Proton NMR studies of bovine serum albumin - Wiley Online Library

Eur. J Biochem. 205, 631 -643 (1992) 4> FERS 1992

Proton NMR studies of bovine serum albumin Assignment of spin systems Petcr J. SADLEK and Alan TUCKER Christophcr lngold Laboratories, Birkbeck College London, England (Received October 18, 1991) - EJB 91 1400

A variety of one- and two-dimensional 'H-NMR methods have been applied to the study of defatted 66.5-kDa bovine serum albumin in solution. 1. The majority of the protons gave rise to broad unresolved resonances and spectral enhancement methods for one-dimensional spectra were investigated in detail. A combination of exponential and sine-bell functions was particularly effective. 2. The presence of contaminating glycoproteins in some commercial samples of bovine serum albumin was readily detectable from their N-acetyl resonances at about 2.1 ppm. 3. The release of bound Cys (from mixed disulphide at Cys34) was observed after addition of dithiothreitol. 4. Through the use of two-dimensional shift-correlated spectroscopy, assignments of some 80 spin systems to amino acid type were made. 5. The pK, of the N-terminal Asp was measured as 7.8 (0.1 M phosphate buffer, 310 K). 6. 'H NMR spectra of bovine, human, porcine and rat serum albumins have been compared. Using sequence comparisons, specific assignments have been made for the N-terminal residues of bovine (Asp-Thr-His), human (Asp-Ala-His), porcine (Asp-Thr-Tyr) and rat (Glu-Ala-His) albumins, and for Thrl89, Tyrl55 and His59/377 of bovine albumin. 7. These NMR data suggest that certain local regions of bovine serum albumin are highly mobile yet structured in solution, and demonstrate that the application of both one- and and two-dimensional NMR methods will allow more detailed investigations of structural transitions in serum albumins induced by, for example, pH, drug and metal binding.

Albumin (66.5 kDa) accounts for about 60% of the total protein in blood serum with a concentration of 42 g I - ' (0.63 mM), a range of 35- 50 g 1-' [l]. I t has a wide range of chemical and laboratory uses, and physiological functions [2]. For example, albumin is involved in the binding, transport and delivery of fatty acids, bilirubin, tryptophan, thyroxine, steroids and a range of pharmaceuticals and dyes [3]. There are also specific metal-binding sites for Cu(II), Ni(II), Ca(II), Cd(II), Zn(I1). Au(I), and Hg(I1). Studies on the structure and dynamics of albumin in solution are crucial for understanding these processes. NMR methods are powerful for the investigation of protein structures in solution. However, severe problems with line-broadening and assignment of 'H-N MR resonances arise for proteins larger than about 15 kDa and there have been -~

Corrc.spondmce to P. J . Sadler, Department of Chemistry, Birkbeck Collcge, Univcrsity of London, Gordon House, 29 Gordon Square. London WClH OPP, England Ahhreviutions. 1D and 2D, one- and two-dimensional; BSA, HSA, PSA and RSA, bovinc, human, porcine and rat serum albumin; BSA2, BSA purified by affinity chromatography; COSY, homonuclear shift-correlatcd spcctroscopy; DQF-COSY, double-quantum-Gltered phase-scnsitive COSY; HOHAHA, homonuclear Hartmann-Hahn correlated spectroscopy; pH*, pH mcter reading in 'H,O; P5 and P95, 5th and 95th perccntile; T I and T,, longitudinal and transverse rclaxation timcs.

relatively few detailed studies of high-molecular mass proteins. Favourable cases for study have been where multidomain proteins exhibit significant internal mobility, e. g. prealbumin (transthyretin 55 kDa) [4], urokinase (54 kDa) [ 5 ] , horseradish peroxidase (42 kDa) [6]. Serum albumin ( x 580 amino acids) is a single-chain protein composed of three structurally similar domains [7]. The arrangement of the 35 Cys residues to give 17 disulphide bridges gives rise to three domains containing long-short-long loop patterns. Each domain (I, 11, and 111) is composed of two sub-domains (A and B). Because of the large number of molecules and ions which can bind tightly to albumin and the possibility of mixed disulphide formation at the free thiol group (Cys34), attention has to be paid to the purity of isolated albumin [8]. There are only a few previously reported high-resolution 'H-NMR studies of serum albumin, all of which are onedimensional, including two studies of His C2H resonances of human serum albumin and its tryptic and peptic fragments [9, 101 in which assignments were proposed for His3 and His464. the use of difference spectra to study fatty acid binding [I I], and investigations of cross-relaxation phenomena [12]. In the present study we have used one-dimensional (ID) and two-dimensional (2D) H-NMK methods to investigate the purity of albumin, heterogeneity at Cys34, an2 the assignment of spin systems for a variety of amino acid types. Com-

632 parison of the spectra of several mammalian albumins (bovine, human, porcine and rat) has allowed sequence-specific assignments to be made for the N-terminal and a few other amino acid residues. These methods will provide a basis for more detailed investigations of the structure and dynamics of albumin in solution and its interactions with drugs, xenobiotics and other molecules and ions.

EXPERIMENTAL PROCEDURES Materials Bovine serum albumin was purchased from Boehringer (cat. no. 775835) as a fatty-acid-free Cohn fraction V (referred to as BSAI). This was further purified by affinity chromatography using Blue Sepharose (Pharmacia) as the immobilising medium [I31 to give BSA2. Other serum albumins were purchased as follows: human (Sigma, A1887), porcine (Sigma, A1 173, globulin-free), rat (Sigma, A6414, globulin-free). Wheat germ lectin, sialidase and dithiothreitol were all purchased from Sigma. Albumin purification An affinity column (30 cm x 1.5 cm) containing blue Sepharose (estimated capacity 450 mg albumin) was equilibrated with 0.1 M phosphate, pH 7.4 and loaded with albumin (BSA1 600 mg/lO ml elution buffer), and eluted with the same buffer (detection at 280 nm) for 7 h and the bound albumin was then displaced by inclusion of 0.2 M KSCN in the buffer. Albumin was collected in 30 - 50 ml volumes, freeze-dried, dissolved in water and desalted on Pharmacia PD-10 columns (Sepkadex G25M) by elution with 50 mM ammonium bicarbonate solution, pH 7.8, which was removed by freeze drying. This purified albumin is hereafter referred to as BSA2. Typically, the free thiol contents were: BSAI 0.59 mol SH/mol and BSA2 0.67 mol SH/mol (2,2’-dithiopyridine titration [14]).

1.1 s for sharp peaks in the aliphatic rcgion (0.5-4.4 ppm) were estimated. Two-dimensional data sets were acquired with 1K or 2K data points in the t 2 dimension, acquisition time 0.085 s or 0.170 s, 256 or 512 increments in the t l dimension, relaxation delay 1-2.5 s, 48-128 transients. Standard 2D pulse sequences were used: absolute-value homonuclear shift-correlated spectroscopy (COSY) [I 51, phase-sensitive double-quantum-filtered COSY (DQF-COSY) [16], absolute-value relayed COSY [17] and homonuclear Hartmann-Hahn (HOHAHA). Routinely, data sets were zero-filled to 4K in t 2 and 1K or 2K in t l and sine-bell functions were used for processing in both dimensions.

