- Email: [email protected]
A method for eliminating Rayleigh scattering from fluorescence spectra
ANALYTICAL
BIOCHEMISTRY
139, 3 16-3 18 ( 1984)
A Method for Eliminating
Rayleigh Scattering
DAVID I. C. KELLS,
JOE D. J. O’NEIL,
from Fluorescence AND THEO
Spectra
HOFMANN
Department of Biochemistry, University of Toronto, Toronto, Ontario M5S lA8, Canada Received October 24, 1983. A numerical method is described for the elimination of Rayleigh scattering from protein fluorescence emission spectra. The method is based upon the observation that Rayleigh scattering is symmetrical about a wavelength at or near the wavelength of excitation. It works best when an automated, computer-based approach can be applied to spectra collected on a data-logging device, although manual correction of chart-recorded spectra is also possible. KEY WORDS: fluorescence; Rayleigh scattering; calcium-binding protein.
Protein fluorescence spectra are often complicated by overlap of the Rayleigh scattering peak and the desired protein emission peaks. In an earlier paper (1) we provided evidence for direct phenylalanine fluorescence in pig intestinal calcium-binding protein (CaBP). In these experiments, it was essential to eliminate the possibility that the shoulder on the tyrosine emission peak, observed when the protein was excited at 263 nm, was indeed due to phenylalanine fluorescence and not to overlapping Rayleigh or Raman scattering. While the Raman scattering can easily be removed by subtracting a solvent baseline from the emission spectra, the Rayleigh scattering peak is usually substantially greater in a protein solution than in the solvent alone, and hence has to be removed by other means. METHOD
AND
RESULTS
We have devised a numerical method for eliminating Rayleigh scattering from spectra and used the following assumptions: (i) the Rayleigh scattering intensity is symmetrical about some wavelength, X,; and, (ii) there is no fluorescence emission from the protein at wavelengths equal to or less than X,. The method can be outlined as follows. First, the wavelength, X,, of the center of the Rayleigh scattering peak is determined. Then, scatter intensity data from the low wavelength side of the scattering peak (wavelengths less than 0003-2697184 $3.00 Copyright 0 1984 by Academic Rey, Inc. All rights of reproduction in any form resewed.
X,) are reflected about X, onto the high wavelength side and subtracted to yield the scattercorrected spectrum (assumption i). In the neighborhood of the Rayleigh peak for wavelengths equal to or less than X,, the fluorescence intensity is set to zero (assumption ii). This procedure is, in principle, suitable for manually removing Rayleigh scattering from fluorescence spectra. However, we have found that the technique is sensitive to the choice of X,, the center of the scattering peak. Erroneous selection of this point of symmetry can lead to positive or negative residual peak spectra. An automated, computer-based approach is therefore preferable and can be readily applied to spectral data collected on a data-logging device. In our system a digital voltmeter is used to measure fluorescence intensities at discrete wavelengths, and the results are recorded on cassette tape for subsequent analysis. The scatter-correction method is as follows. X,, the center of the Rayleigh scattering peak, is determined by fitting a Gaussian curve to the scattering peak data using a nonlinear regression technique described previously (2). The experimental excitation wavelength, the signal intensity at this wavelength, and the spectrofluorometer emission slit width are used as initial estimates for the three unknown parameters (peak center, and height and width at 1/e of this height) that define the Gaussian curve. With these values the program finds the center of the 316
ELIMINATION
OF RAY LEIGH SCATTERING
scattering peak, X,, in 3-8 iterations. Tables of values of AX, the absolute values of wavelength relative to the peak center, and corresponding signal intensities for wavelengths less than or equal to X, and for wavelengths greater than X, are calculated. For each value of AX (X > X,), the signal due to scattering alone is determined from a three-point quadratic Lagrangian interpolation formula (3) applied to the data for wavelengths less than or equal to X,. Actually, the average of two interpolations using the two sets of three points which usually encompass the desired value of AX is used. The scattering signal is then subtracted from the observed signal to produce the “true fluorescence” signal. As with the manual approach, for wavelengths less than or equal to X,, the signal intensity is set to zero. To demonstrate the correction procedure a suspension of glycogen was irradiated at 263 nm. Figure la shows the resulting scatter
290
240
Wavelength
tnml
FIG. 1. Removal of Rayleigh and Raman scattering from a nonfluorescent glycogen spectrum. Glycogen (0.002%) was in 10 mM Tris-HCI buffer, pH 8.0, 150 mM NaCl. Scattering spectra were recorded with an Aminco SPF 500 spectrofluorometer (American Instrument Co., Silver Spring, Md.), in the quantum corrected mode, on cassettetapes at 20 nm min-’ with excitation and emission band widths of 4 nm. (a) Glycogen excited at 263 nm (-); (b) Tris-HCl buffer excited at 263 nm (- - -); (c) Subtraction of the buffer spectrum (b) from the glycogen spectrum (a) eliminates Raman scatter and reduces the Rayleigh scattering slightly (- - -); (d) Rayleigh scattering at 263 nm is eliminated by the numerical procedure described in the text, resulting in a flat baseline (-).
