Bivchimica et Biuphy, ica Acta, I 118 ( 1991 ) 48-58 c 1991 Elsevier Science Publishers BN. All rights re~r~.'ed 0167-483S/ql/$03.50
48
B B A P R O MI)50
M6ssbauer spectroscopic investigation of structure-function relations in ferritins Erika R. Bauminger l Pauline M. Harrison 2, Daniel H e c h e l i, Israel Nowik and A m y r a Treffry z Racah bt~dnm. ,,f Physics. The th'hrew Unic~'rsit~,;Jent~ah'm ~isna'D and 2 Krt,h~ bulitutt" Dt'partmcnt of Mok'c'tdar Biofo~.9" and Bi, m,,'hnobq,% Unilz'rsity o[.~la'flh'ld. $hc[fidd (U.K.J (Received 3 M.,y llltH )
Key ~ords: Ferritin: Iron; Miissbauer specimenS: Iron(liD dimer
Ferritin plays an important role in iron metabolism and our aim is to understand the mechanisms by which iron is sequestered within its protein shell as the mineral ferrihydriteo We present Miissbauer spectroscopic data on recombinant human and horse spleen ferritin from which we draw the following conclusions: (!) that apoferritin catalyses Fe(ll) oxidation as a first step in ferrihydrite deposition, (21 that the catalysis of Fe(li) oxidation is associated with residues situated within H chains, at the postulated 'ferroxidase centre' and not in the 3-fold inter-subunit channels previously suggested as the initial F'e(ll) binding and oxidation site; (3) that both isolated Fe(lil) and Fe(lllJ p-oxo-bridged dimers found previously by M6ssbauer spectroscopy to be intermediates in iron-core formation in horse spleen ferritin, are located on H chains; and (41 that these dimers form at ferroxidase centres. The importance of the ferroxldase centre is suggested by the conservation of its ligands in many ferritins from vertebrates, invertebrates and plants. Nevertheless iron-core formation does occur in those ferritins that lack ferroxidase centres even though the initial Fe(li) oxidation is relatively slow. We compare the early stages of core formation in such variants and in horse spleen ferritin in which only 10-15% of its chains are of the H type. We discnss our findings in relation to the physiological role of isoferritins in iron storage processes.
Introduction
Ferritin pla~'s a key role in iron metabolism as an iron-sequestering molecule acting as both a sink and a source for iron [1-4]. Ferritins, which are widely distributed in animals and plants, are recognised by their molecular architecture, amino acid sequence similarity and (usuallyJ high iron content. Its protein chains are usually of two kinds. H and L. They co.assemble into tetracosameric shells with an 8 nm diameter cavity in which up to 45(!tl ironllll) atoms are deposited as the mineral ferrihydritc. Although amino acid sequences of H and L chains of a given species show only about 55% identity, cross-species identity is much greater: about 85% for H chains and 80% for L. Because most native ferritins are heteropolymer mixtures, H and L chain conformations are expected to be very. similar, and this
Correspondence: P.M. Harrison. Krehs !nstitute. Departmen! of Molecular Biology and Biotechnoh~gv. Uni,.ersity of Sheffield, P.O. 13ox 594. Firth Court. Western Bank. Sheffield Sill 2Utl. U.K.
