Enrichment cultures with cable bacteria were prepared from natural sediment collected on 27-09-2019 within a creek bed from the Rattekaai Salt Marsh (Netherlands; 51.4391°N, 4.1697°E). At this site, earlier studies have documented the presence of cable bacteria in situ (Malkin et al., 2014). After collection in the field, the sediment was brought to the laboratory at Utrecht University, where it was sieved (500 μm mesh size) to remove fauna and large debris, homogenized, and subsequently re-packed into polycarbonate cores (height: 12 cm, inner diameter: 5.2 cm). The sediment cores were submerged in artificial seawater (ASW; salinity of 32, the in situ value) and incubated in the dark for several weeks until an active cable bacteria population developed. The overlying seawater was continuously bubbled with air to maintain 100% air saturation, and the temperature (20°C) and salinity were kept constant throughout the incubation. A total of 19 cores were incubated, all prepared from the same batch of sediment.

Cable bacteria activity was monitored within the incubated cores using microsensor depth profiling (O₂, H₂S, and pH). This so-called geochemical fingerprint provides information about the developmental state and metabolic activity of the cable bacteria population (Nielsen et al., 2010; Risgaard-Petersen et al., 2012; Malkin et al., 2014). The microsensor depth profiles were also used to delineate the oxic, transition, and suboxic zones in the sediment at the time of core sectioning (see Section “Filament Retrieval from the Sediment”).

Microsensors (tip diameters; O₂: 50 μm, H₂S: 100 μm, pH: 200 μm) were purchased from Unisense A/S (Denmark), connected to a four-channel Microsensor Multimeter (Unisense), and mounted in a two-dimensional micro-profiling system that enabled stepwise movement of sensors. The SensorTrace PRO software (Unisense) was used to control the vertical movement of the microsensors and record sensor signals. A general-purpose reference electrode (REF201 Red Rod electrode; Radiometer Analytical, Denmark) was used during pH measurements. Calibration of the microsensors was performed as previously described (Malkin et al., 2014).

Cable bacteria population developed in all but one of the incubated sediment cores. Five cores with the largest ΔpH were selected for the SIP experiment. The quantity ΔpH, defined as the difference between the maximum and minimum pH in the oxic and suboxic zone, respectively, provides a good proxy for comparing cable bacteria activity among different populations (Burdorf et al., 2018).

For the SIP experiment, 10 mL of stock solution was prepared by mixing 2 mL of H₂¹⁸O (Sigma-Aldrich; ¹⁸O atom fraction of 0.99) and 8 mL of artificial seawater (ASW) with a natural abundance of ¹⁸O (¹⁸O atom fraction of 0.002). Hence, the ¹⁸O atom fraction of water in the stock solution was 0.2. The ASW contained no Mg²⁺ and Ca²⁺ ions (to avoid precipitation of Mg¹³CO₃ and Ca¹³CO₃) and no bicarbonate ions (to avoid ¹³C label dilution). The stock solution was additionally labeled in ¹³C by adding ¹³C-bicarbonate (NaH¹³CO₃, Sigma-Aldrich; ¹³C atom fraction of 0.99) to a final concentration of 62 mM. This concentration and labeling were chosen because they were successfully applied in previous SIP experiments (Vasquez-Cardenas et al., 2015; Geerlings et al., 2020, 2021).

Dual labeling (H₂¹⁸O and H¹³CO₃^–) of the sediment cores was done by first inserting a sub-core (inner diameter 1.2 cm, length 5 cm) into each culturing core without disturbing the sediment, and then injecting 500 μL of the labeled stock solution into the sub-core in ten separate and parallel 50 μL injections. To ensure homogeneous spread of the label throughout the sediment, the syringe needle was first inserted to a depth of 5 cm, and then the 50 μL dose of liquid was released while slowly retracting the needle upward. The use of the sub-core ensured that the label was spread within a well-constrained volume. Subsequently, the cores were incubated for 6 h (two replicate cores) and 24 h (two replicate cores), with one additional core chosen as an unlabeled control. Temperature was kept constant at 20°C during all incubations. At the end of the SIP incubation period, the sub-core was carefully pulled out of the sediment core and sectioned based on the redox zonation (see next section). Cable bacteria were retrieved from each sediment section and imaged by scanning electron microscopy (SEM) and NanoSIMS.

During the SIP incubation, each sub-core was overlain with a thin layer of water (∼2 mm) with the same ¹⁸O and ¹³C labeling as the porewater. Additionally, the cores were placed in a sealed container filled with air and the bottom covered with a thin layer of ASW with the natural abundance of ¹⁸O and a similar ¹³C labeling as that of the porewater. This setup ensured similar ¹³C labeling of the porewater and the CO₂ pool in the surrounding atmosphere inside the sealed contained and thus negligible ¹³C label loss from the porewater due to air-water gas exchange. Because the system was stagnant, the loss of ¹⁸O label from the porewater due exchange with the thin layer of ASW at the bottom of the sealed container was also negligible.

