Magnetotactic bacteria and magnetofossils: ecology, evolution and environmental implications

Abstract

Magnetotactic bacteria (MTB) are a group of phylogenetically diverse and morphologically varied microorganisms with a magnetoresponsive capability called magnetotaxis or microbial magnetoreception. MTB are a distinctive constituent of the microbiome of aquatic ecosystems because they use Earth’s magnetic field to align themselves in a north or south facing direction and efficiently navigate to their favored microenvironments. They have been identified worldwide from diverse aquatic and waterlogged microbiomes, including freshwater, saline, brackish and marine ecosystems, and some extreme environments. MTB play important roles in the biogeochemical cycling of iron, sulphur, phosphorus, carbon and nitrogen in nature and have been recognized from in vitro cultures to sequester heavy metals like selenium, cadmium, and tellurium, which makes them prospective candidate organisms for aquatic pollution bioremediation. The role of MTB in environmental systems is not limited to their lifespan; after death, fossil magnetosomal magnetic nanoparticles (known as magnetofossils) are a promising proxy for recording paleoenvironmental change and geomagnetic field history. Here, we summarize the ecology, evolution, and environmental function of MTB and the paleoenvironmental implications of magnetofossils in light of recent discoveries.

Introduction

The terms biomineralization and magnetoreception are invariably associated with magnetotactic bacteria (MTB). Magnetoreception involves the use of Earth’s magnetic field for motility and navigation and biomineralization is a capability by which organisms produce minerals^1,2. MTB are so far the only known group of prokaryotes with the ability to perform both biomineralization and magnetoreception³. MTB have been proposed to represent some of the most ancient organisms capable of biomineralization⁴. This group of bacteria was discovered in 1963 by Salvatore Bellini⁵, and was rediscovered independently in 1974 by Richard Blakemore⁶. The latter study also identified magnetosomal magnetic particles for the first time, which were referred to as “iron-rich particles”. MTB form magnetite (Fe₃O₄) and/or greigite (Fe₃S₄) crystals, generally in bead-like chains⁷ (Fig. 1). Magnetotaxis is one of their primary modes of motility⁷. Many organisms can use the geomagnetic field for navigation. Birds, fish, and certain migratory animals among others top this list, while desert ants, newts, spiny lobsters, snails, Bogong moths, and certain other invertebrates have been recently found to possess a magnetic sense of some sort^8–11. Some mammals, too, can respond to Earth’s magnetic field. Laboratory experiments indicate that wood mice and mole rats make use of magnetic field lines while siting nests; certain cattle and deer use the field for body orientation while grazing; and dogs appear to orient toward north or south when they excrete^12–14. Magnetic orientation is associated with many organisms, and understanding its origin is a subject of extensive research.

Fig. 1. Electron microscope images of MTB.

Transmission electron microscope (a–f) and scanning electron microscope (g–i) images of various MTB. Black arrows indicate magnetosome chains. Scale bars: a, b = 0.5 μm, c–i = 1 μm.

Magnetoreception represents a spectrum of capabilities of which magnetotaxis is a subset. While magnetoreception is mostly associated with higher organisms that use a magnetic sense for mobility, magnetotaxis is associated with microorganisms. It is related to previously discovered methods of taxis such as chemotaxis and aerotaxis, which are common modes of transportation and translocation by bacteria and archaea^15,16. However, unlike chemotaxis or aerotaxis, which are multidirectional, magnetotaxis mostly involves upward/downward movement in search of optimal microenvironments near chemical gradients in water/sediment, aligning passively along Earth’s magnetic field¹⁷. It is thought that magnetotaxis along with chemotaxis/aerotaxis provides an additional benefit to MTB by permitting a one-dimensional search along the oxic–anoxic interface (OAI) in aquatic environments to enable MTB to find optimal oxygen concentrations to carry out necessary physiological functions¹⁸. MTB ecology, diversity, and evolution have been reviewed previously^19–21; however, recent developments of omics, cultivation and magnetic measurements have expanded our understanding of MTB and magnetofossils. Multiple studies have pointed out that magnetotaxis is monophyletic in origin; that is, it originated from a single common ancestor^22–24. This would make it a primordial physiological phenomenon and (probably) the earliest case of magnetoreception and systematic biomineralization on Earth^4,25. Following this discovery, diverse multidisciplinary studies have sought to answer several significant questions. Are MTB widespread across the domain Bacteria? How did magnetotaxis originate and evolve? Do MTB have a significant role in biogeochemical element cycling? These questions are of multidisciplinary interest to microbiologists, geologists, physicists, and chemists. In this paper, we review progress to date in addressing these questions. We discuss the phylogenetic diversity of MTB and their potential role in the biogeochemical cycling of elements, and compare the metabolic pathways of all sixteen phylum-level MTB lineages identified so far. Moreover, we discuss expanding oxygen minimum zones (OMZs) in the oceans; diverse MTB likely live in OMZs and their environmental role in such settings has been understudied.

MTB diversity and ecology

MTB occur in multiple lineages of the bacterial tree of life, although with a patchy distribution. They include cocci, vibrio, rod, spirilla, and multicellular morphotypes^19,20,26,27. Until 2012, both cultured and uncultured MTB could be grouped mostly into the Proteobacteria phylum (including the Alphaproteobacteria, Gammaproteobacteria and Deltaproteobacteria classes) and a few in the Nitrospirae phylum²⁷ (Fig. 2a). In 2012, a novel uncultivated ovoid MTB (designated SKK-01) from Lake Chiemsee, Germany, was discovered and characterized to belong to the candidate phylum Omnitrophica (also known as candidate division OP3)²⁸ (Fig. 2b). In 2015 and 2017, two additional bacterial phyla, the candidate phylum Latescibacteria²⁹ and the phylum Planctomycetes²¹, respectively, were identified to contain MTB based on the presence of magnetosome gene clusters (MGCs, a group of genes responsible for magnetosome biomineralization and magnetotaxis) in their genomes (Fig. 2c). More recently, advances in metagenomic approaches, large-scale field studies and public database mining studies have expanded the total number of MTB-containing phylum-level lineages from five to sixteen^24,30,31, which greatly increases the taxonomic and genomic diversity of MTB across the domain Bacteria (Fig. 2d). These findings highlight a much higher diversity of MTB lineages in the domain Bacteria than previously anticipated. On the basis of MGC types in MTB genomes, putative Fe₃O₄-producing MTB have been identified in all MTB lineages so far except for the Latescibacteria phylum, while putative Fe₃S₄-producing MTB exist in at least five bacterial phyla³¹.

