the study of the structure akd denaturation of heme ... - ACS Publications

School of Chemistry, University of Minnesota, 111 inneapolis, Minnesota ... Department, Minnesota Mining and Manufacturing Company, St Paul, Minnesota...
0 downloads 0 Views 929KB Size
May, 1961

STRUCTURE A S D n E S - 4 T U R A T I o N OF

HEMEPROTEINS

837

THE STUDY OF THE STRUCTURE AKD DENATURATION OF HEME-PROTEINS BY NUCLEAR MAGNETIC RELAXATIOS BY R. LUMRY,H. MATSUMIYA, School of Chemistry, University of Minnesota, 111inneapolis, Minnesota F. A. BOVEYAND Central Research Department, Minnesota Mining and Manufacturing Company, St Paul, Minnesota

A. KOWALSKY Department of Physiological Chemistry, School of Medicine, University of Minnesota, &finneapolis,Ailinnesota Received December 8, 1960

The molar relaxivities toward water protons of the ferric and ferrous forms of simple heme-complexes, myoglobin and hemoglobin have been determined by the direct method. An analysis of the data is given which demonstrates that the controlling correlation time for relaxation by heme-proteins and ferrous heme-complexes is the spin-lattice relaxation time of the iron ions. The experiments thus provide information about this time and about ligand-field effects which influence it. They do not directly measure the position of the iron ions in the protein. The use of relaxivities to provide information about protein denaturation is discussed with sample data. A simple solution for relaxation in a two phase system following saturation is given to demonstrate the use of the methods of chemical-relaxation spectroscopy for nuclear-relaxation problems.

Because of the large size of the electron magneton, paramagnetic substances are very effective in accelerating transitions among nuclear spin states. This effect is readily measured in the longitudinal relaxation time, TI,and in the transverse relaxation time, Tz, and has been exploited to provide information about rates of ligand exchange and electron exchange in complex ions as well as in other useful and interesting ways.' I n principle, the method might be applied to heme proteins to provide information about the position of the paramagnetic iron ions in these compounds, the extent of protein hydration, the spin-lattice relaxation time for the spin vector of the ion and the rates of exchange of protein-bound water molecules. In addition, studies of this sort can provide information about the denaturation of proteins. Although the subject has been investigated by Davidson and co-workers,2 by W i ~ h n i a ,and ~ by the present author^,^ its development is thus far somewhat crude, partly because of inadequacies in the theory but largely because of a lack of suitable experimental data. Previous work has been confined to ferric forms of hemoglobin and myoglobin. In this paper, results on the ferrous forms of these proteins and on reduced hemin will be presented, together with new data for the ferric compounds. It is not yet possible to offer an unequivocal and quantitative interpretation of these results; nevertheless, even a t this stage, some interesting deductions can be drawn, not only concerning the compounds studied, but also concerning the power of the method.

method (method I of Bloembergen, et a1.6), using a Varian V-4300-2 40.00 Mc./sec. spectrometer and Brush recorder. The samples were first saturated by using high r.f. power for 5-10 see.; the power then was reduced quickly to levels corresponding to only slight saturation. T I values were estimated from the slopes of semilog plots of the regrowth of the water proton signal, and were reproducible within 15%. Horse hemoglobin was repared by three recrystallizations from aqueous ethanol; Ruman hemoglobin was recrystallized twice by a modification of Drabkin's salt method which will be described elsewhere.B Hemin was prepared by the method of Matsumiya and Lumry; and reduced with KBH4 or with hydrogen using Pd-on-charcoal. The data for native protein solutions are presented in Table I.

