Enrichment incubations were performed in the laboratory to arrive at novel strains of cable bacteria. To this end, natural sediment was collected at 10 sampling sites in six different geographical locations (Table 1). Both brackish and marine sites were investigated, in order to cover a wide range of salinities. The salinity (S) in the overlying water was recorded at the time of sampling. Marine sediments were collected from the Ebro delta in Spain (S = 33), the Rattekaai salt marsh in the Netherlands (S = 30) and three subtidal sites in the Belgian Coastal Zone (BCZ) of the North Sea (S = 32). Brackish sediments were obtained from three sites in the Magazzolo estuary in Sicily, Italy (S = 2, 3 and 20), one site from the Yarra River estuary in Melbourne, Australia (S = 17), and one site in the Ebro delta in Spain (S = 10). A more detailed site description is given in the Supplementary methods.

Sediments from the deeper sites within the North Sea were collected with a Van Veen grab, which recovered the top ∼30 cm layer of the sediment. All the other sampling sites had shallow water (<0.5 m water depth), and here, the top 5–10 cm of sediment was collected manually with a plastic shovel. Sediments were stored in closed plastic containers with an overlying water layer, in the dark for a maximum of 8 weeks at room temperature (18–20°C) until the laboratory incubations were initiated.

Collected sediments were sieved (1.4 mm mesh size) and homogenized, and then packed into plexiglass core liners (4 cm inner diameter, 10 cm height) as described in Malkin et al. (2014). Three or more replicate sediment cores were incubated per site in the dark at room temperature. Each set of cores was placed in a separate non-transparent plastic box and submerged in artificial seawater (Instant Ocean® sea salt). The salinity of the artificial seawater was made to match the in situ salinity recorded (Table 1). The water was continuously aerated by bubbling with an aquarium pump.

The metabolic activity of the cable bacteria in the sediment cores was monitored through time. Cable bacteria induce a typical geochemical fingerprint in the sediment, which constitutes a suboxic zone in which neither O₂ nor H₂S are detectable (concentration <1μM) and a distinct pH profile that shows a subsurface pH maximum in the oxic zone and a pH minimum in the anoxic zone (Nielsen et al., 2010; Meysman et al., 2015). The development of cable bacteria activity was monitored by recording O₂, H₂S and pH depth profiles with microsensors as described before (Nielsen et al., 2010; Malkin et al., 2014). Once the characteristic geochemical fingerprint was identified in the incubation cores, individual filaments of cable bacteria were retrieved manually. Under a stereomicroscope, single filaments were extracted from the sediment with custom-made glass hooks and washed in drops of Milli-Q water to remove sediment particles (Pfeffer et al., 2012; Burdorf et al., 2017; Li et al., 2022). Cleaned filaments were then individually transferred into sterile PCR tubes (0.2 mL volume, VWR, United States) and stored at −20°C until used for PCR and sequencing. Filaments of species Ca. Electronema aureum GS and Ca. Electrothrix antwerpensis GW3-4 were manually picked from existing clonal cultures (Thorup et al., 2021; Hiralal et al., 2024b) and used exclusively for AFM imaging.

Individual filaments were imaged using Atomic Force Microscopy (AFM) in order to investigate the morphology of the outer surface (i.e., to verify the presence of parallel ridges that harbor the periplasmic fibers that enable LDET in cable bacteria). To this end, cleaned cable bacteria filaments were transferred to a droplet of Milli-Q on a round glass coverslip and dried at room temperature. Coverslips were then mounted to magnetic metal disks of 20 mm diameter using double-sided carbon stickers. Images were recorded on an XE-100 AFM system (Park Systems) operating in tapping mode, using an aluminum SPM probe with a tip radius of <10 nm (AppNano ACTA-200) and with a nominal spring constant of 13–77 N/m. Topography and amplitude images were recorded and processed with the software Gwyddion (Nečas and Klapetek, 2012). To link morphology and phylogeny, filaments retrieved from the same sediment were used for both AFM imaging as well as for full-length 16S rRNA gene sequencing (see section 2.4). Note however that AFM and DNA sequencing were still applied to different filaments, and sometimes, DNA sequences from different strains were retrieved from the same sediment. Therefore, a given AFM morphology was only unequivocally assigned to a strain, when all 16S rRNA gene sequences retrieved from the particular sediment were identical.

Full-length 16S rRNA gene sequences were obtained from individual filaments by implementing a nested PCR protocol. Initial cell lysis was achieved by adding 15 μL of PCR water to the PCR tube containing the isolated filament and heating the sample at 95°C for 5 min. The first PCR reaction (PCR1) was executed to amplify the complete 16S rRNA gene using universal primers 27F and 1492R (Table 2). Each PCR tube contained a total volume of 20 μL: 15 μL of lysed cells and a master mix composed of 1.5 μL PCR water, 2 μL of DreamTaq Buffer (10X), and final concentrations of 0.2 mM dNTPs, 0.5 μM forward primer, 0.5 μM reverse primer and 0.025 U/μL DreamTaq DNA Polymerase (Thermo Scientific, United States). The cycling conditions for PCR1 were: initial denaturation at 95°C for 2 min, 30 cycles of denaturation at 95°C for 45 s, annealing at 46°C for 45 s, extension at 72°C for 1:30 min and final extension at 72°C for 10 min.

