Anda di halaman 1dari 34

Pro~. Blophys. molec. Biol. Vol. 36, pp. 56--86 0079-610718011201-0053505.

00/0
~Pergamon Pre~ Ltd. 1980. Prlnl~d in Greal Britain

FERRITIN" MOLECULAR STRUCTURE A N D


IRON-STORAGE MECHANISMS
G. A. CLEGG, J. E. FITTON, P. M. HARRISON and A. TREFFRY
Department of Biochemistry, The University, Sheffield SIO 2TN

CONTENTS
I. INTRODUCTION 53
1. The Need for an Iron Storage Molecule 53
2. Physiolagical Forms of Storage Iron 54
3. Biosynthesis and Turnover of Ferritin and the Utilization of Storage Iron 56
II. Trm S~uCrURE OFTHEFEmUTINMOLECULE 58
1. The Protein Shell: Molecular Weight and Subunit Structure 58
2. The Three-dimensional Structure of Apoferritin 61
(a) Crystallineforms of ferritin and apoferritin 61
(b) The structure of horse spleen apoferritin at 2.8/~ resolution 62
(c) Conformation of the apoferritin subunit 63
(d) Channels through the apoferritin shell 64
(e) Heavy metal binding sites in apoferritin 64
3. The Iron Core of Ferritin 65
(a) Atomic structure of the iron cores 65
(b) Relationship between iron-cores and protein 67
III. A ~ T I O N OFFEtmrr~ IRON 67
1. Reconstitution of a Ferritin-like Molecule 67
2. Kinetics of Fe(II) Uptake: a Computer Model 67
3. Mechanism of Iron Uptake by Ferritin 69
4. Furt&,r Kinetic Data 71
5. Alternatit~ Mechanism of Ferritin Formation 73
6. Metal Ion Binding and Inhibition of Iron Uptake 75
7. Stoichiometry and Specificity of Oxidation in Ferritin Formation 76
8. Effects of Chemical ModOfcation on Ferritin Formation 77
IV. MOBILIZATION OF FEI~XrlNIRON 79
1. Last-in-first-out 79
2. Mechanism of Iron Release 79
~. ISOFr.gmTiNs 80
VI. CONCLUSIONS 82
~OWLEDGEMENT 83
P~NCr~ 83
I. I N T t ~ O D U C T I O N
I. The Need f o r an Iron-storaoe Molecule
Iron is an essential trace metal for all, or almost all, living species, being a constituent
of many enzymes, electron transfer complexes and oxygen carriers. In these molecules
iron is bound by a variety of organic ligands, which modulate its properties, so that in each
complex it is able to perform a specific biological function in a controlled manner. Iron
which is not bound may be harmful (Aisen, 1977; Willson, 1977). Thus interaction between
Fe(II) and dioxygen yields superoxide by transfer of a single electron, and subsequent
production of hydroxyl and other reactive radicals in a series of chain reactions may lead
to the destruction of cell constituents, including nucleic acid, lipid and protein. Hydrolysis of
Fe0Tl)aq. gives polynuclear Fe(III) species with liberation of protons and lowering of
blood pH is an observed effect of excessive ingestion of ferrous salts (Reissman et al., 1955).
Living organisms need to sequester iron both to conserve supplies for the synthesis of
functional iron compounds and to protect against the potentially harmful effect of
"decompartmentalized" iron. These functions may be combined in the same molecule. The
design of the ferritin molecule for iron-sequestration (Section II) and a discussion of
in vitro experiments on how it accumulates (Section III) and releases iron (Section IV) are

JI~ )~:~/~ A
53
54 G . A . CLEGG,J. E. FITTON,P. M. HARRISON and A. TREFFRY

the main subject of this review. In the remainder of the introduction a brief outline of its
relation to a second iron storage form, haemosiderin, and of its in ~i~'o synthesis and
turnover forms a background to the structure-function studies. The problem of iso-
ferritins is touched on in Sections II and V.

2. Physiological Forms of Storage Iron


The two forms of storage iron found in animals and plants are ferritin and haemosiderin
(Harrison et al., 1974a). Ferritin is a water-soluble molecule consisting of a multi-subunit
protein shell surrounding an inorganic iron-core of approximate composition (FeOOH)s
(FeO:OPO3H~) in the form of microcrystalline particles, sometimes referred to as
"micelles" (Granick, 1946; Farrant, 1954; Harrison et al., 1967; Massover and Cowley,
1973). The electron-dense nature of the iron-cores makes them directly visible in the
electron microscope, see Figs 1-4. The space inside the protein gives it a large capacity for
iron--nearly 30 ~ by weight or up to 4500 Fe(III) atoms--but this capacity is rarely com-
pletely utilized. In most ferritin preparations molecules are on average not more than
three-quarters full, often much less so, while iron-free apoferritin is usually also present.
Maintenance of a reserve capacity and an ability to accumulate and release iron are
important aspects of its iron-storage role (Harrison, 1977), and have stimulated a con-
siderable amount of work.

FJG. 1. Electron dense iron in hepatocyte of thalassaemic patient. Electron dense iron cores
(presumed to be ferritin) are randomly scattered in the cytosol. Molecules segregated within
lysosomes are apparently larger, more electron-dense and tend to form arrays or hexagonal
crystalline arrangements (arrows). H : Haemosiderin, unstained ~ 150,000. By courtesy of T.C.
Iancu.

The other form of storage iron, haemosiderin, was originally isolated as red-brown,
insoluble granules from horse spleens (Cook, 1929). The composition of such granules is
variable with respect both to iron content and orgai~ic constituents, which may include
membrane components as well as some ferritin protein (McKay and Fineberg, 1964).
Haemosiderin iron is also in the form of electron-dense ferric oxyhydroxide-phosphate
"micelles", but their insolubility and the clumping of the "micelles" seen by electron
microscopy suggest that it is formed by breakdown of ferritin with loss of its protein shell
(Matioli and Baker, 1963; Fischbach et al., 1971). It is not always possible to tell from
Ferritin: Structure a n d ' L - s t o r a g e 55

Fla. 2(a) Thin section of yolk platelet of snail, Planobarius corneus L., showing paracrystalline
ferritin after feeding iron sulphate in water. OsO4 fixed. Stained with Pb/U. ×90,000. By
courtesy of W. Bottke. (b) "Bacterioferritin" from .4zotobacter vinelandii By courtesy of E. I.
Stiefel, G. D. Watt and H. E. Clavert.

F[c~. 3. Ferritin from Pianobariu~ corneus L, negatively stained, × 440,000. By courtesy of


W. Bottke.
56 G.A. CLEGG,J. E. FITTON,P. M. HARRISONand A. TREFFRY

Fio, 4. High magnification dark field image of the iron-core of a single molecule of horse
spleen ferritin. Horizontal fringesare 3.8/~ apart. Bar -- 20/~ By courtesy ofW. H. Massover.

electron micrographs which form of iron is present, but semi-regular arrays of electron-
dense particles clearly separated from one another (presumably by their protein shells)
are usually assumed to be ferritin, whereas more irregular arrays are considered to be
haemosiderin, see Fig. 1. The apparent transformation of such ferritin into haemosiderin is
suggested in a study of biopsy material from patients with thalassaemia major (Iancu and
Neustein, 1977).

3. Biosynthesis and Turnover of Ferritin and the Utilization of Storage Iron


One characteristic of human iron metabolism is a lack of a means of excretion of this
element, although small amounts are lost mainly through the exfoliation of dead cells. Iron
absorption is controlled, but there is no absolute block and excess iron may enter the body.
The storage compartments respond to fluctuations in incoming iron as well as to patho-
logical changes in iron metabolism. Iron stimulates the biosynthesis of ferritin protein and
usually there is an overproduction of protein so that a reserve capacity is maintained.
However, high iron loading may result in a ferritin with higher than usual average iron
content and the relative amount of haemosiderin also increases. The stimulation by iron of
ferritin biosynthesis has been well documented in several animals (e.g. Drysdale and Munro,
1966), in plants (David and Easterbrook, 1971) in hepatoma cells in tissue culture (Beck
et al., 1974), and in a cell-free system (Drysdale and Shafritz, 1975). Studies of the in-
corporation of [14C]- or [3H]-leueine into ferritin, stimulated by iron (Drysdale and
Munro, 1966; Lee et al., 1975; Lee and Richter, 1977a, b) have indicated the following
sequence:

amino acids -~ subunits ~ apoferritin shells -~ferritin.
Neither actinomycin D nor cordycepin reduces the stimulation of biosynthesis, suggesting
translational level control (Drysdale and Munro, 1966; Z~ihringer et al., 1976) and iron
Ferritin: Structure and iron-storage 57

seems to make more m-RNA available for synthesis (Ziihringer et al., 1975). Apoferritin is
synthesized predominantly on free polyribosomes (Hicks et al., 1969; Puro and Richter,
1971), although about 15 ~ may be synthesized on bound polysomes and the two popu-
lations may differ both in their associated amino acid pools (Fern and Garlick, 1976) and
in their response to iron (Konijn et al., 1973).
A translational control mechanism for ferritin biosynthesis has been proposed (Z/ihringer
et al., 1976) and is shown in Fig. 5. It is suggested that a ferritin subunit attached to ferritin
m-RNA prevents it binding to ribosomes in the absence of iron. Iron entering the cells
mobilizes these subunits and frees the messenger for translation. It is also proposed that
different m-RNAs may be translated on free and membrane-bound polyribosomes implying
the synthesis of different ferritin subunits. Low concentrations of ferritin are found in
serum and it is possible that the exported protein (presumed to be synthesized on the bound
polyribosomes' m-RNA) is different from cytosolic ferritin. Another proposal that ferritin is
a hybrid of two different subunits is discussed briefly in Sections II and V.
~ DNA

~ t ~Actmomycin
hn RNA
~"~ Cordycepin
~ J,;; + Poly(A)
Cytoplasm 5' I 3'
Ferritin mRNP ~ ' ~ ' A ~
l ................. ~-~!--] Repression

repressed
~'~
,,,",-,~.~A~
3'i
• Holoferritin
Ferritin mRNP ~ Poly(A) I eee

Fe~ritin S~bunit il-Fe.,~'eFe.


I •

,ee-

~ \ •
..
Fernhn . I
Subund
-,o.
~ ~
~.,~Fel~:
Pool • • •
|;
/ ,-Pe
_

Derepression • ""~.~ ,-%


,,,,,
Ferritin mRNA5 ~ ' ' / ~ UG' ' J ~ ' / ~ V ~ ' Apoferritin
Poly(A)

3' ~/~/~/~/~/~/~/~/~.
Translation ~-AU~~-~@--.,~ F.,,,~. Peptid~

FIG. 5. Modelfor translational control mechanismby whichiron slx~cificallyregulatesferritin


biosynthesis(From Z~hringeret al., 1976, with permission).

Studies of ferritin turnover following the incorporation of labelled amino acids give
estimates of half-life varying from about 3 days (Drysdale and Munro, 1966) to 1.3 days
(Glass and Doyle, 1972); the lower value being obtained when amino acid re-utilization
was minimized. Young rats gave a shorter half-life than old (Ore et al., 1972) and female
than male (Linder et al., 1973). Repeated iron injections appear to increase the half-life,
whereas bleeding accelerates turnover (Drysdale and Munro, 1966; Munro and Drysdale,
1970). The observed stimulation of biosynthesis by iron cannot be explained solely by a
decreased rate of degradation (Drysdale and Munro, 1966).
Use of radioactive Fe has shown that iron is incorporated into intact ferritin molecules
(Hoy and Harrison, 1976; Lee and Richter, 1977a, b), and that it may also enter haemo-
siderin without necessarily passing through ferritin (Wyllie and Kaufman, 1971). Both
forms of store provide a source of iron for the pregnant rat foetus (Wyllie and Kaufman,
1971) and may be mobilized by bleeding (Millar et al., 1970). Ferritin may also act as a
short-term acceptor for iron en route for haemoglobin synthesis in erythrocyte precursors
58 G.A. CLEGG,J. E. FITTON,P. M. HARRISONand A. TREFFRY

(Nunez et al., 1978; Speyer and Fielding, 1979). Provided both succinate and FMN or FAD
are present, it is found to act as a source of iron for mitochondrial ferrochelatase (Ulvik
and Romslo, 1978). A microsomal reductase enzyme may also be involved in ferritin iron
mobilization in vivo (Osaki and Sirivech, 1971 ; Sirivech et al., 1977; Zaman and Verwilghen,
1977), although reduced flavins will release the iron in vitro in the absence of enzyme
(Sirivech et al., 1974), see Section IV. It has also recently been suggested that iron may be
made available for reutilization by cytosolic degradation of ferritin molecules (Bomford and
Munro, 1980).

II. THE STRUCTURE OF THE F E R R I T I N MOLECULE


1. The Protein Shell: Molecular Weight and Subunit Structure
Ferritin Fe(III)-iron-cores can be removed by reduction and dialysis leaving intact
protein shells (Granick and Michaelis, 1943; Crichton, 1973). The apoferritin, so produced,
closely resembles "native apoferritin" (the "top component" of differential or density
gradient centrifugation) as judged by circular dichroism and hydrodynamic properties
(Leach et al., 1976) and X-ray diffraction (Harrison, 1959; Fischbach et al., 1969). Most
of the structural work has been done on chemically prepared apoferritin. Horse spleen
apoferritin has a molecular weight of about 447,000_ 17,000 (Bjork and Fish, 1971;
Richter and Walker, 1967; Harrison et al., 1974a) and measurements by sedimentation
equilibrium and SDS-polyacrylamide gel electrophoresis give a subunit weight of
18,500 _ 500 (Bjork and Fish, 1971 ; Bryce and Crichton, 1971) and hence 24 subunits in all.
There have been reports that apoferritin purified from sources other than horse spleen
may be larger than 18,500, e.g. porcine spleen, 503,000 with a 20,000 subunit (May and Fish,
1977); Corbicula sandai, 503,000 with a 23,000 subunit (Baba et aL, 1977), Pisum sativum,
480,000 with a 20,300 subunit and Lens esculenta, 510,000 with a 21,400 subunit (Crichton
et al., 1978) and human heart, 530,000 with a 21,000 subunit (Arosio et al., 1978) although
some of these size variations may be due to differences in experimental technique.
Recently a "bacterioferritin" has been purified from Azotobacter vinelandii, which has a
subunit molecular weight of 17,000 and is unusual in that, in addition to an iron core the
protein contains approximately one haem molecule per two subunits (Stiefel and Watt,
1979). The function of the haem is uncertain. A similar molecule has also been reported in
Escherichia coli (Bauminger et al., 1979; Yariv et al., 1979).
Oligomers (dimers, trimers etc.) of the 24 subunit apoferritin shell have been observed in
most ferritin preparations on polyacryamide gel electrophoresis and also on gel filtration
(Harrison et al., 1974a). The functional significance (if any) of these oligomers is uncertain,
but they may be associated with iron-loading or conversion of ferritin to haemosiderin.
Polypeptides with molecular weights of less than 19,000 (6000-8000, 11,000-12,000 and
14,000-15,000) are frequently seen on SDS polyacrylamide gels of purified ferritin pre-
parations (Niitsu et al., 1973; Linder et al., 1974, 1975). The use of protease inhibitors
during ferritin purification decreases the yield of these polypeptides (Crichton et al., 1977;
Lavoie et al., 1977; Linder et al., 1977) suggesting that they result from a specific proteolytic
cleavage of the 19,000 subunit. The 6000-8000 and 11,000-12,000 polypeptides co-assemble
to give molecules which resemble native apoferritin and their combined amino acid
analyses are virtually identical to that of the 19,000 intact subunit (Ishitani et al., 1975).
The 15,000 polypeptide may be a compact conformational isomer of the 19,000 subunit
(Ishitani et al., 1975) or, alternatively, a dimer of the 7000 polypeptide (Bryce et al., 1978).
The question of subunit composition is further complicated by the suggestion that the
apoferritin shell is a heteropolymer composed of a variable proportion of two subunits of
different molecular weights (19,000 and 21,000) and different primary structures (Drysdale,
1977; Hazard et aL, 1977; Arosio et al., 1978). The two subunits are designated H (for
heavy--21,000) and L (for light--19,000) and ferritin isolated by normal purification
techniques is considered to be a mixture of the 23 possible heteropolymers (L23H1-L1 H23)
and the two homopolymers (L24Ho and LoH24). Evidence for this proposal is based on
isoelectric focusing profiles (Drysdale, 1977) and the observation that two components can
Ferritin: Structureand iron-storage 59

