Cable bacteria at oxygen‐releasing roots of aquatic plants: a widespread and diverse plant–microbe association

Summary Cable bacteria are sulfide‐oxidising, filamentous bacteria that reduce toxic sulfide levels, suppress methane emissions and drive nutrient and carbon cycling in sediments. Recently, cable bacteria have been found associated with roots of aquatic plants and rice (Oryza sativa). However, the extent to which cable bacteria are associated with aquatic plants in nature remains unexplored. Using newly generated and public 16S rRNA gene sequence datasets combined with fluorescence in situ hybridisation, we investigated the distribution of cable bacteria around the roots of aquatic plants, encompassing seagrass (including seagrass seedlings), rice, freshwater and saltmarsh plants. Diverse cable bacteria were found associated with roots of 16 out of 28 plant species and at 36 out of 55 investigated sites, across four continents. Plant‐associated cable bacteria were confirmed across a variety of ecosystems, including marine coastal environments, estuaries, freshwater streams, isolated pristine lakes and intensive agricultural systems. This pattern indicates that this plant–microbe relationship is globally widespread and neither obligate nor species specific. The occurrence of cable bacteria in plant rhizospheres may be of general importance to vegetation vitality, primary productivity, coastal restoration practices and greenhouse gas balance of rice fields and wetlands.


Introduction
Plant root-microorganism interactions are crucial for plant health and increasing crop yields (Berendsen et al., 2012;P erez-Montaño et al., 2014). The root microbiome is composed of microorganisms that inhabit the endosphere (inside of the roots), the rhizoplane (surface of the roots) and the rhizosphere, the soil around the root that is influenced by root processes (Gaiero et al., 2013). In most freshwater and coastal sediments, as well as in rice fields, oxygen only diffuses into the uppermost millimetres (Revsbech et al., 1980;Nicolaisen et al., 2004;Sobek et al., 2009). Below the oxic zone, sulfate is reduced to sulfide, which can accumulate to phytotoxic levels (Lamers et al., 2013). Radial oxygen loss (ROL), which is the release of oxygen from plant roots, is facilitated by root-internal lacunar tissue called the aerenchyma. ROL has been hypothesised to be a defence mechanism against toxic sulfide through abiotic and biotic oxidation (Brodersen et al., 2015).
Filamentous, sulfide-oxidising bacteria, called cable bacteria, uniquely influence aquatic geochemistry as they enable the spatially separated reactions of sulfide oxidation and oxygen reduction through a flow of electrons along their internal wires over centimetre distances (Pfeffer et al., 2012;Meysman et al., 2019). This process is called electrogenic sulfide oxidation and is usually associated with the uppermost 1-3 cm of sediment. Oxygen from the water column has been recognised as one of the dominant controlling factors for successful cable bacteria growth (Burdorf et al., 2018;Marzocchi et al., 2018;Liu et al., 2021). Known cable bacteria belong to the class Deltaproteobacteria (now redefined as phylum Desulfobacterota (Parks et al., 2018) and are divided into two described candidate genera: Candidatus Electrothrix found in marine systems and Candidatus Electronema found in freshwater systems (Trojan et al., 2016). Cable bacteria absolute abundances can be determined using fluorescence in situ hybridisation (FISH), while relative abundances can be derived from 16S rRNA gene sequencing of the microbial community (Trojan et al., 2016;Li et al., 2020;Liu et al., 2021). The latter however can result in unreliable estimates as extraction of DNA from cable bacteria can be difficult due to the rigid filament structure (Trojan et al., 2016), and biases can be introduced 2138 New Phytologist (2021) 232: 2138-2151 during the analysis by technical limitations and possible variations in gene copy number among cells (B alint et al., 2016;Zinger et al., 2019).
Recently, cable bacteria have been found in mangrove sediments  and around oxygen-releasing roots of seagrass (Martin et al., 2018a(Martin et al., , 2019, saltmarsh Salicornia europaea, freshwater plants (including Littorella uniflora and Lobelia cardinalis) and rice (Scholz et al., 2019). It has been proposed that this may protect the roots against sulfide intrusion (Martin et al., 2018a) and indirectly suppress methane emissions . The studies that investigate cable bacteria around roots of aquatic plants have so far been predominantly based on laboratory grown plants (Martin et al., 2018a;Scholz et al., 2019Scholz et al., , 2020 and the natural occurrence of plant-associated cable bacteria is unknown.
Given that oxygen from roots and sulfide from sulfate reduction are common traits in aquatic sediments and waterlogged soils, we hypothesised that the plant-cable bacteria relationship occurs globally and involves diverse plant species and life stages in marine, freshwater and terrestrial environments such as lakes and rice fields. To explore this hypothesis: (1) We used 16S rRNA gene amplicon sequencing of newly generated and public datasets to search for the occurrence of cable bacteria in 14 seagrass species from four continents, five freshwater plants from lakes in Europe, eight salt marsh plants in China and rice from USA, China, India and Vietnam.
(2) We then visualised and quantified cable bacteria around plant roots collected from the field using FISH.
(3) We also used FISH in combination with oxygen planar optodes to determine whether cable bacteria inhabit rhizospheres of seagrass seedlings and whether they co-occur with oxygen release in these developing rhizospheres.
In the meromictic (water layers do not intermix) Lake Cadagno, constant inflow of sulfate-rich groundwater causes high sulfide concentrations in the porewater, up to 900 µM (Putschew et al., 1996;Xiong et al., 2019). In the oligotrophic Lake Hampen (Christensen & Sørensen, 1986) no sulfide was evident, judged from white sandy sediment without any odour. In Lake Knud, the presence of sulfide was recognised by a foul smell and dark-coloured sediment in certain spots caused by the presence of iron sulfides. The sites in Aggersund and Arcachon were both located in marine environments.

