ANALYTICAL

BIOCHEMISTRY

207, 150-156 (1992)

Application of Fourier Transform Infrared to Studies of Aqueous Protein Solutions G. Zuber,*

S. J. Prestrelski,t

and K. BenedekT*l

*Smith Kline and Beecham, King of Prussia, Pennsylvania tdmgen, Inc., Thousand Oaks, California 91320-l 789

Received

May

10406-0939:

and

15, 1992

Modern protein Fourier transform infrared (FT-IR) spectroscopy has proven to be a versatile and sensitive technique, applicable to many aspects of protein characterization. The major practical drawback for the FT-IR spectroscopy of proteins is the large absorbance band of water, which overlaps the amide I resonances. D,O is often substituted for II,0 in infrared experiments. Removal of water from protein samples can be complicated and tedious and potentially lead to denaturation, aggregation, or sample loss. Solvent removal by dialysis is difficult for suspensions and ~01s. A new method called the D20 dilution technique (Ddt) is described which simplifies the sample preparation step and improves the solvent subtraction. The effect of the D,O concentration on the IR spectrum of aqueous solutions of several model proteins was studied. Dilution of aqueous samples with D,O yields good quality spectra. The Ddt has been evaluated for quantitative analysis using standard proteins and its applicability to solutions and suspensions of a genetically engineered malaria antigen is demonstrated. Use of resolution-enhancement techniques with spectra in mixed solvents has also been investigated. o 1992 Academic PW+S, IDC.

Proteins have been studied by infrared (IR)’ techniques as lyophilized powders, solutions, and adsorbed films (1). Assignments of the characteristic infrared bands are well established (2-5). Conventional IR has been used to measure protein adsorption on surfaces (6). However, the use of infrared spectroscopy for detailed studies on protein conformation and band assign-

1 To whom correspondence should be addressed. * Abbreviations used: IR, infrared, FT, Fourier transform; terium dilution technique; PBS, phosphate-buffered saline; man serum albumin; RSD, relative standard deviation.

150

Spectroscopy

Ddt, deuHSA, hu-

ments to characteristic polypeptide structures has been complicated by the high concentration required for conventional IR. The use of signal-averaged Fourier transform IR has greatly extended the utility of this method. FT-IR spectroscopy combined with algorithms designed for band deconvolution, differential, and derivative spectroscopy have shown excellent potential for secondary structure assessment of proteins (7) and are applicable to a wide variety of structural analyses for polypeptides and proteins (8) as well as for studies of the adsorption of proteins on surfaces of biomedical interest (9-11). Despite its high sensitivity and applicability, the interference of water in IR is still a serious problem in protein analysis. IR spectroscopy has a relatively low sensitivity to the amide I vibrational mode (C = 0 stretching, ca. 1650 cm-‘), which is typically examined in conformational studies of polypeptides and proteins. The pronounced water absorbance (HOH bending, ca. 1640 cm-‘) impedes the applicability of IR spectroscopy for the conformational characterization of proteins in their natural aqueous environment. In order to eliminate the problems associated with H,O solutions, studies are typically performed in D,O. Aqueous protein samples have to go through tedious and sometimes complicated and sensitive processes (e.g., lyophilization, drying, solvent exchange). Water removal by these techniques is variable and the treatments themselves can sometimes cause denaturation, aggregation, and recovery loss. Since computer-aided data handling is routinely available on modern instrumentation, solvent subtraction has become a popular approach for the mathematical elimination of the spectral interference due to water. In this case, the spectrum of the solvent is mathematically subtracted from the spectrum of the sample. Water has strong IR absorbance and thus, short path lengths (10 pm or less) are required. Because of the inherent sensi0003-2697192

$5.00

Copyright 0 1992 by Academic Press, Inc. All rights of reproduction in any form reserved.

SPECTROSCOPIC FT - IR H20/D20 lOO%,

75%,

50%,

ANALYSIS

OF

Mixtures; 25%,

AQUEOUS

PROTEIN

SOLUTIONS

151

Myoglobin

0% D20

0%

I 1600

I 1690

I 1560

I 1470

I 1250

1360

Wavenumbers

FIG. 1. Infrared buffers. The spectra

spectra of various mixtures of H,O and were collected at constant path length.

