Magnetotactic bacteria (MTB) refer to a unique group of morphologically, phylogenetically, and physiologically diverse microorganisms thriving mostly at oxic–anoxic boundaries of aquatic habitats. They form intracellular magnetic nanocrystals enclosed within vesicles named magnetosomes that vary between species in size, shape, composition (Fe₃O₄ or Fe₃S₄), number, and organisation (Schüler, 1999; Isambert et al., 2007). The magnetic properties of these magnetosomes allow MTB cells to passively align parallel to the geomagnetic field lines and swim actively towards, or away, attractants and repellents, in chemically stratified environments (Lefevre and Bazylinski, 2013). Most known MTB species are microaerophiles or anaerobes and were essentially described at water-sediment interfaces (Lefevre and Bazylinski, 2013). Because of their ability to trigger the formation of magnetic biominerals into unique prokaryotic organelles and their potential contribution to the Fe biogeochemical cycle (Chen et al., 2014; Lin et al., 2014; Amor et al., 2020), MTB have been for decades the focus of numerous studies in many fields of research, from microbial ecology, evolution (Lin et al., 2018; Monteil et al., 2018), molecular genetics (Uebe and Schüler, 2016), palaeomagnetism (Kobayashi et al., 2006; Li et al., 2013) to geochemistry (Amor et al., 2015) and biophysics (Popp et al., 2014; Keim et al., 2018; Klumpp et al., 2019).

Recent studies of MTB in natural environments revealed additional underestimated biomineralisation processes in this group of bacteria, providing a potential broad impact of MTB on P, S, and C geochemical cycles (Li et al., 2021). In particular, MTB observed at the oxic–anoxic transition zone (OATZ) in the stratified water column and in sediments of the ferruginous Lake Pavin (Massif Central, France) harbour the capability to sequester phosphorus and cations (Ca²⁺, Mg²⁺, or K⁺) in the form of very large inclusions of polyphosphates (PolyP), or carbon and calcium in the form of amorphous calcium carbonates (Rivas-Lamelo et al., 2017; Monteil et al., 2021). Previous studies showed that phosphogenesis in this lake occurs just beneath the OATZ (Busigny et al., 2016) and suggested the involvement of cells in mineral precipitations (Cosmidis et al., 2014; Miot et al., 2016). In this freshwater environment, magnetotactic cocci (MTBc) hyperaccumulating P (up to 90% of their cell volume) as intracellular PolyP were affiliated to the Magnetococcaceae family of the Candidatus Betaproteobacteria class—also proposed as the class Magnetococcia in the GTDB taxonomy (Ji et al., 2017; Parks et al., 2018). Based on the 16S rRNA gene sequence, two different species of MTBc were detected (Rivas-Lamelo et al., 2017). MTB forming 2–4 large granules of intracellular amorphous CaCO₃ (iACC), occupying up to 65% of their cell volume, were affiliated to two undescribed genera within the Rhodospirillaceae family of the Alphaproteobacteria (Monteil et al., 2021). Intracellular carbonatogenesis coupled with magnetosome formation in magnetotactic Alphaproteobacterium was also recently evidenced in sediment of Xingqinggong Lake (Liu et al., 2021). Before these observations, only Cyanobacteria and the non-magnetotactic Gammaproteobacterium Achromatium were known to form carbonate phases in undersaturated solutions (Head et al., 1996; Couradeau et al., 2012; Salman et al., 2015; Cam et al., 2016; Benzerara et al., 2020).