has recently been established [5]. Main chain atoms of recombinant human H chain and rat L chain apoferritin homopolymers arc superposable within + 0. I/rim, and this applies also when native rat and horse apoferritins (largely L chain} are included in the comparison. The three-dimensional structure of the H chain ferritin shows at least one feature distinguishing it from L chain ferritins, however. This is the presence of a metal site within the four-helix-bundle subunit fold of the H chain [5,6]. In earlier work it was found that ferritin can be reconstituted by addition of Fe(ll} to apoferritin and that oxidation and polymerisation leading to ferrihydrite occur more rapidly within apoferritin than do equivalent processes in the absence of apoferritin (which give a different Fe(lil} mineral) [7-10]. Moreover, ferrihydrite deposition is faster in H-rich ferritins than in L-rich. in human H chain homopolymers, this rate is at least 50-fold that in L chain homopolymers Ill-141. For an understanding of iron oxidation-polymerisation processes in ferritin, including the different roles of H and L chains, we need to discover at what sites on
49 the protein shell iron-core intermediates form and in what sequence. In previous work [15,16] we found that MiSssbauer spectroscopy enables several iron species to be distinguished following the addition of Fe(ll) to apoferritin in air: free and bound Fe(ll), solitary Fe(lll) atoms (at least 1.7 nm apart), Fe(lll) g-oxo-bridged dimers, small Fe(Ill) clusters and large clusters of Fe(lil) atoms magnetically coupled at low temperatures. From a variety of evidence it is proposed that initially iron is bound anti oxidised on apoferritin [7,911,13,16,17], but in subsequent Fe(ll) additions, oxidation can occur directly on the mineral surface [16-20]. Although it has been deduced that apoferritin provides a catalytic Fe(ll)-oxidatiun (ferroxidase) centre [7,911], and that this can be identified with the above mentioned metal site on H chains [6-13], strict proof has been lacking because it is the production of cc,lo.~r
due to the final product (Fe(lll) oxo-bridged polymers) that is usually assayed and this results from a complexity of processes. Apofcrritin could accelerate more than one reaction step, e.g. dimer and cluster formation as well as oxidation and it is not clear which of the steps is rate-determining. Recently, it has been suggested from X-ray absorption and MSssbauer spectroscopy that Fe(ll) oxidation in horse spleen apofcr= ritin is relatively slow [21,22]. Here we use M6ssbauer spectroscopy both to measure the extent of Fe(ll)-oxidation in the presence of apoferritin and in protein free controls and to compare the production of core and core-intermediates in both horse spleen and recombinant human H chain ferritins, including variants in which putative oxidation and fcrrihydrite nucleation site tigands have been altered by site-directed mutagenesis. We aim therefore to verify
%75 ~
L____ a
.=,
I
d
e
" .......
......
f
Fig. I. Computer graphics representation of rccamh|nanl H chain ferritin variants used in this study. (a) CdM. alpha carbon plot of a CdM molecule and its neighbours in a crystal face as described in Ref 6. CdM bears the amino acid substitution LysSfi - , Gin on its outer surface, introduced to enable crystallisation. The inset shows an enlarg~.d view of the double metal bridge crystal contact with Gin-86 emphasised. This amino acid substitution does not affect the rate of iron=core formation [14.~]. b and c show close-up, views of parts t)f CdM. (bl the ~ferroxidase centre" of CalM. A metal site (labelled Fe e* ) has ligands Glu-27, Glu-62 and His-65 and a water molecule. There ,~re 24 of these sites per CdM molecule, one per H chain. (c) a threefold channel of CdM with Ca ~'~ ligated to three Asp-131 and three Glu-134 side chains, d, e ,')nd f are modelled from the CdM slruclurc. (d) 222. This variant has Glu-02 --) Lys and His65 ---)Gly las in L chains) in addition to Lys86 ~ Gin. A salt bridge replacing the ferroxidase centre is mode$1ed as in horse and rat L chain [erritin [6]. (el A2. This variant has Glu-61,Glu-64,Glu67 -=) Ala, The three glutamates have been proposed as a ferrihydrite nucleation centre [1,323] and their removal prevents iron binding at this ~sitinn. ([) 206. In variant 206 the 3-fold channel carbo~l groups have been replaced (Asp-131.Glu-134 --) Ala) thus preventing the binding of Fe: + or other metals at these sites.