be resolved using a gradient pore SDS polyacrylamide gel electrophoresis system (Arosio
et al., 1978; Lavoie et al., 1978). The relative proportions of H and L subunits appears to be
organ specific, heart generally giving a higher proportion of H subunit, while liver or spleen
ferritins contain mainly L subunit (Drysdale, 1977; Arosio et al., 1978). Bomford et al.
(1978) consider that H and L subunits differ in charge and not size and it is also possible
that the apparent size differences are a result of SDS binding or conformation changes. A
recent proposal (Bryce et al., 1978) is that, just as the polypeptide of molecular weight
15,000 could be due to the aggregation of two 7000-8000 polypeptides, the H subunit may
be a dimer of the 11,000 fragment. In addition, the discrete bands seen on isoelect'ric
focusing may be methodological artifacts (Shinjo and Harrison, 1979).
Amino acid analyses of ferritin from a wide variety of sources are remarkably similar
although both species differences and organ differences within the same species are observed
(Table 1). Not surprisingly the phytoferritins and "bacterioferritin" show greater dif-
ferences when compared with ferritins from mammalian sources, possibly reflecting their
greater evolutionary divergence.
Small amounts of carbohydrate have been found in ferritin preparations (Shinjo et al.,
1975; Lavoie et al., 1977; Cynkin and Knowlton, 1977). The relative numbers of different
sugars reported varies and does not seem to be stoichiometrically related to the number of
subunits, although in a multisubunit assembly not all subunits need be glycosylated. If,
as has been reported, carbohydrate remains attached to subunits after dissociation of the
apoferritin shell (Lavoie et al., 1977; Cynkin and Knowlton, 1977) it is unlikely that it is due
to contamination with either non-covalently bound carbohydrate or other glycoproteins.
However, Spik (personal communication) has been unable to detect significant amounts of
sugar residues in a purified preparation of horse spleen apoferritin, and it must remain an
open question whether this ferritin, at least, is indeed a glycoprotein. It has recently been
reported that a fraction of human serum but not tissue ferritin (Worwood et al., 1979a, b)
and a fraction of rat liver ferritin (Halliday et al., 1979) bind to concanavalin A. To what
extent the observed subunit heterogeneity is related to carbohydrate content is not yet clear.
Peptide maps derived from proteolytic digestion of apoferritin from a number of species,
from different organs within species, and from foetal, neonatal and maligant tissue (Crichton
et al., 1973; Linder et al., 1975; Vulimiri et al., 1977a; Arosio et al., 1978; Alpert et al.,
1979) suggest that while there is substantial sequence homology between ferritins, there
are also species, and even organ and age specific amino acid sequences.
Amino acid sequence determination of ferritin from several species is now in progress
and that of horse spleen apoferritin is nearing completion (Crichton, personal com-
munication), and this work should help to elucidate the nature of the observed differences in
ferritin subunits.
Amino acid sequence determination has been hampered to some extent by the finding
that the protein contains an acylated amino terminus so preventing straightforward N-
terminal sequence analysis to progress. An N-terminal pentapeptide has been isolated from
horse spleen apoferritin (Suran and Tarver, 1965) and has the sequence: N-acetyl-Ser-Ser-
Gln-Ile-Arg-.
An identical N-terminal tripeptide sequence--N-acetyl-Ser-Ser-Gln--is found in rat liver
apoferritin (Huberman and Barahona, 1978).
Preliminary sequence results (Crichton et al., personal communication, Harrison et al.,
unpublished results) from horse liver, horse spleen, human liver, human spleen and lentil
ferritin show extensive regions of sequence homology as well as organ and species specific
differences, so confirming the results of peptide mapping. Completion of one or more of
the sequences in conjunction with X-ray analysis of the three-dimensional structures should
add greatly to our understanding of the ferritin molecule.
Subunit interactions in apoferritin shells and their assembly have been the subject of
several studies. Although stable between pH 3 and 11 apoferritin may be disaggregated at
extreme pH values, in 66% (v/v) acetic acid/water at 0°C (Harrison and Gregory, 1968);
in 0.25 % SDS at elevated temperatures (Smith-Johannsen and Drysdale, 1969); and in 6 ~!
guanidinium chloride (Listowsky et al., 1972), although it is stable to 10 Murea. Reassembly
60 G. A. C L E G G , J. E. FITTON, P. M. HARRISON a n d A. TREFFRY

Z~ ~ ~ ~
m

U
,.~

.e
.<
;~
:~ ©
0
E
z
~-
.~

<
~ o ~ ~
~
~
~ ~
o ~
~ ~ ~
~ ~
Z

:~

e~

<
0
Z
.~ )~

o ~.-
~ ~ o ~ o ~
~ ~ M ~~ M~--~ M d ~ M ~ ~ --

~ ~ ~O ~
~ ~
<~ ~
~<~O~O<>~ ~
~
~
~ ~
~

E
Ferritin: Structureand iron-storage 61

occurs when the denaturant is slowly removed by dialysis as judged by a variety of physical
methods. A fraction of the reassembled apoferritin is crystallizable (Hoy et al., 1974a). A
comparison of dissaggregation and reassociation by means of ultraviolet difference
spectroscopy shows hysteresis at low pH values (Crichton and Bryce, 1973). Optical rotatory
dispersion and circular dichroism measurements also indicate that reversible unfolding
occurs and that at pH values just below the onset of reassociation the subunit is in a folded
conformation similar to that "~L~the whole protein shell (Wood and Crichton, 1971 ; Leach
et al., 1976). Dimers and tetramers were indicated as assembly intermediates by gel
filtration (Crichton, 1975), but examination of the three-dimensional structure (Banyard
et al., 1978; Section 11.2) suggests a pathway via dimers and hexamers. Use of fluorescent
probes, which bind to subunits and induce their partial polymerization, supports the latter
(Stefanini et al., 1979). Use of differential calorimetry shows that ferritin denatures at a
relatively high temperature (Td= 100°C) with a broad thermal transition indicating little
cooperativity between subunits, so that each appears to melt separately (Donovan, Theft
and Tomimatsu, personal communication).
Ultraviolet difference spectra indicate that four out of five tyrosines and one out of two
tryptophans per subunit move from a more hydrophobic to a more hydrophilic environment
during dissociation (Crichton and Bryce, 1973), and all five tyrosines may be nitrated under
denaturing conditions compared with only one/subunit in native apoferritin (Wetz and
Crichton, 1976). Only eleven out of 22 carboxyls in apoferritin are available for reaction
with glycine amide and carbodiimide in non-denaturing conditions (Wetz and Crichton,
1976) and this modified apoferritin is unable to assimilate iron (see Section III). When all
carboxyls are modified in denatured apoferritin the modified subunits do not reassemble to
complete shells. Nine of the ten arginines and four-five of the nine lysines residues/subunit
are modified by cyclohexanedione and maleic anhydride respectively (Wetz and Crichton,
1976). From these results it has been concluded (Crichton, 1975) that all of the tyrosines,
one tryptophan per subunit and a number of basic and acidic residues probably in salt
linkage are located in interface regions between subunits. Silk and Breslow (1976) found
that all of the tyrosine and cysteine, two of the nine lysines, three out of six histidines and
three of the carboxyl groups are unavailable to titration. These authors also found that in
non-denaturing conditions six-seven lysines react with formaldehyde, and that two-three
histidines and one cysteine are modified with bromoacetic acid. Wetz and Crichton (1976)
found one cysteine reacting with N-ethylmaleimide per subunit. At least two histidines can
be ethoxyformylated by diethyl pyrocarbonzte without impairing the ability ofapoferritin to
accumulate iron (Sowerby and Fitton, unpublished results) and modification of one-two
cysteine is likewise without effect.

2. The Three-dimensional Structure of Apoferritin


(a) Crystalline forms of ferritin and apoferritin
Horse spleen ferritin has been crystallized from solutions containing one of a variety of
different divalent metals (Laufberger, 1937). CdSO,~ has been found to produce the best
crystals for X-ray crystallography and both ferritin and apoferdtin can be crystallized in the
presence of 40-200raM solutions of this salt (Granick and Michaelis, 1943; Harrison
1959). Three crystalline forms have been obtained. By far the most common is the cubic
form, space-group F432, a= 184 A, in which both ferritin and apoferritin grow as
isomorphous octahedra (Fankuchen, 1943; Harrison, 1959). An orthorhombic form
(P22121, a = b = 130 A, c= 184 A) was reported for ferritin, but not apoferdtin (Harrison,
1959, 1963), and tetragonal crystals (P4212, a = b = 1 4 7 A, c= 154 A) have been grown
for apoferritin and for ferritin co-crystallized with apoferritin (Hoy et al., 1974a).
Ferritin molecules are approximately spherical. The cubic unit cell contains four
molecules in face-centred cubic close-packing. This means that molecular centres lie at the
points at which the crystallographic 4-, 3- and 2-fold axes intersect and hence the 24
identical subunits within the molecules must also be related by these symmetry elements. In
orthorhombic crystals molecules pack in a similar manner, a and b are parallel to the cube
62 G.A. CLEGG,J. E. FITTON,P. M. HARRISONand A. TRF;tRY

face-diagonals and the smaller unit cell is approximately body-centred. Molecules are
rotated _ 6 ° about a so that now only this unit cell axis coincides with a molecular sym-
metry (2-fold) axis. The tetragonal crystals are also approximately body-centered, but this
time each molecule is rotated _+27.5° about a 4-fold axis and the unit cell sides are extended
and contracted with respect to those of the orthorhombic cell (Hoy et al., 1974a).
Octahedral crystals have been grown not only from horse spleen and liver ferritin but
from the spleen and/or liver of humans, rats, dogs, jackals, guinea-pigs, dogfish, sharks and
dolphins (Michaelis, 1947; Tecce, 1952; Kato, 1969; Hoare et al., 1975b, and from an
invertebrate, Cryptochiton stelleri (Towe et al., 1963). The similarity of the crystals suggests
a highly conserved three-dimensional structure. Preliminary X-ray diffraction work shows
that ferritins or apoferritins from spleen and liver of humans, rats and dogs crystallize in
unit cells with F432 symmetry isomorphous with those of horse spleen and liver ferritin.
Hence all these crystalline ferritins contain molecules and subunits of essentially the same
size and conformations. This is also true of horse spleen ferritin fractions of different iron
content and of their derived apoferritins, so that the protein shell is hardly affected by the
addition or removal of iron.
Other crystalline forms have been obtained from C. sandai (Baba, 1969), "bacterio-
ferritin" from A. vinelandii (Stiefel and Watt, 1979) and the "iron-storage protein" from
E. coli (Yariv et al., 1979). Early X-ray studies on the latter indicate that, like mammalian
ferritin, it contains 24 subunits in 432 symmetry (Yariv, personal communication). Natural
occurrence of crystalline or quasi-crystalline ferritin is known (see Fig. 2).

(b) The structure of horse spleen apoferritin at 2.8 A resolution


A 6 A resolution electron density map showed horse spleen apoferritin as a hollow shell
penetrated by six channels, lying along 4-fold axes, and containing rods of electron density
thought to be a-helix (Hoare et al., 1975a, b). These features have been confirmed at 2.8 A
resolution (Banyard et al., 1978) and subunit boundaries within the shell have now been
delineated. We can now describe the structure as follows.
The molecule is a hollow shell of external and internal diameters respectively about 130
and 80 A. Each subunit is roughly cylindrical (l= 55 A, d = 27 A) and contains four long,
nearly parallel right-handed helices (A, B, C, D), a shorter helix (E), a helical turn (P),
a long length of extended chain (L) and some regions of irregular conformation, Figs 7a and
b. The long helices lie within the shell, perpendicular with the radius vector. Each subunit
makes contact with five neighbours.
The four long helices have their lengths parallel (or anti-parallel) to the cylinder axes and
these axes lie nearly parallel to molecular 3-fold and nearly perpendicular to molecular 2-
fold axes. Diad related subunits are in contact over the greater part of their lengths.
Operation of the diad produces a system of eight helices in two layers of four. Neighbouring
helices in each layer are anti-parallel and lie at angles of between 151 ° and 171 °. Similar
angles relate neighbouring anti-parallel helices in the second layer, parallel helices beiag at
roughly 20 ° to each other. Also close to the diad, on the outside surface of the molecule,
is the long loop, k, which approaches close enough to the 2-fold axis to make what appears
to be a short length of anti-parallel pleated sheet with the symmetry related loop from its
neighbour. In this region of the electron density map there are two large peaks lying on
intermolecular diads, which we attribute to Cd(II) of crystallization linking loops, L, in
neighbouring molecules. There is also a considerable amount of intersubunit contact
around molecular triads. The subunits form a 3-blade propeller shape with long axes
roughly perpendicular to each other. There is relatively little subunit interaction in the
neighbourhood of 4-fold axes (Fig. 6).
The extended system of long parallel helices produces a rather flat dimer (see Fig. 9)
and the quaternary structure resembles a rhombic dodecahedron truncated at its apices. The
structural analysis suggests that the symmetrical dimer, described above, is likely to be a
relatively stable intermediate in the assembly of the protein shell and that a symmetrical
tetramer would not be stable. Operation of a 3-fold axis could give a symmetrical hexamer
from three dimers. The complete molecule could be assembled by addition of three further
Ferritin: Structure and iron-storage 63

,o,

-- (

1~o. 6. Quaternary structure of horse spleen apoferritin vieweddown a 4-foldaxis. (a) Subunit
arrangement, (b) helical rods in eight subunits with one alternative connectivity. (Bourne,
Clegg, Harrison and Stansfield,unpublished).

such hexamers, or dimers or monomers could be added separately. Operation of a 4-fold


axis would produce a symmetrical octamer, but it is of interest to note that assembly could
not proceed by interaction of three such octamers, since this is topologically impossible.