D,O

problems, a high concentration of protein, usually 510% (50-100 mg/ml) is required for FT-IR. At such concentrations the solution is 3onideal” from a thermodynamic standpoint and protein-protein interactions can lead to artifacts. From a practical point of view, solubility problems, foaming, aggregation, and precipitation can occur, and the required sample size is not always available. tivity

TABLE

1

Fractional Composition of H,O, HOD, and D,O in various mixtures %D,O

I-W

HOD

D,O

25 50 75

0.5625 0.25 0.0625

0.375 0.50 0.375

0.0625 0.25 0.5625

FIG. 2. (A) Infrared spectra of myoglobin in various mixtures of H,O and D,O buffers. The spectra of the protein solutions are overlaid upon the spectra of the buffers. (B) Buffer-subtracted spectra of myoglobin.

In this report, we introduce a simple sample preparation procedure, the deuterium dilution technique (Ddt), which allows one to obtain protein spectra at relatively low concentration with minimal sample preparation. The detection limit, linearity, and reproducibility of the deuterium dilution technique were examined using myoglobin, Lu-chymotrypsin, lysozyme, and human serum albumin; the feasibility and applicability of the tech-

152

ZUBER. HSA A

100%

PRESTRELSKI,

AND

BENEDEK

a - Chymotrypsin

I320

k!k B

100%

D20

i

L!LL

I

I

I

1720

1680

l&l0

I

1600

WAVENUMBER

FIG. 4. Second derivative amide I region. (A) 100% H,O 100% D,O.

infrared spectra of lysozyme buffer, (B) 50% H,0/50% D,O,

in the and (C)

SK&F 105154. The malaria vaccine was prepared by mixing the antigen with aluminum hydroxide gels. Sample Preparation

FIG. 3. Buffer-subtracted and (B) a-chymotrypsin

spectra of (A) human in various mixtures of H,O

serum albumin and D,O buffers.

Aqueous and deuterated PBS solutions were prepared by dissolving the premixed salts with H,O or D,O. Stock solutions of myoglobin, a-chymotrypsin, human serum albumin, and lysozyme were prepared from these. Mixtures with the proper D,O/H,O ratio were prepared by mixing the D,O- and H,O-based stock solutions as required. The prepared protein stock solutions (5 mg/ml) were equilibrated overnight at 4°C prior to FT-IR spectroscopy. Corresponding blank mixtures

nique will be illustrated by applying the method to malaria antigen solution and vaccine. EXPERIMENTAL

TABLE

I Component Peak Frequencies (in cm-‘) for Lysozyme Solutions as a Function of D,O Fraction

Amide

METHODS

Materials The following proteins were used as received from Sigma Chemical Co. (St. Louis, MO): myoglobin, a-chymotrypsin, human serum albumin, and lysozyme. Deuterium oxide was purchased from Aldrich (Milwaukee, WI). Phosphate-buffered saline (PBS) was also purchased from Sigma. The malaria antigen used was

2

%D,O

Frequencies

0 25 50

1642 1641 1641

1657 1656 1654

75

1638

1655

100

1638

1653

D This

band

is only

observed

(1667)”

1665 as a shoulder.

1675 1675 1674

1691 1689

1675

1684 1682

1674

-

SPECTROSCOPIC

ANALYSIS

y.L = 9.9430e-3 y.H = 1.2556e-2

+ 6.5571~.3x + 5.13198-3x

R”2 = 0.977 R”2 = 0.969

y,C = 3.8264e-3

+ 1.6202e-3x

R”2

OF

AQUEOUS

PROTEIN

bands (1456 cm-‘) of the sample and background spectra were aligned; (II) no negative absorbances are allowed in the spectra after subtraction; (III) a straight baseline outside of the range of the amide regions, ~1700 cm-‘; (IV) no major peak at the location of the HOD absorbance (1456 cm-‘).

= 0.992 /

RESULTS

AND

Characterization

10

20

30

CONCENTRATION

FIG. 5. Standard protein solutions.

curves

and linear

40

50

(mg/mL)

regression

analysis

for aqueous

were prepared by mixing the H,O and D,O stock buffer solutions accordingly. The malaria vaccine samples were prepared by combining equal volumes of antigen or vaccine with D,O. Infrared

Spectroscopy

Infrared spectra were acquired using a Nicolet 6000 spectrometer (Nicolet Analytical Instruments, Madison, WI), equipped with a liquid nitrogen-cooled mercury-cadmium-telluride detector or a Nicolet 800 spectrometer equipped with a DTGS detector. The spectra were obtained in a cell with CaF, windows and 0.06- to 0.025mm polyethylene spacers. The spectra were measured at resolution of 2 cm-‘; the number of interferogram scans signal averaged was such that the signalnoise ratio was sufficiently high for the reproducible application of resolution enhancement techniques. Subtraction

153

SOLUTIONS

Criteria

The solvent subtraction method for H,O/D,O mixtures is based on multiple criteria which were identified to minimize the error of the procedure: (I) the HOD