NMR samples The concentration dependence of spectra was investigated in the range 0.5 -6 mM but above 4 mM a significant broadening of peaks was observed. In general, 2 mM solutions of albumins in 0.1 M deuterated phosphate buffer were used subsequently. For the pH titration, individual samples were prepared at each pH* value rather than by the adjustment of the pH* of a single sample, except that additions of N a 0 2 H and *HCIwere made beyond the range of the buffer. The pH* value (meter reading) of each BSA sample was recorded using a Corning 145 pH meter with an Ingold pH electrode (6030-02). Chemicals shifts are referenced to 3-trimethylsilyl(2,2,3,3-’H4)propionate via internal dioxan (3.764 ppm). For wheat germ lectin addition, 90 pl of a solution of the lectin (1 mg/ml ’H20) was added to the BSA solution in the NMR tube, and for sialidase a 10-pl aliquot of a 200 units/ml ’H20 solution was added. Similarly dithiothreitol was added as microlitre aliquots of a stock solution in ’H20.

RESULTS Albumin purity In our initial studies, the effect of the purity of albumin on H-NMR spectra was considered in some detail.

NMR Spectra were recorded on Bruker AM500 or AM400 spectrometers (MRC Biomedical NMR Centre, Mill Hill), using 0.5 ml of sample in 5-mm tubes. Spectra of BSAl were recorded over the temperature range 288-313 K. Only minor changes in the spectra were observed, but some sharpening of peaks occurred at higher temperature and 310 K was chosen for subsequent work. Typical pulsing conditions for I D spectra were: pulse width z 41 ( 5 ps), relaxation delay 2 s, acquisition time 0.68 s, spectral width 12 ppm, data points 8K (zero filled to 16K), 128 transients. Where necessary the residual HO’H resonance was suppressed by presaturation during the relaxation delay. Resolution enhancement was achieved via the use of sine-bell, shifted sine-bell, convolution difference, trapezoidal and Gaussian functions. Exponentialsine-bell multiplication was achieved by the sequential application of exponential and then unshifted sine-bell functions. Spin-echo spectra were acquired using the Hahn sequence and processed using exponential functions with a line-broadening equivalent to 0.5 - 2 Hz. As a guide to the choice of NMR parameters, inversionrecovery spectra of BSA were recorded, from which apparent longitudinal relaxation times ( T , )of 1.2 - 1.9 s for sharp peaks in the His C2H and C4H regions (6.9 - 8.9 ppm), and of 0.5 -

Glycoproteins The spin-echo spectrum of commercial fatty-acid-free Cohn fraction V (defatted by the method of Chen [18]) is shown in Fig. 1. The peaks near 2.1 ppm are similar to those seen in spectra of blood plasma. We confirmed their assignment as N-acetyls of contaminating glycoproteins by addition of wheat germ agglutinin (Fig. 1 B). This lectin is known to bind to N-acetylneuraminic acid and N-acetylglucosamine and the reduction in mobility of these sugars through binding leads to a decrease in transverse relaxation time constant ( T 2 ) and a decrease in their intensities in the spin-echo spectrum. Furthermore, addition of sialidase led to the appearance of a more intense singlet at 2.1 ppm assignable to free N acetylneuraminic acid. Also, in the 2D DQF-COSY spectrum of this sample (BSAI), there were clearly-resolved cross-peaks centred at 1.75/2.7 ppm and 1.75/3.7 ppm assignable to H3,/ H3, and H3,/H4 connectivities of N-acetylneuraminic acid [20]. These glycoprotein impurities were not detectable by PAGE/SDS electrophoresis under reducing conditions using Coomassie blue as a stain. For further NMR work, albumin samples were purified on a blue Sepharose affinity column to remove the glycoproteins.

633 1D Spectral-enhancementmethods Thiol heterogeneity

The purified samples of BSA (BSA2) used for NMR work had free thiol contents of about 0.6 mol SH/mol. Albumin is known readily to form disulphides at Cys34 with, for example, glutathione or cysteine, accounting for the remaining 40%. To investigate the effect of blocking Cys34 on the NMR spectrum, a reducing agent (dithiothreitol) was added. This reagent is known to reduce disulphides of Cys34 but not structural disulphides of albumin in the pH range 5 - 7 IS]. Addition of dithiothreitol led to a few specific changes in the spectrum: the appearance of resonances for free dithiothreitol, and peaks assignable to the AMX spin system of free cysteine (Fig. 2C). N o peaks for glutathione and no significant changes in the observable BSA peaks (aliphatic or aromatic) were seen at the pH* used (5.62). We therefore concluded that the heterogeneity of Cys34 does not significantly affect our interpretation of NMR spectra of BSA.

NAc

Diox

4

The 1D 'H-NMR spectrum of BSA consists of a broad envelope of overlapping peaks with some sharp features (Fig. 3). Notable are the resolved singlets at about 8.28.8 ppm which are likely to arise from imidazole protons of His residues. Other workers [9,10] have attempted to monitor these resonances during pH titrations (of human serum albumin), but due to overlap with other broad features few of them can be followed with confidence over a wide pH range.

2

3

1

0 Uppm

GIPPm

Fig. 1. 500-MHz Hahn spin-echo 'H-NMR spectra (7 = 30 ms) of commercial fatty-acid-free BSA (BSA1) in 0.1 M phosphate pH* 7.55, (A) before and (B) aftcr addition of wheat germ agglutinin. A significant reduction in intensity of the NAc peak at 2.1 ppm is apparent. Assignments: Diox, dioxan (ref.), NAc, N-acetyl (N-acetylncuraminic acid, N-acetylglucosamine) of gl ycoproteins

4

2

3

O

1

Fig. 2. 500-MHz 'H-NMR spectra of BSAl (resolution-enhanccd) (A) before and (B) after addition of 3 mol dithiothreitol/mol, 0.1 M phosphate pH* 5.62, (C) difference spectrum (B-A), and (D) spectrum o f dithiothreitol IHSCH,CH(OH)CH(OH)CH,SHI. Assignments: EtOH, ethanol (contaminant), Cys, cysteine (displaced from disulphide linkage at Cys34 by dithiothreitol)

I _

Wvm

"

I

3

'

I

2

"

I

"

I '

1 0 Fig.3. 500-MHz 'H-NMR spectra of BSAl (2 mM in 0.1 M phosphate pH* 5.7). (A) Aliphatic region, (B) aromatic region. In each case thc upper spectrum was obtained without the use of a window function, and the Iowcr spectrum is the result of the application d'sequential exponential (line-broadening or 2 Hz) and unshifted sine-bell functions to thc frec induction decay before Fourier transformation. 4

634 Tablc 1. Assignment of 'H-NMR spin systems of BSA at pH* 6.95. Only Thr I gives rise to a relayed-COSY peak; the bchaviour of the COSY cross-peaks with changes in p H * for Thr 11-VII suggests that these also arise from Thr residues. Assignment

