We report the use of S-band ferromagnetic resonance (FMR) spectroscopy to compare the anisotropic properties of magnetite particles in chains of cultured intact magnetotactic bacteria (MTB) between 300 and 15 K with those of sediment samples of Holocene age in order to infer the presence of magnetofossils and their preservation in a geological time frame. The spectrum of intact MTB at 300 K exhibits distinct uniaxial anisotropy because of the chain alignment of the cellular magnetite particles and their easy axes. This anisotropy becomes less pronounced upon cooling and below the Verwey transition (TV) it is nearly vanished mainly owing to the change of direction of the easy axes. In a natural sample, magnetofossils were detected by uniaxial anisotropy traits similar to those obtained from cultured MTB above TV. Our comparative study emphasizes that indispensable information can be obtained from S-band FMR spectra, which offers even a better resolution than X-band FMR for discovering magnetofossils, and this in turn can contribute towards strengthening our relatively sparse database for deciphering the microbial ecology during the Earth's history.
Life on the Earth is strongly associated with microbes and the earliest evidence for their presence has been hypothesized from putative morphological microfossils in Archaean rocks of about 3.5 Ga [1,2]. Geological records of microbial biota are sparse, because soft-bodied organisms that are expected to dominate natural environments do not preserve well. Magnetotactic bacteria (MTB) and their chemically stable magnetic remains, known as magnetofossils, have attracted considerable interest as proxy to infer microbial ecology during the Earth's history [3–7]. Although the magnetic properties of MTB are well known [8–13], the detection of magnetofossils in geological samples remains ambiguous. The hallmark of MTB is to form intracellular ferrimagnetic particles encapsulated in membranes, termed magnetosomes. These biominerals consist of stable single domain (SSD) magnetite (Fe3O4) or greigite (Fe3S4), with narrow size and shape distributions, organized along their magnetic easy axes  in chains that are stabilized by cytoskeletal protein filaments [8,9,14–19]. In this configuration, the total magnetic moment of the cell is maximized and the chain can act as compass to guide MTB along the Earth's magnetic field lines towards their favourable habitats. Moreover, the chain configuration causes pronounced magnetic interaction-induced shape anisotropy.
In contrast to intact MTB, the detection of magnetofossils in geological deposits is intricate, because diagenetic processes can lead to the decomposition of the cellular matter and partial collapse of the chain configuration . For the unequivocal identification of magnetofossils, generally two criteria, among others, have to be fulfilled [6,21,22]. First, the magnetite particles must be in a SSD state; second, the particles must show some degree of particle alignment in chains. The grain size criterion can be addressed by rock magnetic measurements such as hysteresis loops and remanence curves [12,23]. Thus far, there are experimental procedures to detect magnetofossils such as first-order reversal curves or field cooled–zero field cooled experiments [12,24], but no standard approach for the unambiguous identification of chain configuration in magnetofossils. Among magnetic methods, however, ferromagnetic resonance (FMR) spectroscopy is the one that probes anisotropy fields in magnetic material by measuring the absorption of a precessional magnetization in an applied magnetic field [25,26].
Different responses of intact and lysed MTB to X-band FMR spectroscopy with a microwave frequency in the 9 GHz range provide compelling evidence that spectral traits (e.g. asymmetry of the absorption spectra) permit a direct way to detect the anisotropy of magnetosome chains [11,27–32]. Mastrogiacomo et al.  showed that FMR experiments of MTB using microwave frequency of 4 GHz, i.e. S-band, yield pronounced resonance patterns for uniaxial anisotropy of magnetite chains. Furthermore, these patterns are narrower and better resolved at lower microwave frequencies , which can be advantageous for studying magnetically heterogeneous natural samples in order to infer magnetofossils.
In this study, we apply S-band FMR to intact MTB between room temperature and 15 K, and compare the spectral response to those from selected Holocene lacustrine sediment samples in order to define spectral criteria for the detection of magnetofossils.
