Electron paramagnetic resonance evidence for a distinction between the two iron-binding sites in transferrin and in conalbumin.

Abstract The electron paramagnetic resonance spectra of aqueous solutions of both transferrin and conalbumin have been studied as a function of increasing sodium perchlorate concentration. Characteristic changes are observed which provide direct evidence for a distinction between the two metal-binding sites under certain conditions in the naturally occurring iron derivatives of each protein. This may be of physiological importance in the case of transferrin. Differences in the behavior of the two proteins were also observed.

From the Inorganic Chemistry Laboratories, Department of Chemistry, Imperial College of Science and Technology, London SW7 2AY, England SUMMARY The electron paramagnetic resonance spectra of aqueous solutions of both transferrin and conalbumin have been studied as a function of increasing sodium perchlorate concentration.
Characteristic changes are observed which provide direct evidence for a distinction between the two metal-binding sites under certain conditions in the naturally occurring iron derivatives of each protein.
This may be of physiological importance in the case of transferrin. Differences in the behavior of the two proteins were also observed.
Conalbumin, from avian egg white and transferrin, obtained from pooled human blood serum, are each capable of binding 2 atoms of iron per molecule to form similar, very stable high spin Fe(II1) complexes which are amenable to study by electron paramagnetic resonance.
This technique is sensitive to small changes in the electronic environment about the paramagnetic atom and much useful information about the properties of these molecules has been obtained by EPR' spectroscopy (1,2).
There is continued interest in the problem of whether or not the two iron-binding sites in each protein have the same properties. The iron-binding constants for each site are nearly identical (1,3). However, Aisen et al. (4) have described EPR experiments which demonstrate a heterogeneity of the sites in the chromium derivative of transferrin. Evidence for dissimilar sites in a bicarbonate-free iron complex of transferrin (5) was not substantiated (6). Ryall (7) suggests by kinetic studies that the two sites in normal transferrin differ in their reactivity towards thioglycollic acid. Fletcher and Huehns (8) have shown a difference between the two sites under more physiological conditions in terms of the rate of uptake of iron by reticulocytes.
When dialysis against sodium perchlorate was used (6) to remove traces of chelating agents from the proteins as purchased, two EPR spectra were distinguished for each protein.
We desig- nate these as broad (B) and sharp (S), and note that the broad signals, as are the sharp, are extremely similar for the two different proteins.
Indeed, we regard them as indistinguishable for the purpose of this paper, in which we show by a study of the (B) and (8) signals under various conditions, both a distinction between conalbumin and transferrin, and differences between the two iron-binding sites in each protein.

EXPERIMENTAL PROCEDURE
Type I apoconalbumin was from the Sigma Chemical Company; lyophilized apotransferrin was prepared by Behringwerke and purchased from Hoechst Pharmaceuticals, Ltd. Aqueous solutions of the apoproteins were purified by dialysis veraua 0.1 M NaCl04, followed by several changes of distilled water to remove small amounts of metal-chelating agents used in the manufacture of the apoproteins.
Other chemicals were of the highest purity available and used without further purification.
Distilled water was used throughout.
EPR spectra were measured at a frequency near 9.2 GHz (X band) using a Varian E-12 spectrometer, operating at temperatures down to 90" K. pH measurements were made with a Vibron model 39A pH meter. Ultraviolet/ visible spectra were recorded on a Cary 15 spectrophotometer.
Fez-conalbumin and Fez-transferrin were prepared by addition of the calculated amount of Fe(C104)3 to a 10% solution of theapoprotein which was usually unbuffered in water.
(The molecular weight was taken as 80,000 in each case.) Solid NaHC03 was added for the dual purpose of increasing the pH to the desired value (generally 7.5) and also to ensure that sufficient bicarbonate was present for complete iron binding (9) ; the desired amount of solid NaClO+ was then added. About 0.3 ml of this solution was transferred to a calibrated EPR tube and frozen; a separate quantity was used to record the electronic spectrum if required.
The usual method of freezing was by slow immersion into liquid nitrogen over a period of about 40 s. Plunging into an acetone-Dry Ice bath was faster but gave the same results.
Experiments were also carried out2 in a rapid freezing apparatus (10)(11)(12) in which the solution was squirted into isopentane at -140" where it froze in about 4 ms. Again the same EPR spectra were obtained as with the other methods of freezing.
Thus we have nitrilotriacetate was present or not. That this broad signal found no evidence that the rate or method of freezing in any way does represent specifically bound iron is shown by the fact that affected the resultant EPR spectra.
Any one method gave re-in all these cases, in the presence or in the absence of perchlorate, producible results.
there was no change in the visible or ultraviolet spectrum.

