Giant multicellular magnetotactic prokaryotes in marine sediments
Elsa C A Turrini, Christian Godon, Marine Bergot, Béatrice Alonso, Sascha Lambert, Emilie Gachon, Nicolas Menguy, Stéphanie Fouteau, Stefan Klumpp, Christopher T Lefèvre, Caroline L Monteil
TL;DR
Scientists discovered a new type of giant multicellular bacteria in marine sediments that use magnetic crystals to navigate, expanding our understanding of prokaryotic multicellularity.
Contribution
The discovery of a new, much larger morphotype of multicellular magnetotactic prokaryotes and its classification into a new genus and species.
Findings
A new giant multicellular magnetotactic prokaryote was found in Mediterranean sediments, containing about 130 cells per consortium.
Each cell produces over 100 greigite magnetosomes arranged to optimize magnetic navigation.
The new species was classified as Magnetogigantoglobus mediterraneus within the Candidatus Magnetomoraceae family.
Abstract
Multicellular magnetotactic prokaryotes represent a unique group of obligately marine multicellular bacteria known for their ability to navigate along magnetic field lines thanks to ferrimagnetic nanocrystals. To date, two distinct spherical and ellipsoidal morphotypes have been described, typically ranging from 3 to 6 μm in diameter and comprising approximately 50 cells of the same species. Although widespread in highly reduced marine sediments, they are represented by solely three genera clustering into a monophyletic group within the Desulfobacterota. In this study, we report a third morphotype in reduced sediments of the Mediterranean Sea in Carry-le-Rouet, France, i.e. approximately 30 times more voluminous than any previously described form. Because their large size, we designated these multicellular bacteria as “giant” and explored their cell ultrastructure, ecological niche and…
Genes, proteins, chemicals, diseases, species, mutations and cell lines named across the full text — each resolved to its canonical identifier and authoritative record.
Click any figure to enlarge with its caption.
Figure 1
Figure 2
Figure 3
Figure 4
Figure 5
Figure 6
Figure 7
Figure 8- —Deutsche Forschungsgemeinschaft10.13039/501100001659
- —France Génomique and French Bioinformatics Institute national infrastructures
- —French National Research Agency10.13039/501100001665
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsGeomagnetism and Paleomagnetism Studies · Protist diversity and phylogeny · Polar Research and Ecology
Introduction
Alongside key innovations, multicellularity represents a major evolutionary milestone that has deeply contributed to the diversification of life on Earth [1–3]. This innovation, characterized by the assembly of often specialized, interconnected cells functioning collectively as a single organism, has emerged independently at least 25 times in the tree of life [4]. Although traditionally associated with eukaryotes, multicellularity also exists in prokaryotes, challenging earlier assumptions about their strictly unicellular nature [5]. James Shapiro stands as a pioneering figure in challenging this prevailing paradigm, highlighting the existence of bacteria that can be categorized as multicellular in terms of their structural cohesion, molecular communication, and division of labor [5]. For example, some filamentous bacteria are structurally linked by septal junctions which are involved in the transfer of nitrogen and carbon sources to ensure filament growth [6]. The fruiting bodies formed by Myxococcus xanthus or biofilms formed by Bacillus subtilis represent multicellular behaviors in response to external stress, manifested by coordination, three-dimensional organization, and chemical communication between cells [7, 8]. The distribution of this trait in the tree of life reflects the selective advantage it provides to the organism [9]. By producing its own internal environment, multicellularity has enhanced self-protection, division of labor, resource acquisition, and dispersal [10].
Multicellular magnetotactic prokaryotes (MMP) are the only prokaryotes known to exhibit an obligately multicellular life cycle [11–13]. MMP were described solely by culture-independent methods. They are found in the vicinity of the oxic–anoxic transition zone of sediments across various aquatic environments worldwide in the Southern and Northern Hemispheres (Brazil, USA, China, Germany, France) including lagoons, brackish waters, marine settings, and hypersaline, sulfide-rich environments [13–17], as well as in the water column of a chemically stratified pond [18, 19]. They are characterized by their distinctive morphology, which consists of a consortium of cells organized in a highly structured manner. This organization is frequently compared to a “raspberry” and can contain between 10 and 86 cells grouped around an acellular center insulated from the environment [13, 17, 20–23]. MMP can be divided into two main categories. The first one includes spherical MMP (sMMP), composed of cells measuring 3–6 μm in diameter, with round to triangular shapes. The second category consists of ellipsoidal MMP (eMMP), whose cells’ shape is more complex, including irregular elliptical conical frustums or H-shaped cells of ~9 μm in length and ~8 μm in width [13, 15, 24]. Each cell of the consortium is in contact with the acellular center, the neighboring cells and the environment. The surface in contact with the environment displays peritrichous flagella and ciliated structures. Likely composed of sugars and proteins, these structures resemble a capsule when observed under scanning electron microscopy (SEM) [15].
Previous ultra-thin section analyses have highlighted the internal complexity of MMP cells when observed by electron microscopy, revealing the presence of polyhydroxybutyrates granules and electron-dense nano-organelles organized in chains, called magnetosomes [15–17, 20, 23, 25, 26]. Like unicellular magnetotactic bacteria (MTB), MMP can synthesize single domain crystals of magnetite (Fe(II)Fe(III)2_O_4) and/or greigite (Fe(II)Fe(III)2_S_4), each enveloped by a lipid bilayer. The formation and alignment of these organelles are associated with numerous genes organized into multiple operons within a magnetosome gene cluster (MGC), and include gene sets specific to each chemical composition [27, 28]. Each cell contains one or more magnetosome chains, contributing to a total magnetosome count that can exceed 1000 per MMP [29]. The magnetic moment created by the chains guides the chemotaxis machinery to facilitate the navigation of magnetosomes-producing bacteria in chemically stratified aquatic environments [30]. However, the complex structure and cell organization raise many questions regarding how the magnetic properties of each cell determine MMP magnetism and motility. They display a distinctive “ping-pong” or “escape” motility pattern in a hanging drop assay, comprising an excursion phase and a return phase, thereby achieving an average motility between 37 and 90 μm.s^−1^ [15, 16, 22, 26]. “Escape” motility timing, distance and frequency can be modified depending on the magnetic field strength, thus suggesting the existence of an active sensory mechanism [31–33]. Disruption of the MMP into individual cells results in a loss of motility and magnetic responsiveness, ultimately leading to the loss of viability [12]. Motility is also influenced by light, as MMP can exhibit a photophobic response to blue, violet or UV light illumination [34, 35].
