Clinical Chemistry 47:4
730 –738 (2001)
Plasma Protein Contents Determined by
Fourier-Transform Infrared Spectrometry
Cyril Petibois,1,2 Georges Cazorla,2 André Cassaigne,3 and Gérard Déléris1*
Background: Fourier-transform infrared (FT-IR) spec-
trometry has been used to measure small molecules in
plasma. We wished to extend this use to measurement of
plasma proteins.
Methods: We analyzed plasma proteins, glucose, lactate,
and urea in 49 blood samples from 35 healthy subjects
and 14 patients. For determining the concentration of
each biomolecule, the method used the following steps:
(a) The biomolecule was sought for which the correla-
tion between spectral range areas of plasma FT-IR
spectra and concentrations determined by comparison
method was greatest. (b) The IR absorption of the
biomolecule at the most characteristic spectral range
was calculated by analyzing pure samples of known
concentrations. (c) The plasma concentration of the
biomolecule was determined using the FT-IR absorp-
tion of the pure compound and the integration value
obtained for the plasma FT-IR spectra. (d) The spectral
contribution of the biomolecule was subtracted from the
plasma FT-IR spectra, and the resulting spectra were
saved for further analyses. (e) The same method was
then applied to determining the concentrations of other
biomolecules by sequentially comparing the resulting
FT-IR spectra.
Results: Results agreed with those obtained by clinical
methods for the following biomolecules when analyzed
in the following order: albumin, glucose, fibrinogen,
IgG2, lactate, IgG1, ␣1-antitrypsin, ␣2-macroglobulin,
transferrin, apolipoprotein (Apo)-A1, urea, Apo-B, IgM,
Apo-C3, IgA, IgG4, IgG3, IgD, haptoglobin, and ␣1-acid
glycoprotein.
1
INSERM U443, Equipe de Chimie Bio-Organique, 2 Faculté des Sciences
du Sport et de l’Education Physique, and 3 Département de Biochimie Médi-
cale et Biologie Moléculaire, Laboratoire de Biochimie, CHU Bordeaux, Uni-
versité Victor Segalen Bordeaux 2, 146 rue Léo Saignat, 33076 Bordeaux,
France.
*Author for correspondence. Fax 33-5-5757-1002; e-mail gerard.deleris@
bioorga.u-bordeaux2.fr.
Received June 5, 2000; accepted January 12, 2001.
730
Automation and
Analytical Techniques
Downloaded from https://academic.oup.com/clinchem/article-abstract/47/4/730/5639132 by guest on 08 March 2020
Conclusion: FT-IR spectrometry is a useful tool for deter-
mining concentrations of several plasma biomolecules.
© 2001 American Association for Clinical Chemistry
Fourier-transform infrared (FT-IR) spectrometry is a
global, sensitive, and highly reproducible physicochemi-
cal analytical technique that identifies structural moieties
of biomolecules on the basis of their IR absorption (1, 2 ).
Because a biomolecule is determined by its unique struc-
ture, each biomolecule will exhibit a unique FT-IR spec-
trum, representing the vibrations of its structural bonds
(3 ). Furthermore, every biomolecule present in the sample
will exhibit more or less specific FT-IR absorption peaks
(4 ). Thus, a plasma FT-IR spectrum will exhibit absorp-
tion peaks related to its major components. FT-IR analyt-
ical applications have allowed determination of blood
contents using various materials and sample prepara-
tions. Concentrations of glucose (1, 2, 5, 6 ), total proteins,
creatinine, urea, triglycerides (1 ), and cholesterol (2 ) in
blood, plasma, or serum have been determined with
clinical accuracy. However, extensive sample preparation,
manipulation of spectra, and mathematical treatments
have been necessary to obtain such results. Clinical anal-
yses require methods that involve minimal sample ma-
nipulations because every handling step may be a source
of quantitative error, affecting the predictive performance
of the method used (7, 8 ).
With FT-IR spectrometry, fluctuations in absorption
related to water content because of environmental condi-
tions (e.g., variations in air temperature, humidity, and
atmospheric pressure) have necessitated manipulation of
samples and/or spectra (or both) (1, 2, 5, 6, 8 ). Recently,
we described FT-IR analysis of glucose concentrations in
dried serum samples (9 ). By avoiding water absorption in
serum spectra, we obtained results that correlated well
with those obtained by the glucose oxidase method (10 );
sample manipulations were limited to a precise dilution
with water and desiccation under moderately reduced
pressure after deposition of 35 ␮L of this solution onto a
multiple sample holder. No other manipulations of the
sample or the spectrum were necessary for obtaining
highly reproducible results. The characteristic IR absorp-
 Clinical Chemistry 47, No. 4, 2001
tion peaks for glucose were determined before those of
biomolecules exhibiting more intense absorption contri-
butions to serum FT-IR spectra (namely, proteins); we
choose this approach because the glucose absorption
appears in a specific spectral area, the ␯COO absorption
region (1300 –900 cm⫺1), in which major protein absorp-
tion is absent.
However, this analytical method is rather limited. Only
⬃80 spectral ranges, i.e., deformations such as peaks,
shoulders, and bands, are found on a FT-IR spectrum,
whereas plasma contains thousands of biomolecules at
various concentrations. We previously had shown that
subtraction of the spectral contribution of a previously
determined metabolite allowed subsequent analysis, e.g.,
subtraction of the glucose spectrum allowed determina-
tion of lactatemia (11 ). These results for lactate were in
good agreement with an enzymatic comparison method
but could not have been obtained before this subtraction
because the glucose absorption partially obscured the
lactate absorption in the complete serum FT-IR spectra of
subjects. Glucose and lactate concentrations could be
determined because the broadest protein IR absorption
peaks (for ␯CAO and ␦NOH: 1700 – 400 cm⫺1) were
outside the spectral region in which most characteristic
absorption peaks for glucose and lactate were found
(␯COO: 1300 –900 cm⫺1). We now apply this methodol-
ogy to the analysis of plasma proteins, the major compo-
nent of plasma (accounting for ⬃75 g/L of the 100 g/L of
dried matter present in plasma).
Materials and Methods
sample collection
Venous blood samples from 35 anonymous healthy blood
donors, admitted to hospital for routine analyses, were
used for this study. Although these subjects did not
present with pathological situations, several presented
with some protein concentrations outside the physiolog-
ical range for healthy adults (Table 1). The mean (⫾ SD)
age of these subjects was 35.4 ⫾ 5.1 years (range, 27– 43
years). Blood was sampled between 0730 and 0830 after
an overnight fast of 10 h. Blood was drawn into sterile and
gel-barrier collection tubes (6-mL red-brown top and
7-mL red top Vacutainer Tubes; Becton Dickinson). One
gel-barrier collection tube was used for FT-IR measure-