DISCUSSION

of the H-O-D

Peak

The basis of the Ddt is the rapid exchange of hydrogen and deuterium in the solvent to generate a new HOD band. This serves as an “internal standard” for the precise alignment of the solvent and the protein spectra. Due to the importance of the HOD band, we carefully investigated some of its relevant characteristics, such as location, intensity, and shape. The spectrum of H,O, D,O, and different mixtures of H,O and D,O are displayed in Fig. 1. The large water band at 1651 cm-’ exceeds the linearity of most infrared detectors and the intensity is outside of the useful portion of the Lambert-Beer region. As the volume fraction of D,O increases, the intensity of the water absorbance decreases. The HOD bending mode is located at 1456.1 +- 0.9 cm-‘, outside the amide I and II frequencies, but sufficiently close to use as an alignment point for background subtraction. The peak intensity at the position of the HOD band does not change (0.98 +- 0.01) in the D,O concentration range of our study. This most likely is due to the overlap of the H,O and D,O modes with the HOD band in such a manner that the peak intensity of the HOD band remains roughly equal despite the difference in HOD concentrations (see Table 1). The H-O-D band is sharp and symmetrical (96 * 2% Gaussian). The D,O band at 1552 cm-’ is not (Fig. 1). A negligible overlap of the HOD band with the amide I and II bands is observed. It is also observed in Fig. 1 that the intensity of the H,O band at 1640 cm-’ decreases disproportionately to the fraction of D,O added. At a 50:50 concentration ratio the H,O band is -25% of the intensity of the pure H,O spectrum. This results from the formation of HOD. Thus, all other factors being equal, a 2~ dilution of an H,O sample will allow a 4~ increase in the maximum usable path length. This corresponds to a net sensitivity enhancement of a factor of two. Table 1 shows the expected fractions of H,O, HOD, and D,O in various H,O/D,O mixtures. Effect of Deuterium Spectra

Content

of the Solvent

on Protein

Figure 2A compares the spectra of 0.5% (5 mg/ml) myoglobin solutions in different volume fractions of D,O with the spectra of the corresponding solvent blanks. The solvent subtracted spectra are displayed in

154

ZUBER,

PRESTRELSKI, TABLE

Statistics

Protein

Note.

Concentration kng/ml)

AND

BENEDEK

3

of Protein FT-IR Data

Wavenumber (cm-‘)

RSD

Amide

I area

RSD

HSA

50.0 20.0 10.0 5.0 2.5 1.3

1651.93 1652.03 1651.40 1651.60 1650.67 1649.80

f + + f + +

0.12 0.06 0.17 0.26 0.06 0.44

0.01 0.00 0.01 0.02 0.00 0.03

0.2617 0.1310 0.0733 0.0363 0.0187 0.0098

f * * k + +

0.0096 0.0111 0.0021 0.0031 0.0015 0.0013

3.67 8.50 2.84 8.41 8.18 13.38

LYS

50.0 20.0 10.0 5.0 2.5 1.3

1651.33 1651.47 1651.67 1652.20 1651.60 1651.63

+ + k k -t f

0.06 0.21 0.15 0.10 0.61 0.21

0.00 0.01 0.01 0.01 0.04 0.01

0.3250 0.1783 0.0710 0.0360 0.0190 0.0123

Y!C0.0095 f 0.0035 f 0.0017 f 0.0046 + 0.0010 f 0.0021

2.94 1.97 2.44 12.73 5.26 16.88

CHY

50.0 20.0 10.0 5.0 2.5

1653.00 1653.00 1650.00 1649.67 1650.00

f f f f +

0.00 0.00 0.00 0.58 0.00

0.00 0.00 0.00 0.03 0.00

0.0907 0.0467 0.0247 0.0127 0.0083

+ + +f +

4.46 7.53 10.20 24.12 18.33

Triplicate

measurements

at each concentration.

Fig. 2B to illustrate the effect of the D,O dilution on the myoglobin spectra. The remarkable overlap of the HOD bands clearly supports the suitability of the HOD band for alignment purposes. Myoglobin has 80% a-helix (12), resulting in a sharp symmetrical amide I band. The background-subtracted spectra in Fig. 2B show that in general the shape and intensity of the amide I region is independent from the D,O fraction of the sample. The maximum of the amide I region shifts slightly from 1652 to 1649 cm-’ as the D,O content increases from 25 to lOO%, corresponding to deuteration of the helical part of myoglobin (13). Larger shifts toward lower wavenumbers have also been observed for some polypeptides and proteins (13-16). The shift in the frequency maximum from 1568 to 1548 cm-’ and change of the shape and symmetry of the amide II band is apparent from Fig. 2B. As the volume fraction of D,O increases, the intensity of the amide II band decreases and the intensity of the amide II’ band (ca. 1450 cm-‘) increases. These changes correspond to different degrees of deuteration of the amide N-H’s as represented by the amide II IR bands (N-H deformations). Effect of Protein Spectra

