Porous Pellicle Formation of a Filamentous Bacterium, Leptothrix

Leptothrix cholodnii SP-6 and OUMS1 cells produce pellicles at the air-liquid interface under a static liquid culture condition (Fig. 1A and see Movie S1 in the supplemental material) (29), which has been reported for several bacteria. However, unlike pellicles consisting of a dense cell assembly, those of L. cholodnii are porous structures with distinctive geometric patterns. To understand how such geometric patterns are formed, we separately inoculated SP-6 and OUMS1 into well dishes and imaged the process of their pellicle formation at high magnification at 1 min time intervals. The time-lapse image sequences indicate that tiny clusters of woven cell filaments sequentially gather at the outer edges of larger filament aggregates for both strains (Fig. 1B and Movie S2). These observations lead us to conclude that the geometric patterning at the air-liquid interface is due to the random accumulation of tiny clusters of short cell filaments, a conserved feature for at least the L. cholodnii strains.

The pellicle formation process in the dish culture is difficult to be viewed vertically from above because the surface-covering pellicles disturb focus stability, making it hard to observe the behavior of the cells and how the cells or cell filaments colonize underneath the liquid surface. For analysis of cellular activity such as the movement of small cell aggregates in the liquid phase, we devised equipment to image the cell culture horizontally as illustrated in Fig. S1. We separately inoculated SP-6 and OUMS1 cells into cuvettes and tracked the process of aggregate formation at 1 min intervals. The image sequences reveal that the cells of both strains predominantly accumulate at the air-liquid interface and produce a thin and rough layer within ~22 h (Fig. 1C and Movie S3), which gradually gets thicker over time. However, we hardly observed small cell clusters in the liquid phases during pellicle formation (Fig. 1C and Movie S3). Indeed, magnified images show that small aggregates of pellicles develop due to the trapping of planktonic cells by cell filaments (Fig. 1D and Movie S4, left), indicating that single motile planktonic cells but not small aggregates may move toward the air-liquid interface. Next, we examined the requirement of flagella motility in pellicle formation.

Leptothrix cells are known to have a single monopolar flagellum, which enables the planktonic cells to explore favorable sites (19, 20). By constructing a deletion mutant of flagellar components, we investigated the role of the monopolar flagellum in the formation of floating pellicles. To construct the flagellum mutant, we prepared a spontaneous rifampin-resistant (Rif^r) mutant originating from SP-6 (the wild type [WT*], used for phenotype comparison to the disruptant) and its disruptant, in which Lcho_2735 and Lcho_2734, which encode putative components FlgA and FlgB, respectively, were deleted (ΔflgAB cells). Since we removed the promoter region of the flg operons encoding putative proteins FlgA through FlgM (Lcho_2724 through Lcho_2736), we expect expression of these proteins to be abrogated.

We confirmed the replacement of the flgA and flgB genes with the kanamycin resistance gene (Km^r) in candidate clones by colony PCR (Fig. S2A). Representative atmospheric scanning electron microscopy (ASEM) images revealed that the ΔflgAB cell loses its polar flagellum (Fig. S2B). Subsequently, we found that the ΔflgAB cells lose their cell motility driven by the flagellum. Movie S5 shows representative swimming planktonic WT* cells. In contrast, we did not observe swimming ΔflgAB cells, suggesting that the replacement of these genes with Km^r abrogates the flagellar function.

In shaking cultures, ΔflgAB cells produced aggregates of cell filaments covered by a sheath, like the WT* (Fig. S2C and E) (34). Contrary to WT*, ΔflgAB cells failed to form porous pellicles in a static dish or cuvette cultures (Fig. 2A and B, Fig. S2D, and Movie S6). Using high-magnification time-lapse images, we found WT* aggregates and swimming planktonic cells at the air-liquid interface (Fig. 2C, left). However, we observed several long cell filaments but no planktonic cells in the ΔflgAB cell culture (Fig. 2C, right). From these observations, we conclude that like Bacillus (16), the flagellum-driven motility of WT* assists planktonic cells to move toward the air-liquid interface and eventually contributes to expansion of the porous pellicles.