Polyphosphate inclusions have been observed within cells of many bacteria, microalgae, diatoms, fungi, yeasts, plants, and animals (Kulaev et al., 2004; Achbergerová and Nahálka, 2011). PolyP are linear polymers of orthophosphate linked together by high-energy phosphoanhydride bonds (Kornberg et al., 1999; Achbergerová and Nahálka, 2011), whose energy is released and used for biological reactions upon breaking (Kulaev, 1975). The PolyP storage usage for microorganisms is multiple, among which are energy storage (Bonting et al., 1991; Kornberg, 1995), motility (Seufferheld et al., 2008; Varela et al., 2010; Möller et al., 2019), chaperone activity (Kampinga, 2014), chelator for metal ions (Kornberg, 1995), virulence, or even antibiotic resistance (Wang et al., 2018; Bowlin and Gray, 2021). Alternatively, Lechaire et al. (2002) hypothesised that PolyP accumulation may serve for oxygen storage. The current paradigm suggests that few bacterial genera, e.g., the marine sulphur oxidisers Thiomargarita and Beggiatoa, are particularly efficient in the sequestration of P and accumulation as PolyP under oxic conditions and hydrolyse them back into orthophosphates under anoxic/sulphide-rich conditions (Mußmann et al., 2007; Brock and Schulz-Vogt, 2011). This process appears to be involved in phosphogenesis at modern water–sediment interfaces in upwelling zones (e.g., continental margin of Peru or Namibia) (Schulz and Schulz, 2005; Crosby and Bailey, 2012; Mänd et al., 2018). Moreover, few genera of PolyP-accumulating microorganisms such as Accumulibacter sp. have been described as key components for P removal in wastewater treatment plants, enriched through anaerobic/aerobic cycling inducing P sequestration at the interface between sludge and sewage sludge (Kawakoshi et al., 2012; Kristiansen et al., 2013; Oyserman et al., 2016). However, recent discoveries of bacteria harbouring large amounts of intracellular PolyP under constant anoxia challenge this paradigm. This is the case for bacteria affiliated to the Sulfurimonas genus in the water column of the Baltic Sea (Möller et al., 2019) and MTB affiliated to the Magnetococcaceae family from the water column of both Lake Pavin and the Black Sea (Rivas-Lamelo et al., 2017; Schulz-Vogt et al., 2019). Although PolyP inclusions were frequently described in MTB cells (Lins and Farina, 1999; Bazylinski et al., 2004, 2013; Oestreicher et al., 2011; Lefèvre et al., 2012), the capability of P hyperaccumulation seems yet to be specific of MTB affiliated to the Magnetococcaceae family (Cox et al., 2002; Eder et al., 2014; Chen et al., 2015; Schulz-Vogt et al., 2019; Li et al., 2020b). Both diversity and unicity of biological and environmental parameters governing these processes of sequestration of large amount of P remain to be better understood.

The genetic and metabolic bases of PolyP and CaCO₃ hyperaccumulation in different MTB groups remain unknown. In each group, the sequestration of these elements may be the response to their stratification according to specific environmental conditions that remain to be determined. MTBc with and without PolyP were both observed by Rivas-Lamelo et al. (2017) in the water column of Lake Pavin. The factors related to this cellular heterogeneity remain unknown. In the absence of cultivated models, an in situ characterisation of the geochemical and physicochemical parameters structuring populations and their bioremediation capabilities can directly or indirectly provide answers, and to a larger extent, it may help in designing a cultivation protocol. Among known MTB habitats, the OATZ that stretch over a few to several metres in some stratified water columns are particularly suitable to conduct such investigations, in comparison with OATZ in aquatic sediments which are usually restricted to a few millimetres in thickness. The permanently stratified water column of Lake Pavin therefore represents a model site (Sime-Ngando et al., 2016), showing oxygen, chemical, and redox gradients extending over a few metres in depth (Michard et al., 1994; Sime-Ngando et al., 2016).

In the present work, we investigated the depth stratification of the diverse MTB populations in relation to environmental parameters in the water column of Lake Pavin. Using a recently developed methodology coupling in situ measurements and water sampling with a high vertical resolution (Busigny et al., 2021), we localised the depth range in the OATZ where the MTB community thrive. We collected water samples from a transect at the redox boundary with a 20-cm interval and examined the MTB morphotype diversity and abundance using magnetic sorting and light microscopy approaches. A classification schema of MTBc is proposed, based on the ultrastructural heterogeneity of PolyP inclusions and magnetosome organisation using transmission electron microscopy. Statistical relationships between the abundance of each MTB morphotype and different physicochemical and geochemical parameters were explored. We identified the biogeochemical niches of MTB populations, in particular PolyP-forming MTBc, and we suggested some key factors associated with metabolisms that could lead to their structuration in natural environments.

In this study, we evidenced a vertical stratification of MTB in the water column of Lake Pavin in October 2019, below the OATZ. It should be noticed that potential seasonal effects were out of the scope of this study. The MTBc, capable of sequestering large PolyP inclusions, dominated the MTB profile with a proportion comprised between 52 and 94% of all MTB at any depth in the MTB zone. The abundance of MTBc was correlated with the particulate organic carbon and nitrogen contents (Figure 5). The C/N molar ratio average value calculated in our samples is 6.63, which corresponds to the Redfield ratio for aquatic biomass (6.625) (Tyrrell, 2001). The ratio of MTBc versus other microorganisms in the water column of Lake Pavin only accounted to ∼1:1,000 (with abundance of bacteria and archaea estimated to 10⁶ cells ml^–1 by Lehours et al., 2005). Rather than a causal impact of MTBc biomass on the organic N and C contents, these fractions of the particulate organic matter are more likely to correspond to a proxy for total biomass, previously evidenced to be represented at these depths by bacteria and archaea (Lehours et al., 2005). The optimal conditions for prokaryotic biomass development in this anoxic layer might also favour the MTBc growth.

