ARCHIVES
Vol.
OF BIOCHEMISTRY
298, No. 2, November
AND
BIOPHYSICS
1, pp. 678-681,
1992
Vibrational Circular Dichroism Studies of Epidermal Growth Factor and Basic Fibroblast Growth Factor Rina K. Dukor,* Petr Steven J. Prestrelski,? of Chemistry, Amgen Inc.,
*Department Department,
Received
May
Pancoska,*‘l Timothy A. Keiderling,*‘2 and Tsutomu Arakawat University Amgen Center,
28, 1992, and in revised
form
of Illinois 1840
July
at Chicago, Dehavilland
Box Drive,
Chemistry
16, 1992
Vibrational circular dichroism (VCD) studies are reported for two unrelated recombinant growth factor proteins: epidermal growth factor and basic fibroblast growth factor (bFGF). NMR, electronic CD, and bFGF X-ray studies indicate that these two proteins are primarily composed of &sheet and loop secondary structure elements with no detectable cY-helices. Two reports on solution conformation of these proteins using FTIR absorption spectroscopy with subsequent resolution enhancement confirmed the presence of a large fraction of a j3-sheet conformation but in addition indicated the presence of large absorption bands in the 1650-1656 cm-’ region, which are typically assigned to a-helices. The VCD spectra of both proteins have band shapes that strongly resemble those of other high &sheet fraction proteins, such as the trypsin family of proteins. Quantitative analysis of the VCD spectra also indicates that these proteins are predominantly in &sheet and extended (“other”) conformations with very little a-helix fraction. These results agree with the CD interpretation and affirm that the FTIR peaks in the region 1650-1656 cm-’ can be assigned to loops. This study provides an example of the limitations of using FTIR frequencies alone for examination of protein secondary structure. Q 1992 Academic Press,
4348, Chicago, Illinois 60680; and tProtein Thousand Oaks, California 91320
Inc.
Epidermal growth factor (EGFj3 and basic fibroblast growth factor (bFGF) are two unrelated mitogenic growth factors each of which plays roles in growth and differr Permanent address: Department of Chemical Physics, Charles University, Prague 2, Czechoslovakia. * To whom correspondence should be addressed. Fax: (312) 996-0431. 3 Abbreviations used: EGF, epidermal growth factor; bFGF, basic fibroblast growth factor; E-CD, electronic circular dichroism; IR, infrared; FTIR, Fourier transform IR; VCD, vibrational circular dichroism; FSD, Fourier self-deconvolution.
entiation in a variety of tissues (1, 2). Recombinantly produced versions of these proteins have been extensively characterized using several biophysical techniques including electronic circular dichroism (E-CD) (3-5), NMR (6), infrared (IR) spectroscopy (7, 8) and X-ray crystallography (9). The X-ray crystal structure of bFGF (9) indicated that the structure consists of 12 antiparallel fistrands arranged in a pattern with approximate threefold symmetry and rather extended loops between some strands. It was also noted that this folding pattern has been previously observed for soybean trypsin inhibitor. NMR (6) and E-CD (3-5) solution studies also indicated that the predominant secondary structure components of both proteins in solution are that of P-sheets and loops with no detectable cu-helices. Two reports on solution conformation of these proteins using Fourier transform IR (FTIR) absorption spectroscopy with subsequent resolution enhancement (7, 8) confirmed the presence of a large fraction of P-sheet conformation. However, in addition, these spectra indicated the presence of large absorption bands (21-39s of the relative integrated area) in the 1650-1656 cm-’ region, which are typically assigned to an o-helical conformation in globular proteins (10,ll). Based on the results from NMR and E-CD spectra, the bands observed in the FTIR absorption spectra were interpreted as due to loops or, possibly, due to other nonhelical components; however, it is possible that the discrepancy may be related to the high concentration conditions used for the IR studies. To answer this question, we have measured vibrational circular dichroism (VCD) spectra of EGF and bFGF in the infrared with sample concentrations similar to those used for the original FTIR studies (7, 8). Vibrational circular dichroism is a measure of the difference in absorption of left- and right-circularly polarized radiation by molecular vibrational transitions (12). VCD combines the advantages of the conformational sensitivity of a measurement dependent on optical activity, such as
678
0003-9861/92
Copyright All
rights
0
1992 of reproduction
by
$5.00 Press, Inc. form reserved.