The Leptothrix sheath is known to consist of nanofibrils and to have multiple functions, such as enabling surface attachment of planktonic cells to solid substrates, holding divided cells in line, mediating interactions between other cell filaments, and facilitating the encrustation of iron oxide-rich particles (22–24). Furthermore, the nanofibrils elongating further outwards from the sheath structure efficiently expand its superficial area (28), which might enable cell filaments to capture swimming planktonic cells as well as to facilitate assembly by the sequential docking with other small cell-filament clusters.

To explore whether nanofibril secretion is critical for the distinctive patterning of the net-like pellicle, we constructed a disruptant lacking the Lcho_3510 gene (hereafter lthA), encoding a putative exopolysaccharide biosynthesis polyprenyl glycosylphosphotransferase whose homolog, sthA, is known to be responsible for the construction of the sheath structure in Sphaerotilus natans (33). After acquiring candidate clones of lthA deletion mutant cells (ΔlthA cells) using colony PCR, we confirmed the proper replacement of the gene (Fig. S3A). We further confirmed the absence of a sheath structure around the periphery of the ΔlthA cell by atmospheric scanning electron microscopy (ASEM) (Fig. S3B) and fluorescent imaging of nanofibrils stained by Alexa Fluor 594 C5-maleimide (Alexa594-SH) (Fig. 3A).

In a shaking liquid culture containing MSVP medium, the WT* strain generates fibrous filament aggregates. In contrast, ΔlthA cells are uniformly distributed in the entire medium, making it cloudy (Fig. S3C). Time-lapse imaging in a static dish culture reveals that the ΔlthA strain forms chains containing 4 or 5 cells at the bottom surface, which break into 1 or 2 cells within ~2 h (Fig. S3D, right). However, for the WT* strain, we observed very little fragmentation of chained cells (Fig. S3D, left). These observations indicate that the ΔlthA cells are not able to form sheathed cell filaments due to a failure of nanofibril secretion regardless of liquid culture condition.

In static dishes and cuvette cultures, unlike ΔflgAB cells, ΔlthA cells produce pellicles. Since planktonic ΔlthA cells maintain their flagellum-driven motility, this could enable them to move up to the interface (Movie S5, right). Notably, ΔlthA cells congregate to form a dense and relatively thick pellicle (Fig. 3B and C, t = 40 h; Movie S7), which consists of cell assemblies that are structurally different from the net-like pellicle formed by WT* cells. Contrary to WT*, ΔlthA pellicles release their resident cells into the liquid phase (Fig. 3C, t = 45 h; Movie S7). We use atomic force microscopy (AFM) to image dried samples from WT* and ΔlthA pellicles. Confirming our expectations, we found that WT* cells are aligned in filaments, whereas the ΔlthA pellicle cells, which lack nanofibrils, have aggregated (Fig. 3D). This difference in aggregation morphology between WT* and ΔlthA cells is likely due to differences in cell surface properties.

To estimate the cell density at the air-liquid interface during pellicle formation, we analyzed the interface area coverage and thickness of WT* and ΔlthA pellicles from image sequences (Movie S7). In dish culture, the percentage of coverage of WT* pellicles plateaus at ~55% after ~30 h. In contrast, the percentage of coverage of ΔlthA pellicles reaches 100% within the first 10 h, before collapsing into the liquid phase (Fig. 4A and Fig. S4A and B). To facilitate measurement of pellicle thickness, we use cuvette cultures; these cultures will have the same qualitative features as dish culture, but will differ in terms of dynamics due to the differences in interface area. From time-lapse sequences, we find that the thickness of WT* pellicles reaches a constant value of 30 μm at ~25 h. In contrast, ΔlthA pellicles demonstrate a long period during which thickness increases slowly, followed by a rapid rise and fall in thickness, which is due to pellicle collapse (Fig. 4B and Fig. S4C and D). We find that ΔlthA pellicles reach a maximum thickness of ~100 μm. From these values and from our surface coverage percentage, we estimate that at collapse, the ΔlthA pellicles have ~6× more mass than WT* pellicles, assuming the same cell density for both strains. The rapid coverage at the air-liquid surface by ΔlthA cells suggests that they have a higher cell surface hydrophobicity than WT* cells. Furthermore, due to the inability to form filaments, it appears that ΔlthA cells can form much denser structures than WT* cells, which are confined to filamentous chains.

