II. RESULTS AND DISCUSSION

Figure 1 shows HAADF-STEM images of individual nonmagnetic bacterial cells acquired before and after exposing the bacterial culture to the iron-replete growth medium, confirming the absence of magnetosome nanocrystals before the induction of biomineralization. While the cultures used in this study are nonmagnetic, occasional, rare magnetosomes could be detected at the initial time point [Fig. 1(a)]. Exposure of nonmagnetic culture to the iron-replete growth medium leads to rapid formation of several magnetosome particles in Figs. 1(b) and 1(c). A bacterium from the culture grown under regular conditions is shown in Fig. 1(d) for comparison. Figure 2(a) shows HAADF-STEM image of a bacterial cell acquired from the sample collected 30 min after the injection of iron-replete growth medium. The larger and strongly scattering magnetosome nanocrystals in the area marked with arrow, are, most likely, present in the cell before the biomineralization induction, as they appear dense and structurally perfect. By comparison, the rest of the nanoparticles in the same bacterial cell, which appear significantly smaller and less ordered, are attributed to the onset of the biomineralization process. The large and small magnetosomes selected for the structural analysis are shown in green and blue rectangles, respectively. The SAED pattern of the larger particle in Fig. 2(b) reveals crystalline lattice with {400}, {220}, and {422} planes indexed to magnetite, while the SAED of the smaller particle in Fig. 2(c) is consistent with an amorphous structure. The plane indexed to {331} in large magnetosome was reported as a structural defect earlier by Taylor et al. ^{Reference Taylor, Barry and Webb51}

FIG. 1.

HAADF-STEM images of individual bacterial cells of M. magneticum, strain AMB-1 grown under the low-iron conditions, sampled (a) before, (b) 30 min after and (c) 1 day after the Fe-replete growth medium injection. Bacterial cell in (d) was grown under regular conditions. Scale bars: 500 nm.

FIG. 2.

(a) HAADF-STEM image of a bacterial cell obtained from the sample 30 min after induction. The bright field transmission electron microscopy (BF-TEM) images of large (green rectangle) and small (blue rectangle) magnetosomes are given in the upper and lower insets, respectively. The larger magnetosomes in the area marked with arrow are likely present in this cell before the biomineralization induction; (b) the large magnetosome is observed to be crystalline as shown in SAED with planes {004}, {220}, and {224}, indexed to magnetite; (c) SAED of the small magnetosome shows continuous diffuse rings, consistent with lack of crystalline order. Scale bars: (a) 200 nm; (b and c) 10 nm; insets: 10 nm⁻¹.

HAADF-STEM in conjunction with EDS was previously used by Baumgartner and coworkers to produce spatially-resolved elemental maps of iron and phosphorus in localized phosphate-rich ferric hydroxide. ^{Reference Baumgartner, Morin, Menguy, Perez Gonzalez, Widdrat, Cosmidis and Faivre44} Notably, while the EDS provides the resolution necessary for the identification of chemical makeup of the formed biominerals, it does not distinguish the oxidation states of their individual constituents, prompting the use of additional analytical techniques. As a result, probing the chemical environment of iron was therefore carried out in their case, which was identified by XAS. ^{Reference Baumgartner, Morin, Menguy, Perez Gonzalez, Widdrat, Cosmidis and Faivre44} However this technique provides a bulk signal, and as such is not applicable to analysis of individual particles. According to the x-ray absorption near-edge spectroscopy (XANES) and extended x-ray absorption fine structure, the three components reported by the coauthors along the induction series were phosphate-rich ferric hydroxide-like and oxyhydroxide-like phases detected 10 and 30 min after induction, respectively, in addition to the growing magnetite phase. This finding was confirmed by the independent study from Fdez-Gubieda and coworkers, who also used XANES to probe the biomineralizaion pathway for M. gryphiswaldense, and reported on this proceeding via the formation of ferrihydrite with high phosphorus content, followed by its conversion to magnetite. ^{Reference Fdez-Gubieda, Muela, Alonso, Garcia-Prieto, Olivi, Fernandez-Pacheco and Barandiaran52} The fast magnetosome magnetite biomineralization, reported by Staniland and coworkers, ^{Reference Staniland, Ward, Harrison, van der Laan and Telling43} is in contrast to the previously proposed slow multistep formation mechanism. ^{Reference Frankel, Papaefthymiou, Blakemore and O'Brien18,Reference Arakaki, Webb and Matsunaga37,Reference Schüler and Baeuerlein53–Reference Pearce, Henderson, Pattrick, van der Laan and Vaughan62}