Academic in any
RECOMBINANT
GROWTH
FACTOR
circular dichroism (CD), with the higher resolution and multiple transitions characteristic of vibrational spectroscopy, such as IR. We have shown that this combination leads to an enhanced sensitivity to protein conformation in solution as compared to either conventional CD or IR studies (13). More recently, we have developed a stable method of using VCD spectra to quantitatively determine the relative fraction of different secondary structure components (14) and have compared the results of such an analysis with a parallel study of E-CD (15). EXPERIMENTAL Recombinant EGF of the natural human sequence was obtained from Amgen Biologicals. Human recombinant bFGF was purified from broken Escherichia coli cells by a series of chromatographic procedures as described previously (16). The bFGF used is an analog with the cysteines at positions 70 and 88 of the native sequence replaced by serines. This analog has been extensively analyzed for activity and stability (4, 16). To achieve hydrogen-deuterium exchange, EGF was dissolved in D20 and lyophilized. The cycle is repeated twice. bFGF was dissolved in 20 mM Tris buffer (in D,O) and allowed to exchange for at least 48 h. For spectroscopic studies, EGF was dissolved in 0.32 mM DC1 (in D,O) solution at a concentration of 50 mg/ml. Small amounts of NaOD solution (freshly prepared by adding D,O to sodium metal) were added dropwise to achieve a final pD of 3.5 for the EGF protein solution. The bFGF solution was prepared at a concentration of 50 mg/ml in Tris buffer (pD 6.8). The pH values were measured with a microelectrode (Ingold) and a Corning 145 pH meter calibrated against standard buffers. To correct for the isotope effect, 0.4 was added to the observed pH to obtain pD values (17). In each case, the spectroscopic samples were made by pressing a small amount of solution between two CaF2 windows separated with a 0.05-mm Teflon spacer. VCD and absorption spectra were obtained on the dispersive instrument at UIC which has been previously described in detail (12). The spectra were recorded at -10 cm-’ resolution, collected with a 10-s time constant, and averaged over four scans. Poly-DL-lysine, in DrO or in Tris buffer, with the concentration adjusted to match the absorbance of the protein sample, was used for the VCD baseline correction. For consistency and as a check that the same structural form was obtained as previously reported, the FTIR spectra of the protein samples were remeasured at UIC and then further analyzed using Fourier selfdeconvolution (18). FTIR absorption spectra were obtained on a Digilab FTS-60 FTIR spectrometer as an average of 1024 scans at 2 cm-’ resolution. After subtraction of the water vapor absorbance spectrum, Fourier self-deconvolution (FSD) calculations were performed, using standard Digilab software, which is based on the algorithm by Kauppinen et al. (18), with the parameters of FWHH = 18 and 14 and an enhancement factor k = 2.5 and 2 for EGF and bFGF, respectively. Quantitative analysis of VCD spectra was done following the method described by Pancoska et al. (14). Briefly, there are three basic steps. First, EGF and bFGF VCD spectra are checked for compatibility with those of a reference set. The training set used consists of 21 globular proteins for which the X-ray-determined structure is known and analyzed using the Kabsch and Sander (19) algorithm for secondary structure content and for which we have obtained amide I’ and II VCD and FTIR data (20). For quantitative studies, a factor analysis method is used which consists of principal component decomposition of the experimental VCD spectra into a linear combination of orthogonal subspectra generated without any a priori parameterization. The spectra to be analyzed (of EGF and bFGF) are coprocessed with those of the training set (20). The similarity of the VCD of the new proteins with any (groups) in the training set is judged using cluster analysis based on the vectors of subspectral coefficients for each protein. For the subsequent quantitative analysis to be reliable, the new spectra should not form a separate cluster
VIBRATIONAL
CIRCULAR
679
DICHROISM
but should instead exhibit similarity to one of the major groups in the set of reference spectra. If this condition is fulfilled, the coefficients are then used in a selective multiple regression scheme (14) to fit the reference set of secondary structure fractions of the 21 “known” proteins. All possible combinations of pairs, triplets, quadruplets, etc., of subspectral coefficients are used for the test regressions, and statistical criteria (correlation coefficients and F tests of significance of coefficient contribution) are implemented for the selection of the final model. In this manner, we select only that part of the spectral information which is relevant for the evaluation of fractional secondary structure of the proteins studied. In the final step, the coefficients obtained from the factor analysis scheme for the EGF and bFGF VCD spectra are input into the final multiple regression equations and their secondary structure fractions are calculated as predicted from the VCD spectra.
RESULTS
AND
DISCUSSION
The VCD, dispersive IR absorption, and resolutionenhanced FSD-FTIR absorption spectra are shown (top to bottom) in Fig. 1 for EGF and in Fig. 2 for bFGF and the corresponding frequencies for the three types of spectra are listed in Table I. The IR and VCD spectra have been normalized to a peak absorbance of A,,, = 1.0 so that the VCD scale reads directly in M/A at A,,,. In both cases,the VCD spectrum is dominated by a negative band on the low frequency side of the absorbance maximum and a small positive one to higher frequency, with the VCD zero crossing for both proteins lying somewhat higher in frequency than their respective absorption maxima which are 1651 cm-’ for EGF and 1643 cm-’ for bFGF (Table I). The VCD spectrum observed in Fig. 1 (top) for EGF is very similar, aside from frequency, to that seen for pro-
AA x105
A
I
1750
1
I
,
1650 Frequency (cml>
1
i0
FIG. 1. VCD (top), IR absorption (middle), and Fourier self-deconvolution (bottom) spectra of EGF in D,O/DCl/NaOD (pD 3.5). The scale for the FSD-FTIR spectrum is arbitrary. The IR and VCD spectra are normalized to a peak absorbance of A,. = 1.0.
