magnetoferritin: characterization of a novel superparamagnetic mr contrast agent

Original Research

Magnetoferritin: Characterization of a Novel Superparamagnetic MR Contrast Agent1

Jeff W. M. Bulte, PhD Trevor Douglas, PhD 6 Stephen Mann, PhD 6 Richard 8. Frankel, PhD Bruce M. Moskowitz, PhO Rodney A. Brooks, PhD Charles 0. Baumgarner, BS Josef Vymazal, MD, PhD Marie-Paule Strub, PhD 6 Joseph A. Frank, MD

A protein-encaged superparamagnetic iron oxide has been developed and characterized by using horse spleen apoferritin as a novel bioreactive en- vironment. The roughly spherical magnetoferritin molecules, 120 A in diameter. are composed of a monocrystalline maghemite or magnetite core 73 A f 14 in diameter. Except for the additional presence of iron-rich molecules of higher molecular weight, the appearance and molecular weight (450 kd) of magnetoferritin are identical to that of natural femtin; the molecules are externally indistin- guishable from their precursor, with a pI (isoelectric point) in the range 4.3-4.6. The measured magnetic moment of the superparamagnetic cores is 13.200 Bohr magnetons per molecule. with TI and T2 relaxivities (rl and r2) of 8 and 175 L-mmol-l (Fe) -sec-'. respectively, at body temperature and clinical field strengths. The unusudiy high r2/rl ratio of 22 is thought to arise from ideal core composition, with no evidence of crystalline paramagnetic inclusions. T2 relaxation enhancement can be well correlated to the field-dependent moleculm magnetization. as given by the Langevin magnetization function, raised to a power in the range 1.4-1.6. With its nanodimensional biomimetic protein cage as a rigid, convenient matrix for complexing a plethora of bioactive substances, magnetofemtin may provide a novel template for specific targeting of selected cellular sites.

Index terms: Contrasl media Iron Relaxometry

JMRI 1994: 4:497-505

Abbreviations: PAGE = polyacrylamide gel electrophoresis. SDS = sodium dodecyi sulfate, SQUID = superconducting quantum interference device.

' From the Laboratory 01 Diagnostic Radiology Research (OIR, OD) 1J.W.M.B.. J.A.F.1, Neuroimaging Branch ININDSl 1R.A.B.. C.D.B.1. Bio- medical Engineering and Instrumentation Program (NCRR) (J.V.). and the Prott4n Expression Laboratory (OR, OD) (M-PSI. Bldg 10. Rm B1N256. National Institutes of Health, Bethesda. MD 20892; the School of Chem- istry. University of Bath. Bath, England (T.D.. S.M.1: the Physics Depart- ment, California Polytechnic State University. San Luis Obispo. Calif IR.B.F.1: and the Department of Geology and Geophysics. University of Minnesota. Minneapohs. Mmn 1B.M.M.). Received January 4. 1994; ac- cepted February 23. Address reprint requests to J.W.M.B.

&'SMR, 1994

SUPERPARAMAGNETIC IRON oxide crystals reduce T2 in water, owing to dephasing of protons as water molecules diffuse through magnetic inhomogeneities created by the iron oxide cores. For use as a magnetic resonance (MR) contrast agent, the crystals are currently prepared by nonbiologic processing (ie, chemical coprecipitatioii of ferrous and femc oxides by using mixed-valence iron oxyhydroxides as an intermediate matrix for synthesis of the superparamagnetic crystals) (1,2). Bacterial dextran is most commonly used to coat and stabilize the iron oxide cores: Ex- amples are AMI-25 (3), AMI-227, USPIO (ultrasmall superparamagnetic iron oxide) (4), and MIONs (monocrystalline iron oxide nanocornpounds) (5); these are chemosynthetic iron oxide colloids that, in general, have broad size distributions because they lack a spatially confined and controlled synthesis. The major application of this generation of nanophase MR contrast agents is in contrast-enhanced imaging of the liver, spleen, and lymph nodes: The crystals are rapidly taken up by the reticuloendothelial system and other cells expressing affinity for poly-D-glucose (dextran).

A novel approach in crystal engineering has re- cently evolved from several interdisciplinary fields. By using biologic principles, novel materials with defined crystal size can be produced by confined biomineralization within specific subunit compartments (6-8). Ferritin, a ubiquitous and structurally highly con- served iron storage protein composed of 24 subunits with an apparent molecular weight of approximately 450 kd (91, is an interesting example of a reaction cage in which nanophase crystals can be processed. By using demetalized ferritin (apoferritin) shells as catalysts and nucleation sites, spatially confined, specific nanophase materials such as iron sulfides (10) and manganese (lo), uranium (10.1 1) and iron (12) oxides can be produced. In particular, the latter compound, magnetoferritin, is of potential interest for use as a nanodimensional MR contrast agent. We re- port here on the MR relaxometric and physicochemi- cal characterization of magnetoferritin and compare these properties of this biosynthetic compound with those of AMI-25, a prototype chemosynthetic iron oxide. It is shown that magnetofemtin has an r2/rl ratio one order of magnitude higher than similarly sized chemosynthetic iron oxides. With its nanodimen-

497

40 50 60 70 80 90 100 110

Core diameter (i)

Figure 1. stained magnetoferritin (a) and core-size histogram (b). The cores average 73 A -t 14 in diameter. (c) Transmission electron micrograph of negatively stained magnetofemtin shows the outer protein shells. The total particle diameter is 120 A and corresponds to that of the (apo)femtin protein cage. Bars in a and c represent 1,000 A.

