 3 Fc conjugate, beta-sheet protein
Protein 4 Growth factor, beta-sheet protein
Protein 5 Cytokine, alpha-helical protein
Protein 6 IgG1 mAb, beta-sheet protein
Protein 7 Fc conjugate, beta-sheet protein
Protein 8 Cytokine, alpha-helical protein
Protein 9 IgG1 mAb, beta-sheet protein
Protein 10 IgG2 mAb, beta-sheet protein
Protein 11 Fc conjugate, beta-sheet protein
buffers at 2◦ C–8◦ C until analysis. All other chemicals
including citric acid, sodium chloride, and guanidine
hydrochloride were obtained from Sigma–Aldrich
(St. Louis, MO) and were of reagent grade. Milli-Q
water was used to prepare all solutions. Millipore
ultracentrifuge filters (Biomax-0.5) with a molecular
weight cutoff of 5 kDa were used to concentrate pro-
tein samples to concentrations of at least 30 mg/mL
for FTIR analysis.
Methods
An instrument suitability test was carried out with a
polystyrene film (3.0 mil) every 3 months by following
the vendor provided procedure. Regular preventative
maintenance of the instrument was performed and
logged to ensure proper calibration and system suit-
ability for the studies.
Fourier transform infrared spectra of protein solu-
tions were recorded at room temperature. The instru-
ments used included the Bruker VERTEX 70, Nicolet
Magna 550, or BOMEM MB 154S FTIR spectropho-
tometers in transmission mode. Protein samples
prepared at at least 30 mg/mL were measured in a
sample cell that employed CaF2 windows separated
by a 6-micron spacer, for example, an FTIR liquid
measurement cell (Spectra-Tech FT04-036), Bruker
AquaSpec Cell, or equivalent. For each spectrum, at
least a 256-scan interferogram was collected in a sin-
gle beam mode, with a 4 cm−1 resolution. Reference
spectra were recorded under identical conditions with
appropriate filtrate/buffer blank in the cell. The spec-
tra for the reference and gaseous water were sub-
tracted from the protein spectra, according to pre-
viously established criteria.18 The second-derivative
spectrum was calculated using the application soft-
ware and the final spectrum smoothed with a nine-
point function to remove white noise. The nine-point
smooth was selected for the following two reasons:
(1) As the spectra of the proteins were collected on
several different instruments, even at the same pro-
tein concentration, the signal to noise (S/N) levels
DOI 10.1002/jps
4633
were quite different, especially at lower concentra-
tions, and a nine-point smooth was necessary for
the noisier spectra and (2) for spectra collected us-
ing the Bruker instrument, the application software
(OPUS) has nine-point smooth as the default setting
when the second-derivative operation is performed. A
seven-point smooth is not possible. So, to be consis-
tent across different instruments, proteins, and pro-
tein concentrations, we chose a nine-point smooth for
all the spectra collected.
The similarity of the second-derivative spectra in
the amide I region was calculated using the Ther-
mal Electron Omnic Software QC Compare function
at 8 cm−1 resolution.17 The QC Compare function cor-
relates the spectral information in a user-specified
region of sample and reference spectra to determine
the similarity between the spectra. The result is a
match value between 0% and 100%. The value shows
how well the sample spectrum matches the reference
spectrum; a value of 100% indicates the spectra are
identical.
For the precision assessment, the repeatability of
the method was assessed by triplicate measurements
of the same proteins (Protein 1–5 and a mixture of
50% Protein 5 plus 50% Protein 9 at the same protein
concentration) on the same day. Interday repeatabil-
ity was determined by collecting three spectra of Pro-
tein 1 over 3 days. The effect of concentration was
assessed using Protein 1 at ±10% of the target con-
centrations at 30 and 150 mg/mL.
For the sensitivity assessment, selected proteins
(Protein 1, 6–8) were analyzed in their respective sta-
ble storage buffers, at pH 3.0 in 20 mM NaCitrate,
140 mM NaCl buffer (C3N), or in the presence of 6
M guanidine hydrogen chloride (Gdn). The changes
caused by the pH 3.0 (C3N) or 6 M Gdn treatment
were assessed by comparing the spectra of the pro-
teins under stable storage conditions to those at pH
3.0 or in 6 M Gdn, respectively.
Blending studies were carried out with native pro-
teins of different structure types [Protein 9—an IgG1
monoclonal antibody (mAb) with Protein 5—an alpha-
helical protein, Protein 10—an IgG2 mAb, or Protein
11—an Fc conjugated protein] at different mass ra-
tios, both experimentally and mathematically to es-
tablish correlations between the signal and the folded
state of the protein. Mathematically simulated blend-
ing studies were then carried out with spectra of na-
tive/control (under stable storage condition) and the
pH 3.0 or 6 M Gdn-treated proteins at different mass
ratios to evaluate the sensitivity of the FTIR method.
Applicability of the mathematical blending results as
a model for the experimental mixing results was es-
tablished based on the results from the native protein
blending study.
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 100, NO. 11, NOVEMBER 2011
4634 JIANG ET AL.