Sample collection at field sites
Intact sediment cores with plants were retrieved with a spade or plastic cores (inner diameter > 10 cm), and either transported back to the laboratory or directly subsampled in the field. The sediment cores were opened and roots with adhering sediment or the sediment from the oxic/anoxic transition zone around oxygen-releasing roots were sampled. Samples for nonplant associated cable bacteria were collected from the oxic/anoxic transition zone at the surface of the bulk sediment (at least 3 cm away from the plant shoots).
Oxygen-releasing roots were identified by red iron oxide precipitates on the root surface, depletion of black iron sulfide around the roots, or by oxygen-sensitive planar optode measurements. Samples taken in the field for DNA extraction were transported on ice or dry ice and stored at À80°C. Samples for FISH were either mixed with ethanol (v/v, 50 : 50) and stored at À20°C in the laboratory until further processing within a few weeks, or fixed in a 4% formaldehyde solution in phosphatebuffered saline (PBS) for 2 h. Fixed roots were then washed twice in 19 PBS and stored in 19 PBS/96% ethanol (v/v, 30 : 70) at À20°C until further processing. The presence of sulfide was qualitatively assessed from the colour of the sediment and the odour.

Collection of seagrass seeds and aquaria design
Fruits of the seagrass Posidonia australis were collected directly from adult plants at a maximum depth of 5 m from Woodman Point, Western Australia in November 2016 (32°08 0 17.7 00 S, 115°45 0 48.2 00 E). Sediment samples were collected from the same site using plastic cores (inner diameter: 10 cm). Fruits were transported back to the laboratory and placed in an aerated aquarium (c. 350 l, S = 34). Fruits were kept under natural light with constant flow and aeration until the fruits naturally dehisced (c. 1 wk; Fig. S1a). Healthy germinated seeds were then selected and planted into three specially designed measuring aquaria containing sieved (2 mm) sediment from the sample site. The sediment was a silty sand with a porosity of 46.8 AE 4.5% v/v (measured by saturation), bulk density of 1.4 AE 0.1 g cm À3 and organic matter content of 2.6 AE 0.05% dry weight (combustion for 4 h at 450°C; Erftemeijer & Koch, 2001). The measuring aquaria, here designated rhizoboxes, were designed to monitor root growth and oxygen dynamics in the rhizosphere of P. australis seedlings and consisted of 10-mm thick acrylic with a larger upper compartment for shoot growth (300 mm W 9 200 mm H 9 150 mm D), and a thinner lower compartment (300 mm W 9 330 mm H 9 20 mm D) containing the sediment ( Fig. 1 (upper right panel), S1b). The lower compartment was angled at 45°and was equipped with a detachable front plate fitted with a thin (c. 0.5 cm) strip of neoprene that was held together with screws to ensure it was watertight. The front plate was fitted with an oxygen planar optode (see section below) and the entire lower compartment was covered in insulating foil to prevent interference from ambient light. The aquaria operated as a recirculating system when temperature and salinity were maintained at 24°C and S = 34 in a larger separate reservoir containing filtered seawater. Plants were grown under an irradiance of 250 mol E m À2 s À1 under a 14 h : 10 h, light : dark cycle for 4 wk.

Rhizobox imaging and sampling of seagrass seedlings
The two-dimensional oxygen distribution in the rhizosphere of P. australis seedlings was mapped using oxygen-sensitive planar oxygen optodes as described in Martin et al. (2018a). The oxygen planar optodes were sufficiently transparent to discern roots growing behind them, which allowed the continual monitoring of root growth and length of oxic area (mm) along the roots. After 3 wk of incubation, the rhizoboxes were imaged each day for 1 wk. The extent of the oxygenated area around the roots of P. australis seedlings was quantified at day 28 of the incubation and expressed as maximum oxygen concentration of the oxic