(a, b) Transmission electron micrograph of un-

sionatl biomimetic protein cage as a rigid, convenient matrix for complexing a wide variety of bioactive substances, magnetofemtin may provide a basis for a novel generation of biocompatible magnetopharmaceu- ticals.

0 MATERIALS AND METHODS

Magnetofemtin Preparation Magnetoferritin was prepared as previously de-

scribed (12), with some modifications. Briefly, horse spleen femtin (197742; Boehringer Mannheim, India- napolis, Ind), cadmium free and crystallized three times. was first depleted of its native femhydrite core by dialysis under nitrogen against 0.1 mol/L thiogly- colic acid (13). Under strictly anaerobic conditions, ferrous ammonium sulfate was added in repeated incre- ments of 50 Fe (11) per protein molecule, with intermit- tent addition of oxidant to allow core formation.

For comparison, chemosynthetic dextran-coated iron oxides in the form of AMI-25 (provided by Ad- vanced Magnetics, Cambridge, Mass) were also stud- ied. These superparamagnetic particles have a particle slze of about 500 A, with an apparent molecular weight of 500-20,000 kd, and are composed o ta magnetite and/or maghemite core averaging 110 A in diameter (3,14:l.

Chemical Analysis The total iron concentration of the magnetoferritin

preparations was determined spectrophotometrically after (oxidation of Fez+ to Fe3+ with hydrochloric acid and hydrogen peroxide, followed by the addition of 1% potassium thiocyanate and measurement of the 480- nrn absorption line of the pink-orange iron-thiocyanate complex. Independently, the iron measurements were validated with atomic absorption spectrophotometry (model AA-275; Varian, Victoria, Australia), by using flame.

The total protein concentration of magnetofemtin was measured spectrophotometrically with Peterson’s

modification (1 5) of the protein assay of Lowry et al (16) by using a kit (P-5656; Sigma Chemical, St Louis, Mo). Bovine serum albumin (supplied by the manufac- turer) and native horse spleen femtin were used as standards; the absorbance was measured at 800 nm to minimize the intrinsic absorbance of the iron oxide core. The “loading factor”-the average number of iron atoms per ferritin molecule-was determined by divid- ing the total iron by the total protein concentration.

Morphologic and Mineralogic Analysis

ritin was assessed with transmission electron microscopy. Magnetoferritin suspensions were deposited on carbon-coated, Formvar-covered copper grids. To as- sess the size of the cores and total molecules, respectively, both unstained and negatively stained (1% ura- nyl acetate) grids were examined in a JEOL (Tokyo, Japan) 1200EX transmission electron microscope op- erating at 120 keV. A magnetofemtin core size histogram was obtained by using measured diameters of 150 individual cores. For mineralogic characterization of the cores, electron diffraction patterns were recorded with a camera length of 0.8 m.

Gel Electrophoresis and Isoelectric Focusing Magnetoferritin and horse spleen ferritin (ie, the ma-

terial from which magnetoferritin was prepared) sarr- ples were separated, under denaturing conditions, with

The morphology and size distribution of magnetofer-

498 a JMRl May I June 1994

32 incremental time intervals and T2 was measured by using a Carr-Purcell-Meiboom-Gill sequence with 100 decreasing spin-echo intensities and an interecho time of 2 msec.

0 RESULTS

Figure 2. Electron diffraction pattern of magnetoferritin. The d spacings for the lines are 3.05, 2.53, 2.10, 1.72, 1.62, and 1.48 A. The first d space (3.05 A] is characteristic of a magnetite or maghemite lattice.

sodium dodecyl sulfate (SDS) polyacrylamide gel (4% 20%) electrophoresis [PAGE), according to Laemmli (17). The gel electrophoresis system (Novex: Novel Ex- perimental Technology, San Diego, Calif) was run at 120 V for 2 hours in a buffer consisting of 25 mmol/L tris and 190 mmol/L glycine and containing 0.1% SDS. The molecular weights of native magnetoferritin and horse spleen ferritin were assessed with a 4% PAGE run at 50 V for 6 hours in a nondenaturing system, with the same running buffer but without SDS. The isoelectric point of magnetoferritin and horse spleen femtin molecules was determined by isoelectrofocusing in an ampholyte system (Ampholine PAG- plate; Pharmacia, Piscataway, NJ) with a pH range of 4.0-6.5. Gels were run in duplicate for both PAGE and isoelectrofocusing, and stained either for protein with Coomassie Blue or for ferric iron with Prussian blue staining.