50 the ferroxidase activity of apoferritin, to locate sites of oxidation, dimer formation and iron-core nucleation with respect to the three-dimensional structure of the protein, and by including parallel studies with horse spleen ferritin, which has about 85% L and 15% H chains, also to obtain insights into the roles of H and L chains in heteropolymer molecules. The structure of human H chain recombinant ferritin containing a single amino acid substitution (Lys86 Gin)on the outer ~urface of the protein has recently been described [5,6], Fig. la-c. This substitution was introduced to enable crystallization [5,6]. Since it does not affect iron incorporation [23], this apoferritin, code-named CdM, is considered as "control'. Another variant, code-named 206, contains two further substitutions: Aspl31 ~ Ala and Glu134 --, Ala, Fig. lf. These carboxylic acids are known metal ligands and their position within the 3-fold inter-subunit channels has suggested these channels as a possible route for Fe(ll) entry into the cavity and as a site for F¢(II) oxidation [24-27]. Substitution of both residues by alanine prevents binding of Fe -'+ or Fe "~* ions within these channels. in variant 222 a different pair of substitutions have been made: Glu62-, Lys and His65 ~ Gly (in addition to those of CdM, Fig. lc). These changes convert two 'ferroxidase centre' ligands to residues found in the inactive L chain (in which the metal centre is replaced by a salt bridge) [5,6,13]. A fourth human H chain variant 'A2' contains three substitutions: Glu-61, Glu-64, Glu-67-, Ala, Fig. le. These glutamates, which are conserved in known sequences of both H and L chains of mammalian ferritins [13] and also in the partial amino acid sequences of two plant ferritins [28], have previously been postulated as nucleation centre ligands [1,3,23]. It has also been proposed that Glu-61, whose side chain seems to be in two alternative positions in the human H chain electron density map, has a role in moving Fe 3÷ from the oxidation site to the cavity [6]. Materials and Methods
Horse spleen ferritin (Cd-free) was obtained from Boehringer Mannheim (Lewes, U.K.). Site-directed mutagenesis, overexpression in Escherichia coil and purification of recombinant human H chain ferritins were performed as described in Ref. 27. Four recombinants were used in the study including those with amino acid substitutions in the putative ferroxidase centre, nucleation centre or 3-fold channels (Fig. 1), as described in the introduction. To provide apoferritin, iron was removed from fer-itin by reduction with sodium dithionite (3 g/10O ml) in deaerated buffers. Reduction was with dithionite in preference to the more usual sodium thioglycollate [29,30] because it was difficul' to remove all traces of
thioglycollate particularly from the human H chain recombinants. The buffered ferritin solution was placed in a 40 ml ultrafiltration cell equipped with a PMI0 membrane (Amicon, High Wycombe, Bucks) and buffered sodium dithionite solution in the reservoir. 0.1 M sodium acetate (pH 4.8) was used for the reduction of horse spleen ferritin and 0.1 M Mes buffer (pH 6.0) for the recombinant human fcrritins. Reduction was continued until the protein solutions were colourless and no more than a trace of Fe(ll) could be detected in the ultrafiltrate. After removal of the protein solution from the ultrafiltration cell bipyridine was added to chelate any remaining Fe(II) before the samples were dial~,ed extensively against borate-buffered saline (BBS:20 mM boric acid/sodium tetraborate, 0.15 M NaCI, pH 7.4) and stored as ammonium sulphate precipitate. Before use the apoferritin was taken up and dialysed against BBS, concentrated by ultrafiltration and dialysed into 0.1 M Mops buffer containing 5 mM NaC1 at pH 7.0, or, in some samples of horse spleen apoferritin, at pH 6.5 or 5.9. Elemental 5~Fe and 5~Fe were obtained from the Atomic Energy Establishment (Harwell, U.K). Preparation of STFe and ~*Fe solutions was carried out as described previously [16]. Just before use the 57Fe(ll) solution was diluted with distilled water to give 15/zg Fe/5 #1. It was then added to the buffered apoferritin solution and the sample transferred to a lucite container (0.5 ml or 1 ml capacity). The first experiment was performed only with horse spleen apoferritin (0.835 mg/ml) at three pH values and S~Fen was added to give 480 Fe atoms/molecule (0.83 mM Fe). Controls had no protein, but the same amount of iron. The solutions were frozen after iron addition at times shown in Table 11. Two different experiments were carried out on the same concentration of -~TFe" (0.537 raM) but at different apoferritin concentrations (21.2 mg/ml or 6.7 mg/ml) giving either 12 or 38 Fe atoms/apoferfitin molecule. Protein-free controls were prepared with the same concentration of 5~Fe in the same buffer. The solutions were frozen at chosen times after iron addition, as shown in Table II. Both recombinant human H chain apoferritins and horse spleen apoferritin were used in these experiments. MOssbauer spectra were obtained with a 100 mCi 57Co(Rh) source at room temperature and a Harwell proportional counter. Velocity calibration was performed with a metallic iron foil at room temperature. Measurements were performed in cryostats at 90 K and at 4.1 K and in two velocity ranges. The spectra were measured at the lower velocity range ( _ 4 ram/s) in order to get a good resolution of the different doublets in the central peaks. The larger velocity range (+ 10 ram/s) was used to measure the relaxation and/or magnetic subspectra. All spectra were analysed