In our study, the particles with sizes <10 nm appeared to be amorphous (n = 4), and with sizes >15 nm appeared to be crystalline. Based on the analysis of the crystalline biomineral phases formed for magnetosomes of varying sizes, particles of 10 nm < size < 15 nm exhibited mixed amorphous (n = 3) and crystalline (n = 2) structures. When dealing with small particles in large bacterial cells, detailed characterization of small nanoparticles is challenging, as in addition to the diminished contrast, there is an increased likelihood of triggering the beam-induced amorphous-to-crystalline transition associated with the accumulation of high electron dose over a very short period of time, as well as the change in electronic configuration of iron. To minimize the ambiguity, we (i) reported here only the results obtained on the magnetosomes residing inside the body of the bacterial cell and (ii) omitted the measurements acquired on the smallest magnetosome particles. Doing so resulted in a relatively small number of samples. The reported results, however, are consistent and representative of most small magnetosomes. Notably, the crystalline magnetosome particles in this transition range could not be indexed directly to just one mineral: based on the interplanar spacings measured in the particles, these particles may consist of ferrihydrite, ^{Reference Stevens, Khasanov, Mabe, Mercader, Gancedo, Cabral Prieto and Baggio-Saitovitch63} magnetite, ^{Reference Swanson, McMurdie, Morris and Evans64} maghemite ^{Reference López-Walle, Reyes-Melo and Rakotondrabe65} and/or hematite. ^{Reference Morris, McMurdie, Evans, Paretzkin, Parker and Panagiotopoulos66} The planes were first indexed to magnetite, and the remaining unmatched peaks were attributed to the mixed mineral phase. Interplanar spacings of 1.51 and 3.41 Å are attributed to 2-line ferrihydrite and maghemite, and those of 1.84, 2.70, and 3.68 Å are attributed to hematite, respectively. The lattice parameter for these magnetosomes was calculated as 8.38 ± 0.08 Å, which is approximately 0.2% off of the lattice parameter of AMB-1 magnetosome magnetite nanocrystals (8.3958 Å) reported by Fischer et al. ^{Reference Fischer, Schmitz, Aichmayer, Fratzl and Faivre67} The lattice parameter for the particles in the transition range is 8.38 ± 0.07 Å. Details of the SAED patterns of the two particles in this transition size range are given in Table SI. SAED analysis of the magnetosomes revealed, in addition to the magnetite, the presence of maghemite, 2-line ferrihydrite, and the occasional hematite in the particles with sizes above 10 nm. Based on the Gibbs Free Energy calculations, we have attributed this result to the oxidation of magnetosome magnetite to maghemite, which could undergo the conversion to hematite under the electron beam. ^{Reference Lagoeiro68} The presence of trace amounts of ferrihydrite in magnetosomes with sizes above 10 nm may also be feasible according to the SAED, as further discussed in the Fe L₃ EELS analysis. Size distribution of magnetosome particles was examined using the cultures sampled at the point of induction, 30 min, and 1 day after the biomineralization induction via the HAADF-STEM imaging. Figure 3 reveals that the sub-10 nm particles were mostly present in the culture collected 30 min after the injection of iron-replete growth medium, which is indicative of this particular time frame including the biomineralization onset. ^{Reference Staniland, Ward, Harrison, van der Laan and Telling43} As the formation of the magnetosomes is a continuous process, the presence of these small particles is expected at all time frames. Even though the samples collected at all three time frames showed magnetosomes with sizes from 0 to 80 nm, the higher sub-10 nm particle content was consistent with the magnetosome formation in the sample collected 30 min after the injection. Moreover, we surveyed size-dependent crystallinity of the magnetosome particles formed during the first 30 min of the induced biomineralization. These results are in good agreement with the reported formation of the nascent 8 ± 4 nm biominerals within the first 10 min after induction, which were identified as a mixture of poorly crystalline iron oxyhydroxide and ferrihydrite, magnetite, and hematite, while magnetite was the major mineral phase 30 min after the induction. ^{Reference Baumgartner, Morin, Menguy, Perez Gonzalez, Widdrat, Cosmidis and Faivre44} However, direct conversion of hematite to magnetite is thermodynamically unfavorable, and under mild conditions oxidation of magnetite proceeds via the formation of maghemite. ^{Reference Lagoeiro68} Here, the detected hematite is likely to form as a result of beam-induced conversion of preoxidized magnetite nanocrystals.

FIG. 3.

Size distribution of magnetosome particles for 0 and 30 min and 1 day after the introduction of iron-replete growth medium shows the highest population of sub-10 nm particles present in the sample collected after 30 min of growth medium. The largest number of magnetosome per bacterium is seen in the sample collected after one day, [n (bacteria for each time frame): 30].