Figure 1. Second-derivative FTIR spectra of the different proteins by the same (Bruker)
instrument (pink, Protein1; purple, Protein 2; green, Protein 3; aqua, Protein 4; black,
Protein 5).

RESULTS AND DISCUSSION
Precision Assessment
The precision of the FTIR measurements was eval-
uated through a multi-instrument, multi-analyst,
and multiprotein study. The representative second-
derivative FTIR spectra of proteins containing dif-
ferent secondary structural types (Protein 1–5) are
shown in Figure 1. The characteristic band frequen-
cies and spectral similarity of the triplicate measure-
ments of the second-derivative FTIR spectra of the
proteins (Protein 1–5 and 50% Protein 5 plus 50%
Protein 9 mixed at the same protein concentration)
are shown in Table 2. For Protein 1, the characteris-
tic amide I bands are at about 1637 and 1690 cm−1 ,
indicating the presence of antiparallel beta-sheet sec-
ondary structure in the protein, typical of an IgG
mAb. For Proteins 2 and 3, the characteristic IR bands
are at 1641/1690 cm−1 and 1643/1688 cm−1 , respec-
tively, suggesting the presence of primarily beta-sheet
secondary structure, as normal for the Fc domain of
the molecules. For Protein 4, the characteristic bands
are at about 1622, 1645, and 1688 cm−1 , again indi-
cating the presence of predominantly beta-sheet sec-
ondary structure as expected. Protein 5 shows the ma-
jor characteristic FTIR band at 1655 cm−1 together
with a weak band at 1634 cm−1 , demonstrating the
presence of mainly alpha-helical secondary structure

Table 2. The Means and Standard Deviations of Characteristic Band Frequencies and Spectral Similarity of the Second-Derivative
FTIR Spectra

Characteristic FTIR
Bandsa (cm−1 ) Overall Spectral Similarity Spectral Similarity Spectral Similarity Overall Spectral
Proteins mean ± SD (%) (Bruker) (%) (Nicolet) (%) (Bomem) Similarityb (%)
Protein 1 1637.5 ± 0.3, 1690.4 ± 1.0 >99 ≥97 >99 ≥ 97
Protein 2 1641.2 ± 0.04, 1689.6 ± 0.4 >96 > 93 >99 >93
Protein 3 1642.8 ± 0.5, 1688.2 ± 1.1 >99 >95 >99 >95
Protein 4 1622.4 ± 0.6, 1644.7 ± 0.6, >99 >96 >96 >96
1687.6 ± 0.3
Protein 5 1633.8 ± 0.6, 1654.7 ± 0.04 >99 >97 >97 >94
Mixture of 50% Protein 5 1638.4 ± 0.0, 1654.8 ± 0.0 >99 NDd NDd NDd
and 50% of Protein 9c
a Band frequencies were averaged among the spectra by the different instruments.
b Overallspectral similarity was determined for replicate measurements using the same instrument.
c Mixed at the same protein concentration.
d Not determined.

JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 100, NO. 11, NOVEMBER 2011 DOI 10.1002/jps
 FTIR SPECTROSCOPIC METHOD FOR PROTEIN STRUCTURAL ANALYSIS
Figure 2. Second-derivative FTIR spectra of computer simulated and experimental mixing of
Protein 9 and Protein 5 (red, 100% Protein 5; purple, 75% Protein 5 + 25% Protein 9; green,
50% Protein 5 + 50% Protein 9; pink, 25% Protein 5 + 75% Protein 9; navy, 10% Protein 5 +
90% Protein 9; blue, 100% Protein 9).
with a small amount of beta-sheet structure in this
protein. A mixture of 50% Proteins 5 and 9 combined
at the same protein concentration was used to mimic
a protein with mixed secondary structure (Fig. 2b,
green trace) and showed the major IR bands at 1655
and 1638 cm−1 as expected. The characteristic band
frequencies for the same proteins are averaged among
the spectra obtained using different instruments. The
characteristic band frequencies are quite reproducible
with standard deviations of 1.1 cm−1 or less on all five
proteins tested, regardless of the instrument, analyst,
or types of secondary structure.
The main reasons for the variability in the char-
acteristic band frequencies are broader and weaker
FTIR bands, which may be improved by increasing
the instrument resolution (e.g., from 4 to 2 cm−1 ).
However, due to the overlapping nature of protein
IR bands and the corresponding increase in noise
level with increasing instrument resolution, this ap-
proach is not recommended. As shown by Dong and
Caughey,18 and Susi and Byler,9 the deconvoluted
amide I band frequencies for the assigned secondary
structures typically have a variability of ±2 cm−1 .
The standard deviations of 1.1 cm−1 or less for the
same proteins across the different instruments and
by different analysts are well within that range and
therefore will not affect the secondary structure de-
termination of proteins. These results indicate that
FTIR analysis is quite precise for characteristic band
frequencies.
The spectral similarities of the replicate mea-
surements using the same instrument are greater
than 93% by OMNIC QC Compare in all cases
(Table 2). The overall spectral similarity for the repli-
DOI 10.1002/jps
4635
cate measurements does not depend on the specific
type of secondary structure. Replicate measurements
of proteins with predominantly alpha-helical (Protein
5) or beta-sheet (Proteins 1–4) secondary structure
or a mixture of both (50% Proteins 5 and 9 mixed at
the same protein concentration) have similar spectral
similarities and measurement precision. This is con-
sistent with the results of intramethod reproducibility
shown by van de Weert et al.,19 where they obtained
at least 94% spectral similarity using the percentage
of area overlap15 for various proteins measured in
transmission mode in water.
The reasons for the variability in the spectral sim-
ilarity include moisture/vapor, water background19
interferences, and lower S/N ratio. Precision can
be improved by minimizing the contribution of wa-
ter vapor to the protein spectra and increasing the
S/N levels by increasing protein concentration, in-
creasing the number of scans, and/or using a
more sensitive detector [MCT(Mercury Cadmium Tel-
luride) instead of DTGS(Deuterated Triglycine Sul-
fate)]. The effect of moisture was demonstrated by the
significant improvement of the similarity of the trip-
licate spectra (to >96% from >93%) using the Bomem
normal transmission cell with a DTGS detector where
moisture interference was decreased due to a closed
system design for the optical components in the beam
path. The effect of less moisture and higher signals
was also demonstrated by results from the Bruker,
where the spectra collected showed decreased contri-
butions from vapor, and increased S/N resulting in
the increased spectral similarity (>96%) of replicate
measurements. This is due to the use of an AquaSpec
cell and an MCT detector that led to the decreased
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 100, NO. 11, NOVEMBER 2011
4636 JIANG ET AL.

Table 3. The Characteristic Band Frequencies and Similarity of the Spectra at Different Protein Concentrations Compared with That at
30 and 150 mg/mL

Nicolet Magna 550, Normal Transmission Cell, DTGS detector 1600–1700 cm−1 , 9 points, 8 cm−1 resolution

Characteristic FTIR Spectral Similarity to Spectral Similarity to

Protein 1 (mg/mL) Bands (cm−1 ) 30 mg/mL (%) 150 mg/mL (%)
30 1638.0, 1691.3 100 ND
27 1637.3, 1691.3 96.5 ND
33 1638.3, 1691.3 94.9 ND
135 1637.3, 1691.3 98.5 99.7
165 1637.3, 1691.3 97.8 99.5
Overall mean ±SD 1637.6 ± 0.5, 1691.3 ± 0.0 >94 >99

ND, not determined.