Adhesion at the air-liquid interface by pellicles can be mediated by surface-active agents (17). However, the surface tension of the MSVP medium was nearly constant (≈71.5 mN/m) when WT* or ΔlthA cells were cultured for 72 h (Fig. S5), suggesting that they were not secreting surface-active molecules. For Bacillus subtilis, prior to pellicle formation, a transition from planktonic cells to nonmotile short cell chains has been reported (16). Its EPS production assists the cell chain in adhesion and alignment that lead to the formation of a dense pellicle (13). Our high-magnification images of the ΔlthA pellicle reveal a loose packing of swimming planktonic cells at the air-liquid interface (Fig. S6), which over time transforms into nonmotile short cell chains (t = 3 h) that eventually become the part of larger assemblies that have loose cell-cell interactions (t = 5 h). This is distinct from SP-6 and WT* pellicles, which are composed of entangled cell filaments (Fig. 1D). These observations indicate that sheathed cell chains are crucial for the formation of porous pellicles consisting of entangled cell filaments. The nanofibril-free, and thus sheathless, ΔlthA cells form thick assemblies without the support of nanofibrils but are unable to form a net-like pellicle.

It has been reported that the glycoconjugates that compose the SP-6 nanofibrils are functionalized with dissociable SH- and NH₂-terminal groups that assist cell filaments in self-entanglement and adsorption of Fe-rich particles (25, 27, 28). We investigated the effect of these terminal groups on the accumulation of tiny filamentous clusters at the air-liquid surface. We incubated SP-6 cells in MSVP containing Alexa594-SH and fluorescein-NH₂ reagents, which specifically block the binding of NH₂ and SH groups, respectively, by conjugation (24, 28, 35) and tracked the formation of floating pellicles (Fig. S7). We observed no significant effect when blocking NH₂- and SH-terminal groups in the net-like SP-6 pellicles. We thus conclude that NH₂- and SH-terminal groups are not required for porous pellicle formation, although these play a primary functional role in interaction with iron oxides and nanofibrils, respectively (27, 28).

We have shown that sheathless ΔlthA cells tend to self-aggregate compared to WT* cells (Fig. 3D). Self-aggregation of bacteria is known to be primarily mediated by the recognition of surface components such as proteins and exopolysaccharides (36). Moreover, cell surface hydrophobicity is also known to be an important factor for self-aggregation (37). We therefore examined cell surface hydrophobicity using hydrophilic fluorescent sicastar-greenF microspheres (100 nm in diameter), which are designed to bind selectively to hydrophilic but not to hydrophobic substrates. We separately incubated sheathed WT* and sheathless ΔlthA cells with 1 mg/mL of sicastar-greenF microspheres and then acquired fluorescence images. As we expected, a strong bonding of the sicastar-greenF to hydrophilic WT* cells resulted in an intense green fluorescence at the filament periphery (Fig. 5A, upper; Fig. S8), while hydrophobic ΔlthA cells showed a weak fluorescence signal, except at the edges of the cell (Fig. 5A, lower). These results indicate that hydrophilic components are widely distributed on the cell surface of WT*. Furthermore, the orange fluorescence of Alexa594-SH, which blocks SH groups of nanofibrils in sheath structures, is perfectly matched with green fluorescence (Fig. 5A), indicating that (i) nanofibrils, which form the sheath structures, are responsible for cell surface hydrophilicity, and (ii) SH groups in nanofibrils have no significant effect on their hydrophilicity. However, we could not determine whether the hydrophilicity of ΔlthA cells is due to polar flagella that contain hydrophilic domains (38).