MMP constitute a small group of organisms with a limited number of species. They cluster into a monophyletic group composed of several genera affiliated with the Desulfobacterales order in the Desulfobacterota phylum [22, 23]. In GTDB [36], they form a family exclusively composed of multicellular organisms with the candidate name Magnetomoraceae. Although most of these multicellular bacteria are magnetotactic, non-magnetotactic forms with an ultrastructure similar to sMMP were observed in low-saline, nonmarine environments [37]. Despite the lack of a model in culture, culture-independent approaches and co-cultivation enabled to investigate the physiology of MMP, showing that they are sulfate reducers and may use acetate, propionate, succinate and/or bicarbonate as carbon sources and/or electron sources [20, 22, 23, 38]. Phylotyping of several MMPs based on their 16S rRNA gene sequences revealed that each MMP consortium is composed of cells belonging to the same species and reproducing by binary fission of the assembly [39]. However, recent analyses coupling metagenomics and nanoscale secondary ion mass spectrometry challenge this paradigm, supporting genetic diversity and metabolic differentiation within the same cell consortium [23].
Despite significant advances in our understanding of MMP functioning, fundamental questions remain unanswered largely due to the absence of available cultured models and the difficulty in observing phenomena at the single cell level. To date, the cellular and molecular mechanisms regulating the formation, differentiation, and coordination of cells within multicellular structures in prokaryotes are still poorly understood. However, the advent of culture-independent approaches has provided an alternative way to study these organisms, offering valuable insights into their ultrastructure, functioning, and the discovery of new diversity. In this study, we report a group of MMP that are nearly 30 times more voluminous than sMMP, referred to as giant multicellular prokaryotes (gMMP). Discovered in seawater sediments from Carry-le-Rouet, France, they represent a previously undescribed genus within the Candidatus Magnetomoraceae family. Through detailed characterization of its ultrastructure, taxonomy, and genome annotation, this study deepens our understanding of both multicellularity and magnetotaxis, shedding light on the evolutionary and functional adaptations of these organisms.
Materials and methods
Site description and sample collection
Sampling was carried out between March and April 2024, collecting surface sediments by freediving in the Rouet Marina, Carry-le-Rouet, France (43.333467°N, 5.172837°E). This marina is sheltered from the currents and waves by stone walls and water temperature fluctuates from 12°C to 25°C according to the season. Sediments are composed of fine sand and small pebbles ranging in color from yellowish to darkish hues. We used this color heterogeneity to track potential organic-rich and reduced zones and to compare the MTB diversity across different types of sediments under the light microscope. Each 1-liter glass bottle was filled with 300–400 ml of sediment and 600–700 ml of overlying water, ensuring no air bubbles were present. In the laboratory, mesocosms were stored with caps loosened under low light conditions at ambient temperature (~20°C). Abundance and diversity of MMP populations in the mesocosms were regularly checked over time.
Magnetic enrichment and light microscope observation
A magnetic stirring bar (~10 mT) was placed adjacent to the bottles for 3 h to concentrate north-seeking magnetotactic cells. The magnet was positioned just above the sediment-water interface and concentrated cells were then examined using the hanging drop technique [40] on a Zeiss Primo Star light microscope with phase-contrast contrast optics. Magnetotactic orientation was confirmed by rotating a stirring bar magnet 180° on the microscope stage to invert the magnetic field. Motility and magnetotactic behavior were also observed and recorded under a Leica LMD6000 light microscope with a Leica DMC 4500 camera (Leica Microsystems GmbH, Germany).
Cell sorting and whole-genome amplification
A dozen gMMP and representative sMMP were sorted from sediment samples as described previously (Supplementary Methods S1) [41]. To obtain enough genomic DNA (gDNA) for 16S rRNA gene sequencing and shotgun metagenomics, whole genome amplification was conducted via multiple displacement amplification using the REPLI-g Single Cell Kit (QIAGEN) following the manufacturer’s protocol. The concentration of double-stranded gDNA was then quantified with a Qubit 4 fluorometer (ThermoFisher Scientific).
Cloning and sequencing of the 16S rRNA gene of magnetically concentrated and sorted cells
The 16S rRNA gene from each DNA sample was amplified using Phusion Hot Start Flex DNA Polymerase according to the manufacturer’s guidelines, with a template concentration of 50–70 ng/μL, and the primers 27F (5′-AGAGTTTGATCMTGGCTCAG-3′) and 1492R (5′-TACGGHTACCTTGTTACGACTT-3′). These blunt-ended 16S rRNA gene amplicons were cloned using the Zero Blunt TOPO PCR Cloning Kit and One Shot TOP10 chemically competent Escherichia coli cells. At least five clones from each single-consortium were sequenced by Eurofins Genomics, Germany, GmbH. Identical sequences were pooled into operational taxonomic units (OTUs). Chimera detection was conducted using the UCHIME2 algorithm [42]. Sequences were aligned using the BLASTN algorithm with default parameters against the NCBI nucleotide database [43] to identify the most similar public sequences. Finally, five 16S rRNA gene sequences representing gMMP and sMMP of Carry-le-Rouet were deposited in GenBank under unique accession numbers PV085504-PV085508.
Fluorescent in situ hybridization
The 16S rRNA gene sequence obtained by PCR or from genome assembly was validated by fluorescence in situ hybridization (FISH) conducted following a previously published protocol [44]. An ATTO488-labeled oligonucleotide probe, DSBAC357p-ATTO 488 (5′- CCATTGCGCAAAATTCCTCAC-3′), was used to specifically target gMMP cells within the Desulfobacterales order [45], while a second probe, EUBp-ATTO 633 (5′-GCTGCCTCCCGTAGGAGT-3′), was employed to broadly target bacterial cells [46, 47]. A complete description of the protocol and probes is given in Supplementary Methods S2.