Based on the volume of the sub-core (5.65 mL) and the porosity of sediments from the Rattekaai salt marsh (0.75; L. Burdorf, 2017 thesis, p. 141), the porewater volume in the sub-core was estimated at 4.24 mL. Since 500 μL of the porewater was replaced by the stock solution, the estimated final ¹⁸O atom fraction of the porewater was 0.025. This is about two-fold greater than the ¹⁸O-labeling of water used in the SIP experiment by Langer et al. (2018) (S. Langer, personal communication). We assumed that all sub-cores in the replicate cores had the same ¹⁸O-labeling, as the sub-core dimensions and the volume of the injected stock solution were identical.

Cable bacterium filaments were retrieved from the sediment matrix under a stereo microscope using fine glass hooks custom-made from Pasteur pipettes. Filaments were retrieved separately from the oxic zone (0-2 mm depth), middle of the suboxic zone (5-10 mm depth), and the oxic-suboxic transition zone, the latter defined here as the zone up to 1 mm below the oxic zone (Supplementary Figure 1). The depth range of these zones were derived from microsensor measurements conducted just before core sectioning. Previous studies showed that in the transition zone, the density (Seitaj et al., 2015) and motility (Bjerg et al., 2016) of cable bacteria are highest. Retrieved filaments were washed several times (> 3) in Milli-Q water (Millipore, Netherlands) to eliminate precipitation of salts, transferred onto polycarbonate filters (pore size 0.2 μm; Isopore, Millipore, Netherlands) that were pre-coated with a 5-10 nm thin gold layer, and air-dried in a desiccator for about 24 h.

Filaments on the polycarbonate filters were imaged with a scanning electron microscope (JEOL Neoscope II JCM-6000, Japan) to identify filament segments suitable for NanoSIMS analysis. Imaging was done under a 0.1-0.3 mbar vacuum and a high accelerating voltage (15 kV) using a backscatter electron detector.

NanoSIMS analysis was performed with the nanoSIMS 50L instrument (Cameca, France) operated at Utrecht University. Fields of view (FOV) selected through SEM imaging were pre-sputtered with Cs⁺ ions until secondary ion yields stabilized. Subsequently, the primary Cs⁺ ion beam (current: 0.5-10 pA, energy: 16 keV, beam size: 130 nm, dwell time: 1-2 ms per pixel) was scanned over the FOV (areas between 10 × 10 μm and 20 × 20 μm in size) while detecting secondary ions ¹²C¹⁴N^–, ¹³C¹⁴N^–, ³¹P^–, ¹⁶O^–, ¹⁸O^–, and ³²S^–.

Initial measurements employed a relatively short pre-sputtering interval (10 min) and a low primary ion current (0.5-2 pA), which resulted in relatively low ¹⁸O^– and ³¹P^– ion yields during the subsequent analysis conducted with the same current. Thus, during the analysis, the primary ion current was increased to 10 pA to enable quantification of ¹⁸O labeling in poly-P granules with a desirable precision and within a reasonable time interval (few hours). This increase in the primary ion current resulted, however, in excessive count rates on the electron multiplier used for the detection of ¹²C¹⁴N^– (> 10⁵ cps). To prevent detector aging due to such high count rates, ¹²C¹⁴N^– ions were therefore not detected during these measurements. As a down-side, ¹³C labeling could not be determined for these initial measurements.

Later in the analysis, we noted that the ¹⁸O^– and ³¹P^– secondary ion yields were low during the initial measurements because the probed volume was too close to the cell surface, whereas the poly-P granules were present deeper within the cell biovolume. Therefore, we changed the measurement protocol by including much longer pre-sputtering intervals (20-30 min), which allowed us to probe the more inner parts of the cells during the subsequent analysis using low primary ion currents (0.5-2 pA). Specifically, it allowed us to detect all target secondary ions and thus simultaneously determine both ¹³C and ¹⁸O labeling of the biomass of cable bacteria, including poly-P granules.

NanoSIMS analysis of most samples focused on the variation of the mean isotopic and elemental composition among cells within filaments. In these analyses, the same FOV was imaged multiple times (100-300 frames) and the resulting ion count images were aligned and accumulated. For selected samples, we aimed to obtain additional insight into the 3D distribution of the isotopic and elemental composition within cells. These measurements were therefore conducted with a substantially larger number of frames (up to 7,000) until the sample material was completely sputtered by the primary ion beam.

