Multicellular magnetotactic bacterial consortia are metabolically differentiated and not clonal

Abstract

Consortia of multicellular magnetotactic bacteria (MMB) are currently the only known example of bacteria without a unicellular stage in their life cycle. Because of their recalcitrance to cultivation, most previous studies of MMB have been limited to microscopic observations. To study the biology of these unique organisms in more detail, we use multiple culture-independent approaches to analyze the genomics and physiology of MMB consortia at single cell resolution. We separately sequenced the metagenomes of 22 individual MMB consortia, representing eight new species, and quantified the genetic diversity within each MMB consortium. This revealed that, counter to conventional views, cells within MMB consortia are not clonal. Single consortia metagenomes were then used to reconstruct the species-specific metabolic potential and infer the physiological capabilities of MMB. To validate genomic predictions, we performed stable isotope probing (SIP) experiments and interrogated MMB consortia using fluorescence in situ hybridization (FISH) combined with nano-scale secondary ion mass spectrometry (NanoSIMS). By coupling FISH with bioorthogonal non-canonical amino acid tagging (BONCAT) we explored their in situ activity as well as variation of protein synthesis within cells. We demonstrate that MMB consortia are mixotrophic sulfate reducers and that they exhibit metabolic differentiation between individual cells, suggesting that MMB consortia are more complex than previously thought. These findings expand our understanding of MMB diversity, ecology, genomics, and physiology, as well as offer insights into the mechanisms underpinning the multicellular nature of their unique lifestyle.

Introduction

Multicellular lifeforms are defined as organisms that are built from several or many cells of the same species (1, 2). Beyond this, other characteristics of multicellularity include a specific shape and organization, a lack of individual cell autonomy or competition between cells, and a display of cell-to-cell signaling and coordinated response to external stimuli (3). The transition from a single cell to a cooperative multicellular organism is an important evolutionary event that has independently occurred at least 25 times across the tree of life (2). This suggests that the development of multicellularity can occur in any species given proper selective pressure (4, 5). Prior research on the transition of unicellular to multicellular organisms has largely focused on eukaryotic model systems such as choanoflagellates (6), fungi (7), and algae (8). Multicellularity within the domain Bacteria is comparatively rare (9), yet this lifestyle likely first evolved approximately 2.5 billion years ago (10). Examples of multicellularity within the domain Bacteria include filamentous cyanobacteria (e.g., Anabaena cylindrica), mycelia-forming actinomyces (e.g., Streptomyces coelicolor), swarming myxobacteria (e.g., Myxococcus xanthus), centimeter-long cable bacteria (e.g., Electrothrix sp.), and the recently discovered liquid-crystal colonies of Neisseriaceae (e.g., Jeongeupia sacculi sp. nov. HS-3) (5, 11, 12). While capable of multicellular growth, each of these microbes undergoes a unicellular stage at some point in their life cycle.

Currently, the only known example of purportedly obligate multicellularity – an organism without a detectable unicellular stage – within the domain Bacteria are several species of multicellular magnetotactic bacteria (MMB; we use the terms ‘MMB consortia’ and ‘MMB’ interchangeably) (13, 14). MMB are symmetrical single-species consortia composed of 15–86 cells (15)of Desulfobacterota (formerly Deltaproteobacteria) arranged in a single layer enveloping an acellular, central compartment (Fig. 1A-B). Consortia range in size from 3–12 μm in diameter (16–18). Within the Desulfobacterota, MMB form an uncultured, monophyletic family that is distinct from several physiologically and genetically well-characterized unicellular relatives, suggesting a common ancestor that achieved a multicellular state (19–21). MMB are globally distributed in sulfidic brackish and marine sediments but typically are of low relative abundance in these habitats (0.001 – 2% (18, 22, 23)). In addition to their unique obligate multicellular lifecycle, MMB have an organelle called the magnetosome (24). The magnetosome is a lipid vesicle that encapsulates biomineralized magnetite (Fe3O4) and/or greigite (Fe3S4, Fig. 1C) and allows MMB to sense and orient themselves along Earth’s geomagnetic field in a phenomenon termed magnetotaxis. Magnetosome formation is controlled by a magnetosome gene cluster (MGC, SI Appendix Text) that encodes several proteins involved in the formation, alignment, and maturation of the organelle (25, 26). The presence of magnetosomes in MMB can be exploited to physically enrich them from environmental samples using a magnet (SI Videos S1 and S2). This is particularly important considering that MMB have not yet been successfully cultured

Fig. 1.

Morphology and structure of MMB. (A) Cartoon depicting the morphology and internal organization of a MMB consortium. At the center of each MMB consortium lies an acellular space that is surrounded by a single layer of cells. Each cell harbors magnetosome organelles (black polygons aligned along cytoskeleton-like filaments), compartments for carbon or energy storage (gray circles), as well as other, currently unidentified structures. Scale bar ca. 1 μm. (B) Scanning electron microscopy (SEM) image of two MMB magnetically enriched from LSSM, possibly undergoing division. Scale bar, 1 μm. (C) Backscatter electron microscopy image of magnetosome chains within MMB cells (arrow). Magnetosome minerals appear to have 4–8 visible facets and are approximately 30–60 nm in diameter. Scale bar, 300 nm. Contrast and brightness of image (C) was increased for better visualization.

MMB are distinctive among bacteria because their life cycle lacks a unicellular stage. Instead, MMB replicate by the entire consortium doubling its cell number and volume before separating into two, seemingly identical consortia (14, 16, 27, 28). Historically, MMB have been described as “aggregates” of cells (29), which could imply that individual cells assemble to form a multicellular aggregate, akin to the early stages of biofilm formation (5, 29). In this study we use the terms “consortium” (singular) and “consortia” (plural) to describe the unique form of multicellularity observed for MMB.