Magne tometry The magnetic properties of samples of water solu-

tions of magnetoferritin and AMI-25 in plastic holders were measured at 300" K with a SQUID (superconducting quantum interference device) susceptometer (Quantum Design, San Diego, Calif). The total magnetic dipole moment M of each sample was measured at discrete magnetic fields between -1 and +1 T. Mo- ment calibration was verified with palladium and nickel standards.

Relaxometry Serial dilutions of magnetoferritin and AMI-25 in

phosphate-buffered saline (10 mmol/L phosphate, pH = 7.4) containing 0.02 %o sodium azide were prepared for MR relaxometry. For the entire range of sample nieasurements, 1/T1 and 1/T2 were verified to have a linear dependence on iron concentration. With a vari- able-field relaxometer (Southwest Research Institutes, San Antonio, Tex), spanning 1-64 MHz, T1 was measured with a saturation-recovery pulse sequence with

Composition, Sue, and Structure Total protein and iron analysis indicated that the

magnetoferritin molecules contained an average of 2,140 iron atoms per protein molecule. Transmission electron microscopy of unstained magnetofemtin revealed that the preparation consisted of monocrystalline cores averaging 73 A f 14 in diameter (Fig la, lb). In negatively stained samples (Fig I c ) ~ the outer protein cage could be identified as a 20-A thick shell. Electron diffraction analysis of the iron oxide cores showed the characteristic d line spacings of a magnetite (Fe,O,) or maghemite (yFe,O,) lattice (Fig 2).

Electrophoretic Properties To compare the molecular weight of magnetofemtin

molecules with that of horse spleen femtin, the proteins were analyzed by means of 4% PAGE under nondenaturing conditions (Fig 3a). Staining for protein (lanes 1 and 2) showed that the magnetoferritin preparation consisted essentially of monomeric protein with a molecular weight of about 450 kd, similar to that of horse spleen ferritin. A polymeric protein complex of high molecular weight, not present in horse spleen ferritin, could be detected in the magnetoferritin preparation. A trace amount of oligomeric protein (ie, dimers and trimers) could be detected in both fractions. Prus- sian blue staining (lanes 3 and 4) showed that the high-molecular-weight fraction contained a relatively large amount of iron.

PAGE under denaturing conditions (Fig 3b) revealed the same two major protein bands (22-24 kd) in both preparations (lanes 1 and 2): No difference between the horse spleen ferritin and magnetoferritin molecules was observed. These subunit sizes are slightly larger than reported values of 19-21 kd (18); however, de- pending on the method of isolation and electrophoretic analysis, the relative femtin subunit size may be any- where between 18 and 24 kd (19). The iron cores, no longer encaged by the protein, showed a similar size distribution for horse spleen femtin and magnetoferr- tin [lanes 3 and 4). Isoelectrofocusing of the proteins (Fig 4) showed a similar microheterogeneous pattern for horse spleen femtin and magnetofemtin: the PI (ie, the pH at which the protein has a neutral electric charge [isoelectric point]) is in the range 4 .34 .6 for both molecules, well in agreement with reported values for horse spleen ferritin (20).

rich, high-molecular-weight fraction in the magnetoferritin preparation, the compound is indistinguish- able from its precursor horse spleen ferritin molecules, both in apparent molecular weight and net electrical charge.

Magnetic and Relaxometnc Properties SQUID magnetometry revealed no hysteresis in the

magnetization of the magnetofenitin and, for comparison, of the AMI-25 molecules [Fig 5a, 5b). The magnetization data were fitted assuming superparamagnetic behavior for the magnetofemtin and AMI-25 molecules, by using the following function:

Analysis of subunit composition with 4°!~20% SDS-

Thus, except for the additional presence of an iron-

Volume 4 Number 3 JMRl 499

M(BJ = M-L(pB,/kT) + aB0,

where L(pB,/kT) is the Langevin function,

L(pB,/kT) = COth(pB,/kT) - (kT/pBJ. (2)

and M, is the saturation magnetization at high field strength, p the magnetic dipole moment per magnetoferritin or AM-25 molecule, B, the applied magnetic field, k the Boltzmann constant, T absolute temperature, and a the diamagnetic and paramagnetic susceptibility of the magnetofenitin and AMI-25 samples, respec tively. The data fitting was done in two ways: (a) fitting the experimental data with Equation (l), with M-, 11, and a as free parameters, and (b) determining a from the high-field-strength (0.7-1 T) data and fitting M - RB, with Equation (2), with B, and p as free parameters (Fig 5). Both procedures yielded similar values for p in both samples (Table 1). Measurements at other temperatures showed that p was essentially constant over the range 280"-320" K. Fits were also tried with a lognormal or uniform distribution of molecular magnetic dipole moments, without evident improvement over the single-moment fits, suggesting a narrow moment distribution in both cases.