NanoSIMS data were processed using the Matlab-based software Look@NanoSIMS (Polerecky et al., 2012). After alignment and accumulation of the measured planes, regions of interest (ROIs) corresponding to the poly-P granules were drawn manually using the combined ¹²C¹⁴N^– (or ¹³C¹⁴N^–), ¹⁸O^– and ³¹P^– ion count images. ROIs were not drawn for cells that appeared damaged. For each ROI, the ROI-specific ¹⁸O atom fraction was calculated as x(¹⁸O) = ¹⁸O^–/(¹⁶O^– + ¹⁸O^–) using the total counts of ¹⁸O^– and ¹⁶O^– accumulated over the ROI pixels. Similarly, the ROI-specific ¹³C atom fraction was calculated as x(¹³C) = ¹³C¹⁴N^–/(¹²C¹⁴N^– + ¹³C¹⁴N^–) from the total counts of ¹²C¹⁴N^– and ¹³C¹⁴N^– (only if the ¹²C¹⁴N^– ions were detected). Note that when detecting secondary ions from a given poly-P granule, the probed volume partly also included the cytoplasm of a cell in which the poly-P granule was embedded (e.g., “above” or “below” the granule; see Results, Figure 1). Thus, the ¹³C atom fraction determined in the ROI drawn around a poly-P granule represents the ¹³C atom fraction in the surrounding cytoplasm.

To quantify ¹³C and ¹⁸O labeling, we used excess atom fractions calculated according to x^E(¹³C) = x(¹³C) – x_i(¹³C) and x^E(¹⁸O) = x(¹⁸O) – x_i(¹⁸O), respectively, where x_i(¹³C) and x_i(¹⁸O) correspond to the atom fractions determined for control cells. Based on these quantities, we classified filaments as ‘inactive’ when x^E(¹³C) < 0.0006 and x^E(¹⁸O) < 0.00025 (the threshold levels correspond to 3 standard deviations of the respective excess atom fractions determined for control filaments), ‘minimally active’ when x^E(¹³C) ≤ 0.001 (reflecting the lowest 25% of measured filament fragments and 2% of the maximum measured excess ¹³C atom fraction) and x^E(¹⁸O) ≤ 0.0004 (which reflects the lowest 10% of measured filament fragments), and ‘active’ when x^E(¹³C) > 0.001 or x^E(¹⁸O) > 0.0004.

To gain insight into the 3D distribution of the isotopic and elemental composition and the position of the poly-P granules within cells, Look@NanoSIMS was additionally used to visualize the depth variation in the nanoSIMS data along a lateral or transversal profile. This analysis was done as previously described (Geerlings et al., 2021).

Overall, data was obtained for 1887 poly-P granules in 884 cells from 203 filament segments (Table 1). Out of these, 126 poly-P granules in 29 cells from 7 filament segments belong to the control samples. Both the ¹⁸O and ¹³C atom fractions are available for 704 cells (from 164 filament segments), while data available for the remaining 151 cells (from 32 filament segments) only includes the ¹⁸O fraction.

Statistical analysis of this dataset focused on the following aspects: (i) the effect of the redox environment and labeling period on the ¹³C labeling of the cytoplasm and the ¹⁸O labeling of the poly-P granules, (ii) variation in the ¹⁸O labeling of the poly-P granules within a cell, within a filament, and among filaments, and (iii) the relationship between the average ¹⁸O labeling of the poly-P granules and the corresponding ¹³C labeling of the surrounding cytoplasm within each filament fragment.

The first two aspects were assessed by fitting the data with a linear mixed model (separately for the ¹³C and ¹⁸O data), which was done in R using the package nlme (Pinheiro et al., 2021). The analysis considered the hierarchical structure of the data (i.e., nesting of cells within filaments for the ¹³C data, and nesting of poly-P granules within cells, which are further nested within filaments, for the ¹⁸O data) and the unbalanced experimental design (Table 2). The ‘redox zone’ (oxic, transition, suboxic) and the ‘labeling period’ (6 h, 24 h) were defined as fixed effects, whereas ‘filament’ (when testing the excess ¹³C labeling of cells) and ‘cell’ and ‘filament’ (when testing the ¹⁸O labeling of poly-P granules) were chosen as random effects in the model. For model selection, a step-up approach was used, which starts with a reference model that contains all fixed components and their interactions. This so-called “beyond optimal model” was then used to find the best variance structure and random structure (Zuur et al., 2009). Using the output of the linear mixed model for the poly-P-specific ¹⁸O data, we calculated the percentage of variance in the data explained by differences among filaments, among cells within the same filament, and differences within the same cell. Similarly, we calculated the percentage of variance explained by differences among filaments and among cells within the same filament for the cell-specific ¹³C data. The third aspect was assessed by calculating the Kendall rank correlation coefficient (τ), which is a robust parameter used for testing correlations in non-normally distributed data (Dalgaard, 2013). Details of the statistical analysis are provided in the Supplementary Methods.

Patterns shown by the ¹³C data in the present study are similar to those observed previously (Geerlings et al., 2020, 2021). ¹³C labeling was mainly restricted to filaments from the transition and suboxic zone, whereas filaments from the oxic zone displayed no, or only very low, ¹³C labeling (SI Appendix, Supplementary Figure 2). Furthermore, the ¹³C labeling was highly variable among filaments from the transition and suboxic zone, but highly similar when compared among cells within the same filament (Figure 1 and Supplementary Figure 2).