Marine Strait
Aggersund, Denmark sp. Fig. 1 Global distribution of samples with confirmed plant-associated cable bacteria (based on 16S rRNA gene sequences) and fluorescence in situ hybridisation (FISH) detection of cable bacteria. Samples from 16S rRNA gene sequences include the rhizosphere, rhizoplane, endosphere or paddy soil. Red arrows in photographs point towards the roots. Blue and yellow arrows in images from FISH point towards thin and thick cable bacteria filaments, respectively. Lower FISH images show hybridisations with probes DSB706, specific for Desulfobulbaceae (red) and EUB-MIX targeting most bacteria (green), counterstained with DAPI (blue). Cable bacteria cells appear whitish/yellow from overlay of the two probes and DAPI stain. The upper right FISH image shows hybridisations with the probes Delta 495a-c specific for Deltaproteobacteria (green), EUB-MIX (red) and DSB706 (blue). Cable bacteria appear whitish from overlay of all three probes. Bars in FISH images, 5 µm. The photograph of the laboratory incubation shows the rhizoboxes that were used for monitoring root growth and oxygen release from P. australis seedlings (Supporting Information Fig. S1b).
area, as well as total width of ROL and oxic length of roots of the same age from three replicate plants (each in a separate rhizobox) using IMAGEJ software (Rueden et al., 2017). Total width of ROL was taken from the widest point and included the diffusion of oxygen from both sides of the root. At day 28 of incubation, three replicate roots (one plant per rhizobox) were collected for FISH analysis. For FISH sampling, rhizoboxes were carefully opened and single roots from a seedling from each separate rhizobox were cut at the root tip (c. 5 mm, which includes the root cap and elongation zone with no root hairs) and at the root hair zone (c. 20-40 mm from the root tip). The unwashed root segments were immediately placed into a 4% formaldehyde (v/v) solution in PBS and fixed overnight at 4°C. Roots were then washed in 19 PBS three times and stored in a 19 PBS/96% EtOH (40 : 60) solution at À20°C (Schmidt & Eickhorst, 2014).

Fluorescence in situ hybridisation and quantification
Samples from Lake Hampen, Lake Knud, Lake Cadagno and Aggersund were analysed by FISH following the protocol as described earlier (Pernthaler et al., 2001;Scholz et al., 2019).
The oligonucleotide probes DSB706 (Loy et al., 2002) labelled with Cy-3 for the detection of filamentous Desulfobulbaceae and EUB-MIX (Amann et al., 1990;Daims et al., 1999) labelled with Atto-488 at a formamide concentration of 35% were used. The negative control probe NON-EUB (Manz et al., 1992) labelled with Atto-550 was run along on each multiwell slide. The samples were counterstained with 4 0 ,6-diamidino-2-phenylindole (DAPI; 1 mg ml À1 ) and the presence or absence of cable bacteria was determined using a fluorescence microscope. Cable bacteria in three replicates of the rhizosphere and surface samples from Aggersund were quantified by measuring the length of all positively stained filaments in each well using the imaging software NIS-Elements (v.4.50; Nikon Instruments Inc., Melville, NY, USA) and the results were expressed as cable bacteria filament densities (m cm À3 ).
FISH and subsequent quantification of cable bacteria on the root segments (c. 0.5 cm) of either root tips or root hair zones of the P. australis seedlings grown in the rhizoboxes was carried out as described in Martin et al. (2018a). The probe Delta 495a-c (Loy et al., 2002) with double-labelled 6-Fam for the detection of Deltaproteobacteria, the probes EUB-MIX with doublelabelled Cy5 and the DSB706 probes labelled with Atto 565 at a formamide concentration of 35% were used. The NON-EUB probe was labelled with Atto 565, 6-Fam-6-Fam or Cy5-Cy5. The root segments were then analysed by confocal laser scanning microscopy on a Nikon Ti-E inverted microscope with a Nikon A1Si spectral detector. Manual counting of cells hybridised in each laser channel was performed using IMAGEJ software and the Cell Counter plug-in (https://imagej.nih.gov/ij/plugins/cellcounter.html). To compensate for nonspecific binding, manual cell counting was also performed on the negative controls, and any cell numbers counted in the negative controls were subtracted from the corresponding sample counts. Each replicate was the average of 15 images and the results were expressed as cable bacteria cell densities (cells mm À2 root).

DNA extraction and sequencing
DNA was extracted from root and bulk samples (total amount of samples: 42) from Lake Hampen, Lake Knud, Lake Cadagno, Aggersund and Arcachon using the DNeasy PowerLyzer PowerSoil Kit (Qiagen), according to the manufacturer's protocol but with 60 µl instead of 100 µl elution buffer. Primers Bac341F (CCTACGGGNGGCWGCAG) and Bac805R (GACTACHVGGGTATCTAATCC) were used to amplify variable regions V3 and V4 of bacterial 16S rRNA genes (Herlemann et al., 2011). The 16S rRNA gene amplicon libraries were prepared according to Illumina's 16S Metagenomic Sequencing Library Preparation guide with three consecutive PCR reactions for amplification of the target regions (1 st PCR: 20 cycles), addition of adapters (2 nd PCR: 10 cycles) and indexes (3 rd PCR: 8 cycles). Sequencing was carried out on a MiSeq desktop sequencer (Illumina, San Diego, CA, USA). The raw sequences can be downloaded from the NCBI/EMBL-EBI/DDBJ Sequence Read Archive under BioProject ID 'PRJNA680155'.