Two main populations of MTBc with different magnetosome organisations were found along the water column of Lake Pavin with a maximum of abundance at different depths. MTBc with two single chains largely dominated the MTB abundance profile (>57%). The maximum abundance of this major morphotype was located at the same depth than the maximum of total MTB abundance (51.3 m). The minor MTBc morphotype with disorganised magnetosomes analogous to those described by, e.g., Koziaeva et al. (2019) and Liu et al. (2020), was most abundant at the same depth than the iACC-forming rods (52 m). Both MTBc morphotypes were located under anoxic conditions. The diversity of magnetosome organisations observed in MTBc suggests that there is more than one species of magnetotactic coccus in the water column of Lake Pavin. This hypothesis is based on a recent study demonstrating that MTBc populations with magnetosomes differing in number, crystal size, axial ratio, and chain configuration belong to different species (Liu et al., 2020). Moreover, this hypothesis is also reinforced by previous phylogenetic analyses made on Lake Pavin MTBc, which showed the presence of two different species based on the 16S rRNA gene sequence (Rivas-Lamelo et al., 2017).

The proportion of MTBc able to accumulate intracellular PolyP (from ∼35% at 51.3 m up to ∼75% at 52-m depth) was independent of the MTBc abundance. In addition, the PolyP sequestration capability was significantly higher at 52-m depth for both morphotypes of MTBc (i.e., two single chains and disorganised magnetosomes). This suggests that phosphorus hyperaccumulation capability depends on environmental conditions affecting both morphotypes.

Significant correlations were found between the concentration of dissolved sulphate and MTBc-related quantitative variables, providing clues about some metabolic traits of these bacteria (Figure 5). According to previous profiles of dissolved sulphate concentration monitored along the whole water column of Lake Pavin (Busigny et al., 2016; Miot et al., 2016), the local maximum of dissolved sulphate concentration observed within the MTB zone in this study corresponds to the global maximum within the water column. A positive correlation between the concentration of dissolved sulphate and the abundance of MTBc (as determined by light microscopy) could suggest that either sulphate is a preferred metabolic substrate for MTBc (e.g., for sulphate reduction) or a by-product of their metabolism (sulphoxidation) and/or of the metabolism of bacteria living in the same biotope. The observation of intracellular S granules suggests a sulphoxidation metabolism (Maki, 2013). At the cell scale, the TEM characterisation of inclusions of the major MTBc population (i.e., with two single chains of magnetosomes) suggests that cells accumulating large PolyP inclusions harbour none or small sulphur granules (Supplementary Figure 5). The reduced size of sulphur granules could be due to intracellular space limitations or to oxidation of the sulphur granules to gain energy for PolyP storage, the latter being previously evidenced by Kornberg (1995). It has been reported that P hyperaccumulators found in the environment are sulphur oxidisers (Brock and Schulz-Vogt, 2011; Möller et al., 2019). They have always been reported in the vicinity of an oxycline or within a gradient of sulphide concentration. For example, bacteria belonging to the Sulfurimonas genus were reported in an euxinic zone (anoxic and >0.1 μM of sulphide) of the water column of the Baltic Sea, where they need both nitrate and sulphide for their energy metabolism and to accumulate PolyP (Möller et al., 2019). In contrast, MTB affiliated to the Magnetococcaceae family from the suboxic layer (<0.1 μM oxygen and <0.1 μM sulphide) in the water column of the Black Sea store PolyP near the depth of oxygen disappearance and keep them under suboxic conditions until they hydrolyse them upon reaching the sulphide-rich zone of the water column (Schulz-Vogt et al., 2019). These two situations are distinct from the current paradigm based on Beggiatoa and Thiomargarita behaviours for which PolyP are accumulated under oxic conditions but are hydrolysed under both anoxic and sulphide-rich conditions (Schulz and Schulz, 2005; Brock and Schulz-Vogt, 2011). Similar to MTBc from the Black Sea and Sulfurimonas from the Baltic Sea, MTBc from Lake Pavin are also capable of keeping a high PolyP content in anoxic water. However, in contrast to the case observed in the Black Sea, the Lake Pavin phosphorus accumulators seem to accumulate PolyP not near the OATZ but rather deeper in the water column, close to the depth where the concentration of dissolved sulphide starts to increase (according to previously published profiles of total dissolved sulphide concentration in the water column, e.g., Michard et al., 1994; Bura-Nakić et al., 2009, 2013; Busigny et al., 2016; Rivas-Lamelo et al., 2017). Therefore, MTBc from Lake Pavin appear to have a metabolism closer to the Sulfurimonas from the Baltic Sea, for which sulphide is required to accumulate PolyP.