In contrast to the low intra-filament variability in ¹³C labeling, there was a clear variation in the ¹⁸O labeling of poly-P granules among cells from the same filament (Figure 1). However, adjacent cells displayed a similar pattern when compared with respect to the size, position, and number of granules per cell, which is congruent with earlier observations on poly-P granules in cable bacteria (Geerlings et al., 2019).

Depending on the diameter of the filament and the intracellular spatial organization of the poly-P granules, two different morphotypes of cable bacteria were distinguished: (i) “thin” filaments (diameter ∼0.5-1 μm), which mostly contained two similarly-sized (diameter ∼100-300 nm) poly-P granules per cell, one at each cell pole (Figure 2A), and (ii) “thicker” filaments (diameter above 1 μm), which contained variable spatial organizations of the poly-P granules within cells. Some cells in the thicker filaments contained many small poly-P granules, while others contained only a few larger ones (Figure 2B). It is unknown whether the different morphotypes represent different species of cable bacteria or they are different manifestations of the same species in a different stage of the life cycle.

Both the ¹³C labeling of the cytoplasm and the ¹⁸O labeling of the poly-P granules were highly variable when compared among different filaments (Figures 1, 3, 4, Table 3 and Supplementary Figure 2). Based on the combined ¹³C and ¹⁸O data, we divided the filaments into four classes (see Methods for details). Class 1 includes filaments considered as metabolically inactive or minimally active during the labeling period. This class contained 4.3% (7/164) of filaments (1 inactive and 6 minimally active), all originating from sediment cores incubated for 6 h. Note that this number does not include filaments without detectable poly-P granules at the end of the nanoSIMS analysis. Class 2 includes filaments that showed ¹³C labeling of the cytoplasm but no, or only minimal, ¹⁸O labeling of the poly-P granules. This pattern was observed in 3.7% (6/164) of filament fragments, all from the 6 h incubation. Class 3 includes filaments that showed no, or only minimal, ¹³C labeling but contained poly-P granules with a significant ¹⁸O labeling (Figure 1E). This pattern, observed in 23% (38/164) of filament fragments, indicates that the filaments did not grow during the SIP incubation but were still active with respect to poly-P metabolism (see Discussion). The highest proportion of filaments (113/164, i.e., 69%) was assigned to class 4, which includes filaments with high labeling in both ¹³C and ¹⁸O. For these filaments, ¹⁸O labeling of the poly-P granules was significantly greater than that of the cytoplasm (Figure 1F). This labeling pattern indicates that these filaments were simultaneously active with respect to growth and poly-P metabolism. Overall, around 78% of filaments from the transition and suboxic zone were actively growing during the 6 h or 24 h incubation period, and around 92% of all measured filaments (from all redox zones) displayed poly-P activity (Table 3).

¹⁸O labeling of poly-P granules was observed in filaments from all redox zones and varied greatly among filaments (Figures 3–5). Despite this high variability, statistical analysis revealed that the labeling values in filaments from the oxic zone were, on average, significantly lower than those from the transition and suboxic zones. In contrast, the transition and suboxic zones showed no significant differences (Figures 4D,E and Supplementary Table 2). This pattern was detected for both labeling periods (6 h and 24 h), although it was more pronounced for the longer period. Furthermore, a comparison made separately for each redox zone revealed a significantly higher ¹⁸O labeling of poly-P granules in filaments incubated for 24 h compared to 6 h (Figures 4D,E and Supplementary Table 1).

Statistical analysis further revealed that most of the variability in the ¹⁸O labeling of poly-P granules (about 52-88%) was explained by differences among filaments, with the percentage being highest in the oxic zone (consistent for each labeling period) and higher for the shorter labeling period compared to the longer one (consistent for each redox zone) (Table 4). The percentage of variance explained by differences among cells within filaments was relatively minor (about 5-8%) and similar for both labeling periods and all redox zones. Finally, a sizeable fraction of the variance (about 4-44%) was unexplained, i.e., was due to unknown differences within cells, with the percentage being lowest in the oxic zone (consistently for each labeling period) and higher for the longer labeling period compared to the shorter one (consistently for each redox zone) (Table 4). For the ¹³C labeling of cell cytoplasm, most of the variability among cells (86-99%) was explained by differences among filaments and only minor differences (1-14%) were observed among cells within the same filament (Supplementary Table 3). Thus, similar to previously observed patterns (Geerlings et al., 2020, 2021), there was only a minimal variability in ¹³C labeling among cells within a given filament from a given redox zone. Additionally, variability among cells within filaments was lower for the ¹³C labeling of the cell cytoplasm compared to the ¹⁸O labeling of poly-P granules.

The filament-averaged ¹⁸O atom fractions of poly-P granules and the corresponding averaged ¹³C atom fractions of the cytoplasm were determined for the same filament fragment for each of the labeling periods. Correlation analysis showed that these quantities were significantly correlated for each of the labeling periods (p = 0.0005, Figure 5). However, this correlation was largely a collinearity effect caused by the dependence of both variables on the redox environment. In all cases, however, the predictive power of the correlation was poor, as indicated by the low value of the Kendall rank correlation coefficient (τ = 0.208 and 0.405 for the 6 h and 24 h labeling period, respectively).

