Detection of Magnetosome-Like Structures in Eukaryotic Cells Using Nonlinear Longitudinal Response to ac Field

Although magnetosomes have been discovered in bacteria since several decades, until today the question remains open whether such biomineralized structures do exist in eukaryotic cells. Herein, evidence was provided for the existence of magnetosome-like Fe-based structures in different viable eukaryotic cells by the registration of second harmonic of magnetization M2(H) of longitudinal nonlinear response to weak ac field. The behavior of the field hysteresis of the M2 response from cells in suspension and/or in pellet indicated a multi-domain state of magnetosome-like structures in certain type of cells, and a single-domain state in other cell lines. The amounts of magnetosomes in cells range from ≤ 1÷2 to 5÷8 per cell. The presence of magnetosome-like structures was analyzed in normal tissue samples obtained from Wistar rats and C57/Bl6 mice. Additionally, the tumor tissue (orthotopic rat C6 glioma and mouse GL261 glioma) were assessed for magnetosomes. Detected magnetosomes in certain tissues (i.e., brain, heart, lungs) matched to a single-domain magnetite nanoparticle, whereas in other organs they exhibited characteristics attributable to a multi-domain state, better corresponding to Fe(0) composition of their magnetic cores. Subsequent studies are necessary to elucidate the role of the Fe-based magnetosome-like structures in the biology and physiology of eukaryotic cells.


Introduction
Magnetosomes are unique organelles which are predominantly found in prokaryotes. Functionally, they are responsible for magnetotaxis in prokaryotes [1,2]. Magnetosomes consist of magnetic mineral crystals, containing either magnetite (Fe 3 O 4 ) or greigite (Fe 3 S 4 ), which are surrounded by a bilayer membrane composed predominantly by phospholipids and a number of proteins that are not present in the cytosol or in the outer membranes (OMs) [3,4]. Magnetotaxis is 1 3 responsible for an optimal positioning of bacteria in vertical chemical concentration gradients by reducing the three-dimensional search problem to a single dimension [5]. Although magnetosomes have been discovered in bacteria since several decades [2,6], the question about the existence of such biomineralized structures in eukaryotic cells has not been answered.
Magnetotaxis which is achieved by the presence of lipid-bound magnetosomes in the magnetotactic bacteria (MTB) aids bacteria to identify optimal growth conditions [7,8]. Although numerous studies clearly demonstrate an influence of low-frequency electromagnetic fields (LF-EMFs) in biological systems [9][10][11][12][13], studies in eukaryotic cell systems mostly remain phenomenological without mechanistic insights. Although ferromagnetism has been described for the magnetotaxis of bacteria and for the navigation of animals (e.g., birds, turtles) [14,15], it remains to be determined whether similar structures are present in eukaryotic somatic cells that could explain the phenomenon of magnetoreception [9] which might influence cell development and differentiation. Presumably the cell-sensing mechanism lies in the existence of magnetosome-like structures that could respond to static and/or oscillating magnetic fields.