Retrieval of public datasets
Criteria to choose public datasets for analysis were studies that involved next generation sequencing of the 16S rRNA gene of the microbial community in root-associated or paddy soil samples. Raw sequences and metadata from 21 public datasets (Nguyen et Table S1).
A list of all samples analysed including the relevant metadata can be found in Table S2. To generalise the different sampling strategies used in each study, we termed 'root samples' to mean all samples that include roots and/or sediment that was washed off or sonicated free from roots in sterile buffer. 'Bulk samples' referred to root-free zones except for samples from rice fields. Rice studies that reported actual root samples were Edwards et al. (2018) and Ma et al. (2020) (Table S2). Samples from all other rice studies were termed bulk samples as they were collected from the bulk paddy soil. However, these samples may have contained rhizosphere soil or root fragments.
The 16S rRNA dataset from Edwards et al. (2018) included 1510 samples from rice fields in California and Arkansas, USA that were taken throughout the growing season over 3 yr (2014)(2015)(2016) consecutively. In this study, the rhizosphere was sampled from roots by vortexing the roots in PBS solution and collecting the slurry. The rhizoplane was sampled from cyclically washed and visibly clean roots by sonication and collecting the pellet after centrifugation. The endosphere fraction was achieved by grounding the washed and sonicated roots (in total 3 9 for 30 s). Bulk samples were collected from a 0.5-m-wide walking lane in the rice field.

Bioinformatic analysis
For both public and our newly generated data, primers were trimmed using CUTADAPT (v.1.16;Martin, 2011) and trimmed reads were processed using DADA2 (v.1.16.0, https://benjjneb. github.io/dada2/; Callahan et al., 2016). Taxonomy was assigned following the standard protocol of the DADA2 pipeline, modifying when necessary to accommodate single reads and to remove poor quality reads. The Ribosomal Database Project (RDP) classifier (Wang et al., 2007) and the SILVA SSURef NR 132 database (Quast et al., 2013) were used. The large dataset from Edwards et al. (2018) was divided into four groups that were run separately through the DADA2 pipeline and then merged into a full-study sequence table before chimera removal using the 'consensus' method and taxa assignment. Sequences assigned as 'Chloroplast' and 'Mitochondria' were removed from each individual study. The assigned amplicon sequence variants (ASVs) were checked for the cable bacteria genera Ca. Electrothrix and Ca. Electronema. The data were analysed using RSTUDIO (v.  Wickham, 2016).
Tentative species assignments were generated by performing a BLAST (BLASTN v.2.2.29+; Camacho et al., 2009) search with the ASV sequences as queries against a database of representative fulllength cable bacteria 16S rRNA gene sequences (compiled from Trojan et al., 2016;Thorup, 2019). ASVs with an identity higher than 97% to a reference sequence across >90% of their length were assigned to that species. ASVs that were classified as Ca. Electrothrix or Electronema by DADA2 but lacked identity above this threshold to reference sequences are listed as unidentified. A phylogenetic tree was constructed from reference sequences using the SINA aligner v.1.2.11 (Pruesse et al., 2012) and PHYML v.3.1 (Guindon & Gascuel, 2003) software with 1000 bootstraps.

Statistics
The following analyses were carried out in R (v.3.6.0, R Core Team, 2013). Differences in variances of the cable bacteria density between the sampling regions of the FISH samples from Aggersund and the rhizobox experiment were tested using an Ftest. The variances of the cable bacteria filament densities between rhizosphere and surface sediment in samples from Aggersund were equal (significance level a = 0.05). Therefore, the differences of the means were tested using unpaired twotailed Student's t-test with a significance level of 0.05, six observations and four degrees of freedom. For the rhizobox experiment, variances of the cable bacteria cell density between the root hair region and the root tip, elongation and immature root region were unequal. Therefore, a Welch test was carried out to test the differences of the means between the root regions with a significance level of 0.05 and six observations. The degrees of freedom were 2, 2.0056 and 2.0451, respectively. Differences for the fractions of samples with cable bacteria per month in the rice fields were tested using one-way ANOVA.