Fig. 2. Schematic representation of the gradual expansion of the phylogenetic tree of MTB in the Bacteria domain.

a Before 2012, only two phyla (Proteobacteria and Nitrospirae) were identified to contain MTB. b Between 2012 and 2015, an additional branch for the Candidatus Omnitrophica phylum was added²⁸. c Subsequently, two bacterial phyla (Candidatus Latescibacteria²⁹ and Planctomycetes²¹) were added to the MTB tree of life based on the presence of magnetosome gene clusters in their genomes. d The latest addition to the MTB tree has expanded its branches to a total of sixteen bacterial phylum-level lineages^30,31. All taxonomic groupings are based on the NCBI taxonomy (https://www.ncbi.nlm.nih.gov/taxonomy).

MTB affiliated with the Nitrospirae phylum are biogeochemically important because many members of this group produce several hundred magnetosomal magnetite crystals per cell rather than the normal 10–50 crystals in most MTB^26,32. Characteristic examples are the Candidatus Magnetobacterium bavaricum, which was discovered by Vali et al.³³ and was subsequently phylogenetically defined by Spring et al.²⁶ as magnetotactic rods that form elongated, bullet-shaped and kinked crystals arranged in a parallel chain-like fashion that appear as rope-like strands or bundles. Hanzlik et al.^34,35 studied the chain architecture of Ca. Magnetobacterium bavaricum in detail. Apart from producing magnetite within magnetosomes, cells of Ca. Magnetobacterium bavaricum also harbor sulphur globules, solely comprising cyclo-octasulphur (S8), which gives them a probable role in both iron and sulphur cycling in nature³⁶. The second example is Ca. Magnetobacterium casensis, which is smaller than Ca. Magnetobacterium bavaricum and produces around 200–500 magnetosomal Fe₃O₄ particles per cell^37–39. Recent 16S rRNA-based and metagenomic surveys suggest that magnetotactic Nitrospirae are more diverse and more widely distributed than previously thought^31,40, which emphasizes their environmental importance in various ecosystems.

Apart from single-celled bacteria, multicellular magnetic prokaryotes (MMPs) are flagella-aided motile organisms that also produce magnetic nanocrystals, as first reported by Farina et al.⁴¹. Despite their multicellular organization, MMPs have coherent magnetosome-chain polarity across their cells⁴¹. Many studies have now been carried out involving especially the spherical MMP morphotype (sMMP)^42–48. A second morphotype, ellipsoidal MMP (eMMP), has been found in the Mediterranean Sea and Pacific Ocean^49–57. MMPs are affiliated within the Deltaproteobacteria and MTB from this group produce Fe₃O₄ and/or Fe₃S₄ nanoparticles^23,31. Studies of MMPs also provide insights into the origin and evolution of bacterial multicellularity, which is one of the most prevalent evolutionary innovations in the bacterial world⁵⁸.

MTB have been identified from diverse biomes with highly variable, including extreme environments. For instance, MTB flourish in shallow hemipelagic sediments (at ~600 m depths) at 8 °C, in Santa Barbara Basin, Eastern Pacific Ocean⁵⁹. Petermann and Bleil⁶⁰ discovered live MTB from pelagic and hemipelagic sediments in the eastern South Atlantic Ocean. They observed assorted cocci, spirilla, vibrioids, and even rod-shaped MTB morphologies in pelagic environments on the African continental margin and on Walvis Ridge at ~3000 m and ~1000 m water depths, respectively. More recently, MTB have been observed from tropical marine environments in Singapore, which adds to equatorial MTB biodiversity⁶¹. A recent phylogenetic analysis of 16S rRNA gene sequences of microbial communities from a seamount in the Mariana volcanic arc (238–2023 m) near Challenger Deep, tropical western Pacific Ocean, revealed 16 novel MTB populations⁶². Abundant MTB have also been observed in seafloor sediments of the Yellow Sea with Nitrospirae being the dominant group⁶³. In more extreme environments, Lefèvre et al.⁶⁴ isolated three MTB strains of Desulfonatronum thiodismutans from hyper-alkaliphilic habitats in California, USA, that are obligate alkaliphiles (with optimal growth at pH 9.0–9.5) and sulphate reducers. Essential genes (i.e., dissimilatory sulphite reductase (dsrAB) and adenosine-5′-phosphate reductase (aprA)) were identified that verify that the novel strains are sulphate reducers. Psychrophilic MTB have been characterized from water and sediment samples from Admiralty Bay, King George Island, Antarctica, where average water temperatures during sampling ranged between 0.1 and 0.8 °C⁶⁵. Nash et al.⁶⁶ characterized a thermophilic MTB variety belonging to Nitrospirae from Little Hot Creek Hot Springs, California, and another Gammaproteobacteria MTB from the hyper-alkaline and hypersaline Mono Lake. Lefèvre et al.⁶⁷ discovered a thermophilic MTB that they named Candidatus Thermomagnetovibrio paiutensis from mud and water samples in hot springs of the Great Boiling Springs geothermal field in Gerlach, northern Nevada. Two rod-shaped Gammaproteobacteria MTB strains, BW-2 and SS-5, from the hypersaline Salton Sea were the first isolated MTB from the Gammaproteobacteria class that biomineralize magnetite⁶⁸. Further Gammaproteobacteria MTB strains FZSR-1 and FZSR-2 were identified recently from a salt evaporation pool in Fuzhou saltern, Bohai Bay, China⁶⁹. Lin et al.³¹ also discovered living MTB cells in a relatively acidic peatland soil with draft genomes dominantly comprising magnetotactic Nitrospirae, Omnitrophica and Deltaproteobacteria.