A previous biodistribution study of a targeted delivery of functionalized superparamagnetic iron oxide nanoparticles (SPIONs) in tumor-bearing rats resulted in the discovery of a small amount of intrinsic magnetic nanoparticles (MNPs) in tissues such as heart, skin, muscle, lungs, and tumors of control animals that had not been injected with SPIONs [16][17][18]. With respect to this finding, the possible presence of magnetosome-like structures was assessed in cell suspensions of viable normal cells [i.e., chondrocytes, osteocytes, and mesenchymal stem cells (MSCs), etc.] as well as in normal tissues such as heart, skin, brain, and tumor tissues (i.e., rat C6 glioma and mouse GL261 glioma). The very low concentrations of the detected magnetosome-like structures inside eukaryotic cells exclude the use of electron microscopy to identify them. However, highly sensitive nonlinear longitudinal responses to a weak ac magnetic field with registration of second harmonic magnetization M 2 (NLR-M 2 ) could be employed for the analysis of these magnetosomes in vitro. This approach was employed earlier for the investigation of the biodistribution of SPIONs in tumor-bearing rats and allowed to reveal intrinsic MNPs in tissues of control animals [16][17][18]. The high sensitivity of this approach is provided, on the one hand, by registration of responses on a frequency that differs from the frequency of ac field which affects the sample during the study. This allows the separation of the signal of a sample from a highlevel voltage-excited field at the input of the receiver using radio-frequency (RF) methods (see details below), and to register the former against the background of the thermal noise of the receiver [19][20][21]. While in ac susceptibility measurements or ESR studies a signal from a sample represents a small change of ac field h(t) level and registers against the background of the amplitude and frequency noise of RF generator that hinders to increase h(t) for enhancement of the signal. On the other hand, the use of nonlinear responses allows the enhancement of signals from MNPs in comparison with ambient paramagnetic species with a large nonlinearity on MNP magnetization M(H) in a weak field H. The latter is followed by the presence of a specific extremum at a weak H in M 2 (H) dependence 1 3 Detection of Magnetosome-Like Structures in Eukaryotic Cells… [22][23][24][25]. It is registered against very weak M 2 -responses from paramagnetic species [since their M(H) nonlinearity is negligible in the weak field H] with the linear dependency on H. This provides large suppression of the contribution to the signal from ambient organelles and components of eukaryotic cells containing species with paramagnetic properties. In measurements of low frequencies ac susceptibility (in this case the latter is close to the static one), a contribution from paramagnetic species will dominate due to their large amount making it very difficult to extract the MNP response. In magnetic resonance measurements MNPs reveal wide spectra due to effective relaxation and inhomogeneous broadening because of the arbitrary orientation of their anisotropy axes. Therefore, the contribution from small amounts of MNPs is not detectable against the background of narrow and intensive ESR signals from paramagnetic species as it was previously found in carbon-based porous nanomaterials [25]. In measurements of static magnetization by SQUID magnetometer an additional problem appears since this method is sensitive also to the diamagnetic contribution of samples as well as of the cavity.

NLR-M 2 Measurements
MNPs possess a large magnetic moment, therefore, magnetization M(H) of MNP's ensemble exhibits saturation in a weak dc field H. As a result the ensemble response to the ac field h(t) in the presence of the parallel to it dc field H reveals distortions relative to the sinosoidal ac field. The latter are non-symmetric for positive and negative signs of h(t) as it is schematically shown in the left panel of Fig. 1. Therefore, mainly the even harmonics of fundamental frequency f will be generated in a sample during the study, and only the largest second harmonic, M 2 , is registered.
The main features of the experimental setup, providing the high sensitivity of this approach can be briefly summarized as follows: (1) the employment of two-mode (f, 2f) resonant sensor (2-mode sensor) providing: (1) the creation of ac field of frequency f = 15.7 MHz and amplitude h up to 50 Oe in radio-frequency (RF) inductance coil L (with sample under study inside) included in both (f), and (2f) modes; (2) effective registration of a sample M 2 (H) response, inducing 2f-voltage in RF coil L, with help of mode (2f); (2) deep suppression of the 2f-voltage from f-generator by low-frequency filter, including the input mode (f); (3) deep suppression of f-voltage from the output of mode (2f) by high-frequency filter, including at the input of the 2f-receiver; (4) the use of in 2-mode sensor elements and materials, which are not produced spurious 2f-signal. Under these conditions it is possible to register the M 2 signal of a sample against the background of thermal noise of the receiver [19][20][21]. Note, Q 2 -factor of 2f-mode is about 200 that provides increase of setup sensitivity by ~ Q 2 1/2 in comparison with non-resonant registration of χ(f) in ac susceptibility measurements.
At M 2 measurements both M 2 components, ReM 2 (H) and ImM 2 (H), are simultaneously recorded as functions of H synchronously with its scan. For detecting the field hysteresis, H was linearly scanned symmetrically relative to the point H = 0 with frequency F sc without constant bias field H 0 . The amplitude of H-scan was 300 Oe and its frequency, F sc , could be varied in the range 8 Hz ≥ F sc ≥ 10 −2 Hz. Sample temperature was stabilized by flow thermostat, using evaporated N 2 . Different magnetics can be investigated by this technique, including systems with strong electronic correlations ( [22][23][24][25] and Refs. therein), its integral sensitivity of M 2 registration being about 10 −10 emu.
