ARCHIVES

OF BIOCHEMISTRY

AND

BIOPHYSICS

Vol. 277, No. 2, March, pp. 324-333,199O

Determination of Tryptophan, Tyrosine, and Phenylalanine by Second Derivative Spectrophotometry’ Yasuhiko

Nozaki

Department

of Biochemistry,

Duke University

Medical Center, Durham,

Received July 24,1989, and in revised form November

6,1989

Second derivative spectrophotometry has been useful for the determination of aromatic amino acids. However, published methods produce erroneous results, because those methods measure second derivative values by the vertical distance between peak and trough which is subject to variation according to the aromatic amino acid composition of proteins. This paper presents a method of second derivative spectrophotometry which measures second derivative absorbance values by means of the vertical distance from baseline to the derivative curve at a wavelength specifically assigned to each aromatic amino acid, and makes corrections for the interference from other amino acids at the same wavelength. The Appendix describes a computational method for obtaining absolute values of second derivative absorbances directly from normal absorbance values without using the spectrophotometer’s derivative mode, because most commercial instruments produce completely arbitrary second derivative values which make comparison of data obtained on two different instruments impossible. o i99o Academic PWS, 1~.

Spectrophotometric determination of tryptophan content of proteins (1) have been found to be difficult in certain cases where interfering absorptions exist. These interferences are due to inherent components, impurities, or turbidity, or to a combination of these factors (2). They tend to increase absorbances, leading to an overestimation of tryptophan residues. Several methods based on second derivative spectrophotometry have been proposed for measuring aromatic amino acids (3-9). When I tried those methods, however, none of them worked satisfactorily. The problem was most acute when the number of tryptophan residues was small compared with that of tyrosine residues, e.g., for

1 This work was supported Public Health Service.

by the Grant GM30900

North Carolina 27710

from the U.S.

BSA’ (2 tryptophans, 19 tyrosines) and HSA (1 tryptophan, 18 tyrosines). The flaws in the published methods include: (i) Use of the vertical distance between adjacent peak and trough as a measure of second derivative absorbance. Peaks, particularly those located near the long wavelength end of a spectrum are not reproducible. For instance, the peak exhibited by tryptophan at 295 nm that has been selected by previous investigators as a reference point for the vertical distance to represent tryptophan is not well-defined, as can be seen in Fig. la. Besides, this mode of measurement is fundamentally wrong, as will be shown below. (ii) Interference from other aromatic amino acids. It has been stated that the second derivative spectra of aromatic amino acids show no contribution in certain wavelength ranges, so that other amino acids can be determined without interference within those ranges (3, 6). However, this sort of no-contribution region exists only for phenylalanine above 280 nm. Both tryptophan and tyrosine show nonzero values of second derivative absorbances in the entire spectral range below 300 nm, and therefore are mutually interfering, as pointed out by Levine and Federici (10). It, therefore, is necessary to make corrections for the interfering amino acids. There have been several proposals (3, 4, 6, 8, 9) to overcome this difficulty, which have been found wanting in one way or another. Despite these flaws, second derivative spectrophotometry has the potential of being an important addition to the methods for determination of the three aromatic amino acids. For example, it is unaffected by gradually changing absorbances, as pointed out by Balestrieri et * Abbreviations used: BSA, bovine serum albumin; HSA, human serum albumin; AcTrpNH,, N-acetyl-L-tryptophanamide; AcTyrNH,, N-acetyl-L-tyrosine amide; AcPheOEt, N-acetyl-L-phenylalanine ethylester; RNase Tl, ribonuclease Tl; CLSO, chicken liver sulfite oxidase; GuHCl, guanidine hydrochloride; AA, amino acid; HWHH, halfwidth at halfheight; SDS, sodium dodecyl sulfate.

324 All

0003-9861/90 $3.00 Copyright 8 1990 by Academic Press, Inc. rights of reproduction in any form reserved.

AMINO

ACID

ASSAY

BY DERIVATIVE

al. (6), such as those due to turbidity and broad bands such as those of cystines. However, from the findings (flaws 1 and 2) shown above, it has become clear that, in order to develop a reliable method, principles underlying second derivative spectrophotometry have to be scrutinized, and a new approach to the mode of measurements has been made. In this paper, a method is presented that utilizes the vertical distance from the baseline to the second derivative spectral curve, at a properly chosen wavelength assigned to each aromatic amino acid residue. The method is primarily aimed at the determination of tryptophan residues in proteins, but, in principle, it can also be applied to the other two aromatic amino acids. Sample proteins have to be in random-coiled form in which all chromophoric side chains are exposed to solvent. Most experiments are, therefore, carried out in 6 M GuHCl. In the Appendix, a computational method of obtaining absolute values of second derivatives directly from normal absorbance values is described, because the derivative values given by most commercial spectrophotometers are on arbitrary scales. These computations allow the comparison of derivative values obtained on different spectrophotometers, and lead to the possible establishment of standard molar second derivative absorptivity. EXPERIMENTAL