To be fully assured of the cell surface hydrophobicity, we next examined the effect of nanofibrils and sheath structure on the cell surface hydrophobicity by measuring the two-phase cell-water-air contact angle by depositing a water droplet on a bacterial lawn (39). We found that the average contact angle (θ) of the water droplet for ΔlthA cells is 60% larger than that of the WT* cells (Fig. 5B). This indicates that the sheath structure contributes to cell surface hydrophilicity, and thus sheathed WT* cells are more hydrophilic than sheath-free ΔlthA cells.

Although several unknown factors may influence this result, cell hydrophobicity is meaningful for the pellicle formation process since it affects adsorption energy at the air-liquid interface. The adsorption energy (E) of a bacterium adsorbed at an interface can be estimated from the equation $E=-\text{γ}S{(1-\left|\cos \text{θ}\right|)}^2$, where γ is the surface tension at the air-liquid interface, S is the bacterium surface area, and θ is the air-liquid contact angle of cells. Here, the negative sign indicates energy is required for a cell to escape the interface (40–42). We measured γ = 71.5 mN/m (Fig. S5) and S = 6 μm² (20). Evidently, for the same air-liquid interface, the adsorption energy of a bacterium mainly depends on its contact angle. Using the measured average value of the contact angle and the adsorption energy equation, we estimate that ΔlthA cells are adsorbed ~5 times more strongly than the WT* cells. This stronger adsorption of ΔlthA cells appears to reduce both swimming motility and Brownian motion, as is visible in Movie S4, which shows a rotational motion of planktonic ΔlthA cells, while aggregates of ΔlthA cells do not show any motion. Furthermore, due to the larger value of ΔlthA cell hydrophobicity, these cells will deform the air-liquid interface more than WT* cells (43–45). Over time, due to the “Cheerios effect,” in which interfacial particles attract each other due to surface tension, ΔlthA cells may attract each other and form a dense pellicle (46), as is visible in Fig. S5. The reason for the rotational motion of planktonic ΔlthA cells (Movie S4, right) is not clear, but we speculate it may be due to a countercoupling of flagellar motile energy and interface adsorption energy. In contrast, weaker adsorption of hydrophilic planktonic WT* cells and their flagellar motility allows them to move and explore the interface, as can be seen in Movie S4 (left). While in motion, they randomly stick to the surface by adhesive nanofibrils and initially develop small aggregates that further enlarge as small aggregates stick together and develop net-like pellicles.

In this study, we investigate the mechanism of pellicle formation by the filamentous bacterium Leptothrix cholodnii SP-6 at the air-liquid interface. Our findings show that the flagellar motility of WT* cells is primarily responsible for transporting cells from bulk liquid to the interface and that cell surface hydrophilicity in part controls the pellicle density and shape. We find ΔlthA cells that were unable to produce sheath structures are 60% more hydrophobic than the sheathed WT* cells, showing that the presence of the sheath structure increases cell hydrophilicity. The estimation of air-aqueous interface adsorption energy shows hydrophobic ΔlthA cells are adsorbed ~5x more strongly than WT* cells. This larger adhesion energy inhibits their movement away from the interface, where they aggregate into the dense pellicle. In contrast, weaker adsorption of WT* cells enables them to swim to, explore, and depart from the interface. During this movement, the planktonic WT* cells stick to one another due to their filamentous structures and form small aggregates that gradually enlarge by the random docking of smaller aggregates, forming branched pellicles.

Bacterial pellicle formation is shown to be triggered by inducers such as iron for Pseudomonas fluorescens, dimethyl sulfoxide or ethanol for E. coli, and arabinose for Vibrio fischeri (47–49). Therefore, it might be possible that the pellicle formation of L. cholodnii is induced by minerals such as iron oxides, but further studies are needed to identify the factors required for pellicle formation. In addition, since we have reported that sheath-forming bacteria Leptothrix, Sphaerotilus, and Thiothrix excrete different types of glycoconjugates (25, 50, 51), we plan to investigate if there are differences in sheath hydrophobicity in future studies.