The nature and function of the empty vacuoles observed within MTBc at every depth in the water column of Lake Pavin are unclear. In the large sulphur-oxidising Beggiatoa, Thiomargarita, and Thioploca, empty vacuoles correspond to a nitrate-storing space for anoxic respiration and energy conservation (McHatton et al., 1996; Schulz, 1999; Zopfi et al., 2001). The closest relative to the Lake Pavin MTBc, the type strain Magnetococcus marinus MC-1 (Bazylinski et al., 2013), has been shown to harbour the set of genes required for denitrification (Philippot, 2002). Such metabolic capability has also been hypothesised for MTBc from the Black Sea (Schulz-Vogt et al., 2019). Moreover, some MTB affiliated to the Nitrospirae phylum also presenting empty vacuoles were suggested to couple sulphoxidation with denitrification (Li et al., 2020a). Thus, the MTBc of Lake Pavin could gain energy from their intracellular sulphur granules (i.e., sulphur oxidation) and potential intracellular nitrate (i.e., denitrification) for the accumulation of P as PolyP. The potential contribution of the MTBc metabolism to the nitrate depletion in the anoxic layer of the Lake Pavin water column (Supplementary Table 1; Miot et al., 2016) remains to be explored.

The sulphur/sulphide-oxidising bacteria (SOB) and sulphur/sulphate-reducing bacteria (SRB) are usually found around the redoxcline, in suboxic or euxinic waters (Vliet et al., 2020). In a recent study using 16S rRNA analyses in the water column of Lake Pavin, it has been shown that MTBc were at their highest abundance at the same depth where the SOB and SRB are both present, the former being dominant (Berg et al., 2019). The SRB become dominant deeper in the water column. Our observations evidenced by TEM the succession as a function of depth of (i) MTBc with two single chains of magnetosomes, followed by (ii) MTBc with disorganised magnetosomes, then more deeply (iii) the Magnetococcaceae rods accumulating PolyP, and (iv) potential sulphate-reducing Deltaproteobacteria, as suggested by their ultrastructure. These observations confirm the vertical succession of SOB and SRB in the water column. An important question is whether the succession of the three Magnetococcaceae morphotypes is due to competition or a difference of metabolism. The three of them accumulate sulphur granules intracellularly, suggesting they might all be SOB. Dissolved sulphate increases slightly around the MTBc abundance peak (Figure 5B). This increase could be due to oxidisation of intracellular sulphur storage by MTBc and species of the Sulfuritalea genus that represent the most abundant SOB in the water column (Berg et al., 2019). The Sulfuritalea are known to oxidise thiosulphate, S°, and hydrogen, but not sulphide, into sulphate as an end product. They can also use nitrate as an electron acceptor. Since some MTBc are known to use sulphide but Sulfuritalea do not, it might explain their coexistence at the same depths as they do not have to compete for electron donor but only for electron acceptor. Nitrate accumulated intracellularly could then be a good electron acceptor for the MTBc to thrive at these depths.

Previous studies of Lake Pavin showed an increase in the concentration of dissolved P below the OATZ reaching up to ∼300 μM deeper in the water column (Rivas-Lamelo et al., 2017). In the present study, this concentration starts to increase with depth immediately below the depth where the proportion of MTBc with PolyP was the highest. After a time of P deprivation when P becomes available again, some organisms such as microalgae have the ability to store more P than necessary in case of future shortage (Solovchenko et al., 2019); this ability is called luxury uptake. In the environmental context of Lake Pavin, we suggest that MTBc harbour a dynamic movement within their habitat. Upon reaching waters with a higher P concentration when moving downward, MTBc might store PolyP to a greater extent, preventing future P deprivation. MTBc might swim back upward towards more optimum conditions (i.e., at the depth where they are the most abundant) where they might hydrolyse their PolyP inclusions, as suggested by a lower proportion of MTBc with large PolyP. The imbalance between synthesis and hydrolysis of PolyP by MTBc remains to be evaluated along the water column. In Li and Dittrich (2019), cyanobacteria were shown to harbour a high content of PolyP during the lag phase due to an “overplus” uptake of P. By contrast, the amount of PolyP was lower during the exponential growth as a result of a competition between the P luxury uptake and its utilisation for growth. Similarly, MTBc could present a lower amount of PolyP at their peak of abundance (e.g., at 51.3 m), due to active growth, whereas they may contain higher amounts of PolyP deeper in the water column (e.g., at 52 m), where they might be in a growth dynamic closer to that of a lag phase. This hypothesis would support the luxury uptake hypothesis for these MTBc in combination with their dynamic movement along the water column.