Material N-AcTrpNH, (mp 198-9°C) and N-AcPheOEt (mp 91°C) were purchased from the Sigma Chemical Co. and used without further purification. Unlike N-AcTyrNH, (see below), they have not shown any spectral changes within two years after purchase. The molar absorptivities of N-AcTrpNH, were slightly lower than the literature values (l), but recrystallization from methanol did not change the value. N-AcTyrNH, was also purchased from the Sigma Chemical Co. and used without further purification. However, after two years’ storage at -20°C even in a tightly sealed container, the molar second derivative absorptivity was found to have decreased. The original values were restored, however, following two recrystallizations from methanol. Melting points were obscured by a distillate condensing on the underside of the cover glass placed over the crystals, yet the process of purification could be followed by observing the narrowing of melting range. The final melting range was 233-234°C. BSA and HSA were purchased from the Sigma Chemical Co. All other proteins were gifts, and their sources are shown in the Acknowledgments. All proteins were shown to be homogenous by SDS polyacrylamide gel electrophoresis, with the exception of CLSO. Purity of CLSO was assured by HPLC. K2Cr,07, the National Bureau of Standards’ Standard No. 136 was used to prepare alkaline K&TO4 for calibrating spectrophotometers. GuHCl, Ultra Pure, was purchased from Schwarz/Mann. Na phosphate buffer (0.02 M; NaPi), pH 6.8, was prepared by dissolving 1.065 g of Na2HP0, and 1.725 g of NaH2P0,. lH,O to make a l-liter solution. Both salts were purchased from Mallinckrodt. Protein stock solutions were prepared in 0.02 M NaPi, pH 6.8, and their concentrations were determined by the dry weight method after exhaustive dialyses (2). Amino acid stock solutions were prepared by weighing out the crystals on the Cahn 28 microbalance and making solutions in 0.02 M Nai’;, pH 6.8. Solutions for spectral analyses were prepared by adding equal weight of GuHCl to a stock solution to make it approximately 6 M in GuHCl (11): The solvent was prepared from 0.02 M Nap,, and its density was determined by pycnometry to be 1.142

SPECTROPHOTOMETRY

325

g/ml at room temperature, 23.5-23.6”C, resulting in 5.969 M GuHCl and approximately 0.0115 M NaPi. Protein- and amino acid-containing solutions were prepared in the same manner, and they were assumed to have the same GuHCl and NaPi concentrations as the solvent. Although the solutions were prepared by weighing, all concentrations were expressed in molarity, using the density value shown above.

Equipment

and Methods

Optical cells, 4 X 10 mm, with tapered Teflon stoppers were type QS from Hellma. One Shimadzu UV260 was used most frequently and three others were used for comparison and they all showed virtually identical performance. Shimadzu UV3000 and UV160 and Perkin-Elmer Lambda 9 were also used for comparison. The spectrum of NAcTyrNH, in 0.02 M NaPi, pH 6.8, was obtained using the 14DS spectrophotometer (Aviv Associates, Lakewood, NJ). Wavelength was calibrated with a holmium oxide filter (Shimadzu) and approximately 0.03 M benzene solution in heptane. Absorbance at 273 nm was calibrated with 2.06 X 10m4M K,CrO, in 0.05 M KOH (0.760A) described by Haupt (12). In addition, it was noted that, when K2Cr04 concentration was varied from 1.4 to 5.238 X 10m4 M, the plot of absorbance, approximately 0.51-1.934, against concentration was linear on either the Cary 15 or the Shimadzu UV260, and thus obeyed the Beer’s law. This allows the utility of this solution above and below the originally assigned absorbance value.

Protocol for Spectrophotometric Determination of Aromatic Amino Acids Using Shimadzu UV2603 1. Establishment of the molar second derivative absorptivity: [(A’c/ Ax’) of N-acetyl amino acid amide or ethyl ester]. The automatic baseline correction mode is used routinely, and other settings are slit 0.3 nm, scan speed (chart paper) slow, and wavelength scan 2.5 nm/ cm. A titration was performed either by adding increments of the amino acid (AA) titrant to the solvent (GuHCl-NaPi) (titrating up), or the solvent to an AA solution in 6 M GuHCl-0.01 M NaPi (titrating down). In the former case, an aliquot of GuHCl-NaPi solution is weighed in the sample cell and baseline is corrected and recorded (once in normal mode; twice in derivative mode). An increment of AA-titrant is weighed in the cell, mixed by several inversions,’ and spectra are recorded, and so on. In the case of titrating down, AA-solution is titrated with the solvent in a similar manner. The concentration ranges were: AcTrpNH, 1.27 X 10m5-2.52 X 10e4 M, AcTyrNHz 1.178.50 X 10e4 M (range wider than shown in Fig. 3b), AcPheOEt 2.10 X 10m4-2.30 X 10m3M. Concentrations were chosen such that none of the absorbances exceeded 1.5A. Measurements on a chart paper are made using a precision ruler (K + E) instead of relying on the printed scale. The plot of second derivative values against concentration reveals a linear relationship, and the molar second derivative absorptivity (A”t/AX’) is obtained as the slope of this line. The 2. Determination of aromatic amino acid residues inproteins. titration of a protein solution is similar to the titration of the solvent with the amino acid. However, in the titration of a protein, the value of the second derivative represents the sum of contributions from the major component, e.g., (A’A/AX’)r,, and those of interfering ones, e.g., (A2A/Ah2)r,,. In the example of titration with tryptophan at 290

3 This instrument produced derivative values far greater in magnitude than the correct ones, and all values, therefore, should be regarded as arbitrary. However, the values are consistent as long as they are obtained with a single instrument or among different instruments of the same model, including a new version UV265. 4 The high viscosity of the 6 M GuHCl solution required extra effort to ensure thorough mixing. Stirring with a small bar and a magnetic stirrer were shown to be inefficient.

