Magnetosomal matrix: ultrafine structure may template biomineralization of magnetosomes

Authors

  • A. P. Taylor,

    Corresponding author
    1. Department of Microbiology and Parasitology and
    2. Centre for Microscopy and Microanalysis, The University of Queensland, Brisbane, QLD, 4072, Australia
      Dr Tony Taylor, Materials and Engineering Science, Australian Nuclear Science and Technology Organization (ANSTO), PMB 1, Menai, Sydney, NSW, 2234, Australia. E-mail: apt@ansto.gov.au
    Search for more papers by this author
  • J. C. Barry

    1. Centre for Microscopy and Microanalysis, The University of Queensland, Brisbane, QLD, 4072, Australia
    Search for more papers by this author

Dr Tony Taylor, Materials and Engineering Science, Australian Nuclear Science and Technology Organization (ANSTO), PMB 1, Menai, Sydney, NSW, 2234, Australia. E-mail: apt@ansto.gov.au

Summary

The organic matrix surrounding bullet-shaped, cubo-octahedral, D-shaped, irregular arrowhead-shaped, and truncated hexa-octahedral magnetosomes was analysed in a variety of uncultured magnetotactic bacteria. The matrix was examined using low- (80 kV) and intermediate- (400 kV) voltage TEM. It encapsulated magnetosomes in dehydrated cells, ultraviolet-B-irradiated dehydrated cells and stained resin-embedded fixed cells, so the apparent structure of the matrix does not appear to be an artefact of specimen preparation. High-resolution images revealed lattice fringes in the matrix surrounding magnetite and greigite magnetosomes that were aligned with lattice fringes in the encapsulated magnetosomes. In all except one case, the lattice fringes had widths equal to or twice the width of the corresponding lattice fringes in the magnetosomes. The lattice fringes in the matrix were aligned with the {311}, {220}, {331}, {111} and {391} related lattice planes of magnetite and the {222} lattice plane of greigite. An unidentified material, possibly an iron hydroxide, was detected in two immature magnetosomes containing magnetite. The unidentified phase had a structure similar to that of the matrix as it contained {311}, {220} and {111} lattice fringes, which indicates that the matrix acts as a template for the spatially controlled biomineralization of the unidentified phase, which itself transforms into magnetite. The unidentified phase was thus called pre-magnetite. The presence of the magnetosomal matrix explains all of the five properties of the biosignature of the magnetosomal chain proposed previously by Friedmann et al. and supports their claim that some of the magnetite particles in the carbonate globules in the Martian meteorite ALH84001 are biogenic. Two new morphologies of magnetite magnetosomes are also reported here (i.e. tooth-shaped and hexa-octahedral magnetosomes). Tooth-shaped magnetite magnetosomes elongated in the [110] direction are reported, and are distinct from arrowhead-shaped and bullet-shaped magnetosomes. Elongation of magnetite magnetosomes in the [110] direction has not been reported previously. A Martian hexa-octahedral magnetite particle was previously characterized by Thomas-Keptra et al. and compared with truncated hexa-octahedral magnetite magnetosomes. Hexa-octahedral magnetite magnetosomes with the same morphology and similar sizes and axial ratios as those reported by Thomas-Keptra et al. are characterized here. These observations support their claim that ALH84001 contains evidence for a past Martian biota.

Introduction

Magnetotactic bacteria are a phylogenetically diverse group of highly motile aquatic Gram-negative eubacteria that are able to use the Earth's geomagnetic field to navigate their environment. Hypothetically, magnetotaxis confers on magnetotactic bacteria the ability to locate the ideal region within the oxic–anoxic transition zone (OATZ) (Blakemore, 1975; Balkwill et al., 1980; Mann et al., 1990a; Bazylinski et al., 1995; Frankel et al., 1997). Magnetotactic bacteria produce intracytoplasmic magnetic inclusions called magnetosomes, which are composed of magnetite (Fe3O4) and/or greigite (Fe3S4) (Frankel et al., 1983; Mann et al., 1990b; Bazylinski et al., 1995; Pósfai et al., 1998). Little is known about the biomineralization of magnetosomes owing to difficulties associated with the culture of these fastidious microbes. Frankel et al. (1983) used Mössbauer resonance spectroscopy to determine the nature and distribution of major iron compounds in Magnetospirillum magnetotacticum (MS-1). They demonstrated that extracellular chelated ferric iron was taken into the cytoplasm as ferrous iron. The cells also contained low-density hydrous iron oxide and high-density hydrous iron oxide (possibly ferrihydrite) and magnetite. They hypothesized that ferrihydrite is a possible precursor in magnetite biomineralization.

Magnetite magnetosomes appear to be surrounded by a magnetosomal membrane and/or an organic matrix believed to be responsible for its biomineralization (Balkwill et al., 1980; Mann et al., 1986; Gorby et al., 1988; Vali & Kirschvink, 1990; Spring et al., 1998; Taylor & Barry, 2000). Gorby et al. (1988) reported an organic matrix surrounding the magnetosomes in MS-1, which was removed with repeated washing in 1.0 m NaCl solution. They reported evidence for a magnetosomal membrane composed of neutral lipids, free fatty acids, glycolipids, sulpholipids, phosphatides and phospholipids. They used two-dimensional gel-electrophoresis to separate two proteins that were unique to the magnetosomal membrane fraction and proposed that the proteins were possibly involved in magnetite biomineralization.

Meldrum et al. (1993) also reported an electron-dense matrix surrounding magnetosomes. Mann & Frankel (1989) reported the coexistence of crystalline and non-crystalline phases in immature magnetite magnetosomes in MS-1. Many of these magnetosomes had outgrowths that appeared to extend into the surrounding material. Taylor & Barry (2000) briefly described the encapsulation of magnetosomes in matrices with lattice fringes aligned with planes in the magnetosomes.

Magnetosomes often sediment and fossilize to become magnetofossils (Kirschvink & Chang, 1984; Chang & Kirschvink, 1989; Chang et al., 1989; Stolz et al., 1989a,b; Vali & Kirschvink, 1989). When they fossilize in carbonates, they are surrounded by an electron-dense halo 10–40 nm thick (McNeill, 1990). Friedmann et al. (2001b) hypothesized that the halo was a fossilized magnetosomal membrane. They also identified five properties of the magnetosomal chain that constituted a biosignature, namely: uniform crystal size and shape; gaps between crystals; orientation of elongation along the chain axis; flexibility of the chain; and a halo surrounding the chain.