Usually the condition M 2 ∝ h 2 is fulfilled in the experiment, and allows the application of the perturbation theory. In this case the susceptibility of second order χ 2 = M 2 /h 2 can be introduced. For the ensemble of MNPs in superparamagnetic (SPM) regime when average anisotropy energy of MNP is less of energy of thermal fluctuations kT (k is the Boltzmann constant and T is temperature) χ 2 (H, T) is described by the simple expression [22][23][24][25]: Here χ (1) = ∂M(H)/∂H, χ (2) = (1/2) ∂ 2 M(H)/∂H 2 , Γ is a magnetic relaxation rate, ω = 2πf, ω 0 = gμ B H/ħ is the Larmor frequency and μ B is the Bohr magneton. The first term in (1) arises due to a nonlinearity of magnetization curve M(H), it exhibits a static limit and provides the main contribution to the real part of the response, is Langevin function. The found value of NP average moment is μ aver ≈ 5.96·10 4 μ B . Very small linear contribution from low-molecular paramagnetic fraction was taken into account at fitting ReM 2 (H), which is proportional to concentration of MNPs. The second one is due to influence of external magnetic field on relaxation process; it has no static limit and provides the main contribution to the imaginary part of the response, ImM 2 (H), which exhibits an opposite sign relative to ReM 2 . Thus, as it is partly seen from Eq. (1), M 2 -response from MNP ensemble exhibits three peculiarities that can be considered as "finger print" of such systems [16,[22][23][24][25]: (1) [23] and allows one to determine their static and dynamic parameters as well as a distribution of magnetic moment values in ensemble. However, this issue is out of the scope of this paper.
In case of single-domain non-interacting MNPs, the presence of field hysteresis evidences a deviation from SPM regime provided by the magnetic anisotropy. The latter includes contributions of magneto-crystalline, form and surface anisotropy of MNPs [24,26]. In widely common case of the simple uniaxial anisotropy the rate of relaxation of MNP ensemble is given by Γ = 1/τ = f 0 exp(− E a /kT) (without taking into account the effect of field on the relaxation processes) [26]. Here anisotropy energy E a = KV, K is anisotropy constant, V is an average volume of a particle and the frequency f 0 is of the order of 10 9 s −1 . The probability that the magnetization remains in its original position during time t after switching off the external field can be written as P(t) = exp(− t/τ), that gives for retentivity M ret (t)/M 0 = exp(− t/τ). Thus, if the registration time is less than τ, the M ret and H-hysteresis will be detected. In our case of the periodic H-scan, the dynamical H-hysteresis will be observed in ReM 2 (H), when a period of H-scan F sc −1 ≤ τ = Γ −1 . When the time of measurements is equal to relaxation time τ* = F sc −1 = τ, the blocking temperature T B can be introduced, above which the behavior of the system will be superparamagnetic [26]: Note, as Eq. (2) shows, T B depends on experimental conditions, namely, on time F sc −1 that system has for the thermal equilibrium before the moment reversal. So varying F sc permits one to alter T B and H-hysteresis of M 2 response of the singledomain MNPs. The example of dynamic H-hysteresis can be seen in the right panel of Fig. 1, where the signal of dextran-coated single-domain magnetite MNPs close to superparamagnetic (SPM) regime is displayed. The field hysteresis disappears at F sc = 1.32 Hz (not shown). In multi-domain state of MNPs the hysteresis should be practically invariable since it is due to pinning of magnetic moment on structure heterogeneities. In this case the magnetization reversal is determined by movement of domain boundaries instead of MNP moment rotation, and the former is much faster than F sc −1 [26]. Thus, the regime of magnetic behavior of MNPs and its possible transformation under effect of environment can be controlled by a simple way on a presence of H-hysteresis in their M 2 -response. The latter can be characterized by ReM 2 "coercivity", H C2 shown in Fig. 2a, which is found from the condition ReM 2 (H C2 ) = 0. Its dependence on F sc and temperature allows one to discriminate single-/multi-domain states of MNPs and to determine T B [24].