The r 1 values of magnetofemtin and AM-25 versus Larnior frequency are shown in Figures 6a and 7a, respectively. For magnetofemtin, r 1 increases rapidly at low field strength, with a maximum between 5 and 10 MHz; this maximum is greater at lower temperatures. At high field strength, r l decreases and becomes temperature independent. For AM-25, the magnitude and dependence of r l on frequency and temperature were identical to that seen with magnetofemtin.

The r2 versus frequency plots are shown in Figures f?b and 7b for magnetofemtin and AM-25, respec- iivelly. For both compounds, r2 increases rapidly at low field strength, approaching saturation near 20 MHz. 'The r2 values of both compounds are greater at lower temperatures over the entire field strength range. Pre- vious work on chemosynthetic magnetite and maghemite particles (14) showed that r2 can be described by a power of the Langevin function:

r2(B0) = r2,LP(pB0/kT), (3)

where r2, is the limiting T2 relaxivity at high field strength and p is the exponential dependence of r2 on the molecular magnetization. In that work, it was shown that p = 1, suggested by the linear dependence of T2 relaxation on frequency for several iron oxyhydroxides including normal femtin (14,21), and p = 2, suggested by some theoretical models of T2 relaxation, both fit the r2 data equally well. However, the corre- sponding p values varied by about a factor of 2. Since these p values are now known (Table 1). we attempted to fit our 1-2 data with r2, and p as free parameters, using the p value determined by magnetometry. For comparison, we also allowed r2, and p to be free parameters, with p values of 1 and 2; finally, we did the same with an added diamagnetic/paramagnetic term yB,,

r2(B0) = r2,Lp(pB0/ kT) + yB,, (4)

to compensate for possible diamagnetic/ paramagnetic contributions that are linear with the magnetic field. The results are summarized in Table 2. We found, for both magnetofemtin and AMI-25, that neither a linear nor EL quadratic dependence of r2 on magnetization (p =

1 or 2) permits a p value that corresponds to the value determined with magnetometry, whether the diamagnetic/paramagnetic term is included or not. By fixing p and varying p, values of p in the range 1.35-1.60 for magnetoferritin and 1.42-1.58 for AMI-25 were found. By including the diamagnetic/paramagnetic term, we found p values in the range 1.40-1.50 for magnetofem-

Figure 3. PAGE of horse spleen ferritin and magnetofemtin, stained for protein with Coo- massie Blue (lanes 1, 2. and M) and for iron with Prussian blue (lanes 3 and 4). (a) Electropho- retic pattern of native horse spleen femtin (lanes 1 and 3) and magnetoferritin (lanes 2 and 4). Thirty (lanes 1 and 2) and 15 (lanes 3 and 4) Fg of each protein were loaded. @) Subunit composition of horse spleen ferritin (lanes I and 3) and magnetofemtin (lanes 2 and 4); lane M contains marker. In each lane, 5 ~g of protein was loaded.

500 JMRl May I June 1994

tin and 1.04-1.18 for AMI-25. For M I - 2 5 , we im- proved the quality of fit (R) when the diamagnetic/ paramagnetic term was included (at fixed p values, R increased from 0.95 to 0.98 at all three temperatures). No such improvement was obtained for magnetoferri- 0 DISCUSSION tin. Finally, we note that the diamagnetic/paramag-

netic factor y increased with temperature for niagneto- femtin regardless of p, while for MI-25 , y decreased with temperature.

Composition of Magnetofemtin In the present study, horse spleen apofemtin was

used to biologically synthesize a nanodimensional and monocrystalline superparamagnetic iron oxide, en- closed by its own processing protein. The PI of the horse spleen ferritin from which the magnetoferritin was prepared ranged from 4.3 to 4.6, well in agreement with literature values (20). The magnetoferritin preparation was found to have an identical PI; the net outer charge of the acidic, anionic protein is not affected by the superparamagnetic core, providing further evidence that the crystals are indeed fully encaged and "hidden" by the surrounding protein matrix. We observed a discrete total molecular diameter of 120 A for magnetoferritin, the same as for its natural precursor. Theoretically, the spatially confined biosynthesis of magnetoferritin results in the formation of cores not exceeding 80-90 A in diameter, as dictated by the in- ternal diameter of the hollow sphere formed by the protein subunits (22). The measured average core diameter of 73 A 5 14 is near this theoretical maximal value.

Electron diffraction analysis indicated that the cores are composed of a magnetite and/or maghemite lattice, but this technique did not allow us to discriminate between the two structures. However, recent Mossbauer spectroscopic studies have indicated that the predomi- nant mineral form in the cores is maghemite rather

Figure 4. Isoelectric focusing of horse spleen ferritin [lanes 1 and 3) and magnetoferritin (lanes 2 and 4). stained for protein with Coomassie Blue (lanes 1 , 2, and M) and for iron with Prussian blue (lanes 3 and 4). Lane M contains marker. Three (lanes I and 2) and 15 (lanes 3 and 4) pg of each protein were loaded.