Chemical shifts of 3

B

r'

2.74 4.07 4.35 4.30 3.94 4.22 4.13 4.16 2.25 2.13 2.09 2.04

2.61 1.14 1.26 0.92 1.04 1.25 1.44 0.52 1.19, 1.09 1.11, 0.96 1.02.0.99 0.66,0.62

(5

other

1.16 1.73 1.68

c 3.06

PPm Asp1 Thr I T h r II Thr 111 Thr IV Thr v Thr VI Thr VII Val I Val I1 Val I l l Val IV Lys 1 Lys I1 Lys I l l

4.22 4.28 3.95 4.14 4.26 3.95 3.88 3.72 4.20 3.81 3.55 4.07

1.43, 1.34

Phe I Phc I1 (pH* 7.85)

His I His II His 111 Ilis IV His V liis VI His VII His VlIl His IX His X His XI (pH* 9.04) Aln I A h II Ala I l l Ale IV Ala V Ala VI Ale VII Ala VllI Ala 1X Ala X Ala XI Ala XI1 A h Xlll Ala XIV Ala XV Ala XVI Ala XVIl Ala XVllI Ala XIX Ala X X Ala XXI Ala XXII Ala XX I I I Ala XXIV Ala XXV Ala XXVI A h XXVlI Ala XXVIIl Ala XXIX Ala XXX Ala XXXl

4.84 4.85 4.62 4.60 4.55 4.55 4.48 4.44 4.47 4.42 4.39 3.52 3.85 3.68 4.10 3.87 3.97 4.04 4.05 4.12 4.20 4.20 4.08 4.13 4.25 4.35 4.16 4.26 4.33 4.31 4.55

1.39 1.25 1.54 1.47 1.33 1.22 1.18 1.28 1.48 1.65 1.69 0.43 0.84 1.39 1.74 1.38 1.43 1.18 1.51 1.57 1.59 1.49 1.33 1.31 1.40 1.47 1.34 1.32 1.29 1.54 1.49

3.01 2.99 ring protons 7.02, 7.24, 1.33 6.69, 6.89, 7.21 ring protons (2H. 411) 8.11, 7.35 7.84, 6.81 8.18, 7.16 8.30, 7.32 1.97 7.92 8.34 8.07 1.72 7.69 7.68 8

I:

635 'Table 1. (continued).

Assignment

Chemical shifts of I

D

Y

S

BiY

UP

S/E

other

PPm Val/Leu I Val/Leu 11 Val/Leu 111 Val/Leu IV Lys/Arg I Lys/Arg I I Lys/Arg 111 LysiArg 1V Lys/Arg V Lys/Arg VI

1.54 1.49 1.82 1.67

0.91,0.78 0.79, 0.76 0.97.0.91 1.10,1.04

I .72 1.64 1.50 1.68 1.58 1.51

Assignment of 2D spectra We recorded a number of 2D spectra, including COSY, DQF-COSY, relayed-COSY, and HOHAHA, of BSA and other albumins (human, rat, porcine) at various pH values. For BSA, under the conditions used, less than a fifth of the amino acids appeared to give rise to significant cross-peaks. Of these, about 80 spin systems were assignable (with some

2.94 2.89 2.86 2.83

ring protons 6.54, 6.39 6.51, 6.71 6.82, 6.47 7.00, 6.52 7.16, 6.78 7.07, 6.79 7.08,6.86 7.62, 7.19 7.41, 6.71 6.13, 5.88 7.40, 6.98 7.54, 7.26 7.02,5.98

PhelTyr 1 Phe/Tyr I1 Phe/Tyr I I1 Phe/Tyr IV Phe/Tyr V Phe/Tyr V1 (pH* 7.85) Phe/Tyr VII (pH* 7.85) Phc/Tyr Vlll Phe/Tyr IX Phe/Tyr X Phe/Tyr XI Phe/Tyr XI1 Phe/Tyr XlII

Also other regions of the spectrum of human serum albumin were not analysed in these reports. For these reasons we investigated in detail methods for resolution-enhancement of 'HNMR spectra of BSA. Firstly, a variety of window functions were applied to spectra acquired in single-pulse mode: convolution difference, trapezoidal, Gaussian, sine-bell and phase-shifted sine-bell functions, as well as sequential application of both exponential and sine-bell functions. The latter was found to be optimal in terms of the balance between resolution enhancement, signal/ noise ratio and ease of application at workable BSA concentrations (up to about 2 mM, see Experimental Procedures and Fig. 3). Secondly, broad resonances were filtered out using multiple pulse methods: Hahn and Carr-Purcell-Meiboom-Gill (CPMG) spin-echo, and spin-locked spectra. In general, these were used to complement normal single-pulse and resolutionenhanced single-pulse spectra and were especially useful for confirming the assignment of multiplets, when phase-modulated, and for detecting specific features, e.g. N-acetyls of contaminating glycoproteins which have long T2 values (see Fig. 1 A).

2.92 2.94

ambiguities) to particular amino acid types and werc labelled with the appropriate three-letter code and an arbitrary Roman numeral (Figs 4 and 5 and Table 1).

Alanine and threonine Twelve pairs of cross-peaks with characteristics expected of Thr residues can be identified in the DQF-COSY spectrum of BSA at pH* 6.95. However, in similar spectra of BSA at pH* 7.85, 7.56 and 6.20, five of these systems no longer show correlations between the potential c$ and B?, cross-peaks. The remaining seven systems consistently exhibit this correlation and are shown in Fig. 5B (by cross-peaks). Further confirmation of these spin systems was sought by the use of relayedCOSY experiments. An ccy relay cross-peak for Thr I was clearly seen, confirming this as a threonine spin system. However, no relay cross-peaks were seen for the other six proposed Thr spin systems. An intense doublet observed in the resolution-enhanced 1D spectrum of BSA at about 1.14 ppm is the yCH3 resonance of the spin system assigned as Thr 1. The long T2 of this resonance may be responsible €or the observation of a relayed ay cross-peak for this threonine alone. For five of the seven proposed Thr spin systems, the chemical shifts of the CI and /3 protons are inverted, i.e. the ,!lresonance is to low field of the c( resonance. Three of the systems, Thr 111, Thr IV and Thr VI1, show a significant high-field shift for one or more resonances, in particular the yCH3 resonance at 0.52 ppm for Thr VII. Apart from the seven threonine fir cross-peaks assigned above, there are 31 other cross-peaks in the region of the spectrum shown in Fig. 5 B. Analysis of this region (at many

636

A

A

r -1

-2

-3

-4

" 1

-

"

"

I

"

"

4

3

2

1

G/PPm

'

"

1.o

1.5 ..........................................

I

0.5

B

B

.

.

l

1.5

"

"

l

"

'

1.o

I

0.5

'

Nppm

Fig. 5. Expansions of 400-MHz DQF-COSY 'H-NMR spectra. Peaks marked NAcNcu arise from glycoprotein impurities. (A) The region of Val /$? and Leu y6 connectivities; (B) the region of Thr / j y and Ala z/3 connectivities.