2. Ferromagnetic resonance
2.1. Basic principles and simulation
A magnetic moment precesses around a magnetic field B with a frequency equal to ge μB B/h (Larmor frequency), where h is the Planck's constant, μB the Bohr magneton and ge the spectroscopic splitting factor of an electron. When the Larmor frequency matches the frequency of incident electromagnetic radiation ν, the energy of that radiation is absorbed by the system and resonance occurs. In FMR experiments, the frequency is generally kept constant and an external magnetic field Bex is changed in order to establish resonance conditions that can be described by hν = gμB Bres. The resonance field Bres and the corresponding g-value are affected by the internal field Bint of the magnetic system. If Bint is anisotropic, Bres and g-value depend on the orientation of the magnetic system with respect to Bex. Furthermore, in magnetic systems, where all orientations with respect to the magnetic field are present, the FMR signal is the spectral superposition of all orientations properly weighted by their individual intensity and line-width. The spectral parameters ΔB, Beff and geff are generally used to describe FMR signals. The effective resonance field Beff and its corresponding effective g-factor (geff), calculated from the resonance equation, are determined at maximum absorption [34,35]. Owing to the use of lock-in amplifiers in FMR experiments the recorded signals correspond to the first derivatives of the absorption profiles. Beff is then determined as the zero-crossing of the curve and the line-width ΔB is measured from peak to peak. In such case, a low-field Beff as a local maximum has been reported in S-band FMR spectra .
Thus far, FMR on MTB was mainly performed using microwave frequencies ν ≈ 9 GHz, i.e. in the X-band range [11,27,29,36]. S-band FMR spectroscopy operates with microwave frequencies in the 4 GHz range. The main difference between S- and X-band FMR experiments on MTB is that in the latter the resonance condition is fulfilled when the magnetization of all chains are saturated, i.e. Beff > Bsat, whereas in S-band, resonance can occur at fields Beff < Bsat. With this in mind, the remanence state of a sample can critically affect the low-field absorption in S-band FMR spectra. Mastrogiaccomo et al.  showed for magnetite chains in MTB that the remanence magnetization has a considerable impact on the absorption profile.
Numerical considerations are presented to evaluate differences between X- and S-band FMR spectra (figure 1a–d). When an external magnetic field is applied to a magnetic system, the magnetization vector tends to align with the field-axis. Depending on the intrinsic anisotropy fields in the system and the strength of the external field, the magnetization vector will dwell in an equilibrium orientation. The equilibrium angle of the magnetization for a certain FMR signal can be calculated by minimizing the free energy density and solving the resonance equation following an iterative procedure . Figure 1c shows an example of this calculation considering a system with uniaxial anisotropy field Buni = 80 mT and a cubic anisotropy field of Bcub = −23.5 mT, values typical for magnetite chains of Magnetospirillum gryphiswaldense . Figure 1c shows the equilibrium polar angle of the magnetization θ as a function of the magnetic field angle θB. The comparison between X- and S-bands makes clear that for the latter, the resonance condition is fulfilled for lower external fields and the magnetization is not saturated along the field-axis when the resonance condition is met, as seen by the strong deviation of θ from the diagonal (figure 1c).
Moreover, for every orientation the intensity of the absorption profile is proportional to the squared projection of the magnetization onto the external magnetic field . Given this, the relative FMR intensity at each field orientation can be calculated as shown in figure 1d, where the normalized FMR intensity is exhibited as a function of θB. At X-band frequencies the system, which is close to saturation, will exhibit nearly the same FMR intensity at all orientations, with deviations of 2–3% from the maximum. For S-band frequencies, however, pronounced deviation will occur for intermediate orientations, where the FMR intensity will be more than 10 per cent lower due to the misalignment of the magnetization vector with the field vector (figure 1d).
Using these results, an FMR signal at S-band can be simulated by considering the relative intensity at each field orientation. Details about the simulation are reported in the study of Charilaou et al. . Figure 1b presents such S-band spectrum with distinctly separated features, i.e. low-field peak due to resonance of chain aligned parallel to the external magnetic field and high-field peak due to resonance of chains aligned perpendicular to Bex. The separation of the anisotropy features is vital for the analysis of MTB samples, because it enables the direct observation of the pronounced uniaxial anisotropy due to stable chain configurations with strong dipole moment coupling. In addition, as observed in experiments at S-band, the low-field peak becomes enhanced owing to the remanence magnetization of chains aligned along the field-axis . This further increases the potential of S-band FMR spectroscopy for the detection of MTB.