moisture interference and increased S/N level. In-
crease in protein concentration and the resulting S/
N ratio can also increase the spectral similarity as
shown in Table 3, where spectral similarity increased
from about 95% at 30 mg/mL to approximately 98%
at about 150 mg/mL.
Effect of Interday and Intraday Precision
and Concentration on FTIR Analysis
The comparison of the interday and intraday tripli-
cate measurements was carried out on Protein 1 us-
ing a Nicolet FTIR spectrophotometer and the results
are shown in Table 4. As can be seen from the ta-
ble, the precisions of the band frequencies of the in-
terday measurements are comparable to or slightly
better than those of the intraday triplicate measure-
ments. The spectral similarities are also comparable
for both the interday (>98%) and intraday (≥97%)
multiple measurements. These results demonstrate
that the precision of the FTIR measurement depends
on the instrument used and the bandwidth of a spe-
cific band, but not on the analyst, the type of protein,
nor the time between the acquisitions of the spectra.
The precision of the FTIR measurement remains un-
changed for the spectra collected on the same day or
during a 3-day period as long as the protein itself is
stable over the time period the data is being collected.
The effect of protein concentration on the preci-
sion of the FTIR measurement was also assessed
using Protein 1 and the Nicolet FTIR instrument.
Protein 1 solutions were prepared at ±10% of the
target concentrations (30 and 150 mg/mL) based
on typical quality control specifications. The char-
acteristic band frequencies and similarity of the
spectra at different protein concentrations com-
pared with that at 30 or 150 mg/mL are shown in
Table 3. The standard deviation of the band frequen-
cies of the spectra of Protein 1 at different concen-
trations is within the range of those of the intraday/
interday multiple measurements of the same protein
sample on the same instrument (Tables 2 and 4). The
spectral similarity is greater than 94%. The results
indicate that the concentration variations of ±10%
from the target protein concentration of 30 mg/mL
and above did not adversely affect the precision of
the FTIR measurement.
However, as shown in Figure 3 and Table 3, the
quality of the second-derivative FTIR spectra and
overall spectral similarity is protein concentration de-
pendent. The higher the protein concentration, the
stronger signal from the protein, the less the interfer-
ence from water and vapor, and the better the spec-
tra obtained. The lower protein concentration limit
of an FTIR measurement is dependent on the S/N of
the instrument and the interference from the water

Table 4. The Characteristic Band Frequencies and Spectral Similarity of the Second-derivative
FTIR Spectra of Protein 1 Measured Using the Same Instrument

1600–1700 cm−1 ,
Nicolet Magna 550, Normal Transmission Cell, DTGS detector 9 points, 8 cm−1 Resolution

Interday Characteristic FTIR Bands (cm−1 ) Spectral Similarity (%)

Day 1 1638.0, 1691.3 1 vs. 2—98.9
Day 2 1637.4, 1691.3 2 vs. 3—99.3
Day 3 1637.4, 1691.3 3 vs. 1—99.2
Mean ± SD 1637.6 ± 0.3, 1691.3 ± 0.0 >98
Intraday
Measurement 1 1638.3, 1691.3 1 vs. 2—97.5
Measurement 2 1637.3, 1691.3 2 vs. 3—98.2
Measurement 3 1637.3, 1691.3 3 vs. 1—97.0
Mean ± SD 1637.6 ± 0.6, 1691.3 ± 0.0 ≥97

JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 100, NO. 11, NOVEMBER 2011 DOI 10.1002/jps
 FTIR SPECTROSCOPIC METHOD FOR PROTEIN STRUCTURAL ANALYSIS
Figure 3. Second-derivative FTIR spectra of Protein 1 at various concentrations (green, 2;
aqua, 5; pink, 10; red, 15; navy, 30 mg/mL) by Bruker FTIR spectrophotometer.
vapor. In general, FTIR analysis requires high con-
centration protein solutions (typically >10 mg/mL) to
obtain a sufficient S/N. For proteins that must be
analyzed at much lower concentrations, FTIR anal-
ysis may be inappropriate, or at a minimum, the
data are much more difficult to interpret quantita-
tively. Samples with protein concentrations of 10 mg/
mL or higher in the same buffer had equivalent FTIR
spectra, and therefore secondary structure indepen-
dent of the protein concentration. Similar results
were also obtained for protein solutions analyzed by
CD spectroscopy from 10 to about 0.5 mg/mL (data
not shown here). However, if the protein is unstable
or partially unfolded under the conditions being an-
alyzed, then both protein concentration and analysis
time can affect the conformation as it unfolds and
aggregates during testing. For the precision data re-
ported in this paper, the proteins analyzed were in so-
lution conditions, which stabilized the conformation.
For an FTIR instrument with good S/N and water
vapor well controlled, a precision of more than 95%
spectral similarity of the measurements can be rou-
tinely achieved with a protein solution at high single
digit protein concentration (data not shown).
The results from the multiple measurements of
the same proteins demonstrate that FTIR analysis
of protein conformation is precise, with characteris-
tic band frequency from replicate measurements typ-
ically within 1 cm−1 .
Sensitivity Assessment
The sensitivity of the FTIR method was evaluated
using a few different approaches. The first was to
compare the spectra of denatured proteins to those
DOI 10.1002/jps
4637
of the same proteins under stable storage conditions.
The FTIR spectra of selected proteins from different
structural categories (Protein 6—IgG1 mAb, Protein
1—IgG2 mAb, Protein 7—Fc conjugate, beta-sheet,
and Protein 8—cytokine, alpha-helical) under stable
storage, C3N, and 6 M Gdn conditions are overlaid in
Figures 4a–4d. From the figures, it is clear that in 6 M
Gdn the secondary structures of the proteins with dif-
ferent structural types were mostly unfolded and un-
ordered with the major FTIR band very broad at about
1648 cm−1 . Gdn is known to absorb strongly in the
amide I region, which may interfere with the protein
spectra if not fully subtracted.20,21 Work performed by
Bowler et al.20,21 has shown that Gdn at more than
3.5 M could not be accurately subtracted due to detec-
tor saturation. The peak of Gdn absorption in aque-
ous solution is at about 1680 cm−1 , whereas the main
band of the unfolded protein is at about 1650 cm−1 .
Mathematical manipulations undertaken to test the
effect of Gdn concentration on data analysis involv-
ing intentionally under or over subtraction of Gdn
resulted in similar spectral similarity values (<30%)
compared with the control and main protein band po-
sition (about 1650 cm−1 ). So even with interference
from Gdn, it will not change the conclusion—proteins
were fully unfolded with the major band at about
1650 cm−1 and the FTIR method is able to detect and
determine the secondary structure changes. Strong
Gdn absorption would be a problem if one wants to
accurately determine the changes in the different sec-
ondary structure components. However, that is out-
side of the scope of this study. High concentrations
of Gdn were used in our experiments simply to pro-
duce spectral changes outside of method variability
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 100, NO. 11, NOVEMBER 2011
4638 JIANG ET AL.

Figure 4. Second-derivative FTIR spectra of Proteins 1, 6, 7, and 8 overlaid under stable

storage, C3N and 6 M Gdn conditions (red, native/control; navy, 6M Gdn; green, C3N).

as determined by OMNIC QC Compare for the over-
all spectra in amide I region.
The spectral similarity results of the selected pro-
teins in C3N or 6 M Gdn as compared with the spectra
of the same proteins under their stable storage condi-
tions (controls) are presented in Table 5. Significant
differences were observed in the FTIR spectra of the
proteins in the presence of 6 M Gdn due to the fact
that the proteins are mostly unfolded in this denatur-
ing buffer. The spectra similarities of Protein 1 and
Protein 7 in C3N as compared with their correspond-
ing controls are <30%, indicating extensive unfolding
of the proteins at pH 3.0. The spectral similarities of
Protein 6 and Protein 8 in C3N as compared with the
controls are greater than 95%. This is not surprising
because Protein 8 is formulated at a pH that is very
close to the pH of C3N buffer and Protein 6 is an IgG 1
mAb, which is more stable to low-pH induced unfold-
ing than IgG 2 (Protein 1) (Jiang et al. unpublished
observations).
Although differences at particular wavenumbers or
peaks could be more sensitive for specific changes
in conformation for specific proteins, the method de-
scribed here can be applied more universally to com-
pare the overall secondary structure, and is applicable
to all proteins regardless of the secondary structure
content or what the actual peak wavenumber and in-
tensity is.
The above FTIR results demonstrate that the FTIR
method is able to detect changes in the proteins
caused by low pH and denaturant and is appropri-
ate for assessing protein secondary structure changes
that may occur during manufacturing processes.

Table 5. Spectral Similarity of Proteins Under Different

Conditions Compared with Those in Stable Storage Buffers
Blending Studies with Native Proteins
(Control) (1600–1700 cm−1 ) Several proteins from different structural categories
Spectral Similarity (%) were used to assess whether computer-simulated
blending can be used to complement experimental
Protein Control to Itself C3N to Control Gdn to Control blending. Both empirical blending and computer-
Protein 1 100 2 19 simulated blending studies using proteins contain-
Protein 6 100 98 22 ing Protein 9 (IgG1, beta-sheet) and Protein 5
Protein 7 100 23 11
(Cytokine, alpha-helical), Protein 9 (IgG1 beta-sheet)
Protein 8 100 100 27
and Protein 10 (IgG2, beta-sheet), or Protein 9 (IgG1,

JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 100, NO. 11, NOVEMBER 2011 DOI 10.1002/jps
 FTIR SPECTROSCOPIC METHOD FOR PROTEIN STRUCTURAL ANALYSIS 4639

Figure 5. Second-derivative FTIR spectra of computer-simulated mixing of Protein 1 in native

state with 6 M Gdn or C3N denatured states (red, 100% native Protein 1; pink, 95% native
Protein 1 + 5% Protein 1 in 6 M Gdn or C3N; light blue, 90% native Protein 1 + 10% Protein
1 in 6 M Gdn or C3N; green, 75% native Protein 1 + 25% Protein 1 in 6 M Gdn or C3N; light
green, 50% native Protein 1 + 50% Protein 1 in 6 M Gdn or C3N; purple, 25% native Protein 1
+ 75% Protein 1 in 6 M Gdn or C3N; yellow, 10% native Protein 1 + 90% Protein 1 in 6 M Gdn
or C3N; blue, 100% Protein 1 in 6 M Gdn or C3N).

beta-sheet) and Protein 11 (Fc conjugate, beta-sheet)
were carried out. Experimentally, one protein sam-
ple was mixed with another at various mass ra-
tios and the spectra of the resulting mixed samples
were collected (spectra not shown). For the computer-
simulated study, the spectra of the original protein
samples were mathematically blended using the same
mass ratios as employed in the empirical study. The
second-derivative FTIR spectra of computer simu-
lated and experimental mixing of Protein 9 and Pro-
tein 5 are shown in Figures 2a and 2b as an ex-
ample. The spectral similarity results from both the
empirical and computer-simulated blending are sum-
marized in Table 6. The results show that the differ-
ences between the spectral similarities of the com-
puter simulated and experimental mixings are 2% or
less for every combination of proteins and mixing ra-
tios, demonstrating that the spectra obtained using
experimental and computer mixing are very similar
at the same mass ratios. The results indicate that the
computer simulated blending may be used as an al-
ternative for the experimental blending to assess the
capability and sensitivity of FTIR analyses to detect
structural changes.
A second approach to evaluate the sensitivity of
the FTIR method is through the blending studies of
proteins in stable storage buffers with those in C3N
or 6 M Gdn by computer simulation. Because of the

Table 6. FTIR Spectral Similarity of the Mixing (Both Computer Simulated and Experimental) Studies with Native Proteins
(1600–1700 cm−1 )

Resulting Spectra Similarity to Protein 9

Protein 10 (IgG2) Protein 11 (Fc conjugate) Protein 1 (IgG2)

Protein 9 (%) (IgG1) The Other Protein (%) Computer Experiment Computer Experiment Computer Experiment
100 0 100 100 100 100 100 100
90 10 99 99 99 99 87 88
75 25 96 96 98 97 47 49
50 50 86 86 89 88 19 19
25 75 72 74 71 72 3 3
0 100 59 59 53 53 <3 <3

DOI 10.1002/jps JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 100, NO. 11, NOVEMBER 2011
4640 JIANG ET AL.

Table 7. The FTIR Spectral Similarity Results of the Computer-Simulated Blending Study

Mixing Ratio Spectral Similarity to Control (%)

Protein 1 (IgG2) Protein 6 (IgG1) Protein 7 (Fc conjugate) Protein 8 (cytokine)

Control 6 M Gdn or C3N 6 M Gdn C3N 6 M Gdn C3N 6 M Gdn C3N 6 M Gdn C3N
100 0 100 100 100 100 100 100 100 100
95 5 100 99 99 N/A 99 100 100 N/A
90 10 98 96 99 N/A 95 98 99 N/A
75 25 83 74 91 NA 70 89 91 NA
50 50 46 40 51 N/A 40 56 63 N/A
25 75 33 19 17 N/A 25 38 42 N/A
10 90 24 6 10 N/A 16 29 34 N/A
0 100 19 2 22 98 11 23 27 100