(c) Conformation of the apoferritin subunit


While helical regions are clear in the electron density map, interhelical connections occur
in weaker regions, which are capable of at least two alternative interpretations. In the
first of these (Banyard et al., 1978), shown in Fig. 7a helices A-E are joined in alphabetical
order and the long external loop, L, connects opposite ends of B and C. The N-terminus
is in a short non-helical region preceding A. Helix A on the outside of the molecule turns
into B on the inner surface. Following the long loop, L, the polypeptide chain runs along C
on the outside of the subunit, and, after a tight corner turns along D on the inside surface,
then via a twisted segment and a single helical turn (P) the chain runs through E and ends
with a non-helical tail, T, inside the spherical shell. The bundle of four long helices is
roughly square in cross-section with neighbouring helices anti-parallel and diagonally
related helices parallel. The symmetry of the bundle is thus approximately 222. In an
alternative arrangement, Fig. 7b, helical positions and directions are as before, but they are
now joined A - - C - - P - - E - - B - - L . Thus the C-terminus of the molecule occurs at the end of
the long loop on the outside of the molecule. Liberation of C-terminal amino acids under
non-denaturing conditions by carboxy peptidase would be possible with this second alterna-
tive, but with human liver ferritin at least this is not found (Sowerby and Fitton, unpublished
observation). Proteolytic splitting of the subunit in native apoferritin molecules to give
7500 (N-terminal) and 11,000 fragments could occur in the regions indicated by the arrows
on the alternative arrangements, Figs 7a and b. Both are in accessible positions on the out-
side surface of the shell. The numbers of residues associated with the various structural
components are indicated in the diagrams. It is of great interest to note that the apo-
ferritin subunit conformation, with its four long helices, closely resembles those of myo-
haemerythrin (Hendrickson and Ward, 1977), haemerythrin (Ward et aL, i975; Stenkamp,
et al., 1976) tobacco mosaic virus (Bloomer et al., 1978), cytochrome b562 (Scott-Mathews
et aL, 1979) and cytochrome c' (P. Weber, et al., 1980). With the second alternative
connectivity, the course of the chain, apart from the lengths of helical and non-helical
regions, is very similar to all of these structures, moreover cytochrome c' is a dimer
resembling those formed within the apoferritin shell (P. Weber, et al., 1980). In the two
cytochrome structures, haem lying between helices bisects the subunit. The bacterial
"ferritin-like" protein contains haem, which may occur only once for every two sub-
units. The fascinating possibility arises that this protein may have a subunit structure
like mammalian ferritin, but with haem placed between helices at the 2-fold axis of the
dimer.
64 G.A. CLEGG,J. E. FITTON,P. M. HARRISONand A. TP,IZlFFRY

(a) E (b) E
p P

FI~. 7. Conformation of the horse spleen apoferritin subunit showing alternative connectivities
between helical sections (rods). (al is the interpretation of the electron density map of Banyard
e¢ al. (1978) and (b) is an alternative based on model building (Ctegg, Stansfield, Bourne and
Harrison, unpublished). Arrows indicate possible sites of proteolysis, ketters A-E and P
indicate helical sections and L a non-helical loop. Approximate number of residues in each
helix; A, 27: B, 25; C, 28, D, 20: P, 7: E, 10.

(d) Channels through the apoferritin shell


One of the most interesting features of the apoferritin quaternary structure is the presence
of six channels, parallel to 4-fold axes, which pass right through the shell and which may
enable iron and molecules necessary for its storage and mobilization to enter or leave the
central cavity (Hoare et al., 1975a; Banyard et al., 1978; Harrison et al., 1978). The channels
have an "hour-glass" shape, surrounded at their narrowest part by four, E, helices, one from
each subunit, forming a square of side about 12 A. Side chains reduce the "waist" still
further to 7-8 A, but there is a diagonal gap of about 14 A which may allow flat molecules
of this dimension to pass through. Molecules which are known to release iron from ferritin
include F M N H z and other dihydroflavins (Sirivech et al., 1974) and it is possible that they
are able to reach the iron-cores (Harrison et al., 1975, 1978; Jones et al., 1978), see Section
IV and Fig. 8.

(e) Heavy metal binding sites in apoferritin


Isomorphous replacement by Hg and U (diffused into the crystals as p-chloro-mercuri-
benzoate and K3UO2F 5 respectively) was used in the X-ray analysis at 2.8 A resolution
and, although not used for phase determination, binding by Tb(llI) has also been reported
(Banyard et al., 1978). Mercuri-benzoate is probably bound covalently to cysteine residues
in apoferritin. (Hoare et al., 1975a~ b) and uranium, probably as UO~+, and Tb(IIl) are
likely to be found at sites containing carboxyl groups (Blundell and Johnson, 1976). Hence
these sites may be of help in fitting the sequence to the electron density, as well as in
(in the case of Tb(III), see Section IV) locating possible iron-binding centres. Figure 9
shows the sites in relation to the known structure. With the exception of the major Hg site,
all of the heavy atom positions are on or near the inside surface of the apoferritin shell and
most of these lie in the region of contact between diad-related subunits. It may be noted
that four of the uranyls bind in two close pairs, possibly alternatively rather than simul-
taneously (compare Gd(IIl) sites in hen egg white lysozyme, Perkins et al., 1979). Tb(III)
is found at positions very close to two of the uranyls and at its major site it forms a close-
pair across the 2-fold axis with its symmetry-related neighbour at 4.3 ]~ distance. Inhibition
of ferritin formation by Tb(lI1) possibly implicating this site is discussed in Section III.
Ferritin: Structure and iron-storage 65

~" ~'<r

| ..... |

~ o 5~,

Fro. 8. Channel through apoferritin shell. (a) End-on view; (b) side view: electron density
sections near centre of channel based on Banyard et al. (1978). (c)-(e) molecules which may
penetrate apoferritin channels; (c) sucrose; (d) FMNH2; (e) FADH2.

External sites thought to implicate Cd(II) of crystallization in intermolecular bridges have


b ~ n mentioned above. Another metal site, at the inner end of the channel, was found when
Nb6Cll, ~ was diffused into the crystals (Harrison et al., 1977, 1978). Nb6Cl~2~- is a nearly
spherical complex of diameter 10 A (Vaughan et al., 1950). Hence either the channels must
open slightly to allow its passage or the complex breaks down on entry.

3. The Iron Core o f Ferritin


(a) Atomic structure o f the iron cores
Granick and Hahn (1944) were the first to show that ferritin iron-cores were particulate or
"micellar" in form. The composition of the iron-cores given originally by these authors as
(FeOOH)a (FeO:OPO3H2) is subject to variation especially with respect to phosphate
content, which tends to increase relative to iron in molecules of low iron content (Fischbach
et al., 1969). Up to three-quarters of the phosphate may be lost when the iron-cores are
released from the protein by precipitation with 1 N NaOH (Granick and Hahn, 1944).
Rat liver perfusion experiments with radioactive tracers have led to the conclusion that
phosphate in ferritin is in dynamic equilibrium with the pool of inorganic phosphate
present in the organ (Van Kreel et al., 1972). Ferritins reconstituted in the absence of
phosphate, then incubated with it, retain a proportion of the anion, probably by adsorption
(Treffry and Harrison, 1978). All these experiments suggest that inorganic phosphate is
bound to iron-core particles at their surfaces and this is further indicated by the similarity
of diffraction patterns (Harrison et al., 1967; Macara et al., 1972) and M6ssbauer hyperfine
spectra (Williams et al., 1978) of native ferritin and ferritin reconstituted without phosphate.
There is now a considerable body of evidence that the iron-cores are not only particulate,
but microcrystalline. This is shown by their X-ray and electron diffraction patterns (Haggis,
1965; Harrison et al., 1967; Girardet and Lawrence, 1968), by their antiferromagnetic
66 G . A . CLEGG, J. E. FITTON, P. M. HARRISONand A. TREPI-RY

©,'(~. <
~o
o

i ° ° o

.,~

NB3~H~IB1 *
O
KEY. PC = PCHB 1~13 %
O
o1:~2 O
u~ : ~UOzF s O
UI~2 1~1~12 • o
TB : TbCI 3 o~
14,4 = I"IeH~CI ue'~:0
M~I HaZe ~ ,
NB = NL~CII~,
~ ~3 *
CD • CaSO~. ~ r%~5
~

Fzo. 9. Stereoview of two diad related subunits of horse spleen apoferritin showing heavy
atom sites. Many of the heavy atoms lie near the intersubunit contact on the inside surface of
the molecule. Diad related TB1 atoms are 0.43 nm apart. Apoferritin was crystallized from
CdSO,~ solution and CD on the outside surface forms an intermolecular bridge. Main chain
joining :~-carbon atoms with one alternative connectivity shown (as in Fig. 7b). Note anti-
parallel pleated sheet linking subunits on outside surface. (Clegg, Stansfield, Bourne and
Harrison, unpublished).

ordering at low temperatures (Blaise et al., 1965: Boad and Window, 1966) and by high
resolution electron microscopy (Massover and Cowley, 1973, 1975), see Fig. 4. Estimates of
crystallite size from X-ray powder diffraction lines gave 75 + 30 A (Harrison et al., 1967)
or 68 A (Girardet, 1969) and as this is approximately the same size as the central hole in the
apoferritin shell, it must be concluded that the iron-cores consist of a single crystallite or a
small number of such crystallites. Single crystals have been seen directly by electron micro-
scopy in some molecules (Massover and Cowley, 1973; Fig. 4). The supraparamagnetic
behaviour found when M6ssbauer spectra and magnetic susceptibility are measured as a
function of temperature (Blaise et al., 1965, 1967) is also consistent with the presence of
small crystalline particles. Iron-core size distributions have been calculated in a number of
ferritin samples (Williams et aL, 1978).
Particles resembling, but not identical with, ferritin iron-cores in their diffraction patterns,
vibrational and electronic spectra, magnetic susceptibility and M6ssbauer spectra may be
obtained by warming or gradually raising the pH of a ferric salt (Spiro et al., 1966; Van
der Giessen, 1966; Towe and Bradley, 1967; Brady et al., 1968). A wide variety of unit
cells and atomic arrangements have been proposed by these authors for ferric polymers or
for ferritin and for the latter by Harrison et al. (1967) and by Girardet and Lawrence (1968).
Most of these alternatives are based on close-packing of oxygens and hydroxyls with iron
in interstitial sites. Although arrangements with tetrahedral (Brady et al., 1968) or with both
octahedral and tetrahedral iron coordination (Harrison et al., 1967) have been suggested,
the presence of only octahedral or six coordinate iron is indicated by analysis of electronic
absorption spectra (Webb and Gray, 1974) or Extended X-ray Fine Structure measurements
of ferritin and its analogues (Heald et aL, 1979; Thiel et al., 1979) as first proposed by Towe
and Bradley (1967).
Ferritin: Structureand iron-storage 67

The iron-storage protein of E. coil exhibits different magnetic properties and does not
show the supraparamagnetism of ferritin (Bauminger et al., 1979). Thus its iron component
may differ in both particle size and atomic arrangement from that of ferritin and perhaps
also in its phosphate content.

(b) Relationship between iron-cores and protein


Although from its role in the formation of the iron-cores (Section III), apoferritin could
be expected to influence the direction of growth of the microcrystals with respect to the
protein structure, this does not seem to be the case. Both low and high angle X-ray scattering
data from ferritin crystals suggest the contrary and it is of particular interest that a Debye-
Scherrer pattern due to the iron-cores may be obtained from single crystals showing the
usual sharp diffraction spots due to the protein lattice (Fischbach et al., 1969). Thus although
it seems to provide heteronucleation sites (Macara et al., 1972) the protein shell determines
only the ultimate size and not the orientation of the iron-core particles. Nevertheless there is
some evidence from neutron scattering (Stuhrmann et al., 1976) and from electron micro-
scopy (Massover; 1978) that the inorganic particles remain attached to the protein respect-
ively in native or even in partially disrupted molecules.

III. ACCUMULATION OF FERRITIN IRON


1. Reconstitution of a Ferritin-like Molecule
Although Ferritin iron-cores contain phosphate, it may not be an essential constituent,
since much of it is los~ when iron-cores are released from protein (Granick and Hahn,
1944), and since a ferritin-like product can be reconstituted in its absence. Reconstitution
can be achieved from intact apoferritin shells, Fe(II) and an oxidant (Bielig and Bayer,
1955; Harrison et al., 1967), whereas attempts to reassemble ferritin from subunits and
iron-cores give apoferritin (Harrison and Gregory, 1968). The most convenient source of
iron for in vitro studies is Fe(NH4) 2 (SO4)2. The oxidant may be either dioxygen or
KIOa/Na2S20 3 (1/4 molar ratio). Gel electrophoresis, X-ray diffraction and electron
microscopy together have shown the iron to be inside the protein shell in the form of
particles of hydrous ferric oxide with atomic structure resembling that of ferritin iron-cores
(Harrison et al., 1967; Macara et al., 1972). A ferritin-like product is obtained even under
conditions in which, in the absence of apoferritin, either ~-FeOOH or ~-FeOOH are
obtained (Harrison et al., 1967). The crystallinity of the reconstituted iron-cores and the
distribution of iron within the population of ferritin molecules, depends on buffer, oxidant,
iron and protein concentrations. In general, those conditions giving more rapid iron uptake
give a more crystalline, more ferritin-like product with a more nearly "all-or-none"
distribution of iron among the ferritin molecules (Harrison et al., 1967; Harrison and Hoy,
1973; Macara et al., 1972; Stefanini et al., 1975). These observations, the finding that
apoferritin itself accelerates Fe(II)-oxidation (Harrison et al., 1967; Niederer, 1970;
Macara et al., 1972, 1973a; Bryce and Crichton, 1973) and the detailed kinetic studies
described in the next section led to a "crystal growth" model of ferritin formation (Macara
et al., 1972, 1973a; Harrison and Hoy, 1973).