Membrane labelling
The labelling procedure was carried out to distinguish cells’ membrane under the confocal microscope. A 20-μL drop of magnetically concentrated cells was placed on a 20 mm × 20 mm coverslip coated with poly-L-lysine to promote cell adhesion. After allowing the cells to settle, the drop was carefully removed, and the adhered cells were fixed overnight with 30 μL of an 8% (w/v) paraformaldehyde solution prepared in PBS. Following fixation, the cells were incubated for 10 min on ice with FM 1-43 FX, a membrane dye analog (Thermo Fisher Scientific), in a solution of Hanks’ Balanced Salt Solution (HBSS) [48] to stain the cellular membranes. The coverslip was then rinsed three times with HBSS to remove excess dye. Finally, the coverslip containing the sample was mounted with a ProLong Diamond Antifade Mountant (Invitrogen) with 4,6-diamidino-2-phenyl-indole (DAPI) and observed using a Zeiss LSM980 confocal microscope.
Confocal microscopy
Imaging was performed using a Zeiss LSM 980 confocal microscope with a 63×/1.4 NA oil immersion lens. DAPI and FM -43 were excited with lasers 405 nm and 488 nm, respectively. Dyes ATTO 488 and ATTO 633 were excited respectively with lasers 488 nm and 639 nm. Images acquired in Airyscan Super-Resolution mode using the Airyscan 2 GaSP detector were processed using the Airyscan joint deconvolution (jDCV) plugin of the ZEN blue software (Zen 3.6)—using “sample structure” option standard—to improve spatial resolution and clarity. The ImageJ software was used to perform brightness and contrast adjustment.
Scanning electron microscopy
SEM was used to get insights into the gMMP morphology. Magnetically concentrated cells were (i) chemically fixed with 2.5% glutaraldehyde, (ii) transferred onto a polycarbonate filter with 0.2 μm pores within a Swinnex support system, (iii) dehydrated in successive ethanol baths (i.e. 50%; 70%; 96%; 100%), (iv) dried through critical point drying (CPD) (Leica EM CPD300), and finally (v) coated with carbon using a Leica EM SCD500 carbon coater. The analyses were performed with a Zeiss Ultra 55 field emission gun SEM. Secondary electron images were acquired using an InLens detector and an accelerating voltage of 1–3 kV with a working distance of 3.5 mm. Backscattered electron imaging (AsB detector) was conducted at an accelerating voltage of 10 to 15 kV and a working distance of 7.5 mm.
Transmission electron microscopy (TEM)
Preparation of transmission electron microscopy (TEM) copper grids consisted of depositing a 2-μL drop of magnetically concentrated bacteria. Using their magnetotactic abilities, bacteria were attracted for 15 min to the edge of the drop by placing a magnet near the appropriate side of the grid. Then, cells were fixed onto the TEM grids covered with a carbon film and poly-L-lysine to improve cell adhesion. Grids were then rinsed with filtered Milli-Q water. Ultrathin (~100 nm thickness) sections of bacterial pellets were also prepared by ultramicrotomy following the protocol detailed previously [41], and described in the Supplementary Methods S3. Selected-area electron diffraction (SAED), Z-contrast imaging in the high-angle annular dark field mode, and elemental mapping by x-ray energy-dispersive spectrometry (XEDS), were carried out using a JEOL 2100 F microscope to get insights into the mineralogy of magnetosomes (Supplementary Methods S4).
Magnetotactic response analysis
Motility was analyzed in response to changing magnetic fields in the water extracted from the samples in which MMP were collected. Consortia were tracked with a customized magnetic microscope equipped with a triaxial Helmholtz coil set and controller (C-SpinCoil-XYZ, Micro Magnetics Inc.), and an Andor Zyla 5.5 high-speed camera [49]. The 3D-axis Helmholtz coils can generate DC magnetic fields with a precision of 5% of Earth’s magnetic field (±2.5 μT). Using this setup, we programmed the switching of the magnetic field between −3.5 and +3.5 mT for the U-turn and measurements were performed as described in the Supplementary Methods S5. Adaptation of the arrangement of the gMMP’s magnetosome chains for magnetotaxis was tested with computational optimization, using the differential evolutionary algorithm [50], see Supplementary Methods S6.
Shotgun genomic sequencing, assembly and binning
Three single consortia of gMMP from a sample collected at the Rouet Marina in March 2024 were sorted for DNA amplification, shotgun genomic sequencing, assembly, and binning as previously described [41], leading to three single-consortium metagenomes (SCM). A complete procedure is given in Supplementary Methods S7. Genome completeness of the different bins from each SCM was evaluated using CheckM v2 lineage workflow [51].
Genome-based taxonomic classification and molecular phylogeny
Phylogenetic trees were built from the whole genomes and 16S rRNA gene sequences as previously described—see the complete procedure in Supplementary Methods S8 [41]. First, GTDB-Tk v2.1.1 [52] was used to assign each bin to a taxonomic group based on the Genome Taxonomy Database (GTDB) classification [36]. Phylogenomic trees were built using the maximum likelihood method implemented in the IQ-TREE v2.2 software [53, 54], a concatenated alignment of conserved markers used by GTDB [36] and a substitution model selected by ModelFinder [55]. The average amino-acid identity (AAI) value was also used for genus delineation [56, 57].
Functional annotation, computation of multicellular magnetotactic prokaryotes pan-genome, and metabolic network modeling
Genomes were also analyzed using the MicroScope platform [58] for further comparative analysis of the magnetosome gene clusters (MGC)s and expert functional annotation. We used well-annotated Desulfobacterota reference genomes, such as Desulfamplus magnetovallimortis BW-1 [27], to reconstruct the MGC in MMP genomes. Protein-coding genes were classified into Clusters of Orthologous Groups (COGs) using eggNOG-mapper v2.1.12 [59] and eggNOG version 5.0.2 [60]. Metabolic pathways were inferred using the MetaCyc database v27.0 [61] and a workflow using the PathoLogic algorithm of Pathway Tools [62] as previously described (Supplementary Methods S9) [41].
Pan-, core-, and variable-genomes of MMP used in this study were inferred with the Microscope platform as described previously [63]. In summary, the platform creates gene/protein families (MICFAMs) that are computed incrementally each time a new genome is integrated. MICFAMs classify proteins in homolog groups using a single-linkage clustering implemented in the SiLiX software [64] of protein sequence pairs sharing at least 80% of AAI and 50% of alignment coverage.