Fig.4. 400-MHz DQF-COSY 'H-NMR spectra of BSAl (2 mM in 0.1 M phosphate pH* 6.95). (A) Aliphatic region, and (B) aromatic region. Resolution-enhanced 1 D spectra (as in Fig. 3) are shown for comparison. The labelling scheme for resonances is shown in Table 1.

different pH* values) failed to produce correlations for any of these cross-peaks either within this region, or to cross-peaks in other regions of the spectrum. Some cross-peaks may arise from spin systems that are not completely observed, e.g. degeneracy of threonine tl and /? resonances would produce an alanine-like system. However, this occurrence is uncommon and would not be expected to account for many of the 29 cross-peaks observed. Similarly, degeneracy of /3 resonances for some of the longer side-chains which might have @/a/?'

cross-peaks in this region (Arg, Glu, Gln, Leu, Lys, Met and Pro) would give rise to a single cross-peak as observed. Val and Ile might also have single connectivities in this region. However, all of the cross-peaks in Fig. 5 B are of reasonable intensity and, if they were ap connectivities, then we would expect to observe the corresponding /?y connectivities since Jxfl and Jay are expected to be of the same order of magnitude [21] and the y protons might be expected to have similar or more favourable relaxation properties than the a and /j protons. Thus, whilst recognising these other possibilities, these cross-peaks are tentatively assigned as Ala ap correlations. The spin systems Ala XI1 and Ala XI11 are notable for their high-field-shifted methyl resonances at 0.43 ppm and 0.84 ppm, respectively.

637

5.0

7.0

9.0

PH’

Fig.6. The pH dependence of thc chemical shifts of the a proton of Aspl. (See Table 1 , Fig. 4.) The solid line is a computer fit for a pK, of 7.8.

resonance [22]) might also occur in this region but a pair of peaks would be expected since Pro 6CHz protons are usually inequivalent. No such pairs of cross-peaks were seen. A similar argument can be applied to the methionine y f i connectivities where again the y resonance would be at the extreme of its expected chemical shift range (Met yCH P95 :2.80 ppm [22]). The possibility of isoleucine a,8 connectivities cannot be discounted (o! P5 3.06 ppm, P95 5.05 ppm; p P5 0.76 ppm, P95, 2.29 ppm) on chemical shift grounds but is again an extreme case. However, no further expected correlation was found for a possible Ile system. Use of the HOHAHA experiment suggests that these cross-peaks are part of Lys spin systems, although Arg with the y resonance at higher shift than the p resonance would also give rise to such a spin system. This type of Arg is rare, but does occasionally occur [23]. The pH behaviour of three of the resonances in 1D spectra provides further evidence for their assignment. Phenylalanine and tyrosine

Aspartate

Of all the cross-peaks seen in the aromatic region of DQFCOSY spectra of BSA at pH* 7.85, 7.56, 6.95 and 6.20 (e.g. A single strong AMX system is observed in the DQFFig. 4B) only two Phe systems can definitely be assigned. COSY spectrum of BSA (Fig. 4 at 4.22/2.74/2.61 ppm). Of the eight amino acids which can give rise to such a spin system, Many correlations of 2,6H/3,5H and 3,5H/4H cross-peaks at peaks for Ser are normally outside this region [22]. The pH one pH* were no longer found at a different pH*. Peaks titration behaviour of these resonances in 1D experiments assigned as Phe I were seen in spectra at each pH* and their (Fig. 6) gives rise to a pK, of 7.8 which is compatible with assignment was also confirmed by an extra cross-peak in the titration of a terminal amino group (i.e. Aspl). This assign- HOHAHA experiment at pH* 6.20. Phe I1 was observable ment is supported by the comparison with spectra of other only in spectra at pH* 7.85 and 7.56. At pH* 6.95 this spin system had either disappeared or was obscurred by overlap. mammalian albumins (vide infra). The remaining cross-peaks in the aromatic region (excluding those assigned to His below) are assigned ambiguously as Valine and leucine Phe/Tyr I - XIII. This ambiguity arises since Phe systems are Eight pairs of cross-peaks characteristic of Val by, By’ or frequently partially degenerate giving the appearance of a Tyr Leu y6, y6’ connectivities can be seen in Fig. 5A. For four of spin system, i.e. a single resolved cross-peak, the other crosspeak being unresolved from the diagonal even in a doublethese pairs, further cross-peaks were found (Table 1) whilst quantum-filtered experiment [24, 251. As the pH* is lowered the remaining four pairs of cross-peaks (Val/Leu I - IV) show from pH* 7.85 to 7.56 and 6.95 many of these systems shift. no obvious further correlation. The distinction between Val and Leu was made on chemi- Notably Phe/Tyr VIII and IX become overlapped and indiscal shift grounds by comparison with a statistical analysis of tinguishable at pH* 6.95. No spin system characteristic of Trp134 or Trp212 was observed chemical shifts in diamagnetic proteins [22] which assignable in the spectra of BSA. It is also notable that none gives 5th and 95th percentiles (P5 and P95 respectively) of of the aliphatic AMX spin-systems of aromatic amino acids 1.53 ppm and 2.33 ppm for Val B resonances and 0.80 ppm were observable in 2D spectra. and 1.93 ppm for Leu y resonances. In Table 1, spin systems Val I - IV are assigned as Val on the basis that the resonances assigned PCH are above P5 for Val ,8 and above for P95 for Histidines Leu yCH. They also show further correlation to an o!CH For BSA, four His C2H-C4H connectivities were clearly resonance that would require the rare occurrence of equival- visible, three from DQF-COSY spectra (e. g. Fig. 4 B), and an ence of ,Y, p’ and y resonances if they were assigned as leucine. additional weaker connectivity in relayed COSY spectra and Spin-systems Val/Leu I - IV are given ambiguous assignments HOHAHA spectra. These four systems are labelled His I since they are incomplete. The spin system labelled Val IV is IV (Table 1). notable for its high-field-shifted yCH3 resonances at 0.66 ppm Seven other resonances which show pH titration behaviour and 0.62 ppm. Confirmation of Val/Leu spin systems was consistent with assignment as His were also resolved in onesought from relayed COSY experiments, but no evidence of dimensional spectra. These are assigned as the C2H proton relay cross-peaks in the expected positions was found. resonances of HisV -XI, but no obvious C4H resonances were identified. The pH titration behaviour of these His protons will be discussed elsewhere (our unpublished work). Lysine and arginine Again, no corresponding AMX spin systems for His residues A group of overlapped cross-peaks (Fig. 4A) can be seen were visible in 2D spectra. to be centred at about 3.00/1.60 ppm. Eight of these can be assigned with some confidence. Only a few amino acid spin systems are likely to fall in this region, the most common Sequence-specific assignments Fig. 7 shows a comparison of the Asp CaH-CPH crossbeing Lys cfi and Arg d y connectivities. The Pro 6 y connectivity (at the extreme of the expected chemical shift range for the 6 peak region 2D COSY spectra for bovine (BSA), human

638

(a)

(b)

Bovine (D-T-H-)

Human (D-A-H-)

- 3.5

I

I

- 4.5 I ppm

3.0

2.0

3.0'

(d)

(c) Porcine (D-T-Y-)

13.5 I

- . -

'

ppm

I

'

2.0

Rat (E-A-H-)

- 3.5

I

I

.

i.0

m

. . .