Secondary

0.0040 0.0035 0.0025 0.0031 0.0015

Structure

on the FT-IR

The spectra of human serum albumin and cr-chymotrypsin were also examined. Albumin and cr-chymotrypsin have lower a-helical content than myoglobin; albumin is 40%, and cy-chymotrypsin is 12% a-helical (18). The solvent-subtracted FT-IR spectrum of human

serum albumin in 50% D,O is displayed in Fig. 3A. The spectrum acquired in H,O is similar to that of bovine serum albumin (17), consistent with the known structural similarities between bovine and human albumins. The slight frequency shift toward lower wavenumbers of the amide I peak is as that with myoglobin. In H,O, only a small amide II band occurs around 1550 cm-’ as a shoulder of a relatively broad amide I band. Incidentally, the peak intensity at 1450 cm-’ also increases with ascending D,O concentration, possibly as a reflection of a slow amide hydrogen exchange (16). The spectrum of cY-chymotrypsin in 50% D,O is shown in Fig. 3B. The protein has the lowest a-helical (12%) and the highest P-sheet (51%) content of the proteins studied. The general appearance of these spectra is similar to prior data available in the literature (7). The minor changes observed on the spectra of the previous examples are also present in this case. The amide I band observed in H,O splits into two peaks. However, an increase in the D,O content of the media sharpens the amide I band and the observed second peak merges into a shoulder on the high wavenumber side of the peak and finally disappears from the spectrum. The amide II band displays a similar trend, namely the separation of the band at 1550 cm-‘, as previously observed with myoglobin and human serum albumin. Band-narrowing techniques are often applied to protein infrared spectra to extract conformational information from the amide bands. The amide bands are known to shift in frequency upon exchange of hydrogen for deuterium. Thus, we examined the spectra of lyso-

SPECTROSCOPIC

ANALYSIS

OF

= 1.6013 = 2.0001

R”2 = 0.992 R”2 = 0.967

+ 5.6509x + 3.5368~

Application Vaccine

2 0.0

0.2

0.4

0.6

CONCENTRATION

FIG. 6. Standard curves antigen and vaccine.

and linear

0.6

1.0

1.2

(mg/mL)

regression

analysis

for malaria

zyme in further detail as a function of volume fraction of D,O. Figure 4 shows the second derivative spectra of lysozyme in H,O, D,O, and a 50:50 mixture. Table 2 gives the frequencies of the amide I components as a function of volume fraction of D,O. These values indicate that the amide I component values shift to lower values upon deuteration, as has been observed previously. Further, the values are shifted in rough proportion to the volume fraction of D,O. Evaluation Aqueous

of Ddt Protein

in the Quantitative Solution

Analysis

PROTEIN

155

SOLUTIONS

mum is independent from the sample size for these proteins and studied concentration range. The absorbance maximum of the amide I frequency was plotted against the protein concentration and displayed in Fig. 5. Regression analysis of the data shows excellent linear response for both proteins. The method also provide good reproducibility of the individual data points, except at the lowest sample concentration as shown in Table 3. However, even relatively high RSD values are acceptable in protein determinations and might be feasible when the protein content of samples in a complicated and unusual environment has to be determined. The reproducibility of the Ddt method can be improved at the expense of the sensitivity as shown in Table 3. Changing the path length from 0.025 to 0.015 mm lead to better reproducibility, but the signal has been lost at the lowest concentration.

10

y,ANTIGEN y,VACCINE

AQUEOUS

of

Next we investigated the validity of the LambertBeer Law under our specific conditions in order to prove its applicability for quantitative analysis. Aqueous solutions of human serum and cu-chymotrypsin were prepared in triplicates. Aliquots of the protein solutions were diluted with an equal volume of D,O and the FTIR spectra were acquired as described under Experimental Methods. The band maximum of the amide I frequency for the HSA samples (n = 18) was 1651.2 + 0.83, 1651.1 ? 1.6 for a-chymotrypsin (n = 15) and 1651.65 k 0.37 for lysozyme. It is apparent that the band maxi-