For the second PCR reaction (PCR2) combinations of universal primers (27F, 1492R) and Desulfobulbaceae-specific primers (DSBB280wF, DSBB+1297R) were used: 27F-DSBB+1297R, and DSBB280wF-1492R (Table 2; Supplementary Figure S1). Because the amplification with primer pair 27F-DSBB+1297R was only successful in 42% of amplified PCR1 products, the alternative pairs 39F-DSBB+1297R and 64F-DSBB+1297R were tested to improve amplification success and sequence length. Each reaction contained a total volume of 20 μL prepared as mentioned for PCR1. PCR2 used 2 μL of tenfold dilutions of PCR1 products and cycling conditions were the same for all primer sets, differing to PCR1 cycling conditions in that annealing was performed at 50°C for 45 s and the extension was shortened to 1 min at 72°C (Table 2). Amplification of PCR2 products was confirmed via 1% agarose gel electrophoresis.

The PCR2 amplicon products were Sanger sequenced (Center for Molecular Neurology, VIB-UAntwerpen, Belgium) and the resulting sequences were manually checked and curated. A multiple sequence alignment of the individual partial sequences was conducted using MUSCLE (Edgar, 2004) before assembling them to consensus sequences of the 16S rRNA gene with lengths varying from ca. 800 to 1,500 bp.

The available high-quality (i.e., sufficiently long, ≥800 bp) 16S rRNA gene sequences in the dataset were grouped into “species-level clades” and “genus-level clusters.” Note that this clustering approach is exclusively based on DNA relatedness, i.e., similarity of sufficiently long 16S rRNA gene sequences. While it allows a good indication of the genus-level and species-level diversity, it is not sufficient to effectively name and describe taxa, as the taxonomic delineation of species and genera requires additional biochemical and other phenotypic data (Rosselló-Móra and Amann, 2015).

Our clustering approach consisted of three separate steps. In a first step, pairwise gene sequence identities were computed with the distance matrix tool in MEGA (version 10.2.5). To define a species-level clade, we applied the conventional 16S rRNA gene sequence identity cutoff of 98.7% (Yarza et al., 2014). The groupings thus obtained were recognized as genuine “species-level clades” when one additional criterium was fulfilled: the cluster should contain at least one nearly complete 16S rRNA gene sequence (sequence of ≥1,200 bp), as to increase the reliability of the taxonomic assignment.

In a second step, we constructed a phylogenetic tree, using the longest sequence of each species-level clade. This selection process was crucial for achieving a reliable phylogenetic tree, as large numbers of sequences can lead to more frequent multifurcations (Boyce et al., 2015). Initially, all cable bacterium sequences were aligned along with reference sequences of closely related Desulfobulbales species (Supplementary Table S1) using the MUSCLE algorithm (Edgar, 2004). The phylogenetic tree was calculated in IQ-TREE v1.6.12 with the automatic best-fit model finder and 1,000 ultrafast bootstrap iterations (Nguyen et al., 2015) using the 16S rRNA gene sequence of Geobacter sulfurreducens PCA (NR_075009) as outgroup. The tree was visualized in FigTree (version 1.4.4).

In a third step, we calculated pairwise gene sequence identities for the 90 reference sequences selected in the species-level clades. The conventional genus-level cutoff value of 94.5% sequence identity (Yarza et al., 2014) was used in combination with the tree topology and grouping of species-level clades to delineate genus-level clusters.

To achieve a reliable taxonomic resolution at the species level, long to full-length 16S rRNA gene sequences (800–1,500 bp) are required (Kim et al., 2011). Such sequences were obtained by targeted incubation of sediments from 10 different marine and brackish locations, followed by retrieval of individual filaments from the sediment, nested PCR application and subsequent Sanger sequencing. The first PCR reaction (PCR1) was always executed using universal primers 27F and 1492R (Table 2). For the second PCR reaction (PCR2), we initially used the primer set DSBB280wF-DSBB+1279R, generating 16S sequences of ca. 900 to 1,000 bp. To attain longer sequences, we adjusted the primer pair (27F/39Fc/64Fc-DSBB+1297R and DSBB280wF-1492R; Table 2), which provided sequences within the range from 1,464 to 1,515 bp (average length: 1,495 bp). In this manner, 125 sequences from single cable bacterium filaments were obtained with ≥800 bp, 26 of which were full-length 16S rRNA gene sequences (≥1,450 bp, Table 3). Overall, 86 sequences were retrieved from marine sediments and 39 sequences were obtained from brackish sediments.

In addition to the novel 16S rRNA gene sequences obtained from single cable bacteria filaments, existing cable bacteria sequences were compiled from the literature and public databases. A total of 206 sequences were collected from literature sources, including the pioneering study by Trojan et al. (2016) and other cable bacteria studies (Pfeffer et al., 2012; Malkin et al., 2014; Marzocchi et al., 2014; Schauer et al., 2014; Larsen et al., 2015; Xu et al., 2021; Yang et al., 2023) as well as from recently published cable bacteria genomes by Thorup et al. (2021), Geelhoed et al. (2023), Sereika et al. (2023), Hiralal et al. (2024a), Hiralal et al. (2024b), and Plum-Jensen et al. (2024). The sequences of ribosomal RNA genes in the available genomes were retrieved using barrnap v0.9.¹ Additionally, 49 full-length 16S sequences were generated through PacBio sequencing as documented in Sachs et al. (2022) (Table 3).