In this article an organic matrix surrounding magnetosomes was studied. Its structure and intimate association with the encapsulated magnetosomes and the presence of an unidentified phase in magnetite-containing immature magnetosomes indicated that the magnetosomal matrix may act as a template for the spatially organized precipitation of the unidentified phase, which later transforms into magnetite.

Materials and methods

Specimen collection, enrichment and harvest

Freshwater specimens were collected from the surface sediments at the OATZ and hypolymnion of an irrigation pond at Howeston Golf Course Pty. Ltd. in south-eastern Queensland, Australia. Marine specimens were collected from approximately 30 cm above the low-water mark at a seagrass bed at Wellington Point in Moreton Bay, south-eastern Queensland. Specimens were collected in enrichment vessels made from 2.0-L polyethylene terephthalate (PETE) Blowpac™ soft-drink bottles with their tops removed, to yield 1.8-L enrichment vessels. They were enriched by incubation in dim natural light at room temperature (22–28 °C) for 3 days. Magnetotactic bacteria were harvested using the modified version of the capillary-racetrack technique (Taylor et al., 2000; Taylor, 2002).

Specimen preparation for TEM

Whole-cell preparations.  Dehydrated whole-cell specimens were made using the method described in Taylor et al. (2000) and were either examined directly in a TEM or ultraviolet-B-irradiated (UV-B-irradiated) for 24 h (Taylor et al., 2000) prior to analysis. Specimens in Fig. 1(A,B,E–G,J,L) and 4 were not UV-B-irradiated and all other specimens were.

Figure 1.

(A) Transmission electron micrograph (80 kV) of an unstained marine rod CRM1 with two contiguous chains of slightly elongated cubo-octahedral magnetosomes (white arrowhead). Note the electron-dense cytoplasmic inclusions (i) and one of its sheathed polar flagella (black arrowhead). (B) High magnification of the magnetosomes in A. Note the electron-dense region surrounding the magnetosomes (arrowhead). (C) Note the S-layer (s) on the cellular envelope. (D) An unstained brackish coccus CCB1, with a single chain of cubo-octahedral magnetite magnetosomes (400 kV). Note the electron-transparent inclusions (i), polar organelles (p) and the electron-dense region surrounding the chain of magnetosomes (arrowhead). (E) A chain of cubo-octahedral magnetosomes in an unstained marine vibrio CVM1 (80 kV). Note the electron-dense region surrounding the chain (arrowhead). (F) An unstained freshwater vibrio DVF1, with a single chain of D-shaped magnetosomes (400 kV). Note the electron-dense region surrounding the chain (arrowhead). (G) A magnetosomal chain in an unstained freshwater vibrio FVF1 (80 kV). Note the irregular arrowhead-shaped projections of the mature magnetosomes and the electron-dense region surrounding them (arrowhead). (H) An unstained marine magnetotactic rod HRM1, with a single chain of truncated hexa-octahedral magnetite magnetosomes aligned with the long axis of the cell (400 kV). Note the decreased electron density (halo) in the region surrounding the magnetosomal chain (arrowhead) and how the electron-transparent inclusions (i) bend the chain. (I) High magnification of the magnetosomes and their surrounding matrix. (J) Transmission electron micrograph (80 kV) of a stained ultrasection of a marine coccus HCM9, with a chain of truncated hexa-octahedral magnetosomes. Note the concentration of stain around the magnetosomes (mm) owing to the magnetosomal matrix and the zigzagged arrangement of the magnetosomes, which yielded a linear chain. (K) A stained ultrasection of a marine magnetotactic coccobacillus HCM12 (80 kV). Note the Gram-negative cellular envelope (ce) with an outer membrane, a periplasmic space and a cytoplasmic membrane. Also note the concentration of stain around the magnetosomal chain (mm). (L) A stained ultrasection of a marine coccus HCM5, showing one of its two chains of truncated hexa-octahedral magnetosomes. Note the Gram-negative cellular envelope with a cytoplasmic membrane (cm), an undulating outer membrane (om) and a periplasmic space (p). Also note the intermagnetosomal space between the parallel faces of adjacent magnetosomes, which were as small as 2.9 nm wide.

Figure 4.

(A) Phase-contrast lattice image of a magnetosome in an unstained freshwater rod BRF1, viewed along its [inline image12] axis (400 kV). Note the discontinuity of the {220} lattice fringes in the narrow (growing) end of the magnetosome and the kinks (arrowheads) in the structure. (B) SAED pattern of the magnetosome in A. Note the elevated intensities of the {222}, {440} and {402} related reflections. (C) Enlarged detail of the {inline imageinline image2} reflection in B. Note the streaking in the {13inline image}*, {inline imageinline imageinline image}* and {inline image0inline image}* directions, which was possibly due to the magnetosomal matrix. (D) Digitally enhanced image of the magnetosomal matrix of the area highlighted in A (box). Note the lattice fringes P and Q aligned with the {13inline image} and {220} planes of the magnetite in A. (E) Filtered Fourier transform of the highlighted area in A.

Ultrasections.  Magnetic concentrates of magnetotactic bacteria were further concentrated in 2.0–20-µL micropipette tips with a low-melting-temperature agarose plug at their narrow ends (Taylor, 2002). Cells were fixed in 3.0% glutaraldehyde in 0.1% cacodylate buffer at 4 °C overnight. Ultrasections were made by substitution of water with acetone and then substitution of the acetone with Epon™ resin, which was polymerized at 60 °C overnight. Ultrasections 60–70 nm thick were cut with a Leica™ UltracutE microtome with a Diatome™ diamond knife. Ultrasections were adhered to 200-mesh Cu grids primed with 2.0% 3-aminopropyltriethoxy silane (APTES) and stained with uranyl acetate and lead citrate.

TEM

Stained and unstained whole-cell preparations were examined in a Hitachi™ H800 STEM, a JEOL™ JEM 1010 TEM and a JEOL™ JEM 1210 high-resolution TEM. ATEM was performed on unstained whole-cell preparations in a JEOL™ 4000 FX high-resolution TEM fitted with an Oxford-Link™ windowless Si X-ray detector. The instrument was operated at 400 kV with a constant objective lens voltage of 6.79 V with 1.0-Ω resistance. A gold standard was used to calculate the magnification.