L...,.
240
I*...c.. 380
240
380
FIG. 2. Removal of Rayleigh and Raman scattering from the emission spectrum of CaBP. Calcium CaBP (68.4 PM; A*,, = 0.115) was in 10.0 mM Tris-HCl buffer, pH 8.0, 150 mM NaCl. Fluorescence emission spectra were recorded on cassettetapes at 40 nm mitt-’ with excitation and emission bandwidths of 4 nm. (a) CaBP excited at 263 nm (-); (b) Tris-HCl excited at 263 nm (- - -); (c) subtraction of the buffer spectrum (b) from the protein emission spectrum (a) eliminates solvent Raman scatter and reduces the Rayleigh scattering peak slightly (- - -); (d) Rayleigh scattering at 263 nm is eliminated by the numerical procedure described in the text, resulting in a scatter-free protein emission spectrum (-).
spectrum for glycogen and Fig. 1b that for the buffer showing the typical Rayleigh and Raman scattering peaks in the absence of fluorescence. Figure lc shows the effect of subtracting the solvent contribution from the solution spectrum to remove the Raman peak, and Fig. Id shows the result of the application of the scatter peak remover program outlined above to eliminate the Rayleigh peak. The residual negative peaks in the totally corrected spectrum have intensities less than 0.5% of the maximum in the scatter peak. Experimental conditions, including glycogen concentration, excitation wavelength, and fluorometer slit widths, were varied to produce spectra which could be corrected by this method to yield results similar to those shown in Fig. 1. This confirms that assumption i (scatter peak symmetry) is valid. Figure 2 shows the effect of these correction procedures on the fluorescence spectrum of pig intestinal calcium-binding protein (CaBP,
318
KELLS, O’NEIL,
M, 8800) with overlapping contribution from Rayleigh scattering [see also Fig. 1, curves b, d, and f of Ref. (l)] The absence of residual positive or negative peaks in the low wavelength range of the protein fluorescence and in the baseline (Fig. 2d) shows that the scattering has been effectively removed to give the “true” fluorescence emission. The numerical method described here is superior to the methods using filters for removing Rayleigh scattering. Regular filters are obviously not capable of selectively removing scatter light signals at the wavelengths where the scattering and emission peaks overlap. Wide-bandpass filters further distort the true emission spectra by reducing the intensity near the scattering peak, whereas interference filters in practice limit the choice of excitation wavelength.
AND HOFMANN
Polarizing filters have been used to reduce light scatter (4); however, the Rayleigh scatter is not completely reduced and there is a considerable loss in sensitivity owing to the loss in light transmission of polarizing filters. The method proposed here does not have these disadvantages. REFERENCES 1. O’Neil, J. D. J., Dorrington, K. J., Kells, D. I. C., and Hofmann, T. (1982) B&hem. J. 207, 389396.
Donington, K. J., Kells, D. I. C., Hitchman, A. J. W., Harrison, J. E., and Hofmann, T. (1978) Cunad. J. Biochem. 56,492-499. 3. Arden, B. W. (1963) in An Introduction to Digital Computing, pp 164-166, Addison-Wesley, Reading, Massachusetts. 4. Chen, R. F. (1966) Anal. B&hem. 14,497-499. 2.