680
DUKOR
4
AA x10
0
5
-4
.5 A 0
1
FIG. 2. VCD (top), IR absorption volution (bottom) spectra of bFGF are plotted as in Fig. 1.
(middle), in 20 mM
and Fourier Tris buffer.
self-deconThe spectra
teins whose secondary structure composition consists mostly of P-sheet and “other” (which in our analysis includes several minor structural types and excludes helix, sheet, turn, and bend). These similar proteins have only a very small fraction of an a-helix (14). On the other hand, this negative couplet VCD is very different from any VCD seen for proteins with mixed (Y and p structures (13, 14). The VCD spectrum of bFGF (Fig. 2, top) is also similar to those proteins that have a high fraction of /3sheet and extended (or other) secondary structures, such as a-chymotrypsin, trypsin, trypsin inhibitor, chymotrypsinogen, and concanavalin A (14). The VCD intensity, measured as AA/A, for EGF and bFGF are almost the same and equal to -1 X 10-4. This value is 20% higher than that obtained for chymotrypsinogen and trypsin inhibitor and equal to the VCD intensity of chymotrypsin. All of these high p proteins have somewhat higher VCD intensities than are obtained for proteins with mixed QI and p character (13, 14). This is due to the tendency of the characteristic a-helix and /3sheet bandshapes to cancel partially in VCD. The FSD-FTIR spectra for EGF and bFGF, shown in Figs. 1 and 2 bottom, respectively, are similar to the results reported previously (7, 8). The similarity of our FSD results and the previously published deconvolved spectra confirms that proteins in this study are in the same conformation as studied earlier, in other words, no changes were induced by our deuteration method or any other minor differences in the handling procedures. Therefore, the VCD results obtained and the conclusions derived from them are applicable to the previous FTIR studies.
ET
AL.
The qualitative bandshape observations described above were put on more quantitative basis using the factor analysis methodology. First, it was observed that the incorporation of EGF and bFGF VCD spectra into the reference set of 28 globular proteins did not substantially change the bandshape of the calculated subspectra. Also their incorporation did not change the number of subspectra (six) necessary for the description of all VCD spectra in the analyzed data set. The cluster analysis using the subspectral coefficients showed the connection of EGF and bFGF VCD spectra to be best clustered with those of chymotrypsinogen and casein, with the next closest protein cluster being composed of trypsin, trypsin inhibitor, rhodanese, chymotrypsin, and superoxide dismutase. This serves to support the applicability of quantitative analysis for the EGF and bFGF VCD spectra. The final multiple regression equations were similar to those found previously for smaller training sets. [In this analysis we used the coefficients of subspectra 2 and 5 for calculation of fraction of a-helix, subspectrum 2 for calculation of P-sheet content, subspectra 3 and 5 for the determination of the bend fraction and subspectra 1 and 3 for the calculation of fraction of other structural types.] The turn fractions remain undetermined. [We have shown recently that for obtaining a statistically reliable estimation of the fraction of turn, the amide I’ spectra should be supplemented by the amide II VCD (20).] The quantitative secondary structural predictions are summarized in Table II. Based on the VCD spectra obtained for EGF and bFGF and comparison of the bandshapes with the previously analyzed set of 28 proteins, we conclude that these two proteins are predominantly in P-sheet and extended (other) conformations. There is very little a-helix, if any, in agreement with all other spectral measurements except the frequency-based FTIR analysis (7, 8). Our previous extensive analysis of accuracy of the VCD method provided an estimate of error limits for evaluation of individual secondary structure types: for P-sheet fraction the absolute error should be less than 15%, and for bend and other fractions less than 10%. These error limits can be
TABLE
I
Summary of Frequencies” EGF
bFGF
IR
FSD
VCD
IR
FSD
VCD
1651
1677 1664 1654 1638
1669 (+) 1642 (-)
1643
1685 1666 1652 1643 1635
1663 (+) 1635 (-)
y In cm-‘.
RECOMBINANT TABLE
Summary
GROWTH
VCD” FTIR* UVCD’ NMRd
0 39 (?) 0
VIBRATIONAL
Results
18 34
a-helix
681
DICHROISM
ACKNOWLEDGMENTS
bFGF p-sheet
CIRCULAR
are much easier to interpret than those obtained by ECD or FSD-FTIR.
II
of Quantitative EGF
a-helix
FACTOR
P-sheet
0 21 (?)
34 42
o-5
54-72
42
a This study. * Refs. (7) and (8). ’ Refs. (3-5). d Ref. (6).