A large (mature) magnetosome in one BRF1 cell was used as a stochiometric magnetite standard orientated along a <211> axis for determining the O/Fe ratio in a smaller (immature) magnetosome, also probed along its <211> axis within the same cell.

Digitization and digital analysis

Negatives were digitized using a Leafscan 45™ negative scanner and Adobe Photoshop™ software. Regions of images were digitally enhanced using NIH Image Analysis™ software. All lattice fringes in the digitally enhanced images presented here are also clear in the original images.

Results

Magnetosomal matrix

Low-magnification images of magnetotactic bacteria at low (80 kV) and intermediate (400 kV) voltages revealed a region of contrast surrounding magnetosomes in many strains of magnetotactic bacteria (Fig. 1). The region of contrast was most easily imaged at 80 kV (Fig. 1A,B,F,G). However, it was also detected at 400 kV (Fig. 1D,H,I). In most cases the contrasted region (magnetosomal matrix) was imaged as a slightly electron-dense region contrasted with a relatively electron-transparent cytoplasm or inclusion (Fig. 1B,D–G). In the strain HRM1, the magnetosomal matrix appeared as an electron-transparent halo contrasted on an electron-dense cytoplasm (Fig. 1H,I). This may have been due to the matrix excluding electron-dense precipitates during sample dehydration.

Stained ultrasections of magnetotactic bacteria also contained evidence of a magnetosomal matrix. An anionic matrix surrounded the magnetosomes in several magnetotactic bacteria, which concentrated cationic stains (Fig. 1J–L). The anionic matrices had widths that were only slightly wider than those recorded in unstained dehydrated whole-cell preparations (Table 1), which was possibly the result of the elevated contrast from the stain.

Table 1.  Data from low-magnification images of the magnetosomal matrix surrounding various maganetosomes reported in this study.
StrainMagnetosome*ReferencePreparationThickness (nm)
  • *

    Cubo-octahedral magnetite (C), D-shaped magnetite (D), irregular arrowhead-shaped magnetite (F), truncated hexa-octahedral magnetite (H).

  • Data not reported (DNR).

  • Dehydrated cell (DC), stained ultrasection (SU).

CRM1CFig. 1(A,B)DC30–40
CCB1CFig. 1(D)DC40–50
CVM1CFig. 1(E)DC40–50
DVF1DFig. 1(F)DC20–30
FVF1FFig. 1(G)DC30–40
HRM1HFig. 1(H,I)DC20–30
HRM3HDNRDC20–30
HCM5HDNRDC20–30
HCM9HFig. 1(J)SU50–70
HCM12HFig. 1(K)SU30–40
HRF1HDNRDC40–50

High-resolution images revealed magnetosomes surrounded by a semicrystalline matrix with lattice fringes aligned with lattice planes in the encapsulated magnetosome (Figs 2–4 and Tables 1 and 2). With only one exception, the lattice fringes had widths that were approximately equal to or twice the width of the corresponding lattice plane in the magnetosome (Table 2). An electron-dense region or halo was detected surrounding magnetosomes in many strains of magnetotactic bacteria, including: BRF1, CRM1, CCB1, DVF1, FVF1, GMB1, HRM1, HCM3, HCM5 and HRF1 (Tables 1 and 2). The matrix surrounded truncated hexa-octahedral magnetite magnetosomes in HRM1, HRM3, HCM3, HCM5, HCM9, HCM12 and HRF1 cells extended 20–70 nm beyond their surfaces (Tables 1 and 2). The matrix was ∼50 nm thick around bullet-shaped magnetosomes in the BRF1 cell (Fig. 4). The matrix surrounded cubo-octahedral magnetosomes in CRM1, CVM1 and CCB1 cells and was 30–50 nm thick. D-shaped magnetosomes in DVF1 cells were surrounded by a matrix 20–30 nm thick, and irregular arrowhead-shaped magnetosomes in FVF1 cells were surrounded by a matrix 30–40 nm thick. Feint contrasted regions were detected around the magnetosomes in many other strains of magnetotactic bacteria, but were not obvious. At least 11 of the 80 strains of magnetotactic bacteria detected in the Moreton-Bay region so far had strong evidence of a magnetosomal matrix and another 10 strains had subtle evidence of one. Furthermore, the matrix was detected in high-resolution images of another three strains of magnetotactic bacteria, where it was not detected at low magnification. The matrix was also detected in stained ultrasections of several strains of magnetotactic bacteria (Fig. 1J–L). Given the low level of contrast produced by the matrix and the high electron density of the cytoplasms of many strains of magnetotactic bacteria, it is likely that the majority of, if not all, magnetotactic bacteria have a magnetosomal matrix encapsulating their magnetosomes, but it is not always detectable using TEM.

Figure 2.

(A) Phase-contrast lattice image of a magnetosome in an HCM3 cell viewed along its [10inline image] axis (400 kV). (B) High magnification of the lattice in the magnetosome in A. Note the {020} and {311} lattice fringes. (C) SAED pattern of the magnetosome in A. (D) Enlarged detail of the {3inline image1} reflection in C. Note the streaking in the {3inline image1}* direction, which may have been caused by the magnetosomal matrix (see Fig. 3).

Figure 3.

(A) Selected area of the image in Fig. 2(A). (B) Fourier transform of the image in (A). (C) Digitally enhanced image of (A). Note the lattice fringes in the matrix X, Y and Z parallel to the {311}, {3inline image1} and {inline imageinline imageinline image} planes in the magnetosome. (D) Filtered Fourier transform of the image in (A).

Table 2.  Data from high-magnification images of the magnetosomal matrix surrounding various maganetosomes reported in this study.
StrainMagnetosome*ReferenceAxisOrientation§d valueCorresponding d**d ratio††
  • *

    Bullet-shaped magnetite (B), cubo-octahedral magnetite (C), greigite (G), truncated hexa-octahedral magnetite (H).

  • Data not reported (DNR).

  • Axis of the magnetosome.

  • §

    Orientation of the lattice fringe in the magnetosomal matrix relative to the magnetosome.

  • Width of the lattice fringe in the magnetosomal matrix. d value not determined (ND).

  • **

    d value of the corresponding lattice plane in magnetite.