Results
Observation of a multicellular magnetotactic organism with a volume approximately 30 times larger than that of spherical multicellular magnetotactic prokaryotes in Carry-le-Rouet
Sediment samples were collected at an average depth of one meter from the Rouet Marina in Carry-le-Rouet, France, to prospect for new magnetotactic organisms. Magnetic enrichments of reduced, sulfide-rich black sediments produced a black pellet made of various MTB morphotypes (Fig. 1A). Abundant north-seeking multicellular bacteria were observed alongside single-celled MTB (e.g. cocci, vibrios). Although sMMP represented the most dominant morphotype, with rare eMMP, several larger, previously undescribed MMP-like organisms caught our attention. Like the sMMP, this morphotype was also spherical (Fig. 1B), but exhibited an unusually large size in comparison, which is why we referred to them as “giant” MMP (gMMP). Contrary to sMMP described in the literature [22, 33], most of gMMP did not accumulate at the edge of the hanging drop during observation. They seemed to be sensitive to air, as most of the cells ceased moving shortly after being transferred to the hanging drop device or when observed between the slide and coverslip (Video S1 & S2). The gMMP abundance was low, with up to 100 consortia observed per hanging drop, compared to sMMP and other single-celled MTB (approximately 10^4^–10^5^ cells/hanging drop), making their observation and characterization more challenging. However, this limitation was partially offset by their large size.
Observation of north-seeking MMP sampled in the sediments of Carry-le-Rouet. (A) A 1-liter bottle was filled with one-third of sediment and completed with seawater collected from the Rouet Marina in Carry-le-Rouet. The sample was magnetically concentrated for 3 h using the south pole of a magnetic stirring bar, resulting in the formation of a black pellet containing MTB. (B) Light microscope image of the edge of a hanging drop harvested from the pellet observed in panel A. Two distinct MMP morphotypes are observed: the spherical sMMP (white arrow) and the giant gMMP (black arrow).
MMP consortia sizes were estimated using light microscope observations. They are 16.2 μm ± 0.99 (n = 20) in diameter, representing an average volume of 2220 μm^3^ if we consider them as perfectly spherical. In contrast, sMMP populations from the same sample are approximately 30 times less voluminous than gMMP, as they are 5.22 μm ± 0.48 (n = 20) wide for an average volume of 73.50 μm^3^. SEM observation of sMMP, eMMP, and gMMP present in sediments of the Rouet Marina gave further insights into the difference between these three morphotypes (Fig. 2). gMMP surface exhibit an irregular topography, characterized by numerous interconnected cells bridged by filamentous structures (Fig. 2A and B). Some cells were disrupted during microscopy preparation, revealing the conical morphology of individual cells shaped like a wallflower with a slightly concave depression in its center (Fig. 2B & Fig. S1A). An increase in voltage revealed electron-dense structures close to the surface reminiscent of magnetosomes (Fig. 2C). All cells are aggregated into a spherical morphology analogous to a raspberry with an acellular core (Fig. 3A and B). Fluorescent labeling of gMMP with FM 1-43 membrane dye, followed by confocal microscopy analysis and 3D reconstruction, revealed a complex cellular architecture comprising 130 ± 22 individual cells (n = 6), each measuring 5.99 μm ± 0.95 (n = 60) in length. Confocal 3D reconstruction and TEM images of thin sectioned cells showed also that each cell is connected to the acellular central compartment 4.21 μm ± 1.35 (n = 10) in diameter. Each gMMP cell contained numerous peritrichous flagella emerging from the peripheral membrane surface (Fig. S1). Transmission electron microscopy observation of gMMP ultrathin sections not only further validated morphological features observed with the light and confocal microscopes but also provided additional information on cell ultrastructure (Fig. 4). Cells’ cytoplasm is structured into multiple compartments, some being of unknown nature. This includes electron-lucent inclusions and high-density nano structures later identified as magnetosomes arranged next to the membrane in contact with the environment. This external membrane is covered by cilia-like structures (Fig. 4C), whereas the internal membrane is associated to many vesicles at the junction between the cells and inside the acellular central region (Fig. 4D). Apart from magnetosomes, none of the cytoplasmic compartments seem to be electron dense. Staining with DAPI suggests that some of them, specifically located in the central region of cells’ cytoplasm, could be composed of DNA (Fig. 3 & Video S3).
SEM images of MMP. (A) SEM image montage of sMMP, eMMP, and gMMP collected in Rouet Marina, Carry-le-Rouet, France, observed at 1 kV, showing the relative size of the different multicellular prokaryotes. (B) SEM image of a partially disrupted gMMP observed at 1 kV, revealing the morphology of individual cells. (C) SEM image of a gMMP observed at 10 kV using the backscattered mode showing the magnetosome chains inside the individual cells of the gMMP consortium. The scale bars represent 2 μm.
Confocal microscopy images of the gMMP complex cellular architecture showing membranes and DNA. (A) Transmitted light images of the same gMMP observed at the same Z-stack when the focus is done at the surface (first line), at the 1/3 of the surface (second line) and at the center of the Z sections (third line). Same images showing membranes stained by FM1-43 (B), and DNA stained with DAPI (C). Panel (D) represents merged confocal and transmitted light microscopy images. The 3D reconstruction of confocal microscope images corresponding to the superposition of 150 Z-stacks of the whole gMMP stained by FM1-43 and DAPI is presented in Video S3.
TEM images of ultrathin sections of gMMP. (A) TEM image of a central section of a gMMP, showing the spatial distribution of gMMP cells surrounding a central acellular region. (B) Detailed view of few gMMP cells, highlighting the presence of various intracellular structures of distinct electron contrast. (C) Enlarged view of a gMMP periphery illustrating key features such as surface cilia (white arrow), magnetosomes (electron-dense and aligned particles), possible electron-lucent granules (asterisk), and vesicles (thick black arrows). (D) TEM image focused on the internal extremities of cells within the acellular region, revealing numerous small vesicles dispersed thorough this area (black arrows).