This work was supported at UIC by a grant from the National Institutes of Health (GM30147, to TAK). The cooperation between Charles University and UIC is aided by an NSF grant for Joint CzechoslovakUSA (INT 91-07588) research. The FTIR instrumentation used was provided in part by grants from the NIH, NSF, and University of Illinois.
REFERENCES 1. Gospodarowicz, D., Massoglia, S., Cheng, J., Lui, G. M., and Bohlen, P. (1985) J. Cell Physiol. 122,323-332. 2. Sporn, M. B., and Roberts, A. B. (1985) Nature 313, 745-747. T., McGinley, M. D., Rohde, M. F., and 3. Narhi, L. O., Arakawa, Westcott, K. R. (1992) Znt. J. Pept. Protein Res. 39, 182-187.
4. Arakawa,
useful for comparisons of VCD quantitative analysis with the outputs of other methods. These results are in substantive agreement with those previously obtained by ECD (3-5), considering the limitations of each technique (15). The FSD-FTIR results can also be seen to be consistent with this analysis if the peaks in the region 16501656 cm-’ are assignedto loop structures as was originally suggestedby Prestrelski and co-workers (7,8). The VCD sign patterns reported here serve to distinguish more definitely between possible interpretations of the FTIR data. CONCLUSION
VCD results only establish the predominant conformation in solution and give no direct evidence that the peaks in the FSD-FTIR spectra in the region 1650-1656 cm-l are indicative of the loop structures. We can say with certainty that these dominant FSD-FTIR bands in the spectra of the two proteins are not due to a-helical structures, since no such conformation was observed in the VCD spectra, but at this point we do not know their exact origin. It is reasonable to assign them to loops (as was done in the two previous FTIR studies) since loop structures are observed in high amounts in the X-ray (9) and NMR (6) analyses. This is further supported by the VCD spectra observed being characteristic of the extended structures (21) that may be seen in loops with only modestly constrained conformations. These results demonstrate the potential errors in interpreting conformational results based on the IR frequencies alone and show the need to use multiple spectroscopic techniques to develop a more accurate picture of protein secondary structure. Qualitatively, VCD results
T., Hsu, Y.-R., Schiffer, S. G., Tsai, L. B., Curless, G., and Fox, G. M. (1989) Rio&em. Biophys. Res. Commun. 161,335-
341. 5. Wu, C.-S. C., Thompson, Chem. 10,427-436. 6. Cooke, R. M., Wilkinson, M. J., Campbell,
S. A., and Yang,
J. T. (1991)
A. J., Baron, M., Pastore, A., Tappin, J. D., Gregory, H., and Sheard, B. (1987) Nature
327,339-341. 7. Prestrelski, S. J., Arakawa, (1991)
Arch.
J. Protein
Biochem.
T., Kenney, W. C., and Byler, Biophys. 285, 111-115.
D. M.
8. Prestrelski,
S. J., Arakawa, T., Wu, C.-S. C., O’Neal, K. D., Westcott, K. W., and Narhi, L. 0. (1992) J. Biol. Chem. 267, 319-322. 9. Zhu, X., Komiya, H., Chirino, A., Faham, S., Fox, G. M., Arakawa, T., Hsu, B. T., and Rees, D. C. (1991) Science 251,90-93. 25, 469-487. 10. Byler, D. M., and Susi, H. (1986) Biopolymers 11. Surewicz, W. K., and Mantsch, H. H. (1988) Biochim. Biophys. Acta 952,115-130. 12. Keiderling, T. A. (1990) in Practical Fourier Transform Infrared Spectroscopy (Ferraro, J. R., and Krishnan, K., Eds.), pp. 203-284, Academic Press, San Diego. 13. Pancoska, P., Yasui, S. C., and Keiderling, T. A. (1989) Biochemistry 28, 5917-5923. 14. Pancoska, P., Yasui, S. C., and Keiderling, T. A. (1991) Biochemistry
30,5089-5103. P., and Keiderling, T. A. (1991) Biochemistry 30,688515. Pancoska, 6895. S. G., Rhode, M. F., Tsai, L. B., Banks, A. R., 16. Fox, G. M., Schiffer, and Arakawa, T. (1988) J. Biol. Chem. 263, 18,452-18,458.
17. Salomaa, P., Schaleger, Sot. 86,1-7.
L. L., and Long,
F. A. (1964)
J. Am. Chem.
18. Kauppinen, J. K., Moffatt, D. J., Mantsch, H. H., and Cameron, D. G. (1981) Appl. Spectrosc. 35, 271-276. 19. Kabsch, W., and Sander, C. (1983) Biopolymers 22, 2577-2637. M., and Keiderling, T. A. (1992) 20. Pancoska, P., Gupta, V., Urbanova, to be published. 21. Dukor, R. K., and Keiderling, T. A. (1992) Biopolymers 31, 17471761.