2. Kinetics of Fe(II) Uptake: A Computer Model


The rate of ferritin formation from apoferritin and Fe(II) can be followed in a spectro-
photometer as the development of red-brown colour (characteristic of polymeric Fe(III)
"hydroxide" species (Spiro and Saltman, 1969)). Although choice of wavelength is not
critical, 420 nm (Elx~c.~ = 100) or 310 nm (E~1.~.
2,~ = 450) have proved suitable (Macara et al.,
1972, 1973a). It is also possible to follow iron uptake as loss of Fe(II), measured as pink
Fe(II)-~, ~'-bipyridine complex (Niederer, 1970). Because of the limited solubility of 02
in buffers (about 0.25 mM), KIO3/Na2S20 3 may be used when excess oxidant is required, or
when a relatively large amount of Fe(II) is to be oxidized. The shape of the progress curve
observed depends on a complex interaction of several factors including the number of iron
atoms added per protein molecule (Macara et ai., 1972), the absolute Fe(II) concentration
68 G.A. CLEGG,J. E. FITTON,P. M. HARRISONand A. TREFFRY

(Treffry and Harrison, unpublished work), and the pH and type of buffers used (Macara
et aL, 1972; Pfiques et al., 1979; Treffry and Harrison, unpublished). At pH values near 7.0,
in buffers which do not strongly chelate iron, progress curves are "hyperbolic" (simple)
at low and "sigmoidal" (complex, S-shaped or biphasic) at higher iron/protein ratios
(Macara et al., 1972, 1973a; Pfiques et al., 1979), see Fig. 10. There seems to be a critical
Fe/protein ratio of about 70 Fe atoms/apoferritin molecule below which sigmoid curves are
not observed. Development of S-shaped curves is also suppressed at low pH and in strongly
chelating buffers like Tris, HC1 or borate-cacodylate (PSques et al., 1979), see Fig. 13.
The S-shape implies that an initial slow phase is followed by incorporation of Fe at ac-
celerating rates. This behaviour is similar to that found in crystal growth or linear polymer-
ization (see for example Nielsen, 1964). Entry into a fast growth phase requires the initial

(a) • (b)
! 0'56mM 1"(3 ( 59

0"16 /(200, 0-8 / /


/ /l'5mM
°°'~ ! ~ /i /
"-
~ / o.,,~ / ~~ 0.75' ~
mN
,
~~ / ~ ~ I ~/ /

~o,//
, 1I / i/ /
L /o~
.....
. ~-.-~'~_E~ ~ ~ ' " I i
1 2 3 1 2 3
Time(min)

1"2 (c)

I
~ 0-4 200
~ 150

1 2 3 4
Time(rain)
FIG. 10. Kinetics of ferritin formation. (a) (b) Rate of Fe(II)-oxidation in presence, - - •

or absence - - - , of apoferritin. Number on graphs represent concentration of Fe(II) and


numbers of Fe atoms/apoferritin molecule. (c) Computer simulation of Fe uptake based on
model of Macara et aL (1972, 1973a) and Macara (1974).
Ferritin: Structure and iron-storage 69

formation of a nucleus (in the slow stage). Once stable nuclei have formed, rates of de-
position depend on free surface area (in general they will increase if particle size increases).
The kinetics of ferritin iron-core formation can be explained by a similar process, but with
the restriction that growth occurs within the confines of the protein shell. A simplified
computer model was developed in which an initial nucleation step, involving binding to the
protein according to the rate equation:
-diS]
dt = ks[AP°] [S]n

was followed by growth of the crystallite by addition of substrate to its free surface ac-
cording to:
i=h
- d i S ] = k~ IS]" Y~ a,
dt i =t
where K s and k~ are rate constants for nucleation and growth respectively, n and m are
orders of reaction with respect to S (the concentration of Fe(II) at time t) and A~ is the
available surface area of the ith molecule in a population of h molecules. To calculate A~ for
the ith molecule at time t a simple model was assumed: that the crystallite could be regarded
as a spherical segment and addition of Fe(II) occurred on its plane surface of area A. Since
not all molecules would nucleate simultaneously, nucleation was started for successive
molecules in the population at intervals:
1
ts=
ks [Apo] IS] n
and growth was continued until all molecules were full. Further details of the model are
given in Macara et aL (1972, 1973a), Macara (1974). Simulated progress curves based on the
computer model showed hyperbolic curves at low [S], transforming to sigmoidal curves at
higher [S]. It also predicted that if iron is added in successive amounts, initial rates of
uptake first progressively increase, then level off, when the molecule is approximately half
full, and finally decrease when growth is restricted by the protein shell. The predicted
bell-shaped plot of V~ vs. iron already incorporated (Macara, 1974) has been verified by
Harrison et al. (1974, 1978), see Fig. 11. Paques et aL (1979) did not observe the predicted
initial rise in V~ at increasing iron content (below 2000 Fe atoms/molecule). However,
their data were obtained at low pH values in complexing buffers (borate-cacodylate at
pH 5.6), conditions in which rates of transfer are very low and sigmoid uptake curves have
not been found. The model does account in a general way for the observed distributions of
iron contents among the population of molecules, when iron additions are made under
different conditions. Thus increasing kN/k~ gives a product withmolecular iron contents
broadly distributed about a mean as well as a decrease in sigmoidicity in the progress of
curves. Such results are found experimentally: 02 as sole oxidant favours the initial step
and gives faster initial rates and a more evenly distributed product; excess KIO3/Na2S203
depresses nucleation, favours growth and gives an "all or none" distribution and larger
crystallite size (Harrison et al., 1~67, 1978; Harrison and Hoy, 1973; Stefanini et al., 1975),
see Fig. 15 and below. Effects of complexing buffers can also be accommodated in the model;
their presence produces a diminished supply of free substrate leading to a dampening of
growth and "hyperbolic" progress curves. Zn(II) has been found to inhibit ferritin
formation: reasonable simulation of the kinetic data is obtained if it is assumed that Zn(II)
both competes for Fe(II) binding sites on apoferritin and acts as a crystal poison, thus
affecting both k s and ko (Macara et al., 1973b, c). The computer model is consistent with
the observed "catalytic" effect of the protein, since apoferritin concentration enters into the
rate equation for the initial "nucleation" step.

3. Mechanism of Iron Uptake by Ferritin


Although the computer model allows simulation of kinetic data, if reasonable as-
sumptions are made about the effects of conditions on various terms in the equations, a
JI~B 36:2/3 B
70 G.A. CLEGG,J. E. FITTON,P. M. HARRISONand A. TREFFRY

more detailed description of chemical events inside the ferritin molecule is required for the
development of a mechanism of ferritin formation. Any proposed mechanism must be
consistent with the relevant chemistry of iron and the chemical and three-dimensional
structure of the protein, as well as with the complex kinetic data. The mechanism of ferritin
formation developed in our laboratory will now be described and assessed below in
relation to experimental evidence, which also allow some details to be filled in. An alterna-
tive view of ferritin iron deposition will be outlined in Section III. 5.
The apoferritin shell contains six channels. These channels are believed to be large enough
to allow the passage of small ions or molecules (Fe2+aq., Fe(OH)+aq, or Fe(OH)2 aq.,
02, IO~ etc.), so that their diffusion into the internal cavity is not rate-limiting, but poly-
meric Fe(III) species cannot penetrate the channels and when formed inside are caged
within the protein shell. Once inside, Fe(II) is chelated at sites on the protein, which favour
Fe(III) binding. Oxy-groups, in particular the carboxylate anions of glutamic or aspartic
acids, would be expected to predominate at the binding sites, although other groups such as

(a)
0.03

'~ 0.02

0.01,

I I i I
10 2O 30 40
Fe ~orns/rnolecule xlO-2
FIG. 11. (a)

(b)

10 20 30 &0
F'e =omslmolecule x'lO-2
F~o. 11. (b)
Ferritin: Structure and iron-storage 71

l (c)

" "-

I I II
10 ~0 ~0 4O
Fe ~ t o ~ l m g l E u l e xl~ 2
Fio. 11. Initial rat~ of iron uptake and ~ 1 ~ by ferritin fractions as a function of iron
content. (a) (c) Average data of Harrison et al. (1978); (b) computer simulation from M a ~ r a
(1974); (a),(b) Fe uptake; (c) Fe release.

histidine may also be present. Such chelation favours Fe(III) thermodynamically and
accelerates Fe(II)-oxidation (as has been found for small chelate molecules such as EDTA,
see Bell, 1977). There may be one or more such site on the inner surface of each subunit,
including sites in close proximity, which could co-operate in the formation of a "hetero-
nucleation centre". It is unlikely that apoferritin provides hexadentate chelation. Some
co-ordination sites are first occupied by water. Subsequently protons are lost and linkage of
further Fe(III) through oxo- and hydroxy-bridges occurs. Thus a nucleus builds up. It is
at this initial phase of ferritin formation that the catalytic effect of the protein is seen
experimentally (Fig. 10a). Provision of heteronucleation sites ensures that ferritin sequesters
the available iron before it is deposited elsewhere. As the "hydrous ferric oxide" crystals,
or polymers, build up, iron is now added directly to them, without the need to pass through
specific oxidation sites on the protein. Indeed OH- and 0 2- groups on the crystallite
surfaces themselves have a higher affinity for Fe(III) and would favour the oxidation of
Fe(II) deposited on them, or preferentially bind Fe(III) already formed. Proximity of Fe(II)
ions on crystallite surfaces may result in more efficient use of oxidant (see below) giving a
further accelerating effect. The presence of several chelating c.entres inside the apoferritin
shell could give formation of more than one nucleus, but the first formed will grow at the
expense of those developing later, so that under conditions of rapid crystal growth some
molecules may contain a single mineral crystal, as observed (see Fig. 4).
It is not certain whether the crystals remain attached to the protein, or whether they
become dislodged (which seems to occur with Fe(III)-EDTA at pH 8 or above, when the
Fe(III) tends to precipitate as insoluble hydroxide, Bell, 1977).
A schematic drawing of the mechanism of ferritin formation developed by Macara et al.
(1972, 1973a, b, c), Harrison et al. (1974b, 1975, 1978) and Treffry et al. (1977, 1978, 1979)
is shown in Fig. 12.
4. Further Kinetic Data
The autoxidation of Fe(II) by O 2 is a complex process, which may involve the formation
of O ~-, HO~" (or O~-) and OH') as intermediates, and which is markedly dependent on
conditions. Thus in perchlorate media George (1954) found it to be first order in 02,
second in Fe01) and nearly independent of pH, whereas Goto et al. (1970) found it to be
first order in Fe(II), first order in O2 and second order in OH- in aqueous solution at
neutral pH. The quadratic dependence of reaction rate on OH- was attributed to hydroxyl-
ation of both Fe(II) and 02. Oxidation of Fe.(II) by H20 2 in perchlorate (pH 2.3-5.3) was
72 G.A. CLEGG,J. E. FITTON,P. M. HARRISONand A. TRI~,FFRY

, •

FIG. 12. Ferritin formation by "crystal growth" m~hanism. After initial nucleation on
apoferritin the rate of ferritin formation depends on the surface-area available for "FeOOH"
deposition (thick line), which first increases, (a) to (c), and then d~reases, (c) to (d). Based on
Macara e¢ al. (1972) and Harrison et al. (1974b).

found to accelerate with increasing pH until a high limiting rate was reached, which was
attributed to the predominance of Fe(OH)2 aq., the most rapidly reacting species, at the
higher pH values (Wells and Salam, 1968). To avoid damage to the protein shell, ferritin
formation may be followed only over a relatively narrow pH range. Initial rates of iron
uptake into apoferritin or ferritin fractions are markedly pH-dependent accelerating
sharply in the range 5.0-7.8 (depending on buffer and Fe(II) concentration) (Macara et al.,
1973a; P~ques et aL, 1979). The order of reaction was found to be 1.5 for the addition of
Fe to iron-cores in the presence of excess KIOa/Na2S203 in the pH range 6.2-7.4 (Macara
et al., 1972). The autoxidation of Fe(II) in the presence of bovine serum albumin (Macara
et al., 1972) or in the absence of any protein (Bryce and Crichton, 1973) showed a very
similar pH dependence to that of iron incorporation into ferritin, emphasizing the
similarity of these processes, and accelerating oxidation is thought to result from increasing
hydrolysis as the pH is raised (Macara et al., 1973a).
The dependence of the initial rate (Vi) of ferritin formation on Fe(II) concentration is
complex, varying with buffer, oxidant and pH, as well as with the range of Fe(lI) con-
centration and the Fe(II)/protein used. At very low Fe(lI) concentrations, at relatively low
pH values or in complexing buffers, "Michaelis-Menten" kinetics are observed with the
apparent order of reaction with respect to Fe(II) varying from unity to zero as [Fe(II)]
is increased (Macara et al., 1973a; Bryce and Crichton, 1973; Russell and Harrison, 1978;
Pftques et aL, 1979), see Fig. 13. This indicates that one or more of the reactions involved in
iron incorporation into the iron-core nucleus (measured as colour change due to the
formation of oxo-bridged Fe(III) polymers, Gray, 1975) is slow compared with the rate of
collision between Fe(II) and ferritin molecules. When the reaction is studied over a wider
range of Fe(II) concentrations in buffers without a strong affinity for iron, complex patterns
are obtained when Vi is plotted against [Fe(II)] (P~ques et al., 1979). It can be seen in
Fig. 13 that as [Fe(II)] increases, Vi first rises steeply, then more gradually, then steeply
again and finally levels off. It is of interest to note that similar "bumpy" curves have been
obtained for a number of multi-subunit enzymes (Teipel and Koshland, 1969). Such patterns
could be generated by enzymes with multiple interacting sites with either binding or catalytic
constants first decreasing and then increasing as more substrate is bound by the enzyme
(negative followed by positive co-operativity) and pseudo-linear regions in Lineweaver-
Burk plots may be obtained with more or less sharp transitions between them depending on
the relative values of the various rate constants (Teipel and Koshland, 1969; Engel and
Ferdinand, 1973). In ferritin the situation is more complicated because the nature and
Ferritin: Structure and iron-storage
73
10 - 2 ,cL,,Ia~o/mi n
'1~0,

I;/0. ~ I ~ ,~,M
~ M~o ~.O
, ........~-

/"
, ~ ~ ~ , ~ o ~ o . ~-~.n

30 ~ •~0~ 0~0~ • ~-O S


~ o ~ ~ * ~ "~ ~ ~
~ ~ .......... ,~
~ ~ ~ ~ 0 ~ 0 0 ~ ~ O, ~ ~N-&~
Me~
(a) 0.~, ,.;4 ~.;~ z~ ~.~ ~ F,~"

10-2 ~ 3 z o / m i n
30

f O.~.~O,..'~O~ • pH- 1.0



J ~,,..-----~ I~N- S.S
~,~

60 ,~ /
o . ~ "- - - ' - ' ~
~.I~,---~_...,-- ,~"
:>-
~Oe,_..o..__,..O~O__ 0....0 ~ 0.__... 0 ~ 0 "o' 0 "0-"-o pH- |.0

(b) °~' , ,4 ~ , ~, ~
~~ mM F¢ ~"
Fz~. 13. Initial velocitiesof iron deposition into apoferritin at different pH values and in
different buffers. (a) In 100 mm MOPS (morpholinepropane sulphonicacid) or MES (2-(N-
morpholino)-ethanesulphonieacid) buffer; (b) in MES buffer. From Phclueset aL (1979)with
permission.

number of the sites (first on the protein and then on the iron-core crystallites) varies. Both
the increasing number of sites generated in accordance with the "crystal growth" model
of Macara et al. (1972) and the increasing efficiency of oxidation at higher Fe(lI) con-
eentrations (see below) may contribute to "positive cooperativity". It may not be necessary
to assume "negative cooperativity" to explain the behaviour at low Fe(I1) concentrations,
although it is possible that if "heteronucleation" involves two or more Fe(II) atoms at
neighbouring sites on the protein, binding of the second (and/or third) positively charged
Fe(I/) may be more difficult. Zn(II) binding data (Macara et aL, 1973c; Treffry et al., 1977)
could also be interpreted as negative cooperativity. However low rates of iron incorporation
at low [Fe(II)] may result from relatively inefficient oxidation (see below).