by computer fits. The deconvolution into subspectra of the experimental spectra, measured at the lower velocity range, was based on computer fits to giving the best fit (expressed by X squared) with a minimal number of doublets. The quadrupole splittings of the various doublets, as well as their isomer shifts, line-widths and relative intensities, were all free parameters in the computer fits. in the computer fits to the experimental spectra measured at the larger velocity range at 90 IL a relaxation spectrum was taken into account, in addition to the doublets observed at the lower velocity range. Relaxation spectra were fitted by using the simplest relaxation model, with one hyperfine field and one relaxation time [31]. In most of the spectra the parameters of the relaxation spectrum were fixed at values obtained in previous measurements [16] and only the relative ime~sity of this subspeetrum was left as a free parameter. In spectra measured at 4.1 K where a magnetic sextet was observed, the hyperfine field was also a free parameter. Results and Discussion
Types of iron species obserced by MiJssbauer spectroscopy Some typical Mrssbauer spectra are presented in Figs. 2 to 7 which also show the computer fits to these spectra. Four different Fe(lIl) species are identified in the various samples. Their Mrssbauer parameters are given in Table 1. Species a is a doublet corresponding to Fe(III) clusters, doublets b~ and b2 correspond to Fe(II). Good fits with Lorentzian lines to the subspectrum corresponding to F¢(li) could in most cases be obtained only if it was assumed to consist of at least two doublets. It is not clear in this case whether this is only a curve fitting procedure or whether indeed different Fe(il) sites exist. There may be a distribution of quadrupole splittings present, due to slightly diffelent environments of the Fe(II) ions, which cause the nonLorentzian line shape of the spectral lines, which can
be approximated by a superposition of two doublets. The Fe(II) doublets have slightly different parameters in ferritin (b I and b 2) and in control samples (b~ and b;). Doublets e~ and e2 correspond to ~-oxo-bridged directs. These two doublets have the same isomer shift, but different quadrupole splittings and they are clearly distinguishable in the spectra. The sextet, d, is a relaxation spectrum due to isolated Fe(III} atoms. The atomic electrons cause an effective magnetic field at the 57Fe nuclei. In paramagnetic materials this field is not stationary, but varies in size and direction due to paramagnetic relaxation. For relaxation frequencies,/', of 10 ") s-t, no magnetic interaction is observed in the Mrssbauer spectra. For f < 10~ s- ~, the spectra consist of a sextet and are similar to spectra observed in magnetic materials with a static magnetic field acting on the Fe nuclei. For 107 s -I
48
B B A P R O MI)50
M6ssbauer spectroscopic investigation of structure-function relations in ferritins Erika R. Bauminger l Pauline M. Harrison 2, Daniel H e c h e l i, Israel Nowik and A m y r a Treffry z Racah bt~dnm. ,,f Physics. The th'hrew Unic~'rsit~,;Jent~ah'm ~isna'D and 2 Krt,h~ bulitutt" Dt'partmcnt of Mok'c'tdar Biofo~.9" and Bi, m,,'hnobq,% Unilz'rsity o[.~la'flh'ld. $hc[fidd (U.K.J (Received 3 M.,y llltH )
Key ~ords: Ferritin: Iron; Miissbauer specimenS: Iron(liD dimer
Ferritin plays an important role in iron metabolism and our aim is to understand the mechanisms by which iron is sequestered within its protein shell as the mineral ferrihydriteo We present Miissbauer spectroscopic data on recombinant human and horse spleen ferritin from which we draw the following conclusions: (!) that apoferritin catalyses Fe(ll) oxidation as a first step in ferrihydrite deposition, (21 that the catalysis of Fe(li) oxidation is associated with residues situated within H chains, at the postulated 'ferroxidase centre' and not in the 3-fold inter-subunit channels previously suggested as the initial F'e(ll) binding and oxidation site; (3) that both isolated Fe(lil) and Fe(lllJ p-oxo-bridged dimers found previously by M6ssbauer spectroscopy to be intermediates in iron-core formation in horse spleen ferritin, are located on H chains; and (41 that these dimers form at ferroxidase centres. The importance of the ferroxldase centre is suggested by the conservation of its ligands in many ferritins from vertebrates, invertebrates and plants. Nevertheless iron-core formation does occur in those ferritins that lack ferroxidase centres even though the initial Fe(li) oxidation is relatively slow. We compare the early stages of core formation in such variants and in horse spleen ferritin in which only 10-15% of its chains are of the H type. We discnss our findings in relation to the physiological role of isoferritins in iron storage processes.
