Based on their features, we classified membrane projections into the following categories: (1) tubular extensions with a uniform diameter and with an internal scaffold (Figure 1a and b and Figure 1—figure supplements 1 and 2); (2) tubular extensions with a uniform diameter and without a clear internal scaffold (Figure 1c–g and Figure 1—figure supplement 3); (3) tubular extensions with a vesicular dilation at the tip (a teardrop-like structure) and irregular dark densities inside (Figure 1h); (4) tubular extensions with irregular diameter or pearling tubes (Figure 2a–g); (5) interconnected chains of vesicles with uniform neck-like connectors (Figure 2h & i); (6) budding or detached vesicles: budding vesicles were still attached to the membrane, while detached vesicles were observed near a cell and could have budded directly or from a tube that pearled (Figure 3a–d and Figure 3—figure supplement 1); (7) nanopods: tubes of S-layer containing OMVs (Figure 3e–i). See Table 1 for a summary of these observations.

Scaffolded membrane tubes were observed only in H. pylori and had a uniform diameter of 40 nm. The H. pylori strain imaged (fliP^*) contains a naturally occurring point mutation that disrupts the function of FliP, the platform upon which other CORE proteins assemble (Fukumura et al., 2017; Fabiani et al., 2017; Minamino et al., 2019). In addition, the dataset contained other mutants in this fliP^* background including additional CORE proteins (ΔfliO and ΔfliQ), flagellar basal body proteins (ΔfliM and ΔfliG), and the tyrosine kinase required for expression of the class II flagellar genes (ΔflgS) (Lertsethtakarn et al., 2011; Figures 1a–b–4, Figure 1—figure supplement 1 and Table 3). This suggests that the H. pylori membrane tubes are unrelated to the CORE-dependent nanotubes that mediate cytoplasmic exchange in B. subtilis and other species (Bhattacharya et al., 2019; Pal et al., 2019).

Previously, H. pylori tubes were described as forming in the presence of eukaryotic host cells (Chang et al., 2018). Here, however, we observed tubes in H. pylori grown on agar plates in the absence of eukaryotic cells, suggesting that they also form in the absence of host cells. We observed some differences, though, from the tubes formed in the presence of host cells: the tube ends were closed, no clear lateral ports were seen, and the tubes were usually straight. While some of these tubes extended more than 0.5 μm, we never observed pearling. However, in some tubes, the internal scaffold did not extend all the way to the tip, and its absence caused the tube to dilate (from 40 nm in the presence of the scaffold to 66 nm in its absence, see Figure 4f and examples in Figure 1—figure supplement 1b and d). In some cases we also observed tubes stemming from vesicles resulting from cell lysis (Figure 4f, Figure 1—figure supplement 1b and d, Figure 1—figure supplement 2), and dark densities could be seen at the base of many of these tubes associated with vesicles (Figure 1—figure supplement 1d).

In Flavobacterium anhuiense and Chitinophaga pinensis, which are both endophytic species extracted from sugar beet roots, in addition to tubes with irregular diameter and OMVs (Figure 2g), tubular extensions with a uniform diameter, and a vesicular dilation (teardrop-like structure) were observed stemming from the sides of the cell in F. anhuiense (Figure 1h). Interestingly, irregular dark densities were observed inside these teardrop-like extensions (Figure 1h). Chains of vesicles connected by neck-like bridges were similarly observed in a single species: Borrelia burgdorferi. The bridges were consistently ~14 nm in length and ~8 nm in width. Where chains were seen attached to the OM, a neck-like connection was present at the budding site (Figure 2h). Vesicles in each chain were of a uniform size, usually 35–40 nm wide (e.g. Figure 2i), but occasionally larger (e.g. Figure 2h).

When both tubes and vesicles were observed in the same species, the tubes generally had a more uniform diameter than the vesicles, which were of variable sizes and often had larger diameters than the tubes (Figure 3—figure supplement 1 and Figure 3—figure supplement 2). In addition, when a tube pearled into vesicles, there was no clear correlation between the length of the tube and the initiation point of pearling, with some tubes extending for many micrometers without pearling while other, shorter tubes were in the process of forming vesicles (Video 1, Video 2, Video 3, and Figure 2). As usually only one (or part of a) cell is present in the cryo-tomogram, we cannot exclude that differences in the extracellular environments, like the presence of a cluster of cells in the vicinity of the individual cells with pearling tubes, might play a role in this observation. Pearling tubes differ from tubes with irregular diameter by the presence of a deep constriction in some part of the tube, while chains of vesicles are entirely made up of semi-circular vesicles connected by thin constrictions suggesting different mechanisms are responsible for the formation of these different extensions. While most pearling was seen at the tips of tubes, pearling occasionally occurred simultaneously at both proximal and distal ends of the same tube (Video 3). With one exception, pearling was seen in all species with tubes of uniform diameter and no internal scaffold. The exception was lysed Pseudoalteromonas luteoviolacea, which had narrow tubes only 20 nm in diameter (Figure 1g). Some lysed P. luteoviolacea contained wider, pearling tubes (Figure 2c). Interestingly, the tubes of various M. xanthus strains (see Materials and methods) and P. luteoviolacea could bifurcate into branches, each of which had a uniform diameter similar to that of the main branch (Video 4 and Figure 1d and Figure 1—figure supplement 3).