326

YASUHIKO

NOZAKI

b

Wavelength

(nm)

, I 24C

Wavelength

I

250

ti

,

scale. (a) AcTrpNH,, 1.10. (c) AcPheOEt,

I

270

260 Wavelength

FIG. 1. Recordings

ti

(nm)

(n m)

of second derivative spectra in 6 M GuHCl-0.011 M Na phosphate, pH 6.8. (A’A/AX’)‘s are presented on an arbitrary 3.82-, 4.5%,5.71-, and 7.60 X 10e4 M; A,,, 0.561.20., 1.43-, 1.77-, 2.15, and 2.51 X 10m4M; A 2800.67-1.41. (b) AcTyrNHz, 0.65-, 0.92., 1.06-, 1.33., and 1.78 X 10m3M; Azs80.13-0.35.

nm, the contribution from tyrosine is eliminated as follows: For the first point, (A*A/AX*)r, = (AxA/AX*),,,,, - (A’A/AX*),,. The last term is obtained by multiplying the molar second derivative absorptivity, (A2dAX2h,r, by the number of tyrosine residues in the protein, N&, and by protein concentration, Ce. For the second point and thereafter, volume increase has to be taken into account: (A’A/AX’),, = (A’A/ X N& X C,; (A*A/Ah*)r, then is corrected by A~*hots, - (A2dAx”),, multiplying by the volume ratio to normalize protein concentration to the original value. The (A*A/AX*) values corrected for volume change, as shown above, are plotted against the ratios (NJ of the concentration of added ) to that of protein (Ce). A linear extrapolation amino acid (CAA.added

intersects with the NAA axis at the point where (AzA/AX’) = 0, and the difference in NAA at this point and the point where NAA = CAA.~~M = 0 corresponds to the number of residues in the protein (N&j. (Figs. 4a, 4b, and 4c.)

RESULTS

AND

DISCUSSION

Baseline Figures la, lb, and lc show the variation of baseline which does not necessarily coincide with the machine

AMINO

ACID

ASSAY

BY DERIVATIVE

zero line. Since the baseline represents spectral zero, this variation makes the method of measuring vertical distances from the machine zero awkward. Instability

327

SPECTROPHOTOMETRY

0.25

of Peaks

Fig. la shows the apparent instability or lack of reproducibility of the peak of tryptophan at about 295 nm. This makes the use of this peak as a reference point inadvisable. On the other hand, the trough at 290 nm appeared stable and reproducible. The same feature was also observed in Fig. lb for tyrosine, i.e., the peak at about 289 nm was variable, while the trough at 283 nm appeared stable. A similar instability of the first peak of phenylalanine was also seen in Fig. lc. While the cause of variability of those peaks is not well understood, the fact remains that peaks in second derivative absorption spectra, particularly those at the long wavelength end, varied in both their magnitude and position. The apparent instability could be decreased by choosing a wider slit, larger Ax, and other means, but with a loss of sensitivity and fidelity. For these reasons, peaks were not suitable reference points. Aside from these practical considerations, the underlying principle of additivity will break down, if peak-to-trough distances are used, as will be discussed below.

%

: a

0.00

“a

-14=T4 Isosbestic Points Fig. la also displays apparent isosbestic points which correspond to inflection points on the original absorption curve. This was the typical pattern exhibited by a single-component system. Actually, in Fig. la, AcTrpNHz was the only light-absorbing component. Likewise AcTyrNHz and AcPheOEt were the sole components in Figs. lb and lc, respectively. Even when concentrations were changed, no changes in the wavelength of inflection point resulted. This explains why isosbestic points are seen on the baseline and suggests the utility of the baseline as a reference for the measurement of second derivative values. If the zero-crossing wavelength is shifted, as an amino acid titrant is added, and if the isosbestic points are located off the baseline, then the involvement of more than one component is indicated, which is the usual situation seen in titrating a protein (Fig. 2). From the findings described above, the titration is shown to be useful in identifying a trough which serves as a measure of a particular amino acid, and to locate the wavelength at which the vertical distance is to be measured. As more titrant is added, the influence of interfering species diminishes and the last few curves normally show a constant trough position (Fig. 2). Stability

of Troughs and Correction for Interference

Unlike peaks, the troughs, at 290 nm for tryptophan (Fig. la), at 283 nm for tyrosine (Fig. lb), and at 264.3-

-0.25L 280

290

300

Wavelength (nm) FIG. 2. Second derivative spectra of human serum albumin titrated with N-acetyltryptophanamide in 6 M GuHCl. Two modes of measuring vertical distances for tryptophan determination are shown: Conventional mode shown as T,-P, and new mode at 290 nm shown as BL-I,, -Ii, . .-Z4 (subscripts indicate the number of titrations). As more tryptophan was added, the relative contribution from tyrosine diminished, and the trough position changed from the initial location (To), approaching 290 nm, which represented the tryptophan contribution (Fig. la).