The filamentous bacteria of the genera Leptothrix and Sphaerotilus are widely distributed in wastewater treatment plants (19, 21). However, their proliferation and aggregation often clog water distribution systems, which reduces the efficiency of the treatment plants (52). The findings of the present study provide a deeper understanding of biological and physiochemical factors that influence pellicle formation. We hope to contribute to finding an effective method to control their proliferation that may be employed in large-scale applications.

In this study, we used Leptothrix cholodnii strains SP-6 (=ATCC 51168) (34) and L. cholodnii OUMS1 (=NITE BP-860) (19). We cultured these strains by transferring frozen stock to an MSVP agar plate (34) followed by incubation at room temperature (RT). After 1 week of culture, we transferred 1 to 3 single colonies to 25 mL of liquid MSVP medium and incubated them in a reciprocating shaker at RT and 70 rpm for 2 days. Escherichia coli strains S17I, DH5α (our laboratory stock), and SN1187 (NBRP-E.coli at NIG), normally used for plasmid manipulations or plasmid transfer, were cultured in LB medium (53) containing 100 μg/mL of ampicillin sodium (Amp) (Nacalai Tesque, Kyoto, Japan) for 12 h in a reciprocating shaker at 37°C and 190 rpm.

For the gene manipulation of SP-6, we first constructed the plasmids pUC-ΔLcho_0008::Km^r-Term-mob and pUC-mob, which were used as the templates for PCR. The 1.3-kb HindIII-SmaI fragment carrying the kanamycin resistance (Km^r) gene cassette and the 1.8-kb BamHI fragment carrying the mob site from pSUP5011, generously provided by Tamaki Hideki and Morinaga Kana (National Institute of Advanced Industrial Science and Technology, Tsukuba), were cloned into the HindIII-SmaI gap and BamHI site of pUC18, resulting in the construction of plasmids pUC-Km^r and pUC-mob, respectively (see Fig. S9 in the supplemental material). The plasmid pUC-ΔLcho_0008::Km^r-Term-mob was constructed using the in vivo E. coli cloning system (iVEC) (54) as follows. Approximately 1.5-kb fragments located at upstream and downstream regions of the Lcho_0008 gene were amplified by PCR using the Tks Gflex DNA polymerase kit (TaKaRa Bio, Inc., Kusatsu, Shiga, Japan). A genomic DNA isolated from SP-6 cells (55) was used as a template, and Lcho0008_F_out/Lcho0008_Km_2 and term_ Lcho0008_1/Lcho0008_R_out were used as primer sets (Table S1). To fully stop the transcription of the Km^r gene downstream of the stop codon, the Km^r cassette and the transcriptional terminator of the Bacillus subtilis spoVG gene (Term) were amplified by PCR using primers Lcho0008_Km_1/Km_term_2 and Km_term_1/term_ Lcho0008_2 from pUC18-Km^r and the genomic DNA isolated from B. subtilis 168, respectively. These four PCR products were then ligated by overlap PCR using pUC_Lcho0008_1 and Lcho0008_pUC_2 primers. The resultant amplified product (ΔLcho_0008::Km^r-Term) and the inverse PCR product of pUC-mob, which excludes the lac promoter region, were amplified using primer set pUC_Lcho0008_2/Lcho0008_pUC_1 and cointroduced into E. coli SN1187 to be circularized by the iVEC system, resulting in the construction of plasmid pUC-ΔLcho_0008::Km^r-Term-mob. This strain is engineered to employ the exonuclease III activity of XthA for exposure of ~40 bp of the homologous region at each end that enables annealing of DNAs, while DNA polymerase I activity of PolA is used for the gap repair of annealed DNAs (54).