Cable bacteria sequences were additionally compiled from two public databases: the SILVA rRNA database and NCBI GenBank. The SILVA 138.1 small subunit rRNA database (release SSU Parc, accession date: 7 July 2024) is a public repository for 16S rRNA sequence datasets (Quast et al., 2012) which implements automatic taxonomic classification of the sequences. To obtain potential cable bacteria sequences, we extracted 1,630 sequences assigned to the genera Ca. Electrothrix and Ca. Electronema (excluding the 71 sequences that were already extracted from literature) as well as 3,839 unassigned Desulfobulbaceae sequences (Table 3). To determine whether 16S sequences were truly belonging to the cable bacteria clade, stand-alone blastn (Altschul et al., 1990; Camacho et al., 2009) was used to screen these sequences against a reference set of known cable bacteria sequences. This reference query set contained five sequences that covered the spectrum of the currently known diversity of cable bacteria: Ca. Electrothrix communis (KR912339), Ca. Electronema palustre (KP728463, Trojan et al., 2016), Ca. Electronema halotolerans (GCA_942493095, Sereika et al., 2023) and the potential genera AR3 (GCA_022765765, Geelhoed et al., 2023) and AR4 (GCA_022765725, Geelhoed et al., 2023). Note that this approach inherently bears an important constraint: we specifically screen the diversity within the main monophyletic Ca. Electrothrix/Ca. Electronema clade. Therefore, cable bacteria sequences that fall outside this clade—if such would exist—will not be retained by the approach.

NCBI GenBank (Release 255, accession date: 7 July 2024) is a genetic sequence repository which contains the nucleotide collection (nr/nt) database with publicly available 16S nucleotide sequences (Benson et al., 2018). Unlike the SILVA database, NCBI GenBank does not automatically classify sequences taxonomically. Thus, to identify potential cable bacteria 16S sequences, blastn was used to screen the default nucleotide collection (nr/nt) database of GenBank against the same reference query set as used above. A sequence identity of ≥92% between any of the five query cable bacteria sequences and the sequences in the database was used as selection criterion to extract potential cable bacteria 16S rRNA gene sequences. This way, 204 additional sequences were identified that were not included in the sequences extracted from the SILVA database nor compiled from literature (Table 3). When available, corresponding metadata was compiled from each sequence based on the information provided in the databases including the habitat type, salinity, and sulfide concentration. Sediments were classified into four categories based on their salinities: freshwater (S < 0.5), brackish (0.5 ≤ S < 30), marine (30 ≤ S ≤ 36) and hypersaline (S > 36).

The compiled database of 16S rRNA sequences was curated based on two criteria. In a first step, the 16S sequence identity was screened: only sequences that had an identity value ≥92% compared to any of the five cable bacteria query sequences (as defined in section 3.2) were retained. Of the 5,469 extracted SILVA sequences, 1802 (33%) met this 92% cutoff criterion (Table 3). This included all sequences previously assigned to Ca. Electrothrix or Ca. Electronema in literature, but with one exception. Sequence GU208270 was retained although it only showed a maximum identity of 90.9% to the query set. However, it is classified as cable bacterium Ca. Electronema in the SILVA database and has been recognized as a putative cable bacterium sequence in previous studies (Trojan et al., 2016; Scholz et al., 2021). Sequences collected from NCBI GenBank already fulfilled the sequence identity ≥92% criterion, so these were not filtered any further. Likewise, all sequences obtained from literature and single filaments fulfilled the sequence identity cutoff criterion. As a result, 2,337 sequences were retained in total as potential cable bacteria sequences (Table 3).

As a second criterion, the sequences had to phylogenetically cluster within the main monophyletic Ca. Electrothrix/Ca. Electronema clade to confirm their taxonomic affiliation as cable bacteria. To this end, the 2,337 retained sequences were aligned with closely related Desulfobulbales taxa (Supplementary Table S1) using MUSCLE (Edgar, 2004). After manually examining the alignment, 71 sequences that contained indels and mutations in conserved regions of the 16S rRNA gene were removed from the dataset. The phylogenetic placement of the aligned sequences was evaluated by constructing a phylogenetic tree in IQ-TREE. A total of 278 sequences were additionally removed as they clustered outside of the cable bacteria monophyletic clade. For the remaining 1,988 cable bacteria sequences (Supplementary Table S2) a sequence length cutoff of 800 bp was applied to ensure sufficient, species-level taxonomic resolution (White et al., 2010). A total of 112 sequences were shorter than 800 bp and were excluded from the phylogenetic analysis (Table 3). The final dataset used for phylogenetic analysis hence included 1,876 sequences in total (Table 3).

The updated phylogenetic tree of cable bacteria contains six distinct genus-level clusters and 90 species-level clades, all within a monophyletic clade forming a sister group to the Desulfobulbus and Desulfogranum genera within the Desulfobulbaceae (Figure 1; Table 3).

Pairwise sequence similarity analysis, using the sequence identity cutoff of 98.7% (Yarza et al., 2014), clustered the 1,876 available 16S rRNA gene sequences in the dataset into 98 separate groups. Among these, 90 groups fulfilled the additional criterion of containing at least one sequence that was sufficiently long (≥1,200 bp), and hence, these groups were designated as genuine species-level clades (85 clusters contained at least one 16S rRNA gene sequence longer than 1,400 bp, while for the other 5, the longest available sequence ranged between 1,200 and 1,400 bp). Note that 38 species-level clades (42%) only contained a single sequence, while the other species-level clades were represented by multiple sequences (up to 1,083 sequences; Supplementary Table S3; Supplementary Figure S2).