Origin and evolution of MTB

Ancient atmospheric molecular O₂ in the Archean, 2.4 billion years ago (Ga), was less than 0.001% of the atmosphere today⁷⁰ and Fe(II) concentrations probably ranged between 0.03 and 0.5 mM^71,72 (modern oceanic iron concentrations range between 0.05 and 2.0 nM⁷³). Archean ocean sulphate concentrations were also less than 1/100 of present levels, with concentrations lower than 200 µM^74,75 and N₂ was a bulk gas with similar levels to today⁷⁶. Oxygen concentrations in the atmosphere and surface ocean first rose at ~2.4 Ga in the Great Oxygenation Event (GOE)⁷⁷ with a second increase in the Neoproterozoic Oxygenation Event⁷⁸ (NOE), which established a more modern ocean redox profile. The GOE is generally thought to have facilitated the emergence of eukaryotes⁷⁹ while Betts et al.⁸⁰ argue that the NOE was associated with emergence of large and complex multicellular organisms. Thus, the GOE and NOE were fundamental pacemakers for evolution of life on Earth, although the beginning of microbial life must have been much earlier than the GOE. There is evidence of microbial life originating sometime during the early Archean at ~3.5–3.8 Ga^79,81. The presence of biogenic carbon has been uncovered in detrital zircon grains from Jack Hills, Western Australia⁸²; if verified by further studies, the beginning of microbial life can be traced to ~4.1 Ga in the late Hadean Eon. Further evidence is required to corroborate this early age, which potentially transforms the way scientists view paleo-environments on Earth.

Fossilized magnetosomal magnetic particles (referred to as magnetofossils) can be preserved in sediments or rocks. Magnetofossil records trace an evolutionary history of MTB to the Cretaceous and with less certainty to the Paleoproterozoic at around ∼1.9 Ga^83–85, which suggests that MTB should have an early origin. Furthermore, phylogenetic and molecular clock analyses suggest an Archean origin for genetic functionality of magnetosome biomineralization (which is needed to perform magnetotaxis); specifically, the emergence of MGCs is estimated to date to the Archean Eon (∼3.38–3.21 Ga)²⁵ or even earlier^24,31. It is important to remember that atmospheric oxygen was most likely sparse in the Archean so that an anoxic environment prevailed^86–89. In addition to magnetotaxis, magnetosomal particles have been proposed to act as iron storage and sequestration organelles, or as gravity detection units or “geobatteries” that provided energy from elemental oxidation-reduction cycling⁹⁰. It has recently been proposed that the initial role of magnetosomal magnetic particles was to mitigate intracellular reactive oxygen species (ROS) toxicity and that, eventually, they were co-opted for magnetotaxis in early Earth environments⁴. A photoferrotrophy-driven origin of magnetotaxis has also been proposed; that is, magnetosome formation was a by-product of Archean iron cycling and magnetotaxis evolved as a result of environmental pressure of co-evolved cyanobacteria and metabolically accumulated magnetite⁹¹. Magnetosomes have also been proposed to provide a protective shield in a metal-stressed environment⁹². To ascertain the evolutionary history of MTB, integration of microbiology, evolutionary biology, geobiology, biogeochemistry, and geochronology is required.

The recent development of metagenomics and single-cell genomics allows direct reconstruction of MTB genomes from environmental samples without cultivation. Comparison of these novel genomes with those from cultivated MTB strains provides insights into the evolution of magnetotaxis. Magnetosome protein phylogeny largely mirrors that of organisms at or above the class or phylum level, which suggests that vertical inheritance followed by multiple independent MGC losses mainly drove bacterial evolution of magnetotaxis at higher taxonomic levels^23,25,31. Subsequent evolutionary trajectories of magnetotaxis at lower taxonomic ranks appear to be much more complicated and multiple evolutionary processes including horizontal gene transfers, gene duplications and/or gene losses may have been involved^93–95. Metagenomic sequences with similarity to known magnetosome genes have been found in the microbiomes of some animals and even humans, which might suggest that MTB sensed by their hosts may produce symbiotic magnetoreception in these organisms^96–98. It has also been suggested recently that the ability to biologically control magnetite precipitation might have been a fundamental feature of eukaryotic biology that was likely present in the last common ancestor of some archaea and extant eukaryotes⁹⁹.

Magnetic characterization and quantification of MTB and magnetofossils

Magnetofossils are distributed widely in freshwater and marine sediments⁸³. Recently, their presence has also been reported in ferromanganese (Fe–Mn) crusts and abyssal manganese nodules^100–103. Several methods exist for identifying magnetofossils. The most direct way is to observe them under a transmission electron microscope (TEM). Kirschvink and Chang¹⁰⁴ first observed magnetofossils under a TEM from deep-sea sediments of the Southwest Atlantic, Equatorial Pacific, and South Atlantic Oceans. Petersen et al.¹⁰⁵ found additional magnetofossil morphotypes from South Atlantic marine sediments, with bullet-, prismatic- and octahedral shapes that they considered to be the predominant natural remanent magnetization (NRM) carrier in sediments with ages from Quaternary to Eocene. Magnetosomal magnetite crystals usually have a [111] elongation direction, while Li et al.¹⁰⁶ found that bullet-shaped biogenic magnetite grows in a two-stage process with the second stage involving elongation in the [100] direction. Kopp and Kirschvink⁸³ proposed a series of criteria to identify magnetofossils. For instance, biogenic magnetosomal magnetite crystals have a narrow size distribution and distinctive morphologies with blunt crystal edges, high chemical purity, crystallographic perfection and chain arrangement.