BIOCHEMISTRY
139, 3 16-3 18 ( 1984)
A Method for Eliminating
Rayleigh Scattering
DAVID I. C. KELLS,
JOE D. J. O’NEIL,
from Fluorescence AND THEO
Spectra
HOFMANN
Department of Biochemistry, University of Toronto, Toronto, Ontario M5S lA8, Canada Received October 24, 1983. A numerical method is described for the elimination of Rayleigh scattering from protein fluorescence emission spectra. The method is based upon the observation that Rayleigh scattering is symmetrical about a wavelength at or near the wavelength of excitation. It works best when an automated, computer-based approach can be applied to spectra collected on a data-logging device, although manual correction of chart-recorded spectra is also possible. KEY WORDS: fluorescence; Rayleigh scattering; calcium-binding protein.
Protein fluorescence spectra are often complicated by overlap of the Rayleigh scattering peak and the desired protein emission peaks. In an earlier paper (1) we provided evidence for direct phenylalanine fluorescence in pig intestinal calcium-binding protein (CaBP). In these experiments, it was essential to eliminate the possibility that the shoulder on the tyrosine emission peak, observed when the protein was excited at 263 nm, was indeed due to phenylalanine fluorescence and not to overlapping Rayleigh or Raman scattering. While the Raman scattering can easily be removed by subtracting a solvent baseline from the emission spectra, the Rayleigh scattering peak is usually substantially greater in a protein solution than in the solvent alone, and hence has to be removed by other means. METHOD
AND
RESULTS
We have devised a numerical method for eliminating Rayleigh scattering from spectra and used the following assumptions: (i) the Rayleigh scattering intensity is symmetrical about some wavelength, X,; and, (ii) there is no fluorescence emission from the protein at wavelengths equal to or less than X,. The method can be outlined as follows. First, the wavelength, X,, of the center of the Rayleigh scattering peak is determined. Then, scatter intensity data from the low wavelength side of the scattering peak (wavelengths less than 0003-2697184 $3.00 Copyright 0 1984 by Academic Rey, Inc. All rights of reproduction in any form resewed.
X,) are reflected about X, onto the high wavelength side and subtracted to yield the scattercorrected spectrum (assumption i). In the neighborhood of the Rayleigh peak for wavelengths equal to or less than X,, the fluorescence intensity is set to zero (assumption ii). This procedure is, in principle, suitable for manually removing Rayleigh scattering from fluorescence spectra. However, we have found that the technique is sensitive to the choice of X,, the center of the scattering peak. Erroneous selection of this point of symmetry can lead to positive or negative residual peak spectra. An automated, computer-based approach is therefore preferable and can be readily applied to spectral data collected on a data-logging device. In our system a digital voltmeter is used to measure fluorescence intensities at discrete wavelengths, and the results are recorded on cassette tape for subsequent analysis. The scatter-correction method is as follows. X,, the center of the Rayleigh scattering peak, is determined by fitting a Gaussian curve to the scattering peak data using a nonlinear regression technique described previously (2). The experimental excitation wavelength, the signal intensity at this wavelength, and the spectrofluorometer emission slit width are used as initial estimates for the three unknown parameters (peak center, and height and width at 1/e of this height) that define the Gaussian curve. With these values the program finds the center of the 316
ELIMINATION
OF RAY LEIGH SCATTERING
scattering peak, X,, in 3-8 iterations. Tables of values of AX, the absolute values of wavelength relative to the peak center, and corresponding signal intensities for wavelengths less than or equal to X, and for wavelengths greater than X, are calculated. For each value of AX (X > X,), the signal due to scattering alone is determined from a three-point quadratic Lagrangian interpolation formula (3) applied to the data for wavelengths less than or equal to X,. Actually, the average of two interpolations using the two sets of three points which usually encompass the desired value of AX is used. The scattering signal is then subtracted from the observed signal to produce the “true fluorescence” signal. As with the manual approach, for wavelengths less than or equal to X,, the signal intensity is set to zero. To demonstrate the correction procedure a suspension of glycogen was irradiated at 263 nm. Figure la shows the resulting scatter
290
240
Wavelength
tnml
FIG. 1. Removal of Rayleigh and Raman scattering from a nonfluorescent glycogen spectrum. Glycogen (0.002%) was in 10 mM Tris-HCI buffer, pH 8.0, 150 mM NaCl. Scattering spectra were recorded with an Aminco SPF 500 spectrofluorometer (American Instrument Co., Silver Spring, Md.), in the quantum corrected mode, on cassettetapes at 20 nm min-’ with excitation and emission band widths of 4 nm. (a) Glycogen excited at 263 nm (-); (b) Tris-HCl buffer excited at 263 nm (- - -); (c) Subtraction of the buffer spectrum (b) from the glycogen spectrum (a) eliminates Raman scatter and reduces the Rayleigh scattering slightly (- - -); (d) Rayleigh scattering at 263 nm is eliminated by the numerical procedure described in the text, resulting in a flat baseline (-).