The 90 species-level clades were used to construct the 16S based phylogenetic tree (Figure 1), as well as a heatmap of 16S rRNA sequence identity (Figure 2). Combining these two types of information, 6 genus-level clusters were delineated. The genus-level cutoff of 94.5% identity was used as a guide value to define the clusters (Figure 2) but final clusters were mainly determined based on the phylogenetic clustering of species-level clades (Figure 1). Therefore, the genus-level clustering does not always follow the 94.5% identity cut-off that has been traditionally proposed to delineate genera based on 16S rRNA gene sequences (Yarza et al., 2014). Comparing sequence identities among the six genus-level clusters, the minimum sequence identity within one cluster (intra-cluster sequence identity) was 89.7% (AUS1_2 vs. KX172796), while the maximum sequence identity between clusters (inter-cluster sequence identity) was 95.9% (AR4 vs. OBEP010014765). Likewise, the species diversity greatly varied among the 6 clusters, and ranged from only one species-level clade (Cluster III) up to 35 species-level clades (Clusters V and VI) per genus-level cluster (Table 4).

In the following sections we provide an overview of the six different genus-level clusters and the species-level clades they contain. Previously described taxa are referred to by their name, while as yet undescribed species-level clades are given a short code name. Note that recently a new cable bacteria taxonomy was proposed following the rules of the Code of Nomenclature of Prokaryotes Described from Sequence Data (SeqCode), thus removing the Candidatus (Ca.) designation for those cable bacteria for which high quality genomes are available (Hedlund et al., 2022; Plum-Jensen et al., 2024). Yet, because of lacking genome data, not all of the named cable bacteria species could be validated under the SeqCode. Hence for consistency, we will use the original species names and keep the Candidatus (Ca.) designation for all previously named species.

Long 16S rRNA gene sequences (≥800 bp) can be used to assign species-level taxonomies (White et al., 2010; Johnson et al., 2019) and attain a high phylogenetic resolution (Kim et al., 2011) of uncultured microbes like cable bacteria. Previously, full-length 16S rRNA gene sequences of cable bacteria have been extracted from (incomplete) genome assemblies of individual cable bacteria filaments after applying whole genome amplification (Risgaard-Petersen et al., 2015; Trojan et al., 2016; Kjeldsen et al., 2019; Geelhoed et al., 2023). More recent methods are full-length 16S PacBio amplicon sequencing of entire sediment communities (Xu et al., 2021; Sachs et al., 2022) or the extraction of 16S sequences from high-quality (meta)genomes obtained from sediment samples (Thorup et al., 2021; Sereika et al., 2023; Hiralal et al., 2024a; Hiralal et al., 2024b). These approaches however require extensive sequencing efforts. An alternative is the isolation of individual, or tufts of, filamentous bacteria from sediments, and the subsequent cell lysis and (RT-)PCR amplification for either clone libraries or direct sequencing, to obtain phylogenetic information of cable bacteria (Pfeffer et al., 2012; Schauer et al., 2014), as well as from other filamentous bacteria (Beggiatoaceae family) (Salman et al., 2011).

However, until now the direct 16S sequencing method has not yielded full-length 16S sequences (≥1,450 bp) of single cable bacteria filaments. The nested PCR method proposed here offers a number of advantages over previous methods. Firstly, applying nested PCR bypasses the need for clone libraries (Pfeffer et al., 2012; Schauer et al., 2014) or extraction and reamplification of amplicons from gel (Salman et al., 2011). Our approach furthermore provides high confidence and minimizes the risk of chimeric sequences by aligning multiple partial sequences from different regions of the 16S rRNA gene obtained by distinct primer pair combinations into consensus sequences. As such, the method consistently produced long (≥800 bp) to full-length (≥1,450 bp) 16S rRNA gene sequences, six of which represent new species-level clades.

To verify that filaments belonged to the cable bacteria, and also assess the morphological diversity of cable bacteria, hand-picked filaments were also imaged microscopically (Figure 3). Filaments showed long cell chains with characteristic continuous, longitudinal ridges, a distinct morphological feature typical for cable bacteria (Pfeffer et al., 2012; Malkin et al., 2014; Trojan et al., 2016; Cornelissen et al., 2018; Geelhoed et al., 2023). Yet, as for now, it is not possible to establish a direct link between phylogeny and morphology, as the same filament cannot be used for both AFM imaging and DNA sequencing. However, the sample of filaments that was imaged here does reveal a rather large morphological variation within the cable bacteria clade, both within as well as between different genus-level clusters (Figure 3).

Finally, one should be aware that sediments may contain different cable bacterium morphotypes at different densities (Marzocchi et al., 2018). Therefore, hand-picking of single filaments may induce a preferential retrieval of filaments that have a larger diameter and length, as well as strains that are relatively abundant. As a result, less data may be retrieved from filaments that have a low abundance, or which are thin and/or short, thus producing a bias in the dataset compiled. Despite this potential bias, the method proposed here offers a straightforward approach to obtain full-length 16S rRNA gene sequences, which also can be applied to other filamentous bacteria and those that are difficult to cultivate.