  • ††

    Ratio of the d value of the lattice fringe in the magnetosomal matrix to the d value of the corresponding lattice plane in magnetite.

BRF1BFig. 4[inline image12]{13inline image}*{220}*ND ND2.532 Å 2.967 ÅND ND
CCB1CDNR[0inline image1]{1inline imageinline image}*ND4.852 ÅND
GMB1GDNR[inline image12]{inline image2inline image}*ND2.4243 ÅND
HRM1HDNR[2inline image1]{13inline image}*ND2.532 ÅND
HCM3HFig. 2[33inline image]{13inline image}*4.0 Å1.9262 Å2.08
HFigs 3 and 4[103]{311}*{3inline image1}*{inline imageinline image1}*2.4 Å 3.9 Å 3.0 Å2.532 Å 1.9262 Å 0.880 Å0.95 2.02 3.41
HDNR[1inline image4]{220}*ND2.967 ÅND
HRF1HFig. 5[01inline image]{inline image11}*4.9 Å4.852 Å1.01
HFig. 6[01inline image]{3inline imageinline image}*{111}*5.1 Å 5.3 Å2.532 Å 4.852 Å2.01 1.09

Lattice fringes were detected in the magnetosomal matrix in unstained dehydrated whole-cell preparations of magnetotactic bacteria that had been UV-B-irradiated (Figs 2 and 3) and those that had not (Fig. 4), which demonstrates that the matrix was not an artefact of UV-B-irradiation. Moiré fringes were generated where the magnetosomal matrix overlayed the lattice of one magnetite magnetosome in an HCM3 cell (Table 2). This demonstrates that the lattice in the background of the image was not caused by astigmatism or specimen drift, so was caused by diffraction. Lattice fringes orientated in three directions were recorded surrounding the magnetite magnetosome in Figs 2 and 3, which also demonstrated that the lattice fringes were not produced by astigmatism or specimen drift. Lattice fringes orientated in multiple directions were recorded in the magnetosomal matrices in BRF1, HCM3 and HRF1 cells (Figs 2–4 and Table 2), which clearly could not have been generated by astigmatism or specimen drift alone. Furthermore, the d values and angles between lattice planes in all magnetosomes reported here were within 1.0% of the values determined for these materials using X-ray diffraction (Skinner et al., 1964; International Centre for Diffraction Data, 1992), so we do not believe that the lattice fringes in the magnetosomal matrix are due to astigmatism or specimen drift.

It is possible that the lattice fringes in the matrix are the product of beam damage, but it seams unlikely if a magnetosomal membrane separates the magnetosome's surface from the matrix. It is more likely if there is no membrane. All beam-damage phenomena so far reported in magnetosomes were in specimens that were not stabilized with UV-B (Taylor et al., 2001; Taylor, 2002). In only one case was the support film affected (melted). This was probably caused by the production of water plasma from ice in the large improperly dehydrated sample. All specimens reported here were properly dehydrated and many were UV-B stabilized.

Pre-magnetite

The narrow end of the bullet-shaped magnetosome in Fig. 4 is of low contrast. The contrast varies across the magnetosome in a continuum in conjunction with the extinction of lattice fringes. The wide end was composed of magnetite, but only {1inline image1} lattice fringes could be seen in the narrow end in Fig. 4. Another image taken along the same axis contains only {220} and {13inline image1} latt} lattice fringes in the narrow end and lacks {1inline image1} latt1} lattice fringes, whereas lattice fringes orientated in multiple directions were imaged in the wide end composed of magnetite. The selected-area electron diffraction (SAED) pattern of the immature magnetosome (Fig. 4B) has no evidence of lattice twist in the form of superperiodicity of reflections and the lattice image has no evidence of lattice strain or boundary planes, so it is unlikely that the lattice was twisted. From this, the extinction of lattice fringes and low electron density of the narrow end is most likely due to it being composed of a different material.

In a personal communication, Mihaly Pósfai proposed that the missing fringes could be absent as a result of a change in thickness or slight disorientation of the two parts of the magnetosome. We have also observed the absence of lattice fringes owing to changes in thickness. However, this only occurred in extremely thin regions of crystals ( 12 nm), whereas lattice fringes became extinct in the crystal in Fig. 4(A) where it was approximately 35 nm thick, assuming that it was symmetrical around the long axis. In every case that we have observed where lattice fringes became extinct in magnetite due to tilting away from the zone axis, more closely spaced lattice fringes disappeared before the wider fringes. For {111} fringes to be absent while {311} and {220} fringes were present contradicts this trend.

X-ray spectra acquired from a BRF1 cell (Fig. 5) had small Okα1,2 peaks, owing to the accumulation of ice on the X-ray detector. Comparison of X-ray spectra acquired from the immature magnetosome in Fig. 4(A) with a mature magnetosome from the same cell (figure 3F of Taylor et al., 2001) taken under the same conditions demonstrated that they had different O/Fe ratios. Using the mature magnetite magnetosome as a stoichiometric magnetite standard (O/Fe = 1.3), the immature magnetosome had an O/Fe ratio of 1.6. Assuming that the immature magnetosome was composed of equal quantities of magnetite and the unidentified phase, and that the wide end was composed of stoichiometric magnetite only, the narrow end was composed of a material with a significantly larger O/Fe ratio of ∼2.0, consistent with an iron hydroxide or a hydrous iron oxide. This is similar to the hypothesis of Frankel et al. (1983), who proposed that the precursor in magnetite biomineralization was ferric (oxy)hydroxide (ferrihydrite). The plots of length versus width for bullet-shaped magnetosomes indicate that they grow unidirectionally at the narrow end, so the unidentified phase that constitutes the narrow end is most likely the precursor in the biomineralization of magnetite.

Figure 5.

X-ray spectra acquired for 100 s each from a BRF1 cell (400 kV). (A) Spectrum acquired from the mature magnetosome in figure 3F in Taylor et al. (2001). (B) Spectrum acquired from the immature magnetosome in Fig. 4(A). (C) Spectrum acquired from the cytoplasm. The Cu peaks in all spectra were from the Cu grid.