Giant multicellular magnetotactic prokaryotes magnetotaxis exclusively based on greigite biomineralization
TEM observations revealed that gMMP cells biomineralize pleomorphic magnetosomes measuring 75 ± 13 nm in length and 64 ± 11 nm in width (n = 159) (Fig. 5). A gMMP cell contains 105 ± 25 magnetosomes on average, which represents 1.7 × 10^4^ ± 3.9 × 10^3^ magnetosomes per gMMP (n = 23). Approximately 79% of the magnetosomes are in the first micrometer of the exterior side of the cell and generally organized into two to three chains perpendicular to the gMMP radius (Fig. 4 & Fig. S2). These observations were further confirmed by SEM analyses (Fig. 2C). The observed standard deviations are relatively high, which can be explained in part by the fact that some cells are in an intermediate state of cell division, while others are not. The nature of magnetosomes has been unambiguously identified through combined SAED measurements (Fig. 5C and D) and STEM-XEDS elemental mapping (Fig. S3).
Characteristics of magnetosomes observed in the gMMP. (A) TEM image showing three gMMP cells detached after the micromanipulation of a single gMMP and deposition onto a TEM grid. (B) TEM image focusing on aligned magnetosome chains located at the periphery of a gMMP cell. (C) TEM image of a magnetosome chain showing at the dashed white circle a single magnetosome analyzed using SAED. (D) SAED pattern of the crystal shown in panel C viewed along the [110] zone axis, consistent with greigite (space group Fd3;‾m* and a = 9.876 Å). (E) Distribution of magnetosome crystals, with their average length, width and shape factor (n = 159). On average, 79% ± 7% of the magnetosomes are in the first micrometer of the exterior side of the cell (n = 22).*
Motility was analyzed in response to changing magnetic fields in the water extracted from the samples in which gMMP were collected. On average, the swimming speed of gMMP consortia was 49.76 ± 5.6 μm/s (n = 43). Consortia aligned and swam along the field’s lines with a rolling motion, which is a swimming behavior similar to what has been previously described for other MMP [17]. The average magnetic moment was estimated by tracking the positioning of a collection of 20 gMMP consortia and monitoring the time required for them to reorient when the direction of a 3.5 mT field is reversed (Video S4). A magnetic moment of m = 1.48 ± 10^−14^ A⋅m^2^ was calculated. This value is greater by a factor of two than any other MTB estimated to date. However, as gMMP experience greater interactions with surfaces than other bacteria, their swimming behaviors can be affected and consequently, their swimming speed and magnetic moment, underestimated.
Organization of magnetosome chains optimizes net giant multicellular magnetotactic prokaryotes magnetization
We predicted that the specific magnetosome chains’ organization at the periphery of each gMMP cell could optimize their net magnetization compared to a radial distribution closer to the center of the consortia. To test this hypothesis, simulations of magnetic dipole–dipole interactions within the gMMP geometry were conducted. Numerical optimization of magnetosome chain position and orientation, was performed based on the assumption that magnetosome chains align with membranes of the gMMP cells either (i) radially in the interior of the gMMP or (ii), parallel to the gMMP consortium’s surface. Calculations show that the energetically optimal configuration has all dipoles gathered near the center of the gMMP, with alternating orientations (Fig. 6). This configuration shows very low net magnetization. However, when constraining the dipoles to the outer layers of the gMMP (Fig. 6), a parallel surface pattern reminiscent of the organization of magnetosome chains in the confocal images of gMMPs was observed (Figs 4, 5 & Fig. S2). Under this constraint, the surface dipoles are energetically favorable compared to the remaining possible interior configurations. This configuration has significantly higher net magnetization than the interior configuration (Fig. 6). The preference for a surface dipole configuration is enhanced by a magnetic field of strength comparable to the geomagnetic field (Fig. 6 & Fig. S4). Based on these calculations, the net magnetization of the gMMP is expected to be between 5% and 40% of the maximal magnetization achievable if the dipole were unconstrained. These observations suggest that the observed organization of magnetosome chains near the exterior gMMP surface supports a higher net magnetization, without requiring the organism to overcome interaction stresses when building new magnetosomes. The required constraint to the exterior (70% of the gMMP radius) exceeds the purely geometric constraint arising from the hollow center of the gMMP, so additional constraining forces are likely at play.
Numerical optimization of the dipole configuration. Simulations of magnetic dipole–dipole interactions within the gMMP geometry (grey sphere) were performed. Two configurations with N = 100 dipoles (red arrows), constrained to be either radially oriented or parallel to the gMMP surface, were optimized for their magnetic energy. A cone represents a gMMP cell. For all magnetic field (B) strengths, the interior configuration (left; generated with a cutoff radius of 0.1, see Supporting Methods S6) has lower magnetization and lower energy than the exterior configuration (right; generated with a cutoff radius of 0.9). The difference in magnetization is particularly pronounced at the highest magnetic field (B = 40, the full B*-dependence is shown in Fig. S4). For all conditions, the distributions were obtained by repeating the optimization 100 times with randomized initial conditions.*
Giant multicellular magnetotactic prokaryotes are affiliated with a new genus in the Candidatus Magnetomoraceae family
gMMP identity was established through sequencing of the 16S rRNA gene and whole genomes of representative consortia. Several consortia of gMMP and sMMP were independently sorted from several magnetic pellets using a micromanipulator, which enabled to specifically associate a 16S rRNA gene sequence and taxonomy to a morphotype. A total of 13 sequences were successfully retrieved (i.e. 10 from gMMP and three from sMMP), ranging in length from 1450 to 1492 bp, representing three single OTU: two OTU of sMMP and one OTU of gMMP. Sequences were aligned against the Core Nucleotide Database and Reference RNA sequences of the NCBI using the BLASTN algorithm with default parameters. While the two sMMP sequences shared more than 95% of identity with each other and more than 97% with described MMP species, gMMP shared <93% with any uncultured Desulfobacterales and 91% with sequences of sMMP observed in the same sample. This affiliation to the Desulfobacterales order was further validated using FISH and the oligonucleotide probe DSBAC357 specific to this group (Fig. S5) [46].