The updated phylogenetic tree of the cable bacteria clade reveals a substantially expanded diversity compared to previous assessments (Trojan et al., 2016; Geelhoed et al., 2023; Sereika et al., 2023; Hiralal et al., 2024a; Hiralal et al., 2024b; Plum-Jensen et al., 2024). Based on 16S rRNA gene comparison, we identified six genus-level clusters and 90 potential species-level clades (Figure 1). However, 38 of these species-level clades contained only a single sequence, while 52 clades contained multiple sequences. This pattern suggests that some cable bacterium species are far more widespread in the environment compared to others, so their chance of detection increases. In this view, the “single sequence clades” then represent a pool of rarely encountered cable bacterium species. Note however that some environments (e.g., river sediments, salt marshes) have been more frequently sampled, which may increase the presence of sequences and species-level clades tied to these environments. Moreover, some single sequence clades could potentially result from sequencing errors such as base substitutions, deletions, insertions or chimeras in metagenome-extracted or full-length amplicon 16S sequences (Schirmer et al., 2016; Delahaye and Nicolas, 2021). However, poor quality sequences were removed from the sequence alignment during data curation in the implemented quality control step.

The six genus-level clusters were primarily defined based on the grouping of species-level clades in the phylogenetic tree (Figure 1). Subsequently, we verified the 16S rRNA gene similarity of representative species sequences within and between genus-level clusters (Figure 2). All species-level clades of Clusters I, II, and III showed similarities of <94.0% compared to other clusters, thus respecting the traditional genus-level cutoff of 94.5% (Yarza et al., 2014). However, species within Clusters IV, V, and VI showed cross-cluster sequence identities of >94.5%, with a maximum of 95.9%, to each other (Figure 2), albeit their clearly separated placement in the phylogenetic tree (Figure 1). Additionally, the 16S rRNA gene sequence identity of the species within a respective cluster often dropped below the 94.5% genus-cutoff (Yarza et al., 2014), indicating a significant genetic divergence within cable bacteria. For instance, the minimum sequence identity within a cluster was observed in Cluster VI, where the species AUS1_2 and KX172796 were only 89.7% similar to each other.

The genus-level clustering is largely consistent with previous genus-level classifications originating from phylogenomic data (Geelhoed et al., 2023; Sereika et al., 2023). Three clusters can be directly linked to previously identified clades. Cluster VI corresponds to the Ca. Electrothrix genus, while Clusters IV and V, respectively, correspond to the potential AR4 and AR3 genera, which have been identified based on comparative genome analysis (Geelhoed et al., 2023), but which have not been named and described yet. A conspicuous observation is that the species that have been previously assigned to the Ca. Electronema genus are here divided over the two separate Clusters I and II. The reason for this is that they emerge as distinct clades within the phylogenetic tree (Figure 1). Moreover, the two most similar species between these clusters only show a 16S sequence identity of 92.0% (Ca. Electronema halotolerans vs. PP554486), which is well below the standard 94.5% genus-cutoff (Yarza et al., 2014). However, based on average nucleotide identity (ANI) and percentage of conserved proteins (POCP) the strain Ca. Electronema halotolerans has been placed in the Ca. Electronema genus (Sereika et al., 2023). This example illustrates the well-known limitation of using 16S phylogenetic information for the delineation of clades at the genus level (Hassler et al., 2022).

In addition to the known genus-level taxa, our dataset reveals one new genus-level clade (SI2) represented by Cluster III (Figure 1), which includes five sequences obtained from the Magazzolo river estuary (Sicily, Italy). Comparing the 16S sequences of SI2 with the representative sequences of the 89 other potential cable bacteria species identified here, provides a maximum sequence identity of 94.0% with Ca. Electrothrix (OBEP010014765), 92.3% with AR3, 91.8% with AR4 and 91.0% with Ca. Electronema (FQ658831) (Figure 2). SI2 may hence constitute a novel genus of cable bacteria, hinting at previously unexplored evolutionary diversity within the cable bacteria clade.

In order to delineate different bacteria species, a cutoff value of 98.7% 16S sequence identity is commonly used (Yarza et al., 2014). However, due to its conserved nature, the 16S rRNA gene cannot always accurately delineate closely related species (Větrovský and Baldrian, 2013; Lan et al., 2016). In order to examine if the 16S rRNA gene is suitable to delineate different cable bacteria species, sequence identities of the 13 described species were compared. Overall, the established species-level cutoff of 98.7% could be applied to successfully delineate most previously described cable bacteria species, but with two exceptions. Based on 16S data, Ca. En. aureum and Ca. En. nielsenii should be designated as one species (99.4% sequence identity). Likewise, Ca. E. laxa falls within the species-level range of Ca. E. communis and Ca. E. aarhusiensis (98.9 and 98.8%, respectively) (Figure 2). However, average nucleotide identities (ANI) of <94% of the available genomes revealed that these are indeed separate species based on the 95% ANI threshold (Jain et al., 2018; Sereika et al., 2023).

By definition, 16S rRNA gene sequences contain less information than whole genomes, with phylogenomic trees providing higher taxonomic resolution than 16S phylogenetic trees. This is known to potentially cause differences in the tree topology (Ankenbrand and Keller, 2016; Hug et al., 2016). Still, comparing our 16S rRNA gene tree with previously published phylogenomic trees (Geelhoed et al., 2023; Sereika et al., 2023; Hiralal et al., 2024a; Hiralal et al., 2024b; Plum-Jensen et al., 2024), there appears only one major difference in topology. In the 16S tree, the AR3 genus (Cluster V) is placed between the AR4 genus (Cluster IV) and Ca. Electrothrix (Cluster VI), while in recent genome trees, AR3 is placed between AR4 and Ca. Electronema (Clusters I/II). This different placement may be due to poor bootstrap support of the AR3 and AR4 clades (Figure 1), as also seen in previous 16S rRNA gene trees (Dam et al., 2021; Geelhoed et al., 2023; Plum-Jensen et al., 2024), or may result from the expanded number of sequences adjacent to AR3 in Cluster V. The assignment of bacterial genus-level clades is conventionally based on genomic information—a commonly used threshold for genus classification is 65% amino acid identity (AAI) (Konstantinidis et al., 2017). At present, high quality (unfragmented) genomes are only available from three clusters (Clusters I, II, and VI). Thus additional, high-quality genomes are needed from Clusters III, IV, and V to obtain a more accurate resolution of the taxonomy of the cable bacteria. Providing whole genomes is also an essential requirement to describe and name any of the new species proposed here.