In another simulation, the effect of uniaxial anisotropy field Buni on the spectral response is tested using the same model . As shown in figure 2, Buni generated by the chain configuration of magnetic particles determines the width of the FMR signal of the ensemble, i.e. the two peaks shift closer together. Decreasing Buni while keeping Bcub constant at −23.5 mT, the two peaks are shifted closer together (figure 2). At some point, the peaks become indistinguishable and a broad feature at relatively low-field will be observed. Moreover, considering the inhomogeneity in natural samples, a distribution of effective Buni for each chain can be assumed and in such case additional broadening would be imposed on the overall FMR signal. From this simulation, it can be argued that FMR signals with different anisotropy fields can provide an insight into chain configurations, because the alignment along the chains promotes stronger Buni.
2.2. Experimental set-up
For S-band experiments, cultured cells of M. gryphiswaldense strain MSR-1 were used in the earlier studies [11,30,31]. The MTB were cultured following the procedure by Heyen & Schüler . The grown bacteria show chains of nearly isometric SSD magnetite particles with sizes of about 40 nm. Superparamagnetic particles with sizes less than 30 nm were found at the ends of the chains . The sediment samples SOP258 and SOP300 were taken from a sediment piston core of Lake Soppen in central Switzerland covering the Holocene. Sedimentological and rock magnetic properties of these samples were previously described . The cultured MTB and the sediment sample SOP300 contain SSD magnetite particles that saturate in fields Bsat ≈ 130 mT ; a similar value was determined for sample SOP258.
Freeze-dried cells of MTB and lake sediment samples, embedded in paraffin to protect them from alteration, were fixed in ESR quartz glass tubes, and the S-band FMR spectra were measured on a custom-built spectrometer controlled by Spec-Man software operating at a microwave frequency of 4.02 GHz, power of 2 mW and modulation amplitude of 0.1 mT in Bex sweeping from 5 to 400 mT. The data acquisition for an FMR spectrum is fast (within minutes) and can be performed on routine basis. All samples were recorded after they were exposed to fields >Bsat. For low-temperature measurements, the temperature of the sample in the cavity was controlled by a helium gas-flow cryostat ESR 910 (Oxford Instruments). The measurements were performed after zero-field cooling to 15 K and subsequent heating to room temperature. The spectral parameters, Beff, geff and ΔB were derived from first derivative absorption spectra.
3. Results and discussion
3.1. S-band spectrum of magnetotactic bacteria at room temperature
The S-band FMR spectrum obtained from intact cells of MTB at room temperature exhibits an asymmetric signal with two distinct resonance fields and a shoulder at about 170 mT (figure 3). The high-field resonance Beff = 158.4 mT, corresponding to geff = 1.81 and the low-field Beff = 38.03 mT (geff = 7.55).
Detailed multi-frequency FMR study on intact MTB by Mastrogiacomo et al.  showed that the low-field Beff arises from magnetosome chains that are aligned parallel to Bex. In such array, these chains are saturated in contrast to those with orientations different from Bex. The broad peak at 140 mT, the minimum at 203 mT and the shoulder at about 170 mT originate from the contribution of magnetocrystalline anisotropy fields and the anisotropy fields owing to chains perpendicular to Bex (figure 3). The relatively high intensity of the low-field compared with the high-field resonance obtained from the MTB sample is due to the remanence magnetization of the bulk sample . If magnetite has been exposed to a saturation field Bsat, which is less than 300 mT, the remanence magnetization is about 50 per cent of magnetization at Bsat for SSD particles . Such remanence markedly contributes to fulfil the resonance condition of the magnetite chains parallel to Bex, i.e. in the case, where the bulk sample is not magnetically saturated. It is worth noting that the effect of the remanence state is neglected in the simulation, and, therefore, the low-field resonance is less pronounced compared with the measured MTB sample (figure 1b).