ments. Venous blood was sampled by an antecubital
venipuncture of the right arm through a Teflon catheter.
Samples collected in gel-barrier collection tubes were
centrifuged immediately for 10 min at 4000g; plasma was
collected in the red-brown top tubes and stored at ⫺20 °C
before analysis.
Venous blood samples from 14 age-matched patients
(37.1 ⫾ 4.4 years; range, 30 – 44 years) admitted to hospital
for severe infection (n ⫽ 4), renal disorder (n ⫽ 4), or
anemia (n ⫽ 6) were used for validation of the method-
ology implemented for the 35 blood samples from the
relatively healthy subjects.
731
comparison methods
Glucose and urea concentrations were measured with an
accepted enzymatic method on a Dax 48 analyzer and
using calibrators and controls from Bayer. Lactate was
also measured by an enzymatic method (Microzym-L;
SGI). Protein concentrations were determined immuno-
nephelometrically with a BN II apparatus and calibrators
and controls from Dade Behring, except that apolipopro-
teins A1, B, and C3 (Apo-A1, Apo-B, Apo-C3) and IgD
were evaluated by radial immunodiffusion with reagents
from Daiichi Pure Chemicals Co. (for the apolipoproteins)
Downloaded from https://academic.oup.com/clinchem/article-abstract/47/4/730/5639132 by guest on 08 March 2020
and Dade Behring (for IgD).
acquisition of ft-ir spectra
The method for acquiring spectra has been described
elsewhere (9 ). Briefly, after samples returned to room
temperature (⬃15 min at 20 –25 °C), 20-␮L aliquots were
diluted with 80 ␮L of water. The diluted samples were
homogenized with an agitator (Vortex Reax 2000; Hei-
dolf) at 1000g for 10 s. Thirty-five microliters of each
solution was deposited exactly within the cell limits of a
zinc selenide wheel with 15 sample cells (Bruker). The
loaded wheel was subsequently placed in a low-pressure
[266 Pa (2 mmHg)] drier to evaporate the water in the
sample (45 min), after which a KBr cover was attached to
protect the wheel. The sample-bearing wheel was then
inserted into the analysis compartment of a Bruker IFS
28/B spectrometer equipped with a Globar (MIR) source
(7 V), a KBr beam splitter, and a DTGS/B detector
(18 –28 °C) from Bruker. Beam diameter at the sample
location was 6 mm. In all experiments, we used a resolu-
tion of 2.0 cm⫺1 and performed 32 scans for data acqui-
sition. All analyses were performed in triplicate on three
successive wheels.
comparison ft-ir spectra of biomolecules
FT-IR spectra obtained for pure components (98 –99%
purity; Sigma-Aldrich) were used as comparison FT-IR
spectra. These were the spectra used for all subtractions of
biomolecule absorption peaks. The concentration of the
dried matter in plasma is ⬃100 g/L. We previously
demonstrated (9, 11 ) that FT-IR analyses of plasma di-
luted fourfold with water provided a higher signal-to-
noise ratio (determined in the 2100 –2200 cm⫺1 region,
where no absorption attributable to plasma biomolecules
is found) and greater signal reproducibility for successive
analyses. Because these fourfold-diluted solutions con-
tained ⬃20 g/L dried matter, we used pure component
solutions ranging in concentration from 12 to 28 g/L to
obtain FT-IR spectra that would be comparable in absorp-
tion values to those of the plasma FT-IR spectra. Appli-
cation of pure component solutions onto the analytical
wheel, desiccation, and FT-IR spectrum acquisition were
performed exactly as described for plasma FT-IR spec-
trum acquisitions. For every biomolecule, linearity be-
tween the absorption signals detected and these concen-
trations was assessed by total area integration over 4000 –
732 Petibois et al.: Plasma Proteins Determined by FT-IR Spectrometry
500 cm⫺1, representing the whole FT-IR absorption range
for the biomolecules (with P ⬍0.01).
analyses and calculations
The method for iterative determination of biomolecule
concentrations (Fig. 1) was assessed as follows:
(1) Spectral ranges (peaks, bands, shoulders) were chosen
as approximate absorption bands within the plasma
FT-IR spectra. The precise abscissa limits for the
spectral ranges were assigned using a subroutine of
OPUS 3.0 software (Bruker). In brief, once two wave-
numbers had been selected outside the absorption
band, the software found the lowest absorption values
that allowed the drawing of a line, no other point on
which was in common with the spectrum curve.
(2) For each defined spectral range common to the 35
plasma FT-IR spectra of our experiments, the area
between the line drawn and the spectrum was inte-
grated by the software and expressed in arbitrary
units (U).
(3) A correlation matrix was used to compare all integra-
tion results with the concentrations of all biomolecule
measured with the comparison clinical analytical
methods. The spectral range that led to the greatest
correlation between integration values and the results
obtained by the comparison method for each of the
biomolecules considered was selected for subsequent
use.
(4) A series of accurately determined calibration solutions
of the selected biomolecule allowed us to acquire and
integrate the spectral area of the pure compound over
the previously determined spectral range. These re-
Fig. 1. Flow diagram of the steps in the iterative procedure for predicting concentrations of plasma compounds from plasma FT-IR spectra.
sults were expressed as the biomolecule concentration
per arbitrary unit of spectral area: g 䡠 L⫺1 䡠 U⫺1.
(5) For each FT-IR plasma spectrum, the concentration of
the selected biomolecule was then calculated using its
absorption value, based on the spectrum of the pure
component, and the spectral area integration result
obtained in step 2.
(6) The contributing spectrum of the biomolecule, deter-
mined from its pure component spectrum and its
plasma concentration (as measured with a compari-
Downloaded from https://academic.oup.com/clinchem/article-abstract/47/4/730/5639132 by guest on 08 March 2020
son method), was subsequently subtracted from the
plasma FT-IR spectrum.
(7) The resulting spectrum was then saved and subjected
to the same iterative method (steps 1 to 6) to deter-
mine the concentration of the next biomolecule to be
analyzed. This process was repeated for each of the
biomolecules being evaluated.
statistics
Data are expressed as the mean ⫾ SD. In each clinical
reference data series, the dispersion [mean dispersion
(CV), expressed as a percentage] around the mean value
was calculated. When the CV exceeded 5%, the concen-
tration values were considered significantly heteroge-
neous (P ⬍0.05). Linear regressions were performed to
compare the data series. The results obtained by the
clinical analytical method and the spectral data were
considered comparable when P was ⬍0.05. When the data
were comparable, dispersion data around the regression
lines were estimated by the mean standard error of
prediction:
 Clinical Chemistry 47, No. 4, 2001 733