Vol.
OF BIOCHEMISTRY
298, No. 2, November
AND
BIOPHYSICS
1, pp. 678-681,
1992
Vibrational Circular Dichroism Studies of Epidermal Growth Factor and Basic Fibroblast Growth Factor Rina K. Dukor,* Petr Steven J. Prestrelski,? of Chemistry, Amgen Inc.,
*Department Department,
Received
May
Pancoska,*‘l Timothy A. Keiderling,*‘2 and Tsutomu Arakawat University Amgen Center,
28, 1992, and in revised
form
of Illinois 1840
July
at Chicago, Dehavilland
Box Drive,
Chemistry
16, 1992
Vibrational circular dichroism (VCD) studies are reported for two unrelated recombinant growth factor proteins: epidermal growth factor and basic fibroblast growth factor (bFGF). NMR, electronic CD, and bFGF X-ray studies indicate that these two proteins are primarily composed of &sheet and loop secondary structure elements with no detectable cY-helices. Two reports on solution conformation of these proteins using FTIR absorption spectroscopy with subsequent resolution enhancement confirmed the presence of a large fraction of a j3-sheet conformation but in addition indicated the presence of large absorption bands in the 1650-1656 cm-’ region, which are typically assigned to a-helices. The VCD spectra of both proteins have band shapes that strongly resemble those of other high &sheet fraction proteins, such as the trypsin family of proteins. Quantitative analysis of the VCD spectra also indicates that these proteins are predominantly in &sheet and extended (“other”) conformations with very little a-helix fraction. These results agree with the CD interpretation and affirm that the FTIR peaks in the region 1650-1656 cm-’ can be assigned to loops. This study provides an example of the limitations of using FTIR frequencies alone for examination of protein secondary structure. Q 1992 Academic Press,
4348, Chicago, Illinois 60680; and tProtein Thousand Oaks, California 91320
Inc.
Epidermal growth factor (EGFj3 and basic fibroblast growth factor (bFGF) are two unrelated mitogenic growth factors each of which plays roles in growth and differr Permanent address: Department of Chemical Physics, Charles University, Prague 2, Czechoslovakia. * To whom correspondence should be addressed. Fax: (312) 996-0431. 3 Abbreviations used: EGF, epidermal growth factor; bFGF, basic fibroblast growth factor; E-CD, electronic circular dichroism; IR, infrared; FTIR, Fourier transform IR; VCD, vibrational circular dichroism; FSD, Fourier self-deconvolution.
entiation in a variety of tissues (1, 2). Recombinantly produced versions of these proteins have been extensively characterized using several biophysical techniques including electronic circular dichroism (E-CD) (3-5), NMR (6), infrared (IR) spectroscopy (7, 8) and X-ray crystallography (9). The X-ray crystal structure of bFGF (9) indicated that the structure consists of 12 antiparallel fistrands arranged in a pattern with approximate threefold symmetry and rather extended loops between some strands. It was also noted that this folding pattern has been previously observed for soybean trypsin inhibitor. NMR (6) and E-CD (3-5) solution studies also indicated that the predominant secondary structure components of both proteins in solution are that of P-sheets and loops with no detectable cu-helices. Two reports on solution conformation of these proteins using Fourier transform IR (FTIR) absorption spectroscopy with subsequent resolution enhancement (7, 8) confirmed the presence of a large fraction of P-sheet conformation. However, in addition, these spectra indicated the presence of large absorption bands (21-39s of the relative integrated area) in the 1650-1656 cm-’ region, which are typically assigned to an o-helical conformation in globular proteins (10,ll). Based on the results from NMR and E-CD spectra, the bands observed in the FTIR absorption spectra were interpreted as due to loops or, possibly, due to other nonhelical components; however, it is possible that the discrepancy may be related to the high concentration conditions used for the IR studies. To answer this question, we have measured vibrational circular dichroism (VCD) spectra of EGF and bFGF in the infrared with sample concentrations similar to those used for the original FTIR studies (7, 8). Vibrational circular dichroism is a measure of the difference in absorption of left- and right-circularly polarized radiation by molecular vibrational transitions (12). VCD combines the advantages of the conformational sensitivity of a measurement dependent on optical activity, such as
678
0003-9861/92
Copyright All
rights
0
1992 of reproduction
by
$5.00 Press, Inc. form reserved.