We used whole genomes to define more accurately the genetic relationships between gMMP, other sequenced MMP, and single-celled Desulfobacterales species. Three amplified DNA samples from single consortia of gMMP were randomly picked for shotgun high-throughput sequencing. Three nearly identical 4.2–4.3 Mb high-quality genomes of gMMP (i.e. completeness of 91% with 0.65% redundancy) were fully assembled, each containing a single copy of 16S rRNA gene 100% identical to OTU sequences obtained from PCR (Table S1). GTDB-tk affiliated the genome of three SCM (i.e. gMMP-1, gMMP-13, and gMMP-15) to an undescribed genus within the Ca. Magnetomoraceae family based on the relative evolutionary divergence value. A maximum-likelihood tree of the Desulfobacterales order was built using the 115 out of the 120 informative markers used in GTDB and using the group C00003060—also known as the lineage SEEP-SRB1c [65]—as an external group based on a previous high-quality Desulfobacterota phylogeny (Fig. 7 & Fig. S6) [66]. The resulting tree topology is well supported, particularly at the deep internal nodes, allowing clear resolution of the relationships among families. In this tree, gMMP genomes form a divergent group within the Ca. Magnetomoraceae family apart from the Magnetoglobus and Magnetomorum species, with which they share <59% average AAI of coding sequences (Fig. S7). Given the novelty, we propose that genomes of the gMMP from the Mediterranean Sea serve as the nomenclature type for the species Magnetogigantoglobus mediterraneus according to the rules of the SeqCode [67] (Supplementary Results S1). Our phylogenomic analysis provides compelling evidence for its classification as a member of the Magnetomoraceae (fam. nov.) in accordance with the rules of SeqCode, corresponding to the Ca. Magnetomoraceae [68]. Since no eMMP genome is available, we also built a 16S rRNA gene-based tree to get insight into the relationships between gMMP and species related to Candidatus Magnetananas (Fig. S8) [13, 14, 21]. Despite the limitations of using such a marker for resolving genetic relationships, we also obtained well-supported internal branches linking the three morphotypes to each other. The tree shows that eMMP and sMMP share a more recent ancestor than either does with gMMP.
Taxonomic affiliation of gMMP and analysis of their MGCs. (A) Minimal maximum-likelihood tree of the Desulfobacterales inferred from the concatenation of 115 conserved marker proteins. The tree was inferred from representative strain genomes only and one single-amplified genome of the magnetic epibiont Ca. Desulfarcum epimagneticum. Branch lengths represent the number of substitutions per site. Tree topology was tested using an ultrafast bootstrap approximation approach with 1000 replicates [89]. The gray circles represent bootstrap values >95% and are drawn to scale. Branch colors indicate family-level classification according to GTDB [36]. An expanded version including metagenome-assembled genomes representing all Desulfobacterales families, together with full taxonomy assignments and accession numbers, is given in Fig. S6. (B) MGC synteny in a representative gMMP sequenced in this study. Each arrow represents a gene. Homologue families were determined based on sequence similarity and the presence of conserved domains using the Microscope platform [58]. gMMP genes were classified according to their homology with mam, mad, and greigite genes in the reference genome BW-1 [27]. Most of the BW-1 magnetite gene cluster located upstream of the greigite gene cluster is not shown, as no homologues were detected in gMMP. Grey genes are genes of unknown function or not conserved in MTB. Slashes represent truncations mainly, and sometimes regions spacing two operons. Sequence identities between reciprocal best hits were estimated with MMseqs2 [90] and are represented by bands, with their intensity reflecting the percentage of identity. A full version of this figure with representative sMMP and eMMP species is given in Fig. S9. (C) Distribution of MMP with the capacity to biomineralize magnetite only, greigite only, both minerals or none of them. Groups are organized according to the phylogenetic tree based on the 16S rRNA gene given in Fig. S8.
Giant multicellular magnetotactic prokaryotes genomes harbor genes involved in multicellularity and sulfate reduction coupled to both heterotrophy and autotrophy
We performed a functional annotation of the three gMMP genomes with EggNOG mapper [59] (Table S2). The most represented proteins with a known function, were classified in the categories “Signal transduction mechanisms” (~8%), “Cell wall/membrane/envelope biogenesis” (~7%), “Energy production and conversion” (~5%), and “Motility” (~5%). Several proteins classified in the COG categories “Cytoskeleton” or “Extracellular structure” (Z and W, respectively) could be associated with the multicellular lifestyle. To go further, a comparative genomics analysis was conducted to identify additional genes conserved in all MMP and sMMP genomes, or specific to gMMP-1 (Supplementary Results S2 and Table S3). As for the sMMP, gMMP genomes encode multiple proteins likely involved in cell–cell adhesion, cell signaling, and extracellular matrix organization in many bacteria, that were not found in representative free-living Desulfobacterales used in the whole genome phylogeny. The genes encoding many of these proteins were found in the Ca. Magnetoglobus multicellularis genome previously [38], and are putative adhesion-like proteins (ENOG502NRZP; ENOG502778B), fibronectin type III domain-containing proteins (containing the IPR036116 domain), filamentous hemagglutinin outer membrane proteins (ENOG504PM4Q; IPR011049, IPR008640, IPR030392), cadherin-like proteins (IPR015919), H-type lectin domain-containing proteins (IPR019019), cell surface glycoproteins (e.g. IPR002035), and integrin-like repeat-containing proteins (IPR010221). However, we identified additional core protein families with no predicted known domains, functional sites, repeats, or motifs (Supplementary Results S2 and Table S3) that are likely at play. We also identified an MGC with the Microscope platform [58] with a full greigite operon and mad genes as described earlier in other greigite producers such as Ca. Magnetoglobus multicellularis (Fig. 7B) [28]. The absence of a magnetite operon implies that they should not be able to form magnetite. Gene synteny is consistent across MMP and other Desulfobacterales forming greigite such as BW-1 (Fig. S9).