Theory Relaxation depends on the magnitude of the magnitude of the magnetic dipoles, their number, the distance from these to the measuring protons, and the relative motion of the magnetic dipoles. Since spontaneous transitions among nuclear spin states occur very infrequently, all transitions which are observed are induced only by those magnetic fields at the measuring nuclei which vary at the resonant frequency. In treating the problem thus presented, Bloembergen, Purcell and Pounds analyzed the several forms of relative motion of the magnetic dipoles, ie., rotational and tra~slational motion, by Fourier analysis to determine the amplitude of the resonant component thus generated for various values of the parameters. The general expression for longitudinal relaxation has been modified by Solomon7 and extended to include a scalar term suggested by Bloembergen.8 The latter term need not concern us here, since for ferric compounds it is small and for ferrous compounds, Experimental The values of TIfor water protons under the influence of although we have none of the information necesthe paramagnetic substances were measured by the direct sary to determine its importance, its presence can have but a minor influence on the arcgments pre(1) "High Resolution Nuclear Magnetic Resonance," by J. A. sented below. Pople, W. G. Schneider and H. J. Bernstein, McGraw-Hill Book Co.. For our present purposes we may express the Toronto, 1952, chapter 9; J. E. Wertz, Ann. Rev. Phya. Chem., 9,93 longitudinal relaxation rate constant as9 ( 1958). (2) N. Davidson and R. Gold, Biochzm. Biophys. Acto, 26, 370 (1957); H. Kon and N. Davidson, J . Molecular Biol., I, 190 (1959). (3) A. Wishnia, J . Chem. Phys., Sa, 871 (1960). (4) F. A. Bovey and R. Lumry, Discussions of the Faraday Society, "Energy Transfer with Special Reference to Biological Systems," 1960 p. 247.

( 5 ) N. Bloembergen, E. M. Purcell and R. V. Pound, Phys. Reo., 73, 679 (1948). (6) H. Matsumiya, to be published. (7) I. Solomon, Phgs. Rea., 99,559 (1955). (8) N. Bloembergen, J . Chem. Phya., 27, 572 (1957).

TABLE I MOLARRELAXIVITY R FOR AQUO-FERROUS A K D AQCO-FERRICIoss, FERROA N D FERRIHEMIN, A N D SATIVE HEMEPROTEINS IN WATERAT 25' Compound

R,1. set.-' mole-'

Concn. (based on Fe)

Conditions

10-3-10-6 M

pH 2 . 0 pH 2 . 0 I n 0 2 A' KOH; reduced with HZ and Pd on charcoal or with KBH, In 0 . 2 ;V KOH I n 0.10 AT KC1, pH 7 . 2 In 3 x 10-3 AT KC1, pH 7 . 2 S o salt, pH 7 . 2 X o salt, pH 7 . 2 I o salt, pH 7 . 2

Fe( H Z O ) ~ + ~ Fe( H20)6+ ++ FPrroheniin

1-2 x 10-2M

Ferrihemin Horse hemoglobin Human hemoglobin Horse methemoglobin Whale myoglobin Whale metmyoglobin

1-3 x 10-4~i 3-6 X 31 5 x 10-3 M 45 x 10-3~ i x 10-3~ 1.5 x 1 0 - 3 ~

10-4-10-6 M

We distinguish the calculated rate constant k from the experimental quantity 1/Tl; k is the rate constant for the reaction: spin state l+spin state 2 , or for the reverse. The effective paramagnetic moment pee is that calculated from static susceptibility measurements; 71 is the nuclear magnetogyric ratio and us is the electron Larmor frequency. The interdipole distance is r and the correlation time is rC. This expression is not correct for proton-proton relaxation which, together with the cffect of oxygen, is treated as a corrective quantity measured in control experiments. A further corrective quantity is the "diamagnetic" relaxing effect due to the protein but not to be attributed to the protein-bound paramagnetic ions. This correction was established by measuring T1 for oxyhemoglobin and oxymyoglobin solutions, since these forms of the proteins are diamagnetic. This effect probably is due to the slowing down of the motion of the water molecules when they are in the hydration shell of the proteinlo; we shall consider it more fully in a later publication. Theoretical expressions for T~ hare been discussed elsewhere."ll There are usually several sources of motion and, since the times of these motions combine in reciprocal fashion, the shortest time will dominate T ~ . In considering the effects of ferrous and ferric ions, particular account must often be taken of the fact that the electron magnetic moment may undergo transitions among its available states more rapidly than the atomic motions take place. In this case, the electronic transition time T~ will dominate rC. A solution of ferric ions provides two environments or phases for nuclear relaxation: phase A is the solvent in which relaxation by paramagnetic ions may be controlled by translational motion or hy electron spin-lattice relaxation; phase B is the water bound tightly to the ferric ion. Broersma's equations11 show that for sufficiently long re the relaxation rate in A will be a small fraction of that in B. This probably is true, but at present it does not appear that equation 1 can be used to predict (9) R. A. Bernheim, T. H. Brown, H. S. Gutowsky and D. E. Woessnor, z b d , 30, 950 (1959). (10)F. A. Bovey, paper presented t o Div. of Biol. Chem., 138th Meeting of American Chemical Society, New York, Sept. 11-15, (1960). (11) N. Broerama, J . Cham, Phys,, 24, 650 (1066): 87, 481 (1957).