Under external stress, an MMB consortium becomes dismantled, followed by an immediate loss of magnetic orientation and motility and eventual loss of membrane integrity, leading to cell death (30). MMB consortia consistently exhibit a high degree of magnetic optimization, excluding the possibility that the consortium is a mere aggregation of cells without underlying self-organization (31, 32). Each cell within the consortium has multiple flagella, resulting in the whole consortium being peritrichously flagellated (17, 33). When environmental conditions change, such as alterations in light exposure or magnetic fields, a coordinated response in motility occurs within fractions of a second (33, 34). This collective response implies inter-cellular communication among individual cells, which is hypothesized to occur through the central acellular volume that the cells surround (16). Previous work has hypothesized that the absence of a single cell stage in MMB might be necessary to maintain the acellular volume at the center of each MMB or that their larger size is needed to evade predation by protists (14). Currently, there is no evidence to support or refute these hypotheses. While past studies have presented fascinating insights into the cellular organization of MMB and their diverse abilities to sense the environment via light and electron microscopy (20, 34, 35), their recalcitrance to cultivation has hindered progress towards a better understanding of their physiology and genomics. With the exception of a study that demonstrated chemotactic response of MMB consortia to small molecular weight organic acids (35), questions about their physiology remain unaddressed, and hypotheses about the potential for metabolic differentiation or a division of labor between individual cells within a consortium have not been experimentally tested.

To address these knowledge gaps, we investigated the taxonomic diversity, genomics, physiology, metabolic differentiation, and clonality of MMB inhabiting a tidal pool. To investigate the diversity of MMB within this environment, we sequenced the Single Consortium Metagenomes (SCMs) of 22 MMB consortia, representing eight distinct species of MMB. Comparing the SCMs we were able to quantify the extent of single nucleotide polymorphisms (SNPs) between cells composing individual MMB consortia. Our analyses showed that MMB exhibit genetic diversity within a single consortium, indicating that they are not composed of clonal cells. Physiological predictions were established through the reconstruction of species-specific metabolic models. We tested these predictions by performing stable isotope probing (SIP) experiments and analyzing individual consortia using fluorescence in situ hybridization (FISH), nano-scale secondary ion mass spectrometry (NanoSIMS), and bioorthogonal non-canonical amino acid tagging (BONCAT). Our results demonstrate that MMB are mixotrophic sulfate reducers and that individual cells within MMB consortia exhibit dramatically different rates of substrate uptake, indicating metabolic differentiation, as well as localized protein synthesis activity.

Results and Discussion

Genomic features and phylogenetic analysis of MMB

MMB were recovered from sulfidic sediments collected from a tidal pool in Little Sippewissett Salt Marsh (LSSM; Falmouth, MA, Fig. S1A-B). This sample site was selected based on the ability to magnetically enrich (SI Videos S1 and S2) relatively large quantities of MMB, as previously demonstrated (34, 36). Individual MMB consortia were sorted from a magnetically enriched pellet using fluorescence-activated cell sorting and the DNA of individual sorted MMB was amplified by multiple displacement amplification before Illumina sequencing. From this sample, the SCMs of 22 individual MMB were recovered (Fig. 2, SI Appendix Table S1). The GC content of the SCMs ranged from 36.2 to 38.4%, which is similar to the GC content observed in previously published MMB draft genomes (20, 37, 38). The average and median size of the 22 new SCMs was 7.7 Mb, with a range from 6.1 to 9.1 Mb (SI Appendix Table S1). Prior to this study, only three draft genomes of MMB had been sequenced. These genomes exhibited significant variations in size, ranging from 14.3 Mb for Ca. Magnetomorum sp. HK-1 (37), 12.5 Mb for Ca. Magnetoglobus multicellularis (20), and 8.5 Mb for MMP XL-1 (38), although the MMP XL-1 genome is not publicly available. The genome sizes of Ca. M. multicellularis and Ca. M. sp. HK-1 could be conflated due to contamination or the combination of sequence data into the same final bin, as discussed in the respective studies (20, 37) and evidenced by our own evaluations of genome contamination (Fig. 2A)

Fig. 2.

Genomic and phylogenetic analysis of all publicly available MMB MAGs and the 22 SCMs generated in this study. (A) Maximum-likelihood tree, inferred with FastTree, using a concatenated set of six conserved COGs (Table S3) present in all entries. Ultrafast bootstrap support values and selected genome statistics are listed. The color codes for the SCM Groups remain the same throughout all figures. (B) Average full length 16S rRNA gene identity and (C) average genome nucleotide identity heat maps of the eight newly identified MMB species compared to two available MMB reference genomes (Ca. M. multicellularis and Ca. Magnetomorum sp. HK-1). For a phylogenetic tree of all publicly available MMB 16S rRNA gene sequences, see Fig. S2. For an exhaustive sequence identity analyses of 16S rRNA and whole genomes of MMB see Figs. S3–5.

Only 14 of the 22 SCMs contained 16S rRNA genes (SI Appendix Table S1). These sequences, together with publicly available 16S rRNA sequences of MMB as well as those of their single-cell relatives Desulfosarcina variabilis and Ca. Desulfamplus magnetomortis BW-1, were used to construct a phylogenetic tree (SI Appendix Table S2). This analysis revealed the presence of five phylogenetically distinct genera of MMB in LSSM with high bootstrap support (>75%) (Fig. S2). Analysis of amplicon sequence data obtained in this study and sequences from a previous study at LSSM (36) showed that Group 1 MMB was most abundant in the sample site, constituting 61% of all 16S rRNA genes. Groups 2, 4, 5, and 3 accounted for 21%, 6.5%, 6.5%, and 5% of the 16S rRNA genes, respectively (Fig. S2, S3).

