Int.
3.
Biochem., 1973, 4, 365-371.
CONFORMATIONAL FROM
ANALYSIS
CIRCULAR W.
H.
Department
[Scientechnica
(Publishers)
OF BOVINE
DICHROISM
BANNISTER, of Physiology
J.
V.
Ltd.]
365
ERYTHROCUPREIN
AND INFRA-RED
BANNISTER*,
and Biochemistry,
Royal
P. University
SPECTRA
CAMILLERIt, of Malta,
Msida,
Malta
AND A. Department
LEONE
of Mathematics,
Royal
GANADO University
of Malta,
Msida,
Malta
ABSTRACT I. The analysis of protein conformation from circular dichroism (C.D.) spectra by curve-fitting and matrix-rank determination is examined with reference to spectra of bovine erythrocuprein. 2. Poly-a-amino-acid reference spectra are not applicable to bovine erythrocuprein C.D. spectra in so far as they give least-square fits with negative percentages of secondary structure, in contrast to reference spectra derived from the C.D. spectra of the proteins The latter indicate the presence of a small ribonuclease, lysozvme, and myoglobin. percentage of a-helix and preponderance of unordered structure over 8-structure in the protein. 3. The presence of these structural modes and the preponderance of unordered structure is supported by the i&a-red spectrum of the holoprotein in the amide I region. 4. Matrix-rank analysis of a set of five independent C.D. spectra of bovine erythrocuprein in the far ultra-violet supports the three-component fit of the spectra by least squares. is
ERYTHROCUPREW (Carrico
and
a
Deutsch,
cupro-zinc 1970;
protein Bannister,
and Wood, x97x) with superoxide dismutase activity (McCord and Fridovich, I 969). The profile of the far ultra-violet circular dichroism (C.D.) spectrum of bovine erythrocuprein suggests that the protein contains little or no a-helix and a mixture of antiparallel p-pleated sheet and unordered structure (Wood, Dalgleish, and Bannister, 197 I; Weser, Bunnenberg, Cammack, Djerassi, FlohC, Thomas, and Voelter, x971). This view has been examined and put on a quantitative basis by a combined approach of curve-fitting and matrix-rank analysis of C.D. spectra of the protein. This kind of approach to the conformational analysis of Bannister,
* Present address: Clinical Biochemistry, OX2 6HE.
Nuffield Denartment of Radcliffe Infirmary, Oxford
t Present address: Department of Chemistry, University College of Swansea, Swansea SA2 8PP.
proteins from C.D. spectra has not been exploited so far. The results for bovine erythrocuprein are described in the present paper to indicate the potential of the method of analysis as well as its assumptions and limitations. The results are also compared with the qualitative picture of conformation given by the infra-red (I.R.) absorption spectrum of the protein in the amide I region. THEORETIC& The analysis of C.D. spectra for the number of contributing components by curve-fitting and matrix-rank methods depends on the vectorial representation of the spectra as the mean residue ellipticity at a series of wavelengths in the far ultraviolet. As a working hypothesis, the C.D. spectrum is assumed to be given by a linear combination of properly chosen spectra of the a-hehcal, 8structural, and unordered conformations (Greenfield and Fasman, 1968). In functional notation:
A(h)c = do.), (1) where A(i) is a vector function whose components are the spectra of the three conformations; c is
y36
BANNISTER
ht. 3. Biochem.