Introduction
Ferritin pla~'s a key role in iron metabolism as an iron-sequestering molecule acting as both a sink and a source for iron [1-4]. Ferritins, which are widely distributed in animals and plants, are recognised by their molecular architecture, amino acid sequence similarity and (usuallyJ high iron content. Its protein chains are usually of two kinds. H and L. They co.assemble into tetracosameric shells with an 8 nm diameter cavity in which up to 45(!tl ironllll) atoms are deposited as the mineral ferrihydritc. Although amino acid sequences of H and L chains of a given species show only about 55% identity, cross-species identity is much greater: about 85% for H chains and 80% for L. Because most native ferritins are heteropolymer mixtures, H and L chain conformations are expected to be very. similar, and this
Correspondence: P.M. Harrison. Krehs !nstitute. Departmen! of Molecular Biology and Biotechnoh~gv. Uni,.ersity of Sheffield, P.O. 13ox 594. Firth Court. Western Bank. Sheffield Sill 2Utl. U.K.
has recently been established [5]. Main chain atoms of recombinant human H chain and rat L chain apoferritin homopolymers arc superposable within + 0. I/rim, and this applies also when native rat and horse apoferritins (largely L chain} are included in the comparison. The three-dimensional structure of the H chain ferritin shows at least one feature distinguishing it from L chain ferritins, however. This is the presence of a metal site within the four-helix-bundle subunit fold of the H chain [5,6]. In earlier work it was found that ferritin can be reconstituted by addition of Fe(ll} to apoferritin and that oxidation and polymerisation leading to ferrihydrite occur more rapidly within apoferritin than do equivalent processes in the absence of apoferritin (which give a different Fe(lil} mineral) [7-10]. Moreover, ferrihydrite deposition is faster in H-rich ferritins than in L-rich. in human H chain homopolymers, this rate is at least 50-fold that in L chain homopolymers Ill-141. For an understanding of iron oxidation-polymerisation processes in ferritin, including the different roles of H and L chains, we need to discover at what sites on
49 the protein shell iron-core intermediates form and in what sequence. In previous work [15,16] we found that MiSssbauer spectroscopy enables several iron species to be distinguished following the addition of Fe(ll) to apoferritin in air: free and bound Fe(ll), solitary Fe(lll) atoms (at least 1.7 nm apart), Fe(lll) g-oxo-bridged dimers, small Fe(Ill) clusters and large clusters of Fe(lil) atoms magnetically coupled at low temperatures. From a variety of evidence it is proposed that initially iron is bound anti oxidised on apoferritin [7,911,13,16,17], but in subsequent Fe(ll) additions, oxidation can occur directly on the mineral surface [16-20]. Although it has been deduced that apoferritin provides a catalytic Fe(ll)-oxidatiun (ferroxidase) centre [7,911], and that this can be identified with the above mentioned metal site on H chains [6-13], strict proof has been lacking because it is the production of cc,lo.~r
due to the final product (Fe(lll) oxo-bridged polymers) that is usually assayed and this results from a complexity of processes. Apofcrritin could accelerate more than one reaction step, e.g. dimer and cluster formation as well as oxidation and it is not clear which of the steps is rate-determining. Recently, it has been suggested from X-ray absorption and MSssbauer spectroscopy that Fe(ll) oxidation in horse spleen apofcr= ritin is relatively slow [21,22]. Here we use M6ssbauer spectroscopy both to measure the extent of Fe(ll)-oxidation in the presence of apoferritin and in protein free controls and to compare the production of core and core-intermediates in both horse spleen and recombinant human H chain ferritins, including variants in which putative oxidation and fcrrihydrite nucleation site tigands have been altered by site-directed mutagenesis. We aim therefore to verify
%75 ~
L____ a
.=,
I
d
e
" .......
......