Magnetic techniques provide complementary approaches that are commonly used to initially identify magnetofossils before direct TEM observation. These methods include low-temperature remanence measurements, first-order reversal curve (FORC) diagrams, and ferromagnetic resonance (FMR) spectroscopy. Moskowitz et al.¹⁰⁷ suggested that low-temperature isothermal remanent magnetization (IRM) measurements can be used to distinguish magnetite magnetosome chains. They defined δ as δ = (IRM_80K-IRM_150K)/IRM_80K and the δ ratio = δ_FC/δ_ZFC, where δ_FC and δ_ZFC are the difference between field cooled (FC) and zero-field cooled (ZFC) IRM curves at 150 K and 80 K, respectively. Moskowitz et al.¹⁰⁷ concluded that biogenic magnetite chains are present when the δ ratio >2. However, when magnetosome chains are oxidized, the δ ratio becomes less diagnostic. The test suggested by Chang et al.¹⁰⁸, based on low-temperature cycling of a saturation IRM imparted at room temperature, is more robust for oxidized magnetosomes. Chang et al.¹⁰⁹ reported that a double Verwey transition temperature signal can also be used to identify magnetofossils and suggested that ~100 K and ~120 K are the Verwey transition temperatures of biogenic and detrital magnetite, respectively, although it has also been suggested that the two Verwey transitions could result from other factors¹¹⁰.

FORC diagrams are normally interpreted in terms of the magnetic coercivity distribution and magnetostatic interactions among single-domain magnetic particles^111–113. When interactions among uniaxial single-domain particles are negligible or weak, FORC diagrams will have a central ridge along the horizontal B_c axis, with a small vertical distribution along the vertical B_u axis. Although magnetic particles in magnetosome chains have strong interactions among them, the entire chain will act as a single needle with uniaxial magnetization that does not interact with other chains¹¹⁴. Therefore, MTB samples produce a central ridge in FORC diagrams^115–119. However, when magnetofossil chains are broken, magnetic particles clump together so that interactions become strong; a central ridge usually persists and FORC distributions spread vertically in the B_u direction¹²⁰. Inorganic magnetite with weak interactions also produce a central ridge FORC signature. Thus, if a FORC central ridge is observed, further TEM observations are needed to demonstrate the presence of magnetofossils.

FMR spectroscopy is sensitive to the magnetic anisotropy of the chain configuration¹²¹. The magnetic anisotropy of MTB arises mainly from dipolar interactions among adjacent magnetic particles in a chain, which behaves like an elongated single-domain particle¹¹⁴; therefore, FMR is used to indicate the presence of magnetosome chains^122,123. The main FMR parameters are as follows: the effective g-factor (g_eff), asymmetry ratio (A), and empirical parameter (α), where g_eff = hv/βB_eff, h is Planck’s constant, v is the microwave frequency, β is the Bohr magneton and B_eff is the maximum absorption field, A = ΔB_high/ΔB_low, where ΔB_high = B_high − B_eff, and ΔB_low = B_eff − B_low. The full width at half maximum is defined as ΔB_FWHM = B_high + B_low (Fig. 3e). Weiss et al.¹²² proposed that a magnetosome chain will have A < 1 and g_eff < 2.12. Kopp et al.¹²³ proposed an empirical parameter, defined as α = 0.17 A + 9.8 × 10⁻⁴ ΔB_FWHM/mT, and concluded that A < 1 and g_eff < 2.12 are insufficient to ensure the presence of magnetofossils. However, all of their measured MTB had α < 0.25, and magnetofossil-bearing samples have α = 0.25–0.30. When α is larger, magnetofossil contents are lower. Kodama et al.¹²⁴ suggested that detrital and extracellularly produced authigenic magnetite (D + Ex) dominate sediments with α > 0.40, while magnetofossil-rich sediments have α < 0.35 and A < 1. However, inorganic elongated single-domain particles in a volcanic tuff also have 0.3 < α < 0.35 and A < 1¹²⁵. Kind et al.¹²⁶ investigated magnetic components in Holocene Lake sediments by combining anhysteretic remanent magnetization (ARM), IRM, FORC diagrams and FMR spectra and suggested a combination of FORC and FMR measurements to detect magnetofossils in natural environments. Blattmann et al.¹²⁷ applied quantitative FMR to analyze magnetofossil variations in Lake Constance, which records sediment–water interface redox changes. FMR spectroscopy is also used widely to detect magnetofossils in pre-Quaternary marine sediments^116,128.

Fig. 3. Methods for identifying magnetofossils in samples from core MD01-2444 at 26.74 m depth.

a Low-temperature magnetic measurements (blue and red curves: normalized zero-field cooled (ZFC) and field cooled (FC) curves, respectively; black curve: normalized first derivative of the FC curve; peaks with arrows (~100 K and ~115 K) are indicative of biogenic and detrital magnetite, respectively). b Coercivity distribution from IRM acquisition curve unmixing (gray dots: IRM acquisition data; orange curve: spline fit based on measured data; purple and red curves: biogenic soft and biogenic hard components, respectively; shaded areas for each curve are 95% confidence intervals). c, f TEM images of magnetic extracts from the same sample. Scale bars: c = 200 nm, f = 100 nm. The main magnetofossil morphotypes are prisms, octahedra and bullets. d FORC diagram with a central ridge signature that is indicative of non-interacting single-domain magnetic particles. e Ferromagnetic resonance (FMR) and FMR absorption spectra (black curve: measured FMR spectrum; red curve: fitted FMR spectrum; black humped curve: normalized FMR absorption spectrum). Definition of commonly used magnetofossil-indicative parameters are shown in (e)¹²³.

To better understand the magnetic properties of magnetofossils, micromagnetic modeling has been applied to build a link between magnetofossil size distributions and magnetic properties. Muxworthy and Williams¹²⁹ found that the largest reported interacting magnetosomal magnetite particles still have single-domain behavior based on micromagnetic simulations. Chang et al.¹³⁰ used micromagnetic models to build a link between magnetic properties and magnetofossil size distributions from Ocean Drilling Program (ODP) Site 1263 during the Paleocene–Eocene thermal maximum (PETM). Moreover, micromagnetic modeling demonstrates that magnetosome-chain structures play a more important role in controlling magnetic particle coercivity than morphology¹³⁰. Different MTB species have different chain structures and magnetic particle numbers. Therefore, magnetofossil diversity probably affects the magnetic properties of bulk samples, especially when magnetofossils are the main NRM carrier. Berndt et al.¹³¹ modeled magnetosome chain morphologies (including crystal size, elongation, chain length, and intra-chain spacing) and found that intra-chain spacing can control the magnetism of magnetosomal magnetic particles such as coercivity, coercivity of remanence, and saturation magnetization.