In Caulobacter crescentus tomograms, we identified structures very similar to the ‘nanopod’ extensions previously reported in D. acidovorans (Shetty et al., 2011). These structures consist of a tube made of the S-layer encasing equally spaced OMVs (Figure 3e–h and Video 5). The diameter of the S-layer tubes was ~45 nm and vesicles exhibited diameters ranging from ~13 to 25 nm. The nanopods were seen either detached from the cell (Figure 3e–g) or budding from the pole of C. crescentus (Figure 3h).

Next, we examined protein complexes associated with OMEs and OMVs that we could identify in our cryo-tomograms. These complexes fell into three categories: (1) seemingly randomly located complexes found on OMEs, OMVs, and cells; (2) seemingly randomly located complexes observed only on OMEs and OMVs; and (3) complexes exclusively located at the tip of OMEs/OMVs.

In the first category, we observed what appeared to be the OM-associated portion of the empty basal body of the type IVa pilus (T4aP) machinery in OMEs of M. xanthus. These complexes, which were also found in the OM of intact cells, did not exhibit a preferred localization or regular arrangement within the tube at least within the fields of view provided by our cryo-tomograms (Figure 5).

The second category of protein complexes, observed only on MEs and not on cells, contained two structures. The first was a trapezoidal structure observed on purified OMVs of S. oneidensis. The structure was ~11 nm wide at its base at the membrane and was seen sometimes on the outside (Figure 5c) and sometimes the inside of vesicles (Figure 5d). The second structure was a large crown-like complex. We first observed these complexes on the outer surface of MVs associated with lysed M. xanthus cells (Figure 6a). Occasionally, they were also present on what appeared to be the inner leaflet of the inner membrane of lysed cells (Figure 6b). The exact topology is difficult to determine, however, since the arrangement of IM and OM can be confounded by cell lysis. The structure of this complex was consistent enough to produce a subtomogram average from nine examples, improving the signal-to-noise ratio and revealing greater detail (Figure 6c). These crown-like complexes are ~40 nm tall with a concave top and a base ~35 nm wide at the membrane (Figure 6c). No such complexes were seen on OMEs and OMVs associated with intact M. xanthus cells. We identified a morphologically similar crown-like complex on the outside of some tubes and vesicles purified from S. oneidensis (Figure 6d–f). However, this complex was smaller, ~15 nm tall and ~20 nm wide at its base. As these MEs/MVs from S. oneidensis were purified, we cannot know whether they stemmed from lysed or intact cells. Interestingly, we found a similar large crown-like structure associated with lysed cells of two other species in which we did not identify MEs, namely Pseudomonas flexibilis and P. aeruginosa (Figure 6g–j and Figure 6—figure supplement 1).

In the third category, we observed a secretin-like complex in many tubes and vesicles of F. johnsoniae. Secretins are proteins that form a pore in the OM and are associated with many secretion systems like type IV pili and type II secretion systems (T2SS) (Chang et al., 2016; Ghosal et al., 2019; Gold et al., 2015). In tubes attached to the cell, the complex was always located at the distal tip (Figure 7, Figure 7—figure supplement 1, and Video 6). From 35 membrane tubes seen attached to cells, we identified a secretin-like complex at the tip of 25 of them (~70%). In OMEs disconnected from the cell, the secretin-like complex was always located at one end (Figure 7b & e). In total, we identified 88 secretin-like particles in 198 tomograms, none of which were located in the middle of a tube. As the MEs are less crowded than cellular periplasm and usually thinner than intact cells, we could clearly distinguish an extracellular density and three periplasmic densities in side views (red and purple arrows, respectively, in Figure 7a). Top views showed a plug in the center of the upper part of the complex (yellow arrows in Figure 7g & h). Subtomogram averaging revealed details of the complex, including the plug and a distinct lower periplasmic ring (Figure 7i & j & Figure 7—figure supplement 2). While the upper two periplasmic rings were clearly distinguishable in many of the individual particles (e.g. Figure 7a), they did not resolve as individual densities in the subtomogram average (Figure 7i). The extracellular density was not resolved at all in the average, suggesting flexibility in this part.