S xⱍy ⫽ 冑冋 冘 共 x i ⫺ ␮兲 / n
2
册 45 U (integration range, 4000 –500 cm⫺1). Results were
where xi corresponds to the predicted analyte value, ␮ is
the value determined by the comparison method, and n is
the total number of samples in the prediction data set. For
results to be acceptable, the regression line had to show a
slope close to 1 and an intercept close to 0. Differences
between methods were tested using a standard t-test for
paired data.
comparable for the set of FT-IR spectra from the plasma
samples from 14 patients, except that the mean plasma
FT-IR spectrum area for the latter was 461 ⫾ 76 U
(integration range, 4000 –500 cm⫺1).
ft-ir spectra of biomolecules
For each series of pure component solutions, we studied
the correlations between the corresponding 4000 –500
cm⫺1 spectral area and the biomolecule concentration.
The correlations were 0.95– 0.99 for all analytes evaluated

Downloaded from https://academic.oup.com/clinchem/article-abstract/47/4/730/5639132 by guest on 08 March 2020

Results
clinical data
The plasma concentrations of glucose, lactate, urea, and
proteins for the first set of 35 venous blood samples are
presented in Table 1. Immunoglobulins, haptoglobin, ␣1-
acid glycoprotein, and Apo-C3 exhibited the broadest
distribution of concentrations (CV ⫽ 8 –10%). This heter-
ogeneity in plasma protein content allowed us to assess
the analytical potential of FT-IR spectrometry for deter-
mining concentrations of successive biomolecules.
ft-ir spectrum acquisition
For the set of 35 plasma FT-IR spectra, the mean signal-
to-noise ratio in the 2100 –2200 cm⫺1 region was 687 ⫾ 19.
The baselines between repeated samples were also homo-
geneous, varying by 0.011 ⫾ 0.006 at 4000 cm⫺1. The
between-run CV was 1.3% for triplicate FT-IR measure-
ments. The mean plasma FT-IR spectrum area was 472 ⫾
in the concentration range 12–28 g/L (Table 2). The FT-IR
spectra for the biomolecules presented important IR ab-
sorption dissimilarities, even for biomolecules belonging
to the same families, e.g., apolipoproteins (Fig. 2) and IgG
(Fig. 3).
analyses
The iterative procedure for determining the concentra-
tions of biomolecules was tested for the set of 35 venous
blood samples. Table 3 presents, in the order of analysis of
the biomolecules, details on the spectral ranges used for
integration (left and right abscissa limits), the FT-IR
absorption data for the pure components in the same
spectral ranges used to measure the concentrations of the
biomolecules in plasma, the statistical significance of any
correlation with results by the comparison methods, and
a summary of the statistical indices for linearity between
methods. After multiple analyses of the highest statistical