Academic in any
RECOMBINANT
GROWTH
FACTOR
circular dichroism (CD), with the higher resolution and multiple transitions characteristic of vibrational spectroscopy, such as IR. We have shown that this combination leads to an enhanced sensitivity to protein conformation in solution as compared to either conventional CD or IR studies (13). More recently, we have developed a stable method of using VCD spectra to quantitatively determine the relative fraction of different secondary structure components (14) and have compared the results of such an analysis with a parallel study of E-CD (15). EXPERIMENTAL Recombinant EGF of the natural human sequence was obtained from Amgen Biologicals. Human recombinant bFGF was purified from broken Escherichia coli cells by a series of chromatographic procedures as described previously (16). The bFGF used is an analog with the cysteines at positions 70 and 88 of the native sequence replaced by serines. This analog has been extensively analyzed for activity and stability (4, 16). To achieve hydrogen-deuterium exchange, EGF was dissolved in D20 and lyophilized. The cycle is repeated twice. bFGF was dissolved in 20 mM Tris buffer (in D,O) and allowed to exchange for at least 48 h. For spectroscopic studies, EGF was dissolved in 0.32 mM DC1 (in D,O) solution at a concentration of 50 mg/ml. Small amounts of NaOD solution (freshly prepared by adding D,O to sodium metal) were added dropwise to achieve a final pD of 3.5 for the EGF protein solution. The bFGF solution was prepared at a concentration of 50 mg/ml in Tris buffer (pD 6.8). The pH values were measured with a microelectrode (Ingold) and a Corning 145 pH meter calibrated against standard buffers. To correct for the isotope effect, 0.4 was added to the observed pH to obtain pD values (17). In each case, the spectroscopic samples were made by pressing a small amount of solution between two CaF2 windows separated with a 0.05-mm Teflon spacer. VCD and absorption spectra were obtained on the dispersive instrument at UIC which has been previously described in detail (12). The spectra were recorded at -10 cm-’ resolution, collected with a 10-s time constant, and averaged over four scans. Poly-DL-lysine, in DrO or in Tris buffer, with the concentration adjusted to match the absorbance of the protein sample, was used for the VCD baseline correction. For consistency and as a check that the same structural form was obtained as previously reported, the FTIR spectra of the protein samples were remeasured at UIC and then further analyzed using Fourier selfdeconvolution (18). FTIR absorption spectra were obtained on a Digilab FTS-60 FTIR spectrometer as an average of 1024 scans at 2 cm-’ resolution. After subtraction of the water vapor absorbance spectrum, Fourier self-deconvolution (FSD) calculations were performed, using standard Digilab software, which is based on the algorithm by Kauppinen et al. (18), with the parameters of FWHH = 18 and 14 and an enhancement factor k = 2.5 and 2 for EGF and bFGF, respectively. Quantitative analysis of VCD spectra was done following the method described by Pancoska et al. (14). Briefly, there are three basic steps. First, EGF and bFGF VCD spectra are checked for compatibility with those of a reference set. The training set used consists of 21 globular proteins for which the X-ray-determined structure is known and analyzed using the Kabsch and Sander (19) algorithm for secondary structure content and for which we have obtained amide I’ and II VCD and FTIR data (20). For quantitative studies, a factor analysis method is used which consists of principal component decomposition of the experimental VCD spectra into a linear combination of orthogonal subspectra generated without any a priori parameterization. The spectra to be analyzed (of EGF and bFGF) are coprocessed with those of the training set (20). The similarity of the VCD of the new proteins with any (groups) in the training set is judged using cluster analysis based on the vectors of subspectral coefficients for each protein. For the subsequent quantitative analysis to be reliable, the new spectra should not form a separate cluster
VIBRATIONAL
CIRCULAR
679
DICHROISM
but should instead exhibit similarity to one of the major groups in the set of reference spectra. If this condition is fulfilled, the coefficients are then used in a selective multiple regression scheme (14) to fit the reference set of secondary structure fractions of the 21 “known” proteins. All possible combinations of pairs, triplets, quadruplets, etc., of subspectral coefficients are used for the test regressions, and statistical criteria (correlation coefficients and F tests of significance of coefficient contribution) are implemented for the selection of the final model. In this manner, we select only that part of the spectral information which is relevant for the evaluation of fractional secondary structure of the proteins studied. In the final step, the coefficients obtained from the factor analysis scheme for the EGF and bFGF VCD spectra are input into the final multiple regression equations and their secondary structure fractions are calculated as predicted from the VCD spectra.
RESULTS
AND
DISCUSSION
The VCD, dispersive IR absorption, and resolutionenhanced FSD-FTIR absorption spectra are shown (top to bottom) in Fig. 1 for EGF and in Fig. 2 for bFGF and the corresponding frequencies for the three types of spectra are listed in Table I. The IR and VCD spectra have been normalized to a peak absorbance of A,,, = 1.0 so that the VCD scale reads directly in M/A at A,,,. In both cases,the VCD spectrum is dominated by a negative band on the low frequency side of the absorbance maximum and a small positive one to higher frequency, with the VCD zero crossing for both proteins lying somewhat higher in frequency than their respective absorption maxima which are 1651 cm-’ for EGF and 1643 cm-’ for bFGF (Table I). The VCD spectrum observed in Fig. 1 (top) for EGF is very similar, aside from frequency, to that seen for pro-
AA x105
A
I
1750
1
I
,
1650 Frequency (cml>
1
i0
FIG. 1. VCD (top), IR absorption (middle), and Fourier self-deconvolution (bottom) spectra of EGF in D,O/DCl/NaOD (pD 3.5). The scale for the FSD-FTIR spectrum is arbitrary. The IR and VCD spectra are normalized to a peak absorbance of A,. = 1.0.