f
Fig. I. Computer graphics representation of rccamh|nanl H chain ferritin variants used in this study. (a) CdM. alpha carbon plot of a CdM molecule and its neighbours in a crystal face as described in Ref 6. CdM bears the amino acid substitution LysSfi - , Gin on its outer surface, introduced to enable crystallisation. The inset shows an enlarg~.d view of the double metal bridge crystal contact with Gin-86 emphasised. This amino acid substitution does not affect the rate of iron=core formation [14.~]. b and c show close-up, views of parts t)f CdM. (bl the ~ferroxidase centre" of CalM. A metal site (labelled Fe e* ) has ligands Glu-27, Glu-62 and His-65 and a water molecule. There ,~re 24 of these sites per CdM molecule, one per H chain. (c) a threefold channel of CdM with Ca ~'~ ligated to three Asp-131 and three Glu-134 side chains, d, e ,')nd f are modelled from the CdM slruclurc. (d) 222. This variant has Glu-02 --) Lys and His65 ---)Gly las in L chains) in addition to Lys86 ~ Gin. A salt bridge replacing the ferroxidase centre is mode$1ed as in horse and rat L chain [erritin [6]. (el A2. This variant has Glu-61,Glu-64,Glu67 -=) Ala, The three glutamates have been proposed as a ferrihydrite nucleation centre [1,323] and their removal prevents iron binding at this ~sitinn. ([) 206. In variant 206 the 3-fold channel carbo~l groups have been replaced (Asp-131.Glu-134 --) Ala) thus preventing the binding of Fe: + or other metals at these sites.
50 the ferroxidase activity of apoferritin, to locate sites of oxidation, dimer formation and iron-core nucleation with respect to the three-dimensional structure of the protein, and by including parallel studies with horse spleen ferritin, which has about 85% L and 15% H chains, also to obtain insights into the roles of H and L chains in heteropolymer molecules. The structure of human H chain recombinant ferritin containing a single amino acid substitution (Lys86 Gin)on the outer ~urface of the protein has recently been described [5,6], Fig. la-c. This substitution was introduced to enable crystallization [5,6]. Since it does not affect iron incorporation [23], this apoferritin, code-named CdM, is considered as "control'. Another variant, code-named 206, contains two further substitutions: Aspl31 ~ Ala and Glu134 --, Ala, Fig. lf. These carboxylic acids are known metal ligands and their position within the 3-fold inter-subunit channels has suggested these channels as a possible route for Fe(ll) entry into the cavity and as a site for F¢(II) oxidation [24-27]. Substitution of both residues by alanine prevents binding of Fe -'+ or Fe "~* ions within these channels. in variant 222 a different pair of substitutions have been made: Glu62-, Lys and His65 ~ Gly (in addition to those of CdM, Fig. lc). These changes convert two 'ferroxidase centre' ligands to residues found in the inactive L chain (in which the metal centre is replaced by a salt bridge) [5,6,13]. A fourth human H chain variant 'A2' contains three substitutions: Glu-61, Glu-64, Glu-67-, Ala, Fig. le. These glutamates, which are conserved in known sequences of both H and L chains of mammalian ferritins [13] and also in the partial amino acid sequences of two plant ferritins [28], have previously been postulated as nucleation centre ligands [1,3,23]. It has also been proposed that Glu-61, whose side chain seems to be in two alternative positions in the human H chain electron density map, has a role in moving Fe 3÷ from the oxidation site to the cavity [6]. Materials and Methods
Horse spleen ferritin (Cd-free) was obtained from Boehringer Mannheim (Lewes, U.K.). Site-directed mutagenesis, overexpression in Escherichia coil and purification of recombinant human H chain ferritins were performed as described in Ref. 27. Four recombinants were used in the study including those with amino acid substitutions in the putative ferroxidase centre, nucleation centre or 3-fold channels (Fig. 1), as described in the introduction. To provide apoferritin, iron was removed from fer-itin by reduction with sodium dithionite (3 g/10O ml) in deaerated buffers. Reduction was with dithionite in preference to the more usual sodium thioglycollate [29,30] because it was difficul' to remove all traces of