Next, plasmids used to disrupt Lcho_3510 (here lthA), and Lcho_2735 and Lcho_2734 (here flgAB) genes by double-crossover integration were constructed as follows. In the case of lthA, ~1.5-kb fragments located at upstream and downstream regions of the open reading frame were amplified by PCR. We used Lcho3510_F_out/Lcho3510_km_2 and km_Lcho3510_1/Lcho3510_R_out as primer sets (Table S1) and the genomic DNA of SP-6 as a template. The Km^r-Term gene was amplified using Lcho3510_km_1/km_Lcho3510_2 as a primer set and pUC-ΔLcho_0008::Km^r-Term-mob as a template. These three PCR fragments that contain upstream and downstream of sections of the lthA gene, respectively, were ligated with the Km^r-Term gene by overlap PCR using pUC18_Lcho3510_1/Lcho3510_pUC18_2 as a primer set. The resultant fragment (ΔLcho_3510::Km^r-Term) and the inverse PCR product of pUC-mob amplified using primer set pUC_Lcho3510_2/Lcho3510_pUC18_1 were circularized by the iVEC system. The resultant plasmid, pUC-ΔlthA::Km^r-Term-mob, contains an lthA disruption cassette for replacement with the Km^r-Term gene and mob site for conjugation. The sequence of the constructed plasmid was confirmed using seq primers before gene disruption (Table S1). Another plasmid, pUC-ΔflgAB::Km^r-Term-mob, used for the disruption of flgA and flgB genes located next to each other in the SP-6 genome, was constructed in the same way.

To generate gene disruptants of SP-6, we first obtained spontaneous rifampin-resistant SP-6 strain (WT*) by successive transfer of SP-6 cells to MSVP plates supplemented with rifampin (Rif [10, 15, and 20 mg/L]). We further transformed the resultant WT* cells with exogenous DNA via conjugation with the E. coli S17I strain. The contact opportunity between S17I and WT* cells is important to enhance the efficiency of plasmid transfer. However, the sheath consisting of entangled nanofibrils that cover the cell filaments of SP-6 (34) diminishes such a contact opportunity. Previously we discovered that SP-6 cells failed to excrete nanofibrils under a Ca-deficient condition and produce short cell filaments without covering with a sheath (29). Based on this finding, we cultured WT* cells in routine MSVP for 2 days and then cultured them in MSVP lacking Ca for an additional 6 h. We concentrated these cells ~30-fold by centrifugation (7,000 × g for 5 min) and passed the cell filament aggregates through a 23G needle three times to break it apart into individual cells. In parallel, S17I cells carrying plasmid pUC-ΔlthA::Km^r-mob were inoculated into 4 mL of LB medium containing 100 μg/mL Amp and incubated at 37°C for 12 h. Subsequently, S17I cells were then centrifuged (3,000 × g for 1 min) and washed three times in 0.9% NaCl solution (normal saline). The mixed cellulose ester (MCE) membrane (diameter, 25 mm; pore size, 0.22 μm; MF-Millipore, Burlington, MA, USA) was used as a scaffold for the conjugation process. One-half milliliter of the concentrated WT* suspension was spread on an MCE membrane placed on nutrient broth plates (Becton, Dickinson and Company, Franklin Lakes, NJ, USA) and dried for 1 h. An equal volume of S17I suspension was spread on the same membrane and incubated to accelerate mating at RT for an additional 16 h. The donor/recipient ratio was ~1:2. The MCE membrane was transferred to a 15-mL tube and soaked in 10 mL of normal saline, and bacterial cells placed on the membrane were collected by centrifugation (7,000 × g for 5 min). The resultant cell suspension was spread on MSVP plates containing 20 μg/mL Rif and 50 μg/mL Km and incubated at RT for 4 to 5 days. The obtained colonies were selected and streaked on the same plates to allow for single colony isolation. The single colonies were used for further analyses in an antibiotic-free MSVP medium since the parental WT* cells used as the control for deletion mutants are Km sensitive (Km^s).