Phylogenomic analysis of six bacterial single copy genes found in all recovered MMB SCMs yielded a topology consistent with the phylogeny derived from the 16S rRNA gene sequences (Fig. 2, SI Appendix Table S3, Fig. S2). Similarly, whole genome and 16S rRNA specific ANI analyses resolved eight unique species of MMB with >96% average nucleotide identity. We assigned type genomes for each new MMB species and named them after scientists who have greatly advanced our knowledge of MMB (SI Appendix Text, SI Appendix Table S4).

Clonality within MMB

MMB have historically been assumed to be clonal due to the synchronized replication of cells during division, which should result in genetically identical daughter cells in the same consortium (14, 27). Additionally, obligate multicellularity has traditionally been thought to perpetuate a clonal population (39). Although MMB maintain an obligate multicellular lifecycle, the degree to which clonality exists within a single consortium has never been experimentally tested. Currently, the only evidence suggesting that cells within MMB are closely related comes from analyses of the 16S rRNA genes from cells of a single genome amplified MMB consortium (37) and a FISH study demonstrating that cells within individual MMB have identical 16S rRNA sequences (36).

We set out to test the hypothesis of clonality using comparative genomics of the 22 MMB SCMs recovered in this study. Reads from each individual SCM were mapped to the corresponding genome bins to quantify single nucleotide polymorphisms (SNPs) within a single MMB consortium. As a procedural control, 10, 30, 60, and 100 cells of a clonal culture of Pseudomonas putida were sorted to construct a mock multicellular consortium. The DNA of MMB consortia and P. putida controls were amplified using multiple displacement amplification and sequenced using Illumina short read sequencing. Our analysis of the SCMs revealed for the first time that MMB consortia are genomically heterogeneous and thus do not fit the model of clonality for obligate multicellular organisms (Fig. 3A). MMB from LSSM contain up to two orders of magnitude more SNP differences within a single consortium as compared to the same number of cells from the clonal control (p < 7.3 × 10⁻⁹), with an estimated range of 157–789 SNPs in individual SCMs (Fig. 3, SI Appendix Table S5). Other environmental microbes co-sorted with MMB showed a SNP rate similar to the clonal control and a SNP rate statistically different from the MMB (p < 2.4 × 10⁻⁶), illustrating the uniqueness of MMB. Wielgoss et al. performed a similar analysis on fruiting bodies of the aggregative multicellular bacterium Myxococcus xanthus in which a comparison of the genomes of cells in fruiting bodies revealed 30 SNP differences between lineages originated from a recent single ancestral genotype (40). Furthermore, nearly half the mutations detected in the M. xanthus genomes occurred in the same six genes, suggesting there was a strong selection for socially relevant genes, such as a histidine kinase (signal transduction) and methyltransferase (gene expression). Positive selection upon cooperative genes may promote diversity within the organism as a mechanism to increase fitness within spatiotemporally variable environments and protect against social cheaters (41).

Fig. 3.

Clonality analysis of individual MMB consortia. (A) Individual reads were mapped to the same genome bin for each of the 22 SCMs. This analysis revealed that the genomes of cells within MMB consortia have a higher single nucleotide polymorphism rate (SNP expressed as Variations per kb) as compared to a clonal Pseudomonas sp. control (p < 7.3 × 10⁻⁹, n = 10, 30, 60, and 100 Pseudomonas cells) and other environmental cells (p < 2.4 × 10⁻⁶, e.g. “Other”). (B) The three sample categories showed no statistically significant difference in terms of their ratio of non-synonymous to synonymous substitutions (dN/dS). Values near 0 indicate that substitutions are neutral and there is no positive selection of the protein-coding genes in which the SNPs reside. The color of each SCM corresponds to the color identifying each unique species in Fig. 2.

To investigate if the genetic heterogeneity within MMB contributes to an increased fitness of the organism, we identified the genes containing SNPs and calculated the corresponding ratio of non-synonymous (dN) to synonymous (dS) substitutions. This analysis showed that the SNP differences within the SCMs of MMB appear to be random with no single gene or category of genes exclusively impacted by the SNPs within or across MMB consortia (Fig. 3B, SI Appendix Table S6). SNPs with a high dN/dS ratio were predominantly found in unannotated genes, such as hypothetical proteins (Fig. S6). Such unannotated genes that are subject to stronger positive selection could ultimately drive functional divergence within the consortium. Other benefits of genomic heterogeneity within MMB are not readily apparent and could be attributed to errors during DNA replication or damaging effects of mutagens. However, it has been shown that a single mutation can lead to a division of labor in bacteria (42). At this point, it is unclear whether any of the changes we observe in the genomes contained within individual MMB would lead to phenotypic differentiation between the adjacent cells.