Before we interpret our data, we briefly review possible pathways through which ¹⁸O derived from H₂¹⁸O can enter poly-P granules.

The first step involves ¹⁸O exchange between water and the inorganic (ortho)phosphate (P_i) pool, which occurs when inorganic pyrophosphate (PP_i) is hydrolyzed. This process can occur both inside and outside of a cell (Figure 6, reactions 1 and 2). The PP_i required in this step is produced in a metabolically active cell by many different processes, such as hydrolysis of ATP into AMP or breakdown of large biomolecules. At temperatures below 80°C, abiotic phosphoryl-transfer reactions such as PP_i hydrolysis are very slow but are significantly accelerated by enzymes (Lassila et al., 2011). Therefore, in natural systems, the ¹⁸O signature of the P_i pool (δ¹⁸O_p) is dominated by enzyme-mediated ¹⁸O exchange with water (Blake et al., 2001, 2005). When this process occurs intracellularly, H₂¹⁸O first enters the cell via diffusion and the ¹⁸O exchange is then catalyzed by the enzyme pyrophosphatase (PPase). In contrast, the ¹⁸O exchange outside of a cell occurs, e.g., due to the activity of microorganisms that use extracellular phosphatases (e.g., phosphomonoesterases) as a strategy to acquire P_i from organic compounds (Liang and Blake, 2006; Von Sperber et al., 2014), and the ¹⁸O-labeled P_i then enters the cell via diffusion. We assume that in both cases the intracellular ¹⁸O isotope equilibrium between H₂O and P_i is established within a few hours (Blake et al., 2005; Chang and Blake, 2015).

Once the intracellular P_i pool is labeled with ¹⁸O, the label can enter the poly-P granule via poly-P synthesis. This process is catalyzed by two families of poly-P kinases, PPK1 and PPK2 (Rao et al., 2009), and involves incorporation of P_i into a nucleoside triphosphate (NTP, such as ATP or GTP) followed by a transfer of P_i from NTP to a poly-P chain (Figure 6, reactions 3 and 4). As these reactions consume NTP, poly-P synthesis hence requires the investment of metabolic energy. Cable bacteria carry the genes necessary to produce enzymes from both of these families (Kjeldsen et al., 2019).

Poly-P breakdown was suggested as another mechanism how ¹⁸O could enter the poly-P granule (Langer et al., 2018). Poly-P breakdown can proceed via hydrolysis, which is catalyzed by an exopolyphosphatase (PPX) and leads to the cleavage of the terminal P_i from the poly-P chain (Kornberg et al., 1999). However, evidence suggests that this pathway proceeds via a concerted mechanism with a single transition state where the nucleophilic water attacks the P-atom at the end of the poly-P chain. As a result, the O-atom from the water molecule eventually ends up in the P_i residue leaving the chain, and so there is no labeling of the poly-P (Lassila et al., 2011). Another possible pathway of poly-P breakdown involves phosphorylation of nucleoside monophosphates (AMP or GMP) or diphosphates (ADP and GDP) (Ishige et al., 2002; Nocek et al., 2008, 2018). This process is aided by PPK2-type enzymes, and most likely also proceeds via a concerted mechanisms with a single “loose” transition state (Nocek et al., 2018) where the O-atom from the water molecule ends up in the NDP or NTP molecule rather than in the terminal P_i group on the remaining poly-P chain. Thus, poly-P breakdown in the presence of H₂¹⁸O is unlikely to increase significantly the ¹⁸O atom fraction of an unlabeled poly-P granule (a minor effect, expected to occur due to kinetic isotope fractionation, is neglected in this study).

Poly-P breakdown could, however, decrease the ¹⁸O atom fraction of a previously ¹⁸O-labeled poly-P granule, if the ¹⁸O-labeled P_i groups added to the granule during poly-P synthesis were removed preferentially during the subsequent phase of poly-P breakdown. This could occur, for instance, if the newly synthesized poly-P was incorporated into a specific location within a granule, such as the outer surface, and the breakdown involved poly-P from the same location. Our detailed 3D nanoSIMS analysis revealed, however, that there is no significant variability in ¹⁸O labeling within a poly-P granule (Figure 1), indicating that, on the scale of hours, the poly-P pool within a granule becomes well mixed. This mixing is likely the result of structural organization of a granule where the negatively charged poly-P chains are connected to one another via ion bridges involving divalent cations which causes the tertiary structure to be rearranged and the ¹⁸O label to become well-mixed. This hypothesis is supported by the recently documented association between Ca²⁺ and Mg²⁺ ions and poly-P granules in cable bacteria (Geerlings et al., 2019) and a modeling study on the conformation and organization of poly-P molecules and Ca²⁺ ions (Müller et al., 2019). Thus, due to the well-mixed nature of the poly-P granule, poly-P breakdown has no significant effect on the ¹⁸O atom fraction of an ¹⁸O-labeled poly-P granule.