reversible nature of the structural changes induced
both at low pH and in the presence of denaturant,
mixing of the actual proteins under these different
conditions will result in reversal of some of the
changes as the protein reaches equilibrium under
these new unfolding conditions. Assuming certain
population of the unfolded species remains after mix-
ing, to assess the sensitivity of FTIR in detecting the
low pH or 6 M Gdn induced structural changes, the
FTIR spectra of the selected native proteins (Protein
1—IgG2, Protein 6—IgG1, Protein 7—Fc conjugate,
and Protein 8—cytokine) were mathematically mixed
with those of the denatured forms (pH 3.0 or 6M Gdn)
at various mass ratios. The second-derivative FTIR
spectra of computer-simulated mixing of Protein 1 in
native state with 6 M Gdn or C3N denatured states
are shown in Figures 5a and 5b as an example.
The spectral similarities of the resulting mixed
spectra compared with those of the controls are sum-
marized in Table 7 and plotted as a function of the
percentage of the denatured protein in Figures 6 (6 M
Gdn) and 7 (C3N). The FTIR analysis is the most
sensitive in detecting structural changes for those
proteins where the degree of change in the spectra
is greatest under the different conditions being as-
sessed. For example, based on the plots, for Protein 7
the 6 M Gdn induced unfolding and for Protein 1
the C3N induced unfolding can be easily detected
(spectral similarity falls below 90%) when the mix-
tures contain 15% of the unfolded species. However,
when structural changes are less, resulting in smaller
changes in the spectra, spectral changes will only be
observed at a higher level of the denatured species.
This is demonstrated by the case of Protein 8 exposed
to 6 M Gdn and C3N. In the presence of 6 M Gdn,
there were significant structural changes in Protein
8, which caused a high degree of changes in the FTIR
spectrum with 27% spectral similarity compared with
the control. In this case, it takes more than 25% of
the 6 M Gdn denatured protein to cause significant
spectral changes (<90% spectral similarity) in the
mixture. In C3N, where there were minimal changes
to the structure of Protein 8 compared with the sta-
ble storage condition (spectral similarity >95%), com-
puter mixing of spectra was unable to detect differ-
ences in the structure at any mixing ratios (Table 7
and Fig. 7). The results clearly show that the sensitiv-
ity of the FTIR method to detect structural changes

Figure 6. Spectral similarity as a function of percentage Figure 7. Spectral similarity as a function of percentage
of denatured protein by 6 M Gdn (navy diamond, Protein of denatured protein by C3N (navy diamond, Protein 1; pink
1; pink square, Protein 6; yellow triangle, Protein 7; aqua square, Protein 6; yellow triangle, Protein 7; aqua cross,
cross, Protein 8). Protein 8).

JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 100, NO. 11, NOVEMBER 2011 DOI 10.1002/jps
 FTIR SPECTROSCOPIC METHOD FOR PROTEIN STRUCTURAL ANALYSIS
is dependent on the nature of the structural changes
and extent of the spectral changes caused by those
structural changes.
CONCLUSION
The above reported results described the assessment
of precision and sensitivity of the FTIR method on the
analysis and characterization of protein secondary
structures. Five proteins from different structural
categories were selected for the precision assessment.
The study demonstrates that the method is precise
with the standard deviations of the characteristic
FTIR band frequencies (<1.1 cm−1 ) and the spec-
tral similarity (>93%) for replicate measurements.
The precision of the FTIR measurement remains un-
changed for the spectra collected on the same day or
during a 3 day period (assuming the protein being as-
sessed is stable over this time) and when the protein
concentration deviates from the target concentration
(≥30 mg/mL) by 10% or less. The precision obtained
can be used to justify method performance parame-
ters. The results apply to FTIR measurements of pro-
teins regardless of the maker of the instrument used
or the protein structural categories.
The sensitivity of the FTIR method in detect-
ing conformational changes in proteins from differ-
ent structural categories (alpha-helical, beta-sheet;
mAbs, cytokines; etc.) resulting from common bio-
pharmaceutical manufacturing stresses such as pH
and denaturant was analyzed. The results demon-
strate that the method is sensitive and appropriate for
assessing protein secondary structure, and changes in
conformation resulting from stresses that can be en-
countered during common manufacturing processes.
A mixing study was carried out to further assess the
sensitivity of the method, which demonstrated that
FTIR sensitivity is dependent on the extent of the
structural changes induced and the magnitude of the
resulting changes in the spectra.
For both the precision and sensitivity assessment,
the OMNIC QC compare algorithm was used to quan-
titatively compare the spectra in this study. However,
other methods such as area of overlap by Kendrick
et al.15 may also be used, and we continue to evaluate
additional methods as appropriate. The choice of the
method will be based on several factors, including the
capability and sensitivity of the method, the ease of
access and use of the method, and the purpose of the
study, and so on, which we will discuss in a separate
paper.
ACKNOWLEDGMENTS
The authors would like to thank Shengwu Wang,
Bob Bailey, Brent Kendrick, Vladimir Razinkov, and
David Brems for valuable input and helpful discus-
sions.
DOI 10.1002/jps
4641
REFERENCES
1. Dong A, Huang P, Caughey WS. 1990. Protein secondary struc-
tures in water from second-derivative amide I infrared spectra.
Biochemistry 29(13):3303–3308.
2. Mantsch HH, Chapman D. 1996. Infrared spectroscopy of
biomolecules: Theoretical analyses of the amide I infrared bands
of global proteins. New York: Wiley-Liss, Inc.
3. Carpenter JF, Prestrelski SJ, Dong A. 1998. Application of in-
frared spectroscopy to development of stable lyophilized protein
formulations. Eur J Pharm Biopharm 45(3):231–238.
4. Fu K, Griebenow K, Hsieh L, Klibanov AM, Langer R. 1999.
FTIR characterization of the secondary structure of proteins
encapsulated within PLGA microspheres. J. Control Release
58:357–366.
5. Haris PI, Chapman D. 1995. The conformational analysis of
peptides using Fourier transform IR spectroscopy. Biopolymers
(Peptide Science) 37:251–263.
6. Vonhoffa S, Condliffeb J, Schiffterb H. 2010. Implementation of
an FTIR calibration curve for fast and objective determination
of changes in protein secondary structure during formulation
development. J Pharm Biomed Anal 51(1):39–45.
7. Susi H, Byler M. 1987. Fourier transform infrared study
of proteins with parallel b-chains. Arch Biochem Biophys
258(2):465–469.
8. Johnston MJW, Nemra K, Hefford MA. 2010. Influence of bovine
serum albumin on the secondary structure of interferon alpha
2b as determined by far UV circular dichroism spectropolarime-
try. Biologicals 38(2):314–320.
9. Susi H, Byler M. 1986. Resolution enhanced fourier trans-
form infrared spectroscopy of enzymes. Methods Enzymol
130:290–311.
10. Mannall GJ, Titchener-Hooker NJ, Dalby PA. 2007. Factors
affecting protein refolding yields in a fed-batch and batch-
refolding system. Biotechnol Bioeng 97(6):1523–1534.
11. Hahn R, Schlegel R, Jungbauer A. 2002. Comparison of protein
A affinity sorbents. J Chromatography B 790:35–51.
12. ICH guideline Q6B Specifications: Test procedures and ac-
ceptance criteria for biotechnological and biological products.
September, 1999.
13. Chirino AJ, Mire-Sluis A. 2004. Characterizing biological prod-
ucts and assessing comparability following manufacturing
changes. Nat Biotechnol 22 (11):1383–1391.
14. ICH guideline Q5E Guidance for industry: Comparability of
biotechnological/biological products subject to changes in their
manufacturing process. June, 2005.
15. Kendrick BS, Dong A, Allison SD, Manning MC, Carpenter
JF. 1996. Quantitation of overlap of infrared second derivative
spectra to determine structural similarity between proteins.
J Pharm Sci 85:155–158.
16. Prestrelski S, Tedeschi N, Arakawa T, Carpenter J. 1993. Bio-
phys J 65(2):661–671.
17. Cover TM, Hart PE. 1967. Nearest neighbor pattern classifica-
tion. IEEE Trans Inf Theor 13(1):21–27.
18. Dong A, Caughey WS. 1994. Infrared methods for study
of hemoglobin reactions and structure. Methods Enzymol.
232:139–175.
19. van de Weert M, Haris PI, Hennink WE, Crommelin DDA.
2001. Fourier transform infrared spectroscopic analysis of pro-
tein conformation: Effect of sampling method and stress factors.
Anal Biochem 297:160–169.
20. Bowler BE, May K, Zaragoza T, York P, Dong A, Caughey
WS. 1993. Destabilizing effects of replacing a surface ly-
sine of cytochrome c with aromatic amino acids. Biochemistry
32:183–190.
21. Bowler BE, Dong A, Caugheys WS. 1994. Characteriza-
tion of the guanidine hydrochloride-denatured state of iso-
1-cytochrome c by infrared spectroscopy. Biochemistry 33:
2402–2408.
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 100, NO. 11, NOVEMBER 2011