5. Alternative Mechanism of Ferritin Formation


It is clear from the above discussion that the kinetic data such as shown in Figs. 10, 11
and 13 are consistent with the mechanism we have proposed. Further evidence described in
subsequent sections supports and extends this mechanism. However, an alternative
mechanism has recently been proposed (Crichton et aL, 1977; Crichton and Roman, 1977,
•978; Wauters et al., 1978; Pfiques et al., 1979), see Fig. 14. The major difference between
74 G.A. CLEGG,J. E. FITTON,P. M. HARRISONand A. TREFFRY

2 Fe'n" 02

,~ ~
0
Fe-" %0,,,F~]~
i1~

2Fe "n"

Fe rr Fe"m" ~ Fe "n'r Fe~___~


Fe \ 0 "~ Fe
// \ // \ // \ //\
O OH 0 ON 0 O~ 0 ON

4H++- I / 2 02 3H20

FIO. 14. Mechanism of ferritin formation involvingspecific Fe(H)-oxidationsites. Alternative


to Fig. 12. Based on Crichton and Roman (1978) with permission.

this mechanism and that described above and in Fig. 12 is that it invokes the presence of
active sites on apoferritin, which catalyze the oxidation of all Fe(H) atoms enterin# the
molecule. Such a mechanism has previously been considered to be unlikely (Macara et al.,
1972, 1973a; Harrison et al., 1974, 1975; Harrison et al., 1978). The mechanism pro-
posed by Crichton and colleagues proceeds in the following steps:
(1) fixation of Fe(II) atoms on pairs of adjacent polypeptide chains;
(2) fixation of dioxygen between the Fe(II) atoms;
(3) reduction of dioxygen to give a peroxo-complex in which the 022- anion is held between
two Fe(III) atoms;
(4) hydrolysis of this peroxo-complex followed lay migration of the ferric oxyhydroxide
to hetero-nucleation sites in the interior of the protein shell due to its displacement by
incoming Fe(II) atoms.
Stage (4) is considered to be reversible. Iron is thought to equilibrate between internal and
catalytic sites, so that its removal from the latter to the external medium displaces the
equilibrium towards iron release.
There are several assumptions implicit in this mechanism which lead to difficulties. The
catalytic sites promoting oxidation must have a higher affinity for Fe(II) than for Fe(III),
although the reverse is usually the case. The reaction seems to be driven towards ferritin
formation (stage 4) by the higher affinity of the iron cores for Fe(III) than for Fe(II), but
this would make the catalytic sites redundant and there is no evidence that ferritin acts as a
ferroxidase except in its own formation. The reaction in the proposed scheme seems to be
quite specific for dioxygen and yet ferritin formation can proceed in the absence of dioxygen
if the alternative oxidizing system (KIOa/Na2S~O3) is present (see below). The overall
reaction is given as a four-electron transfer from four Fe(II) to one molecule of dioxygen,
but it is not clear how this is achieved. Initially (stage 3) dioxygen accepts two electrons
from two Fe(II). Yet the stoichiometry of oxidation is found experimentally to be variable
and transfer of a single electron from one Fe(II) to dioxygen occurs at low [Fe(II)] (see
Ferdtin: Structureand iron-storage 75

below). The mechanism requires the continued presence of Fe(II) binding sites on apo-
ferfitin for ferritin formation to proceed, but when such sites are blocked by chemical
modification or by Zn(II), iron accumulation can proceed provided an iron-core nucleus is
present (see below). P,~ques et al. (1979) attempt to relate the complex kinetics of iron
uptake to the state of protonation of a group on the protein of pK about 6 (which they
conclude is histidine). They suggest that at pH below about 6 the simple form of the progress
curves and V~vs. [Fe(II) ] plots results from deprotonation of the catalytic site as the rate-
limiting step, whereas the complex curves at pH above 6 are due to this first-order rate-
limiting step gradually being displaced by a second-order reaction corresponding to the
fixation of two iron atoms at the catalytic site. However progress curves and plots of V~vs
[Fe(II)] remain simple, or hyperbolic, at low iron concentration or in complexing buffers
even at pH 7.4, when the hypothetical group of pK 6.0 should be completely deprotonated,
and in imidazole buffer the transition from hyperbolic to sigmoid progress curves may
clearly be observed at pH 7.4 as Fe(II)/apoferritin is increased (Macara et aL, 1973a).

6. Metal Ion Bindin# and Inhibition of Iron Uptake


Metal ions may be used as probes of the number and nature of binding sites, which may
be involved in ferritin formation. Non-linear Scatchard plots of binding data for Zn(II),
Cd(II), Cu(II), Mn(II) and Tb(III) to apoferritin could be interpreted as multiple classes of
independent site, negative cooperativity or a mixture of interacting and non-interacting sites
(Macara et al., 1973c). Data for Zn(II), Cd(II), Tb(III) and probably Cu(II) could be fitted
to two classes of independent site, containing a total of two high affinity sites and two-three
sites of lower affinity per subunit (Macara et aL, 1973c; Treffry et al., 1977). Tb(III) and
several other cations were found to compete with Zn(II) for both classes of site. Anomalous
binding behaviour was observed for Zn(II) at pH 6.3 which may have been due to the
presence of a group with pK in this region. This could be a histidine or a carboxyl group of
unusually high pK (perhaps owing to the proximity of another ionized carboxyl). Alterna-
tively the anomalous behaviour may have resulted from hydroxylation of Zn(II). The
N-terminal amino group is acetylated (Harrison et al., 1962; Suran and Tarver, 1965).
Several metal ion binding sites have been located in horse spleen apoferritin crystals
(crystallized from approx. 40 mu CdSO4) see Fig. 9. Sites found in the outside of the
protein shell (including Cd(II) at the intermolecular contacts) are probably of no interest in
relation to mechanism of iron uptake. No major sites were identified far inside the protein
subunits or deep within any of the channels; however, several sites were found on or near
the internal surface of the shell. Nearly all of these are located near the region of contact
between diad-related subunits. As described above the main internal Tb(IIl) and UO2(II)
sites form two clusters with the major Tb(III) site only 4.3 A from its diad-related neighbour
(i.e. Tb(III) on another subunit) and a similar distance from a UO2(II) ion on the same
subunit).
Such a site, if occupied by Fe(II), could represent the catalytic site proposed by Crichton
and colleagues. However, it could also represent a nucleation site in the mechanism we have
proposed.
Unfortunately neither Zn(II) nor Fe(II) are easy to locate by X-ray anal3,sis, because of
their low scattering power, but both binding and kinetic studies indicate these ions and
Tb(III) have sites in common (Treffry et al., 1977 and unpublished work). Stefanini et al.
(1976) found the intrinsic fluorescence of apoferritin was quenched by iron. With Fe(II)
lack of a clear end-point in the quench curve indicated relatively weak binding as compared
with Fe(III), consistent with the mechanism of Macara et al. (Fig. 12). Stefanini et al.
(1979) have shown that the enhanced fluorescence of Tb(III) bound to apoferritin is
quenched by an approximately equal concentration of Fe(II) and have interpreted their
data as due to competition of these ions for the same binding sites. Both Zn(II) and Tb(III)
inhibit iron uptake into ferritin (Macara et al., 1973b, c; Treffry et al., 1977; Harrison et al.,
1978; Stefanini et al., I979). Zn(II) is more tightly bound by apoferritin than is Fe(II)
(Harrison et al., 1978), as it is by a number of chelators containing oxy- and nitrogen ligands
such as EDTA and nitrilotriacetate (NTA) (Bell, 1977). At pH 7.0, 9 0 ~ inhibition of
76 G.A. CLOGS,J. E. FITTON,P. M. HARRISONand A. TRI~FFRY

incorporation is found with only three Zn(II) per subunit and an equal concentration of
Fe(II). At pH 6.4 complete inhibition is found with Tb(III) at Tb(lII)/Fe(II) ratios near
unity (Stefanini et al., 1979), but at pH 7.0 Tb(lI1) is less effective than Zn(lI) (Treffry and
Harrison, unpublished work). Wauters et al. (1978) report that binding of 0.54-0.77 Cr
atoms/apoferritin subunit causes 80 ~ inhibition of iron uptake. An earlier study indicated
the following order of effectiveness of metal ion inhibitors: Z n > Ni> Hg> Cd> Co>
Mg with Mg(II) having little effect (Niederer, 1970).
These studies confirm that Fe(II) binding sites on apoferritin are important for ferritin
formation as both mechanisms predict. Zn(II) binding data with ferritin indicate the presence
of several low affinity sites in addition to those found in apoferritin, which must be on the
iron-cores (Treffry et al., 1977). The finding of Harrison et al. (1978) that under conditions
in which only the high affinity sites are saturated, only 50 ~ inhibition of iron uptake was
found in a ferritin fraction, compared with almost complete inhibition with apoferritin, is
clearly inconsistent with the mechanism of Crichton and Roman (Fig. 14). Both mechanisms
also involve Fe(III) sites on apoferritin. No such binding has been found in equilibrium
dialysis experiments, but this may have been prevented by the presence of either citrate or
NTA added to avoid the formation of polynuclear Fe(III) complexes (Yreffry and Harrison,
1979). Up to about 200 Fe(III) atoms were, however, incorporated in ferritin in these
experiments, although this number was reduced in the presence of Tb(llI). Fe(IlI) ions
appeared to bind at iron-core surfaces. They did not prevent the further incorporation of
iron as Fe(II) and behaved like other iron-core Fe(III) atoms.

7. Stoichiometry and Spec(ficity o f Oxidation in Ferritin Formation


Although iron chelating centres on apoferritin may have some stereochemical specificity
(they may be less than hexadentate, there may be sites in proximity, carboxyls are probably
involved and so on), the mechanism developed by Macara, Hoy, Treffry and Harrison,
(as in Fig. 12) so far makes no special requirements about the oxidant specificity or the
number of electrons transferred per Fe(II), which could vary with conditions. In contrast
the mechanism of Crichton and Roman, Fig. 14, seems to require a two-electron transfer
specifically to dioxygen, giving a peroxo intermediate. Several workers have looked for
intermediates of Fe(II) oxidation by dioxygen in ferritin. Crichton et al. (1975) reported that
superoxide dismutase accelerates the reaction by 30-40~, but in subsequent work no
superoxide or peroxide could be detected (Melino et al., 1978; Wauters et al., 1978; Treffry
et al., 1978). The previously observed effect of superoxide has been attributed to non-
specific oxidation of Fe(II) by Cu(lI), since it was also found with the denatured enzyme
(Wauters et al., 1978). Melino et al. (1978), working at Fe(II) concentrations of 1.5-6.3 raM,
have found that a single Oz molecule oxidizes nearly four Fe(II) atoms, whether apo-
ferritin was added or not. Treffry et al. (1978), however, showed that either in the presence or
absence of apoferritin, the overall stoichiometry of oxidation of Fe(ll) by dioxygen is
dependent on the buffer, pH and Fe(II) concentration used and, in the case of ferritin
formation, on the relative concentrations of Fe(ll) and protein. Ratios down to 1.4 Fe(II)
oxidized/O2 were observed at low Fe(ll) concentrations and the ratio appeared to tend to
unity (see Fig. 15). These results indicate that oxidation mechanisms in the presence or
absence of apoferritin are similar and probably involve successive steps of one-electron
transfer. It was also noted by Treffry et al. (1978) that conditions favouring increased
Fe(H)/02 ratios are also those givin 9 rise to sigmoidal iron uptake curves with ferritin and
that both seemed to be a consequence of the presence of FeOOH nuclei on which Fe(II)
could be deposited and oxidized. In the absence of an adequate supply of Fe(lI) to reduce
O2 to water, the oxy or hydroxy radical intermediates may be dissipated by interaction with
groups on buffer or protein.
A comparison of the effectiveness of dioxygen and the KIO3/NazSzO3 systems as
oxidants has provided further evidence for the "crystal growth" mechanism of ferritin
formation (Treffry et al., 1979). Thus in the initial stage, oxidation proceeds more rapidly
with dioxygen than with KIO3/Na2S203 alone and the latter is inhibitory of oxidation by
02. However, when a small iron-core is present, oxidation proceeds just as, or more,
Ferritin: Structure and iron-storage 77

rapidly with KIO3/NaaS20 3 as with dioxygen, even when the latter is excluded (see Fig. 15).
These results strongly suggest that oxidation occurs at different sites in the two stages of
ferritin formation in agreement with the mechanism in Fig. 12, but not that in Fig. 14. The
superiority of 02 at the initial stage may be due to its ability to act as a one-or two-electron
acceptor when few iron atoms are bound to the protein.

8. Effects of Chemical Modification on Ferritin Formation


Of the amino acid side chain modifications described above only those involving carboxyl
groups have been found to effect iron uptake into apoferritin. Although not all residues are
reactive, modifications of histidine, cysteine, lysine or tyrosine are all without effect
(Niederer, 1970; Bryce and Crichton, 1973; Macara 1974; Wetz and Crichton, 1975;

(o) AF,
'1-6 ~ pH ~4

I
5 10 15
Time(rain)
FIG. 15. (a)

(b)

40

30
._=

'~" 20

10

Fe alomslmol~:ule xlO-2
FIG. 15 (b)
78 G.A. CLEGG,J. E. FITTON,P. M. HARRISONand A. TREFFRY

i ii

l (c)

I F -
AF

A~
0"06
O2

g
~ 0.0~

121
O

0-02

| ~" . . . . . ~.~A~

20 40 60 80
Time(rain)
FIG. 15. (c)

(d)

~ ~ m ~

~ 2

11 m !
~ Ov~ ~
mM Fe~)
FIG. 15 (d)
FIG. 15. Iron uptake by apoferritin and ferritin fractions. (a) and (c) progress curves with
different oxidants; (b) distribution of Fe in product after reconstitution; (d) variable stoichio-
rnetry of Fe(II)-oxidation. Arrows in (a) indicate remixing. In (c) F is ferritin fraction con-
taining 500 Fe atoms/molecule. In (d) aDoferritin concentrations adjusted to give 22, 222 or
2220 Fe atoms/molecule (no apoferritin in - - - ) Data of Treffry et al. (1978, 1979) aad
Harrison et aL (1978)

Treffry e t aL, 1977). Wetz and Crichton (1976) found that reaction o f eleven carboxyl
groups in apoferritin with glycine amide and carbodiimide completely abolishes its ability
to take up iron. F o u r o f these are protected in full ferritin and the derived apoferritin has the
same ability to incorporate iron as the unmodified protein. This implies that the un-
modified carboxyls are on the inside surface o f the protein shell. Examination of ferritin
fractions by Harrison e t al. (1978) showed that increasing iron content gave progressive
Ferritin: Structureand iron-storage 79

protection against loss of activity after reduction. The modified ferritins were also able to
incorporate iron, presumably on the existing iron-cores. These results lend further support to
the mechanism of Macara, Hoy, Treffry and Harrison (Fig. 12). Carboxyl modification also
prevents the binding of Zn(II) ions to apoferdtin confirming that they share common sites
with FeflI).
The combination of metal ion binding, chemical modification and kinetic data, strongly
suggest that one or more of the Tb(III) binding sites seen in the three dimensional structure
by isomorphous replacement also represents the site or sites at which iron is bound and that
carboxyl groups are involved in this binding. One or more of these carboxyls could be
shared by two iron atoms in a close-pair or neighbouring Fe(III) might possibly be/~-oxo
bridged through an 0 2- atom (alternatives suggested for methemerythrins, see Stenkamp
et al., 1976; Hendrickson, 1978).