AND OTHERS
the vector representing the contributions of the smoothed by a quadratic five-point least-squares conformations: and d(i.) is the vector function approximation (Savitzkv and Golay. 1961,. representing the protein spectrum. .4ssuming A(i.), Iterated interpolation {Neville, tg34! was used to c may be obtained by a continuous least-squares construct spectra digitized at intervals of I nm. approximation ( LIagar, I 968). Alternatively, a Curve-fitting was performed on these spectra. collocation method at equal intervals can be used Matrix-rank analysis was performed on the basic on equation (I) to obtain the discrete approximaexperimental spectra digitized at intervals of tion : 2’j nm. C.D. spectra of poly-L-lysine in a-helical, 8c = (AA)-‘A’d, (2) structural, and unordered conformations were where A and d are obtained by sampling A()<) and constructed from the data reported by Greenfield d(i), and A’ is the transpose of A (Searle, x966). and Fasman (1969). Reference spectra of the aGood results are obtained when the spectra helix, ,%tructure, and unordered structure, chosen for the matrix A truly form a basis for the obtainable from the C.D. spectra of ribonuclease, protein spectrum. The spectra of poly-~-aminolysozyme, and myoglobin as proposed by Saxena acids in the a-helical, p-structural, and unordered and Wetlaufer (x971), were calculated from the conformations give good approximations in certain data reported by these authors. The spectra of the cases (Greenfield and Fasman, rg6g; Rosenkranz reference conformations proposed by Rosenkranx and Scholtan, 1971). The difficulties in the use and Scholtan ( I g7 I )-the standard a-helical of these homopolymer spectra have been pointed conformation of poly-r&sine, the 8-structural out by Greenfield and Fasman (1969). As an conformation of poly-r-lysme in the presence of alternative, Saxena and Wetlaufer (1971) have I per cent SDS at neutral pH, and the unordered proposed the use of a matrix A obtained from the conformation of poly-L-serine in the presence of relationship : 8 ilf LiCl-were kmdly supplied by Dr. W. AC = D, Scholtan. (3) The infra-red absorption of a deposited protein where C is the matrix composed of the vectors film prepared according to the procedure of representing the a-helical, ~-structural, and unTimasheff and Susi (1966) was measured with a ordered structural content of three reference UnicamSP~oo gratingspectrophotometeroperated proteins as revealed by X-ray crystallography, at a mean speed of approximately 4 cm.-1 per and D is the matrix of the C.D. spectra of the second. The wavenumber scale of the instrument This assumes that the spectra of the proteins. was calibrated by means of a polystyrene film. reference proteins have the same basis set and The resolution in the amide I region was apthat the matrix A which is obtained is a proximateh 1.; cm.-i and the wavenumber reasonable approximation to the basis set of the ;eproducibility approximately 0.5 cm.-‘. .4 despectra of the proteins on which curve-fitting is mountable cell with CaF, windows was used. The carried out. spectra of Nujol mull, -fluorocarbon film, and For a set of protein C.D. spectra having the same KBr disk preparations of the protein were also basis the rank of the matrix of the spectra gives the measured. number of vectors forming the basis. When the The I.R. spectrum between 16oo and matrix is composed of the spectra of a number of I 700 cm.-i was read at intervals of 0.5 cm.-’ by different proteins it may be difficult to attach a means of a mechanical co-ordinate digitizer. The general meaning to the rank of the matrix. ,4 data were smoothed by a quadratic five-point rank of four (or higher) in these cases (Dalgleish, least-sauares annroximation and the centre of x972) does not necessarily disprove the threeabsorption bands was confirmed by numerical component representation of protein C.D. spectra differentiation (Savitzky and Golay, 1964). in the far ultra-violet. Although a rank greater All computations were performed on a Hewlettthan three is conceivable in view of the expectation Packard giooB Calculator fitted with a g1or.4 of contributions from chromophores other than Extended Memory. amides in certain cases (Greenfield and Fasman,
1969; Magar, 197x), it is desirable to demonstrate this for a matrix of spectra of a single protein under conditions which alter the proportions of the structural modes, since it would then be safer to assume that the spectra have the same basis set. METHODS
The C.D. spectra analysed here have been reported previously (Wood and others, I 971I. For the purposes of the present work the spectra were read at intervals of 2.5 nm. ?\lolar C.D. was converted to mean residue ellipticity assuming a mean residue weight of I IO, and the data were
RESULTS The C.D. spectra of bovine erythrocuprein analysed in the present work are given
in
Table I.
of
The
rcsuhb
of fitting
the
spectra
the reference conformations proposed by Rosenkranz and Schoitan ( I 97 I j, Saxena and 1Vetlaufer ( I 97 I 1: and Greenfield and Fasman
i 1969; to the spectra apoprotein, apoprotein
and between
the
of the
holoprotein,
the
S-carboxymerhylated 200 nm.
and
240 nm.
CONFORMATIONAL
‘9735 4
ANALYSIS
given in Table II. It is seen that continuous least-squares approximation gave I.-CIRCULAR
WAVELENGTH (nm.j
_
DICHROISM SPECTRA OF BOVINE ER~THROCUPREIX PREPARATIONS MEAN RESIDUE ELLIPTICIN* (degree cm* per dmole)
-
S-Carboxymethylated Xpoprotein
Holoprotein
- 2904 -4432 -54’3 -j856 -j8Io
202’j 205 207'j
210 212’j 21j 217.j 220 222’j 22j 227’j 230 232’j
-5492 -4931 -4120 -3180 -2364 - I 800 - 1426 -1138 861
235 237’5 240
-
* Experiment
--
__
__
.a1 values
604 410 300
367
virtually the same percentages of r-helix, and unordered structure as $-structure,
are
Table
OF ERYTHROCUPREIN
-
- ‘994
-3822 -5192 -5962 -6154 -5980 - j486 -4771 -3951 -3187 -2523 -2013 - 1627 -1301 z
:F$
-
459
-
1402
- ‘235 -1051 868
smooth ed by quadratic
15,
-
703 573 475 399 3j* 312 286
-
2j9 231 181 132
-
89 59
Apoprotein in 8M Urea
Holoprotein in 8M Urea
-4603 -3695
-4385 -4888 -4938 -4661
-2922
-2296 - 1820 -1438 - 1167 - 989 - 889 - 841 - 796 - 6go - 540 - 407 - 301
-4’05 -333’ -2516 - 1878 1480 -1220 - 1001 766
-
t-point least-squares
568 403 302
approxima
n.