900 f 100 11500 3z 1000

21 f 5 2170 f 380 11 f 4 15 i: 4 200 i:20 250 f 40 1200 32 100

unambiguously the absolute values of proton relaxation rates in solutions of transition metal ions, mainly because of complications dependent on the rate of migration between phases. The general problem of migration among phases has been treated using a stochastic method by Zimmerman and Brittin. l 2 (-4 straight-forward mass law derivation is given in the Appendix.) It is found that if the exchange rate is fast with respect to the rate of relaxation in both phases, then all protons average their environments and a simple limiting condition applies where na, nB and nT are numbers of protons in phase A and phase B, and the total number, respectively. If, on the other hand, relaxation is much more rapid than exchange, one finds

In the latter case, there are two distinct relaxation times. If the rate of migration of protons through the inner ligand shell of ferric ion were dependent on the rate of water exchange in this shell, equation 2b would apply since the exchange time is of the order of 10-3 sec.11J3 and relaxation in the ligand shell will be much faster.14 Migration may take place through the hydrolysis reaction

+

khyd

Fe f3(H20),0H-

Fe +3( H Z O ) ~ HzO k'hyd

+ H30

+

(3)

which must have a forward rate constant of about lo8 liter mole-' sec-1 1 3 ; in this case, (2a) would apply. Many common examples will have intermediate exchange rates and will require a fuller treatment not confined to these two limit's. Experimental Results and Interpretation 1. Aquo-ferrous and Aquo-ferric Ions ; Ferroand Ferrihemh-If we assume the rapid exchange limit indicated above and take T, for aquo-ferric ion as 2 X sec., an estimate based on King'sI5 (12) J. R. Zimmerman and Tv. E. Brittin, J . Phys. Chem., 61, 1328 (1967). (13) H. Wendt and H. Strehlow, 2. Elektrochem., 64, 131 (1960); J. F. Below, R. E. Connick and C. P. Copgel. J . Am. Chem. Soc., 80, 2961 (1958). (14) M. Eigen, pereonal communication. (15) J. King, thesis, California Institute of Tecnnology, 1958, quoted in ref. 3.

M q , l9(il

S T R U C T U R E AND

DENATURATION OF HEMEPROTEINS

lower limit of 1.33 X sec., then the appropriate correlation time T~ is Ttrans (translational) in A and Trot (rotational) in B. We express our rebults in terms of the molar relaxivity which for this caqe for ?mall concentrations of paramagnetic io11 is givcm by

T L B being

the number of moles of protons bound in phase B and p a designation of paramagnetic effect. The calculated value, Rcalcd,must be brought, into agreement n-ith the experimental value