264.5 nm for phenylalanine (Fig. lc) are stable and reproducible, suggesting the advantage of measuring second derivatives by the vertical distance from the baseline to the minimum of a trough. But troughs were not free from mutual interference. Therefore, a correction for the contribution from an interfering species was made to obtain the correct value for the major amino acid. However, the correction could not be made properly by using the peak-to-trough distance method, because the positions of peak and trough change according to the varying composition. For instance, the tyrosine peak at 289 nm (Fig. lb) affected the trough of tryptophan, which was normally located at 290 nm, such that

328

YASUHIKO

NOZAKI

0.5

E 0.4 0.4 -

5 P4 0.3 ‘;

I

b

E

C

:

“x 9 0.2 4 “a

G

N; 0.2< a “a

’ 0.1

0.c CTvr Ix 10-4M) 05

1

I

d E 0.4 4 $ ‘D 0.3 2-c < 0.2 Q “a ’ 0.1

10 CTvr (x 10-4M) s are presented on an arbitrary scale. FIG. 3. Linear plots of second derivative absorbances in 6 M GuHCl-0.011 M NaPi, pH 6.8. (A’A/AX’)’ (a) AcTrpNHp at 290 nm, five sets of experiments, slope -1.95 X lOi cmm3 Mm’. (b) AcTyrNH, at 283.1 nm, four sets of experiments, slope -4.95 X lOi crnm3M-l. (c) AcTyrNH, at 290.0 nm, three sets of experiments, slope +2.7 X 1O’a crne3 Mm’. (d) AcPheOEt at 264.3-264.4 nm, four sets of experiments, slope -1.95 X IO’a cmm3 Mm’.

the trough position shifted toward longer wavelength as the tyrosine content increased. In extreme cases such as of HSA (containing 1 tryptophan and 18 tyrosine residues), the trough became very shallow and its minimum shifted by more than 1 nm, as indicated by To in Fig. 2. As a consequence, when the trough shifted from 290 nm, it was no longer a reliable feature for measuring tryptophan. On the other hand, such an effect of interference

could be corrected, if the residue number in the protein and the molar derivative absorptivity of the interfering amino acid were known. But such a correction had to be made at a fixed wavelength, so that the rule of additivity applied. The logical choice of such a wavelength for tryptophan was 290 nm, as can easily be seen in Fig. la. The basis for the correction for the effect of the tyrosine contribution on the tryptophan content is that the

AMINO

ACID

ASSAY

BY DERIVATIVE TABLE

329

SPECTROPHOTOMETRY

I

Molar Second Derivative Absorptivities in 6 M GuHCl” Major trough Amino acid AcTrpNHs AcTyrNH, AcPheOEt

Contributions

( Azc/AX2)

X (nm)

( A2c/AX2)

-1950 -495 -195

290.0 283.1 264.3-264.5

-760 +270 0

at other wavelengths ( A’e/Ax’)

X bm)

A bm)

~60 +40

283.1 290.0 above280nm

264.3-264.5 264.3-264.5

’ 1Or4cm -3 Mu ’ on Shimadzu UV260.

total value of the second derivative absorbance is the algebraic sum of contributions from tryptophan and tyrosine, but this is true only at the same wavelength, in the present case, at 290 nm. If peak-to-trough distances are used, with concomitant changes of composition causing the change of trough and peak positions, the principle of additivity will no longer be applicable, because, e.g., the molar derivative absorptivities will change. In this regard, changes in the spectral positions of the peaks and troughs are hidden by the choice of a compressed wavelength scan (rim/cm). The almost exclusive use of the vertical distance between a peak and its adjacent trough by previous authors may have originated in the adoption

TABLE

of this measure by O’Haver and Green (13) in analyzing mathematically constructed spectra. While the results of changing peak and trough positions were shown to be important factors in the present report, they were not important in the treatment of mathematical models with fixed compositions. Measurement

of Second Derivative

Absorbances

The vertical distance from the baseline to the second derivative curve at 290 nm, therefore, is adopted as the measure of tryptophan. These reference points are the least ambiguous in the second derivative spectral curve.