Genome annotation

Metabolic reconstructions of the MMB SCMs (Fig. 4, SI Appendix Table S7) revealed that all MMB are capable of heterotrophic sulfate reduction and can use acetate, succinate, and propionate as carbon donors and/or electron sources, consistent with previous genomic analyses (20, 37). The SCMs show that LSSM MMB have highly similar metabolic potential. One exception is Ca. M. sippewissettense, which lacks the ability to utilize acetyl-coenzyme A (CoA) synthetase and is unable to use acetate, instead likely relying on lactate dehydrogenase to metabolize lactate, a substrate the other species are not capable of using. None of the SCMs contain acetaldehyde dehydrogenase, indicating that MMB are not capable of alcohol fermentation. We resolved a complete glycolysis pathway and TCA cycle as well as reductive CoA pathway in all SCMs. The presence of these genes suggests that MMB in LSSM are capable of both heterotrophic and autotrophic growth using sulfate reduction coupled to hydrogen metabolism, by means of hyaA/B and hybA/B complexes and oxidative phosphorylation. MMB are genetically capable of shuttling electrons using complexes I, II, and V of the oxidative phosphorylation pathway using F-type ATP synthase complexes, although partial V/A type ATP synthase were found in Ca. Magnetoglobus martinsiae and Ca. Magnetomorum sippewissettense. In addition, they encode a full Nqr (Na⁺- transporting NADH:ubiquinone oxidoreductase) complex that can move electrons from NADH to ubiquinone with the translocation of a Na⁺ across the membrane. Cytochrome bd oxidase subunits I and II are present in all SCMs, except Ca. Magnetoglobus farina, and could be used to respire molecular oxygen (O₂) using electrons from cytochrome c or quinols (43). All species of MMB from LSSM encode rubrerythrin and superoxide reductase, suggesting the possibility that O₂ could instead be detoxified by the cytochrome bd oxidase (SI Appendix Table S7) (20, 44). Electrons can also be removed by the reduction of protons to molecular hydrogen (H₂) by group 1 nickel-iron hydrogenases. The H₂ can then diffuse across the membrane where HybA/B could oxidize the H₂, yielding two electrons and two protons. From there, cytochrome c can shuttle the electrons to the Dsr and Qmo complexes for dissimilatory sulfate reduction.

Fig. 4.

Metabolic potential of the eight MMB species in LSSM. Arrows without circles indicate presence of the respective enzyme or pathway in all bins. Circles indicate complete presence (black), partial presence (gray), or missing (white) genes in each species. A full list of genes used to construct this figure can be found in Table S5.

The MMB SCMs encode several divalent metal transporters, including FoaAB ferrous iron and FepBDC ferric iron transport proteins, indicating they are capable of using both Fe(II) and Fe(III). All SCMs encode phosphate transporters as well as oligopeptide and branched-chain amino acid transporters. Genes for polyamine transport were recovered in the SCMs and may provide resistance to environmental stress such as osmotic pressure and reactive oxygen species (45). Additionally, each SCM encodes a glycine betaine transporter but does not encode a betaine reductase, indicating that MMB do not use glycine betaine as a nitrogen source but as an osmoprotectant (46). All MMB species in LSSM, except Ca. M. sippewissettense, encode an Amt transporter to transport ammonia into cells that can then be converted into glutamine or glutamate and fed into anabolic pathways. Additionally, each species encodes the NitT/TauT system for nitrate, sulfonate, and bicarbonate transport into cells. The SCMs showed that MMB are capable of synthesizing all canonical amino acids except cysteine and lack cysteine prototrophy genes. Cultures of single celled magnetotactic bacteria have been found to require the addition of cysteine for growth, suggesting that many magnetotactic bacteria, including MMB, cannot synthesize their own cysteine (47). The inability to synthesize a sulfurous amino acid is surprising given that most magnetotactic bacteria, including all known MMB, live in sulfur-rich environments.

Previous studies using transmission electron microscopy have found large vesicles within MMB cells that have been attributed to carbon/energy or phosphate storage (48). Metabolic analysis of the SCMs showed that acetyl-CoA could be condensed and polymerized to polyhydroxybutyrate (PHB) for storage. Furthermore, all necessary genes were identified for β-oxidation using triacylglycerol synthesized from the acylation of glycerol-3P with acyl-CoA (Fig. 4, SI Appendix Table S7). Using Raman microspectroscopy applied to individual MMB, we demonstrated the presence of PHB and lipids, along with Nile Red staining of carbon-rich droplets within cells (Fig. S7, SI Appendix Table S8). This is, to our knowledge, the first time carbon and energy storage compounds in MMB have been unambiguously identified. Carbon storage has been shown to support the multicellular reproductive life cycles in Vibrio splendidus through the specialization of cells during resource limitations (49), suggesting that MMB may utilize a similar mechanism to support their multicellular growth.

Altruistic behavior in biological systems is often favored when relatedness among species is high and the benefit is comparatively large compared to the cost, as has been observed in multicellular myxobacteria (41). The SCMs revealed that MMB encode mazE/F, hicA/B, and yefM/yefB type II toxin-antitoxin (TA) systems (Fig. 4, SI Appendix Table S7). TA systems represent an extreme example of altruism in multicellular systems, as individual cells that contribute to the organism by sacrificing themselves through death do not directly benefit from the organism’s multicellularity. But, selection favoring altruistic traits occurs due to the fitness benefits those traits impart on relatives (50). Detection of CRISPR (clustered regularly interspaced short palindromic repeats) systems I, III-A, and III-B (SI Appendix Table S7) suggest the TA systems could be used in response to viral infection (51). The evolution of altruistic cooperation in multicellular organisms has been proposed as a response to environmental stressors (50), indicating the presence of TA systems likely confer increased fitness for MMB in the environment.