The last three terms are the empirical corrections previously mentioned and d refers to the diamagnetic effect of the protein.I6 It is readily shown that k ~ is, a small fraction of [mi(2 X 35.5)] X LB,. C4ng the measured linear diffusion coefficient of aquo-ferric ion17t o provide a, the radius of the hydrated shell can be estimated from the measured linear diffusion coefficient of aquo-ferric ion17 using Stokes’ equation. This valuc. is then uhed to estimate Trot by means of Stokes’ equation for rotational diffusion, yielding a value of 5 X 10-11 sec. U.$iig Rroersma’s form of equation 1 with r = 2 . 2 A. and a degree of hydration equal to 6, we find Rcalcd = 1.2 X io41. sec.-! mole-’ (water molecules in the second hydration $hell are essentially uninfluenced becaux of the r6factor). This agreement with the experimental value of 1.2 x 104 1. sec.-l mole-! (Table I) must be regarded a 4 largely fortuitous hilice r is not well known and l i ~ ,(approximately 0.2 X lo4 1. set.-' mole-’) has been ignored. This experimental value niay be used as a teqt to see if the fast-exchaiige limit assumed is justified, for if so exchange must be fast with respect to R c a l c d corrected for the higher coiicentration of bulk water, z.e., 1.2 X lo4 X 55.5; 6z105 sec.-l. This is certainly too fast for exchange of water molecules, as has been mentioned, hut is less than the expected order of magnitude for the hydrolysis exchange rate in lo-? mole 1 . ~ ’ (16) T h e complete foim of RelDti for heme protein solutions is

in n-hich n.4 is the niiniher of moles of water protons in phase A, etc. and nT is t h e total nnnibrr of moles of water protons; ( l / T i ) i s , o corrects f o r t h e spin-lattice relasation in pure water. ‘Io the extent t h a t this iesitlts from intermolecular dipolar interaction it must he corrected for reduced water concentration a t high protein concentration; (1/ is the rjriantity dPterniined for pure phase A a t [Olz = 1 M . Similarly, (1/Ti)02,n is measured for phase R , hilt will differ little from (l/TL)oltj.. ( I / TL)l,rot d , % .is the diamagnetic rontribution. It will depend on t h r re1atix.r x-aliies of n.4 and n g . Further treatment will require t h a t thc protpin hydration water be treated as a separate phase; m p r O t . is the niimber of iron atonis per protei11 niolecu~e. F o r separate phases this cont.rihrition would be

(17) R . 9. Robinson and R . H. Stokes, “Electrolytp Foliitions,” Academic Press Tnc , New E’ork. 1955,p . 450.

839

acid, our experimental conditions for the ayuo ions. This is shown in the following way for inner shell ligands: the constant k’hyd for the diffusion controlled back reaction of ( 3 ) should be about 7 X lo9 1. sec.-l m o k 1 . 1 3 Since pKa = 2.7 for aquoferric ion, the second-order forward rate constant k h y d appropriate to ( 3 ) is ( 2 X 1 0 - ~ X 7 X 109),’55.5 = 2.5 X lo5 1. set.-' mole-‘ and this number multiplied by the number of water moles present yields 1.4 X lo7 set.-'. The later number controls the exchange rate zia (3) so that the assumption of the fast-exchange limit is justified. Although R e x p t l may be dependent on pH above 5, since the hydrogen-ion concentration mill control the extent of deviation from the fast-exchange limit, below pH 5 R should not be dependent on hydrogen-ion concentration. Wishnia3 finds that in 0.1 Jf acid, Rexptl= 0.93 x lo41. R C C . - ~ mole-I, which is in fair agreement with OUT experiment a1 value. B r ~ e r s m a ’ finds ~ results at higher pII values which may indicate a deviation from the fast-exchange limit, although he does not interpret them in this way. For aquoferrous ion pKa E 917and exchange z~ia hydrolysis is negligible. However, the rate of exchange of inner-shell mater molecules is now considerably faster-rate constant lo6 set.-'than for ferric ions;. The rapid exchange limit appears justified in this case, but as is well kiion-n (Gee also Table I), the ferrous-ferric molar relaxation ratio is much less than would be expected from static susceptibilities. This effect cannot be attributed to an inadequate exchange rate and hah been attributed to a w r y small T ~ ,expected 011 the basis of ligand-field c~oiisiderations and $pinorbital coupling in the divalent ion. For rapid exchange T, can be estimated roughly from T,(ferrous) =

3 p2(ferric) ~(ferric) X -1.5 p2(ferrous) r6(ferrous) +(ferric)