II

Molar Absorptivities of Three Aromatic Amino Acids Dilute salt solution Amino acid Trp

AcTrpOEt

t

6 M GuHCl

A (nm)

Author

5550 5550 5600

278 max 279.9 279.8

B&H” S&H’ G’

53908

280

Nf

1340 1390 1420 1420

274.5 274.6 274.5 274.6

B&H B&Kh S&H G

1390 1185 1185’

275 280 280

N N N

257.5 257.4 257.5

B&H G N

AcTrpNH, Tyr AcTyrOEt

AcTyrNH,

Phe AcPheOEt

195 197 195

e 5800

X (nm) 280

8 Author

M

urea

e

X (nm)

Author

5870

max

H&Sd

1420

max

H&S

Eb

5690 5610”

280.8 280

E N

1475,1515 1450 1500

275.5 275.3 max 275.5

E B&K E

1490 1470 1280 1280

275.5 275.5 280 280

E N E N

195

258

N

’ B&H, Beaven and Hiliday (14) Data in N/10 HCl listed. b Edelhoch (1). ’ S&H, Sol1 and Herskovits (16). ’ H&S, Herskovits and Sorensen (15). ’ Gratzer, W. G. (1973) in Handbook of Biochemistry (Sober, H. A., Ed.), Chemical Rubber Co. Cleveland, OH. f Nozaki, this paper. ’ Maximum in 0.02 M Na phosphate (pH 6.8) is located at 279.5 nm, maximum in 6 M GuHCl is at 280.8 nm, and their absorbances are higher than the values at 280 nm shown above, but the differences are not significant. Therefore, the values at 280 nm are listed for practical purposes. h B&K, Brants and Kaplan (17). ‘ Spectrum was obtained by J. Aviv on 14DS spectrophotometer.

330

YASUHIKO TABLE

NOZAKI

III

tration of tyrosine is linear and the slope, which represented the molar second derivative absorptivity, is -4.95 X 1016cmp3 M-’ (Fig. 3b). The second derivative spectrum of AcPheOEt is shown in Fig. lc. Of the five troughs, the two at 264.3N& 264.5 and 258.1 nm were larger than others. The former, N&Y NL however, was selected to represent phenylalanine, beProtein Normal” 2nd derivative 2nd derivative 2nd derivative cause the latter was less symmetric and likely a composite band. The molar second derivative absorptivity of BSAb 3.1, 3.4 2.0 19.1 27.1 phenylalanine is listed in Table I. (Lit.: 2 19 27) HSAb 2.0.2.7 1.0 18.0 29.0,32.0 The success of the present method has depended upon (Lit.: 1 18 31) strict adherence to a single, properly chosen wavelength RNaseTl” 1.2, 1.1 1.0 9.0,8.5 4.0 in the measurement of second derivative absorbance and (Lit.: 1 9 4) the correction for the effect of interfering contributions. Enolased 5.2 9.1 16.0 This has been possible only in 6 M GuHCl solution, (Lit.: i’” 9 16) CLSO’ 12 15 which, by denaturing the protein, exposes all chromo(Lit.: ti.5 12.5 15) phoric side chains to solvent, and also eliminates intraand intermolecular electrostatic effects by providing a ’ Edelhoch’smethod (1). high concentration of ions. Thus the resemblance of a * BSA, HSA, National Biomedical Research Foundation Protein protein and the mixture of constituent amino acids (aIdentification Search,Release15, 1988. ’ RNaseTl: ribonucleaseTl, Takahashi, K. (1985)J. B&hem. 98, blocked) was nearly perfect, so that additivity of absor815-817. bances, both normal and derivative, was realized. The dEnolase, Chin, Ch. C. Q., Brewer, J. M., and Wold, E. (1981) J. effect of the presence of an interfering amino acid is Bid. Chem. 256,1377-1384. merely algebraic and no physicochemical interactions ’ CLSO, chicken liver sulfite oxidase;Kessler, D. L., and Rajagopaare involved. lan, K. V. (1972)J. Biol. Chem. 247,6566-6573. Aromatic Amino Acid Residues on Proteins (NiA) Determined by the Present Method and Edelhoch’s Method in Comparison with Literature Values

The linear relation between the second derivative absorbance value and the concentration of tryptophan is illustrated in Fig. 3a, with the slope of this line representing the molar second derivative absorptivity (listed in Table I). Another trough at 282.7-282.8 nm is much shallower and variable, so that it is not used as a measure of tryptophan. However, the contribution of tryptophan near this wavelength plays a crucial role in determining tyrosine concentration, because the minimum almost coincides with that of tyrosine located at 283.1 nm. The tryptophan contribution at 282.7 nm should not be used for correction, since it would lead to too large correction values. Tryptophan also shows nonzero contribution at 264.3-264.5 nm, which is the location of the major trough of phenylalanine (Table I). These contributions from tryptophan are corrected during the determination of tyrosine and phenylalanine, respectively. The same arguments are applicable to tyrosine. As can be seen in Fig. lb, there is no major region showing a zero value. Therefore, the contribution from tyrosine cannot be ignored in evaluating the second derivative value for the other aromatic amino acids at any wavelength. Thus the molar contribution from tyrosine to the measurement of tryptophan at 290 nm (Fig. 3c) and to the measurement of phenylalanine at 264.3-264.5 nm have been listed in Table I. The trough at 283.1 nm is chosen as the reference point for measuring the tyrosine contribution. The plot of second derivative, thus obtained, against the concen-