The immature tooth-shaped magnetosome in Fig. 6 was also composed of magnetite and an unidentified phase. The unidentified phase encapsulated a magnetite core. The core was clearly composed of magnetite, whereas the outer crystal was composed of a less electron-dense and less crystalline material, diffracting in the {inline imageinline image1} latt1} and {0inline image2} lattice planes (Fig. 6B,E). There was a region of intermediate contrast on the (1inline image1} latt1) and (inline imageinline image1) faces of the magnetite core, which contained {inline imageinline image1} lattice fringes as well (Fig. 6A, arrow heads). This may have been a third phase, which was a transition phase in the solid-state transformation of the encapsulating phase into magnetite.

Figure 6.

Figure 6.

(A) Phase-contrast lattice image of a tooth-shaped magnetosome in a freshwater vibrio TVF1, imaged along the [011] axis (400 kV). Note the two crystals (1 and 2), the jagged boundary between them, composed of (21inline image) (1inline image1) (inline imageinline image1) and (inline image00) faces and the region of intermediate contrast on the (1inline image1) and (inline imageinline image1) faces (arrowheads). (B) High magnification of the lattice image in A (boxed area). Note the alignment of the {0inline image2} lattice fringes in both crystals and the lack of lattice strain associated with the boundary. (C) SAED pattern of the magnetosome in A. Note the elevated intensities of the {311}, {022}, {333} and {600} related reflections, which was possibly due to crystal 2 being composed of pre-magnetite. Also note the extra reflections from the adjacent magnetosome, orientated near the [inline image1inline image] axis. (D) Enlarged detail of the {0inline image2} reflection. Note the slight streaking in the {inline image00}* direction. (E) Phase-contrast lattice image of the magnetosome in A tilted less than 1° relative to the axis of orientation in A. Note the faint {inline image1inline image} and {0inline image2} lattice fringes in crystal 2 and their perfect alignment with the same lattice planes in crystal 1.

Figure 6.

Figure 6.

(A) Phase-contrast lattice image of a tooth-shaped magnetosome in a freshwater vibrio TVF1, imaged along the [011] axis (400 kV). Note the two crystals (1 and 2), the jagged boundary between them, composed of (21inline image) (1inline image1) (inline imageinline image1) and (inline image00) faces and the region of intermediate contrast on the (1inline image1) and (inline imageinline image1) faces (arrowheads). (B) High magnification of the lattice image in A (boxed area). Note the alignment of the {0inline image2} lattice fringes in both crystals and the lack of lattice strain associated with the boundary. (C) SAED pattern of the magnetosome in A. Note the elevated intensities of the {311}, {022}, {333} and {600} related reflections, which was possibly due to crystal 2 being composed of pre-magnetite. Also note the extra reflections from the adjacent magnetosome, orientated near the [inline image1inline image] axis. (D) Enlarged detail of the {0inline image2} reflection. Note the slight streaking in the {inline image00}* direction. (E) Phase-contrast lattice image of the magnetosome in A tilted less than 1° relative to the axis of orientation in A. Note the faint {inline image1inline image} and {0inline image2} lattice fringes in crystal 2 and their perfect alignment with the same lattice planes in crystal 1.

We dismiss the hypothesis of Mihaly Pósfai (personal communication) that the magnetosome in Fig. 6 was composed of a small crystal overlaying a second larger one, on the grounds that the lattice fringes of both crystals were in focus simultaneously through many focal planes, yet the focal plane at 400 kV is very narrow; similar tooth-shaped magnetosomes with electron-dense cores surrounded by a relatively electron-transparent material were imaged at low voltage and magnification in other vibrios (curved rods). Slight tilting of the crystal revealed no sign of the proposed overlap (Fig. 6E).

Because the unidentified phase was only detected in immature magnetosomes and, in particular, at the narrow end of the immature bullet-shaped magnetosome (Fig. 4), it is most likely the precursor in the biomineralization of magnetite, so in this paper it is termed pre-magnetite. This is consistent with the results of Frankel et al. (1983), who predicted that high-density hydrous iron oxide (ferrihydrite) was the precursor in magnetite biomineralization. The structure of the unidentified phase is inconsistent with that of ferrihydrite and the d values of the material only match those of magnetite and maghemite (γ-Fe2O3) and no other iron oxide or hydroxide. The structure of the material is inconsistent with that of maghemite as the SAED pattern imaged along the [110] axis (Fig. 6C) lacks the {100} and {110} reflections reported by Banfield et al. (1994). It is also inconsistent with the structure of magnetite as indicated by the extinction of lattice fringes that were present in all other magnetite magnetosomes imaged along the same axes and by direct comparison with the magnetite components of the respective magnetosomes in the two immature magnetosomes (Figs 4 and 6).

Tooth-shaped magnetite magnetosomes

The tooth-shaped magnetosomes in TVF1 (Fig. 6) cells were elongated in the [110] direction. Tooth-shaped magnetosomes were not imaged along their long axes, so the morphology of the crystals is uncertain. However, elongation in the [110] direction has not been reported in other magnetite magnetosomes, so these tooth-shaped magnetosomes have different morphologies to arrowhead-shaped and bullet-shaped magnetosomes, which are elongated in the [100] and [211] directions, respectively (Mann et al., 1987; Vali & Kirschvink, 1990). Blakemore et al. (1989) reported a morphologically similar magnetotactic vibrio with tooth-shaped magnetosomes. However, they did not analyse the structure of the magnetosomes using ATEM.

Hexa-octahedral magnetite magnetosomes

Truncated hexa-octahedral magnetite magnetosomes have been reported in terrestrial magnetotactic bacteria (Mann et al., 1990a; Buseck et al., 2001; Thomas-Keptra et al., 2001; Weyland et al., 2002), but hexa-octahedral ones have not. Immature and mature hexa-octahedral magnetite magnetosomes are produced by HSM4 cells (Fig. 7; Taylor, 2002). Note the lack of {110} related truncations on the ends of the magnetosome. This magnetosome has the same morphology and approximately the same size and axial ratio as the Martian magnetite particle reported by Thomas-Keptra et al. (2001). The {1inline image1} and {inline image1inline image} faces of the magnetosome appear to be covered with an unidentified phase, which appeared as lattice fringes orientated parallel to the {02inline image} plane of the magnetite, extending ∼5.1 Å beyond its surface. This observation is consistent with the observed structure of pre-magnetite. However, these fringes may have been the result of a thickness effect. Because the immature magnetosome was in the final stage of growth, where magnetosomes from HSM4 cells widen via growth of {111} related faces on the sides of the magnetosome (Taylor, 2002), it is likely that the lattice fringes were from pre-magnetite.