680
DUKOR
4
AA x10
0
5
-4
.5 A 0
1
FIG. 2. VCD (top), IR absorption volution (bottom) spectra of bFGF are plotted as in Fig. 1.
(middle), in 20 mM
and Fourier Tris buffer.
self-deconThe spectra
teins whose secondary structure composition consists mostly of P-sheet and “other” (which in our analysis includes several minor structural types and excludes helix, sheet, turn, and bend). These similar proteins have only a very small fraction of an a-helix (14). On the other hand, this negative couplet VCD is very different from any VCD seen for proteins with mixed (Y and p structures (13, 14). The VCD spectrum of bFGF (Fig. 2, top) is also similar to those proteins that have a high fraction of /3sheet and extended (or other) secondary structures, such as a-chymotrypsin, trypsin, trypsin inhibitor, chymotrypsinogen, and concanavalin A (14). The VCD intensity, measured as AA/A, for EGF and bFGF are almost the same and equal to -1 X 10-4. This value is 20% higher than that obtained for chymotrypsinogen and trypsin inhibitor and equal to the VCD intensity of chymotrypsin. All of these high p proteins have somewhat higher VCD intensities than are obtained for proteins with mixed QI and p character (13, 14). This is due to the tendency of the characteristic a-helix and /3sheet bandshapes to cancel partially in VCD. The FSD-FTIR spectra for EGF and bFGF, shown in Figs. 1 and 2 bottom, respectively, are similar to the results reported previously (7, 8). The similarity of our FSD results and the previously published deconvolved spectra confirms that proteins in this study are in the same conformation as studied earlier, in other words, no changes were induced by our deuteration method or any other minor differences in the handling procedures. Therefore, the VCD results obtained and the conclusions derived from them are applicable to the previous FTIR studies.
ET
AL.
The qualitative bandshape observations described above were put on more quantitative basis using the factor analysis methodology. First, it was observed that the incorporation of EGF and bFGF VCD spectra into the reference set of 28 globular proteins did not substantially change the bandshape of the calculated subspectra. Also their incorporation did not change the number of subspectra (six) necessary for the description of all VCD spectra in the analyzed data set. The cluster analysis using the subspectral coefficients showed the connection of EGF and bFGF VCD spectra to be best clustered with those of chymotrypsinogen and casein, with the next closest protein cluster being composed of trypsin, trypsin inhibitor, rhodanese, chymotrypsin, and superoxide dismutase. This serves to support the applicability of quantitative analysis for the EGF and bFGF VCD spectra. The final multiple regression equations were similar to those found previously for smaller training sets. [In this analysis we used the coefficients of subspectra 2 and 5 for calculation of fraction of a-helix, subspectrum 2 for calculation of P-sheet content, subspectra 3 and 5 for the determination of the bend fraction and subspectra 1 and 3 for the calculation of fraction of other structural types.] The turn fractions remain undetermined. [We have shown recently that for obtaining a statistically reliable estimation of the fraction of turn, the amide I’ spectra should be supplemented by the amide II VCD (20).] The quantitative secondary structural predictions are summarized in Table II. Based on the VCD spectra obtained for EGF and bFGF and comparison of the bandshapes with the previously analyzed set of 28 proteins, we conclude that these two proteins are predominantly in P-sheet and extended (other) conformations. There is very little a-helix, if any, in agreement with all other spectral measurements except the frequency-based FTIR analysis (7, 8). Our previous extensive analysis of accuracy of the VCD method provided an estimate of error limits for evaluation of individual secondary structure types: for P-sheet fraction the absolute error should be less than 15%, and for bend and other fractions less than 10%. These error limits can be
TABLE
I
Summary of Frequencies” EGF
bFGF
IR
FSD
VCD
IR
FSD
VCD
1651
1677 1664 1654 1638
1669 (+) 1642 (-)
1643
1685 1666 1652 1643 1635
1663 (+) 1635 (-)
y In cm-‘.
RECOMBINANT TABLE
Summary
GROWTH
VCD” FTIR* UVCD’ NMRd
0 39 (?) 0
VIBRATIONAL
Results
18 34
a-helix
681
DICHROISM
ACKNOWLEDGMENTS
bFGF p-sheet
CIRCULAR
are much easier to interpret than those obtained by ECD or FSD-FTIR.
II
of Quantitative EGF
a-helix
FACTOR
P-sheet
0 21 (?)
34 42
o-5
54-72
42
a This study. * Refs. (7) and (8). ’ Refs. (3-5). d Ref. (6).