3.2. S-band spectra of magnetotactic bacteria at low temperature
Figure 3 shows the low-temperature spectral behaviour of the MTB sample. The signal intensity reaches a maximum at 260 K followed by a decrease down to 15 K. Upon cooling, the spectra exhibit a shift to higher geff values, respectively, lower Beff and a broadening of the line-width ΔB. These parameters suggest changes in the internal fields and the overall anisotropy of the sample. The geff values of the high-field feature decrease from 1.81 at 300 K to 1.99 at 230 K with corresponding Beff-values of 158.4 and 144.3 mT. Values of geff < 2 are indicative of distinct demagnetization fields that in the case of MTB originate from the interaction-induced shape anisotropy . At 200 K, geff = 2.20 (Beff = 130.3 mT) and increases to 2.61 (Beff = 109.9 mT) at 80 K. Considering Bsat ≈ 130 mT, the high-field features down to 200 K occur at Beff ≥ Bsat, i.e. the high-field absorption is obtained from the magnetically saturated bulk sample. At lower temperature, where Beff shifts to lower fields, absorption arises from the bulk sample with partly unsaturated magnetite particles. This situation is even more pronounced for signals obtained in the low-field absorption range between 300 and 170 K with Beff ≪ 130 mT and geff values between 7.55 and 15.31. To obtain Beff in very low external fields two conditions are required. The particles should be easily magnetized and their magnetization should substantially contribute to the internal field. Considering the chain configuration in intact MTB, these conditions are given because magnetite is a soft magnet and the linear alignment of magnetic particles along their easy axes maximizes the magnetic moments of the ensembles. The shift of the low-field feature to lower fields upon cooling is affected by the change of the magnetocrystalline anisotropy constant |K1| down to the isotropic point at about 130 K, where K1 = 0 [25,31]. With decreasing strength of the cubic anisotropy field, the uniaxial field Buni of the chain is compromised, because the magnetic moments of the individual particles are not strongly confined along the chain axis anymore. Superparamagnetic contributions must also be taken into consideration when interpreting these spectra, because with decreasing temperature the superparamagnetic magnetosomes become blocked and contribute to the total anisotropy and also to the total magnetization of the sample [11,13,37].
Below 120 K, the disappearance of a clear low-field absorption feature and the high-field shoulder is mainly affected by the Verwey transition (TV). It is worth noting that for magnetite particles in M. gryphiswaldense TV ≈ 100 K was found . At this intrinsic transition, magnetite undergoes a structural change from cubic to monoclinic with a shift of the easy axes from  to , which causes a prominent change in the magnetization properties [41–45]. The switching of the magnetic axes is associated with an increase of |K1| as calculated from monoclinic anisotropy constants and a decrease of magnetization [43,46,47]. Considering the chain configuration in MTB, the switching of easy axes from  to  weakens the net magnetic moment of the chains and randomizes their contribution to Bint. Thus, the effect of the interaction-induced shape anisotropy, which is the main source for the uniaxial anisotropy contribution to the FMR signal of MTB, is averaged out . Given this, the broad and weak absorption in fields of less than 60 mT in the spectra below TV can be interpreted as the remaining contribution of the interacting magnetite particles in chains without alignment of magnetic moments along the chain axis (figure 3).
The decrease of the signals below 80 K can be explained by the large monoclinic anisotropy constants  that hamper the alignment of the magnetization vector along the external magnetic field (figure 1a). This, in turn, causes reduction of the FMR intensity, as discussed in §2.1. Moreover, the decay of the effective dipole moment owing to the switching of the easy axes below TV, also causes further reduction of the FMR intensity.
3.3. S-Band spectra of natural samples
Spectral responses of magnetofossils in sediment samples, where the decay of cellular matter during diagenesis disintegrated the chain configuration of the magnetic particles and their alignment of magnetic moments, are expected to be similar to responses obtained from the low-temperature series shown above (figure 3). To verify this hypothesis, two lacustrine sediment samples from Lake Soppen were taken [32,39].