Cells
Rat C6 glioma and mouse GL261 glioma cell lines were grown at 37 °C in RPMI-1640 cell medium supplemented with 10% fetal bovine serum (FBS), Detection of Magnetosome-Like Structures in Eukaryotic Cells… 2 mM l-glutamine and antibiotics (100 units/mL Penicillin G and 100 μg/mL Streptomycin). Cell lines were regularly tested negative for mycoplasma contamination. Cells were passaged twice a week and single-cell suspensions were derived by short-term (less than 1 min) treatment with 0.25% (w/v) Trypsin-0.53 mM EDTA.

Isolation of Primary Chondrocytes
Rabbit cartilage obtained from articular surfaces was minced and sequentially digested by the following Sigma-Aldrich enzymes, 0.05% hyaluronidase, 0.25% trypsin and 0.4% collagenase to harvest primary chondrocytes for monolayer culture. The isolated cells were washed and suspended in DMEM-F12 complemented with 15% FBS (GIBCO/Life Technology, NY, USA) and 1.0% penicillin-streptomycin solution. Prepared primary chondrocytes were seeded in 25-cm 2 flasks with 8 × 10 5 /mL cells and cultured in DMEM-F12 medium till 80% confluence followed.

Primary Osteocytes
Purified osteocyte and primary osteoblasts were derived from both parietal bones. Briefly, rabbit's parietal bones of the calvaria and tibiotarsal bones were cleared of the surrounding soft tissue and periostea. Medullary cavities were flushed with a Dulbecco phosphate-buffered saline (Invitrogen, Paisley, UK), Three sequential digestions of bone segments with 1 mg/mL collagenase type 1 (Clostridium histolyticum, Sigma, Dorset, UK) in PBS followed by 4 mM ethylenediaminetetraacetic acid (Sigma) in PBS. Digestion was stopped by incubation with 10% FBS in a Hankbalanced salt solution (HBSS) (Invitrogen), and each of these three consecutive fractions were centrifuged (800g, 4 °C for 5 min) and then resuspended in heat-inactivated chicken serum in an HBSS on ice. A single-cell suspension was resuspended in PBS containing 4 mM ethylenediaminetetraacetic acid and 0.5% bovine serum albumin (BSA; Fraction V, Sigma) (PEB) to produce a mixed bone-derived cell population. Bone-derived cells were cultured in Dulbecco Modified Eagle Medium minus phenol red, 5% FBS, 2 mM l-glutamine, 50 μg/mL gentamicin (Invitrogen); 50 μg/mL l-ascorbic acid; 5.6 mM glucose (Sigma).

Results and Discussion
We All the samples exhibit M 2 (H) signals revealing extremum in a weak field, opposite signs of signals in ReM 2 and ImM 2 components (the latter is basically not shown) and H-hysteresis, which was appreciably larger than that of single-domain Fe 3 O 4 MNPs close to SPM regime presented in Fig. 1. This suggests the presence of magnesome-like structures, including magnetic (probably Fe-based) core in studied cells similar to that in prokaryotes [1,2]. Note, as a rule, the field hysteresis of ReM 2 (H) signal from suspensions of studied cell cultures was larger or equal to this of samples with precipitated cells assuming some potential effects of cell contacts on the response parameters (see Figs. 2, 3). The presence of considerable H-hysteresis in the M 2 (H) response of magnetosome-like particles evidences blocking regimes or multi-domain state of their magnetic core. To clarify this issue we have studied the dependency of the M 2 (H) response on the frequencies of H-scan, F sc .  [17] with an effective Fe concentration of C(Fe) ≈ 0.01 mg/mL. The shape of ReM 2 (H) hysteretic curves of the latter is close to that of cell suspensions at F sc = 8 Hz, (see panels a, c) assuming the Fe nature of the magnetic cores of magnetosome-like structures, which generate the M 2 response of the cell samples.