Cell-to-cell adhesion

One of the most intriguing features of MMB is their multicellular lifecycle. But how these bacteria maintain their multicellular shape is not entirely known. Previous genomic and microscopic analysis of MMB suggested that exopolysaccharides, adhesion molecules, and Type IV pili could be involved in cell-to-cell adhesion (20, 52). Extracellular matrices, specifically those composed of polysaccharides, have been shown to be important for the development and maintenance of bacterial multicellularity, resulting in several emergent properties that benefit the organism, including the reduction of maintenance energy for individual cells (53). Myxobacteria sp. and Escherichia coli have both been shown to use exopolysaccharides to maintain macroscopic biofilms, (7, 54). The SCMs recovered in this study encode genes for extracellular polysaccharide biosynthesis, including family-2 glycosyltransferases (GT2), which have been shown to secrete diverse polysaccharides such as cellulose, alginate, and poly-N-acetylglucosamine (55, 56). Specifically, the genes identified in the SCMs were homologous to GT2 Bcs proteins, a bacterial protein complex that synthesizes and secretes a β−1,4-glucose polymer (e.g., cellulose) during biofilm formation (SI Appendix Table S7) (57, 58). The LSSM MMB encode enzymes that catalyze the production of cellulose for biofilm formation (bcsA, bcsQ, bcsZ, pilZ, and bglX), but lack the co-organization of genes at a single locus as observed for other bacteria (57). Furthermore, the bcsB and bcsC subunits were not identified, but additional GT2 as well as wza genes that may be involved in the synthesis of exopolysaccharides were present (59). The catalytic activity of BcsA has been shown to be influenced by the concentration of cyclic dimeric guanosine monophosphate (c-di-GMP) which is in turn affected by environmental oxygen levels (60, 61). Under oxic conditions the cellular level of c-di-GMP has been shown to increase and bind to BcsA, leading to increased cellulose synthesis (61). Because MMB commonly exist in oxygen-deficient sediments, cellulose synthesis may be triggered under oxic conditions to stimulate biofilm formation, which has been observed in cultivation attempts of MMB (20).

Filamentous hemagglutinin has been shown to recognize and bind to carbohydrates to facilitate cell-to-cell adhesion in a biofilm (62, 63). The presence of filamentous hemagglutinin genes in our SCMs suggests MMB could use these protein complexes as a mechanism for cell-to-cell adhesion, as previously suggested (20). Furthermore, the SCMs encode genes for OmpA/F porins, proteins with adhesive properties that have been suggested to interact with exopolysaccharides leading to aggregation of cells (64). Type IV pili, which have been shown to be involved in cell-to-cell adhesion by interacting with exopolysaccharides (65), were also identified in the SCMs. The pili could alternatively be used for motility, chemotaxis, organization, and DNA uptake (66). Further investigation into the use of the Type IV pili within MMB is warranted as only predictions can be made from the available genomes.

Previous studies on the membrane of MMB using Ruthenium Red dye and calcium cytochemistry have shown that the consortia are coated in a polysaccharide that extends between cells into the acellular central compartment but the exact composition and structure of this polysaccharide remains unclear (16, 67). Using Raman microspectroscopy we identified peaks corresponding to exopolysaccharides, confirming the presence of an exopolysaccharide within or surrounding MMB (Confocal Raman does not have enough z-resolution to distinguish the in- and out-side of cells; Fig. S7, Appendix Table S8). Cellulase hydrolysis of the MMB resulted in eroded surfaces of the consortia, demonstrating that MMB are indeed covered by a cellulose layer (Fig. S8). Together, these analyses highlight the structural and functional significance of exopolysaccharides required for the multicellular morphotype of MMB.

Abundance, distribution, and in situ activity of MMB in LSSM

Temporal shifts in MMB groups at LSSM have previously been documented (68) but the abundance of MMB correlated to sediment depth has not yet been analyzed. MMB in the LSSM subsurface were quantified by retrieving a 15 cm core from the tidal pond and determining the fractional abundance of each of the five MMB groups recovered throughout the core at centimeter-scale resolution using newly designed fluorescence in situ hybridization (FISH) probes (Fig. S9, SI Appendix Table S9). In the top five centimeters of sediment, Group 1 MMB accounted for >75% of all MMB while the other groups accounted for 1–25%, depending on sediment depth. The total abundance of MMB dropped sharply below 5 cm, where the sediment horizons transitioned from sandy to dense clay sediment containing plant roots. This could be due to MMBs preference for low oxygen conditions, under which sulfate reduction is favored (35, 69). A similar depth-abundance profile was previously observed for the closely related MMB Ca. M. multicellularis (69).

Bioorthogonal noncanonical amino acid tagging (BONCAT) was used to determine the anabolic activity of MMB Group 1 in the top 6 cm of the LSSM core, which hosted the majority of MMB. Using this approach, we identified a statistically significant difference in MMB activity from 1 cm depth compared to the 2–3 cm (p < 3.4×10⁻⁴) and from 3 cm compared to 4–5 cm (p < 3.9×10⁻³), below which the MMB population diminished (Fig. S10). The increase of activity of MMB in the first 5 cm of the sediment could be attributed to the circumneutral pH and low redox potential (−260 to −460 mV), as previously observed to be important for the bioavailability of iron and sulfur species for MMB (37).