Table 1. Plasma concentrations of glucose, lactate, urea, Table 2. Relationships between concentrations and
and various proteins as determined by corresponding 4000 –500 cmⴚ1 total spectral areas
comparison methods. obtained from series of pure component solutions.
Concentration,a g/L Biomoleculea r Slope Intercept
Albumin 0.97 18.8 0.22
Biomolecule Mean ⴞ SD Min Max CV, %
␣1-Acid glycoprotein 0.98 19.8 ⫺0.41
Glucose 5.20 ⫾ 0.16 5.0 5.6 0.84 ␣1-Antitrypsin 0.99 22.2 ⫺0.30
Lactate 1.72 ⫾ 0.31 1.11 2.23 2.4 ␣2-Macroglobulin 0.99 19.9 0.28
Urea 6.07 ⫾ 1.26 4.13 9.33 7.7 Apo-A1 0.96 19.5 0.17
Albumin 40.9 ⫾ 3.8 36.8 50.5 5.6 Apo-B 0.95 24.7 ⫺0.02
Fibrinogen 3.37 ⫾ 0.43 2.34 4.15 2.4 Apo-C3 0.98 20.4 0.11
Transferrin 2.44 ⫾ 0.30 1.8 3.0 7.4 Fibrinogen 0.99 22.1 0.21
Haptoglobin 0.94 ⫾ 0.45 0.16 2.02 20 Glucose 0.97 28.5 0.19
␣1-Acid glycoprotein 0.79 ⫾ 0.22 0.48 1.47 13 Haptoglobin 0.95 18.7 ⫺0.22
␣1-Antitrypsin 2.89 ⫾ 0.63 1.7 4.0 6.7 IgA 0.96 21.5 ⫺0.37
␣2-Macroglobulin 2.05 ⫾ 0.50 1.1 3.3 8.4 IgD 0.99 18.5 ⫺0.39
IgA 1.88 ⫾ 0.69 0.26 2.99 11 IgG1 0.97 20.0 ⫺0.13
IgD 0.37 ⫾ 0.13 0.18 0.71 12 IgG2 0.99 18.5 ⫺0.10
IgG1 7.51 ⫾ 1.91 4.1 13.2 8.7 IgG3 0.96 21.4 0.28
IgG2 2.53 ⫾ 0.54 1.1 3.4 9.6 IgG4 0.96 24.4 0.31
IgG3 0.61 ⫾ 0.21 0.2 1.2 12 IgM 0.97 19.7 ⫺0.12
IgG4 0.53 ⫾ 0.25 0.1 1.2 21 Lactate 0.94 29.4 ⫺0.44
IgM 1.20 ⫾ 0.66 0.20 2.69 14 Transferrin 0.96 21.9 0.19
Apo-A1 1.52 ⫾ 0.21 1.2 2.1 6.8 Urea 0.99 30.5 ⫺0.23
Apo-B 0.93 ⫾ 0.17 0.55 1.27 4.7 a
n ⫽ 15 analyses per biomolecule concentration. For each biomolecule, the
Apo-C3 0.48 ⫾ 0.17 0.17 0.81 9.1 mean of the spectra for the 20 g/L solution was used as the comparison
a
Concentrations for glucose, lactate, and urea given as mmol/L. spectrum for spectra subtractions.
734 Petibois et al.: Plasma Proteins Determined by FT-IR Spectrometry
Fig. 2. FT-IR spectra (1850 –500 cm⫺1) of IgG1–IgG4 in 20 g/L
solutions.
Several absorption peaks differ among these biomolecules: 802 cm⫺1 for IgG1,
1087 and 982 cm⫺1 for IgG2, 1185 cm⫺1 for IgG3, and 893 cm⫺1 for IgG4.
values found in the correlation matrices, only one se-
quence of biomolecule analysis was found to give results
that provided sufficient agreement with the 35 results
obtained with the clinically used comparison methods.
The closest relationship between integration results
and the values obtained by the clinical comparison meth-
ods was for albumin, the biomolecule present in the
highest concentration in plasma. This relationship in-
volved the spectral area located between 1600 and 1488
cm⫺1 (␦NOH region; 23.3 ⫾ 0.3 U vs 40.9 ⫾ 3.8 g/L; r ⫽
0.99; P ⬍0.001). Albumin represents 65% of plasma pro-
teins and 45% of the total plasma mass. The FT-IR
absorption of albumin between 1600 and 1488 cm⫺1 was
1.76 g 䡠 L⫺1 䡠 U⫺1. The albumin concentration measured by
FT-IR spectrometry was 41.0 ⫾ 3.9 g/L (correlation with
the comparison method, r ⫽ 0.99; P ⬍0.001). Regression
Fig. 3. FT-IR spectra (1850 –500 cm⫺1) of Apo-A1, -B, and -C3 in 20 g/L
solutions.
Several absorption peaks are highly specific for these biomolecules: 1466 and
1164 cm⫺1 for Apo-A1, 1218 and 984 cm⫺1 for Apo-B, and 997 cm⫺1 for Apo-C3.
analysis of the two methods showed a slope of 1.00 with
an intercept of 0.17 g/L (Fig. 4). The spectral contribution
of albumin (4000 –500 cm⫺1) represented 39.9% of the total
plasma FT-IR spectral area. Subtracting the albumin ab-
sorption left a spectral area of 230 ⫾ 21 U (Fig. 5).
Evaluation of the remaining area showed that glucose
was the second biomolecule whose concentration closely
correlated with absorption area integration values (5.2 ⫾
0.4 mmol/L; ␯ COOOC absorption at 1033–997 cm⫺1 ⫽ 7.27
mmol 䡠 L⫺1 䡠 U⫺1; plasma spectral area at 1033–997 cm⫺1
⫽ 0.71 ⫾ 0.05 U). The glucose concentration measured by
Downloaded from https://academic.oup.com/clinchem/article-abstract/47/4/730/5639132 by guest on 08 March 2020
FT-IR spectrometry was 5.22 ⫾ 0.17 mmol/L (correlation
with the comparison method, r ⫽ 0.97; P ⬍0.001; regres-
sion slope ⫽ 0.99; intercept ⫽ 0.04 mmol/L). Regardless
of whether the albumin absorption was subtracted before
or after glucose concentration analysis in the resulting
spectra, the results for glucose were in accord with those
from our previous study (9 ).
For subsequent spectra obtained after the spectral
contributions of specific biomolecules were subtracted,
the greatest correlations were found between the concen-
trations measured by the comparison method and the
results obtained by spectral integration in the following
descending order: fibrinogen, IgG2, lactate, IgG1, ␣1-anti-
trypsin, ␣2-macroglobulin, transferrin, Apo-A1, urea,
Apo-B, IgM, Apo-C3, IgA, IgG4, IgG3, IgD, haptoglobin,
and ␣1-acid glycoprotein (Tables 2 and 3). Urea was
included in this analysis because it exhibits strong absorp-
tion bands between 1800 and 1100 cm⫺1, a spectral range
in which IgM, IgG3, and IgG4 are also examined. In the
1500 –1300 cm⫺1 region, several of the spectral ranges
used for determining biomolecule concentrations were
very similar, i.e., those for fibrinogen, haptoglobin, IgG1,
IgA, and IgM. Utilization of these ranges was possible,
however, because the absorption bands for the pure
compounds were not found in exactly the same locations
(␭max and ⑀) on the spectrum for each of the other pure
components (Fig. 6).
Haptoglobin was one the last biomolecules analyzed in
our sequence and presented important concentration vari-
ations among subjects. Nevertheless, although fibrinogen,
IgG1, IgA, and IgM concentrations were determined be-
fore those of haptoglobin, we obtained very good rela-
tionships between spectral and clinical reference values
for these five biomolecules. Finally, after subtraction for
20 biomolecules, which corresponded to 69 ⫾ 4 g/L of the
dried matter in plasma, the mean total spectral area of the
remaining spectrum was 114.5 ⫾ 11.2 U, i.e., 24.2% of the
initial plasma FT-IR spectral area (Fig. 7).
validation with plasma samples from 14 patients
The results obtained for plasma samples from the 14
patients are presented in Table 4. Patients presented with
broad concentration heterogeneities for albumin, IgA,
IgD, IgG1, IgG2, IgG3, IgG4, IgM, urea, Apo-B, Apo-C3,
and haptoglobin. By comparing these differences in
 Clinical Chemistry 47, No. 4, 2001 735