An important question is whether cable bacteria truly comprise a monophyletic clade. Currently, all sequences attributed to cable bacteria belong to one monophyletic clade, apart from one possible exception. Putative cable bacteria detected in aquifers form a genetically distinct sister clade to Desulfurivibrio spp. within the Desulfobacterota that does not fall within the 6-cluster clade in the Desulfobulbaceae family examined here (Müller et al., 2015; Müller et al., 2020; Sachs et al., 2022). These aquifer strains appear to be physiologically distinct from the known Electrothrix/Electronema strains, as a they have a single cell stage and are capable of sulfur disproportionation (Müller et al., 2015; Müller et al., 2020). It should be noted that the approach taken here cannot resolve whether the cable bacteria clade is monophyletic or not, as both the primers used in the nested PCR as well as the database search only targets sequences within the Desulfobulbaceae family. At present, little microscopic, electrochemical and/or spectroscopic data are available for the aquifer strains, and hence it is not clear whether these organisms are truly capable of long-range electron transport and electrogenic sulfide oxidation. This is also the case for many of the potential cable bacteria sequences obtained from public databases for which no metabolic or geochemical data is available. The capacity for long-range electron transport can be assessed in a number of ways: either directly by measuring the current through filaments (Meysman et al., 2019; Bonné et al., 2020; Pankratov et al., 2024) or indirectly, by microscopy or spectroscopy. Detailed microscopy can document the unique topography of their cell surface, which reveals a series of ridges that run in parallel along the full length of the bacterial filament, and which each embed a single fiber (ca. 50 nm diameter) that carries the electron current (Cornelissen et al., 2018; Meysman et al., 2019). Detailed spectroscopy can document the Raman fingerprint of cable bacteria, which originates from the Ni cofactor involved in long-range electron transport, and hence is highly specific and unique to cable bacteria (Boschker et al., 2021; Smets et al., 2024). Hence, in future studies, when describing and naming new bacterial clades, it is highly recommended to explicitly assess these physiological, microscopic and spectroscopic features that are unique to cable bacteria (Hiralal et al., 2024a; Hiralal et al., 2024b).

Cable bacteria belong to the Desulfobulbaceae family, which was recently taxonomically reclassified and currently encompasses the genera Desulfobulbus, Desulfogranum, Desulfolithobacter, Ca. Electrothrix, and Ca. Electronema (Waite et al., 2020; Hashimoto et al., 2022). They were suggested to have evolved from a group of strictly anaerobic, sulfate-reducing bacteria (Kjeldsen et al., 2019), although their exact evolutionary origin, and whether they evolved from a saltwater or freshwater environment, remains uncertain. Recently, molecular clock estimates suggested that cable bacteria may have diverged ca. 0.56 billion years ago, either during or after the Neoproterozoic oxygenation event (Cui et al., 2024).

A marine origin of cable bacteria was suggested previously (Klier et al., 2018; Dam et al., 2021), and in this view, the ability of long-distance electron transport evolved once in the Desulfobulbaceae family before a divergence into a freshwater and a marine lineage took place (Risgaard-Petersen et al., 2015). The recently described species Ca. En. halotolerans has been suggested to be an evolutionary link between the two previously described genera Ca. Electrothrix and Ca. Electronema (Sereika et al., 2023). Additionally, the ca. 1 Mbp smaller genome in Ca. En. aureum compared to Ca. Electrothrix species suggests a loss rather than a gain of osmoregulation genes (Na⁺/H⁺ antiporter NhaA) required for life in saltwater environments (Kjeldsen et al., 2019; Dam et al., 2021; Sereika et al., 2023). Our results here support this hypothesis. The majority of the species (7 of 11 species) of the genera Desulfobulbus, Desulfogranum and Desulfolithobacter as well as 84% of the cable bacteria species-level clades that we identified (76 of 90), originate from brackish/marine environments (Figure 4B). Together these findings indeed suggest a longer evolutionary history of cable bacteria in saltwater environments and a later occupation of freshwater sediments.