Physiology of MMB

Previous genome- and chemotaxis-based studies suggested that MMB live by heterotrophic sulfate reduction using small organic acids as electron donors (20, 35, 37). However, no direct observation of the use of such organics has been reported. Our metabolic reconstructions revealed that all MMB species in LSSM are genetically capable of coupling sulfate reduction to the use of acetate, propionate, and succinate as well as inorganic carbon fixation via the reductive acetyl-CoA pathway. To test whether MMB use these carbon sources to support their growth, we incubated sediment samples with ¹³C-labeled substrates (acetate, bicarbonate, propionate, and succinate) in situ and analyzed individual MMB using Nano-scale secondary ion mass spectrometry (NanoSIMS). MMB that had been incubated with ¹³C-acetate exhibited higher ¹³C labeling as compared to the other substrates, which could suggest a preference for acetate (Fig. 5, SI Appendix Table S10). To identify specific MMB groups, FISH was performed prior to NanoSIMS analyses. Group 1 MMB showed the highest incorporation of ¹³C from acetate as compared to Groups 3 and 4 (p < 1.5×10⁻³, Fig. S11). We also observed a significant difference between Group 1 and 4 for ¹³C-bicarbonate and ¹³C-propionate uptake (p < 3.9×10⁻³ and 5.8×10⁻⁵, respectively). At least three genera of MMB (i.e., Groups 1, 2, and 3) assimilated both bicarbonate and propionate (Fig. S15). We were unable to magnetically enrich MMB from a sediment sample incubated with ¹³C-acetate and molybdate, an inhibitor of sulfate reduction, indirectly demonstrating that MMB are in fact sulfate reducers. In summary, our analyses demonstrated that LSSM MMB are capable of assimilating both inorganic and organic carbon, indicating autotrophic and heterotrophic growth, and that different Groups of MMB demonstrate variable affinities for carbon sources.

Fig. 5.

NanoSIMS analysis of the cellular ¹³C-content of MMB consortia after in situ incubation with isotopically light or heavy carbon sources, specifically 1,2-¹³C₂-acetate, ¹³C-bicarbonate, 1,2-¹³C₂-propionate, or 1,2-¹³C_2-succinate, for 24 hours. The kill control contained magnetically enriched MMB that had been fixed in 4% paraformaldehyde prior to ¹³C-acetate addition. The negative control was sediment containing MMB without substrate addition. The dotted line shows the natural abundance of ¹³C. For further description of boxplots, see SI Appendix Text. Inset images show representative NanoSIMS hue saturated images (HSI) for each ¹³C-labeled substrate analyzed. Color scales in HSI images are 1.1% - 5% atom percent ¹³C. Scale bars are 5 μm. Fig. S1CD show the incubation setup. For a comparison of the anabolic activity of MMB groups 1, 3, and 4 see Fig. S11. Fig. S12 provides an example for correlative microscopy analysis of MMB. SI Materials and Methods detail the calculation of atom percent. For ROIs, see Fig. S13.

Metabolic differentiation as studied by SIP-NanoSIMS

A hallmark of multicellularity is the existence of a division of labor (5), however, because of their recalcitrance to cultivation, this hypothesis has never been addressed in MMB. To investigate whether MMB are metabolically differentiated, a magnetic enrichment of MMB was incubated in vitro with ¹³C-labeled acetate and deuterium oxide (²H₂O), with cellular labelling from the latter being a general proxy for metabolic activity (70). Samples analyzed using NanoSIMS showed variation of isotopic signal across cells within individual consortia, indicating different metabolic activity within MMB (Fig. 6, SI Appendix Table S11). The mass ratio for each isotope label was quantified and areas of high anabolism (referred to as “hotspots”) within the consortium compared to the value of the same isotope label for the whole consortium. This analysis demonstrated a statistically significant difference of anabolic activity between hotspots and whole consortium for both ¹³C and ²H₂O (p < 1.3×10⁻³ and < 2.2×10⁻⁸, respectively). Comparison of SEM and NanoSIMS imaging shows that the extent of SIP labeling varies within a single cell as well as across the entire MMB consortium (Fig. S12). The hotspots do not exhibit localization in any specific region of an MMB. However, they are not uniformly distributed throughout the consortium, demonstrating variations in metabolic activity with some areas displaying lower metabolic activity than others. To further investigate the localization of the isotope within the individual consortium, we applied a median filter ratio to the hue saturated images (HSI) using different kernel radii (71). This method averages the isotopic ratio over the given pixel radius, revealing sub-consortium localization across the MMB (Fig. S15). Together, our analyses shows that metabolism of ¹³C-acetate and ²H-water is not uniform across the MMB, suggesting a differentiation in metabolic activity within individual consortia. Similar differences in the uptake of isotope-labeled substrate have also been reported for cellularly and metabolically differentiated cells of filamentous cyanobacterium Anabaena oscillarioides (72).

Fig. 6.

NanoSIMS analysis of MMB consortia incubated with 1,2-¹³C₂-acetate and ²H₂O. Hotspots within individual consortia were auto-segmented in ImageJ and the isotope ratios of hotspots compared to the value for the whole consortium and negative controls. The ¹³C and ²H hotspots showed significantly higher isotopic enrichment when compared to the values for the respective whole consortium (p <1.3×10⁻³ and <2.2×10⁻⁸, respectively), indicating they are metabolically differentiated. For further description of boxplots, see SI Appendix Text. Inset images show NanoSIMS HSI of the same MMB consortium analyzed using mass ratio ¹³C¹²C/¹²C₂ and ²H/¹H, revealing cell-to-cell differentiation. The HSI are scaled to show the atom percent of the respective isotope. For an example of the correlative microscopy workflow used to study MMB see Fig. S12. For ROIs, see Fig. S14.