IV. MOBILIZATION OF FERRITIN IRON


1. Last-in-first-out
It has been observed experimentally that if iron labelled with 5gFe is added in vitro
either as Fe(II) (Hoy et al., 1974b) or as Fe(III) (Treffry and Harrison, 1979) to ferritin
molecules already containing a small iron-core, and all the iron present is gradually released,
then the ~gFe label appears predominantly in the first iron to be mobilized. A similar result
has been obtained recently when 59Fe is injected into rats and iron released from a purified
ferritin fraction isolated 5 hr later from liver (Treffry and Harrison, unpublished work).
These results have been explained by the particulate nature of the iron-cores and by the
proposed mechanism of ferritin formation (Macara et al., 1972; see Section III) in which
iron-cores are built by addition of incoming Fe atoms to surface sites on their crystallites.
Indeed this view of the ferritin molecule prompted the experimental work. A corollary to
iron addition on surface layers is that it is also released first from these layers, since this
iron is more available to attack by reagents than iron in internal positions. A second
corollary is that relative iron release rates of ferritin fractions should bear a similar
relation to molecular iron content as found for rates of iron uptake (Fig. 11) and that per
iron atom, iron in ferritin of low iron content is more readily available, since a relatively
high fraction of it is in surface layers. Both of these predictions have been confirmed
experimentally (Harrison et aL, 1974b, 1978; Harris, 1978; Jones et ~71., 1978). An alternative
view of ferritin iron-cores is that they should be considered as coiled up linear Fe(III)-
hydroxy polymers (Brady et al., 1968; Heald et al., 1979) rather than as micro-crystals. In
this case "last-in-first-out" might be explained by removal of iron from the end of the
polymer, but, to account for the observed relation between release and iron content, further
assumptions would have to be made. Inorganic phosphate, which is present in native
ferritin (Granick, 1946), and, which is retained after incubation by reconstituted ferritin
(Treffry and Harrison, 1978) is also found largely to be among the "first out", when iron is
released by reductants, thus suggesting it is adsorbed on iron-core crystallite surfaces
(Treffry and Harrison, 1978).

2. Mechanisms o f Iron Release


Iron may be removed from ferritin by action of either reductants, e.g., dithionite (Granick
and Michaelis, 1943), thioglycollate (Crichton, 1973) or reduced flavins (Sirivech et aL,
1974) or of chelators without reduction, e.g. desferrioxamine (Wohler, 1963; Harris, 1978),
nitrilotriacetate (Pape et al., 1968) on rhodotorulic acid (Crichton et al., 1980). In some
cases, e.g. with thioglycollate or 1,10-phenthroline (Jones and Johnston, 1967), both re-
duction and chelation may play a part. Rates of release vary enormously with reagents and
conditions, but action of Fe(III)-chelators alone is generally slow. Broadly there are two
proposed alternative means of attack of these reagents. One is that they penetrate the shell
and act directly on the iron-cores (Harrison et aL, 1974, 1975, 1978; Jones et al., 1978)
and the other is that iron-core Fe is in equilibrium with Fe-binding sites on the protein
shell and release is from the latter sites. The second mechanism was first proposed by Jones
80 G.A. CLEGG,J. E. FIT~rON,P. M. HARRISONand A. TREFFRY

and Johnston (1967) to explain the slow release by 1,10-phenanthroline and has since been
elaborated by Crichton (1976) and Crichton and Roman (1978), who further proposed that
Fe was removed by ~,~'-bipyridine or by reduced flavin mononucleotide by attack at the
putative active centre containing two Fe atoms (see Section III). The action of the latter
reagents appears to be biphasic. With c~,~'-bipyridine there is a short fast "burst" as a few
atoms are released, which is then followed by more gradual release. Crichton and Roman
(1978) consider that the "burst" is due to removal of Fe from the "active centres" and that
slow release is determined by the time taken for iron to reach the active centres from the
iron-core. With F M N H 2 an initial slow phase is observed followed by faster release.
Working with F M N / N A D H Crichton et al. (1975) attributed the slow phase to dissipation
of reducing equivalents by Oz present initially in the system. F M N H z is considered by these
authors to bind to the protein and act as a "coenzyme". However, Jones et al. (1978)
observed an initial lag with several dihydroflavins under anaerobic conditions, which they
thought might be due to hindered access to the core due to the small space initially avail-
able to reductant in the central cavity. Saturation kinetics were found at high concentrations
of dihydroflavins and this might be because rates were partially determined by slow
diffusion through the channels, since no evidence for binding could be found. Relatively
slow reduction of ferritin iron (but not of synthetic Fe(llI)-polymers) by polyanionic
dihydroflavins might reflect slower rates of penetration into the ferritin shell by these
derivatives. However, relative rates of reduction of FeC13: Riboflavin > FMN > FAD
in presence of N A D H have been reported (Michelson, 1977) and it is possible that rates of
attack on isolated ferritin iron-cores by these molecules could vary in a similar manner.
It is difficult to reconcile the 105-fold differences in rates of release observed experimentally
(see Harrison et al., 1980) with the view that the rate-determining step is release from iron-
core to protein sites. On the other hand very slow release has been observed to transferrin
molecular weight of about 76,000, which clearly cannot penetrate the channels (Harris,
1978). However, a chelating buffer, Tris, HC1, was used and this, rather than transferrin,
may have been the direct mediator of release. Flavins may just be able to penetrate the 4-
fold channels (see Section II). Penetration is suggested by the finding that their attachment
to agarose beads prevents reduction of ferritin iron (Jones et al., 1978) but not of isolated
iron-cores. Maximum release rates observed with F M N H z (Jones et al., 1978) and with
thioglycollate (Harrison et al., 1978) correspond to about six Fe/mol./sec. With dithionite at
pH 4.8 release rates of up to eight times this may be found and similar maximum rates of
iron incorporation have been observed (Harrison et al., 1978), or about 0.3/k/sec. Glucose
and amino acids have also been reported to take less than 1 sec to enter the apoferritin
cavity as judged by low angle neutron scattering (Stuhrmann et al., 1975). Penetrations by
ions in X-ray diffraction analysis (Sections II. 2(e)) by several reagents causing chemical
modification inside the shell (Sections II. 1 and (III. 8) and by negative stains (e.g. Van
Bruggen et al., 1960) is also evident. These observations and the lack of specificity of
releasing agents seem to make complex mechanisms involving coenzyme sites and/or
shuttling of Fe atoms or electrons through the protein shell unnecessary.
Moreover the finding of a bell-shaped dependence of initial release rates on iron content
similar to that for iron uptake, Fig. 11, for a wide variety of reagents including 1,10-
phenanthroline (Harrison et al., 1974), cysteine (Crichton et al., 1975), thioglycollic acid
(Harrison et al., 1978) and dihydroflavins (Jones et al., 1978), despite very large differences
in absolute rates, strongly suggests that diffusion through the protein channels is not
significantly rate-determining and that all these reagents directly attack the iron-cores after
penetration of the shell.

V. I S O F E R R I T I N S
In the functional studies discussed in Sections Ill and IV ferritin has been treated as if it
were a homogeneous protein composed of identical molecules. Strictly speaking this is not
the case. Evidence for heterogeneity has already been indicated (Section |I. 1), although it is
not clearly established whether it is post-translational or genetic in origin (or both).
Molecular heterogeneity might reflect heterogeneity in cell populations in a given tissue
Ferritin: Structureand iron-storage 81

each with its distinctive ferritin, or all cells might contain hybrids. Microheterogeneity has
been observed by electrofocusing within ferritin from a uniform population of tadpole
red cells (Brown and Thiel, 1976), indicating subunit charge heterogeneity. There is some
evidence for metabolic differences in isoferritins from the same animal. Thus, for example,
the two electrophoretically separable ferdtins in rat heart seem to have different rates of
synthesis and turnover (Vulimiri et al., 1977a, b) and also to respond differently to iron
loading (Powell et al., 1975). Hence they may belong to different cell populations. In liver
perfusion experiments Halliday et al. (1979) have shown that ferritin from rat serum, heart,
liver and kidney are cleared from perfusion media at decreasing rates, which may be related
to carbohydrate content. Cancer tissues may produce abnormal ferritins often low in iron
content (Linder et al., 1972) and in leukaemic leucocytes ferritin biosynthesis seems not to
be stimulated by iron (White et al., 1974). In malignant lymphoma ferritin appears to be a
tumour-associated antigen (Moroz et al., 1977) and when isolated from tumour-involved
spleen in Hodgkin's disease it can be distinguished from normal ferritin in the leucocyte
migration inhibition test (Hancock et al., 1979).
Here, however, we are concerned mainly with the structure and iron-storage properties
of ferritin. When ferritins isolated from horse or human spleen or liver are fractionated with
respect to iron content or isoelectric point, the two properties are found to correlate roughly,
except that pI increases with iron content in the horse and decreases in the human ferritins
(Ishitani et al., 1975; Russell and Harrison, 1978; Wagstaff et al., 1978). When iron is
removed and rates of iron incorporation into the fractions are compared; those originally
containing least iron take up iron at the slowest rates (Russell and Harrison, 1978; Wagstaff
et al., 1978). It is not clear whether these apparent functional differences are properties of
the native protein shell, perhaps related either to "H subunit" content or to some in vivo
activation or degradation process, or whether they are artifacts produced during preparation
of the fractions due to differences in stability related to iron content. Recent unpublished
data of Treffry and Harrison on rat liver ferritin show no clear correlation of iron uptake
rates and pI and hence do not support the view that rates of iron uptake are related to the
proportion of a catalytically active subunit presumed to vary through the pl range.
The heteropolymer hypothesis has structural implications. Comparison of crystalline
ferritins from several species by X-ray diffraction (see Section II.2) suggests a considerable
degree of structural conservation (certainly within the ferritins from liver or spleen of rat,
horse, dog and man--horse and dog ferritins have also been co-crystallized, Harrison,
unpublished work). Examination of the three-dimensional structure of horse spleen apo-
ferritin shows a tightly organized structure with 60 % helix content. It seems likely that
allowable amino acid substitutions within the shell would be conservative and most unlikely
that many additional amino acids per subunit could be incorporated within the existing
shell volume. Thus while hybrid molecules of two subunits with a limited number of
substitutions (especially at surface positions) may be possible, LH hybrids of the 19,000
(L) and reported 21,000 (H) molecular weights, are structurally improbable unless the
approximately twenty additional amino acids could be accommodated on the surface of the
compact shell structure. Since molecules, thought to be H-rich, may be of high iron content
(Arosio et al., 1978; Wagstaff et al., 1978), additions on the inside of the shell seem un-
likely, hence the supposed additional peptide, or peptides, would have to be placed at the
outside surface of the molecule. Whereas the subunit shown in Fig. 7a has its C-terminus
inside the shell, it may be noted that in the subunit conformation shown in Fig. 7b both
N-(acetyl) and C-termini are on the external molecular surface. In such case it is con-
ceivable that L-subunits could be formed from H by proteolytic removal of a C-terminal
peptide after shell assembly. Detailed comparisons of primary structure of the isoferritins
are needed to verify this and other possible inter-relationships between the subunit types.
Limited evidence for differences in molecular volume has come from gel-filtration behaviour
which suggests that "H-rich" molecules (e.g. heart ferritin) may be larger than L-rich
ferritin (Drysdale et al., 1977). Differences in surface antigens (not, necessarily connected
with size) are indicated by cross-reactivity studies with antibodies to heart and liver or
spleen ferritins (Hazard et al., 1977; Wagstaff et al., 1978). Electrophoresis of heart and liver
82 G.A. CLEGG,J. E. FITTON,P. M. HARRISONand A. TREFFRY

ferritins co-assembled from subunits has provided evidence that hybrid molecules may form
under laboratory conditions (Lavoie et al., 1977), although this does not prove their natural
occurrence. Hazard et al. (1977) report that in human liver ferritin only the most basic
species (considered by them to be a L24H o homopolymer) crystallizes from the mixture,
but, in the case of horse spleen ferritin, crystallization of all species in the isoelectric range
occurs (Russell and Harrison, 1978). It is difficult to see how this is possible, if molecules
containing significant amounts of H subunits are present, and if the H subunits contain a
2000 molecular weight insertion (at C-terminus or elsewhere in the sequence) located on the
molecular outside surface. It may be that the apparent size differences are exaggerated by the
method used. Not all workers have been able consistently to detect two subunits in SDS-
polyacrylamide gel electrophoresis and it is clear that further experimentation is necessary
for this problem to be solved.

VI. C O N C L U S I O N
The last ten years has seen a stride forward in our understanding of the ferritin molecule
and how it functions as an iron-storage protein. X-ray analysis at 2.8 ~, resolution shows
apoferritin as a hollow shell formed from bundles of long helices (in 24 subunits) interwoven
to form a symmetrical (432) bilayer covering for the ferritin iron-core. Apoferritin allows
access to the iron-core, the storage centre of the molecule (which has space for over 4000
Fe(III) atoms), through six identical channels around molecular 4-fold axes, each roughly
10/~ across. Each channel is circumscribed by four short helices. The helix content of horse
spleen apoferritin estimated from the electron density map agrees well with previous
estimates from ORD and CD measurements and analysis of subunit packing suggests
possible pathways of self-assembly for the shell.
The assembly of ferritin is an important part of its function. It was observed more than
30 years ago that iron stimulates the production of ferritin molecules and subsequent work
indicates that the first biosynthetic step is to make apoferritin. Apoferritin shells then
accumulate iron giving ferritin. Inside the protein shell iron is stored in the form of micro-
crystals of hydrous-ferric oxide-phosphate and details of the structure of these particles
has been observed directly by high resolution electron microscopy, and deduced from X-ray
diffraction and EXAFS measurements and from electronic spectra. In vitro studies of the
kinetics of ferritin formation indicate that microcrystal surfaces may be an important
determinant of the rate at which ferritin accumulates its iron. The only really successful
method of ferritin reconstitution in vitro starts from apoferritin and ferrous iron, apo-
ferritin is found to act as a catalyst at least in the first stages of Fe(II) oxidation. This is
probably an effect of chelation by groups on the protein, although an alternative mechanism
has been postulated in which the proximity of two apoferritin-bound Fe(II) atoms allows
transfer of two electrons to an Oz molecule. Such a mechanism cannot account for the
anaerobic formation of ferritin in the presence of oxidants other than dioxygen, which has
been observed. To provide a reserve function ferritin iron must be available. In vitro iron
can be released from intact molecules by action of reductants and/or chelators. Use of
radioactive isotopes has shown that ferritin iron is not uniformly available, but a "last-in-
first-out" principle is followed, in accordance with the particulate nature of the iron-core.
Details of ferritin biosynthesis, turnover and iron mobilization in vivo remain to be
determined, although studies in animals and in cell free systems have suggested a mechanism
for the stimulation by iron of apoferritin biosynthesis, which postulates relief of inhibition
of translation by apoferritin subunits due to the involvement of iron in the assembly of
ferritin molecules.
In the next few years we may anticipate a further step towards the elucidation of
structure-function relationships. The amino acid sequence of horse spleen ferritin is nearing
completion and others are on the way. A fully interpreted and refined three-dimenSional
structure is now foreseeable. Chemical modification, metal ion binding and kinetic studies,
together with X-ray analysis of heavy metal derivatives have already shown the presence
of cation binding sites on the inner surface of the protein shell and implicated carboxyl
Ferritin: Structure and iron-storage 83

groups in their binding. It is hoped that such residues can be identified in the three-
dimensional structure and the role of the protein in iron accumulation ascertained.
A number of aspects of ferritin structure and metabolism have been referred to only
bdetty, if at all, in this review. One outstanding question, which must be solved, is the
nature and role of isoferritins. While there is a considerable body of experimental evidence
to suggest that tissue ferdtin protein is heterogenous, the nature of this heterogeneity
remains elusive. Thus, while ferdtin appears to be a hybrid of two subunits, which are
electrophoretically distinguishable in some experiments, doubts remain as to whether these
subunits have different primary structures and, if so, to what extent they differ, or whether
the differences are due, wholly or in part, to post-translational modification. X-ray analysis
only shows us the average of all subunits in the crystal, but their structural equivalence, the
ability of ferritin molecules from different tissues (spleen, liver, heart) to crystallize in
similar forms and the co-crystallization of molecules with isolectric points varying over
about one pH unit suggest that the subunit differences may not be large. It has been
suggested that ferdtins are glycosylated and that the extent of glycosylation may very with
tissue or cell type. A glycosylated form of ferdtin has been found in serum, but serum
ferritin itself seems not to be a uniform molecule. The nature, origin and physiological
role of serum ferdtin has not been discussed here and alteration in ferritin properties or
levels in various pathological states has received only passing mention. These questions are
important because of the increasing use of serum ferritin assay as a monitor of disease.
However, a thorough knowledge of the structure and metabolism of ferdtin under normal
conditions provides the foundation on which such alterations can ultimately be understood.