I I

- 6 E'

c1 s o E -2 u .* c1 2 -4

2 E -6

I I I I -0.5 0.0 0.5 1 .o

Magnetic field (Tesla) a.

h a

v

Table 1 Measured Magnetic Dipole Moments per Molecule

Bohr Magnetons Sample p(x 10-17 mJT-') per Molecule

Magnetoferritin 12.2 5 0.14 13,161 AM-25 11.2 2 0.17 12,082

0.12 I I I 1

0.08

0.04

0.00

-0.04

-0.08

-0.12 ' 1 I I I -1.0 -0.5 0.0 0.5 1 .o

Magnetic field (Tesla) b.

Figure 5. Superparamagnetic magnetization of magnetoferritin (a) and MI-25 (a) molecules at 300" K. obtained by correcting the total sample magnetization for the diamagnetic and paramagnetic background susceptibility. The solid lines represent a best fit of the data with the Langevin magnetization function, MB,) = M,IP(pB,/kT).


, 0 15 30 45 60

0 ~

Frequency (MHz)

a.

400 T

_____ 0 15 30 45 60

Frequency (MHz)

b. Figure 6. T1 (3 and T2 [a) relaxivity of magnetoferritin versus Larmor frequency at 3°C (A), 25°C (W), and 37°C (0). The solid Lines in b represent a best fit of the r2 data with the Langevin T2 relaxation function, r2_Lp(pB0/ kT) (3). and a value of 13,200 Bohr magnetons, as obtained with SQUID magnetometry.

1ci

0

.- 5 200

a

> X m a,

.- -

p 100

0 &- --@I c ,

0 15 30 45 60 0 15 30 45 60

Frequency (MHz) Frequency (MHz)

a. b. Figure 7. T1 (a) and T2 @) relaxivity of AM-25 versus Larmor frequency at 3°C (A), 25°C (H), and 37°C (0). The solid lines in b represent a best fit of the r2 data with the Langevin T2 relaxation function, r2,Lp(pB0/kT) (3), and a p value of 12.100 Bohr magnetons, as obtained with SQUID magnetometry.

than magnetite (23). While it is possible that, under strict anaerobic conditions, a magnetite core is initially formed, it appears that on exposure to the atmosphere a significant fraction of the ferrous ions in magnetite are rapidly oxidized and converted to a femc state compatible with the mineral form of yFe,O,.

Since the femmagnetic inverse spinel st-mcture of magnetite is composed of cubic cells 8.39 A in length, containing 24 iron atoms each (24). a diameter of 73 A for the magnetite or oxidized maghemite core in magnetofemtin corresponds to a theoretical number of 8,277 iron atoms per crystal. This value is about four times larger than the experimentally determined average value of 2,140 iron atoms. Also, the magnetic moment of 13,200 Bohr magnetons suggests that the ac- tual “loading factor” of the magnetofemtin cores is about five to six times greater than the average bulk value of 2,140, which can be calculated as follows. The saturation magnetization values for maghemite and magnetite at room temperature are 76 and 93 mJl-’/g,

respectively (25). or 2.18 Bohr magnetons ([76 mJT-l/g .160 g/moll/[6.02 x loz3. 9.27 x lO-”’]) per yFe,O, molecule and 3.87 Bohr magnetons ((93 mJT-’/g . 232 g/mo1]/[6.02 x loz3. 9.27 x per Fe,O, molecule. Hence, a magnetic moment of 13,200 Bohr magnetons corresponds to 12,110 iron atoms for a maghemite lattice and 10,233 iron atoms for a magnetite lattice. This discrepancy between the theoretical and experimentally determined “average” number of iron atoms per protein molecule suggests that only about 20% of the total demetalized femtin population was effectively in- corporating iron and being converted to magnetofemtin. This is not that surprising, given that horse spleen femtin is composed of 15% H-chain and 85% L-chain subunits. In L-chains, the channel of the ferroxidase site is thought to be blocked (26). and these subunits show-compared with H-chains with their specific ferroxidase activity--markedly reduced iron uptake kinet- ics under an “all or nothing” regime, with comprised cores of increased diameter and regularity (27). This is

502 JMRl May / June 1994

Table 2 T2 Relaxivity of Magnetofemtin versus That of AMI-25 at Three Temperatures

Magnetoferritin AMI-25 Temperature

("K) P P y r2, R P P y r2, R

Equation (3) 276 1.35 13.200 Q 332 0.983 1.58 12.100 Q 212 0.952 298 1.52 13.200 Q 229 0.982 1.57 12.100 0 148 0.948 3 10 1.60 13,200 Q 173 0.962 1.42 12.100 Q 112 0.956 276 - 1 9,900 Q 332 0.980 I 7,100 - 0 217 0.970

0 153 0.970 298 - 1 8,000 Q 230 0.979 1 7,000 310 - 1 7,900 Q 175 0.961 1 7,900 - 0 115 0.968

14,700 Q 213 0.947 276 - 2 19,000 Q 331 0.983 2 298 - 2 15,800 Q 228 0.981 2 14,700 Q 149 0.941 310 - 2 16,200 Q 173 0.954 2 16,400 0 112 0.948