Based on these arguments, we conclude that in the presence of H₂¹⁸O and an ¹⁸O-labeled P_i pool, the ¹⁸O atom fraction of a poly-P granule can only increase (during poly-P synthesis) or stay constant (during poly-P breakdown). Thus, a significant enrichment of poly-P granules in ¹⁸O indicates that poly-P synthesis occurred at some point during the SIP incubation with H₂¹⁸O, but it may be insufficient to provide insights into the actual rate of poly-P synthesis or the cycling between poly-P synthesis and breakdown.

Our poly-P-specific ¹⁸O data indicate that on the time scale of 24 h, cable bacteria cells turn over a significant amount of poly-P. This conclusion derives from the comparison of ¹⁸O data between the 6 h and 24 h labeling intervals. Specifically, if poly-P synthesis and incorporation into existing poly-P granules at a constant rate were the only processes during the incubation, the expected excess ¹⁸O atom fraction of poly-P granules after 24 h of incubation would be, on average, around 5 to 7-fold greater than after 6 h of incubation (Supplementary Model). This estimate considers that, for the incubation temperature of 22°C, it takes about 27 h for the ¹⁸O exchange between the P_i pool and H₂¹⁸O to reach an equilibrium (Blake et al., 2005). However, the excess ¹⁸O atom fractions of poly-P granules only increased, on average, by a factor of 1.8-2.7 between the 6 h and 24 h incubation periods (Figure 4 and Supplementary Table 1). When combined with the ¹⁸O labeling asymmetry between poly-P synthesis and breakdown proposed above, this discrepancy suggests that poly-P breakdown occurred during a significant fraction of the 24 h labeling interval. Present data is, however, insufficient to constrain the rates of poly-P synthesis and breakdown or the number of turnover cycles. Research on shorter time scales (∼1 h) is required to gain more insight into the dynamics of poly-P turnover in cable bacteria. Since SIP using ¹⁸O-labeled water requires an exchange between the P_i pool and H₂¹⁸O to produce ¹⁸O-labeled P_i, research on shorter time scales may be hampered and it is therefore advised to directly add labeled ¹⁸O-labeled phosphate. This can be prepared either using a minimal abiotic system with commercially available PPase (Blake et al., 2005) or by the hydrolysis of PCl₅ using H₂¹⁸O (Versaw and Metzenberg, 1996).

Poly-P metabolism in cable bacteria is governed by the redox environment, with higher levels of synthesis in the suboxic (and transition) zone and lower levels of synthesis, or higher levels of breakdown, in the oxic zone. This conclusion is supported by the statistical analysis of the poly-P-specific ¹⁸O labeling, which revealed, consistently for both labeling intervals (6 h and 24 h), a significantly lower mean value in the oxic zone compared to the transition and suboxic zones and no significant differences between the transition and suboxic zones (Figures 4C-D and Supplementary Table 2).

The dependence of poly-P metabolism on the redox environment, as observed in the present study for cable bacteria, seems to follow an opposite pattern to that observed for other poly-P accumulating microorganisms. For most of these organisms, phosphate is stored as poly-P under aerobic conditions, whereas poly-P is broken down under anaerobic conditions (He and Mcmahon, 2011; Saia et al., 2021). This pattern was observed, e.g., for the giant sulfur-oxidizing bacteria Thiomargarita namibiensis (Schulz and Schulz, 2005), for the large filamentous sulfide oxidizing bacteria Beggiatoa spp. (Brock and Schulz-Vogt, 2011), or when investigating the phosphorus cycling in wetland sediments from South Gippsland (Southern Australia) (Khoshmanesh et al., 1999, 2002).

This apparent inconsistency between cable bacteria and other poly-P accumulating organisms can be explained by considering the peculiar energy metabolism of cable bacteria facilitated by long-distance electron transport along the filament. Although poly-P appears to be synthesized predominantly by cells residing in the anoxic sediment, these cells can only gain energy if part of the filament is connected to oxygen (Geerlings et al., 2020). Thus, the energy metabolism of the poly-P synthesizing cable bacteria cells is de facto aerobic.

Another theoretical possibility could be anoxic poly-P synthesis. Based on genomic data, cable bacteria have a metabolic potential for sulfur disproportionation (Kjeldsen et al., 2019), which could be a mechanism for gaining energy required for poly-P synthesis when a filament is disconnected from oxygen. However, the function of cells in their native habitat often cannot be reliably predicted from genomic data (Hatzenpichler et al., 2020). More direct methods need to be employed to test whether poly-P synthesis in cable bacteria is coupled to anaerobic energy metabolism such as sulfur disproportionation.

Our measurements revealed significant ¹⁸O labeling of poly-P granules in filaments retrieved from the oxic zone (Figures 3, 4), indicating that these cells must have synthesized poly-P at some point during the SIP incubation. Because this process requires energy, preferentially in the form of ATP (Kornberg et al., 1999; Nocek et al., 2008; Rao et al., 2009; Wang et al., 2018), the cells from the oxic zone therefore must have been able to generate energy at some time during the incubation. However, these cells do not possess any known terminal oxidases (Kjeldsen et al., 2019) and thus cannot generate energy from oxygen reduction, as supported by the observation that cells in the oxic zone do not grow (Geerlings et al., 2020, 2021; this study, Figure 5).