ACKNOWLEDEGMENTS
We thank the Medical and Science Research Councils, the Royal Society and the Blood
Research Fund for financial support. We are also grateful to Mr. Graeme Dolderson and
Mrs. Janet Sowerby for expert assistance. We are greatly indebted to those who have
provided illustrative material (see Figure legends) and to our office staff for typing the
manuscript.

REFERENCES
,Ms~, P. (1977) In; Ciba Foundation Symposium, 51 (New Series). Iron Metabolism, pp. 1-17, Elsevier,
Amsterdam.
At.p~'r, E., Coszo~4,R. L. and D~tVS~ALe,J. W. (1979)Nature 242,194-195.
A~osIo, P., ADEI~t~N,T. G. and D~tYSDAL~.,J. W. (1978)J. biol. Chem. 253, 4451-4458.
BAe^, A. (1969)J. Biochem. 65, 915-923.
BAI~^,A., MAY, M. E. and F~-I, W. W. (1977) Biochim. biophys. ,4cta 491, 490-496.
BA~VARD,S. H., STAM~RS,D. K. and HA~JUSO~,P. M. (1978)Nature 271, 282-284.
BAulvn~ctE~t,E. R., Con~s, S. G., OFER,S. and RAcn~m.~q'rz, E. A. (1979) Proc. hath. ,4cad. Sci. U.$.,4.
76, 939-943.
BAUm~t3E~t,E. R., Dxctso~, P. P. E., Cox~e~,S. G., I~w, A. and Or~t, S. (1980) In publication.
BECK,G., BOLUaCK,C. and BEet, J. D. (1974) FEBS Lett. 47, 314-317.
BELL,C. F. (1977)Metal Chelation Principles and Applications, Clarendon Press, Oxford.
l~I.IO, H. G. and Bayer, E. (1955) Natarwissenschachaften. 42, 125-126.
]]Jest, I. and Frail, W. (1971) Biochemistry 10, 2844-2848.
BL~s~, A. ~ P ~ R ' r , J. and Gno~D~t', J. L. (1965) C.r. ,4cad. Sci. Paris 261, 2310-2313.
BLAISe,A. F~toN, J. GIRARDFr,J. L. and L A N C E , J. J. (1967) C.r. Acad. Sci. Paris 265B, 77-84
BLOOm~, A. C., ~ C t ~ s s , J. N., BI~COGNE,G., ST~3EN,R. and KLu~, A. (1978) Nature 276, 362-368.
BLtmD~LL,T. L. and Jol-n~o~, L. N. (1976). In Protein Crystallooraphy, Academic Press, London.
BOAS;J. F. and WI~ct~ow,B. (1966) Aust. J. Phys. 19, 573-576.
BOMrORD,A., LIS,Y., MCF,*,~J~^~r~,I. G. and WILHAMS,R. (1977)Biochem. J. 167, 309-312.
BO~ORD, A., BERGS, M., L~s, Y. and WILL~J~S,R. (1978) Biochem. biophys. Res. Comrmm. 83, 334-341.
Bourom3, A. and Mul~o, H. N. (1980) In: Iron in Biochemistry and Medicine 2nd ed. (eds. A. JAcoaa and
M. WoRwooo), Academic Press, London (in press).
BRADY,G. W., KURKJIAN,(2. R., LYDEN, E. F. X., RosIN, M. B., SAL3"MAN,P., SPIRO, T. and TeltZIS, A.
(1968) Biochemistry 7, 2t55-2191.
B~tovv~,J. E. and Ttxe~ E. C. (1976) Br. ~. Haematol. 34, 663-665.
B~tYc~,C. F. A. and CIlca'ro~, R. R. (1971)J. biol. Chem. 246, 4198-4205.
BRYCE,C. F. A. and Cltlc:Irrols, R. R. (1973) Biochem. J. 133, 301-309.
BltYc~, C. F. A., M^G~SaO~, G. M. and C~cwro~, R. R. (1978) FEBSLett. 96; 257-262.
COL~r-CAssAItT,D. (1977) PhD 'thesis. Universit6 de Louvain-la-Neure, Belgium.
Coo~, S. F. (1929)-,/. biol. Chem. 82, 595-609.
CIUCH'rON,R. R. (1973) Struct. Bond. 17, 67-134.
84 G. A. CLEC~, J. E. FITTON,P. M. HARRISONand A. TREFFRY

CRICI-rrON, R. R. (1975). In: Proteins of Iron Storatte and Transport in Biochemistry and Medicine (ed.
R. R. CRICHTON)pp. 253-260.
CRicrrrON, R. R. (1976). In: Iron Metabolism and its Disorders (ed. H. KIEF.) pp. 81-89. Excarpta medical
Amsterdam.
CRICaTON, R. R. and BRYCE,C. F. A. (1973) Biochem. J. 133, 289-299.
CRICr~TON,R. R. and ROMAN,F. (1977) Biochem. Soc. Trans. 5, 1126-1128.
CRierrroN, R. R. and ROMAN,F. 0978) J. Mol. Catal. 4, 75-82.
CRICHTON,R. R., MILLAR,J. A. CUMMING,R. L. C. and BRYCE,C. F. A. (1973) Biochem. J. 131, 51-59.
CRicrrroN, R. R., WAtrr~RS, M. and ROMAr%F. (1975) In: Proteins of Iron Storat~e and Transport in Bio-
chemistry and Medicine (ed. R. R. CRICaTON)pp. 287--294. North Holland, Amsterdam.
CRICHTON,R. R., COLLET-CASSART,D., PONCE-ORTIZ,Y., WAUTERS,M., ROMAN,F. and P.~QUES,E. (1977)
In: Proteins oflron Metabolism (eds. E. B. BROWN,P. AISEN,J. FIELDINGand R. R. CRICHTON)pp. 13--22
Grune and Stratton, New York.
CRICHTON,R. R., PONCE-ORTIZ,Y., KOCH, i . H. J., PARFAIT,R. and STUHRMANN,H. B. (1978) Biochem. J.
171, 349-356.
CRICHTON,R. R., ROMAN,F. and ROLAND,F. (1980) FEBS Left. 110, 271-273.
CYNKIN, i . A. and KNOWLTON,i . (1977) In: Proteins of Iron Metabolism (eds. E. B. BROWN, P. AISEN,
J. FIELDINGand R. R. CRICHTOS)pp. 115-120. Grune and Stratton, New York.
DAVID, C. N. and EASTERBROOK,K. 0971)J. cell Bid. 48, 15-28.
DRYSDALE,J. W. (1977) In: Ciba Foundation Symposium 51 (new series) Iron Metabolism, pp. 41-57. North
Holland, Amsterdam.
DRYSDALE,J. W. and MUNRO,H. N. (1966)J. biol. Chem. 241, 3630-3637.
DRYSDALE,J. W. and SHAFRITZ,R. A. (1975) Biochim. biophys. Acta 383, 97-105.
DRYSDALE, J. W., ADELMAN, T. G., AROSIO, P., CASSAREALE,D., FITZPATRICK,P., HAZARD, J. T. and
YOKOTA,M. (1977) Semin. Haematol. 14, 71-88.
EL-SHOBAKI,F. A. and RUMMEL,W. (1977) Res. exp. Ned. 171, 243-253.
ENGEL, P. C. and FERDINAND,W. (1973) Biochem. J. 131, 97-105.
FANKUCHEN,1. (1943)J. biol. Chem. 150, 57-59.
FARRANT,J. L. (1954). Biochim. biophys. Acta 13, 569-576.
FERN, E. B. and GARLICK,P. J. (1976) Biochem. J. 156, 189-192.
FIELDING J. and SPYER, B. E. (1977) In: Proteins of Iron Metabolism (eds. E.B. BROWN, P. AISEN, J.
FIELDINGand R. R. CRICHTON)pp. 311-318. Grune and Stratton, New York.
FlSCHBACH, F. A., GREGORY,D. W., HARRISON,P. M., HoY, T. G. and WILLIAMS,J. M. (1971) d. Ultra-
struct. Res. 37, 495-503.
FISC'rlnAClt,F. A. HARRISON,P. M. and HoY, T. G. (1969) J. molec. Biol. 39, 233-238.
GEORGE,P. (1954) J. chem. Soc. 4349-4359.
GmAnXmT,J. L. (1969) Thesis, University of Grenoble.
GIRAROET,J. L. and LAWRENCE,J. J. (1968) Bull. Soc. Fr. Mineral, Crystillogr. 91, 440-443.
GLASS, R. n . and DOYLE,D. (1972) J. biol. Chem. 247, 5234-5242.
GOTO, K., TAMURA,H. and NAGAYAMA,i . (1970) Inory. Chem. 9, 963-964.
GRANICK,S. (1946) Chem. Rev. 38, 379-403.
GRANICK,S. and HAHN, P. (1944) J. biol. Chem. 155, 661-669.
GRANICK,S. and MtCHAELIS,L. (1943). J. biol. Chem. 147, 91-97.
GRAY, H. B. (1975) In: Proteins of Iron Starage and Transport in Biochemistry and Medicine (ed. R. R.
CRicI-rro~) pp. 3-13. North Holland, Amsterdam.
HAGGIS,G. H. (1965) J. molec. Biol. 14, 598-602.
HALLIDAY,J. W., MACK, O. and POWELL,L. W. (1979) Br. J. Haematol. 42, 535-546.
HANCOCK,B. W., BRUCE,L., MAY, K. and RICHMOND,J. (1979) Br. J. Haematol. 43, 223-233.
HARRIS, D. C. (1978) Biochemistry 17, 3071-3078.
HARRISON,P. M. (1959) J. molec. Biol. 1, 69-80.
HARRISON, P. M. (1963) J. molec. Biol. 6, 404-422.
HARRISON, P. M. (1977) Semin. HaematoL 14, 55-70.
HARRISON,P. M. and GREGORY,D. W. (1968) Nature Lond. 220, 578-580.
HARRISON, P. M. and HoY, T. G. (1973) In: Inorganic Biochemistry, (ed. G. G. EICHHORN)pp. 253--279.
Elsevier, New York and London.
HARRISON,P. M. and TREFFRY,A. (1979) Specialist Periodical Report on Inoryanic Biochemistry, 1, 120-158,
Chemical Society, London.
HARRISON,P. M., FISCHBACK,F. A., HoY, T. G. and HAGGIS,G. H. (1967) Nature Lond 216, I 188-1190.
HARRISON,P. M., HOFMANN,T. and MAINWARING,W. I. P. (1962)J. molec. Biol. 4, 251-256.
HARRISON,P. M., HOARE,R. J., HOY, T. G. and MACARA,I. G. (1974a) In: Iron in Biochemistry andMedi-
cine. (eds. A. JACOBSand M. WORWOOD)pp. 73-114. Academic Press, New York.
HARRISON,P. M., HOY, T. G., MACARA,I. G. and HOARE,R. J. (1974b) Biochem. J. 143, 445-451.
HARRISON, P. M., HoY, T. G. and HOARE, R. T. (1975) In: Proteins of Iron Stora#e and Transport in BiD-
chemistry and Medicine (ed. R. R. CRICHTON)pp. 271-278 North Holland, Amsterdam.
HARRISON, P. i . , BANYARD,S. H., HOARE, R. J., RUSSELL,S. i . and TREFFRY,A. (1977) In: Ciba Found-
ation Symposium 51 (new series). Iron Metabolism, pp. 19-35. Elsevier, Amsterdam.
HARRISON,P. M., BANYARD,S. H., CLEGG,G. k., STAMMERS,n . K. and TREFFRY,A. (1978) In: Transport
by Proteins (eds. G. BLAUERand H. SUND)pp. 260-272. Walter de Gruyter, Berlin and New York.
HARRISON,P. M., CLEGG,G. A. and MAY, K. (1980) In: Iron in Biochemistry and Medicine (eds. A. JACOBS
and M. WORWOOD)Academic Press, London (in press).
HAZARD,J. T., YOKOTA,M., AROSIO,P. and DRYSDALE,J. W. (1977) Blood49, 139-146.
HEALO, S. i . , STERN,E. A., BUNKER, B., HOLT, E. M. and HOLT, S. L. (1979) J. Am. chem. Soc. 101,
67-73.
Ferritin: Structure and iron-storage 85

HENDRICKSON,W. A. and WSmD, K. B. (1977) J. biol. Chem. 2$2, 3012-3018.