Breadth of the plant-cable bacteria relationship
Cable bacteria were found by 16S rRNA sequencing roots, rhizosphere or paddy soil in 36 out of 55 investigated sites around the world and encompassing four of the five investigated freshwater plant species, 10 of the 14 investigated seagrass species and one of the eight saltmarsh plants. Cable bacteria were present in rice fields in eight out of 13 investigated sites ( Fig. 1; Table S1).
In Lake Knud, the two described freshwater cable bacteria species, Ca. Electronema palustris and Ca. Electronema nielsenii, and undefined Electronema species were identified around roots of L. uniflora and P. globulifera (Fig. 2). Both known cable bacteria species were also present in the surface sediment close to the L. uniflora meadow.
By contrast, Ca. Electronema nielsenii and the undefined Electronema species occurred in Lake Cadagno around the roots of Equisetum sp. and Potamogeton sp. (Fig. 2). Cable bacteria associated with rice and paddy soil included both known Ca. Electronema species, and two sequences classified as Ca. Electronema in the SILVA database (accession nos. GU208270 and FQ658891), as well as undefined Electronema and Electrothrix species.
Cable bacteria around seagrass and associated bulk samples were more diverse than in the freshwater environments and included all four described Ca. Electrothrix species, 15 newly proposed species (Thorup, 2019), as well as unclassified Electrothrix species. Cable bacteria diversity was higher in root samples compared with the bulk samples (16 vs 3 defined species) for H. ovalis. However, this trend was not observed in samples from Z. marina and Z. noltii, which showed high diversity in both root and bulk samples. Interestingly, undefined Electronema species appeared in root and bulk samples of H. ovalis and Z. marina.
Moreover, only undefined cable bacteria species were found around roots of the saltmarsh plant Spartina alterniflora. Taken together, no specific cable bacteria species appeared to dominate in the root or bulk samples of the freshwater, terrestrial and marine environments.
Thick and thin cable bacteria filaments were found on seagrass roots, with cable filament widths of around 4 µm and 1 µm on roots of P. australis seedlings and filament widths of around 6 µm and 1 µm in the rhizosphere of Z. marina (Fig. 1, FISH images).
In the oligotrophic Lake Hampen no cable bacteria were found associated with roots of L. uniflora nor with the bulk sediment by means of FISH and 16S rRNA gene sequencing (Fig. S2). In Lake Knud, cable bacteria were found in association with the roots of L. uniflora and P. globulifera, as well as in the bulk sediment. The maximum relative abundance of cable bacteria in the root microbiome of both plants was 0.02% and 0.01%. (Fig. 3). Phylogenetic diversity and fraction of cable bacteria species. The fraction represents the amount of samples that contain the specific cable bacteria species over the total amount of root or bulk samples per plant species. Root samples include samples from the rhizosphere, rhizoplane or endosphere. Bulk samples were collected from root-free zones or paddy soils. The absence of cable bacteria in bulk samples from certain species indicates that cable bacteria were either absent in the bulk samples or that no bulk samples were analysed (this is evident in Fig. 3). Reference 16S rRNA gene sequences without an accession number (GM3-4, LOE-1, AU1-5, ATG1, 30AUS4, EH2, 30AUS1-2, LOE-2, AR3, AR4) are from Thorup (2019 In the meromictic Lake Cadagno, cable bacteria were present around roots of Equisetum sp. and Potamogeton sp. and the highest relative abundances of cable bacteria in the root microbiomes were 0.9% and 4.07%, respectively. By contrast, roots of Isoetes sp. that were sampled from the wetland in immediate vicinity of Lake Cadagno, did not show cable bacteria by FISH or 16S rRNA gene sequencing (Fig. S2). Cable bacteria relative abundances appeared to be higher around the roots of the fastgrowing, seagrass species Zostera spp. and Halophila ovalis compared with other seagrass species. Cable bacteria relative abundances tended also to be higher on the roots compared with the bulk samples, particularly for H. ovalis. However, it was difficult to compare abundance information from 16S rRNA gene sequencing across studies and we therefore specifically quantified the cable bacteria abundances in the rhizosphere and at the surface bulk for Z. marina in Aggersund by FISH (see section below). Cable bacteria were also found in both tropical (Syringodium filiforme and Cymodocea nodosa) and temperate seagrass species (Posidonia spp. and Amphibolis antarctica; Fig. 3).
In the marine strait Aggersund (Fig. 1), cable bacteria were found in the root samples of Z. marina with a maximum relative abundance of 9.85%. The maximum relative abundance of all samples was found in a root sample of H. ovalis from Pelican Point, Australia (dataset: J.A. Middleton et al., unpublished) and was 24.21%.
Only 0.02% (12 reads total) of Ca. Electrothrix were found in one out of three root samples from Z. noltii in Arcachon, France by 16S rRNA sequencing. In comparison, the other seagrass samples from our field study showed 2-10% (1180-4574 reads total) cable bacteria for Z. marina from Aggersund in three out of four samples. Furthermore, the Ca. Electrothrix species of the sample from Z. noltii in Arcachon could not be confirmed by the BLAST search. This combination of low abundance, a single positive sample, and poor sequence identity to known cable bacteria have led to the conservative judgement not to include Z. noltii roots from Arcachon in our overview of samples with cable bacteria present (Fig. 1).
The highest relative abundance of cable bacteria around rice roots was 3.73% found in the USA (Figs 3, S3). Cable bacteria were also found associated with paddy soil in China, India and Vietnam ( Fig. 1; Table S1) when the maximum relative abundance of cable bacteria was 0.29% (see asterisk in Fig. 3).