Particulate Mg and P were found to be significantly correlated with the MTBc abundance (Figures 5B,C). PolyP are composed of negatively charged molecules and are commonly associated with counter ions with a valence of either +1 or +2 (Peverly et al., 1978; Kulaev et al., 2004). Our XEDS results showed that magnesium was a major component of PolyP in Lake Pavin (Supplementary Figures 2, 3). Therefore, at least at these depths, PolyP might be a major carrier of particulate Mg explaining the correlation of particulate Mg with the MTBc. This association has already been reported in MTBc from Lake Pavin (Rivas-Lamelo et al., 2017). The association of PolyP with Mg has also been shown for hyperaccumulators of P, e.g., Sulfurimonas in the Baltic Sea (Möller et al., 2019) and the giant sulphur-oxidising Beggiatoa (Brock et al., 2012). In addition to Mg as counter ion for PolyP in our samples, two other counter ions with a lower abundance were found to be associated with the PolyP, i.e., potassium and calcium. These two counter ions were shown to be significantly and positively correlated together.

In Lake Pavin, particulate P below the OATZ was previously evidenced to be hugely dominated by extracellular iron–phosphate particles, driving the turbidity profile (Cosmidis et al., 2014; Busigny et al., 2016; Rivas-Lamelo et al., 2017). The correlation between the MTBc abundance and the particulate P observed in the present study indicates that the MTB habitats overlap with the phosphogenesis zone. The systematic observation of cells encrusted within Fe–P minerals at these depths was formerly hypothesised as the result of several mechanisms, including PolyP hydrolysis or iron oxidation by microorganisms (Cosmidis et al., 2014; Miot et al., 2016; Berg et al., 2019). The observation of MTBc with low capability to accumulate intracellular PolyP in the phosphogenesis zone supports this hypothesis of PolyP release. The contribution of the dynamics of the P metabolism of MTBc to the phosphogenesis, occurring in the anoxic layer of the water column of Lake Pavin, remains to be determined.

Our results illustrate that magnetotaxis in MTBc of the Lake Pavin water column may not function in conjunction with aerotaxis, unlike for most known microaerophilic MTB that find efficiently their optimal oxygen concentration thanks to a magnetically assisted aerotaxis, the so-called magneto-aerotaxis (Frankel et al., 1997; Bennet et al., 2014; Lefèvre et al., 2014). MTBc would rather take advantage of their orientation with the Earth’s magnetic field to swim in one direction using the redox gradient to swim back and forth between their biotopes of interest, in particular to support their P and S metabolisms. In MTBc, magnetosome formation would thus represent a selective advantage to use magnetotaxis in conjunction with redoxtaxis (i.e., magnetically assisted redoxtaxis, or magneto-redoxtaxis) to efficiently migrate to and maintain the position at their preferred redox conditions (Figure 7).

It would be interesting in the future to better understand the P fluxes between the MTBc and the local environment and thus determine the conditions of active storage or hydrolysis of intracellular PolyP by MTBc. Their PolyP metabolism could also be explored by establishing the gene expression profiles in the water column using transcriptomics and evaluating the potential relations with P sequestration or hydrolysis. Lake Pavin represents a very promising natural laboratory to understand not only how PolyP are formed but also other types of biomineralisation such as iACC or magnetosomes. This study cannot identify a specific biogeochemical niche for the iACC-forming rods or the other minor morphotypes of MTB. The reasons could be that (i) a high biomass concentration is necessary for efficient statistical analyses along the entire MTB profile and (ii) parameters which structure their vertical distribution were not measured (e.g., pCO₂). Future metagenomic studies will be helpful in order to reveal the different populations of MTB present in the water column of Lake Pavin and their functional diversity and will also give insight into the molecular pathways involved in the MTBc P and S metabolisms. We thus believe that our results are paving the way for future studies on other magnetotactic microorganisms that have this unique advantage to be magnetically concentrated and present a metabolic, taxonomic, and ultrastructural diversity of interest for a large community of scientists.

The original contributions presented in the study are included in the article/Supplementary Material, further inquiries can be directed to the corresponding author.

CB, CM, CL, NM, and ÉD designed the experiments. All authors contributed to water sampling and manuscript revision. CB and NM acquired and interpreted the STEM and XEDS data. CL and VB described the MTB morphotype diversity and abundance using light microscopy. DJ and ÉV acquired the physicochemical profiles and performed the chemical analyses of dissolved and particulate fractions. CM performed the statistical analyses. CB, CM, and ÉD wrote the manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.