Table 3. Statistical indices between comparison method values and FT-IR spectrometry values.
FT-IR spectrometry data Statistics

Abscissa
(wavenumber)
limits, cmⴚ1 Mean ⴞ SD

Absorptivity, Concentration,
Biomoleculea Left Right g 䡠 Lⴚ1 䡠 Uⴚ1 Area, U g/L rb Sxⱍy Slope Intercept
Albumin 1600 1488 1.76 23.3 ⫾ 0.3 41.0 ⫾ 3.9 0.98 0.67 1.00 0.67
Glucose 1062 997 7.27c 0.71 ⫾ 0.05 5.22 ⫾ 0.17d 0.95 0.05c 0.99 0.04d
Fibrinogen 1428 1363 2.42 1.38 ⫾ 0.17 3.36 ⫾ 0.46 0.97 0.10 1.04 ⫺0.16
0.74 ⫾ 0.11 2.57 ⫾ 0.57

Downloaded from https://academic.oup.com/clinchem/article-abstract/47/4/730/5639132 by guest on 08 March 2020

IgG2 1652 1622 3.72 0.97 0.18 1.01 0.02
Lactate 1134 1115 65.46c 0.026 ⫾ 0.007 1.80 ⫾ 0.36d 0.94 0.11c 1.09 ⫺0.08d
IgG1 1419 1361 9.07 0.83 ⫾ 0.23 7.66 ⫾ 2.08 0.98 0.35 1.07 ⫺0.35
␣1-Antitrypsin 712 691 48.4 0.059 ⫾ 0.015 2.89 ⫾ 0.68 0.96 0.18 1.04 ⫺0.14
␣2-Macroglobulin 1116 983 1.98 1.05 ⫾ 0.26 2.11 ⫾ 0.51 0.98 0.22 1.01 0.03
Transferrin 1338 1292 12.39 0.19 ⫾ 0.02 2.45 ⫾ 0.31 0.97 0.08 0.99 0.02
Apo-A1 1484 1427 1.81 0.91 ⫾ 0.11 1.55 ⫾ 0.22 0.97 0.05 1.06 ⫺0.06
Urea 1652 1632 35.01c 0.17 ⫾ 0.04 6.25 ⫾ 1.39d 0.97 0.33c 1.07 ⫺0.25d
Apo-B 2862 2835 6.95 0.13 ⫾ 0.01 0.96 ⫾ 0.17 0.96 0.05 0.96 0.07
IgM 1428 1360 1.96 0.61 ⫾ 0.25 1.27 ⫾ 0.65 0.98 0.13 0.97 0.11
Apo-C3 1676 1660 8.09 0.062 ⫾ 0.028 0.52 ⫾ 0.19 0.97 0.04 1.06 0.004
IgA 1418 1372 5.31 0.36 ⫾ 0.11 1.88 ⫾ 0.70 0.97 0.16 0.99 ⫺0.003
IgG4 1538 1505 6.89 0.075 ⫾ 0.031 0.54 ⫾ 0.26 0.95 0.08 0.93 0.08
IgG3 1652 1628 4.29 0.14 ⫾ 0.05 0.63 ⫾ 0.23 0.95 0.07 1.03 0.01
IgD 1479 1464 21.8 0.019 ⫾ 0.006 0.38 ⫾ 0.16 0.92 0.05 1.11 ⫺0.03
Haptoglobin 1428 1378 3.35 0.29 ⫾ 0.14 0.98 ⫾ 0.47 0.99 0.08 1.04 0.01
␣1-Acid glycoprotein 1337 1279 2.77 0.33 ⫾ 0.08 0.82 ⫾ 0.23 0.96 0.07 0.99 0.02
a
Data are presented in order of analysis of the biomolecules in the iterative method.
b
P ⬍0.001 if r ⬎0.99; P ⬍0.005 if r ⬎0.95.
c
mmol 䡠 L⫺1 䡠 U⫺1.
d
mmol/L.