MMP genera have similar repertoires of predicted metabolic pathways, most of them having the putative ability to use the same electron donors and acceptors (Fig. 8, Table S4 & Table S5). sMMP and gMMP genomes encode pathways for the utilization of various organic substrates as carbon sources or electron donors (e.g. fumarate, pyruvate, acetate, glucose, citrate, or L-cysteine), implying a potentially chemoheterotrophic metabolism during respiration and fermentation processes. The detection of an aerobic [NiFe]-hydrogenase (EC 1.12.99.6) support that they are H_2_-oxidizing in microaerophilic conditions (Table S4). All MMP are sulfate-reducing organisms, as evidenced by the presence of genes coding for key enzymes such as sulfate adenylyltransferase (sat) (EC 2.7.7.4), adenylyl sulfate reductase (aprAB) (EC 1.8.99.2), and genes coding for dissimilatory sulfite reductase (dsrAB) (EC 1.8.99.22) organized in a operon with drsD (Table S5) [69]. However, selenate could serve as an alternative terminal electron acceptor. We also identified genes involved in the carbonyl branch (acsAB genes; EC:1.2.7.4) and the methyl branch (fdh, folD, metF, and acsE genes) of the Wood–Ljungdahl (WL) pathway [70] (Table S4). Although a key gene coding for the formyltetrahydrofolate synthetase (fhs gene; EC 6.3.4.3) is missing, the H_4_folate-dependent methyl branch is almost complete. Together with the presence of a H^+^/Na^+^-translocating ferredoxin-NAD^+^ oxidoreductase (EC 7.1.1.11/7.2.1.2), it supports that gMMP have the genetic potential to perform reductive CO_2_ fixation. However, the WL pathway is also likely able to operating in reverse supporting central metabolism through the oxidation of organic compounds during organo-heterotrophic growth. The autotrophic capabilities of gMMP could be supported by a non-canonical form of the TCA cycle that uses the 2-oxoglutarate synthase (EC 1.2.7.3) and can operate in reverse to fix CO_2_ in some environmental conditions [71].
Comparative analysis of metabolic pathways predicted in all MMP, including one of the gMMP genomes, using the MetaCyc database. Heatmaps representing metabolic pathways involved in energy metabolism and assimilation (A), utilization or degradation of organic and inorganic compounds (B), respectively. Matrices of metabolic pathways are organized according to a maximum-likelihood tree of the Magnetomoraceae family built as described in Fig. 7. Only high-quality genomes representative of MMP species were used for the comparison. The Desulfatibacillaceae family was used as an outgroup based on the tree in Fig. S6. Metabolic pathways are also organized using a hierarchical clustering analysis (Euclidean distance) according to their pairwise distance. Absence of prediction can be linked to the absence of a single reaction/enzyme/gene mandatory for the pathway prediction. Yet, this absence can be a false negative and be linked to the quality of the draft genome assembly. The full analysis of metabolism is given in the Table S4.
Few non-essential predicted metabolic pathways were unique to gMMP or sMMP*.* For instance, Magnetogigantoglobus seems to be unable to use succinate as carbon source and electron donor unlike Ca. Magnetoglobus related species. In contrast, certain pathways were solely predicted in gMMP, including the ability to utilize D-mannose, acrylate, or propanoyl-CoA as carbon sources and the ability to degrade some amino acids (Fig. 8, Table S3). Magnetogigantoglobus also stands out for its potential ability to convert atmospheric nitrogen into ammonia anaerobically through a multiprotein nitrogenase complex that transfers electrons from a ferredoxin to a dinitrogenase [72] (EC 1.18.6.1). In addition, they seem to be able to ferment L-lysine as a source of carbon and nitrogen or produce NADH and CO_2_ via the 2-oxoisovalerate decarboxylation to isobutanoyl-CoA pathway (Table S3).
Discussion
MMP represent a remarkable exception among the various identified multicellular forms in the Bacteria domain, both by their structure and their development devoid of a unicellular phase. Since the early 1980s, studies have been largely limited to the unique uncultivated models within the Desulfobacterales: the spherical type (i.e. sMMP), which was the first characterized, and the ellipsoidal type (i.e. eMMP), which was identified more recently [73, 74]. Here, we shed light on a undescribed group of MMP represented by the species Magnetogigantoglobus mediterraneus. Characterized by their unusually large cellular volume, gMMP reach ~2220 μm^3^, about 30-fold higher than that of previously characterized sMMP [15, 75] (Figs 1B & 2A). This feature makes gMMP the biggest magnetotactic bacterium described to date, but not the biggest bacterial species, as much bigger microorganisms have already been described, such as Achromatium oxaliferum (5–100 μm) or Ca. Thiomargarita magnifica [76]. The distinctive raspberry-like architecture formed by individual cells is characteristic of MMP, as well as the presence of lipid bilayers, peritrichous flagella, surface-associated cilia, intracellular electron-lucent inclusions, and magnetosome organelles at the cell periphery (Figs 2–4). Size and cell numbers are thus almost the only distinctive morphologic traits differentiating them from sMMP, to which they resemble the most.
gMMP complexity brings up the same fundamental questions that were raised about sMMP decades ago: how do they operate as cohesive units in their environment, particularly with respect to magnetotaxis? A model was previously proposed in M. multicellularis, in which each sMMP cell contained hundreds of magnetosomes organized in multiple chains within each cell, positioned at a distance >70% of the radius of the microorganism and close to the cell surface [75], which are even closer to the exterior in gMMP. In contrast to sMMP, no striated structure was observed near magnetosomes in gMMP [75], but we observed the same patterns of magnetosome organization into clusters at the cell surface (Figs 2C, 4A–C & Fig. S2). This suggests that they are stabilized and anchored through comparable interactions with the membrane, thereby generating a net magnetic moment and a defined axis of movement [75]. The magnitude and orientation of the magnetic dipole moments in the cells were not determined in any of the MMP genera. Nevertheless, in the intact organism, they would add vectorially to give a net magnetic dipole moment that would orient the organism in the ambient field [77]. Here, we showed that this configuration theoretically optimizes the net magnetization and is greater than that required for alignment with the Earth’s magnetic field (Fig. 6), as we observed previously in magnetotactic holobionts [78].