of Ddt

in the Quantitative

Analysis

of a

Vaccines are the mixtures of antigens and adjuvants. Aluminum hydrogels were used as the adjuvant for the present study. The antigen adsorbs onto the surface of the solid phase as a function of the bulk protein concentration and media parameters such as pH and ionic strength. A concentration series of antigen and vaccine samples was prepared. After background subtraction the area of the amide I band was integrated in the spectral region from 1800 to 1550 cm-‘. Adsorption can cause band broadening and/or shift in the peak maximum as a reflection of adsorption-induced conformational reorganization (6). The integrated areas were then used to construct the standard curves by plotting them as a function of the antigen concentration and the curves are displayed in Fig. 6. Again, regression showed excellent linearity for both samples. The difference in the slope and the intercept of the standard curves should be related to environmental effects. It can be hypothesized that since the vaccine is a cloudy, nontransparent suspension, light scattering occurs and consequently the signal intensity decreases. The importance of the linearity is that it shows that the same method can be used for the quantitative analysis of proteins in solutions and in rather complex environments, without going through a time consuming and complicated sample preparation procedure. The sensitivity and reproducibility of the Ddt methodology suggests that the technique should be applicable for the detailed study on the effects of D,O on the structure of proteins. Secondary structure-frequency assignments can be accomplished and by that facilitate the application of FTIR spectroscopy in a variety of studies on the structure of proteins.

156

ZUBER,

PRESTRELSKL

CONCLUSIONS

A simple FT-IR method for the quantitative analysis of complex protein containing aqueous samples has been developed. The deuterium dilution technique results in a wider spectral range (2800-940 cm-‘) than the previously used methods and consequently is able to cover the whole “biological fingerprint” region, which is usually from 2000 to 900 cm-‘. The method provides a simple sample preparation and consequently, rapid analysis. Since the protein stays in the solution state, no major conformational disturbance, protein unfolding or denaturation, is expected. Previous methodologies had to account for mass recovery at every sample handling step, since this step is eliminated small amounts of samples can be measured (lo-100 pg/ml). This latter point can be extremely important when only a small amount of sample is available for characterization purposes, which often occurs in a research environment, especially working with biopharmaceuticals or novel genetically engineered protein constructs. The Ddt appears applicable to the analysis of samples present in complicated aqueous environments, such as vaccines and other formulated pharmaceutical products. Utilizing the resolution enhancement and deconvolution techniques, conformational analysis of proteins is possible at the higher concentrations. Precise analysis of band shifts as a function of D,O concentration although should be performed in order to establish correct structure-frequency assignments.

AND

BENEDEK

REFERENCES 1. Susi, H. (1969) Structure cules, Dekker, New York. 2. Elliott, A., and Ambrose,

and Stability R. J. (1950)

3. Elliott, A. (1953) Proc. R. Sot. A221, 4. Miyazawa, T. (1960) J. Chem. Phys.

of Biological Nature

Macromole-

165,921.

104. 32, 1647.

5. Miyazawa, T., and Blout, E. R. (1961) J. Am. Chem. Sot. 83,712. 6. Morrissey, B. W., and Stromberg, R. R. (1974) J. Colloid Interface Sci. 46,152-164. 7. Byler, D. M., and Susi, H. (1986) Biopolymers 25, 469-487. 8. Surewicz, W. R., and Mantsch, H. H. (1988) Biochim. Biophys. Acta 952, 115-130. 9. Fink, D. J., and Gendreau, R. M. (1984) Anal. Biochem. 139, 140-148. 10. Chittur, K. K., Fink, D. J., Leininger, R. I., and Hutson, T. B. (1986) J. Colloid Interface Sci. 111, 419-433. 11. Chittur, K. K., Fink, D. J., Hutson, T. B., Gendreau, R. M., Jakobsen, R. J., and Leininger, R. I. (1987) in ACS Symposium Series: Proteins at Interfaces (Horbett, T., and Brash, J. L., Eds.), Am. Chem. Sot., Washington, DC. 12. Provencher, S. W., and Glockner, J. (1981) Biochemistry 20,3337. S. N., and Stevens, L. (1967) J. Biol. Chem. 13. Susi, H., Timasheff, 242, 5460-5466. E. G. (1966) Biopolymers 4, 561-577. 14. Bendit, 15. Moore, W. H., and Krimm, S. (1976) Biopolymers 15,2456-2483. 16. Olinger, J. M., Hill, D. M., Jakobsen, R. J., and Brody, R. S. (1986) Biochem. Biophys. Acta 869, 89-98. 17. Kato, K., Matsui, T., and Tanaka, S. (1987) Appl. Spectrosc. 41, 861-865. 18. Blevins, R. A., and Tulinsky, A. (1985) J. Biol. Chem. 260,4264.