plasma contents with the results obtained for the first set
of plasma samples, we could test the validity of the
sequence used for biomolecule analysis. For this set of 14
plasma samples, we used exactly the same spectral ranges
and sequence order that were used for the previous
analyses of 35 plasma samples. As shown in Table 4, all
concentration values obtained correlated well with the
concentrations determined by the comparison methods.

Fig. 5. Subtraction of IgG1 absorption spectrum (for 7.66 g/L IgG1

Fig. 4. Relationship between albumin concentration measured by the sample) from plasma FT-IR spectrum.
comparison method and by FT-IR spectrometry in plasma samples from The mean spectra (1500 –500 cm⫺1) for samples from 35 healthy subjects
35 healthy subjects. before (plasma) and after (resulting) subtraction are shown.
736 Petibois et al.: Plasma Proteins Determined by FT-IR Spectrometry
Fig. 6. FT-IR spectra of 20 g/L fibrinogen, haptoglobin, IgG1, IgA, and
IgM in the 1500 –1300 cm⫺1 spectral region.
In each spectrum, vertical bars give abscissa limits and top of spectral ranges
used for determination of biomolecule concentrations. The upper limits of the
spectral ranges were 1398 cm⫺1 for fibrinogen, 1401 cm⫺1 for haptoglobin,
1405 cm⫺1 for IgG1, 1411 cm⫺1 for IgM, and 1408 cm⫺1 for IgA (see Table 3
for abscissa limits).
Discussion
We previously demonstrated that FT-IR spectrometry
could determine glucose and lactate concentrations with
clinical accuracy in the 1300 –900 cm⫺1 absorption region
(9, 11 ). These measurements were facilitated by the ab-
sence of marked absorption by proteins, mainly albumin,
within this spectral region. This opportunity has to be
related to the absence of glycosylation of albumin, this
protein being by far the most abundant compound in
plasma. However, until now, no more than a few charac-
teristic absorption peaks have been used successfully for
measuring the individual concentrations of several bi-
omolecules in a sample. Only statistical methods, such as
multivariate regression, allowed determination of concen-
trations of other biomolecules in plasma, using FT-IR
Fig. 7. Plasma FT-IR spectrum (4000 –500 cm⫺1) after subtraction of
absorption spectra of 20 biomolecules.
The FT-IR spectrum is the mean for samples from 35 healthy subjects.
spectra. These methods looked for a strong relationship
between a wide spectral region (500 cm⫺1 or higher) that
contains several absorption bands for a biomolecule and
the concentration of the molecule in the analyzed sample
(1, 2 ). The aim of multivariate regression methods is to
correlate variations in blood FT-IR absorption spectra
with the known changes in the absorption spectrum of a
pure component when the concentration of the pure
component varies within samples. Such methods have
been used to determine the plasma concentrations of total
proteins (1 ) and cholesterol (2 ). However, a strong limi-
Downloaded from https://academic.oup.com/clinchem/article-abstract/47/4/730/5639132 by guest on 08 March 2020
tation of these methods is that a direct analysis, i.e., one
showing any characteristic absorption of the biomolecules
in the spectrum, has been unavailable. Indeed, defined
concentrations of biomolecules could not be determined
as had been done for glucose or lactate. In the FT-IR
spectra of complex biological samples, such as plasma, the
absorption spectra of some biomolecules may first be
subtracted to uncover other absorption patterns. This has
been clearly shown for the absorption pattern of lactate in
the 1300 –900 cm⫺1 spectral region, which is obscured by
glucose absorption (11 ). Lactate absorption in this spectral
region could be exploited only after the absorption attrib-
utable to glucose had been subtracted.
Our aim in this study was to determine the concentra-
tions of various proteins in plasma on the basis of their
most characteristic IR absorption peaks. For albumin, the
best correlation with results obtained by a comparison
method was found using the ␦NOH absorption region
(1600 –1480 cm⫺1) common to all plasma proteins. How-
ever, the physiological concentration of albumin is 37– 45
g/L, and the biomolecule that contributes the most in-
tense absorption to plasma FT-IR spectra after albumin is
IgG1, the concentration of which usually is 6 – 8 g/L. We
found a high correlation between IgG1 concentrations
measured by the comparison method and spectral data
for the 1419 –1361 cm⫺1 absorption region, which is out-
side of the spectral region used for determining the
albumin concentration.
We determined an absorption-related sequence for the
biomolecules to be subtracted from plasma FT-IR spectra;
that is, the concentrations of several biomolecules could
be determined only after the absorption spectra of other
biomolecules had been subtracted, even when the absorp-
tion contributed to the plasma FT-IR spectrum by the
molecules subtracted earlier was several-fold less intense
than the absorption contributed by the remaining biomol-
ecules. This was clearly illustrated by the plasma concen-
trations of the biomolecules determined in the following
order: glucose (⬃1 g/L), fibrinogen (⬃3 g/L), lactate
(⬃0.3 g/L), and IgG1 (⬃8 g/L). Despite its high concen-
tration, IgG1 could not be determined before the glucose,
fibrinogen, and lactate IR absorption peaks were sub-
tracted. Glucose absorbs strongly in the 1419 –1361 cm⫺1
spectral region, which is also the most specific region for
IgG1 absorption in plasma. Similarly, fibrinogen exhibits a
strong absorption centered at 1395 cm⫺1, and lactate
 Clinical Chemistry 47, No. 4, 2001 737