Absorptivities

in Dilute Salt Solutions

During the course of studies of second derivative absorbances of the three aromatic amino acids, a number of normal molar absorptivities in dilute salt solution at pH 6.8 were also compiled (Table II). It was noted that, as expected, the effect of different solvents was virtually nil for AcPheOEt, while the molar absorptivities of AcTyrNH, and AcTrpNH, were greater in 6 M GuHCl or 8 M urea than in water or dilute salt solutions. When the same kind of comparison was made for second derivative spectra, AcTyrNH, showed a 10% greater value in 6 M GuHCl (-4.95 X 1016cmp3 M-l) at 283.1 nm than in 0.02 M NaPi (-4.4 X 1016cmW3M-l) at 282.8 nm. On the other hand, AcPheOEt showed a 10% smaller magnitude of value in 6 M GuHCl (-1.95 X 1016cm-3 M-‘) at 264.4 nm than in NaPi (-2.15 X 1016 cm-3 M-‘) at 263.7 nm, and AcTrpNH, shared the same trend, i.e., larger magnitude in 0.02 M NaPi (-2.12 X 1017cmm3M-‘) at 288.9 nm than in 6 M GuHCl (-1.95 X 1017 cm- M-l) at 290 nm. Thus AcTrpNH, showed a larger normal molar absorptivity at 280 nm in 6 M GuHCl (Table II), but the larger magnitude of the second derivative value was indicative of the sharper5 constituent band near the shoulder at 289 nm in 0.02 M NaPi than in 6 M GuHCl at 290 nm. When a comparison of normal absorptivities was made at the same wavelength, 289 nm, instead of 280 nm, the same 5The depth of a secondderivative curve reflects the narrownessof an absorption band. It may be better understoodby using a modelsuch as describedin the Appendix.

AMINO

ACID

ASSAY

BY DERIVATIVE

-20

?2c NTrp tCTrp.added

'CP,,,)

NT,r iCTyr.added

331

SPECTROPHOTOMETRY

0

20

40

NPhe KPhe.added

"Prot)

60

80

"Prot)

FIG. 4. Estimation of the numbers of aromatic amino acid residues on CLSO. Distance from the common intercept to N = 0 corresponds to the number of residues per subunit (M, 55,000). (a) Fifteen tryptophan residues. Initial C rrot: top, 1.156 X lo-” M; middle, 5.91 X 10m6M; bottom, 5.58 X 10e6 M. (b) Twelve tyrosine residues. Initial Crrot: from top, 1.385 X lo-’ M, 1.332 X 10m5M, 8.87 X lo-” M, 7.29 X 10m6M, 4.06 X 10m6M. (c) Fifteen phenylalanine residues. Initial C rrot: top, 1.344 X 10e5 M, middle, 1.135 X 10e5 M, bottom, 1.067 X 10-“M.

trend prevailed (data not shown). These variations of normal and derivative absorbances may not be directly related with each other, because those two absorbances are located at different wavelengths and, therefore, their origins may be different. But further discussion is beyond the scope of this paper. Application

to Proteins

The application of second derivative spectrophotometry to a number of proteins is tabulated in Table III, but the analyses of chicken liver sulfite oxidase are described in detail. As a heme-containing metalloprotein, it offered an interesting test case for the second derivative spectrophotometric analyses, because it showed nonzero absorption in the near uv region. This absorption directly interfered with the application of the normal absorbance method of Edelhoch (1) for determining tryptophan residues and indirectly by having broad absorption bands in the visible region which virtually precluded the correction for turbidity (2) to obtain true absorbances at 280 nm and 288 nm. Second derivative spectrophotometry appeared to be immune to these difficulties. The results of determinations of the three aromatic amino acid residues are shown in Figs. 4a, 4b, and 4c.

The first question to be answered is how small the wavelength differential should be to approach the theoretical values of the second derivative within an allowable limit. In order to answer this question, we have to consider a model absorption curve which is represented by a simple, mathematical expression whose derivatives can be computed unambiguously. Such a model, frequently used to represent absorption curves, is a Gaussian distribution curve. We then compare the theoretical values and the calculated ones with various discrete values of wavelength differential. The second question is how to obtain theoretical values of the second derivative of a Gaussian curve. We assume that the values obtained with a hand-held calculator (HP 28C) are sufficiently accurate. A simple expression of a Gaussian distribution is

TABLE

Comparison of the Molar Second Derivative Absorptivities Obtained on Shimadzu UV260 and Those Calculated Directly from Absorbances in 6 M GuHCl

(lOi

APPENDIX

Calculation of Absolute Second Derivative Absorbance Values This Appendix describes an attempt to reach reasonable values of absolute second derivative absorbances of the three aromatic amino acids using simple procedures.