thioglycollate particularly from the human H chain recombinants. The buffered ferritin solution was placed in a 40 ml ultrafiltration cell equipped with a PMI0 membrane (Amicon, High Wycombe, Bucks) and buffered sodium dithionite solution in the reservoir. 0.1 M sodium acetate (pH 4.8) was used for the reduction of horse spleen ferritin and 0.1 M Mes buffer (pH 6.0) for the recombinant human fcrritins. Reduction was continued until the protein solutions were colourless and no more than a trace of Fe(ll) could be detected in the ultrafiltrate. After removal of the protein solution from the ultrafiltration cell bipyridine was added to chelate any remaining Fe(II) before the samples were dial~,ed extensively against borate-buffered saline (BBS:20 mM boric acid/sodium tetraborate, 0.15 M NaCI, pH 7.4) and stored as ammonium sulphate precipitate. Before use the apoferritin was taken up and dialysed against BBS, concentrated by ultrafiltration and dialysed into 0.1 M Mops buffer containing 5 mM NaC1 at pH 7.0, or, in some samples of horse spleen apoferritin, at pH 6.5 or 5.9. Elemental 5~Fe and 5~Fe were obtained from the Atomic Energy Establishment (Harwell, U.K). Preparation of STFe and ~*Fe solutions was carried out as described previously [16]. Just before use the 57Fe(ll) solution was diluted with distilled water to give 15/zg Fe/5 #1. It was then added to the buffered apoferritin solution and the sample transferred to a lucite container (0.5 ml or 1 ml capacity). The first experiment was performed only with horse spleen apoferritin (0.835 mg/ml) at three pH values and S~Fen was added to give 480 Fe atoms/molecule (0.83 mM Fe). Controls had no protein, but the same amount of iron. The solutions were frozen after iron addition at times shown in Table 11. Two different experiments were carried out on the same concentration of -~TFe" (0.537 raM) but at different apoferritin concentrations (21.2 mg/ml or 6.7 mg/ml) giving either 12 or 38 Fe atoms/apoferfitin molecule. Protein-free controls were prepared with the same concentration of 5~Fe in the same buffer. The solutions were frozen at chosen times after iron addition, as shown in Table II. Both recombinant human H chain apoferritins and horse spleen apoferritin were used in these experiments. MOssbauer spectra were obtained with a 100 mCi 57Co(Rh) source at room temperature and a Harwell proportional counter. Velocity calibration was performed with a metallic iron foil at room temperature. Measurements were performed in cryostats at 90 K and at 4.1 K and in two velocity ranges. The spectra were measured at the lower velocity range ( _ 4 ram/s) in order to get a good resolution of the different doublets in the central peaks. The larger velocity range (+ 10 ram/s) was used to measure the relaxation and/or magnetic subspectra. All spectra were analysed
by computer fits. The deconvolution into subspectra of the experimental spectra, measured at the lower velocity range, was based on computer fits to giving the best fit (expressed by X squared) with a minimal number of doublets. The quadrupole splittings of the various doublets, as well as their isomer shifts, line-widths and relative intensities, were all free parameters in the computer fits. in the computer fits to the experimental spectra measured at the larger velocity range at 90 IL a relaxation spectrum was taken into account, in addition to the doublets observed at the lower velocity range. Relaxation spectra were fitted by using the simplest relaxation model, with one hyperfine field and one relaxation time [31]. In most of the spectra the parameters of the relaxation spectrum were fixed at values obtained in previous measurements [16] and only the relative ime~sity of this subspeetrum was left as a free parameter. In spectra measured at 4.1 K where a magnetic sextet was observed, the hyperfine field was also a free parameter. Results and Discussion
Types of iron species obserced by MiJssbauer spectroscopy Some typical Mrssbauer spectra are presented in Figs. 2 to 7 which also show the computer fits to these spectra. Four different Fe(lIl) species are identified in the various samples. Their Mrssbauer parameters are given in Table 1. Species a is a doublet corresponding to Fe(III) clusters, doublets b~ and b2 correspond to Fe(II). Good fits with Lorentzian lines to the subspectrum corresponding to F¢(li) could in most cases be obtained only if it was assumed to consist of at least two doublets. It is not clear in this case whether this is only a curve fitting procedure or whether indeed different Fe(il) sites exist. There may be a distribution of quadrupole splittings present, due to slightly diffelent environments of the Fe(II) ions, which cause the nonLorentzian line shape of the spectral lines, which can
be approximated by a superposition of two doublets. The Fe(II) doublets have slightly different parameters in ferritin (b I and b 2) and in control samples (b~ and b;). Doublets e~ and e2 correspond to ~-oxo-bridged directs. These two doublets have the same isomer shift, but different quadrupole splittings and they are clearly distinguishable in the spectra. The sextet, d, is a relaxation spectrum due to isolated Fe(III} atoms. The atomic electrons cause an effective magnetic field at the 57Fe nuclei. In paramagnetic materials this field is not stationary, but varies in size and direction due to paramagnetic relaxation. For relaxation frequencies,/', of 10 ") s-t, no magnetic interaction is observed in the Mrssbauer spectra. For f < 10~ s- ~, the spectra consist of a sextet and are similar to spectra observed in magnetic materials with a static magnetic field acting on the Fe nuclei. For 107 s -I