Metabolic differentiation as studied by BONCAT

To determine if protein synthesis was localized to specific or individual cells within the consortium, we combined BONCAT with confocal laser scanning microscopy. Our analysis revealed an apparent gradient of newly synthesized proteins within each cell of the consortium, showing localization around the acellular center of individual consortia (Fig. 7). This distinct pattern of protein synthesis was observed in all 57 MMB we examined (Fig. S16). The localization of newly synthesized protein around the acellular center of the consortium suggests this area is highly active, however the reason is currently unknown. Cells within the consortium could engage in a division of labor by metabolizing specific substrates (e.g., acetate) and then sharing those resources with other cells through the acellular space, possibly by the utilization of membrane vesicles (52). A prime example of a division of labor in multicellular bacteria is the filamentous cyanobacteria Anabaena. This organism has established a mutually beneficial interaction between the heterocyst and vegetative cells via intercellular exchange of metabolites through septal junctions (5, 73). However, there is no evidence that such pores or channels exist in MMB, although an alternative route for metabolite transfer could be the acellular space within the consortium. This space has been hypothesized to be used for communication and metabolite exchange because it provides the shortest distance between any two cells (52). The localization of newly synthesized protein around the acellular center of the consortium suggests this area is highly active, possibly for exchange of metabolites from cells that are hotspots for anabolic activity. This implies cells within the consortium could metabolize specific substrates (e.g. acetate) and then share those resources with other cells through the acellular space, possibly by the utilization of membrane vesicles (52).

Fig. 7.

Heterogeneity in anabolic activity within individual MMB consortia as revealed by BONCAT. (A) The averaged intensity profile across the diameter of 57 rotationally averaged BONCAT-labeled MMB with standard deviation shown in gray. Relative fluorescence intensity (RFI) and diameter of each MMB was scaled as a ratio (0 to 1) to account for differences in fluorescence intensity between consortia and size of consortia. The boxplots show the averaged RFI for each quarter section of the radius with a pairwise statistical difference of p < 1.0×10⁻¹⁰. For further description of boxplots, see SI Appendix Text. (B) Gray scale confocal microscopy image of a BONCAT labeled MMB showing proteins that had been synthesized over a 24-hour period. (C) Image of the MMB shown in (B) that has been rotationally averaged prior to quantification in Eman2. The red dotted line shows each quarter analyzed for the boxplots shown in (A). For raw and rotationally averaged images of all 57 MMB, see Fig. S16.

Conclusion

In summary, our study demonstrated that cutting-edge culture-independent approaches can reveal fundamental biology of yet uncultured multicellular microorganisms. We showed that MMB exhibit a higher level of complexity than previously thought by maintaining genomic heterogeneity and metabolic differentiation amongst the individual cells of a consortium. Moreover, we provided a detailed analysis of the genetic potential of eight newly discovered species of MMB as well as their ecology, ecophysiology, and in situ activity. We hope that these results will eventually lead to MMB representatives to be brought into culture. In addition, our results provide the basis for future experiments to further explore the mechanisms of cell-to-cell heterogeneity. Specifically, we expect mRNA-FISH (74, 75) studies to reveal to what extent gene expression levels differ from cell to cell, and SIP-NanoSIMS and spatial metabolomics (76) to reveal the molecular underpinnings of cellular interactions. Given that the biology of MMB is, as far as we know, unique in the bacterial domain, we propose MMB should, despite their recalcitrance to cultivation, receive higher attention by researchers interested in the evolution and biology of bacterial multicellularity.

Materials and Methods

MMB sorting, single consortia genomic sequencing and clonality analyses

A sediment sample from LSSM was shipped overnight to the Joint Genome Institute (JGI, then Walnut Creek, CA) where a magnetic enrichment was performed to obtain a pellet of MMB (see SI Appendix Methods for details). The enriched MMB were stained with SYBR Green (ThermoFisher, Eugene, OR) and sorted using a BD Influx fluorescence-activated cell sorter based on size (448 nm excitation of SYBR vs. side scatter; Fig. S17) to obtain individual MMB consortia in single wells of a 384 well plate. In addition, replicates of 10, 30, 60, and 100 cells from a culture of Pseudomonas putida KT2440 that had been grown in LB media were sorted into single wells as a mock control for clonal multicellularity. The P. putida culture liquid culture was initiated from a single colony picked from an LB agar plate. Sorted MMB and P. putida were then lysed and DNA amplified via the WGA-X protocol (77). Amplified SCMs were screened using 16S rRNA gene PCR according to DOE JGI standard protocols (78). Next, sequencing libraries were generated from amplified DNA using the Nextera XT v2 library preparation kit (Illumina), and sequenced on the Illumina NextSeq platform. Assemblies were derived from the IMG/M database (79). Contigs larger than 2kb were organized into genome bins based on tetranucleotide sequence composition with MetaBat2 (80) with default settings. Metagenome assembled genome (MAG) completeness and contamination were estimated with CheckM (v1.012) (81). Gene calling was performed with Prodigal (82) using the bacterial code (translation table 11). Average nucleotide identities (ANI) between MAGs were calculated with FastANI (v1.1) (83), filtered at 95% sequence identity and 30% aligned fraction, and then clustered using mcl (v14–137) (84).

We assessed clonality of sorted MMBs, single sorted and amplified Pseudomonas controls and other MAGs derived from sorted MMBs by mapping the reads from the respective libraries to the contigs larger than 5kb in assemblies derived from the same library using BBMap (v38.79) (https://sourceforge.net/projects/bbmap/, (85)) with the flags minid=0.95 minaveragequality=30. Variants were called with the BBTools script callvariants.sh using the flags minreads=2 minquality=30 minscore=30 minavgmapq=20 minallelefraction=0.05 and identified variants were then annotated as synonymous (s), nonsynonymous (ns) or intergenic depending on their position. Variants made up by one or more Ns were excluded from the analysis. To investigate differences between MMB, all libraries were also mapped to contigs with a size of at least 5kb infrom the longest MMB assembly (3300034493).