Table 4. Results obtained for the plasma samples from 14 patients.

Concentration, g/L

Comparison method FT-IR spectrometry data Statistics

Biomoleculea Mean ⴞ SD Min Max Mean ⴞ SD Min Max rb Sxⱍy

Albumin 43.1 ⫾ 8.5 32 61 42.8 ⫾ 8.6 31.4 61.3 0.98 1.66
Glucose 5.06 ⫾ 1.26c 3.6c 8.4c 4.93 ⫾ 1.21c 3.18c 7.99c 0.98 0.22c
Fibrinogen 2.61 ⫾ 0.84 1.4 4.2 2.59 ⫾ 0.80 1.51 4.01 0.97 0.19
IgG2 2.91 ⫾ 2.31 0.5 8.0 2.82 ⫾ 2.27 0.47 7.46 1.00 0.21
Lactate 1.24 ⫾ 0.60c 0.5c 2.8c 1.18 ⫾ 0.57c 0.45c 2.61c 0.97 0.15c
IgG1 7.16 ⫾ 5.19 1.8 22.0 7.40 ⫾ 5.60 1.61 23.6 0.96 0.56

Downloaded from https://academic.oup.com/clinchem/article-abstract/47/4/730/5639132 by guest on 08 March 2020

␣1-Antitrypsin 1.69 ⫾ 0.37 1.09 2.90 1.67 ⫾ 0.57 0.89 2.61 0.94 0.20
␣2-Macroglobulin 1.79 ⫾ 0.59 0.83 2.94 1.79 ⫾ 0.63 0.86 3.05 0.98 0.12
Transferrin 1.87 ⫾ 0.49 0.81 2.58 1.91 ⫾ 0.51 0.81 2.67 0.94 0.14
Apo-A1 1.17 ⫾ 0.49 0.3 2.18 1.17 ⫾ 0.52 0.24 2.33 0.99 0.08
Urea 11.7 ⫾ 7.9c 1.7c 25.5c 11.7 ⫾ 7.8c 1.54c 25.4c 0.97 0.78c
Apo-B 0.90 ⫾ 0.44 0.19 1.87 0.90 ⫾ 0.43 0.22 1.71 0.96 0.13
IgM 2.73 ⫾ 3.10 0.31 10.40 2.64 ⫾ 3.16 0.33 11.1 0.93 0.35
Apo-C3 0.16 ⫾ 0.10 0.03 0.33 0.16 ⫾ 0.10 0.02 0.37 0.96 0.03
IgA 3.41 ⫾ 1.95 0.69 6.06 3.23 ⫾ 1.90 0.56 6.04 0.94 0.23
IgG4 0.24 ⫾ 0.23 0.01 0.80 0.23 ⫾ 0.23 0.01 0.86 0.97 0.05
IgG3 1.07 ⫾ 0.60 0.22 2.00 1.08 ⫾ 0.65 0.3 2.21 0.95 0.13
IgD 0.67 ⫾ 0.37 0.13 1.11 0.69 ⫾ 0.42 0.11 1.23 0.94 0.15
Haptoglobin 1.79 ⫾ 1.26 0.49 4.35 1.73 ⫾ 0.30 0.45 4.89 0.97 0.23
␣1-Acid glycoprotein 1.48 ⫾ 0.48 0.80 2.31 1.44 ⫾ 0.49 0.61 2.19 0.95 0.16
a
Data are presented in order of analysis of the biomolecules in the iterative method.
b
P ⬍0.01 if r ⬎0.98; P ⬍0.05 if r ⬎0.92.
c
mmol/L.