IV

Amino acid AcTrpNH, AcTyrNH, AcPheOEt

A%/AX’ cm-3M-‘)

X (nm)

UV260

calcd

290.0 283.1 264.3-264.5

-1950 -495 -195

-230 -58 -38

332

YASUHIKO

A = c exp(-X21n 2/B2)

111

NOZAKI

rected by Steiner et al. (19). They used a polynomial the form

in

where A is absorbance, c is concentration, X is wavelength, B is halfwidth at halfheight (HWHH), and an optical path of unit length is assumed. A primitive method of calculating a second derivative applicable to both a Gaussian equation and an experimental absorption curve is the following:

to approximate a section of the curve scanning from i = -m to m with raw Ai, at the central point, where A,, is the value of absorbance at the central point of the At (i - 1)th wavelength, smoothed section, ai’s are weighting coefficients and N AA/AX = (A, - Aip2)/2AX PI is a normalization coefficient. The values of ais and N’s are given in Refs (18) and (19) and the method has been At (i + 1)th wavelength, critically reviewed by Enke and Nieman (20). It must be remembered that smoothing is normally acAA/AX = (Ai+ - Ai)/2Ah [31 companied by distortion (18), which minimizes the magThen at ith wavelength, nitude of the resulting second derivative values. The distortion can be controlled by choosing such parameters A2A/AX2 = (Ai+z - 2Ai + Aipz)/(2AQ2, 141 as the number of data points per passing, the number of where A’s are absorbances and Ax is assumed to be of passes through data points, the size of intervals (Ax), and the degree of polynomial to be used, etc. Several sets the same values throughout. Maximum allowable value of wavelength increment depends on the size of B.6 It has of parameters have been tried empirically for smoothing been shown that for a Gaussian curve with B = 20 A and the absorption curve of each of the three aromatic amino acids without losing details and particularly without Ax = 1 A or B = 2.4 A and Ax = 0.1 A, the data obtained shrinking. Numerical calculation was carried out at first by the present method and by the calculator are indistinwith a hand calculator (HP41CV) on a program with Ax guishable. Therefore, it may be concluded that, when B = 1 A and data points 25. Later calculations were peris as large as 20 A, Ax of 1 A is small enough to produce formed with a computer on a more flexible program correct second derivatives. A Ax of 0.1 A is small enough in any circumstance usually encountered in the case of (YAS.C) which could handle any value of Ax, e.g., 0.1 A of Eq. those aromatic amino acids, B values being between 2.4 and even smaller, data points per pass (-m-+m A and 5.3 A (see below). When B is 5.3 A and Ax is 0.1 A [5]) 7-25, and several modes of smoothing and derivatithe agreement is perfect. But if Ax is 1 A, difference of zation, separately and combined. The mode which was used most frequently was the combination of smoothing the two sets of second derivatives becomes 4.8%. When (once, twice, thrice, etc.) plus the primitive derivatizaB is further reduced to 2.4 A, Ax of 1 A produces a 21% tion by Eq. 141, because it allowed the evaluation of the smaller value in trough depth than the correct one. effect of smoothing for each number of smoothings, unSince most commercial spectrophotometers’ printout like the Savitsky method, which smoothed and derivashows no wavelength intervals smaller than 1 A, a B tized simultaneously. value of 5-6 A may be the lower limit. The essential re1. An example of tryptophan, 2.512 X 10e4 M. quirement for a model curve to be adequate in the presComparison of 2 passes of quartic-quintic7 smooth with ent case is that it produces a second derivative value 25 data points and 6 passes of quadratic-cubic7 smooth close to the similar value produced by the experimental absorption curve at the trough minimum. It has been with 11 data points showing almost identical results on derivatization with a trough depth ( A2A/AX2) -0.0575 X shown above that, for a Gaussian curve, the combination 101* cmp3 MP1 at 290 nm. (Table IV contains the molar of a Ax of 1 A and a B value as small as 5.3 A gives tolerable closeness. To obtain tolerable theoretical values of value.) Since the half-width of a second derivative spectrum second derivatives from experimental absorbance valon the baseline corresponds to 85% of the half-width at ues, the latter is usually accompanied by too much noise, and smoothing of the curve is necessary. A typical half-height of the original curve, if it is Gaussian, the method is proposed by Savitsky and Goley (18) and corexperimental value of 17 A of the former corresponded to 20 A of HWHH of the latter. However, a Gaussian “B values shown subsequently were obtained by trial-and-error curve with B (=HWHH) of 20 A produced a value of secmethod. For AX, 1 A, and the value of 20 A, which coincided with the ond derivative -0.0040 X 101* cmp3, much too low comHWHH of the apparent constituent peak of the tryptophan absorppared with the experimental value. Instead, a B value of tion curve at 289 nm, was tried first. Then, since the trough depth calculated was too small, the B value was reduced gradually until calculated and experimental values of trough depth coincided reasonably well at B = 5.3 A. The same B values also worked well for tyrosine. Further reduction of the B value to 2.37 A to simulate phenylalanine produced only a modest agreement.

7 Those numbers designate the degrees of the polynomial being used. The reason why they appear pair-wise is that the members of a pair share the same set of weighting coefficients.