Figure 7.

(A) Phase-contrast lattice image of an immature hexa-octahedral magnetite magnetosome imaged along the [011] axis in a HSM4 cell (400 kV). Note the {111} and {100} related faces and the lack of {110} related truncations on the ends. (B) Selected-area electron diffraction pattern of the magnetosome in A. (C) High magnification of the {inline image00} reflection in the diffraction pattern in B. Note the slight elongation in the {inline image00} direction, which was interpreted to be from slight disorder in the (02inline image) lattice plane, possibly from pre-magnetite growing on the (1inline image1) and (inline image1inline image) faces (see following). (D) Fourier transformation of a selected area of the (1inline image1) face of the magnetosome in A (boxed area). (E) Filtered Fourier transform of the selected area in A. (F) Digitally enhanced image of the selected area in A. Note the {02inline image} lattice fringes that continue past the (1inline image1) face into the background matrix. These lattice fringes were possibly due to pre-magnetite growing on the face.

Discussion

Magnetosomal matrix: possible role in the biomineralization of magnetosomes

Lattice fringes in the magnetosomal matrix parallel to the {311} magnetite were most frequently imaged. {220}, {111} and {331} lattice fringes were next most frequent and {inline imageinline imageinline image} lattice fringes were only imaged once (Table 2). The hypothetical material pre-magnetite has {311}, {220} and {111} lattice fringes (Figs 4, 6 and 7), which is consistent with the structural data obtained from the magnetosomal matrix. This indicates that the magnetosomal matrix may act as a template for the precipitation of pre-magnetite, which transforms into magnetite, possibly via an intermediate (Fig. 6A, arrowheads). The magnetosomal matrix may be a semicrystaline gel (possibly a polysaccharide) and may control the chemiosmotic properties of the solution surrounding the growing crystal (Fig. 8). For the magnetosomal matrix to have survived the dehydration process with some of its structure intact, it must have been relatively anhydrous in the first place. It may control the infiltration of Fe2+ via chelation and/or provide an oxidizing environment via the removal of H+ by the electron transport system (Fig. 8), which was always near to the magnetosomal chain and its surrounding matrix in ultrasections (e.g. Fig. 1J–L). The magnetosomal matrix may explain the structural integrity of magnetosomal chains and the ultrafine-structural coordination between adjacent magnetosomes (e.g. mirroring seen in Fig. 1J). The d values for the {200} lattice plane of greigite and the {111} lattice plane of magnetite are 4.94 Å and 4.85 Å, respectively, and when these minerals are co-precipitated and co-organized in the same magnetosomal chain, as reported by Bazylinski et al. (1995), these lattice planes are parallel. This is consistent with a magnetosomal matrix being the template for both magnetite and greigite.

Figure 8.

Schematic diagram of a hypothetical mechanism for the controlled biomineralization of a chain of magnetite magnetosomes. The electron transport system is coupled to the magnetosomal matrix to make oxidizing conditions of pH and Eh within it, via a controlled eflux of protons. The matrix is relatively impermeable to the flow of protons perpendicular to its (111) lattice plane, which enables the oxidizing conditions to be confined to the region where the magnetosome is growing. The intermagnetosomal spaces and regions where magnetosomes are mature are in chemical equilibrium with the cytoplasm. By arranging these spaces at regular intervals relative to the cell wall and by regulating the activity of the electron transport system(s) within specific regions, chains of crystals with a narrow size range and regular intermagnetosomal spaces may be formed. NAD+ may be able to permeate the 5.9-Å-thick (111) lattice planes of the matrix to act as a reducing agent to precipitate pre-magnetite. This function could also be conducted directly by the matrix itself or by the excretion of H2. Fe2+ may also be able to permeate parallel to the (111) plane via co-ordination with the matrix. The continuous removal of protons from the matrix would promote the influx of cations (e.g. Fe2+). The proton efflux is coupled to ATP generation. The stoichiometric equation for the biomineralization of magnetite by this hypothetical reaction sequence is: 3Fe2+ + 3H2O + 0.5O2 + 3ADP + 3Pi → Fe3O4 + 6H+ + 3ATP. Solutes are displayed in black and solids in white.

Meldrum et al. (1993, their figure 5b) reported a magnetosomal matrix ∼12 nm thick surrounding truncated hexa-octahedral magnetosomes. Sparks (1990) reported an organic matrix associated with cubo-octahedral magnetite extracted from ethmoid tissue from sockeye salmon that was 10–15 nm thick. Gorby et al. (1988) reported organic matter surrounding magnetosomes from MS-1 that was removed by extensive washing in 1 m NaCl. In a personal communication, Mihaly Pósfai reported an accumulation of iron in the region surrounding magnetite magnetosomes. Spring et al. (1998) reported a magnetosomal membrane 20–50 nm thick, surrounding truncated hexa-octahedral magnetosomes in cells analogous to the HCM12 cells studied here (Fig. 1K). The magnetosomes in HCM12 cells had intermagnetosomal spaces (ISs) ∼3–4 nm wide (e.g. Fig. 1K) or 5–6 nm wide (Farina et al., 1994; Spring et al., 1998), which is too narrow for two membranes each 5.7 nm thick, as described by Gorby et al. (1988) in MS-1 cells. In some cells the ISs were consistently 1.8–3.0 nm wide (e.g. Fig. 1L), which is also too small to contain two magnetosomal membranes. There was a contrast variation between the {111} faces of the adjacent magnetosomes in Fig. 1(L). This may have been evidence of two extremely thin membranes ∼1.4 nm thick. However, the specimen was quite thick and the voltage was low (80 kV), so the contrast was most likely the result of Fresnel fringes. It is unlikely that a lipid bi-layer composed of molecules only 0.20–0.25 of the length of those in other biological membranes could be stable in an aqueous environment at ambient temperatures. The ISs in most of the strains of magnetotactic bacteria studied here were < 6.0 nm thick, so they probably did not possess a magnetosomal membrane. The most common strains studied in the Moreton-Bay region (Taylor et al., 2001) that produce truncated hexa-octahedral magnetite magnetosomes had very narrow ISs (2–4 nm).