The structural integrity of a multicellular organism depends on the maintenance of stable cellular connections with precise spatial arrangements [79]. Although each cell within the MMP appears independent of each other, several features suggest the presence of intercellular cohesion, potential cell-to-cell communication, and adhesion mechanisms, for which we provide candidate genes. For example, vesicles have been identified at intercellular junctions and near the acellular core region, resembling those observed in other MMP that may contribute to maintaining the organism’s cohesion (Fig. 4C and D). In Ca. M. multicellularis, carbohydrate and calcium cytochemistry analyses showed that a polysaccharide layer holds cells together and that calcium-dependent adhesion molecules are present between the cells [75]. Here, we identified lectin domain-containing proteins (IPR019019) and other proteins in gMMP genomes that could be involved in the same processes. However, the exact set of genes driving MMP multicellularity remains unknown, as some of the mechanisms are likely controlling broad interactions with other bacterial species too.
The high-quality genomes of M. mediterraneus obtained in this study enabled to predict an overall physiology comparable to that previously deduced from sMMP genomes (Fig. 8). Although MMP form distinct genomic genera, the family Magnetomoraceae contains a globally identical functional inventory regardless of the sampling location. This finding suggests that all MMP, including those from Carry-le-Rouet, could exploit the same niche. In addition, the three morphotypes were observed in the same samples, all recovered from black sediments emitting a characteristic sulfurous odor, indicative of an anoxic, reduced environment. These conditions are similar to those found in saline and brackish environments, where other MMPs have been observed previously, and where sediments are characterized by high sulfide concentrations ranging from 0.1 to 1 mM [19]. As other Magnetomoraceae genera, M. mediterraneus is a sulfate reducer using small organic acids and hydrogen as electron donors, but can also perform inorganic carbon fixation. Such metabolism releases H_2_S and could explain the systematic assembly of a second bin of low-quality corresponding to an undescribed species of sulfur oxidizing bacteria, Thiomicrorhabdus (Table S1) [80]. Although no structure reminiscent of an intracellular symbiont was observed, the nature of this association, symbiotic or not, and the origin of the DNA presence could not be properly investigated in this study. Ca. Magnetoglobus multicellularis has previously been described to be in association with Pseudoalteromonas species. Cells of Pseudoalteromonas species were observed attached at the surface and in the intercellular space of the sMMP although the precise nature of this association was unclear [81].
Despite this diversity pattern and the co-occurrence of the three MMP morphotypes in the same samples, one can hypothesize that there is a subtle difference in the lifestyle, leading eventually each group occupying different depths, that remain to be determined. One clue can be found in predicted metabolic pathways repertoire indicating that gMMP could generate energy from alternative carbon sources compared to other MMP. For example, they likely degrade acrylate, a toxic compound found in certain environments, which can be reduced by anaerobic bacteria like Desulfovibrio, particularly in anoxic conditions [82]. More insightful, however, is the capacity to degrade a broader range of amino acids compared to other MMP, which may suggest a greater propensity for a heterotrophic lifestyle based on the breakdown of complex organic compounds. Such ability would provide a selective advantage by enhancing adaptation to chemical gradients rich in decomposing organic matter, like in some locations of the sediments in Carry-le-Rouet where gMMP were found. This ability contrasts with that of other Magnetomoraceae that could rely more on autotrophic pathways or degradation of simpler organic molecules [20, 23]. Such difference could have triggered geographical separation even at such a millimeter/centimeter scale, creating a barrier to gene flow. Mutations and differential selective sweeps in each group increase population differentiation [83], and ultimately, lead to the emergence of divergent genetic groups from the same MMP ancestor.
Whether or not each MMP genus occupies a different niche, it remains unclear how size matters in gMMP adaptation history to their environment. Breaking larger molecules can imply a larger set of enzymes to achieve complex metabolic pathways, as well as specialized microcompartments, including those for storage of metabolic intermediates. Size could have been a consequence of their trophic specialization. On the other side, selective advantage could have arisen from the gain of buoyancy that supports navigation within chemically stratified environments. This mechanism was described in the multicellular cyanobacterium Microcystis flos-aquae, which adjusts its position in the water column in response to environmental stimuli [84] and was proposed in other large bacteria such as Achromatium [85]. Moreover, in environments characterized by high levels of predation, increasing cell size may represent a strategy to avoid predation by eukaryotic predators, as MTB are common targets of protists, ciliates, biflagellates, and dinoflagellates phagocytosis [86–88]. This aggregate-like structuring could serve as a protective barrier against chemical and environmental stresses, as discussed previously [10]. Although culture-independent methods offer valuable insights into the overall organism complexity, the absence of cultivated models continues to hinder a comprehensive understanding of MMP functioning.
Supplementary Material
wrag017_Supplemental_Files
The reference list from the paper itself. Each links out to its DOI / PubMed record.
- 1Lyons NA, Kolter R. On the evolution of bacterial multicellularity. Curr Opin Microbiol 2015;24:21–8. 10.1016/j.mib.2014.12.00725597443 PMC 4380822 · doi ↗ · pubmed ↗
- 2Rose CJ, Hammerschmidt K. What do we mean by multicellularity? The evolutionary transitions framework provides answers. Front Ecol Evol 2021;9:730714. 10.3389/fevo.2021.730714 · doi ↗
- 3Knoll AH . The multiple origins of complex multicellularity. Annu Rev Earth Planet Sci 2011;39:217–39. 10.1146/annurev.earth.031208.100209 · doi ↗
- 4Grosberg RK, Strathmann RR. The evolution of multicellularity: a minor major transition? Annu Rev Ecol Evol Syst 2007;38:621–54. 10.1146/annurev.ecolsys.36.102403.114735 · doi ↗
- 5Shapiro JA . Bacteria as multicellular organisms. Sci Am 1988;258:82–9.2847312 10.1038/scientificamerican 0288-82 · doi ↗ · pubmed ↗
- 6Zeng X, Zhang C-C. The making of a heterocyst in cyanobacteria. Ann Rev Microbiol 2022;76:597–618. 10.1146/annurev-micro-041320-09344235671534 · doi ↗ · pubmed ↗
- 7Vlamakis H et al. Sticking together: building a biofilm the Bacillus subtilis way. Nat Rev Microbiol 2013;11:157–68. 10.1038/nrmicro 296023353768 PMC 3936787 · doi ↗ · pubmed ↗
- 8Thiery S, Kaimer C. The predation strategy of myxococcus xanthus. Front Microbiol 2020;11:2. 10.3389/fmicb.2020.0000232010119 PMC 6971385 · doi ↗ · pubmed ↗