Roxptl(ferrou,S) ReXpti(ferric)

the factor 3 ’4.5 arising from the brac;keted factor in equation 1. Taking rferric = 2.2 A., Tferrous = 2.5 A., /*ferrous = 5.4 Bohr magnetons. and Trot (ferric) = 5 x 10-l’ sec., we find that T~ 4 . 3 X sec., which is not unreasonable for aquoferrous ion but cannot be appropriate for the ferrous heme compounds as judged by the data of Table I. To see this more fully, let 11;. compare first the ferric compounds. Since the effective moment and T d u e s are the same for ferrihemin and aquoferric ion, the molar relaxivity for the ferrihemin, again estimated on the basis of phase B contribution, should be kBplirem,n)

n (hemin) n (ferric)

= ---

hemin) x a3( __a3(ferric)

kBn(ferrlc)

the rotational diffusion coefficients Drat being assumed to be the same for both, thus balancing unknown factors of charge agaiiist unknown effects of size and shape. The proper value of a ( h e m l n ) is not known but it cannot be much larger than the hydrodynamic value of 3.6 8.for ferric ion. In 0.2 N KOH (as in our experiments) one of the two (18) L G. Sillen, Quart. Ret.. 13, 146 (1959). ( 1 0 ) S. Broersme, J . Chem. P h y s . , 30, 707 (1959).

R. LUMRY,H.

840

R'fATSUMIY.4,

F. A. BOVEY AND A. KOTVALSKY

solvent ligands probably is ionized so that n = 3, and we have: k~~ = '/4 X 1.2 X lo4 = 3000 1. set.-' mole-', which is in satisfactory agreement with the experimental value. 2. Heme Proteins.-The factors determining the relaxation for the heme proteins are somewhat different from those governing the behavior of the smaller molecules just considered, since Trot is greater than T ~ . Hence, 7, is T ~ . The proteins introduce a third phase i n the hydration water, but since the exchange rate between this water and the solvent water is probably much greaterz0 sec.) than the relaxation rate in either phase, it is not uqeful to recognize the new phase as separate. The total relaxation in this combined phase is calculated from

in which (L+) refers to the average distance from a given water molecule at the origin to the ferric ion buried in the protein at a fixed distance from the water molecule; Npe is the number of ferric ions per ~ m . regardless ~ , of whether the protein has 1, 2 or 4 such ions per molecule;21 rw is the water radius and rp the protein radius. The distance from heme iron to the center of the protein sphere is b. Treating the protein as a sphere rp for fixed T and averaging over the khell at r

+

(L-6) =

so'

sin ede

[(r

+ r d 2 + b2 - 2b(r + rP) Jo"

1

z (1

+

S)(r,, + 1

TP2

b3

COS

n-

(rp

e 7

sin ede

1 rw - b)' +

+ rw - b )

- $In

rp

(rp

+

rW

+ rw - b )

Since the density of water in t'he surface layer will not differ greatly from the bulk density, this neglect of any distinction between bulk and hydration water does not appear to be unreasonable. Myoglobin, shown by Kendrew, et a1.,23to be a lozenge-shaped body 45 X 35 X o 2 5 A.,can be treated as a sphere of radius 18 A. wit'hout any error greater than those of the other approxima(20) T h e possibility of chemical exchange of protons with basic protein groups is for t h e present ignored. I n any event this effect will be small in most solutions of paramagnetic proteins. T h e matter will be considered in a subsequent publication. (21) We will throughout neglect a n y magnetic interaction among t h e heme groups of t h e same protein molecule since these can hardly he significant if the distances are a s large a s shown by the X-ray diffraction studies of Peruta a n d co-workers.** Heme-heme interactions mediated by the protein fabric m a y influence t h e relaxation effectiveness of the iron ions, b u t this will not h e a variable when all heme gronrJs are in the same state, Le., all oxygenated or deoxygenated. (22) 31. F. Peruta. e t a!., Nafure, 186, 416 (1960). 123) .J. C. Kendrew, ibid., 182, 704 (1958); J. C. Kendrew, et al., ibid.. 186. 422 (1960).