The chemical environment of iron in magnetosomes was probed via Fe L₃ EELS. To determine the nature of minerals formed in this alleged transition range and in the sub-10 nm range, EELS was carried out on both Fe L₃ and O K-edges and compared with that obtained from the magnetosomes with sizes larger than 15 nm. The reference HAADF-STEM image of selected magnetosomes and analysis of Fe L₃ peaks are shown in Fig. 4. The EEL spectra from large magnetosome in Fig. 4(a) (52 nm) marked with green square, and small magnetosome (10 nm) marked with blue circle, are shown in Figs. 4(b) and 4(c), respectively. To mitigate the low signal-to-noise ratio associated with small particles embedded in the bacterial matrix, we used 3-channel intensity smoothing. After that, the acquired Fe L₃-edge spectra were fitted with four different reference peaks corresponding to different oxidation states and coordination numbers of iron. For the large magnetosome, the fitting peaks have positions (and %area under the fitting peaks) at 707.5 eV (24%), 708.5 eV (23%), and 709.3 eV (27%) [Fig. 4(b)], which correspond to octahedral Fe²⁺, tetrahedral Fe³⁺, and octahedral Fe³⁺, and an additional peak at 710.3 eV (25%), which was attributed to FeO(OH). ^{Reference Ochoa, Bello, Sancristóbal, Balsamo, Albornoz and Brito49} Similarly, the experimental spectrum of the sub-10 nm magnetosome was fitted with peaks at 708.2 eV (19%), 708.9 eV (31%), 710.0 eV (33%), and 710.9 eV (16%) corresponding to the same environment as above [Fig. 4(c)]. The shift of approximately 0.7 eV between these two experimental spectra [Figs. 4(b) and 4(c)] was attributed to the instrument drift during the consecutive EELS acquisition. The interpeak positions between first three peaks were around 0.9 eV for both cases. The ratio of peak 1 to peaks (2 + 3), corresponds to the relative concentrations of Fe²⁺ to Fe³⁺, which are 1:2.1 (matching the ratio expected for Fe₃O₄) and 1:3.3 (1:2.9 for n = 2, Fig. S1, Table SII) for these large and small magnetosome particles, respectively. For the particles in the transition region, this ratio is 1:3.4 (n = 2, Fig. S1, Table SII). This indicates that the chemical environment of the iron in amorphous particles is dissimilar to that of iron in magnetite, and tends toward a more ferric-rich situation instead. The relative ratio of the areas under the tetrahedral Fe³⁺ and octahedral Fe³⁺ peaks in the amorphous to crystalline transition range and sub-10 nm range varies, which can be attributed to continuous conversion of Fe³⁺ to Fe²⁺ during the biomineralization. The Fe²⁺ and Fe³⁺ concentrations in Magnetovibrio blakemorei strain MV-1 was described by Kalirai et al. in terms of formed premagnetosomes, magnetosomes, and gaps along the magnetosome chains, and while observed in a different microorganism, this could be indicative of the relevant magnetosome biomineralization pathway. ^{Reference Kalirai, Bazylinski and Hitchcock69} The presence of FeO(OH) peak in this fitting was attributed to the presence of ferric oxyhydroxide, specifically ferrihydrite, which was observed previously both in case of small ^{Reference Baumgartner, Morin, Menguy, Perez Gonzalez, Widdrat, Cosmidis and Faivre44} and large magnetosome nanocrystals. ^{Reference Frankel, Papaefthymiou, Blakemore and O'Brien18} Trace amount of ferrihydrite detected in some of the large particles was attributed to the incomplete conversion of nascent ferrihydrite precursor to magnetite, in addition to the newly formed magnetite. Fe L₃ edge EEL spectrum outside the magnetosome region in the bacterium did not reveal any iron signal in the bacterium, which eliminates the possible presence of iron outside the magnetosome (Fig. S2).

FIG. 4.

(a) HAADF-STEM image of a bacterial cell obtained from the sample 30 min after induction. Magnified images of the magnetosomes in the red rectangle are shown in the insets. Fe L₃ core EEL spectra of the particles shown in the inset 1 with size: 52 nm and in the inset 2 with size: 10 nm are given in (b) and (c), respectively. The ratio of the areas of peak 1 to peaks (2 + 3), corresponding to Fe²⁺ (octahedral) to Fe³⁺ (tetrahedral + octahedral) ratio, is around 1:2.1 for large particle and 1:3.3 for small particle. Peak 4 is attributed to the FeO(OH). [Scale bars: (a): 500 nm, inset 1 and 2: 20 nm.]