As shown in the Fig. 2a, b, decreasing of the F sc is followed by a substantial decrease of H-hysteresis in signals of the single-domain Fe(0)@MCM41 MNPs that are in the blocking regime, while the latter is practically absent in the response of MSC + CC samples both in suspension and in pellet. This is indicative for a mainly multi-domain status of magnetosome magnetic cores in this sample. These multi-domain magnetosome-like patterns of eukaryotic cells are different from chainlike single-domain magnetosomes described for bacteria and protists [27,28]. At the same time, the response of the MSC_OC cells reveals a decreasing H-hysteresis similar to that of Fe(0)@MCM41 MNPs (at F sc decreasing the positions of ReM 2 "coercive force", H C2 , move to lower field for both samples and only their relative positions approximately conserved), at least, for pelleted MSC_OC cells, see Fig. 2c, d. This implies some contribution of the The assumption of the Fe(0) nature of magnetic cores in magnetosome-like structures allows one to estimate Fe(0) concentration inside according to the amplitude of the latter compared to that of the Fe(0)@MCM41 (Fig. 2a, c). This results in: (1) C(Fe) ~ (1.2 × 10 −7 /7.9 × 10 −4 )·[C(Fe) = 0.01 mg/mL] ≈ 1.5 × 10 −3 μg/mL = 1.5 ng/ mL for MSC + CC sample; and (2) C(Fe) ~ (7.4 × 10 −8 /7.9 × 10 −4 )· [C(Fe) = 0.01 mg/ mL] ≈ 9 × 10 −4 μg/mL = 0.9 ng/mL for MSC_CO one. The volume of the samples was always 0.5 mL and the concentration of cells in suspension was ~ 10 5 cells/mL which allows an estimation of the Fe content per cell of ~ 1.5 × 10 −5 ng and ~ 0.9 × 10 −5 ng for considered samples accordingly. Transforming these quantities in moles of Fe(0) we obtain ~ 2.7 × 10 −16 mol(Fe)/cell and ~ 1.6 × 10 −16 mol(Fe)/cell, respectively. The latter corresponds to ~ 1.6 × 10 8 (Fe atoms)/cell and ~ 9.6 × 10 7 (Fe)/cell accordingly.
The critical diameter of single-domain Fe nanoparticles is estimated as ~ 20 nm [29]. Therefore, the diameter of of multi-domain magnetosome-like structures of MSC + CC and MSC_OC is not less than the double-critical size of the Fe magnetic core of ~ 40 nm. Approximately 2 × 10 7 Fe-atoms should correspond to this minimal size of the magnetic core. The latter estimation matches to the maximum of ~ 5 ÷ 8 "magnetosomes" per cell (for MSC + CC and MSC_OC cells). M 2 response from other cell lines is at least 3 ÷ 5-times less indicating the presence of ≤ 1 ÷ 2 magnetosomes per cell. This fact hampers the application of electron microscopy for the magnetosomes' identification in the cells. Figure 3 displays the results of NLR-M 2 measurements carried out on human osteosarcoma (HOS) cells and chondrocytes in suspension and precipitated samples. As shown in panels (a), (b) the response of HOS cells is practically independent of their aggregate state (single-cell suspension or precipitated sample) similar to that of MSC + CC cells, although the "magnitosome" amount in them (∝ amplitude of their signal) is drastically lower. At the same time the aggregate state of other cell cultures, chondrocytes, strikingly affects on H-hysteresis of their M 2 response, as it can be seen in Fig. 3c, d. Therefore, Fig. 3 supports the above-postulated generalizations.
Biodistribution studies of control tissues of glioma-bearing rats (tumor, heart, etc.) without MNP revealed hysteretic M 2 (H) signals in NLR-M 2 experiments implying the presence of internal MNPs in them [16][17][18]. In this work we performed similar measurements in probes of rats with C6 glioma and mice with GL261 glioma to confirm the presence of the magnetosome-like structures in tissues of several organs and in tumor cells, to compare the data with those of cultured cells. The results are displayed in Fig. 4.