Vol. 65

tions involved. If the heme lies with its plane along the radius of the protein and Jvith one edge on the protein surface, b = 13.5 A. and taking rw as 1.5 A.,k~~ = 50. We conclude that R e x p t l , equal to 1200, must be due primarily to phase B. In this phase again only the single water molecule or hydroxyl ion in the sixth ligand position will make a major contribution because of the r-6 factor. Thus, "crevice" water may be ignored. Although there are no ligand-field effects in ferric ion complexes the nature of the bonding in the complex will influence the magnetic properties of the iron ion. Thus T~ may vary with the kind of ligand. For example if T~ were the same for aquo-ferric ion and metmyoglobin and T~ = T~ for the latter, the expected value of the relaxation rate constant for the protein would be given by

3k g , nT

n(met Mb) n( aquof erric ) 2 x 10-10 (met Mb) = 5 x 1 X ~1 lX

x

~

~(aquoferric) ~ ~

1.2 X lo4 = 8 X lo3

The experimental value, R e x p t l , is 1200 and the ratio 1200/8000 may well be a measure of the decrease in electron spin-lattice relaxation time in metmyoglobin relative to aquo-ferric ion. On the other hand the small value of the observed relaxivity may also be due in part to a slow rate of exchange or to a decreased effective susceptibility related to the anisotropy of the magnetic susceptibilityz4 (the observed static susceptibility has the expected value for 5 unpaired electrons but the situation is known to be complicated by an orbital contribution.) hIethemoglobin is less adequately treated by the spherical model than metmyoglobin. According to X-ray studiesz5oxyhemoglobin consists of four myoglobin-like molecules arranged roughly tetrahedrally so that there are appreciable voids for water in the structure. Even if this structure, which was determined for oxyhemoglobin in the crystalline state, is correct for deoxyhemoglobin and methemoglobin in solution, the spherical model is inadequate. Howeyer, l i ~ ,for myoglobin was shown to be small eyen for extremely close positioning of the heme iron to the protein surface (3 A.); for methemoglobin it mill be even smaller because of the increased value of rp and the fact that some water of hydration from the inner faces of the separate myoglobin units is displaced to larger distances in hemoglobin. We continue to ignore interaction among heme units and any detailed consideration of the anisotropy of the magnetism in the heme unit. The latter factor may be of importance in reducing the effective value of the magnetic moment in the direction of the nearest water molecule, but there is little reason for its serious conFideration at present. It would appear then that for methemoglobin (24) D.J. E.Ingram and J. E. Bennett, Faraday SOC.Disc., No. 19, 140 (1955): D.J. E. Ingram. J. F. Gibson a n d J. F. Perutz, N a t u r e , 178, 906 (1956); J. F. Gibson a n d D. J. E. Ingram, Sbzd., 180, 29 (1957): J. F. Gibson, D. J. E. Ingram and D. Schonland, Faraday SOCDisc, No 26, 72 (1958).

t

l

May, 1961

STRUCTURE AND D E N h T U R A T I O N OF

must again govern the relaxation. This conclusion is at variance with that of 'IVishnia,3 who assumed the same rc for aquo-ferric ion and methemoglobin. This we believe to be improbable, as , on shown in the previous discussion. k ~ depends rw - p , whereas k g , does not. the distance r p Therefore, k g , should be the same for metmyoglobin and methemoglobin if all factors other than the distance rp - rw - p are the same. But for methemoglobin, Rexptl is only 200 (Table I) and the factors responsible for the deficiency in Rexptl for metmyoglobin muqt again be responsible. The most rea5onable of these might appear to be a failure of rapid exchange, but the very rapid rate of oxygen uptake characteristic of hemoglobin (at least 6 X lo6 1. see.-' mole-', according to Roughton and which must be a lower limit for the rate of m t c r exchange) makes this seem rather improbable. Thus attention must be focussed on the anisotropy of the heme g-value and on re. It appears that although 7, may be constant within about an order of magnitude for the ferric compounds, it is not constant within a much smaller factor. Returning now to the ferrous compounds we find a similar situation exists. The structural similarity between ferrohemin and ferrihemin is so close that they might be expected to have nearly the same ratio of molar relaxivities as the aquo ions, 900/11500, or even higher since n is probably larger for the ferrohemin. The experimental ratio is 2112470 and must indicate a further decrease in in ferrohemin if n remains unchanged. For all these compounds, k ~ will , be completely negligible hecause of the small values of 7, (