Using single colonies freshly formed on the Rif- and Km-containing MSVP plates, we verified gene disruption by colony PCR using Lcho3510_F_out and Lc3510_Km_2 as ΔlthA primer set 1, Km_Lc3510_1 and Lcho3510_R_out as ΔlthA primer set 2, Lcho1694_F_BamHI and Lcho1694_R_BamHI as the internal control set for the ΔlthA strain, Lcho2735_out_F and Km_Lchp2734_F as the forward primers, and Km_Lcho2734_R and Lcho2734_out_R as the reverse primers for the ΔflgAB strain (Table S1). We also confirmed the disruption by sequencing genomic DNAs prepared from the individual candidate cells using del3510_seq primers for the ΔlthA strain and primer pUC18_Lcho2735_F/Km_Lcho2734_R/Km_Lcho2734_F for the ΔflgAB strain (Table S1).

To image aggregated cell filaments on the bottom glass surface, we inoculated 600 μL of WT* and ΔlthA cultures into a polymer coverslip bottom dish (μ-Dish 35-mm Quad; Ibidi GmbH, Gräfelfing, Germany) and imaged them with an LSM780 confocal microscope (Carl Zeiss, Oberkochen, Germany) at 15-min intervals. To image the formation of pellicles at the air-liquid interface, we incubated 2 mL of culture in a 24-well dish (Iwaki, Tokyo, Japan) and imaged the pellicles using an Axio Zoom.V16 microscope (Carl Zeiss) equipped with a heater unit (Tokai Hit, Fujinomiya, Japan) at 0.5- to 1-min intervals. To examine cellular motility (Movie S5), we separately pipetted 1 mL of WT*, ΔflgAB, and ΔlthA cultures into 3.5-mm glass bottom dishes (Iwaki) and immediately imaged them at 18.322-ms intervals using a SpinSR10 microscope (Olympus, Tokyo, Japan). The surface area coverage and thickness of WT* and ΔlthA pellicles were analyzed from image sequences using ImageJ. For estimation of surface area coverage, the area covered by cells at the air-liquid surface was divided by the total area. For estimation of thickness, the height of the water surface alone (measured at t = 0) was subtracted from the total height of the pellicle at respective times.

To image the floating pellicles in liquid culture horizontally, we incubated 3 mL of culture in clear cuvettes (As One, Osaka, Japan) and imaged them at 1-min intervals from a horizontal direction using the equipment illustrated in Fig. S1. This equipment was constructed using a Raspberry Pi 4 Model B (Raspberry Pi Foundation, Cambridge, England), a Raspberry Pi HQ camera (Raspberry Pi Foundation), and an official CGL 6-mm wide-angle lens for the Raspberry Pi camera (CGL Electronic, Hong Kong, China) equipped with a 27-mm circular polarized light filter (Hakuba Photo Industry, Tokyo, Japan). A Python script for controlling the equipment was written.

The sheaths of SP-6 contain abundant terminal functional groups such as SH and NH₂ that serve as cross-linkers for stabilizing entangled nanofibrils and direct interaction between nanofibrils and Fe(III) minerals (25–28). To clarify whether these reactive end groups are required for the accumulation of porous pellicle, we incubated SP-6 cells in MSVP containing the fluorescein-NH₂ reagent (1:50 dilution) (fluorescein labeling kit-NH₂; Dojindo, Kumamoto, Japan) (24) and Alexa Fluor 594 C5-maleimide (Alexa594-SH [(30 μM]) (Thermo Fisher Scientific, Waltham, MA, USA) (29) for masking end groups NH₂ and SH, respectively. To visualize the sheath consisting of entangled nanofibrils, the WT*, ΔflgAB, and ΔlthA cells were incubated in 3.5-mm glass bottom dishes for ~4 h and stained with Alexa594-SH for the sheath and 4′,6-diamidino-2-phenylindole (DAPI [0.5 μg/mL]) (Dojindo) for DNA, respectively. After being washed once in MSVP, the cells were imaged using a SpinSR10 microscope.