Cable bacteria distribution and dynamics in rice fields
The re-analysis of the published 16S rRNA gene dataset from rice fields in California and Arkansas, USA (Edwards et al., 2018), indicates that the cable bacteria population followed similar growth patterns interannually throughout the growing seasons, which started in May when fields were flooded. From plotting the relative abundances of cable bacteria it appears that cable bacteria were less abundant in the months May to June compared with July to September (Fig. S3). However, this trend was strongly influenced by the high number of zero values. To unravel a clearer succession of the cable bacteria population, relative abundance was reduced to the presence or absence of cable bacteria and reported as the fraction of samples that contained at least one read of cable bacteria. The fraction of samples with cable bacteria increased from May to August in all years. The increasing fraction implies the spread of a dense localised population to a wider area and/or the growth of cable bacteria over time to densities above the detection limit of the extraction and sequencing of the cable bacteria 16S rRNA gene. In September, the fraction dropped in the years 2014 and 2015, but further increased in 2016 (Fig. 4). The differences of cable bacteria fractions between the months of the growing season were significant (one-way ANOVA; F(4,10) = 7.711, P < 0.01). Furthermore, the fraction of samples with cable bacteria decreased from the rhizosphere to the endosphere compartment. The bulk soil showed the lowest fraction of cable bacteria-positive samples (Fig. 4).

Occurrence of cable bacteria in seagrass seedlings (Posidonia australis) and spatial organisation around roots
The rhizobox experiment showed that newly emerged roots of P. australis had the greatest oxygen release, compared with older roots. Oxygen was also leaking from newly developing shoots or rhizomes at day 1 and day 5 of the incubation experiment (Fig. 5). The growing, leaking root segments were on average 10 AE 1 mm long (mean AE SE, n = 3) and the maximum oxygen concentration was 26 AE 3% (mean AE SE, n = 3) air saturation (Table S3). When the light was turned off, oxygen around most of the roots was consumed within 30 min (Fig. 5b).
The cable bacteria cell density appeared to decrease from the root hair zone (1250 AE 1020 cells mm À2 , mean AE SE, n = 3) to the zone of small, newly developed root hairs (100 AE 100 cells mm À2 , mean AE SE, n = 3) and the root tip (40 AE 40 cells mm À2 , mean AE SE, n = 3; Fig. 6a). No cable bacteria cells were found on the root cap. However, the abundance in the root hair zone was not significantly different compared with the 'small root hair zone' (P = 0.38, n = 6, Welch test), the root tip (P = 0.36, n = 6, Welch test) and root cap (P = 0.34, n = 6, Welch test). Moreover, Deltaproteobacteria followed a similar distribution and abundance to cable bacteria. By contrast, the counts of all bacteria showed no clear partitioning along the roots (Fig. S4).
The cable bacteria filament density in the oxic/anoxic transition zone in the rhizosphere of Z. marina (20 AE 4 m cm À3 , mean AE SE, n = 3) was as high as the one found in the oxic/ anoxic transition zone of the bulk surface sediment (20 AE 7 m cm À3 , mean AE SE, n = 3; P = 0.83, n = 6, unpaired two-tailed t-test; Fig. 6b).

Diversity
Our study expanded the known habitats of cable bacteria by reporting the in situ occurrence of cable bacteria around roots of various plants in a diverse set of environments, including freshwater and marine environments, as well as rice fields. The breadth of this relationship across different plant types and ecosystems, as well as the high diversity of cable bacteria associated with the roots, bulk sediments and paddy soils (Figs 1-3; Table S1), showed no evidence for species-specific associations between plants and cable bacteria. All six described candidate cable bacteria species, which have been found in various nonvegetated locations (Trojan et al., 2016), can also grow around plant roots (Fig. 2). Overall, no cable bacteria species, including the proposed and undefined candidate species, was found to be exclusively root specific.
The high cable bacteria diversity around seagrass roots appeared to match our observation of two different cable bacteria morphologies co-existing around seagrass roots. The thick and thin filaments could be genotypic features of different marine cable bacteria species. Interestingly, around roots of H. ovalis and Z. marina, undefined Electronema species were detected that would be usually expected to be found in freshwater environments (Trojan et al., 2016;Dam et al., 2021). Groundwater inflow might locally decrease the salinity and create site-specific habitats when freshwater and marine cable bacteria could co-exist. Alternatively, these undefined Electronema species may tolerate salinity.