The selected lacustrine sediment samples contain magnetite particles in an SSD state that is a required criterion for magnetofossils . The two samples differ in their magnetite content according to their bulk susceptibility with 1.75 × 10−7 m−3 kg−1 for SOP300 and 1.79 × 10−6 m−3 kg−1 for SOP258, respectively. The higher content of the latter has been interpreted as magnetite accumulation owing to MTB . The different FMR intensity of the two samples is also manifested in the S-band spectra. For the sake of comparison, the FMR spectrum obtained from SOP300 was multiplied by a factor of 10 in figure 4. This spectrum is relatively broad with Beff = 126.5 mT, corresponding to geff = 2.27, ΔB ≈ 69 mT, and a slightly stronger absorption intensity in the high-field than in the low-field range. Considering that Beff ≈ Bsat the asymmetry of the spectrum is most likely owing to the fact that in the low-field range the bulk sample is not fully saturated, i.e. not all magnetization vectors are aligned along the external field. Despite this effect, the shape of the broad FMR signal is similar to those reported for SSD magnetic particles disperse in a matrix or in clumps [28,49]. The low-field feature at g = 4.2 originates from paramagnetic Fe(III) cations, structure-bound in a host mineral , and they do not contribute to the magnetic anisotropy of the bulk sample. Furthermore, a signal of a radical with g = 2 was found.
The sample SOP258 exhibits an asymmetric spectrum with four features, two peaks at 58 and 112 mT, a shoulder at about 155 mT and minimum at 212 mT (figure 4). The shape of the spectrum is similar to those of the intact MTB at about 170 K with respect to the peak at 112 mT, the shoulder and the high-field minimum (figures 3 and 4). These features can be assigned to absorptions due to magnetocrystalline anisotropy contribution and chains perpendicular to Bex. Furthermore, the low-field feature can be attributed to chains with parallel orientation to Bex. Compared with intact MTB, this feature is broader and occurs at higher fields. Considering the simulations in figure 2 the shift to higher fields is probably owing to lesser Buni. The lesser Buni can be explained by uneven alignment of the easy axes and/or by shorter chains. With the departure of the magnetic easy axes from the magnetite chain axis, higher fields are needed to fulfil resonance conditions. It is worth noting that fields in the range of 30–80 mT are found to align or switch the magnetic moments in MTB chains [33,51,52].
Comparing the FMR spectra obtained from the two natural samples, SOP258 exhibits clear evidence in the low-field range for uniaxial anisotropy as expected for magnetofossils. The comparison of the low-field absorption of this sample with those from low-temperature series and numerical simulation of intact MTB (figures 2 and 3) suggests that magnetofossils in sediments are preserved in chain configuration with partially broken alignments of the magnetic moments.
We have proved that the preservation of magnetofossils in geological samples can be detected by the anisotropy traits obtained from S-band FMR spectra and the comparison with spectral features of intact MTB. The disappearance of the uniaxial anisotropy contribution of the chain configurations in intact MTB below the Verwey transition can be taken as fingerprint to estimate the spectral response of magnetofossils, where the decay of cellular matter during diagenesis disintegrates the configuration of the magnetic particles and their alignment of the easy axes. The natural sample expected to contain magnetofossils exhibits S-band FMR spectrum similar to that of intact MTB recorded at low-temperature at about 170 K. Broad features attributed to parallel and perpendicular contributions to the uniaxial anisotropy provide clear evidence for magnetite particles in chain configuration and in turn for magnetofossils. The other natural sample reveals a relatively symmetric signal with a low- and a high-field peak that is expected for dispersed or clumped single domain magnetite particles. Our study emphasizes S-band FMR spectroscopy as a powerful tool to detect magnetofossils, and it can therefore make a valuable contribution to decipher the microbial ecology during Earth's history.
The authors would like to thank the three reviewers for their helpful comments. The research was financially supported by the Swiss National Science Foundation (grant no. 121844) and CHIRP1 project of ETH Zurich.
- Received October 1, 2012.
- Accepted November 30, 2012.
- © 2012 The Author(s) Published by the Royal Society. All rights reserved.