.

. . . . . ppm

'

i.0.

t

- 4.5 *

4.5 3.0

ppm

.

i.0

Fig. 7. Comparisons of the aCH-/?CH* Asp/Glu regions of 500-MHz 2D 'H-NMR spectra of various albumins. (a) BSA (pH* 6.95, DQF-COSY), (b) HSA (pH* 7.40, DQF-COSY), (c) PSA (pH* 6.25, COSY) and (d) RSA (pH* 6.19, relayed-COSY). Assignments for Asp-1 (BSA, HSA and PSA) are shown. Cross-peaks for Aspl are missing in thc spectrum of RSA and are replaced by thosc for Glul with the assignments shown (the cross-peak at 2.25/3.95 ppm arises from the aCH-yCH2 relay). The N-terminal sequence is given in brackets above each spectrum (onc-lctlcr codes).

(HSA), porcine (PSA) and rat (RSA) serum albumins. For BSA, HSA and PSA an intense set of cross-peaks characteristic of an AMX spin system (CxH-CpH2) is seen in the range 2.1 -2.514.3 -4.0 ppm. As shown above, the titration behaviour of these resonances for BSA gives rise to a pK, of 7.8 typical of an N-terminal amino group, in this case Aspl. Furthermore these cross-peaks are absent for RSA (Fig. 7d) for which the terminal residue is Glul . The set of cross-peaks at 2.25, 2.02/3.98 ppm for RSA can be assigned to the latter residue. Fig. 8 shows a comparison of 2D COSY spectra for the Ala/Thr CH3-CH region. Both BSA and PSA give rise to intense doublets in 1D spectra at 1.14 ppm and 1.09, ppm respectively. Their assignments to Thr CH3 groups was confirmed by the presence of yCH3/bCH and PCH/crCH COSY cross-peaks and yCH 3/(xCHrelayed-COSY cross-peaks. For HSA and RSA the sharp doublet assigned to Thr in BSA (Thr 1) and PSA was replaced by a similar doublet at 1.28 ppm and 1.29 ppm, respectively, with 2D COSY crosspeaks at 1.28/4.25 ppm and 1.29/4.29 ppm (Fig. 8 b, d). No associated COSY or relayed-COSY cross-peaks were seen, consistent with assignment of these HSA and RSA resonances to Ala.

Comparison of the sequences of BSA [26 - 281, HSA [29], PSA [30] and RSA [31] shows that only at position 2 is the amino acid substitution pattern Thr -+ Ala observed for both HSA and RSA. Hence these resonances are assigned to Thr2 and Ala2 in the respective proteins. The sharpness of these resonances (relatively long T 2 values) is also consistent with their assignment to protons in a highly mobile region of the protein, as might be expected for the N-terminus. Similarly the imidazole C2H and C4H resonances of His3 can be assigned by comparison of 2D relayed-COSY spectra of BSA, HSA, PSA and RSA (Fig. 9). The spin system labelled as His I1 for BSA is also present in the spectra of HSA and RSA, but absent in that of PSA. The only sequence position which exhibits this substitution pattern is amino acid 3, which for PSA is Tyr. His I1 is therefore assigned to His3 of BSA, HSA and RSA. In the spectrum of PSA (Fig. 9c) a new, intense cross-peak assignable to Tyr3 is present a t 6.50/6.80 PPm, The coupling between the C2H and C4H protons of His is usually small ( z3 Hz) and the delayed-COSY sequence is often used to enhance the intensities of cross-peaks in such circumstances. In the present case, the relayed-COSY sequence gave a better signallnoise ratio for these cross-peaks.

639

(a)

(b)

Bovine (D-T-H-)

q

Human (D-A-H-)

- 3.5

13.5

- 4.5 88

'

..

It

#-Alan

(c) Porcine (D-T-Y-)

(d)

Rat (E-A-H-)

- 3.5

I '

I Thr-2

0. 1

'0 1 .I5

- 4.5

- 4.5

It

I'

PPm

0'.5

:

115

PPm

0.5

Fig. 8. Comparisons of the CH-CH3 regions of 500-MHz 2D 'H-NMR spectra of various albumins. (a) RSA (pH* 6.95, DQF-COSY), (b) HSA (pH* 7.40, DQF-COSY), (c) PSA (pH* 6.15, COSY) and (d) RSA (pH* 6.19, relayed-COSY). Assignments labelled refer to those in Table 1. Note the absence of Thr2 and the presence of an intense cross-peak (Ala2) in the spectra of HSA and RSA.

However, of the 11 resonances which we have tentatively assigned to C2H protons for BSA, only four give rise to relayed-COSY cross-peaks (Fig. 9a). All of these have sharp C2H resonances. Of the other resonances, for which assignments to amino acid type are listed in Table 1, only two unambiguous sequence-specificassignments can be made on the basis of their presence or absence in 2D spectra of BSA, HSA, PSA and RSA. These are Thr189 (labelled Thr IV) and Tyr155 (labelled Phe/Tyr V). When the occurrence patterns for Phe and Tyr are examined for all four proteins, only for Tyr155 does it match the absence of the Phe/Tyr V resonances only in the spectrum of HSA (see Table 2 legend). For Thrl89 the occurrence pattern for resonances is matched by the substitution patterns at positions 2 and 189 oCBSA. Since we have assigned resonance Thr I to Thr2 in BSA and PSA, then peak Thr IV is assignable to Thr189. Of the remaining resonances for which ambiguous assignments are possible on the basis of sequence comparisons (Table 2) the most useful is perhaps His I/His IV assignable to His59/His377.

DISCUSSION Albumin purity A large number of albumin preparations are available commercially and in our initial work we attempted to identify contributions from impurities to defatted albumin spectra. Other than ethanol (used in Cohn fractionation [33]), the major impurities detected were the mobile carbohydrate sidechains of serum glycoproteins such as a1-acid glycoprotein. These contain a high percentage of N-acetylglucosamine and sialic acid (up to 25% by mass) and their N-acetyls give rise to sharp 'H-NMRpeaks [34]. There appeared to be no significant, resolvable, contribution to spectra from the protein component of glycoproteins. Although DEAE-cellulose chromatography has been recommended as a standard method for the purification of albumin [8], Ikehara et al. [35] have shown that at least one glycoprotein impurity can be separated from albumin samples prepared by this method using a concanavalin-A - Sepharose column. In this study, chromatography on blue Sepharose was shown to be an effective method for removing glycoprotein

640

(a)

(b)

Bovine (D-T-H-)

Human (D-A-H-)

t

- 6.2 - 7.2 -

ppm

(c)

l

8.2

'

.

.

.

I

7.2

'

.

.

.

7.2

l

(d)

Rat (E-A-H-) I

7.2.