Equation (4) 276 1.44 13.200 -0.39 347 0.987 1.18 12.100 6.85 178 0.988 298 1.50 13,200 0.05 227 0.982 1.17 12.100 0.60 124 0.981 3 10 1.40 13.200 0.38 157 0.965 1.04 12.100 0.48 93 0.997 276 - 1 8,800 -0.55 355 0.986 1 9,600 0.76 182 0.990 298 - 1 8,400 -0.07 233 0.979 1 9,200 0.48 130 0.984 310 - 1 7,900 0.40 175 0.961 1 11.300 0.47 94 0.997 276 - 2 18,200 -0.35 344 0.986 2 20,100 0.88 176 0.882 298 - 2 17,500 0.08 225 0.982 2 20,100 0.61 123 0.973 310 - 2 18,900 0.41 155 0.965 2 23,100 0.51 91 0.993

Note.-The fitting parameters for Equations (3) and (4) (see Results) used to fit the r2 data are the following: exponential dependence of r2 on magnetization (P), magnetic moment per domain [p, in Bohr magnetons), diamagnetic/paramagnetic factor (y), 1-2 at limiting field strength (r2-, in L.rnmol-' LFe1.sec-l). and coefficient of quality of fit (R). Underlined values represent fmed parameters.

also consistent with the PAGE results (Fig 3), which showed the presence of a relatively iron-poor monomeric fraction in magnetofemtin, tentatively represent- ing a significant amount of residual apofemtin (which remained ineffective in iron incorporation during synthesis of magnetofemtin). Also, a higher-molecular- weight fraction was observed, which appeared to con- tain a relatively larger iron-to-protein ratio. In normal horse spleen femtin, oligomeric and polymeric fractions generally have a larger iron content (28) than do monomers. In these iron-rich polymers, the proteins are cross-linked by disulfide bridges, a process that is thought to be mediated by iron, since apofenitin is found only in the monomeric form. A similar process could have taken place in our magnetoferritin preparation, but it is more likely that the heating step in our synthesis induced some minor protein degradation with subsequent aggregation, which appeared as the iron-rich higher-molecular-weight fraction.

Relaxometry and Comparison with Conventional Chemosy nthetic Agents

It is of interest to compare the relaxometry of magnetofemtin with that of chemosynthetic iron oxides currently used as MR contrast agents. Although AMI- 25 has been investigated intensively by others and was included in our measurements, the size and composition of magnetofemtin more closely resembles those of AM-227, which also shows a more narrow size distribution. These properties are summarized in Table 3.

For magnetofemtin, there is no direct contact between bulk water protons and the cores, since the protein shell is essentially impermeable to water molecules (the T1 relaxation effects of natural femtin are believed to originate only from a small number of noncore iron atoms bound to the protein shell [29]). The peak r l for magnetoferritin at room temperature (30 L.mmol-'.sec-') (Fig 6a) is about three orders of

magnitude higher than that of natural femtin, which is about 0.03 L.mmo1-'.sec-l (14). On the other hand, the magnetic moment of 13,200 Bohr magnetons per magnetofemtin molecule is about 80 times larger than the reported value for natural femtin (1 62 Bohr magnetons) (30). Hence, T1 relaxation enhancement by magnetofemtin seems to arise primarily from the strong magnetic field inhomogeneities created by the cores, which affect the diffusiqg bulk water protons over a distance of at least 20 A (that is, beyond the physical barrier of the protein shell). The field dependence of the magnetization and the magnitude of the magnetic moment thus appear to be major determi- nants of T1 relaxation enhancement by superparamagnetic iron oxide cores; this has also been discussed in an adapted Freed-Ayant model of longitudinal relaxation (3 1). We observed a similar pattern of longitudinal relaxometry for magnetofemtin and AMI-25 (Figs 6a, 7a), both in frequency and magnitude of peak relaxation enhancement, a finding consistent with the near identical magnetic moment of these two molecules (Table l).

T2 relaxometry shows that transverse relaxation enhancement by the superparamagnetic cores of magnetofemtin and AMI-25 can be well described by their field-dependent molecular magnetization, as given by the Langevin function. The exponent of the Langevin T2 relaxation function needed to quantitatively fit the r2 data-being in the range 1.4-1.6 for magnetofemtin and AMI-25-cannot be explained by any current the- ory of T2 relaxation. Interestingly, the exponent for AMI-25 decreases to 1.0-1.2 if diamagnetic/paramagnetic effects are included. For magnetofemtin, the p value does not appear to change when the yB, term is added, suggesting that diamagnetic/paramagnetic effects are not important and may not even be present. The diamagnetic/paramagnetic factor y increases with temperature; the origin of this dependence is not


Table 3 Comparison of Magnetoferritin and n o Conventional Chemosynthetic Iron Oxides I

Property Magnetofemtin AM-25*

Core sue 70 (60-90) A 110 (80-140) A Core crystallinity Monocrystalline Oligocrystalline Surrounding matrix Protein Dextran Molecular size 120 A 500 (200-800) A Apparent molecular weight (kd) 450 500-20,000 r1t 8 10 r2t 175 110 r2 / r l t 22 11

Note.-Units for r l and r2 values are L.mmol-'.sec-'. * Data from reference 14.