Table II.-STRUCTURAL CONTENT OF BOVINE ERYTHROCUP~UEINPREPARATIONS AS ESTIMATEDBY FITT~G THE REFERENCE SPECTRA PROPOSEDBY ROSENKRANZ AND SCHOLTAI ( I g7 I ) (d) t SAXENA ANY LVETLAUFER (1971) ;B>, AND GREENFIELD AND FANAN (1969, (CJ TO THE C.D. SPECTRUM BETWEEN 200 nm. AND
240 nm.
/ REFERENCE SPECTRA
I
%-HELIX
S-STRUCTURE
(per centj
(per cent)
/ UNORDERED STRKXXURE (per cent)
RMSt
Holoprotein ;: C
.4poprotein
A CB
o-2 (o-3) IO (10.6) 17.3 (12.8)
22’1
(22.2)
31.8 28.7
(31.5) (32.3)
1’7 (1.2) 12.j (12.4) 18.8 (14.6)
24’3 (24.3) 31’9 (32.2) 3 * ‘: (34.6)
74’1 (74.j) jj’j (jj.4) 49’5 (jo.7)
j94 670 IOIj
SCarboxymethylated apoprotein ;: c
2.5 (2.6’1 Ij.9 (16’1 17.3 (16)
3.7* (4-o*) (18.5)
18.2 ‘3
( ‘4.3)
* Segative with respect to least squares. t Root mean square of residuals for fitted spectrum. The results of discrete and continuous least-squares approximation
93.8 (93.4) 65.9 (65.5) 69.7
(69.6)
30
281 86
are given, the latter in parentheses.
BANNISTER AND OTHERS
368
approximation. The reference discrete spectra of Rosenkranz and Scholtan ( I 97 I j indicated virtually no z-helix, those of Saxena and Wetlaufer (1971) and Greenfield and Fasman (1969) indicated a small percentage Table
of r-helix. All reference spectra indicated a preponderance of unordered structure over $-structure and further increase in unordered structure in the A’-carboxyrnethylated apoprotein.
ANALYSIS OF CD. SPECTRA OF Table Z BETWEEN 205 PROCEDURE OF WALLACE AND KATZ (1964)
ZZZ.--MATRIX-RANK
-6rj4
Reduced data matrix*
0 0 0 0
Reduced 5 per cent error
308 0 0 0 0
matrix*
*
Int. J. Biochem.
Original
data transposed:
-jg62 -2138
o 0
0 0 0
0 0 submatrix
AND
240 nm.
BY THE
- j486 785
0
217 -119 28
299 271 4% 643
274 243 461 681
0
8j
-156
260 307 262
298
314
leading
-jg8o j43 82
-j792 -864 626
0 0 0
nm.
after four reduction
steps.
Table IV.-STRUCTURAL CONTENT OF BOVINE ERYTHROCUPREIN PREPARATIONSAs ESTIMATED BY FITTING THE REFERENCE SPECTRA PROPOSEDBY ROSENKRANZ AND SCHOLTAN ( Ig7 Ij (A), SAXENA AND WETLAUFER (1g7I), (B), AND GREENPIELD AND FA~MAN (1969) (C) TO THE C.D. SPECTRUM BETWEEN 205 nm. AND 240 nm.
REFERENCE
SPECTRA
8-STRUCTURI (per cent)
I-HELIX (per cent)
__
UNORDERED sTxucrURz (per cent)
RMSt
Holoprotein d
I’9 9’4 3.2*
CB
__
18.3 32’9 38.4
80 j7’7 j8.3
328 506 822
20’7 32’4 40.2
78.6 jj’b j8.1
3jg jj 736
Apoprotein .4
0.6* 12’1 1’7*
: SCarboxymethylated
__
apoprotein
A
j.8*
3-o Ij'I
CB Holoprotein
3’7
27
62.2 68.4
42 66
18.0
80.2 j8.I j8.4
334 4jI 698
92’4
155
in 8 M urea I.8* IO
: c .4poprotein
__
91'1
22.6 27’9
in 8
M A CB
3.3*
_.