Heslop et al.¹³² calculated the vector difference sum (VDS) of components in NRM demagnetization curves to quantify magnetofossil contents from sediments offshore of northwestern Australia. In addition, IRM unmixing and principal component analyses applied to FORCs (FORC-PCA) are used to quantify magnetic components in sediments. Two categories of methods are used to unmix IRM acquisition curves. One is single sample IRM unmixing, which includes fitting of log-normal, skewed generalized Gaussian, and Burr-XII distribution models^133–135. Egli¹³⁶ concluded that the dispersion parameter of biogenic magnetite is <0.2, while values for detrital magnetite particles are between 0.3 and 0.4. Chang et al.¹³⁰ analyzed magnetic components from PETM sediments using a cumulative log-Gaussian function and decomposed IRM acquisition curves into three components, including biogenic soft and hard magnetite, and a soft detrital component. They suggested that magnetofossil components contribute 76% of the total remanent magnetization during the PETM onset. A further IRM unmixing approach is endmember-based IRM unmixing, with non-negative matrix factoring¹³⁷. Usui et al.¹³⁸ reconstructed the biogenic proportion of particles using this approach and obtained a two biogenic magnetite endmember solution by combining χ_ARM/IRM ratios (a magnetic mineral grain size proxy) and FORC diagrams. They concluded that the two biogenic magnetite endmembers represent isotropic and bullet-shaped magnetofossils.

Lascu et al.¹³⁹ and Harrison et al.¹⁴⁰ demonstrated that FORC-PCA can help to determine biogenic and detrital magnetite abundances. FORC diagrams enable the characterization of magnetic minerals with different domain states and interaction fields. Channell et al.¹⁴¹ applied this method to trace variations of three magnetic endmembers (EMs) in the Rockall Trough (NE Atlantic), including a magnetofossil EM. Roberts et al.¹⁴² used FORC-PCA to detect magnetic property variations during early diagenetic reduction, including loss of a magnetofossil component due to dissolution and to estimate the proportions of different minerals in different diagenetic systems. Yamazaki et al.¹⁴³ applied FORC-PCA to two cores from the western North Pacific Ocean and the South Pacific Ocean and identified two non-interacting single-domain EMs, which represent biogenic soft and biogenic hard components, respectively¹⁴⁴. Qian et al.¹⁴⁵ used FORC-PCA analysis of eastern Mediterranean sediments to demonstrate that elevated magnetofossil abundances occur at oxidation fronts above organic-rich intervals. FORC-PCA is becoming increasingly common for quantifying sedimentary magnetofossil contents in sediments^146–148. Combining magnetic measurements and TEM or scanning electron microscope (SEM) observations can provide information to identify, characterize and quantify magnetofossils in natural samples (Fig. 3).

Paleoenvironmental and paleomagnetic implications of magnetofossils

Varying proportions of different magnetofossil morphotypes can reflect the paleoredox level of sediments^{100,101,103,149–151}. Abundant bullet-shaped magnetofossils have been detected in reducing environments with higher total organic carbon (TOC), while octahedral magnetofossils appear to dominate relatively oxic environments. However, Lean and McCave¹⁵² found that more elongated magnetofossils occur in low-TOC horizons. Therefore, the paleoenvironmental implications of magnetofossil morphologies need further investigation, especially the relationship between magnetofossil abundance and sediment nutrient content. Moreover, if magnetofossils undergo diagenetic modification, the relationship between their abundance and paleoenvironment will change¹⁵³. For example, moderate TOC availability promotes mild diagenetic iron reduction that reduces sedimentary iron so that it becomes bioavailable to MTB¹¹⁶. However, high TOC produces sulphidic diagenetic environments in which magnetite magnetofossils dissolve¹⁵⁴. Equant magnetofossils dissolve more easily than bullet-shaped forms¹⁵⁵, and bullet-shaped magnetofossils are more easily dissolved than hexagonal prisms and octahedral forms¹⁵³. Magnetofossils are ideal ancient magnetic field recorders. In environments where magnetofossils are preserved, Ouyang et al.¹⁵⁶ and Chen et al.¹⁵⁷ demonstrated that that they have a higher magnetic recording efficiency than detrital magnetite, while Li et al.¹⁵⁸ found that biogenic magnetite can be less efficiently magnetized than detrital magnetite. The debate also remains about whether magnetofossils record a biogeochemical remanent magnetization (BGRM). Tarduno et al.¹⁵⁹ suggested that magnetofossils that survive burial below the Fe-redox boundary can produce delayed BGRM acquisition. Roberts et al.¹¹⁶ noted that paleomagnetic signals carried by magnetofossils must be acquired at shallow depths based on comparison between two nearby records. Tests for offsets between paleomagnetic signals carried by detrital and biogenic magnetite indicate no depth lag, so the evidence is still lacking for the BGRM mechanism^156,157. Yamazaki et al.¹⁶⁰ found that elongated magnetofossils inhabit lower parts of a redox zonation, while isotropic forms occupy shallower levels.

Although there are few reports of greigite magnetofossils, it has been suggested that they can be reliable paleomagnetic recorders¹⁶¹. Pósfai et al.¹⁶² proposed that greigite in Miocene sedimentary rocks is similar to biogenic crystals produced by MMP and Chang et al.¹⁶³ concluded that greigite magnetofossils are potentially widespread in ancient sediments. The favored MTB habitat depth relates to their ability to contribute to a BGRM; future detailed studies of the vertical distribution of MTB populations within the water column or sediments are required to constrain the habitat range of MTB.

Ecosystem functions of MTB

MTB are distributed globally in aquatic ecosystems²¹. They can comprise up to ~30% of the microbiome in some habitats²⁶ and their biomineralizing capability is assumed to be important for biogeochemical cycling of iron in sediments⁹⁰. The recent discovery of an Archean origin for MTB further suggests that they may have contributed to iron cycling throughout Earth history. With reports of the ability of MTB to fix nitrogen^164,165, sequester carbon¹⁶⁶, and to incorporate elements like sulphur, phosphorus, selenium, calcium, etc., into biominerals (Fig. 4)^{28,32,39,57,64,95,166–171}, evidence is growing to indicate that MTB play important roles in global biogeochemical cycling. Here, we compared the major KEGG pathways of representative MTB genomes across all known 16 bacterial phylum-level lineages, which suggest that MTB have phylogenetically diverse metabolic features (Fig. 5). What is not yet clear is the extent to which they impact the total cycling budget of each element.