This work was supported at UIC by a grant from the National Institutes of Health (GM30147, to TAK). The cooperation between Charles University and UIC is aided by an NSF grant for Joint CzechoslovakUSA (INT 91-07588) research. The FTIR instrumentation used was provided in part by grants from the NIH, NSF, and University of Illinois.
REFERENCES 1. Gospodarowicz, D., Massoglia, S., Cheng, J., Lui, G. M., and Bohlen, P. (1985) J. Cell Physiol. 122,323-332. 2. Sporn, M. B., and Roberts, A. B. (1985) Nature 313, 745-747. T., McGinley, M. D., Rohde, M. F., and 3. Narhi, L. O., Arakawa, Westcott, K. R. (1992) Znt. J. Pept. Protein Res. 39, 182-187.
4. Arakawa,
useful for comparisons of VCD quantitative analysis with the outputs of other methods. These results are in substantive agreement with those previously obtained by ECD (3-5), considering the limitations of each technique (15). The FSD-FTIR results can also be seen to be consistent with this analysis if the peaks in the region 16501656 cm-’ are assignedto loop structures as was originally suggestedby Prestrelski and co-workers (7,8). The VCD sign patterns reported here serve to distinguish more definitely between possible interpretations of the FTIR data. CONCLUSION
VCD results only establish the predominant conformation in solution and give no direct evidence that the peaks in the FSD-FTIR spectra in the region 1650-1656 cm-l are indicative of the loop structures. We can say with certainty that these dominant FSD-FTIR bands in the spectra of the two proteins are not due to a-helical structures, since no such conformation was observed in the VCD spectra, but at this point we do not know their exact origin. It is reasonable to assign them to loops (as was done in the two previous FTIR studies) since loop structures are observed in high amounts in the X-ray (9) and NMR (6) analyses. This is further supported by the VCD spectra observed being characteristic of the extended structures (21) that may be seen in loops with only modestly constrained conformations. These results demonstrate the potential errors in interpreting conformational results based on the IR frequencies alone and show the need to use multiple spectroscopic techniques to develop a more accurate picture of protein secondary structure. Qualitatively, VCD results
T., Hsu, Y.-R., Schiffer, S. G., Tsai, L. B., Curless, G., and Fox, G. M. (1989) Rio&em. Biophys. Res. Commun. 161,335-
341. 5. Wu, C.-S. C., Thompson, Chem. 10,427-436. 6. Cooke, R. M., Wilkinson, M. J., Campbell,
S. A., and Yang,
J. T. (1991)
A. J., Baron, M., Pastore, A., Tappin, J. D., Gregory, H., and Sheard, B. (1987) Nature
327,339-341. 7. Prestrelski, S. J., Arakawa, (1991)
Arch.
J. Protein
Biochem.
T., Kenney, W. C., and Byler, Biophys. 285, 111-115.
D. M.
8. Prestrelski,
S. J., Arakawa, T., Wu, C.-S. C., O’Neal, K. D., Westcott, K. W., and Narhi, L. 0. (1992) J. Biol. Chem. 267, 319-322. 9. Zhu, X., Komiya, H., Chirino, A., Faham, S., Fox, G. M., Arakawa, T., Hsu, B. T., and Rees, D. C. (1991) Science 251,90-93. 25, 469-487. 10. Byler, D. M., and Susi, H. (1986) Biopolymers 11. Surewicz, W. K., and Mantsch, H. H. (1988) Biochim. Biophys. Acta 952,115-130. 12. Keiderling, T. A. (1990) in Practical Fourier Transform Infrared Spectroscopy (Ferraro, J. R., and Krishnan, K., Eds.), pp. 203-284, Academic Press, San Diego. 13. Pancoska, P., Yasui, S. C., and Keiderling, T. A. (1989) Biochemistry 28, 5917-5923. 14. Pancoska, P., Yasui, S. C., and Keiderling, T. A. (1991) Biochemistry
30,5089-5103. P., and Keiderling, T. A. (1991) Biochemistry 30,688515. Pancoska, 6895. S. G., Rhode, M. F., Tsai, L. B., Banks, A. R., 16. Fox, G. M., Schiffer, and Arakawa, T. (1988) J. Biol. Chem. 263, 18,452-18,458.
17. Salomaa, P., Schaleger, Sot. 86,1-7.
L. L., and Long,
F. A. (1964)
J. Am. Chem.
18. Kauppinen, J. K., Moffatt, D. J., Mantsch, H. H., and Cameron, D. G. (1981) Appl. Spectrosc. 35, 271-276. 19. Kabsch, W., and Sander, C. (1983) Biopolymers 22, 2577-2637. M., and Keiderling, T. A. (1992) 20. Pancoska, P., Gupta, V., Urbanova, to be published. 21. Dukor, R. K., and Keiderling, T. A. (1992) Biopolymers 31, 17471761.