8.2

7.2

-

6.2

6.2

8.2

6.2

- 8.2

8.2

Porcine (D-T-Y-)

PPm

t-

b.2

pprn

8.2

7.2 ' '

'

'

'

-

6.2

-

7.2

-

8.2

'6.2

Fig. 9. Comparisons of the His C2H/C4H regions of 500-MHz 2D 'H-NMR relayed-COSY spectra of various albumins. (a) BSA (pH* 5.66), (b) HSA (pH* 6.19), ( c ) PSA (pH* 6.15) and (d) RSA (pH* 6.19). Assignments for His I, His I1 (assigned as His3), His 111 and His IV arc shown. Thc cross peak for His3 is missing in the spectrum of PSA for which the assignment of Tyr3 is shown.

impurities. The latter do not stain well with Coomassie blue on gels and NMR is a therefore useful method for detecting their presence. Heterogeneity at Cys appeared to have little effect on BSA spectra. We have demonstrated that NMR can be used for investigating disulphide formation at Cys34. Addition of a reducing agent, dithiothreitol, led to the displacement of cysteine, but no glutathione was seen. The samples of BSA contained about 0.6 mol SH/per mol BSA. We have also detected cysteine release from BSA on binding antiarthritic gold phosphine drugs [36]. In the latter case, gold binding to Cys34 appears to perturb the C2H peak of His3. Spectral enhancement techniques The application of a sequential exponential-sine-bell resolution-enhancement routine was found to be helpful for the interpretation of 1D spectra. This resolution-enhancement method appears to have been little used in the past [37, 381. The choice of an enhancement method for BSA spectra was dictated partly by the severe broadening at higher concentration (picsumably due to aggregation) and accompanying decreased thermal stability (gelling) at higher temperatures.

Large proteins such as albumin would be expected to give rise to broad 'H-NMR resonances with corresponding short T , values and it is therefore remarkable that about 80 sets of cross-peaks are observable in 2D COSY experiments. For example, human prealbumin (55 kDa) does not give rise to significant cross-peaks in COSY spectra [4], whereas spectra of urokinase (54 kDa) are remarkably well resolved [5]. For the latter, sharp resonances arise from highly mobile kringle and EGF domains [5]. Our spectra of BSA are indicative of the existence of highly mobile, yet structured regions, accounting for about a fifth of the total number of amino acids in a1bumin. The recent X-ray crystal structure of HSA at 0.4-nm resolution has confirmed the existence of three structurally similar domains. Each domain consists of two sub-domains linked by a long section of a-helix. There appear to be non-helical segments connecting the domains. This arrangement may allow considerable segmented mobility especially for each domain. Indeed, 'H-NMR spectra of tryptic (domains i I + 111) and peptic (domains 1 + 11) fragments [lo], and of isolated domain I (A. Tucker, P. J. Sadler, J . M. Viles, A.V. Quirk, unpublished results) have very similar features to intact albumin.

64 1 Table 2. Sequential assignments for BSA. An X indicates the absence of cross-peaks corresponding to the spin-system indicated for BSA. Thc numbering scheme is that for BSA; HSA and RSA have an additional amino acid inserted after residue 116, and HSA, RSA and PSA have an insertion after residue 155. It has recently been suggested that such an insertion (of Y) is also present in BSA [28]. The occurrence pattern of spin system Phe/Tyr V then matches that of both 155 and 155A Spin system

Occurrence pattern BSA

HSA

PSA

RSA

Aspl His I His I1 His 111

D H H H

D X H

H

D X X H

X X H H

His IV Val I Val 11 Val I11 Val IV Thr I Thr I1 Thr 111 Thr VII Thr IV Thr V Phe/Tyr 1V PheiTyr V Ala I1 Ala V Ala XI Ala XI11 Ala XV

H V V V V T

X X X V X X X X X X X Fly X X X X X A

X V X X X T X X X T X X Fly X X A X A

X X X X X X X) X) X) X T Fly FiY XI X) A

T T T T T Fly FiY A A

A A

A

Types of assigned spin systems Only a limited number of certain types of amino acid spin systems give rise to sharp NMR resonances and observable 2D cross-peaks in spectra of BSA. In part this arises from the normally more favourable relaxation behaviour of methyl groups (Ala, Val, Thr) and His imidazole protons. As can be seen from Table 1, the types of spin systems we have assigned are due mainly to hydrophobic amino acids (Ala, Val, Leu, Phe/Tyr), and these residues might be expected to be internalised in the protein structure [39]. In contrast, we were able to assign few cross-peaks to hydrophilic residues. For the 59 Lys and 23 Arg residues present in BSA, only nine spin systems were identified. Similarly, for the 60 Glu and 40 Asp residues present, only one spin system, Aspl, was assigned. These charged residues are normally on the surface of the protein but the existence of a large number of salt bridges in albumin has been previously proposed [40]. No Trp spin systems were assignable in BSA spectra, suggesting that the two Trp residues, Trp134 and Trp212, have only a limited mobility. While Trp212 is thought to be buried in the interior of the protein, Trp134 has been reported to be more solventexposed [8]. There do not appear to be any previous reports of the determination of the pK, of the terminal amino group of albumin. The value obtained here (7.8) is similar to the values reported for other proteins and peptides [41], and suggests that the amino-terminal Asp residue is in a freely solventaccessible environment. High-field shifted resonances Many of the high-field shifted resonances (below 0.8 ppm) can be assigned to methyl groups in close proximity to the

A

X

Sequence positions with matching mutation pattern

Sequential assignments

1,257,373 59,311 3 9,39,67, 105, 145,245, 336, 462, 508, 533 59,377 187, 568 227,343, 550, 575 77,229,291 227,343, 550, 575 2,189 121, 182, 304, 369,447 517,571. 579

Aspl

2,189 230,433,437 no match 155 60, 128,295, 308, 323 340,499 353,566, 582 no match 169, 192, 199, 509, 551

Thr189

His 3

Thr2

Tyr155

Number 81

"

0.4 0.3 0.2 0.1 0 -.I -.2 -.3 -.4 .-5 -.6 -.7 -.8 -.9

a/PPm Fig. 10. Plot of the distribution of the shifts of a proton resonances (observed-random coil (211) for assigned spin systems of BSA.

faces of aromatic rings. If Phe or Tyr is assumed to be the source of the ring current, then the maximum distances for these methyl groups from the centre of the aromatic rings can be estimated [42] to be 0.37 nm for Ala XII, 0.42 nm for Thr VII, and 0.47-0.49 nm for Thr 111, Val IV and Ala XIII. For Trp, or for a combination of two or more aromatic ring systems, similar shifts of methyl resonances could occur at larger contact distances. These interactions can be assumed to arise from highly structured but mobile regions of the protein. An analysis of the chemical shifts of the resonances assigned as a-proton signals reveals a strong bias to high field as shown in Fig. 10. Such a high-field shift of a resonances has been correlated with the presence of a-helical structure in proteins [43]. In contrast, the-@and amide protons'lin fl-sheets are often shifted to low field. A number of other studies