L...,.
240
I*...c.. 380
240
380
FIG. 2. Removal of Rayleigh and Raman scattering from the emission spectrum of CaBP. Calcium CaBP (68.4 PM; A*,, = 0.115) was in 10.0 mM Tris-HCl buffer, pH 8.0, 150 mM NaCl. Fluorescence emission spectra were recorded on cassettetapes at 40 nm mitt-’ with excitation and emission bandwidths of 4 nm. (a) CaBP excited at 263 nm (-); (b) Tris-HCl excited at 263 nm (- - -); (c) subtraction of the buffer spectrum (b) from the protein emission spectrum (a) eliminates solvent Raman scatter and reduces the Rayleigh scattering peak slightly (- - -); (d) Rayleigh scattering at 263 nm is eliminated by the numerical procedure described in the text, resulting in a scatter-free protein emission spectrum (-).
spectrum for glycogen and Fig. 1b that for the buffer showing the typical Rayleigh and Raman scattering peaks in the absence of fluorescence. Figure lc shows the effect of subtracting the solvent contribution from the solution spectrum to remove the Raman peak, and Fig. Id shows the result of the application of the scatter peak remover program outlined above to eliminate the Rayleigh peak. The residual negative peaks in the totally corrected spectrum have intensities less than 0.5% of the maximum in the scatter peak. Experimental conditions, including glycogen concentration, excitation wavelength, and fluorometer slit widths, were varied to produce spectra which could be corrected by this method to yield results similar to those shown in Fig. 1. This confirms that assumption i (scatter peak symmetry) is valid. Figure 2 shows the effect of these correction procedures on the fluorescence spectrum of pig intestinal calcium-binding protein (CaBP,
318
KELLS, O’NEIL,
M, 8800) with overlapping contribution from Rayleigh scattering [see also Fig. 1, curves b, d, and f of Ref. (l)] The absence of residual positive or negative peaks in the low wavelength range of the protein fluorescence and in the baseline (Fig. 2d) shows that the scattering has been effectively removed to give the “true” fluorescence emission. The numerical method described here is superior to the methods using filters for removing Rayleigh scattering. Regular filters are obviously not capable of selectively removing scatter light signals at the wavelengths where the scattering and emission peaks overlap. Wide-bandpass filters further distort the true emission spectra by reducing the intensity near the scattering peak, whereas interference filters in practice limit the choice of excitation wavelength.
AND HOFMANN
Polarizing filters have been used to reduce light scatter (4); however, the Rayleigh scatter is not completely reduced and there is a considerable loss in sensitivity owing to the loss in light transmission of polarizing filters. The method proposed here does not have these disadvantages. REFERENCES 1. O’Neil, J. D. J., Dorrington, K. J., Kells, D. I. C., and Hofmann, T. (1982) B&hem. J. 207, 389396.
Donington, K. J., Kells, D. I. C., Hitchman, A. J. W., Harrison, J. E., and Hofmann, T. (1978) Cunad. J. Biochem. 56,492-499. 3. Arden, B. W. (1963) in An Introduction to Digital Computing, pp 164-166, Addison-Wesley, Reading, Massachusetts. 4. Chen, R. F. (1966) Anal. B&hem. 14,497-499. 2.