A possible explanation for the observed ¹⁸O labeling of poly-P in cells from the oxic zone is that these cells generated energy via endogenous catabolism. This process is a stress response often observed in microorganisms entering a non-growing stage, where energy is generated from the breakdown of large biomolecules (e.g., DNA, RNA, ribosomes, phospholipids) (Bergkessel et al., 2016). For example, in stressed E. coli cells, this stringent response redirects cellular phosphorus toward synthesis of poly-P, which is essential for adaptation to various stresses (including oxidative stress) and survival during periods of no growth (Rao and Kornberg, 1996; Ault-Riché et al., 1998; Rao et al., 1998). For cable bacteria, endogenous catabolism could theoretically be directed toward poly-P synthesis. The ¹⁸O labeling of poly-P in the cells from the oxic zone could then be explained by assuming that the energy released by the endogenous catabolism is directed toward the synthesis of ATP, which is then used to “propagate” the ¹⁸O-labeled P_i toward poly-P (Figure 6; reactions 5 & 6). Yet why direct energy from endogenous catabolism to poly-P rather than using the energy directly for maintenance and stress release?

An alternative, and perhaps more likely, explanation for the observed ¹⁸O labeling pattern of poly-P is that the cells retrieved from the oxic zone spent part of the SIP incubation in the suboxic zone, where they synthesized poly-P and thus increased the ¹⁸O labeling of the poly-P granules, and then migrated to the oxic zone before the end of the incubation. This hypothesis is consistent with the gliding motility of cable bacteria (Bjerg et al., 2016), which is most prominent at the oxic-suboxic boundary, where cells move transiently in and out of the oxic zone such that cell abundance in the oxic zone remains relatively constant (Scilipoti et al., 2021; Yin et al., 2021). It is also consistent with their ability to quickly switch between sulfide oxidation and oxygen reduction depending on the surrounding redox conditions (Geerlings et al., 2020). This hypothesis suggests that the average ¹⁸O labeling of poly-P granules in cells from the oxic, transition, and suboxic zone reflects the average residence time of the cells in each zone.

For each combination of the labeling interval and redox zone, ¹⁸O atom fractions of poly-P granules showed considerable variability among granules (Figures 4A,B). Most of this variability is explained by differences among filaments in combination with differences among granules within the same cell rather than by differences among cells within the same filament. Here, we discuss possible reasons for these patterns.

Although both ¹³C labeling of the cell cytoplasm and ¹⁸O labeling of poly-P granules are, on average, lower in the oxic zone compared to the transition and suboxic zones, the ¹³C labeling does not predict ¹⁸O labeling for individual cells (Figure 5). Thus, cell growth and poly-P synthesis are linked on the population level due to their common dependence on the redox environment. We cannot exclude that a connection between active growth and poly-P synthesis also exists on the filament and cell level (e.g., in the suboxic zone). However, we observed a sizeable fraction (27/129 or ∼21%) of filaments from the transition and suboxic zones that showed no, or very little, ¹³C labeling of the cytoplasm but clear ¹⁸O labeling of the poly-P granules (Figure 1E, Table 3 and Supplementary Figure 2), which shows that, on the filament and cell level, growth and poly-P synthesis in cable bacteria can operate independently of each other.

A similar large heterogeneity in growth among filaments from the suboxic zone was observed previously in SIP experiments using DIC labeled with ¹³C (Geerlings et al., 2020, 2021) or ¹⁴C (Kjeldsen et al., 2019), which also found no significant growth in about 30-50% of filaments from the suboxic zone after 24 h of incubation. In previous studies, filaments displaying no significant growth were considered as inactive, and it was hypothesized that these filaments had no contact with oxygen during the incubation interval and thus could not generate ATP.

Our present results refine this interpretation, because most of the filaments that showed no activity in terms of growth were still actively synthesizing poly-P. Because poly-P synthesis requires energy, most likely in the form of ATP, these filaments must have had the capacity for energy generation, and were therefore most likely connected to oxygen, at least during part of the labeling interval. However, instead of directing this energy toward growth and reproduction, these filaments remained in a state of growth arrest and directed the energy, or at least part of it, toward poly-P synthesis (Figure 7). This way, cable bacteria appear to have two phases in their life cycle when connected to oxygen (explaining the two clusters in Figure 5B): period of growths, in which the suboxic cells divert the energy gained from sulfide oxidation toward biomass synthesis and poly-P production, and periods of growth arrest, where the suboxic cells divert the energy gained from sulfide oxidation exclusively toward poly-P production. In both cases, only the cells that perform sulfide oxidation are capable of ATP formation (Figure 7). The synchronized growth and division observed in previous research indicates that cells regularly lose their connection to oxygen as part of their life cycle (Geerlings et al., 2021). When disconnected from oxygen, both poly-P synthesis and growth cannot happen via long-distance electron transport. Thus, the variability in the ¹⁸O labeling of the poly-P granules between filaments can be explained by access to oxygen and differences in the cell-cycle stage (growth vs. non-growth). The cells that perform oxygen reduction do not show energy conservation, and as a result show vanishing rates of biomass synthesis (¹³C labeling) and poly -P formation (¹⁸O labeling) (Figure 7).