Conalbumin-
The EPR spectrum of a 10% solution of Fezconalbumin in water at pH 7.3 is shown in Fig. IA. Addition of NaC104 to this sample caused the features at b and B to diminish while the features at s and S grew larger and better resolved; see the spectrum shown in Fig. lB, corresponding to 0.4 M perchlorate.
Further addition of NaC104, up to 3 M, caused no further change. The features marked S and s increased by a factor of 3.4 from the sample with no perchlorate to the sample containing 0.4 M perchlorate; in the same sequence, the features marked B and b disappeared.
Small but reproducible changes in the spectra at fields greater than 4 kilogauss were also observed; these are shown in Fig. 2. Clearly the broad upwardgoing absorption between 4 and 6.5 kilogauss is associated with B and b while the shallow depression at about 7 kilogauss is associated with X and s. Z'runsferrin-Similar expriments were performed with transferrin; the effect of perchlorate, however, is the reverse of that on conalbumin.
The EPR spectrum of a 10% solution of Feztransferrin in water at pH 7.3 is shown in Fig. 5A. The EPR spectrum of the same transferrin solution containing 0.4 M NaC104 is shown in Fig. 5B. Taking account of the gain settings of the spectrometer, it can be seen that the height of the sharp signal labeled S has in this case decreased and by a factor of 2 on the addition of 0.4 M NaC104; also the S peak has become less well resolved.
At the same time, a broad shoulder (B) has appeared on the low field side of the sharp signal and a more intense low field absorption may be observed (Feature b). The shoulder marked s in Fig. 5B is part of the peak labeled s in Fig.  5A. Changes in the region between 4 and 10 kilogauss were also observed with transferrin, the absorption positions corresponding closely to those for conalbumin.
Addition of NaC104 EPR spectra for the fully loaded Fez-conalbumin, recorded as a function of increasing NaGlO concentration are shown in Fig. 3. Isosbestic points are observed, the lower one being expanded in the inset.
If the conalbumin was less than fully loaded with iron, the increase in the sharp signal on adding perchlorate was greater. For example, Fig. 4 shows the S signal in 0.4 M perchlorate to be about 5.5 times greater than in water when the iron present corresponds to $5 of that required for saturation.
The broad signal was always present in the absence of perchlorate whether the conalbumin was dissolved in water, Tris buffer, or phosphate buffer (1 = 0.1, pH = 7.6) and whether I; (   3  4  5 spectrum (B or b). This is shown in Fig.  6, where the EPR spectrum of one sample of transferrin was recorded as a function of NaC104 concentration.
Isosbestic points were observed at field values of 720 and 1340 gauss for v = 9.137 GHz.
Again, addition of NaGlO at concentrations greater than 0.3 M (up to 3 M) had virtually no further effect on the EPR spectrum of a 10% solution.
Changes induced in the EPR spectrum by addition of NaC104 were studied for transferrin solutions which were fully, x, and >$ saturated with iron and at values of pH in the range 7 to 9. In all cases, the height of signal S decreased at the same rate as the NaC104 concentration increased.
Under no conditions, however, was its height less than M that observed in the absence of NaC104. Also, iron incorporated in the presence of strong perchlorate gave an equal distribution between the B and S EPR signals.
The same results were obtained when the transferrin was buffered with phosphate buffer (I = 0.1, pH 7.6), but the broad signal was not observed when Tris buffer was used nor when nitrilotriacetate was present in the phosphate-buffered solutions. Again, in none of these cases, was there a change in the visible or ultraviolet absorption spectrum.
The EPR spectrum of transferrin as it occurs naturally in blood serum3 was also recorded.
In this sample which was approximately half-saturated with iron, no evidence for any broad signals was observed. No change occurred when either we added more iron to this sample to saturate the transferrin, or when we added perchlorate to it.
Both ProteinsThat the effect of the sodium perchlorate is reversible and not harmful to the proteins was shown as follows. A 25% solution of protein, which contained 0.4 M NaC104, a portion of which was diluted 10 times, and a 2.5% protein solution which was made 0.04 M in NaC104 all give identical EPR spectra. If added perchlorate was removed either by passing the protein solution down a column of Sephadex G-25 equilibrated with water and reconcentration by membrane filtration, or by dialysis against two changes of water, the EPR spectrum characteristic of the protein in water was obtained.