Fig. 4. Magnetotaxis, combined with chemotaxis and aerotaxis, allows MTB to efficiently locate and maintain an optimal position for survival and growth in habitats with vertical redox concentration gradients in water columns and sediments around an oxic–anoxic interface (OAI).

MTB are widely distributed in aquatic environments from marine to freshwater ecosystems and are thought to play important roles in the cycling of various elements (e.g., Fe, C, N, S, P, and some heavy metals, such as tellurium and selenium).

Fig. 5. Metabolic pathways of representative MTB genomes across 16 bacterial phylum-level lineages.

The color gradient represents the completeness of major metabolic pathways inferred from the presence or absence of genes. Dark represents a complete or nearly complete pathway, and white represents a pathway that is absent or mainly incomplete. Note that most available MTB genomes are draft or incomplete, so we cannot rule out that the absence or incompleteness of metabolic pathways maybe due to the fragmented nature of draft genomes. All taxonomic groupings are based on the GTDB taxonomy (Release 89, https://gtdb.ecogenomic.org).

To better understand iron biomineralization, the processes and products of biomineralization, the relevant genetics, and magnetotaxis need to be analyzed. The ability to navigate by magnetotaxis depends on whether magnetosomal magnetite particles are fully developed¹⁹. Magnetotaxis is best achieved with fully developed crystals (~30–150 nm)^7,59,172 rather than with immature crystals (~30 nm or smaller). Cornejo et al.¹⁷² studied the biomineralization mechanism for mature octahedral crystal formation inside magnetosome membranes in Magnetospirillum magneticum strain AMB-1 (AMB-1), whereas Li et al.¹⁷³ studied the mechanism for mature bullet-shaped crystal formation in Deltaproteobacterial strain WYHR-1. These and further studies could help to better understand crystal biomineralization in various MTB strains. Magnetic particle growth is important because assemblages of mature magnetosomal magnetite particles in chain-like structures produce an intracellular magnetic dipole that interacts with Earth’s magnetic field to assist MTB cells in magnetic navigation^{114,174–176}. Therefore, research on the growth stage of magnetosomal crystals in different MTB populations would help to better understand underlying magnetotaxis mechanisms.

Bioavailable iron is scarce in many environments today, which may limit MTB population growth¹¹³. In particular, iron is scarce in high-nitrate low-chlorophyll (HNLC) oceans. Addition of iron from, for example, eolian dust or volcanic ash inputs, can fertilize surface ocean primary productivity, which increases the chlorophyll content in iron-limited biomes^177–179. Solubilization of iron from such inputs plays a key role in carbon and nitrogen fixation into biomass, which makes iron bioavailability responsible for primary production in these ecosystems^180–183. Export of carbon from the surface ocean to the seafloor can also stimulate mild diagenetic release of iron in the uppermost sediment to release a limitation on MTB productivity in pelagic environments and allows MTB populations to grow¹¹⁶. Considering the global MTB distribution in aquatic ecosystems, it has been suggested that MTB communities play significant roles in present-day global iron cycling^20,184.

Apart from MTB, magnetic inclusions have been reported within the purple-sulphur bacterial genera, Rhodopseudomonas and Ectothiorhodospira¹⁸⁵. These inclusions have no prismatic crystal structure and they lack a lipid-membrane to enclose magnetosomes. Unlike the microaerophilic model strains, AMB-1 and Magnetospirillum gryphiswaldense strain MSR-1 (MSR-1), recent analysis of an anaerobic sulphate-reducer, Desulfuvibrio magneticus strain RS-1 (RS-1), is gaining popularity as a model MTB organism^186,187. This species constructs sub-cellular electron-dense Fe-rich granules (sometimes including phosphorus and oxygen) that are encased by a lipid-like membrane. These “ferrosome” (iron body) granules are independent and separate entities from magnetosomes and are proposed to play a vital role in storing iron under anaerobic respiration during extreme iron starvation conditions¹⁸⁸. This aligns with the findings of Amor et al. that magnetite in AMB-1 only represents about ~25–30% of the total iron mass in a bacterial cell¹⁸⁹. Broader iron storage remains to be explored in relation to other MTB species to explore their probably crucial role in global iron cycling.

Apart from the well-studied Fe₃O₄ and Fe₃S₄ biomineralization products within MTB, diverse other inclusions contain elemental sulphur, polyphosphate, calcium, etc. Sulphur-rich inclusions have been identified in Magnetovibrio strains MV-1 and MV-2 of the Alphaproteobacteria class and in many Nitrospirae MTB^{7,28,32,57,190,191}. It was initially proposed that some MTB, like Nitrospirae MTB, are capable of adapting to chemical gradients near the OAI by adjusting their metabolic strategies—either by moving downward to accumulate reduced sulphur species or upward to oxidize stored sulphur with oxygen¹⁹². This redox-controlled response is supported by analysis of the vertical distribution of Ca. Magnetobacterium bavaricum in sediments³² and by genomic characterization of Ca. Magnetobacterium casensis³⁸. Ca. Magnetobacterium casensis and Ca. Magnetobacterium bavaricum are proposed to conduct sulphur oxidation with nitrate and/or oxygen as electron acceptors in the upper micro-oxic layer and to perform sulphate reduction in the anoxic lower layer, suggesting a complex metabolic strategy of Nitrospirae MTB and electron shuttling depending on redox conditions^38,39,193.

Bazylinski and Blakemore¹⁶⁴ revealed that Magnetospirillum magnetotacticum strain MS-1 (MS-1) fixes nitrogen at rates equivalent to those of Azospirillum lipoferum when cultured microaerobically under nitrogen limiting conditions, where Azospirillum species were among the earliest known potential nitrogen fixers for cereal plants¹⁹⁴. Apart from MS-1, two Magnetospirillum strains, including MSR-1 and AMB-1, when proliferated in nitrogen limiting, semi-solid media can also fix nitrogen via nitrogenase¹⁶⁴. MSR-1 contains indispensable nitrogenase structural genes required for nitrogen fixation in the presence of the DRAT-DRAG nitrogenase post-translational regulatory system¹⁶⁵ (Fig. 5).