We propose that the branching of Clusters I, II, and III could be linked to a freshwater-based adaptation (Figure 1). Cluster I encompasses species exclusively from genuine freshwater environments. This is in concordance with the results of a recent survey, in which no Ca. Electronema 16S sequences (excluding Ca. En. halotolerans) were detected at salinities higher than 0.3 (Dam et al., 2021). Sequences from Cluster I also never co-occur with (salt-tolerant) sequences from other clusters, thus lending support to a specific restriction to freshwater environments and dedicated metabolic adaptations that prevents them from thriving at brackish and marine salinities (Kjeldsen et al., 2019; Dam et al., 2021; Sereika et al., 2023). In contrast, sequences in Clusters II and III have a “brackish” signature, as they are retrieved from sediments in estuaries, which had intermediate salinities (range 2–20) at the time of sampling and during incubations (sites: ME2, ME3, ME20, and YR17, Table 1). Estuaries are dynamic systems, with spatial and temporal fluctuations of environmental parameters such as salinity, temperature, sediment grain size and oxygen availability (Borsuk et al., 2001; Cloern et al., 2017), as well as temporal variation in microbial community composition (Bernhard et al., 2005; Zheng et al., 2014). In addition to the Ca. En. halotolerans, we identified three novel species-level clades (Cluster II: SI1 and YR; Cluster III: SI2) from sites in the Magazzolo and Yarra River estuaries (Table 4). These two sites are known to exhibit strong salinity fluctuations (Roberts et al., 2012) (personal observations). Therefore, one hypothesis could be that Clusters II and III are not so much adapted to steady “brackish” conditions, but have specific adaptions to cope with strong and fast salinity fluctuations.

Clusters IV, V, and VI have a “marine” signature as they all contain sequences retrieved from sediments with full marine conditions (salinity ≥30). However, a high salinity does not seem to be a necessary requirement. For instance, within Cluster VI, species Ca. E. aarhusiensis and communis can survive at a wide range of salinities (0.3–35, and 0.3–21, respectively) (Trojan et al., 2016; Dam et al., 2021). Ca. E. aarhusiensis was also found in a freshwater lake in Denmark (Karst et al., 2018), whereas our data extends the salinity tolerance of Ca. E. communis to 32 (BCZ130). Moreover, Ca. Electrothrix clades SI3 and EB2 (Cluster VI) were detected at lower salinities (S = 2 and 10, respectively) than Ca. En. halotolerans and clades SI1 and YR (Cluster II). The detection of EB2 at salinities 10 and 33 suggests a potential broad salinity tolerance of this clade, from brackish to marine systems. The specific physiological adaptations of the “brackish” Cluster groups II-III versus the “marine” Cluster groups IV-VI require further elucidation. As noted above, the absolute salinity level may not be the only important factor, but also the particular temporal pattern of salinity fluctuations may play a role.

Interestingly, sequences within Cluster V were mainly retrieved from deep sea environments. In fact, 28 of the 35 identified species-level clades in Cluster V originate from sites deeper than 700 m, across geographically distant areas, including the Gulf of Mexico, the Atlantic, the Indian and the Pacific Ocean (Figure 4A; Supplementary Table S5). While deep-sea strains are present in Clusters IV and VI, they are less prominent: one species-level clade in Cluster IV was retrieved from deep sea sediments, and three species-level clades from Cluster VI. This “deep sea” signature of Cluster V is remarkable, as cable bacteria have been predominantly associated with shallow or coastal environments, which typically have shown high sulfate reduction and hence high sulfide production rates. Cluster V species have mostly been retrieved from sites near hydrothermal vents and cold seeps (Grünke et al., 2011; Ruff et al., 2013), which are local hotspots of sulfide production in the deep sea, which hence enables electrogenic sulfur oxidation. In comparison to other deep sea environments, these sites are likely overrepresented. Nonetheless, the cable bacteria from Cluster V represent an enigmatic group that could be specifically adapted to deep sea sulfidic environments. Such a separation of lineages with water depth (deep sea versus coastal environments) was observed for the gammaproteobacterial order Woeseiales, which occurs globally in marine surface sediments (Hoffmann et al., 2020).

Our data indicate that coexistence of multiple cable bacteria species is possible in diverse habitats from freshwater lakes to deep sea sediments. Different morphologies of cable bacteria (e.g., filaments with varying diameters) are frequently observed in the same sediment sample (Scholz et al., 2021; Geelhoed et al., 2023), thus suggesting strain diversity. Direct evidence of coexistence of different Ca. Electrothrix species was first reported for a Baltic Sea site by fluorescence in situ hybridization (Marzocchi et al., 2018). Cable bacteria amplicon sequences (ca. 400 bp) from multiple clusters (IV, VI, VI) were detected in the same sediment near fish farms in Iceland (Vasquez-Cardenas et al., 2022), filaments belonging to multiple clusters (IV, V, VI) and to multiple species within the Ca. Electrothrix cluster (VI) were detected at marine or brackish sites by whole genome sequencing (Geelhoed et al., 2023) and two Ca. Electronema species were conjointly detected in a freshwater sediment in California (Yang et al., 2023).

Our survey here indicates that coexistence of multiple cable bacteria species is very common. Our dataset of 16S rRNA gene sequences includes 74 sampling sites across the world with sequences longer than 800 bp (Table 3; Figure 4). Seven out of these 74 sites harbored sequences from more than one genus-level cluster, whereas half of the sites contained two or more species-level clades. Ten sites showed an even higher diversity with 4 to 12 species species-level clades being present (Figure 4C; Table 5; Supplementary Table S5).