INTERPRETATION AND DISCUSSION
That there is an equilibrium involving at least two paramagnetic high spin ferric protein species, and which is affected by perchlorate concentration, is clearly shown by the isosbestic points in the EPR spectra. That there is a third species detectable by EPR is most unlikely since (a) its molar extinction coefficient at the isosbestic points would have to be the same as that of the other two species and (b) a careful search over a wide field range at temperatures down to 2" K revealed no new features.
The effect of the perchlorate is most interesting. A change in the ligand binding the iron is thought to be most unlikely since Cl04 is a very weak complexing agent and also because there is no change in the ultraviolet/visible spectrum. Admit-803-l tedly, there is a change in the EPR spectrum, but this is known to be very sensitive to small changes in the strength and symmetry of the ligand field (13). Addition of any substance to an aqueous solution may alter the hydrogen-bonding properties of the medium, the term chaotropism having been coined for this effect. Perchlorate is a well known chaotropic anion (14, 15) and the conformation adopted by a protein is well known to be sensitive to the hydrogen-bonding properties of the solvent. We suggest, therefore, that the EPR changes we have recorded are caused by conformational changes which result in slightly different ligand fields at the ferric ion sites.
Both conalbumin and transferrin contain two independent iron-binding sites with remarkably similar properties (16) and, therefore, we seek a single scheme for the two equilibria.
There has been some discussion as to whether transferrin has a subunit structure (17) or whether there is only one polypeptide chain (18,19). However, it seems fairly clear that at least some duplication occurs (19, 20) and, in this sense, the protein may be thought to consist of two halves (whether or not these are linked is irrelevant for our purposes), each of which binds one iron atom. Furthermore, a conformational change in the protein has been proposed (8) to account for the ease with which transferrin releases iron in vtio compared with the difficulty of doing this in vitro at physiological pH. With these facts in mind we depict the equilibria for the two proteins as follows. c3L- The implications of this diagram are (a) in one conformation the iron atoms are bound identically to give the sharp EPR signal (88) as in cl; (b) a conformational change in one half causes just that iron to give the broad EPR signal (SB) as in c2; (c) a conformational change in the other half causes the other iron also to give the broad EPR signal (BB) as in c3.
Consider first conalbumin. In water the equilibrium SS + SB = BB is balanced towards the right but perchlorate drives it fully to the left. If the protein is less than fully loaded with iron the B site is preferred (b2) ; and when a second iron atom is added, the B site is again preferred (~3). Thus we account for the greater-than-2-fold drop in the S signal when perchlorate is removed. This is an alternative description to Woodworth's crevice model (21) and does retain the different iron-binding affinities for the two sites, the B site being more strongly binding, as well as incorporating our new data. That there should be a preference for the more firmly bound B site is in keeping with the apparent function of conalbumin, to bind metal firmly with little or no provision for later release. For transferrin, the S signal dropped no more than 2-fold on addition of perchlorate; therefore, in our experiments, only the equilibrium XX =: SB is important; this was true whether the protein was fully loaded or not.
Aisen et al. (2), like us, found that transferrin in isolated human 1 blood serum gave an EPR spectrum identical with that of the commercial sample of pure transferrin. We further found our sample to be half-loaded and unaffected by perchlorate, therefore, as in al, the conformation in u-hich iron is in the weaker binding site. We envisage, therefore, that when fully loaded, the conformation of the protein might change, under the influence of the red cell precursor, from SS to SK This would then give up 1 iron atom from the weaker S site in preference to 2, in accord with the findings of Fletcher and Huehns (8). Thus the important distinction between conalbumin and transferrin is upheld, the function of the latter being to bind and later release the iron it is carrying.
Of course our measurements on frozen solutions provide no direct evidence of two conformations under physiological conditions.
However, we have found artificial conditions under which the two sites differ and make the points that the sites could differ in viva probably as a result of conformational changes. It is known that protein conformations can alter markedly with salt concentration gradients set up during the freezing process (22) and it could be argued that such changes are being induced by perchlorate in our experiments.4 This does not affect our interpretation in terms of two conformations but it does imply that the room temperature concentrations of such conformers might be different from those of the frozen solutions. Salt gradient effects presumably do not affect conalbumin where the two EPR signals are seen in the absence of added perchlorate since it is hard to see how acid or salt concentration gradients can be set up in the simple aqueous protein solution.
Transferrin may be affected in this way but we think this is unlikely because: (a) the same spectra resulted from different rates and conditions of freezing; (b) its two EPR spectra are so similar to those of conalbumin in the absence of added salt; (c) the intensity of the sharp signal dropped by just 35, strongly indicating that just one of the two binding sites was being affected.
Ultracentrifuge experiments have shown the presence of an equilibrium (23) when conalbumin is exposed to certain buffer solutions but that the forms in equilibrium may be monomeric or aggregated according to the prior absence or presence of sulfate or perchlorate (24).
Other experiments (25) have shown that EPR spectra may be affected by dipolar broadening through solute aggregation on freezing.
This is expected to be important for small molecules but not for proteins, unless the paramagnetic site is on the outside of the molecule, because the broadening being proportional to l/r3 diminishes rapidly with increasing distance between the paramagnetic sites. We know of no evidence to suggest that the iron-binding sites are near the surface of the molecule in transferrin or conalbumin; indeed the EPR spectra of the lyophilized proteins are not significantly different from those of the frozen solutions and this suggests that they are not.
Aggregation in either of these two senses is not expected to affect our interpretations.