Measured at 1.5 T and 37°C.

AM-227*

70 (50-90) A Monocrystalline Dextran 110 (80-150) A 700 20 60 3

known but could be related to the presence of diamagnetic protein, which has a negative magnetic susceptibility. For AM-25, however, the yvalues decrease with temperature, which is consistent with paramagnetic behavior (Curie's law), and possibly suggests the presence of a small amount of paramagnetic impurity in this material.

Unlike magnetofemtin, in which core synthesis is precisely regulated by the protein, without intermediate formation of non-superparamagnetic structures, the cores of AM-25 are prepared in bulk precipitation with iron oxyhydroxides (eg, goethite) as intermediate crystal lattices. Hence, it is not completely unlikely that small paracrystalline inclusions of iron oxyhydroxides are present in AMI-25, which could be re- sponsible for the suggested paramagnetic impurity. 'fie fact that, for a constant amount of iron, fewer, larger cores (up to about 50 nm [32]) are believed to have a higher T2 relaxivity than a larger number of smaller cores (32-34) could also explain why, despite Its larger core sue, AMI-25 has a lower T2 relaxivity than magnetofemtin. As a corollary, the major difference between relaxation enhancement by magnetofer- iitin and the chemosynthetic iron oxides is the high r2/rl ratio of magnetofemtin, which is one order of magnitude higher than that of the similar-size crystals of AM-227 (Table 3). As an MR contrast agent and at low tissue concentrations, magnetofemtin would therefore offer a wider margin at which negative signal enhancement could be obtained with standard spin- echo sequences.

I n summary, we have developed a novel, biomimetic, superparamagnetic iron oxide with ideal core characteristics for use as an MR contrast agent. More- over. the rigid, spherical protein cage provides a convenient template to which a wide variety of bioactive substances can be easily conjugated. Putative complexing molecules include antibody fragments and synthetic peptides, which may further direct the use of magneto- fenit in for tissue-specific imaging. Here, we have used an equine protein as the mineralizing precursor; however, an abundant choice of precursor molecules is available because femtin is ubiquitously distributed t hroiighout most prokaryotic and eukaryotic organ- isms. Through the use of established recombinant tleoxyribonucleic acid techniques, the functional expression of human ferritin homopolymers (35,36) or hetei-opolymers (37) may offer a biotechnologic, different approach to further magnetopharmaceutical devel- opment. 0

Acknowledgments: We gratefully acknowledge Vu Tran and Paul T. Wingfield, PhD. for their help and stimulating discussions.

References 1. Hasegawa M, Hokukoku S. U.S. patent 4,101,435, 1978. 2. Molday RS, MacKenzie D. Immunospecific ferromagnetic iron-

dextran reagents for the labeling and magnetic separation of cells. J Immunol Methods 1982; 52:353-367.

3. Josephson L, Lewis J, Jacobs P, Hahn PF, Stark DD. The effects of iron oxides on proton relaxivity. Magn Reson Imaging

4. Weissleder R, Elizondo G, Wittenberg J, Rabito CA. Bengele HH, Josephson L. Ultrasmall superparamagnetic iron oxide: characterization of a new class of contrast agents for MR imaging. Radiology 1990; 175:489-493.

5. Shen T, Weissleder R, Papisov M, Bogdanov A, Brady TJ. Monocrystalline iron oxide nanocompounds (MION): physico- chemical properties. Magn Reson Med 1993: 29:599-604.

6. Heuer AH, Fink DJ, Laraia VJ, et al. Innovative materials processing strategies: a biomimetic approach. Science 1992;

7. Mann S , Biomineralization: the hard part of bioinorganic chemistry. J Chem Soc Dalton Trans 1993; 1: 1-9.

8. Mann S . Achibald DD, Didymus J M , et al. Crystallization at inorganic-organic interfaces: biominerals and biomimetic synthesis. Science 1993; 261:1286-1292.

9. Harrison PM. The structure of apofemtin: molecular size, shape and symmetry from x-ray data. J Mol Biol 1963; 6:404-422.

10. Meldrum FC, Wade VJ, Nimmo DL, Heywood BR, Mann S. Synthesis of inorganic nanophase materials in supramolecu- lar protein cages. Nature 1991; 349:684-687.

11. Hainfeld JF. Uranium-loaded apofemtin with antibodies at- tached: molecular design for uranium-capture therapy. Proc Natl Acad Sci U S A 1992; 89: 11064-1 1068.