Stable isotope probing

To empirically test the use of carbon substrates as predicted by the functional annotation of MMB SCMs and determine the anabolic activity of MMB cells, we employed performed both in situ and in vitro incubations of MMB with ¹³C- and ²H-labeled substrates (all 99.9%, Cambridge Isotopes Laboratories). The in situ incubations were performed in duplicate on August 28^th 2022 at LSSM by amending 200 mL top sediment slurry with 2 mM ¹³C-1,2-acetate, 2 mM ¹³C-1,2-succinate, 5 mM ¹³C-1,2-propionate, 5 mM ¹³C-bicarbonate, or 2 mM ¹³C-1,2-acetate plus 8 mM molybdate (a competitive inhibition of sulfate reduction). A negative control to which no amendment was made as well as a killed control in which biomass had been pre-incubated with 4% paraformaldehyde (PFA) for 60 minutes at ambient temperature prior to addition of 2 mM ¹³C-1,2-acetate were also performed. Samples were stored in 200 mL Pyrex glass bottles (Corning, Glendale, AZ) and incubated for 24 hours in situ at the sample site where they were buried 4–6 cm below the sediment in a basket (Fig. S1C-D). The in vitro incubations were performed by incubating magnetically enriched MMB in 10 mL of 0.22 μm filter sterilized (Millipore, Burlington, MA) LSSM water amended with the same amendments as the in situ incubations, as well as 50% deuterium oxide (D₂O), for 24 hours at ambient lab temperature (~23 °C) in the dark. At the end of each incubation period, MMB were magnetically enriched and fixed with 4% PFA for 60 minutes at ambient temperature. Cells were centrifuged for 5 minutes at 16,000 g, after which the supernatant was removed, and the cell pellets resuspended in 50 μL 1× PBS and stored at 4 °C.

NanoSIMS

Samples were prepared for NanoSIMS on stainless steel coupons as previously described (86); for details see SI Materials and Methods. To quantify cell-to-cell differences in isotope uptake within individual consortia, ROIs were selected around localized densities (i.e., hotspots) of masses corresponding to the respective substrate and compared to whole consortia values for the same isotope of interest. To select ROIs, Fiji (https://imagej.net/software/fiji/) was used to convert the mass image to an 8-bit image for which the brightness and contrast adjusted to help identify the localized densities for the mass of interest (e.g. ¹²C²H 14.02, ¹²C¹³C 25.00).

BONCAT

BONCAT was performed as previously described (87); for details see SI Materials and Methods. To evaluate cell-cell differences in anabolic activity of individual consortia, MMB were imaged by taking z-stacks (approximately 300 nm per image) of the entire consortia using an Inverted DMI8 Stellaris 8 Confocal Microscope (Leica Microsystems). Images focused on the center of the consortia were selected and Eman2 (88) was used to select individual MMB for particle analysis. Each image was then filtered using an edge mean normalization, center of mass xform, and rotational average math settings (Fig. S16). Because of varying sizes of consortia, a Python script was used to determine the radius of each consortium by calculating the number of pixels from the center of mass, as determined by the filter, to where the standard deviation of the pixels is < 0.01. The radius of all consortia was standardized by dividing 1 by the radius. Additionally, the average fluorescence intensity was normalized by calculating $I_{norm}=\frac{I_{ori}-I_{min}}{I_{max}-I_{min}}$, where I_ori is the original fluorescence intensity value and I_min/I_max are the minimum and maximum relative fluorescence intensity values for the individual consortia. The average and standard deviation of data was calculated and plotted using R. All code used for analysis is deposited on GitHub (https://github.com/georgeschaible/MMB-BONCAT).

Supplementary Methodology

Sample collection, phylogenetic analysis, genome and magnetosome analyses, FISH, BONCAT, NanoSIMS, Raman microspectroscopy, and SEM experiments, geochemical analysis, and statistical analyses are described in the SI Materials and Methods.

Significance statement.

The emergence of multicellular lifeforms represents a pivotal milestone in Earth’s history, ushering in a new era of biological complexity. Because of the relative scarcity of multicellularity in the domains Bacteria and Archaea, research on the evolution of multicellularity has predominantly focused on eukaryotic model organisms. In this study, we explored the complexity of the only known bacteria without a unicellular stage in their life cycle, consortia of multicellular magnetotactic bacteria (MMB). Genomic and physiological analyses revealed that cells within individual MMB consortia are not clonal and exhibit metabolic differentiation. This implies a higher level of complexity than previously assumed for MMB consortia, prompting a reevaluation of the evolutionary factors that have led to the emergence of multicellularity. Because of their unique biology MMB consortia are ideally suited to become a model system to explore the underpinnings of bacterial multicellularity.

Data availability

The single consortia metagenomes of MMB generated in this study are available on JGI’s IMG/M under the genome numbers 3300028595, 3300034483–3300034486, and 3300034488–3300034505. The genome sequences of Ca. M. multicellularis and Ca. M. HK-1 are available at NCBI Genbank under accession numbers GCA_000516475 and JPDT00000000, respectively. Magnetosome sequences for Ca. Desulfamplus magnetomortis BW-1, Ca. Magnetananas rongchenensis RPA, and MMP XL-1 are available at GenBank under accession numbers HF547348, KY084568, and ON204283:ON204284, respectively. Python and R code used to analyze BONCAT data are available on GitHub (https://github.com/georgeschaible/MMB-BONCAT).