The comparison of ReM 2 (H) amplitudes in panels (a), (b) of Fig. 4 shows increased amounts of magnetosome-like structures in mouse glioma. Besides, a weak "coercivity" signal detected in rat glioma (Fig. 4a) suggests the state of its magnetosome-like magnetic cores close to single-domain one in the blocking regime. While in mouse glioma, the multi-domain state of the latter is definitely realized since its ReM 2   ReM 2 (H) signals from C6 glioma and normal tissues (brain, skin, heart) of the rat (left column, panels a, c, e, g) and GL261 glioma as well as normal tissues (brain, skin, heart) of the mouse (right col- umn, panels b, d, f, h). In all the panels the presented signals were obtained at F sc = 8 Hz while in panels (b) and (f) the signal from tumor and skin tissues of mouse recorded at F sc = 0.25 Hz is shown as well. In panels (b, e) the ReM 2 (H) signals from suspension of Fe(0)@MCM41 MNPs [with effective concentration of Fe C(Fe) ≈ 0.1 mg/mL] are displayed for comparison. Panels (d, h) present also the results of ReM 2 (H) fit by ∂ 2 L(H)/∂H 2 , here L(H) is the Langevin function temperature (in agreement with Eq. 1) similar to this for Fe(0)@MCM41 MNPs (see Fig. 2b). The form of the ReM 2 (H) response from mouse tumors is similar to that of the Fe(0)@CMC41 MNPs, implying close to Fe(0) composition of its magnetosome-like magnetic core. Similar features are found in the ReM 2 (H) response of mouse skin tissue [panel (f)], assuming the same state of magnetic core in it. The close or even larger value of H C2 in signals of the rat skin tissue (Fig. 4e) is indicative for similarities in their magnetosomes-like structures although, their concentrations are smaller which is in agreement with a lower signal amplitude.
Signals from mouse brain sample [i.e., ReM 2 (H) responses] (Fig. 4d) definitely correspond to that of single-domain MNPs close to the SPM regime. Indeed, the decreasing of F sc down to 0.25 Hz leads to a collapse of the ReM 2 field hysteresis (data not shown). The best fit of ReM 2 (H) curve by corresponding functions obtained from the Langevin is shown in Fig. 4d with a value of MNP average moment of μ aver ≈ 1.24 × 10 5 μ B . This value is larger than 6 × 10 4 μ B obtained from the same best fit of ReM 2 (H) response of dextran-coated Fe 3 O 4 MNPs with an average diameter of ~ 10 nm while the "coercivities" and forms of both signals are similar (Fig. 1). These results argue for a magnetite core in mouse brain magnetosome-like structures with about two times larger volumes of their magnetic cores (d aver ~ 13 nm). The description of the signal by fitting curves is not suitable since the considerable affect of magnetic field on relaxation processes was not taken into account for the fitting. This is evidenced by the presence of the ImM 2 (H) component of the weakly hysteretic signal which is comparable with the ReM 2 (H) amplitude and an opposite (positive in H > 0 region) sign which is due to this affect [23,24]. The signal from the rat brain is almost, by an order of magnitude, smaller and exhibits larger H-hysteresis (H C2 ) whereas the position of its extremum is in a similar weak field H. This indicates the smaller amount of magnetosome-like structures in rat brain and does not allow to assume a magnetite nature of magnetosome-like cores in rat brain. However, drastically larger magnetic anisotropy is observed in rats compared to mice. The latter should be accompanied by blocking regime of their behavior leading to larger H-hysteresis in ReM 2 (H) in agreement with the data.
Comparative analysis of the M 2 responses from normal brain and glioma tissues could indicate a plausible transformation of magnetosome-like magnetic core state from magnetite to Fe(0) in tumor cells.
The features of ReM 2 (H) signals from heart tissue of the control mouse are similar to those of brain (Fig. 4h) and indicate the plausible magnetite composition of magnetosome-like magnetic cores in it which are close to SPM regime behavior. The fit of response by the same function obtained from Langevin includes the slightly larger value of the MNP average moment μ aver ≈ 1.38 × 10 5 μ B , which is also presented in this Fig. 4. The observed discrepancy of experimental signal and fitting curve is explained again by the presence of the non-considered contribution to the response from the effects of the external magnetic field on relaxation processes [23,24]. The more hysteretic ReM 2 (H) signal from heart tissue of the control rat exhibits a comparable amplitude with extremum position in a weak field close to that of the mouse response. This shows the presence of a comparable amount of magnetosome-like structures with magnetite core in heart tissues of both rodents but lower oxygen supply revealed hysteretic M 2 (H) signals more corresponding to Fe(0) multi-domain magnetic core.