Electron spectroscopic imaging is a good technique for selectively enhancing contrast according to the elements present. The images of magnetotactic bacteria reported by Lins et al. (2000) contain numerous examples of the magnetosomal matrix surrounding magnetosomes in cells with electron-dense cytoplasms, which prevented clear imaging of their magnetosomes without filtration. The magnetosomal matrix appeared as a region of low contrast (halo) surrounding the magnetosomes, which was very clear in figures 6 and 8 of their work. The cells in these two images resemble the strains HCM12, HCM7 and HCM14 (Fig. 1K; Taylor, 2002) and had magnetosomal matrices ∼15, 20 and 55 nm wide, respectively. The low voltages that they used also enhanced the contrast of the images, so their techniques (Lins & Farina, 1998) should be used more extensively to determine the presence of the matrix throughout magnetotactic bacteria.

Taylor (2002) reports two types of spatial arrangements of cubo-octahedral magnetite magnetosomes in many strains of uncultured magnetotactic bacteria. The first type of chain development has magnetosomes nucleating and growing in predefined positions relative to the cell ultrastrusture, where small magnetosomes have large intermagnetosomal spaces and large magnetosomes have small intermagnetosomal spaces, and the spacing between the geometrical centres of adjacent magnetosomes (magnetosomal displacement) are relatively constant. The magnetosomes appear to grow into the spaces provided for them. The second type of magnetosomal chain development has magnetosomes nucleating at the ends of the chain with a small constant IS, and grows to nearly maximum length before the next magnetosome is nucleated. The magnetosomal displacement increases with the average length of the two adjacent magnetosomes. The first method is consistent with the magnetosomal membrane theory and the second is consistent with the magnetosomal matrix theory (Fig. 8).

The spatial transformation of the magnetosomal matrix during magnetosomal development may be accounted for by two hypothetical mechanisms. The first involves the detachment of the matrix components (monomers) from the matrix at the surface of the crystal and their migration to the outer surface of the matrix, where they rejoin the matrix. This could be achieved by coupling the oxidation of Fe2+ in the precipitation of pre-magnetite to the reductive ligation of the monomer, which is re-oxidized at the surface of the matrix during repolymerization. This last step may be catalysed by NAD+ and be coupled to the electron transport system. The second involves the matrix moving to expose a region of a facet to promote the precipitation of an island, which grows across the face as the matrix moves outward (Fig. 8).

In Fig. 8 the matrix is attached to the cytoplasmic membrane (CM) to prevent the efflux of protons from the IS. The elongation of magnetosomes may be controlled by regulating the attachment and detachment of the matrix to/from the CM. Zones of the matrix could be detached during crystal growth and re-attached at maturity. This may explain how twinned arrowhead-shaped magnetosomes could develop into a crystal elongated in the [100] direction despite being twinned in a {111} plane, as described in Taylor et al. (2001).

Magnetosomal matrix: possible role in the fossilization of magnetosomes

The magnetosomal matrix could be composed of polymerized tricarboxylic acids or siderophores co-ordinating Fe2+. Once the magnetotactic cell dies, and is placed in a basic environment containing Ca2+, the organic component of the matrix may become hydroxylated and transform into siderite. The spatial arrangement of the matrix would most likely provide a template for the precipitation of the siderite such that its structure was aligned with that of the encapsulated magnetosome.

The biomineralization of magnetosomes within a magnetosomal matrix is consistent with the observations of electron-dense regions surrounding magnetofossils in carbonates, as reported by McNeill (1990). The electron-dense regions were 10–40 nm thick and surrounded arrowhead-shaped, cubo-octahedral and truncated hexa-octahedral magnetite particles with characteristic sizes, shapes and axial ratios of magnetite magnetosomes. The electron-dense regions surrounding the magnetofossils were narrower than the magnetosomal matrices observed here. This may be due to corrosion during sedimentation and/or fossilization. However, the electron-dense regions surrounding the magnetofossils were too wide to be artefacts of membranes.

Friedmann et al. (2001b) reported electron-dense regions surrounding chains of magnetofossils in carbonate globules in the Martian meteorite ALH84001. The electron-dense regions were consistent with the magnetosomal matrices reported here and the electron-dense regions surrounding terrestrial magnetofossils reported by McNeill (1990), so were most likely fossilized magnetosomal matrices. The magnetosomal-matrix theory provides an explanation for all five properties proposed by Friedman et al. (2001b) for the biosignature of the magnetosomal chain, whereas the magnetosomal-membrane theory does not. There are only two ways in which terrestrial magnetotactic organisms can deviate from the biosignature proposed by Friedmann et al. (2001b), namely: D-shaped magnetosomes in DVF1 cells are elongated perpendicular to the chain (Fig. 1E; Taylor et al., 2001; Taylor, 2002); and some magnetotactic rod-shaped bacteria produce chains of magnetosomes with a diverse mixture of crystal morphologies (Vali & Kirschvink, 1990; Bazylinski et al., 1995). However, these deviations do not invalidate the biosignature. Furthermore, consecutive type 3 truncated hexa-octahedral magnetite magnetosomes are twinned relative to each other, so that they zigzag to form a linear chain (Fig. 1J,L). Thomas-Keptra et al. (2000, 2001) reported magnetite particles with five features of magnetosomes that constitute a biosignature, namely: size range and length-to-width ratios that cluster in the superparamagnetic and single-domain range; chemical purity; structural perfection; hexa-octahedral morphology; and elongation in the [111] direction. Because the magnetite particles in ALH84001 possess all 10 of the properties of magnetosomes and their chains, and inorganic and synthetic magnetites possess none or only a few, it is most likely that ALH84001 contains magnetofossils.

Golden et al. (2001) produced magnetite particles by thermal decomposition of synthetic Mg–Fe-carbonate concretions. They reported two magnetite particles as being parallelepipedal (truncated hexa-octahedral) based on a single projection, but detailed analysis was not performed. They also reported a chain of magnetite particles, but the particles had different shapes to each other, were not preferentially elongated parallel to the chain's axis and lacked ISs, so the chain did not match the biosignature proposed by Friedmann et al. (2001b). In a public detate referred to here as the abstract: Friedmann et al. (2001a), Kathie Thomas-Keptra argued that Golden et al. (2001) had used analytical reagents to produce the concretions, so they lacked significant quantities of trace elements such as Cr and Al. Therefore, their model does not account for the chemical purity of the hexa-octahedral magnetite particles and the concurrent chemical impurity of the irregularly shaped and structurally flawed magnetite particles in ALH84001.