642 including CD [40,44,45], Raman spectroscopy [46], hydrogen exchange [47], secondary structure prediction [40,45,48] and model building [3] have indicated that albumin has a high helical content (50 - 60%) with a small amount of B-sheet (1 6%). Sequence-specific assignments On the basis of line-widths, chemical shifts and sequence comparisons, we have assigned all the non-exchangeable proton resonances for Aspl, Thr and the imidazole ring protons of His3 for BSA, and related resonances for HSA, PSA and RSA. There are a number of potential problems to be borne in mind when using such procedures. Firstly, changes in amino acid sequence between different species could lead to significant structural changes and, in particular, to changes in groups which have shielding or deshielding roles, e.g. aromatic rings. Secondly, the absence or presence of such a group is likely to lead to significant shifts of peaks for nearby protons. The sequences of these albumins can be aligned to give 200% conservation of the 35 Cys residues, with the following adjustments using the BSA numbering scheme: HSA and RSA have an amino acid insertion after residue 116, and HSA, PSA and RSA an insertion after residue 155. Such an alignment gives a 58% positional identity and 77% conservation (i.e. including substitutions for a similar amino acid type). The numbering scheme adopted is that of BSA (i.e. the amino acid residue numbers for HSA, PSA and RSA do not directly correspond to their amino acid sequences). For the four proteins used here, there are 14 positions in the sequence where a substitution occurs involving an aromatic residue. Of these, three (residues 36, 155 and 393) involve replacement of Phe by Tyr or vice versa, and at eight of the positions only one of the four species has a different amino acid (positions 30, 115, 139, 163, 209, 372, 375, 573). At position 18, only PSA has Tyr, BSA and RSA have His, whereas HSA has Asn. From the alignment of the published amino acid sequences, HSA has an insertion of Phe, and PSA and RSA have an insertion of Tyr after position 155, i. e. 155A (although a recent report [28] suggests BSA also has an extra Tyr residue at position 155A). At position 134, BSA and PSA have Trp, whilst in HSA and RSA this residue is Leu. The latter substitutions (155A and 134) might perhaps have the greatest potential for localised shift changes between the various species. Overall, therefore, for the albumins considered here, the number of amino acid changes where major perturbations of the NMR spectra might be expected, appears to be small and the method may well lead to reliable assignments. We have indde our comparisons of albumin spectra under as similar solution conditions (ionic strength, temperature, concentration and pH) and experimental N M R conditions as possible. Where differences in these were unavoidable, crosschecks with BSA were made, for which more complete sets of data (e.g. pH range, a variety of 1 D and 2D experiments) were available. Other workers have drawn up species comparisons to aid sequence-specific assignments of NMR resonances of proteins. These include cytochrome c [49], plasminogen kringle domains [50] and HPr (phospho-transfer proteins) [51]. The method has proved valuable where NOE data are limited, and where site-directed mutagenesis is unavailable. For albumin, 2D NOESY spectra are complicated by the broadness, low intensity and overlap of many cross-peaks (P. J. Salder, A. Tucker and J . Viles, unpublished results). Only about a fifth of the amino acid residues in albumin give rise to resonances

sharp enough to be observed in either 1 D or 2D COSY spectra. This provides a further limitation to the procedure. The proposed assignments for the resonances of three Nterminal amino acids are consistent with the following observations. Firstly the peaks concerned are noticeably sharper than most other resonances in the spectra of all four albumins. Both the N- and C-terminal regions might be expected to be the most mobile regions of the protein. In the present case the types of amino acid concerned correspond with only the Nterminus, and to a region which secondary-structure predictions suggest is random coil [48] (and A. Mills, P. J . Sadler, J. Thornton and A. Tucker, unpublished results). Secondly the pH titration behaviour of the Aspl resonances correspond to a pK, of 7.8 consistent with an N-terminal residue. Also the PCH resonance assigned to Thr2 reflects a similar pK, (our unpublished results). The only previous attempts to make sequence-specific assignments for 'H-NMR resonances of albumin are those of Bos et al. [lo]. They assigned the C2H resonances of His3 and His464 of HSA by a consideration of pH titration curves of HSA, fragments consisting of domains I 11 and I1 111, nicking experiments and Cu(I1) binding. The pH titration profiles of the His3 resonance for HSA [lo] and BSA are very similar. Their shapes and associated pK, values are markedly different from those expected for His in a random coil peptide [21] and those reported for His3 in the isolated 1-24 peptide of HSA (found to have a conformation closely related to random coil [52]). This suggests that, unlike the isolated peptide, the properties of His3 in the intact protein are detcrmined by interactions with other parts of the protein. The available 0.4-nm X-ray crystal structure of HSA [53] and structure predictions [48] (and our unpublished results) are also consistent with this suggestion.

+

+

Conclusion We have shown that a large number of resonances from a variety of types of amino acid residues in BSA can be resolved and tentatively assigned with the aid of resolution-enhanced 1 D and 2D NMR spectra. Comparisons between I D and 2D 'H-NMR spectra of albumins from bovine, human, porcine and rat sera have allowed assignments to be made for resonances from the three N-terminal amino acids ofeach protein. These assignments are based on the presence or absence of appropriate amino acid spin systems for the sequences concerned. The linewidths and pH dependence of these resonances are also consistent with their assignment to a mobile N-terminal region. Only about a fifth of the protein is detected by NMR under these conditions, suggesting that large portions are relatively immobile. However, the residues which are seen d o not appear to be clustered in a single segment of the protein and may thus provide useful reporter groups for the study of changes in albumin structure induced by pH, drugs etc. Our demonstration that the release of bound cysteine (as a disulphide at Cys34) in the presence of dithiothreitol can be detected by 'H-NMR should greatly aid attempts to define the heterogeneity of albumin samples. Also the presence of contaminating glycoproteins is readily detectable by H-NMR spectroscopy. The solvent accessibility of the amino terminus of albumin is relevant to its ability to bind to Cu(l1) and Ni(I1) ions. We have studied such binding and these results will be reported elsewhere. The combination of NMR data, structure predictions and the recent low-resolution X-ray crystal structure (of HSA)

643 allows a deeper insight to be gained into the structure and dynamics of albumin in solution. We will describe this in a subsequent paper.