The question as to how several species with apparently the same metabolic requirements coexist thus arises. It was previously proposed that higher electron donor availability, in freshwater sediments, may reduce inter-species competition among cable bacteria species, thus allowing less dominant species to co-occur (Xu et al., 2022). This could explain why organic-rich and high-sulfide sites (RSM30, Aarhus Bay and YR17) harbor multiple cable bacteria species (Trojan et al., 2016; Geelhoed et al., 2023). Conversely, our analysis suggests that a shortage of the electron donor sulfide may also stimulate diversity. Effectively, there were two sites (BCZ130 and ME2) that harbored species-level clades from three different clusters, and these two sites showed low free sulfide concentrations (Table 5; Supplementary Table S5). Cryptic sulfur cycling, i.e., simultaneous reduction of sulfate and oxidation of sulfide, may be taking place in these sediments which can prevent the accumulation of sulfide in the porewater (Van de Velde et al., 2016; Burdorf et al., 2017). Cable bacteria species inhabiting these low-sulfide sediments may have a high affinity for sulfide (Meysman et al., 2015), and hence, sulfide availability could be a factor driving diversity (in addition to salinity). Different substrate affinities between species can lead to micro-niche differentiation, as seen for different coexisting Achromatium species, which show a shift in relative abundances in anoxic slurry incubations compared to oxic ones (Gray et al., 2004). Indeed, we found no overlap between species inhabiting non-sulfidic sediments compared to sulfidic ones, except for Ca. E. communis, which could indicate species-specific adaptations to different levels of sulfide. It should be noted that only few studies explicitly reported H₂S concentrations of investigated sediments, which impedes conclusions about the effect of H₂S on cable bacteria species compositions. Ten sites showed sulfide concentrations of <10 μM. Six of these were brackish sites, incubated in this study (ME2, ME3, ME20, ED33, BCZ130, and BCZ700), which interestingly revealed potential novel or undescribed cable bacteria taxa from clusters II (SI1, SI2), IV (AR4), and VI (SI3) (Table 4).

Overall, the relation between coexistence and niche differentiation remains to be further examined for cable bacteria. Presently, it is unknown which environmental conditions facilitate or impede the coexistence of cable bacteria and how different cable bacteria species might interact or compete with each other. Targeted time series sediment manipulation with treatments such as different electron donor (Xu et al., 2021) and acceptor availabilities may further illuminate the conditions selecting for a dominant species of cable bacteria or the coexistence of multiple species.

Cable bacteria 16S rRNA gene sequences (≥800 bp) were compiled from 92 different sampling sites across the world (Table 3; Figure 4A). More than half of the locations (n = 53) were coastal sites (≤200 m water depth), 23 were deep sea locations (1,000–3,843 m) and 16 were freshwater sites (Figure 4B). These sites include a broad range of temperatures in incubations with subpolar to tropical sediments from 0 to 25°C, indicating a broad temperature tolerance (Burdorf et al., 2017). The absence of long cable bacteria 16S sequences in African, South American and (sub)polar locations in the present dataset is likely due to the lack of studies investigating surface sediments in these geographical regions (Figure 4A).

Coastal sites encompassed a diverse array of habitats, including mud flats and mud banks, (intertidal) salt marshes, seagrass beds, mangroves, bivalve reefs, seasonally hypoxic basins, marine lakes and estuaries (Malkin et al., 2014; Burdorf et al., 2017; Dam et al., 2021; Scholz et al., 2021; Geelhoed et al., 2023). Deep sea habitats included hydrothermal vents, mud volcanoes and cold seeps (Grünke et al., 2011; Ruff et al., 2013; Vigneron et al., 2014) and some cable bacteria sequences even originated from the surfaces of deep sea fauna, such as gastropods and yeti crabs (Goffredi et al., 2004; Thurber et al., 2011). Freshwater sites constitute lakes, streams and aquifers (Trojan et al., 2016; Sachs et al., 2022), and one potential Ca. Electronema (Cluster I) sequence was detected in a prairie soil in Canada (MG394614), thus introducing terrestrial environments as a possible cable bacteria habitat (Figure 4B).

When zooming into the geographic distribution of species-level clades, two patterns emerge: global versus local distributions (Figure 4A; Supplementary Table S4). Ca. E. gigas (Cluster V) shows a global distribution as it was found in intertidal salt marsh sediments in the Netherlands (S ∼ 30), a brackish sediment in Denmark (S ∼ 18–23) and an estuary with high salinity fluctuations (5–30) in Australia (Burdorf et al., 2017; Geelhoed et al., 2023; Sereika et al., 2023). High sulfide-tolerance may be a species-specific characteristic of Ca. E. gigas as all three locations are sulfide-rich coastal sediments (Geelhoed et al., 2023). Likewise, Ca. En. palustre and nielsenii (Cluster I) were found across geographically distant locations, from a stream and an artificial pond in Denmark to rivers in both southern Germany and eastern China (Trojan et al., 2016; Xu et al., 2021; Sachs et al., 2022). Ca. E. communis also appears to be a generalist clade, occupying a broad ecological niche (BCZ130, RSM30, Aarhus Bay, Little Sippewissett salt marsh, MA, United States), coping with a wide range of salinities and free sulfide concentrations (from below detection limit to >1,000 μM), rather than being specifically adapted to environmental conditions of the Baltic Sea as has been suggested previously (Dam et al., 2021).

In contrast, certain species-level clades (such as SI1, SI2, SI3, EB2, and YR) show an “endemic” occurrence, being only detected at a single site (Supplementary Table S4). In total, 63 out of 98 potential species-level clades were found at a single site, including 31 assigned to Cluster V which mostly originate from deep locations of ≥700 m depth (Table 4; Figure 4A). This high number of potential “endemic” species indicates that a link between the phylogeny and the geographical distribution of cable bacteria may exist and that some cable bacteria species may have evolved site-specific adaptations. Further screening of aquatic sediments is however needed to confirm the “endemic” biogeographic character of certain strains.