exhibits a weak absorption centered at 1399 cm⫺1. Fur-
thermore, we were unable to find another specific IgG1
absorption peak in any other spectral region. Indeed,
before we could ascertain which absorption region could
be used for determining the IgG1 concentration in plasma,
the glucose, fibrinogen, and lactate absorption spectra had
to be subtracted. The sequence we found for determining
biomolecule concentrations was the one yielding the best
correlation with the spectral range areas calculated with
respect to the resulting spectra for the other biomolecules.
Multivariate calibration methods, such as partial least
squares and principal component regression, effectively
remove the spectral features of each determined analyte
sequentially, similar to the method we have proposed.
However, by subtracting the spectral contribution of each
biomolecule determined from the complete FT-IR spec-
trum for plasma, we were able to determine the concen-
trations of many more biomolecules than could be mea-
sured by other FT-IR spectrometry methods.
Importantly, the FT-IR spectra of biomolecules belong-
ing to the same families contained important differences,
as was observed for the IgGs. This reflects structural
specificities for these biomolecules; e.g., between 1300 and
900 cm⫺1 (␯COO region), spectral differences reflect dif-
ferences in the sugar content. Thus, we were able to
determine the concentration of each IgG form in plasma
according to the FT-IR spectrum. This was also true for
apolipoproteins, for which three different spectral regions
were used, representing the different FT-IR absorption
spectra of these biomolecules.
Another important finding of this study is that the
concentrations of several biomolecules, notably those for
fibrinogen, haptoglobin, IgG1, IgA, and IgM, could be
determined using spectral ranges that were very close to
one another, i.e., 1500 –1300 cm⫺1, but that differed as to
the exact locations and intensities of the absorption peaks.
Even when haptoglobin concentrations exhibited several-
fold variations between subjects, we were able to deter-
mine the concentrations of the other biomolecules with
clinical accuracy in the previous steps. Indeed, the spec-
tral ranges we used exhibited slight but sufficiently char-
acteristic differences in the 1500 –1300 cm⫺1 absorption
region, leading to determination of the concentrations of
these biomolecules. For the last biomolecule determined
in this series, ␣1-acid glycoprotein, the results correlated
well (r ⫽ 0.96) with the values obtained by the compari-
son method.
We tested the sequence order we found for determin-
ing biomolecule concentrations by assaying plasma sam-
ples from 14 patients in which the concentrations of
several proteins varied widely. We also assayed these
samples with the comparison methods for all biomol-
ecules determined by the FT-IR method. The results
demonstrated that the spectral ranges we used to deter-
mine biomolecule concentrations reflected structural ab-
738 Petibois et al.: Plasma Proteins Determined by FT-IR Spectrometry
sorption peaks sufficiently characteristic of these biomol-
ecules.
Moreover, the FT-IR absorption spectra of the biomol-
ecules analyzed on the basis of sequential plasma spectra
were not altered by successive subtractions of absorption.
After the individual absorption peaks for 20 biomolecules
were subtracted from the complete FT-IR spectrum of
plasma, the resulting spectra were not noise. We therefore
consider it possible to continue to use this method to
determine the concentrations of additional biomolecules,
namely, triglycerides, cholesterol esters, amino acids, and
fatty acids.
In conclusion, the present study has demonstrated that
FT-IR spectrometry is a useful tool for determining con-
centrations of multiple biomolecules in microsamples of
plasma.
We are indebted to the Conseil Régional d’Aquitaine and
the Fédération Française de Rubgy for financial support
and technical assistance.
References
1. Shaw RA, Kotowich S, Leroux M, Mantsch HH. Multivariate serum
analysis using mid-infrared spectroscopy. Ann Clin Biochem 1998;
35:624 –32.
2. Budinova G, Salva J, Volka K. Application of molecular spectros-
copy in the mid-infrared region to the determination of glucose and
cholesterol in whole blood and in blood serum. Appl Spectrosc
1997;51:631–5.
3. Gremlich HU. Infrared and Raman spectroscopy. In: Ullmann’s
encyclopedia of industrial chemistry, Vol. B5. Weinheim, Switzer-
land: VCH, 1994:429 – 69.
4. Sockalingum GD, Bouhedja W, Pina P, Allouch P, Bloy C, Manfait
M. FT-IR spectroscopy as an emerging method for rapid character-
ization of microorganisms. Cell Mol Biol 1998;44:261–9.
5. Ward KJ, Haaland DM, Robinson MR, Eaton RP. Quantitative
infrared spectroscopy of glucose in blood using partial least-
square analyses. Proc SPIE Int Soc Opt Eng 1989;1145:607– 8.
Downloaded from https://academic.oup.com/clinchem/article-abstract/47/4/730/5639132 by guest on 08 March 2020
6. Werner GH, Böcker D, Haar H-P, Kuhr H-J, Mischler R. Multicom-
ponent assay for blood substrate in human sera and haemolysed
blood by mid-infrared spectroscopy; in Infrared spectroscopy: new
tools in medicine. Proc SPIE Int Soc Opt Eng 1998;3257:91–100.
7. Ward KJ, Haaland DM, Robinson MR, Eaton RP. Post-prandial
blood glucose determination by quantitative mid-infrared spectros-
copy. Appl Spectrosc 1992;46:959 – 65.
8. Heise HM, Bittner A. Investigation of experimental errors in the
quantitative analysis of glucose in human blood plasma by ATR-IR
spectroscopy. J Mol Struct 1995;348:21– 4.
9. Petibois C, Rigalleau V, Melin A-M, Perromat A, Cazorla G, Gin H,
Déléris G. Determination of glucose in dried serum samples by
Fourier-transform infrared spectroscopy. Clin Chem 1999;45:
1530 –5.
10. Bergmeyer HU. Methods of enzymatic analysis. New York: Aca-
demic Press, 1974:496pp.
11. Petibois C, Melin A-M, Perromat A, Cazorla G, Déléris G. Glucose
and lactate concentrations determination on single microsamples
by Fourier-transform infrared spectroscopy. J Lab Clin Med 2000;
135:210 –5.