Drivers
The comparisons of the relative abundances of root and bulk samples from the 16S rRNA gene data indicated higher relative abundances of cable bacteria at the roots of various plant species, including H. ovalis and S. alterniflora (Fig. 3). However, the analysed public datasets included root and bulk samples that did not specifically target oxygen-releasing plant roots and the top 2 cm surface sedimentthe sites at which the most of the cable bacteria cells would be expected. Therefore, the data from the public datasets may underestimate the relative abundance of cable bacteria associated with roots of aquatic plants and in bulk samples.
Furthermore, we quantified cable bacteria in the rhizosphere of Z. marina and at the oxic-anoxic interface at the sediment surface. The results revealed similar absolute abundances of cable bacteria around the roots of Z. marina and at the top surface sediment (Fig. 6b), which underlined the idea that the rhizosphere is a true expansion of the classical cable bacteria habitat. Typically, cable bacteria have been found at oxic-anoxic interfaces of sediment surfaces where they reduce the oxygen from the water column. Our results, in combination with recent studies that showed cable bacteria around worm and shrimp burrows (Aller et al., 2019;Li et al., 2020), extend the cable bacteria habitat to a 3D grid deeper down into the sediments and soils.
Roots of seagrass seedlings, P. australis, oxygenated the rhizosphere at the growing root tip (Fig. 5) whereas the root hair region did not leak oxygen. This pattern of rhizosphere oxygenation has previously only been shown for adult seagrass species (Jovanovic et al., 2015;Martin et al., 2018). However, cable bacteria were more abundant in the root hair region than at the oxygen-releasing tips of P. australis (Fig. 6a), which was also observed for roots of the seagrasses Z. muelleri and H. ovalis (Martin et al., 2018a). The observed higher cable bacteria cell abundance in the root hair region could be explained in two ways. First, cable bacteria grow along the root, with one end in the oxygenated top sediment layer or in the oxygenated rhizosphere around the root tip region. Here, the latter is supported by the observations that cable bacteria can adjust their position  with a speed of 2.9 mm h À1 . This rate of movement would be sufficiently fast to trace the downward growing, oxygen-releasing, root tip. The sulfide-oxidising cable bacteria cells located along the root surface could be stimulated by root exudates feeding sulfate reduction in the root hair region. Second, the cable bacteria may penetrate the root to directly tap into the oxygen-rich aerenchyma, while the other end consumes sulfide on the outside of the root. This second explanation would qualify the parts of cable bacteria

Research
New Phytologist filaments embedded in the roots as endophytes. Indeed, the reanalysis of the 16S rRNA gene dataset from Edwards et al. (2018) indicated cable bacteria cells in the inside of rice roots (Fig. 4). Endophytic bacteria are known to penetrate the epidermis at the site of lateral root extensions (Dong et al., 2003), which agrees with observations from Scholz et al. (2019), showing cable bacteria running along root hairs pointing towards the root surface. However, the rice root samples were separated into the root compartments (rhizosphere, rhizoplane and endosphere) by washing and sonication procedures and subsequently analysed by 16S rRNA gene sequencing (Edwards et al., 2018). Therefore, sampling artefacts cannot be entirely excluded and microscopic investigations are needed to reveal if cable bacteria are true endophytes. In rice, the cable bacteria sample fraction decreased from the rhizosphere to the endosphere compartment (Fig. 4). The observations that the abundance of cable bacteria cells decreased at the sites where cable bacteria would source the oxygen (endosphere compared with the outside of rice root; at the root tip of seagrass compared with the root hair region) is in accordance with observations from cable bacteria in nonvegetated sediments. In these sediments less than one-tenth of each filament is in the oxic zone (Scilipoti et al., 2021) and most of the cable bacteria cells are located in the reduced sediment layer below the oxic surface (Schauer et al., 2014;van de Velde et al., 2016).
In addition to the microscale oxygen availability, we found that sulfide availability in the environment is also likely to control the abundance of cable bacteria. No cable bacteria were found to be associated with L. uniflora in the oligotrophic Lake Hampen, but their presence was confirmed in the mesotrophic Lake Knud. The relative abundance of plant-associated cable bacteria was higher in sulfide-rich Lake Cadagno sediment. The similar relative abundance of cable bacteria around rice roots could be explained by the generally high turnover of organic matter in rice fields that stimulates sulfate reduction (Wind & Conrad, 1997 continuous source of substrate for cable bacteria. More detailed sulfide data are needed to disentangle how the availability of electron donor sources (free sulfide vs iron sulfide) for cable bacteria influences their abundance and activity. However, the absence or near absence of cable bacteria around roots in sulfide-rich sediments, for example Isoetes sp. at Lake Cadagno or Z. noltii in Arachon, France, indicates that the establishment of a plant-cable bacteria relationship is also determined by factors other than oxygen and sulfide availability.
The annual re-establishment of cable bacteria in rice fields after flooding (Fig. 4) as well as the presence of cable bacteria on roots in the pristine, high alpine Lake Cadagno (Fig. 1), show that cable bacteria populations are robust and not limited to open systems such as marine environments (Burdorf et al., 2017) and freshwater streams (Risgaard-Petersen et al., 2015).