To examine the cell surface hydrophobicity of Leptothrix cells, we performed the contact angle measurements of water droplets on a bacterial lawn using Kyowa model no. DMs-401 followed by a similar method reported by van Loosfrecht et al. (39). Sixty milliliters of a 2-day culture of WT* or ΔlthA cells in MSVP was collected by centrifugation at 7,000 × g for 5 min and washed three times in ultrapure water. The aggregates of WT* cells were passed through a 23G needle three times to break them apart into individual cells. Subsequently, ~20 mL of the cell suspension (optical density at 600 nm [OD₆₀₀] of ~0.5 to 0.75) was passed through 5-μm-pore filters (Sartorius AG, Göttingen, Germany) to remove the bacterial chunks. Next, WT* or ΔlthA cells were collected on polyvinylidene difluoride (PVDF) hydrophilic filters (0.45-μm-micropore; Sartorius AG) to prepare the bacterial lawn. The filters with a bacterial layer were attached to glass slides using double-sided tape and dried for 2 to 3 h at 30°C. We deposited 5-μL droplets of ultrapure water on the filter and measured its contact angle. The cell’s contact is the angle of the droplet 2 s after depositing it on the bacterial surface.

WT* and ΔlthA cells inoculated in 25 mL MSVP medium (t = 0). At respective times, 1.5 mL culture was picked up and passed through a 0.2-μm-pore syringe filter (Sartorius AG). The surface tension of the resultant cell-free medium was performed at room temperature using a surface tensiometer (DY-700, Kyowa Interface Science, Saitama, Japan). As the control, ultrapure water was employed. All the experiments were performed in triplicate.

We employed fluorescent microsphere sicastar-greenF (100 nm in diameter) (micromod Partikeltechnologie GmbH, Rostock, Germany), designed to adhere to a hydrophilic substrate by terminal Si-OH groups. sicastar-greenF was suspended in MSVP (2 mg/mL) and sonicated for 1 h at 40 kHz to prepare a homogenous suspension. In parallel, cell suspensions of WT* were fluorescently stained with DAPI and Alexa594-SH for 5 min followed by a single wash with MSVP. The cell suspensions were added separately to an equal volume of the microsphere suspension, vortex mixed vigorously for 1 min, and transferred to glass bottom dishes. The samples were imaged immediately using a SpinSR10 microscope.

We used atmospheric scanning electron microscopy (ASEM) (JASM-6200, JEOL, Tokyo, Japan) to directly visualize the distribution of flagella and nanofibrils on SP-6 cell surfaces (24). We inoculated the cells into an ASEM membrane dish (35 mm in diameter) (JEOL, Tokyo, Japan) by transferring 3 mL of a 2-day culture. After 4 to 16 h of incubation, the cells were washed in sterile phosphate-buffered saline (PBS), and the remaining surface-attached cells were fixed with 1% (vol/vol) glutaraldehyde. After quenching in 50 mM ammonium chloride, the nanofibrils secreted from cell surfaces were visualized by staining them with positively charged gold nanoparticles (Nanoprobes, Yaphank, NY, USA). To observe nanofibrils, the cells were further stained with 2% (vol/vol) phosphotungstic acid (TAAB Laboratories Equipment, Aldermaston, Berkshire, England). Finally, specimens were soaked in 1% (vol/vol) ascorbic acid and observed using ASEM at an acceleration voltage of 30 kV (56).

We use an SPM-Nanoa atomic force microscope (Shimadzu, Kyoto, Japan) driven by the Nano 3D Mapping Fast software to visualize the self-aggregation of WT* and ΔlthA cells. These cells were independently inoculated into 8 mL of culture in a 6-well dish (Iwaki, Tokyo, Japan) and statically incubated for ~24 h to generate pellicles. Cells comprising pellicles were then scooped onto coverslips and air dried for ~30 min before atomic force microscopy (AFM) observation.

Statistical analyses were carried out using unpaired Welch's t tests.

The data supporting this study’s findings are available at https://figshare.com/s/428ed5ea5bc08b984474.