$3
urea I.6 ‘3’9 I .6
6* 22.3 28.5
* Negative with respect to least squares. f Root mean square of residuals for fitted spectrum. The results of discrete least-squares approximation are given.
63.8 69.7
‘94 325
‘973>4
CONFORMATIONAL
ANALYSIS
Rank analysis of the matrix given by the spectra of Table I between 205 nm. and 240 nm. indicated a three-component fit of the spectra. The procedure of Wallace and Katz (I 964) was followed, a 5 per cent error matrix being first set up as a reasonable
OF ERYTHROCUPREIN
369
by matrix-rank analysis was best corroborated by the fit given by the reference spectra of Saxena and Wetlaufer ( 197 I). These were the only reference spectra which did not give negative percentages of secondary structure in the best fit of the experimental
-6000 200
220 nm
240
FIG. I.-Far ultra-violet C.D. spectrum of bovine erythrocuprein (-) and least-square fits (. . .) given by the reference spectra of Saxena and Wetlaufer (1971) in the wavelength regions 200-240 nm. and 205-240 nm.
estimate. The results of the four reduction steps possible with the original matrix and the corresponding propagated errors are given in Table III. The elements of the principal diagonals of the reduced and the error matrices indicate three non-zero rows in the reduced matrix. A similar finding was made with a IO per cent error matrix. The results of fitting the five spectra of Table I between 205 nm. and 240 nm. by the reference spectra of Rosenkranz and Scholtan (1971), Saxena and LVetlaufer (1971), and Greenfield and Fasman (I 969) are shown in Table IV. It is seen that the three-component
fit of the
spectra
indicated
1600
1650 cm-l
1700
FIG. I.-I.R. spectrum of a deposited film of bovine erytbrocuprein in the amide I frequency region.
by least squares. The fit of the holoprotein spectrum by the reference spectra of Saxena and Wetlaufer ( 197 I) is shown in Fig. I. The presence of r-helix was supported by the I.R. spectrum of the holoprotein, whether a deposited film, a Nujol mull, a fluorocarbon film, or a KBr disk preparation was examined. The deposited film spectrum in the 1600r 700 cm.-’ region is shown in Fig. 2. The spectra
BANNISTER AND OTHERS
370 observed cm.-I
amide and
Relevant
I band has a peak at 1641
considerable
to the analysis
fine
structure.
of protein
conforma-
at 1632, 1654, and 1685 cm.-‘. The peak at 1641 cm.-’ indicates a predominance of unordered structure; the band at 1632 cm.-1 indicates the presence of p-structure; and the smaller band at 1685 cm.-l suggests that it is the antiparallel P-structure which is present. The maximum at I 654 cm.-l is in the region of the amide I frequency thought to be characteristic of the a-helix (Susi, Timasheff, and Stevens, I 967). tion
are
the partially
resolved
bands
make a significant difference to the computations given here. For example, discrete least-squares fitting of the erythrocuprein spectrum in the stringent 200-240 nm. region, using basis spectra calculated from the revised C.D. spectra ofribonuclease, lysozyme, and myoglobin given by Chen, Yang, and Martinez (I 972) and the values for the X-ray structural composition of these proteins given by Saxena and Wetlaufer ( 197 I ,,. indicates 10.4 per cent a-helix, 31.2 per cent P-structure, and 58.4 per cent unordered structure and gives a value of 350 for the root mean square of the residuals. ACKNOWLEDGEMENT
DISCUSSION The present analysis indicates three structural modes for bovine erythrocuprein: a-helix, $-structure, and unordered structure. This conclusion was reached by curvefitting and matrix-rank analysis of far ultraviolet C.D. spectra of the protein and examination of the I.R. spectrum in the amide I region. It is of interest that the reference spectra of Saxena and Wetlaufer (I 97 I) derived from three reference proteins did not give negative percentages of secondary structure in least-squares fitting of the experimental CD. spectra in contrast to the poly-aamino-acid reference spectra of Rosenkranz and Scholtan ( 1971) and Greenfield and Fasman ( 1969). In principle, negative percentages of secondary structure should not be found if the reference spectra truly form a basis for the protein spectrum and equation !I) is satisfied. Magar (1971 j has suggested hnear programming instead of least-squares curve-fitting where negative percentages occur. However, it would be better in these cases first to inquire into the &ability of the reference spectra. Probably bovine erythrocuprein and the reference proteins of Saxena and Wetlaufer (197x ), i.e., ribonuclease, lysozyme, and myoglobin, have the same or a closely similar three-component basis for their far ultra-violet C.D. spectra. NOTE ADDED AT PROOF STAGE: Chen, and Martinez (1972) inadvertent error in
have
Saxena and Wetlaufer
(197 I).
the
ht. J. Biochem.