Qualitative comparison of the acquired O K-edge EEL spectra of magnetosomes with sizes of 7, 11, 48, and 67 nm in Fig. 5 focused on the evolution of distinct spectral features as a function of the magnetosome size. This analysis involved 10-channel intensity smoothing to mitigate the high noise level of smaller particles. The peaks at 531 eV (blue area) and 540 eV (purple area) correspond to the transition of electrons of O 1s to unoccupied hybridization peaks of O 2p with Fe 3d and Fe 4s/4p, respectively. ^{Reference Bischoff and Motta50} It was suggested by Colliex et al. that the peak around 548 eV (green area) can serve as fingerprint for iron oxides. ^{Reference Colliex, Manoubi and Ortiz45} The presence of this peak indicates that the structure resembles either γ-Fe₂O₃ or Fe₃O₄, while its deficiency indicates that the structure is either α-Fe₂O₃ or FeO. As the size of the magnetosome increases, the initially broad and noisy peaks at 531, 540, and 548 eV appear clearly defined and become better resolved. Moreover, the peak at 540 eV becomes narrower, and the peaks at 531 and 548 eV are more pronounced in larger particles. The presence of characteristic 548 eV peak, along with the peak around 531 eV in larger particles, is in an agreement with the presence of fully developed magnetite structure, while the lack of these two peaks for smaller particles is consistent with the absence of magnetite. Considering the observed ratios of spectral peaks attributed to Fe(II) and Fe(III) discussed above, the smallest magnetosome particles cannot be magnetite, but as explained above, it is also known that ferric oxyhydroxides, specifically ferrihydrite, can be present during the early stages of biomineralization. ^{Reference Baumgartner, Morin, Menguy, Perez Gonzalez, Widdrat, Cosmidis and Faivre44} The O K-edge EEL spectrum of the smallest particle in this work resembles to that of 2-line ferrihydrite, FeOOH·0.4H₂O. ^{Reference Gloter, Zbinden, Guyot, Gaill and Colliex48}

FIG. 5.

O K-edge core EEL spectra of magnetosomes of sizes 7, 11, 48, and 67 nm. The peaks at 531 eV (blue area), 540 eV (purple area), and 548 eV (green area) are the fingerprints for the determination of the type of iron oxide. The spectra become noisier as particle size decreases. The peaks at 531 and 548 eV are more pronounced in large particles. The width of the 540 eV peak decreases as the particle grows.

The structures up to 10 nm are viewed as disordered ferric hydroxide with the Fe³⁺/Fe²⁺ ratio higher than 2.0 and the presence of the peak in Fe L₃ EEL spectra is assigned to FeO(OH). The particles in the 10–15 nm range are partially crystalline and the planes in their SAED analyzed cannot be directly indexed to one specific mineral because they are single crystals distributed in the bacteria with different zone axes parallel to the electron beam, which inhibits visualization of some characteristic planes of these minerals in SAED. Thus, the planes of these minerals are indexed to a combination of 2-line ferrihydrite, magnetite, maghemite, and hematite suggesting the transformation of amorphous ferric hydroxide to crystalline 2-line ferrihydrite and magnetite in this range. In this region, too, the Fe³⁺/Fe²⁺ ratio remained higher than 2.0, and the presence of all these four minerals was detected via SAED. Above 15 nm, through SAED analyses, the structures were again indexed to 2-line ferrihydrite, magnetite, and maghemite, but the Fe³⁺/Fe²⁺ ratio of 2.0 indicates that the structure is mostly magnetite with trace amounts of both 2-line ferrihydrite and maghemite. Furthermore, the O K-edge EEL spectrum of the large particle verifies the magnetite mineral in magnetosomes, while the lack of the high intensity 531 and 548 eV indicates that the structure resembles that of 2-line ferrihydrite for sub-15 nm particles.

Based on the detailed analysis, we outline the following pathway to the induced magnetite magnetosome biomineralization in AMB-1. Upon the exposure of the bacterial culture grown under the low-iron conditions to the iron-replete growth medium, the magnetosome biomineralization proceeds via the formation of small and amorphous ferric oxyhydroxide nanoparticles. These particles grow and transform to more ordered 2-line ferrihydrite and to magnetite, possibly via the intermediate maghemite. While we are unable to follow the evolution of the crystalline structure in a specific particle, we note that the increase in particle size of the biomineral phase with time is strongly correlated with its ordering, and that the underlying phase transformation, from amorphous (ferric hydroxide) to poorly crystalline (ferrihydrite) to ordered, crystalline (maghemite, magnetite), is, most likely, linked to the particle growth. The observed change in the Fe²⁺:Fe³⁺ concentration ratio, from 1:3 in the initial disordered phase to 1:2 in the intermediate ferrihydrite phase, points to the commencing production of ferrous iron during the growth process, and is in excellent agreement with the earlier reports on reproduction of Fe²⁺ and reduction of Fe³⁺, responsible for the formation of magnetite via ferrihydrite. ^{Reference Frankel, Papaefthymiou, Blakemore and O'Brien18}