Carbon sequestration is a key consideration when analyzing pollution mitigation potency of a plant or microorganism, both in terrestrial and aquatic habitats. Microorganisms and other autotrophs drive the carbon cycle because they are at the forefront of carbon cycle feedback mechanisms and their primary production is critical in carbon cycling¹⁹⁵. Marine phytoplankton have a well-known role in CO₂ sequestration¹⁹⁶; chemolithoautotrophs do the same under dark conditions in deeper, less-oxic/anoxic parts of ocean water or sediment redox gradients¹⁹⁷. If not obligate, most MTB are facultative chemolithoautotrophs, which might suggest an underlying role in CO₂ fixation¹⁹. Most terrestrial plants and microorganisms contain genes that encode a key metabolic pathway known as the Calvin–Benson–Bassham pathway (CBB) that allows the plant/microbe to sequester bioavailable carbon from the environment. Ribulose-1,5-bisphosphate carboxylase/oxygenase (RuBisCO) is an abundant protein that catalyzes the CO₂ fixation reaction¹⁹⁸. The cbbM gene that encodes form II RuBisCO was identified in magnetotactic chemolithoautotrophs Magnetovibrio strains MV-1 and MV-2 and Magnetospirillum strain MS-1; these species are proposed to be important in carbon cycling and primary productivity¹⁶⁶. Gammaproteobacteria MTB strains BW-2 and SS-5 from the hypersaline Salton Sea have a partial RuBisCO gene sequence and potentially use the CBB pathway for autotrophy⁶⁸. MTB from the Etaproteobacteria (Magnetococcus marinus strain MC-1) use the reverse tricarboxylic acid (rTCA) cycle for autotrophy¹⁹⁹, while magnetotactic Nitrospirae are proposed to fix CO₂ via the reductive acetyl-CoA (Wood-Ljungdahl or WL) and/or reductive tricarboxylic acid (rTCA) pathways^38,193,200 (Fig. 5).

Uncultured MTB cells from the Seine River, France, contain vesicles rich in barium and calcium oxide¹⁷¹ and MTB from Lake Pavin, France, biomineralize CaCO₃ in addition to magnetosomal magnetite²⁰¹. Instances of phosphorous sequestration by Magnetococcaceae alongside magnetite chains have also been demonstrated in ferruginous Lake Pavin¹⁶⁸ and in the anoxic Black Sea¹⁶⁹. Schulz-Vogt et al.¹⁶⁹ suggested that magnetotactic cocci act as a shuttle (using magnetotaxis) to transport phosphorous from the upper to the lower stratum of the suboxic zone, which prevents eutrophication resulting from excess phosphorus accumulation in the upper stratum.

Elements can also deposit on MTB cell surfaces, which is important in analyzing the role of MTB in heavy-metal bioremediation and magnetic separation of metal-loaded MTB cells from aquatic bodies. For example, Arakaki and colleagues²⁰² observed electron-dense Cd²⁺ deposits enveloping RS-1 cell surfaces under TEM when cells were cultured in growth media containing 1.3 ppm Cd²⁺. Mono-dispersed crystalline inclusions have 20–40 nm sizes and were easily distinguished from magnetic particles in TEM images²⁰². A novel Alphaproteobacterium MTB species grown in a cobalt supplemented medium has efficient biosorption competence with 89% cobalt removed by magnetic separation of the biomass²⁰³. Tellurite (O₃Te²⁻) uptake by MTB followed by biomineralization of discrete tellurium crystals alongside and separate from magnetite crystals was observed by Tanaka et al.²⁰⁴. These authors then demonstrated that when a growth medium is supplemented with selenite (SeO₃²⁻), elemental selenium nanoparticles with no definitive form gradually built up and some precipitated within cells, in conjunction with and autonomous from magnetite particles¹⁷⁰. Quantitative analysis revealed that MTB accumulate selenium at a higher rate per cell than other non-iron element accumulations. For example, Se accumulated by MTB is higher than O₃Te²⁻ uptake and Cd adsorbed by factors of 2.4 and 174, respectively¹⁷⁰. Elemental adsorption on cell surfaces is not specific to MTB, but they can contribute to metal pollutant removal via magnetic separation. Heavy-metal recovery from diodes and resistors of waste printed circuit boards using MTB is a latest discovery in terms of MTB applications in bioremediation²⁰⁵. Although such high metal concentrations are usually not encountered in natural environments, analysis of samples from polluted water bodies will help to identify the usefulness of MTB for bioremediation, e.g., MTB morphotypes have been observed in hydrocarbon-contaminated microcosms from the Gulf of Fos, France, and hypothesized to be involved in degrading hydrocarbons or other aromatic compounds²⁰⁶. It remains to be shown if MTB extracted from such microcosms can be grown in defined induction experiments using aromatic substrates under anaerobic conditions. MTB could, therefore, be potent model organisms for bioremediating contaminated water bodies and surface sediment. As new MTB genomes are obtained by metagenomics and single-cell genomics of environmental samples, identification and characterization of key proteins involved in these metabolic pathways will provide an important way to understand the ecological functions of MTB.