12. Meldrum FC, Heywood BR, Mann S. Magnetofemtin: in vitro synthesis of a novel magnetic protein. Science 1992;

13. Treffry A, Harrison PM. Incorporation and release of inorganic phosphate in horse spleen ferritin. Biochem J 1978;

14. Bulte JWM, Vymazal J, Brooks RA, Pierpaoli C, Frank JA. Frequency dependence of MR relaxation times. 11. Iron oxides.

15. Peterson GL. A simulification of the urotein assay method of

1988; 6~647-653.

255:1098-1105.

257:522-523.

17 1131 3-320.

J M R I 1993; 3:641-648.

Lowry et a1 which is more generally gpplicable. &a1 Biochem 1977; 83:346-356.

16. Lowry OH, Rosebrough N J , Farr AL, Randall R J . Protein measurement with the Folin phenol reagent. J Biol Chem 1951;

17. Laemmli U. Cleavage of structural proteins during the assem- bly of the head of bacteriophage T4. Nature 1970; 227:650- 685.

18. Arosio P, Adelman TG. Dwsdale JW. On ferritin heterogene-

193~265-275.

ity: further evidence for heteropolymers. J Biol Chem 1578; 253:445 1-4458.

19. Drysdale J W , Adelman TG. Arosio P, et al. Human isofemtins in normal and disease states. Semin Hematol 1977; 14:71- 87.

504 JMRI May/ June 1994

20. Malamud D. Drysdale JW. Isoelectric points of proteins: a table. Anal Biochem 1978; 86:62@647.

21. Vymazal J. Brooks RA, Zak 0, McRill C, Shen C, DiChiro G. 'rl and T2 of ferritin at different field strengths: effect on MRI. Magn Reson Med 1992; 27:368-374.

22. Ford GC, Hamson PM, Rice DW. et al. Ferritin: design and formation of an iron-storage molecule. Philos Trans R SOC Land 1984; 304:551-565.

23. Pankhurst QA, Betteridge S , Dickson DPE. Douglas T. Mann S, Frankel RB. Mossbauer spectroscopic and magnetic studies of magnetofemtin. Hyperfine Interactions (in press).

24. Verwey EXW, Heilmann EL. Physical properties and cation arrangement of oxides with spinel structures. I. Cation arrangement in spinels. J Chem Phys 1947; 15:174-180.

25. Chevallier R. Magnetic properties of minerals. In: Thewlis J, ed. Encyclopaedic dictionary of physics. Vol4. Oxford, En- gland: Pergamon, 1961; 430-434.

26. Lawson DM, Artymiuk PJ, Yewdall ST, et al. Solving the structure of human H femtin by genetically engineering in- termolecular crystal contacts. Nature 1991; 349:541-544.

27. Wade VJ, Levi S. Arosio P. Treffry A, Hamson PM, M a n n S. Influence of site-directed modifications on the formation of iron cores in ferritin. J Mol Biol 1991; 221:1443-1452.

28. Niitsu Y, Listowsky I. Mechanisms for the formation offer- ritin oligomers. Biochemistry 1973; 12:4690-4695.

29. Koenig SH, Brown RD Ill, Gibson JF, Ward RJ , Peters TJ. Relaxometry of femtin solutions and the influence of the Fe" core ions. Magn Reson Med 1986; 3:755-767.

30.

31.

32.

33.

34.

35.

36.

37.

Blaise A, Feron J, Girardet J-C. Lawrence J-J. Contribution a l'etude des proprietes magnetiques de la ferritine. C R Acad Sci B 1967; 265:1077-1080. Roch A. Muller RN. Longitudinal relaxation of water protons in colloidal suspensions of superparamagnetic crystals (abstr). In: Book of abstracts: Society of Magnetic Reso- nance in Medicine 1992. Berkeley, Calif: Society of Mag- netic Resonance in Medicine, 1992; 1447. Muller RN, Gillis P, Moiny F, Roch A. Transverse relaxivity of particulate MRI contrast media: from theories to experi- ments. Magn Reson Med 1991; 22:178-182. Gillis P, Koenig SH. Transverse relaxation of solvent protons induced by magnetized spheres: application to femtin, erythrocytes, and magnetite. Magn Reson Med 1987: 5:323- 345. Majumdar S, Gore JC. Studies of diffusion in random fields produced by variations in susceptibility. J Magn Reson 1988: 78:41-55. Levi S, Cesareni G, Arosio P, et al. Characterization of human ferritin H-chain synthetized in Escherichia colt. Gene

Levi S, Salfeld J, Franceschinelli F, Cozzi A, Dorner M, Arosio P. Expression and structural and functional properties of human L-chain from Escherichia coli. Biochemistry

Santambrogio P, Levi S , Cozzi A, Rovida E, Albertini A, Arosio P. Production and characterization of recombinant heteropolymers of human femtin H and L chains. J Biol Chem 1993; 268: 12744-12748.

1987; 51:269-274.

1989: 28:5179-5184.


magnetoferritin: characterization of a novel superparamagnetic mr contrast agent

Documents