23. Weber, P. L., Drobny, G. & Reid, B. R. (1985) Bioclzemistry 24, 4549-4552. 24. Holak, T. A. & Prestegard, J. H. (1986) Biochemistry 25, 57665774. 25. Klevit, R. E. & Waygood, E-. B. (1986) Biochemistry 25, 7774We thank the Wellcome Trust, Science and Engincering Rescarch 7781. Council, Medical Research Council, RhBne-Poulenc, Delta Biotech- 26. Brown, J. R. (1977) in Albumin structure, function and uses nology, and the Wolfsen Foundation for their support for this work. (Roscnocr, V. M., Oratz, M. & Rothschild, M. A., eds) pp. We are grateful to the Biomedical NMR Centre Mill Hill for the 27- 51, Pergamon, Oxford. provision of NMR facilities and to Drs J. Feeney, C. Bauer and T. 27. Reed, R. G., Putnam, F. W. & Peters, J., Jr (1980) Biochem. J . Frenkiel (Mill Hill) and Dr G. Williams (Univcrsity College London) 191,867- 868. for helpful discussions. 28. Hirayama, K., Akashi, S., Furuya, M. & Fukuhara, K. (1990) Biochem. Biophys. Res. Commun. 173, 639 - 646. 29. Dugaiczyk, A,, Law, S . W. & Dennison, 0. E. (1982) Proc. Nut1 Acad. Sci. USA 79, 71 -75. REFERENCES 30. Baldwin, G. & Wcinstock, J. (3988) Nucleic Acids Res. 16, 9045. I . Peters, T., Jr (1975) in The plusmaproteins (F. W. Putnam, ed.) 31. Sargent, T. D., Yang, M. & Bonner, J. (1981) Proc. Nu11 Acud. Sci. U S A 78, 243 -246. 2nd edn, vol. I, pp. 133-181, Academic Press, New York. 2. Peters, T., Jr & Reed, R. G. (1978) in Proc. FEBS Meet. 50, 32. Bax, A. & Freeman, R. (1981) J. Mugnet. Resonance 44, 542544. 11 -20. 3. Brown, J . R. & Shockley, P. (1982) in Lipidprolein interactions 33. Cohn, E. J., Strong, L. E., Hughes, W. L., Mulford, D. J., Ashworth, J. N., Melin, M. & Taylor, H. L. (1946) J . Am. (P. C. Jost & 0. H. Griffith, cds) vol. 1, pp. 25-68, Wiley, Chem. Soc. 68,459 - 475. Ncw York. 34. Clamp, J. R . (3975) in Theplusmaproteins (F. W. Putnam, ed.) 4. Reid, D. G. & Saunders, M. R. (1989) J . Biol. Chem. 264,2003 vol2, pp 163 - 21 1, Academic Press, New York. 2012. 5. Oswald, R. E., Bogusky, M. J., Bamberger, M., Smith, R. A. G. & 35. Ikehara, Y., Oda, K. & Kato, K. (1977) J. Biochem. (Tokyo) 81, 1293- 1297. Dobson, C. M. (1989) Nature 337, 579- 582. 6. Veitch, N. C. & Williams, R. J. P. (2990) Eur. J . Biochem. 189, 36. Ni Dhubhghaill, 0.M. & Sadler, P. J. & Tucker, A. (1992) J . Am. Chem. Soc. 114, 1118 - 1120. 351 - 362. 37. De Marco, A. & Wiithrich, K. (1976) J . Magnet. Resonance 24, 7. Carter, D. C., He, X., Munson, S. H., Twigg, P. D., Gernert, K. 201 - 204. M., Broom, M. B. & Miller, T. Y. (1989) Science 244, 119538. Gueron, M. (1978) J . Magnet. Resonance 30, 51 5 - 520. 1198. 39. Kyle, J. & Doolittle, R. F. (1982) J. M o f . Biol. 157, 105-132. 8. Peters, T., Jr (1985) Adv. Protein Chem. 37, 161-245. 9. Labro, J. F. A. & Janssen, L. H. M. (1986) Biochim. Biophys. 40. Foster, J. F. (1977) in Allmmin structure, function and uses (Rosenocr, V. M., Oratz, M. & Rothschild, M. A,, eds) pp. A C ~ 873, U 267-278. 53 - 84, Pergamon Press, Oxford. 10. Bos, 0. 3. M., Labro, J. F. A., Fischcr, M. J. E., Wilting, J. & 41. Martell, A. E. & Smith, R. M. (eds) (1982) in Critical stability Janssen, L. H. M. (1989) J . Biol. Chem. 264,953-959. constunts, vol. 5 , Plenum Press, New York. 31. Oida, T. (1986) J . Biochem. (Tokyo) 100, 1533-1542. 42. Johnson, C. E. & Bovey, F. A. (1958) J . Chem. Phys. 2Y, 101212. Era, S., Itoh, K. B., Sogami, M., Kuwata, K., Iwama, T., Ydmada, 1014. H. & Watari, H. (1990) Int. J . Peptide Protein Res. 35, 1-11. 13. Travis, J., Bowen, J., Tewkesbury, D., Johnson, D. & Pannell, R. 43. Williams, R. J. P. ( 3 989) Eur. J . Biochem. 183,479 - 497. 44. Sjoholm, I. & Ljungstedt, I. (1973) J . B i d . Chem. 248, 8434(1976) Biochem. J . 157, 301 -306. 8441. 14. Grassetli, D. R. & Murray, J. F. (1967) Arch. Biochem. Biophys. 45. Reed, R. G., Feldhoff, R. C., Clute, 0. L. & Peters, T., Jr (1975) 119,41-47. Biochemistry 14, 4578 -4583. 15. Aue, W. P., Bartholdi, E. & Ernst, R. R. (1976) J . Chem. Phys. 46. Chen, M. C. & Lord, R. C. (1976) J . Am. Chem. Soc. 98, 99064,2229 - 2246. 992. 16. Rance, M., Ssrensen, 0. W., Bodenhausen, G., Wagner, G., Ernst, R. R. & Wiithrich, K. (1983) Biochem. Biophys. Res. 47. Benson, E. S., Hallaway, B. E. & Lumry, R. W. (1963) J. Biol. Chem. 239,122 - 129. Commun. 117,479-485. 17. Bolton, P. H. & Bodenhausen, G . (1982) Chem. Phys. Lett. 89, 48. McLachlan, A. D. &Walker, J. E. (1978) Biochim. Biophys. Actu 536,106-111. 139-144. 49. Gao, Y . , Lee, A. E. J., Williams, R. J. P. &Williams, G. (1989) 38. Chen, R. F. (1967)J. Biol. Chem. 242, 173-181. Eur. J . Biochem. 182, 57-65. 19. Bell, J. D., Brown, J . C. C., Nicholson, J. K. & Sadler, P. J. (1987) 50. Petros, A. M., Gyenes, M., Patthy, L. & Llinas, M. (1988) Eur. FEBS Lett. 215, 31 1 - 31 5. J . Biochem. 170, 549 - 563. 20. Breg, J., Kroon-Batenburg, L. M. J., Strecker, G., Montreuil, 51. Kalbitzer, H. R., Hengstenberg, W., Rosch, P., Muss, P.. J. & Vliegenthart, J. F. G. (1989) Eur. J . Biochem. 178, 727Bernsmann, P., Engelmann, R., Dorschug, M., Deutscher, J. 739. (1982) Biochemistry 21,2879-2885. 21. Bundi, A. & Wuthrich, K . (1979) Biopolymers 18,285-297. 52. Laussac, J.-P. & Sarkar, B. (1984) Biochemistry 23,2832-2838. 22. GroO, K.-H. & Kalbitzer, H. R . (3988) J. Magnet. Resonance 76, 53. Carter, D. C. & He, X. (1990) Science 249, 302-303. 87-99.