Ecological consequences
From our synthesis we showed that cable bacteria form a generic member of root microbiomes of aquatic plants across a diversity of ecosystems. However, we found that cable bacteria were not associated with all roots (e.g. cable bacteria were absent in 65% of the rhizosphere samples from rice fields in the USA; Fig. 4) and that for instance, L. uniflora can live with and without cable bacteria (Figs 1, S2). The negative findings and the high diversity of plant-associated cable bacteria indicated that the association is not obligatory for either of the partners. But the positive findings with cable bacteria densities as high as in the bulk surface sediment, where their impact on geochemical cycling is recognised (Risgaard-Petersen et al., 2012;Seitaj et al., 2015), make further investigations into their potential impact on plant health worthwhile. For example, the cable bacteria-plant relationship could alleviate critical sulfide toxicity in seagrass meadows (Borum et al., 2005) and rice fields (Joshi et al., 1975). Furthermore, electrogenic sulfide oxidation can lead to acidification by 1 pH unit (Burdorf et al., 2018;Sandfeld et al., 2020), which may mobilise the essential plant nutrients iron and phosphorous (Brodersen et al., 2017). By contrast, excess iron is toxic for plants (Mehraban et al., 2008) and the potential cable bacteria-driven mobilisation of trace metals, such as arsenic in contaminated Asian rice fields, may result in increased levels of arsenic in rice grain and arsenicrelated human diseases upon intake (Meharg, 2004;van de Velde et al., 2017). However, cable bacteria may also promote the formation of iron oxide-hydroxide plaques on oxygen-releasing plant roots, which would be a firewall against toxic sulfide and arsenic uptake into roots, but also be a sink for phosphorus (Liu et al., 2004;Seitaj et al., 2015;Sulu-Gambari et al., 2016;van de Velde et al., 2017). Furthermore, cable bacteria may increase nitrogen availability for the plants as they stimulate dissimilatory nitrate reduction to ammonium (Kessler et al., 2019) and have the potential to fix N 2 (Kjeldsen et al., 2019). These biogeochemical effects of cable bacteria around the roots of aquatic plants and rice, which are likely to change on a diurnal basis, should be studied for the different plant, sediment and soil types.
The cable bacteria-mediated impact on plant health could be particularly important for seedling vigour. Here, we show for the first time both oxygen loss and cable bacteria on a Posidonia species and seedlings. This suggests a close association throughout the life of seagrasses and lays the groundwork for potential improvements in seed-based seagrass restoration techniques. Seagrass meadows are in decline across the globe, which has subsequently spearheaded significant investment in seagrass restoration. However, restoration of marine ecosystems is still considered a developing area of science and is associated with a relatively high cost compared with terrestrial restoration (Tan et al., 2020). In this context, roots precoated with naturally enriched cable bacteria could be a low-cost solution to increase seedling vigour upon transplantation.
Furthermore, inoculation of cable bacteria in rice pots has been found to reduce methane emissions . We show that cable bacteria grow in rice fields in Asia and the USA, which expands the known habitats of cable bacteria. Not only were cable bacteria found to be closely associated with rice roots (rhizosphere, rhizoplane and endosphere) in the USA, but their occurrence was also linked to the annual growing season (Fig. 4). It is vital to better understand the drivers behind the repeated peak abundances in the second half of the growing season. By contrast with rice fields in the USA, which usually remain continuously flooded throughout the growing season, common practices in Asia include the intermittent drainage of rice fields. The question whether the short-term aerations of the paddy soil would impede the establishment of peak abundances, as observed for rice fields in the USA, still remains open. Understanding the cable bacteria dynamics in rice fields and their impact on plant health, as well as metal mobilisation (e.g. arsenic), will help to develop novel management practices to stimulate continuous high cable bacteria abundances throughout the entire growing season. These practices will, therefore, potentially reduce emissions of the potent greenhouse gas methane, while avoiding negative side effects.
the Integrated Coastal Analyses and Sensor Technology (ICoAST) project with funding from the Indian Ocean Marine Research Centre, a joint partnership between The University of Western Australia (UWA), the Australian Institute of Marine Science (AIMS), The Commonwealth Scientific and Industrial Research Organisation (CSIRO) and The Department of Primary Industries and Regional Development (DPIRD) WA, USA. The authors declare no competing interests.

Author contributions
VVS, BCM, MWF, AS, LPN, GAK, NRP and IPGM conceived, designed and planned the work. VVS, BCM, RM, AS, MWF and LDWB carried out the field work. VVS, BCM, MWF and RM carried out FISH experiments. VVS and BCM carried out 16S rRNA gene sequencing. VVS, BCM, IPGM analysed the 16S rRNA gene sequencing data. BCM and MWF carried out the planar optode measurements and analysis. All authors contributed to the interpretation of the data and production of the manuscript.

Fig. S4
Abundance of all bacteria, cable bacteria and Deltaproteobacteria along roots of P. australis seedlings.

Table S1
Overview of all 16S rRNA gene datasets.

Table S2
Overview of all 16S rRNA gene samples analysed.