Barber & Scott (2002) reported the preferential orientation of perclase (MgO) with respect to the carbonate lattice. The crystals were often associated with voids, which they proposed to be due to the escape of CO2 generated from the thermal decomposition of Mg–Fe-carbonate. They also reported magnetite particles that were aligned with the carbonate lattice and were associated with voids. Euhedral magnetite particles that were entirely embedded in the carbonate were topotactically orientated with respect to the carbonate lattice. The magnetite particles in the Fe-rich rims where irregularly shaped and were not well orientated with respect to the carbonate lattice. They reported that the occurrence of magnetite and perclase were consistent with in situ growth by solid-state diffusion, as a result of carbonate decomposition during impact heating, and dismissed all claims that the magnetite particles are biogenic. This high-temperature origin for the magnetite and carbonates is inconsistent with the palaeomagnetic history of the meteorite (Kirschvink et al., 1997).

Bradley et al. (1998) reported the alignment of lattice fringes in the carbonates surrounding magnetite particles in ALH84001, with lattice planes in the magnetite particles. The {inline image1inline image} and {11inline image} lattice planes in the magnetite were aligned with the {1inline image0} and {00inline image} lattice planes in the carbonates, respectively, with mismatched alignments of 13% and 11%, respectively. They interpreted this as evidence of vapour-phase growth of both minerals via epitaxy. The observation of lattice fringes in the magnetosomal matrix aligned with the {311} and {111} planes of magnetite magnetosomes in dehydrated magnetotactic cells (Table 2) and observations of the magnetosomal matrix (Table 1) and magnetofossils in carbonates surrounded by an electron-dense region (McNeill, 1990), collectively indicate that Bradley et al. (1998) and Barber & Scott (2002) may have detected more evidence of a fossilized magnetosomal matrix surrounding the Martian magnetite particles. Some arguments claiming evidence of a high-temperature (inorganic) origin for the magnetite in ALH84001 have been dismissed (Devouard et al., 1998; Taylor et al., 2001). However, multiple origins for the magnetite in ALH84001 would not exclude a biological origin for some (Thomas-Keptra et al., 2002), namely the chains of chemically and structurally pure hexa-octahedral magnetite. Although only about one-quarter of the magnetite particles in ALH84001 are chemically and structurally pure with a hexa-octahedral morphology and only a few of these are arranged in chains, magnetosomes can be corroded during sedimentation and fossilization and terrestrial magnetofossils are rarely found arranged in chains, so the arrangement of the Martian magnetite particles is consistent with terrestrial magnetofossils. Furthermore, Taylor et al. (2001) clearly demonstrated that all of the structurally imperfect and morphologically irregular magnetite particles in ALH84001 have terrestrial biogenic analogues. All of the chemically pure magnetite particles in ALH84001 could have been produced by a population of just a few strains of magnetotactic bacteria. Taylor (2002) also reports chemically impure magnetite magnetosomes produced by HCF3 cells, which contain traces of Mn. This indicates that some of the chemically impure magnetite particles in ALH84001 may be biogenic also and that chemical impurity cannot be used to exclude a biogenic origin for magnetite particles.

Pre-magnetite: possible precursor in the biomineralization of magnetite

Pre-magnetite is a hypothetical material, and evidence for its existence is based on images of two magnetosomes with immature sizes, imaged along the [inline image12] and [011] axes, respectively (Figs 4 and 6). Neither of the magnetosomes had evidence of lattice twist, as their SAED patterns lacked superperiodicity as found in twinned magnetosomes or in serpentine rolled microstructures of chrysotiles (Yada, 1971; Amelinckx et al., 1996), nor did phase-contrast lattice images contain Moiré fringes or evidence of lattice strain (Figs 4 and 6). Lattice twist therefore seems unlikely.

Frankel et al. (1983) hypothesized that ferrihydrite may be the precursor in the biomineralization of magnetite. However, the unidentified phase reported in Figs 4 and 6 here has lattice planes with d values that are inconsistent with those of two-line ferrihydrite [Fe4(O,OH,H2O)12] and six-line ferrihydrite [Fe4.6(O,OH,H2O)12] (Eggleton & Fitzpatrick, 1988). Maghemite (γ-Fe2O3) has a structure that is isomorphous to that of magnetite. Banfield et al. (1994) reported faint {001} and {110} reflections in SAED patterns from maghemite imaged along the [110] axis. No such extra reflections were observed here (Fig. 6C), so it is unlikely that the unidentified phase was maghemite. It may therefore be a new material and, if so, a biosignature.

Mann & Frankel (1989) reported extensive non-crystalline outgrowths overlying the surfaces of immature magnetite magnetosomes. The outgrowths had a superperiodicity when imaged along the [01inline image] axis. Only {022} lattice fringes were imaged in the outgrowths, which is consistent with it being composed of pre-magnetite.

Conclusions

The works of McKay et al. (1996), Uwins et al. (1998), Steele et al. (1998), Thomas-Keptra et al. (2000, 2001, 2002), Friedmann et al. (2001b), Gibson et al. (2001), Taylor et al. (2001) and this study collectively provide solid evidence for past life on Mars, the techniques to detect and analyse such fossils and the terrestrial analogues to compare them with. They also provide the techniques to detect, culture and analyse possible present Martian life-forms such as nanobes (if they exist). It follows that sample retrieval missions to Mars to confirm the results of McKay et al. (1996), Thomas-Keptra et al. (2000, 2001) and Friedmann et al. (2001b) are warranted. More work is needed to elucidate the structures and compositions of the magnetosomal matrix and pre-magnetite, as well as the possible intermediate in the solid-state transformation of pre-magnetite to magnetite. Electron spectroscopic imaging will be useful to elucidate the prevalence of the magnetosomal matrix throughout magnetotactic bacteria. More work is also needed to elucidate the composition and structure of the halo surrounding magnetofossils in terrestrial and Martian carbonates.

Acknowledgements

We wish to thank Mr Duncan Waddell for assistance with software, Mr Richard Webb for assistance with ultrasections, Dr A. Chris Hayward for assistance with funding, the management, staff and members of Howeston Golf Course Pty. Ltd and Tamarind Taylor for assistance with sample collection, and Dr Ann Kemp, Mrs Kathleen Taylor and the Astronomical Association of Queensland for moral support.

Ancillary