For most poly-P accumulating organisms, poly-P is typically thought of as a “back-up” energy reservoir, but in many species it can also be essential for survival of the non-growing stage of the cell cycle or in response to oxidative stress (Rao and Kornberg, 1996; Ault-Riché et al., 1998; Kornberg et al., 1999; Rao et al., 2009; Gray et al., 2014; Gray and Jakob, 2015). We hypothesize that poly-P also plays these roles in cable bacteria, where the specific role depends on both the redox environment and cell cycle stage.

Cells in the oxic zone need to perform oxygen reduction to facilitate energy supply for the rest of the filament in the suboxic zone with no known method of energy conservation (Geerlings et al., 2020). Being in an oxic environment, these cells are likely in a constant state of oxidative stress, which is supported by the high concentrations of antioxidant proteins (e.g., catalases, superoxide reductase, rubrerythrin, and GroEL/ES chaperonins) detected by proteomic analysis of cable bacteria (Kjeldsen et al., 2019). Thus, a viable hypothesis is that cable bacteria cells use poly-P as a “back-up” energy reservoir and/or as a method of protection against oxidative stress when they (temporarily) reside in the oxic zone. Poly-P can prevent oxidative stress by acting as an ATP-independent primordial chaperone with minimal substrate specificity to stabilize a wide variety of proteins (Gray et al., 2014; Gray and Jakob, 2015). Thus, the use of poly-P as a chaperone would be ideal for cells in the oxic zone as they cannot build extra antioxidant proteins due to their inability to generate energy. Furthermore, no phosphate groups are removed from poly-P when it functions as a chaperone, so once the poly-P chain is separated from the accompanying protein, it can still be used as an energy reserve to support other tasks (e.g., to aid motility).

Microbial cells spend most of the time in a prolonged non-growing stage, with very little to no metabolic activity and growth occurring due to a limited availability of energy or nutrients, but are prepared to undergo rapid division cycles once these resources become available (Bergkessel et al., 2016). Conversely, depletion of a resource leads to a transition from a growing to a non-growing stage. The cell cycle is not arrested randomly during this transition, since this could halt key processes, especially DNA replication, at stages where severe and irreparable damage could occur (Kolter, 1993; Bergkessel et al., 2016). Thus, although the non-growing stage of the bacterial life cycle is poorly understood, even for model organisms, it is required and the transition between the growing and non-growing stage is regulated to ensure cell viability (Bergkessel et al., 2016). Previous research showed an involvement of poly-P during the non-growing stage of bacteria. For the model bacterium Escherichia coli, for example, mutants lacking ppk and thus the ability to synthesize poly-P failed to survive the state of growth arrest and lacked resistance to several stressors (e.g., heat, H₂O₂, or osmotic stress) (Rao and Kornberg, 1996; Shiba et al., 1997; Ault-Riché et al., 1998; Grillo-Puertas et al., 2016). For Pseudomonas aeruginosa, poly-P accumulation was also maximized during the non-growing stage (Kim et al., 1998).

Based on this evidence, we hypothesize that poly-P plays an important, perhaps regulatory, role during the non-growing stage of the cable bacteria cell cycle. This hypothesis is supported by the significant ¹⁸O labeling of poly-P granules in filaments with no, or very little, labeling in ¹³C, which was observed for filaments retrieved from all redox zones (Figure 5 and Table 4). This observation indicates that, in the suboxic zone, poly-P synthesis is one of the metabolic activities performed by cable bacteria during the non-growing stage of their cell cycle. For the filament fragments retrieved from the oxic zone, this observation suggests that cell growth is arrested earlier than poly-P synthesis. Recent research on E. coli showed that the ability of a cell to survive prolonged periods of starvation was dependent on the conditions prior to starvation where cells experiencing slower growth prior to starvation were shown to have consistently lower death rates (Biselli et al., 2020). In general, when a cell adjusts it physiology toward growth, it becomes less adapted for survival and vice versa. Synthesizing proteins that protect cells and increase their survival chances comes at the expense of synthesizing proteins needed for growth. The result is a trade-off between a fitness benefit in starvation and a fitness cost during growth (Scott et al., 2010). Thus, it appears that before cells of a filament residing in the suboxic zone enter the oxic zone and therefore a non-growing stage, they are already “prepared” to deal with oxidative stress and shortage of energy by entering into a stage of growth arrest which might enhance the chances of survival and increase the time a cell can survive in the oxic zone and thus provide energy for the cells in the suboxic zone below.