HICKS, S. J., DRYSDALE,J. W. and MUNRO,H. N. (1969) Science 164, 584-585.
HOARI~,R. J., HSmRmON,P. M. and HeY, T. G. (1975a) Nature 255, 653-654.
HOARE, R. J., HARRISON,P. M. and HeY, T. G. (1975b) In: Proteins oflron Storage and Transport in Bio-
chemistry and Medicine (ed. R. R. CRICHTON)pp. 231-236. North Holland, Amsterdam.
HeY, T. (~. and HARRISON,P. M. (1976) Br. J. Haematol. 33, 497-504.
HeY, T. G., HARRISON,P. M. and HOARE,R. 3. (1974a) J. molec. Biol. 86, 301-308.
HeY, T. G., HARRISON,P. M. and SHARBm,M. (1974b) Biochem. J. 138, 603-607.
HtmEm~AN, A. and BARAHONA,E. (1978) Biochim. biophys. Acta. $33, 51-56.
HUEBERS,H., HUEBERS,E., RUMMEL,W. and CRICHTONR. R. (1976) Eur. J. Biochem. 66, 447-455.
IANCU,T. C. and NEUST~IN,H. B. (1977) Br. J. Haematol. 37, 527-535.
IANCU, T. C., LICHTERMAN,L., NUESTEIN,H. B. (1978) Israel J. reed. Sci. 14, 1191-1201.
ISHITANI,K., NIITSU, Y. and LISTOWSKY,I. (1975) J. biol. Chem. 250, 3142-3148.
JON~, M. M. and JOHNSTON,D. O. (1967) Nature 216, 509-510.
Jos~s, T., SPENCER,R. and WALSH,C. (1978) Biochemistry 17, 4011--4017.
KATO, T. (1969) Seikagaku. 41, 27-37.
KONIJN, A. M., BALIGA,B. S. and MUNRO,H. N. (1973) FEBS Lett. 37, 249-252.
LAU~ERGER, M. V. (1937) Bull. Soc. chim. Biol. 19, 1575-1582.
LAVOIE, D. J., MARCUS,D. M., ISHIKAWA,K. and LISTOWSKY,L (1977) In: Proteins of Iron Metabolism
(eds. E. B. BROWN, P. AISEN,.J. FIELDINGand R. R. CRICHTON)pp. 71-78. Grune and Stratton, New
York.
LAVOIE,D. J., ISHIKAWA,K. and LISTOWSKY,I. (1978) Biochemistry 17, 5448-5454.
LEACH, B. S., MAY, M. E. and FISH, W. W. (1976) J. biol. Chem. 251, 3856-3861.
LEE, J. C. K., LEE, S. S. C., SCHLESINGER,K. J. and RICHTER,G. W. (1975) Am. J. Pathol. 80, 235-239.
LEE, S. S. C. and RicmmR, G. W. (1977a) J. biol. Chem. 252, 2046-2053.
LEE, S. S. C. and RICHTER,G. W. (1977b) J. biol. Chem. 252, 2054-2059.
LINDER, M. C., MOOR,J. R., MUNRO,H. N. and MORRIS,H. P. (1972) In: Isozymes andEnzyme Regulation
in Cancer (eds. WEINHOUSE,S. and ONO, T.), Vol. 13, pp. 299-313, Gann Monograph on Cancer
Research.
LINDER, M. C., MOOR,J. R., SCOTT,L. E. and MUNRO,W. H. (1973) Biochim biophys. Acta 297, 70-80.
LINDER,M. C., MOOR,J. R. and MUNRO,H. N. (1974)J. biol. Chem. 249, 7707-7710.
LINDER, M. C., MOOR,J. R., MUNGO,H. N. and Momus, H. P. (1975) Biochim. biophys. Acta 386, 409-421.
LINDER, M. C., Z~.HRINGER,J., BALIGA,B. S., DRAKE, R. L., BARITES,B. and MUNRO, H. N. (1977) In
Proteins of Iron Metabolism (eds. E. B. BROWN,P. AISEN,J. FIELDINGand R. R. C~CHTON)pp 121-129.
Grune and Stratton, New York.
LISTOWSKY,I., BLAUER,G., ENGLAND,S. and BETHEIL,J. J. (1972) Biochemistry 11, 2176-2181.
MACARA,I. G. (1974) Ph.D. Thesis, University of Sheffield.
MACARA,L G., HOY, T. (~. and HARRISON,P. M. (1972) Biochem. J. 126, 151-162.
MACARA,]. G., HeY, T. G. and HARRISON,P. M. (1973a) Biochem. J. 135, 343-348.
MACARA,I. G., HeY, T. G. and HARRISON,P. M. (1973b) Biochem. Soc. Trans, 1, 102-104.
MACARA,I. G., HeY, T. G. and HARRISON,P. M. (1973c) Biochem. J. 135, 785-789.
MASSOVER,W. H. (1978) J. molec. Biol. 123, 721-726.
MASSOVER,W. H. and COWLey,J. M. (1973) Prec. hath. Acad. Sci. U.S.A. 70, 3847-3851.
MASSOVER,W. H. and COWLEY,J. M. (1975) In: Proteins of Iron Storage and Transport in Biochemistry and
Medicine (ed. R. R. C~CHTON) pp. 237-244. North Holland, Amsterdam.
MATIOLI,G. T. and BAKER,R. F. (1963) J. Ultrastruct. Res. 8, 477-490.
MAY, M. E. and FISH, W. W. (1977) Archs Biochem. Biophys. 82, 396-403.
McKAY, R. H. and FINEBERG,R. A. (1964) Archs Biochem. Biophys. 104, 487-508.
MELINO,G., STEFANINI,S., CH~ANCONE,E. and ANTONINI,E. (1978) FEBS Left. 86, 136-139.
MICHAELIS,L. (1947) Adv. Protein Chem. 3, 53-66.
MICH~LSON,A. M. (1977) In: Superoxide and Superoxide Dismutase (eds. A. M. MIChELSON,J. M. McCORD
and I. G. FRIDOVlCH)pp. 87-106. Academic Press, London.
MILLAR, J. A., CUMMING,R. L., SMITH,J. A. and GOLDBERG,A. (1970) Biochem. J. 119, 643-649.
MOROZ, C., LAHAR,N., BUNIANOV,M. and RAMOT,B. (1977) Cli,. exp. Immunol. 29, 30-35.
MUNRO, H. N. and DRVSDALE,J. W. (1970) Fed. Prec. 29, 1469-1473.
NIEDERER, W. (1970) Experientia 26, 218-220.
NIITSU,Y., ISHITANI,K. and LISTOWSKV,I. (1973) Biochem. biophys. Res. Commun. 55, 1134-1140.
NIELSEN, A. E. (1964) Kinetics of Precipitation, pp. 1-151. Pergamon Press, Oxford.
NUNEZ, M. T., GLASS,J. and ROBINSON,S. H. (1978) Biochim. biophys. Acta 509, 170-180.
OSAKI, S. and SmlVECH,S. (1971) Fed. Prec. 30, 1292 abs.
OVe. P., OBENRADER,M. and LANSING,A. (1972) Biochem. biophys. Acta 277, 201-210.
PAPE, L., MULTANI,J. S., STriCt, C. and SALTMAN,P., (1968) Biochemistry 7, 606.
P~.QUES, E. P., P.~QUES,A. and CRICHTON, R. R. (1979) J. molec. Cat. 5, 363-375.
PERKINS,S. J., JOHNSON,L. N., MACroN,P. A. and PmLUPS, D. C. (1979) Biochem. J. 181, 21-36.
POW~LL, L. W., MCKEERING,L. V. and HALUDAV,J. W. (1975) Gut 16, 909-912.
PURO, D. G. and RICHTER,G. W. (1971) Prec. Soc. exp. Biol. Med. 138, 399-403.
REISSMANN,K. J. and COLEMAN,T. J. (1955) Blood 10, 46-51.
RICHTER, J. W., and Walker, G. F. (1967) Biochemistry 6, 2871-2881.
RUSSELL, S. M. and HARRISON,P. M. (1978) Biochem. J. 175, 91-104.
SCOTT-MATHEWS,F., BETHGE,P. H. and CZERW~NSKI,E. W. (1979)J. biol. Chem. 254, 1696-1706.
SHINJO, S., ABE, H. and MASUDA,M. (1975) Biochem. biophys. Acta 411,165-167.
SHINJO, S. and HARRISON,P. M. (1979) FEBS Lett. 10~, 353-356.
SILK, S. T. and BRESLOW,E. (1976) J. biol. Chem. 251, 6963-6973.

JPB 36:2/~ C
86 G.A. CLEGG,J. E. FITTON,P. M. HARRISONand A. TREFFRY

SIRIVECH,S., FRIEDEN,E. and OSAKI,S. (1974) Biochem. J. 143, 311-315.


SIR1VECH,S., DRISKELL,J. and FRIEDEN,E. (1977) J. Nutr. 107, 739-745.
SMITH-JOHANNSEN,H. and DRYSDALE,J. W. (1969) Biochem. biophys. Acta 194, 43-49.
SPEYER, B. E. and FIELDING,J. (1979) Br. J. Haematol. 42, 255-267.
SPIRO, T. G. and SALTMAN,P. (1969) Struct. Bond. 6, 116-156.
SPIRO, T. G., ALLERTON,S. E., RENNER,J., TERZlS,A., BILS, R. and SALTMAN,P. (1966) J. Am. chem. Soc.
83, 2721-2726.
STEFANINI,S., CHIANCONE,E., VECCHINI,P. and ANTON1NI,E. (1975) In: Proteins o f Iron Storage and Trans-
port in Biochemistry and Medicine (ed. R. R. CRICHTON)pp. 295--302. North Holland, Amsterdam.
STEFANINI,S., CHIANCONE,E., ANTONINI,E. and FINAZZI-AGRo,A. (1976) FEBS Left. 69, 90-94.
STENKAMP,R. E., SIEHER, L. C., JENSEN, L. H. and LOEHR, J. S. (1976) J. molec. Biol. 100, 23-34.
STIEFEL,E. I. and WA'rT,G. D. 0979) Nature 279, 81-83.
STUHRMANN,H. B., HAAS,J., IBEL, K., KOCH, M. H. J. and CRICHTON,R. R. (1975) In: Proteins o f l r o n
Storage and Transport in Biochemistry and Medicine (ed. R. R. CRICHTON)pp. 261-266. North Holland,
Amsterdam.
STUHRMANN,H. B., HAAS,J., IaEL, K., KOCH, M. H. J. and CRICHTON, R. R. (1976) J. molec. Biol. 100,
399-413.
SURAN,A. A. and TARVER,H. (1965) Archs. Biochem. 111, 399-406.
TE¢CE, G. (1952) Nature 170, 75.
TEIPEL, J. and KOSHLAND,D. E., Jr. (1969) Biochemistry 8, 4656M663.
THEIL, l~. C. (1973)J. biol. Chem. 248, 622-628.
THEIL, [~. C., SAYERS,D. E. and BROWN,M. A. (1979) J. biol. Chem. 254, 8132-8134.
TOWE, K. M. and BRADLEY,W. F. (1967) J. Colloid Interface Sci. 24, 384-392.
TOWE, K. M., LOWENSTAM,H. A. and NESSON, M. H. (1963) Science, N.Y. 142, 63-64.
TREEFRY,A. and HARRISON,P. M. (1978) Biochem. J. 171, 313-320.
TREFFRY,A. and HARRISON,P. M. (1979) Biochem. J. 181,709-716.
TREFFRY,A., BANYARD,S. H., HOARE, R. J. and HARRISON,P. M. 0977) In: Proteins o f l r o n Metabolism
(eds. E. B. BROWN, P. AISEN,J. FIELDINGand R. R. CRIeHTON)pp. 3-11. Grune and Stratton, New
York.
TREFFRY,A., SOWERBY,J. M. and HARRISON,P. M. (1978) FEBS Left. 95, 221-224.
TREFFRY,A., SOWERBY,J. M. and HARRISON,P. M. (1979) FEBS Lett. 100, 33-35.
UVLIK, R. and ROMSLO,I. (1978) Biochim. biophys. Acta 541,251-262.
VAN BRUGGEN,E. F. J., WIEaENGA,E. H. and GRUBER,M. (1960) J. molec. Biol. 2, 81-82.
VAN DER GIESSEN,A. A. (1966) J. Inory. nucl. Chem. 28, 255-259.
VAN KREEL,B. K., PHNENBERG,A. M. C. M., VAN EIJK, H. G. & LEIJNSE,B. (1972) Biochim. biophys. Acta
273, 243-247.
VAUGHAN,P. A., STURDIVANT,J. H. and PAULING,L. (1950) J. Am. chem. Soc. 72, 5477.
VULIMIRI, L., LINDER, M. C., MUNRO, H. N. and CATSIMPOOLAS,N. (1977a) Biochim. biophys. Acta 491,
67-75.
VULIMIRI,L., LINDER,M. C. and MUNRO,H. N. (1977b) Biochim. biophys. Acta 497, 280-287.
WAGSTAEF,M., WORWOOD,M. and JACOBS,A. (1978) Biochem. J. 173, 969-977.
WARD, K. B., HENI~RIKSON,W. A. and KLIPPENSTEIN,G. L., (1975) Nature 257, 818-821.
WAUTERS,M., MIeHELSON,A. M. and CRICHTON,R. R. (1978) FEBS Lett. 91,276-279.
WEan, J. and GRAY, H. B. (1974) Biochim. biophys. Acta 251, 224-229.
WEBER, P. C., BARTSCH,R. G., CUSANOVICH,M. A., HAMLIN,R, C., HOWARD,A., JORDAN, S. R., KAMEN,
M. D., MEYER, T. E., WEATHERFORD,D. W., NGUYENHUU XUONGand SALEMME,F. R. (1980) Nature
286, 302-304.
WELLS, C. F. and SALAM,M. A. (1968) J. chem. Soc., 24-29.
WETZ, K. and CRICI4'TON,R. R. 0976) Eur. J. Biochem. 61,545-550.
WHITE, G. P., WORWOOD,M., PARRY,D. H. and JACOBS,A. (1974) Nature 250, 584-585.
WILLIAMS,J. M., DANSON,D. P. and JANOT,C. H. R. (1978) Phys. reed. Biol. 23, 835-851.
WILLSON, R. L. (1977) In: Ciba Foundation Symposium 51 (new series). Iron Metabolism, pp. 331-354.
Elsevier, Amsterdam.
W6HLER, F. (1963) Acta Haematol. 30, 65-87.
WOOD, G. C. and CRICHTON,R. R. (1971) Biochim biophys. Acta 229, 83-87.
WORWOOD,M., CRAGG,S. J., WAGSTAFF,M. and JACOBS,A. (1979a) Clin. Sci. 56, 1-5.
WORWOOD, M., CRAGG, S., JONES, B. M. and WAGSTAFF,M. (1979b). Fourth International Conference on
Proteins o f Iron Metabolisms. Davos, Switzerland.
WYLLIE,J. C. and KAUFMANN,N. (1971). Br. J. Haematol. 20, 321-327.
YARIV,J., SPERLING,R., BAUMINGER,E. R., COHEN,S. G. and OFER,S. (1979) in publication.
ZAMAN,Z. and VERWILGHEN,R. L. (1977) Biochem. Soc. Trans. 5, 306-308.
Z.~HRINGER,J. KONIJN, A. M. BALIGA,B. S. and MUNRO, H. N. (1975) Biochem. biophys. Res. Commun.
65, 583-590.
Z~,HRINGER,J. BALIGA,B. S. and MUNRO,H. N. (1976) Proc. HatH. Acad. Sci. U.S.A. 73, 857-861.

Anda mungkin juga menyukai