AMINO

ACID

ASSAY

BY DERIVATIVE

5.3 A produced a value of -0.058 X 1014cme3, which was almost identical with the experimental value. The large discrepancy in the widths of the actual spectrum and the model, 21 A against 5.3 A, may suggest the possibility of the spectral band consisting of several constituent subbands. This suggestion may appear to be farfetched, but it is quite reasonable, since a normal absorption band is likely to be composite, each constituent subband having HWHH of 5-6 A. One of them at the absorption maximum is responsible for producing the trough minimum of the second derivative spectrum. This view is supported by the absorption spectra of benzene, which show several rather sharp bands for its heptane solution, but each band is shown to consist of four or five subbands which are resolved in vapor phase. 2. An example of tyrosine, 6.8 X lo-* M. The normal absorption curve was best smoothed with 2 passes of quartic-quintic7 smooth with 25 points leading to a smooth curve with the second derivative values of -0.039 X lOI cmd3 at 283.1 nm. (Table IV contains the molar value.) The halfwidth of the derivative curve on baseline 21 A corresponds to HWHH 24.7 A of a normal spectrum. Like tryptophan, a Gaussian curve with B = 20 A gave a second derivative of -0.0027 X 1Ol4 cmP3. This was too low, but another Gaussian curve, with B = 5.3 A, gave a good fit with (A2A/AX2) of -0.039 X 1014 cmm3. The result again gives a picture similar to that of tryptophan, i.e., the absorption band could consist of several constituent subbands with HWHH of 5-6 A. 3. An example of phenylalanine, 2.983 X 1Oe3M. The original absorption spectrum was best smoothed by 2 passes of quartic-quintic7 smooth with 11 data points, leading to a second derivative -0.11 X 1014 cm-3 at 264.3-264.5 nm. (Table IV contains the molar value.) It was difficult, as shown earlier, to find a Gaussian model with a reasonable B value. The closest fit was shown with B = 2.37 A and yet the calculated trough depth of -0.088 X lOI cmP3 is only 80% of the experimental value. The ratio of HWHH of the model curve and that of the real original curve again shows a value close to 5, as exhibited by two other amino acids. Table IV summarizes the molar second derivative absorptivities on Shimadzu UV260 and those calculated directly from absorbance values as described above. It is noted that the molar second derivative values of the three aromatic amino acids calculated from the figures published by Levine and Federici (10) are fairly close to the calculated values in Table IV: tryptophan -190

333

SPECTROPHOTOMETRY

X 1014cme3 M-‘, tyrosine -50 X 1014cme3 M-‘, phenylalanine -12 X 1014Cme3 M-l. ACKNOWLEDGMENTS The author is indebted and expresses his thanks to Drs. I. Fridovich, H. M. Steinman, and B. Kaufman and Mr. P. Gardner who read and improved the manuscript, Dr. T. Takagi who gave enolase, Dr. N. C. Pace who gave RNase Tl, Dr. K. V. Rajagopalan who gave the sample of CLSO, Mr. M. Stump who loaned his calculator (HP41CV) and programmed quartic-quintic smoothing, Dr. G. Neece who programmed YAS.C, which was essential to this work, and Mr. J. Aviv who volunteered to run the sample of AcTyrNHs on his 14DS spectrophotometer.

REFERENCES 1. Edelhoch,

H. (1967) Biochemistry

6,1948-1954.

2. Nozaki, Y. (1986) Arch. Biochem. Biophys. 249,437-446. 3. Ichikawa, T., and Terada, H. (1977) &o&em. Biophys. Acta 494, 267-270. 4. Ichikawa, T., and Terada, H. (1981) Chem. Pharm. Bull. (Tokyo) 29,438444. 5. Terada, H., and Ichikawa, T. (1984) Chem. Pharm. Bull. (Tokyo) 32,585~590. 6. Balestrieri, C., Colonna, G., Giovane, A., Irace, G., and Servillo, L. (1978) Eur. J. Biochem. SO, 433-440. 7. Balestrieri, C., Colonna, G., Giovane, A., Irace, G., and Servillo, L. (1980) Anal. Biochem. 106,49-54. 8. Servillo, L., Colonna, G., Balestrieri, C., Lagone, R., and Irace, G. (1982) Anal. Biochem. 126,251-257. 9. Ragone, R., Colonna, G., Balestrieri, C., Servillo, L., and Irace, G. (1984) Biochemistry 23,1870-1875. 10. Levine, R. L., and Federici, M. M. (1982) Biochemistry 21,2600-

2606. 11. Nozaki, Y. (1972) in Methods in Enzymology (Hirs, C. H. W., and Timasheff, S. N., Eds.), Vol. 26, pp. 43-50, Academic Press, New York. 12. Haupt, W. G. (1952) J. Res. N&l. Bur. Stand. US 48,414-423; J. Opt. Sot. Amer. 42,441-447. 13. O’Haver, T. C., and Green, G. L. (1976) Anal. Chem. 48,312-318. 14. Beaven, G. H., and Hiliday, E. R. (1952) in Advances in Protein Chemistry (Anson, M. L., and Edsall, J. T., Eds.), Vol. 7,pp. 319377, Academic Press, New York. 15. Herskovits, T. T., and Sorensen, Sr. M. (1968) Biochemistry 7,

2525-2532. 16. Soll, N. J., and Herskovits,

T. T. (1973) Anal. Biochem. 54, 370-

378. 17. Brants, J. F., and Kaplan, L. (1973) Biochemistry 12,2011-2024. 18. Savitsky, A., and Goley, M. J. E. (1964) Anal. Chem. 36, 16271639. 19. Steiner, J., Termona, Y., and Deltour, J. (1972) Anal. Chem. 44, 1906-1909. 20. Enke, C. G., and Nieman, T. A. (1976) Anal. Chem. 48, 705A712A.