MTB as a potential important constituent of OMZ microbiomes

Oxygen minimum zones (OMZs) are oxygen delimiting/nutrient-rich sections of an oceanic gradient²⁰⁷. Waters above and below the OMZ have higher oxygen concentrations compared to the OMZ interval. OMZs are global sinks for reactive nitrogen because they have the highest microbial activity that conserves available oxygen and produces N₂ and N₂O as respiration by-products²⁰⁸ and have been termed “microbial reactors” with global significance²⁰⁹. OMZs have expanded gradually over the past 50 years due to ocean warming, which reduces oxygen solubility^210–212. Microbiological analyses of oceanic OMZs reveal that the most abundant phyla are Proteobacteria, Bacteroidetes, Actinobacteria, Planctomycetes, and Marinimicrobia (previously known as Marine Group A)²¹³. A characteristic feature of the OMZ microbiome is its predominant role in the biogeochemical cycling of marine nitrogen^214,215, sulphur^216–219, and carbon²¹⁵. Rhoads et al.²²⁰ showed that MTB can occur within an OMZ at oxygen levels as low as <4 mg L⁻¹. The range of dissolved oxygen contents in an OMZ appears to be optimal for MTB to thrive (0.1–0.5 mg L⁻¹). It has been hypothesized that marine dysaerobic zones (with 0.1–0.5 mL L⁻¹ dissolved oxygen concentrations) have a biogenic magnetite inventory based on TEM results in which fine-grained (80–100 nm) subrounded cubic magnetite predominates²²⁰. High-nitrate concentrations in OMZs suggest that these oxygen-limiting microenvironments should harbor greater denitrifying MTB populations²²⁰. Symbiotic occurrences have been proven by studying the “cryptic” sulphur cycle in OMZs around the world²²¹. OMZ biogeochemistry is important in the greater oceanic system with implications even outside the oceans. For example, consortia of ectosymbiotic, sulfate-reducing, chemolithoautotrophic Deltaproteobacteria, and Excavata protists have been reported²²². Sulphur-reducing bacteria with magnetosomes have also been observed in microbial mats on carbonate concretions from the Black Sea alongside other archaea that aid methane oxidation²²³. Sulphur-reducing gammaproteobacterial MTB even occur as extracellular symbionts on marine bivalves²²⁴. OAIs and OMZs are ecological niches in which MTB thrive within chemical gradients; tweaking the chemistry of such environments will likely affect much of the microbial community. The increasing importance of OMZs in the modern ocean could imply that global MTB populations will increase over time with accelerating ocean warming and that they could help to limit deterioration of ocean ecosystem health.

Outlook

MTB are distributed in aquatic ecosystems globally. Understanding of the phylogenetic taxonomy and metabolic flexibility of MTB has expanded greatly over the past decade, which suggests an unexpected natural diversity of these microorganisms. Continued discovery of novel MTB lineages from different environments highlights our limited knowledge of their diversity and ecology. Continuing cultivation-dependent and -independent MTB analyses will be crucial to understand their diversity and taxonomy more fully. We anticipate great progress in defining the phylogenetic diversity and evolutionary origin of MTB in the coming years. MTB incorporate elements from their surroundings to produce several biomineral types, with magnetic iron-bearing minerals being the primary biomineralization product. In addition to iron biomineralization, MTB can fix nitrogen, oxidize/reduce sulphur and sequester carbon and phosphorus from the environment. These findings suggest that MTB play important roles in biogeochemical elemental cycling through time, although their contributions are yet to be evaluated quantitatively. Moreover, MTB in marine OMZs/OAIs may play a role in ocean biogeochemical cycling and, in turn, in trimming eutrophication. If this is demonstrated by further work, MTB could be a natural eutrophication repressor that could become important with ongoing global warming. In addition, MTB can be used in pollution bioremediation by accumulating heavy metals on their surfaces by adsorption (e.g., Cd) and intracellularly (e.g., O₃Te²⁻ and Se). The magnetism of MTB ensures that metal-loaded MTB cells can be separated magnetically from contaminated waters. Magnetofossils are important in paleoenvironmental, paleoclimatic and paleomagnetic interpretations of sediments and sedimentary rocks. Further research is needed to understand the environmental adaptability of MTB and limiting factors that affect magnetofossil preservation to develop robust magnetofossil proxies for past environmental changes.

Author contributions

W.L., A.P.R., and P.G. conceived the review. All authors contributed to the writing and editing.

Data availability

All NCBI accession numbers of MTB genomes used in Fig. 5 are as follows-HG794546.1: Magnetospirillum gryphiswaldense MSR-1; CP000471.1: Magnetococcus marinus MC-1; CP032508.2: Gamma proteobacterium SS-5; JADFZC010000123.1: Oligoflexales bacterium isolate nHA4_bin4; JADFZW010000010.1: Oligoflexia bacterium isolate nKLK_bin12; AP010904.1: Desulfovibrio magneticus RS-1; FWEV01000001.1: Candidatus Desulfamplus magnetomortis BW-1 PRJEB14757; NVTF01000001.1: Elusimicrobia bacterium isolate NORP122; JADFYW010000100.1: Fibrobacteria bacterium isolate nGR_bin1; DUZM01000001.1: Candidatus Hydrogenedentes bacterium isolate MAG_17963_hgd_111; ASWY01000001.1: Latescibacteria bacterium SCGC AAA252-B13; JADGAO010000100.1: Nitrospinae bacterium isolate nNGH_bin12; JMFO00000000: Candidatus Magnetobacterium casensis; JADGBR010000010.1: Candidatus Omnitrophica bacterium isolate nW3_bin14; DUZJ01000001.1: Planctomycetes bacterium isolate MAG_11118_pl_115; JADGAS010000010.1: Zetaproteobacteria bacterium isolate nPC_bin1; JADFZL010000010.1: Candidatus Riflebacteria bacterium isolate nHGR_bin4; JADGAT010000100.1: SAR324 cluster bacterium isolate nPCR_bin10; JADGCS010000100.1: Deltaproteobacteria bacterium isolate nXX_bin1. The original data for Fig. 3 can be found at 10.6084/m9.figshare.17122277.v1.

Code availability

MagCluster (https://github.com/RunJiaJi/magcluster), code: magcluster prokka 19mtb_genomes; KofamKOALA server (https://www.genome.jp/tools/kofamkoala/) 10.1093/bioinformatics/btz859KEGG-Decoder10.1038/s41396-018-0091-3, code: KEGG-decoder -input 19mtb.txt -output out.list -vizoption static -m myorder.txt. Fig. 3b was produced using by Max UnMix and Fig. 3d was drawn using FORCinel v3.06 (https://wserv4.esc.cam.ac.uk/nanopaleomag/?page_id=51).

Competing interests

The authors declare no competing interests.