Yang,
pointed out an C.D. spectra of This does not
One of us (W. H. EL) thanks the Nufficld Foundation and the Wellcome Trust for research grants.
REFEREXCES BANNISTER,J., BANNISTER,\V., and WOOD, E. (ICJ~I), ‘ Bovine erythrocyte cupro-zinc protein. I. Isolation and general characterization ‘, Eur. j. Biochem., 18, 178-186. CARRICO. R. J., and DEUTSCH,H. F. ( Ig7o), ‘ The presence of zinc in human ytocuprein and some properties of the apoprotem ‘, 3. biol. Chem., y159 723-727. CHEN, Y-H., YANG*J. T., and LMARTINEZ, H. &I.
(x972), ‘ Determination of the secondary structures of proteins by circular dichroism and optical rotatory dispersion ‘, Biochemistry, II, 4’2~4131. DALGLEISH. D. G. i I 072,. .,, ,. ‘ The analvsis of the farultraviokt circular dichroism ’ spectra of proteins ‘, FEBS Letters, q, I 34-13 j. GREENFIELD, R’. J., and FASMAN, G. D. i 1969,. ‘Computed circular dichroism spectra for the evaluation of protein conformation ‘, BioChckStTy,8, 4108-4116. MCCORD, J. M., and FRIDOVICH, I. (Ig6g,;, ‘ Superoxide dismutase. An enzymic function for erythrocuprcin (hemocuprem) ‘, 3. biol.
Gem., *
6049-605j.
MACAR, M. E. (rg6f), ‘ On the analvsis of the optical rotatory dispersion of proteins ‘, Biochemish-/, 7, 6 17-620. MAGAR, M. E. ( 197 I ), ‘ On the possibility of determining the secondary structure of proteins in solution ‘, 3. theor. Biol., 33, 1oj-11g. SEVILLE, E. H. (Ig34), ‘ Iterative interpolation ‘, 3. Indian math. SOL., no, 87-120. ROSENKRANZ. H., and SCHOLTAN, W. (Ig7I,, ‘ Eine Verbesserte Methode zur Konformationsbestimmung van Helicalen Proteinen aus Messungen des Circulardichroismus ‘, Hoppe-
St$er’s <. physiol. Gem., 352, 896-904.
‘9731
4
CONFORMATIONAL ANALYSIS OF ERYTHROCUPREIN
SAVITZKY, A., and GOLAY, M. J. E. i ig64), ’ Smoothing and differentiation of data by simohfied least scmares nrocedures ‘. .&oivf. 4 C/r&., 36, x627-1639. * SAXENA. V. P.. and WETLAUFER. D. B. f 107I ). ‘ .q new basis for interpreting the c~ir&&r dichroism snectra of oroteins ‘. Proc. natn. Acad. Sci. C:S.A.,*68, g6g&. ’ SEARLE. S. R. (1466). ‘ Matrix Alecbra for the Biolo&al &ien&~e9, ‘pp. 23x-233. “New”York: Wiley. !%.I, H., TIMASHEFF, S. N., and STEVENS, L. ( I 967)) ’ Infrared spectra and protein conformations in aqueous solutions. I. The amide I band in He0 and DsO solutions ‘, J. biol. Gem., ZZ~Z,j46wj466. TIMASHEFF, S. N., and Susr, H. (rg66), ‘ Infrared investigation of the secondary structure of 3lactoglobuhns ‘, 3. biol. Chem., ~41, 24g-250.
371
WALLACE, R. M., and KATZ, S. M. (x964). ‘ A method for the determination of rank in the anaivsis of absorntion soectra of muiticomnonent syste’ms ‘, 3. phy;. Chcm:, 68, 3890-3892. ’ BUNNENBERG. E.. CAIHMACK. R.. WESER. U.. DJE~I,’ C., FLOHI?, i., THOMAS, G.; and VOELTER, W. (t971), ‘ A study on purified bovine erythrocuprein ‘, Biochim. biophys. Acta, zrq3, 203-2 13. WOOD, E., DALGLEISH, D., and BANNISTER, W. (197 I ), ‘ Bovine erythrocyte cupro-zinc protein. 2. Physicochemical properties and circular dichroism ‘, Eur. 3. Biochcm., 18, 287-193.
Kq Word Index: Protein conformation, circular dichroism spectra, basis spectra, curve fitting, matrix rank analysis, infra-red spectra, amide I band, bovine erythrocuprein.