Previous studies showed that a species which belongs to the same phylum as F. johnsoniae, namely Cytophaga hutchinsonii, uses a putative T2SS to degrade cellulose (Wang et al., 2017). Since F. johnsoniae also degrades polysaccharides and other polymers, we BLASTed the sequence of the well-characterized V. cholerae T2SS secretin protein, GspD (UniProt ID P45779), against the genome of F. johnsoniae and found a hit, GspD-like T2SS secretin protein (A5FMB4), with an e-value of 1e^–9. This result and the general morphological similarity of this secretin to the published structure of the T2SS (Ghosal et al., 2019) suggested that the complex we observed might be the secretin of a T2SS. We therefore compared our subtomogram average with the only available in situ structure of a T2SS, a recent subtomogram average of the Legionella pneumophila T2SS (Ghosal et al., 2019; Figure 7i–l). The two structures were generally similar in length and both had a plug in the upper part of the complex. However, we also observed differences between the two structures. In L. pneumophila, the widest part of the secretin (15 nm) is located near the plug close to the OM, and the lower end of the complex is narrower (12 nm). In F. johnsoniae, this topology is reversed, with the narrowest part near the plug and OM (Figure 7i–l). Additionally, the lowest domain of the L. pneumophila secretin did not resolve into a distinct ring as we saw in F. johnsoniae and no extracellular density was observed in L. pneumophila, either in the subtomogram average or single particles (Ghosal et al., 2019).

Hylemonella gracilis cells were grown as described in Kaplan et al., 2020. P. luteoviolacea were grown as described in Shikuma et al., 2014. Magnetospirillum magneticum were grown as described in Cornejo et al., 2016. P. flexibilis 706570 were grown in lactose growth medium. S. oneidensis MR-1 cells were grown, as detailed in Phillips et al., 2020, in Luria Bertani (LB) media under aerobic conditions at 30°C with shaking at 200 rpm until they reached OD₆₀₀ of ~3. M. xanthus PilY1.3-sfGFP, M. xanthus ΔtsaP, and M. xanthus SA6892 strains were grown as described in Chang et al., 2016. B. burgdorferi B31 ATCC 35210 and H. hepaticus ATCC 51,449 cells were grown in standard media (see Briegel et al., 2009 and references therein).

C. crescentus was cultured in M2G and M2 media (prepared as described in Schrader and Shapiro, 2015); 5 mL of M2G were inoculated with a frozen stock of C. crescentus NA 1000 (wild-type and DpleD mutant cells; see Kaplan et al., 2021c) and grown overnight at 28°C; and 5 mL of the overnight culture was diluted in 15 mL M2G and grown at 28°C with a shaking speed of 200 rpm for ~2 hr until mid-log phase (OD₆₀₀ 0.4–0.5). The sample was then centrifuged at 5200 × g for 6 min at 4°C (same temperature for all subsequent centrifugation steps) and the pellet was resuspended in 1 mL M2 solution. The resuspended cells were transferred into a 2 mL microcentrifuge tube and centrifuged at 5200 × g for 5 min. All but ~250 μL of supernatant was removed, 650 μL M2 was added and the pellet was resuspended, and 900 μL cold Percoll (Sigma Aldrich) was added and the sample was centrifuged at 15,000 × g for 20 min. Samples were taken from the bottom of the tube to select swarmer cells.

Cells of F. johnsoniae strain CJ2618 (a wild-type strain overexpressing FtsZ, ATCC 17061) were taken from a glycerol stock, streaked onto a CYE plate with 10 µg/mL tetracycline and grown at 25°C. Subsequently, 5 mL of motility medium (MM) was inoculated with colonies from the plate and the culture was incubated at 25°C with 80 rpm shaking overnight. Then another 5 mL MM was inoculated with 80 µL of starter culture and placed at 25°C with no shaking until the next day when the cells were harvested and prepared for plunge-freezing.

H. pylori mutants (ΔfliM fliP*, ΔfliO fliP*, ΔflgS fliP*, ΔfliG fliP*, ΔfliQ fliP*) were grown from glycerol stocks on sheep blood agar at 37°C with 5% CO₂ for 48 hr and then either plunge-frozen directly or the cells were spread on another plate and left to grow for 24 hr before plunge-freezing. No difference could be discerned between the two samples by cryo-ET.

F. anhuiense (strain 98, see Carrión et al., 2019) and C. pinensis (strain 94, see Carrión et al., 2019) cells were grown overnight in 1/10 TSB at 25°C and 300 rpm shaking in 50 mL cultures. For sample preparation, cells were first concentrated by centrifugation; 3 μL aliquots of the cell suspension were applied to glow-discharged R2/2, 200 mesh copper Quantifoil grids (Quantifoil Micro Tools), the sample was pre-blotted for 30 s, and then blotted for 2.5 s (F. anhuiense) and 1 s (C. pinensis). Grids were pre-blotted and blotted at 20°C and at 95% humidity. Subsequently, the grids were plunge-frozen in liquid ethane using an automated Leica EM GP system (Leica Microsystems) and stored in liquid nitrogen.

S. oneidensis OMVs were purified as described in Phillips et al., 2020. First, S. oneidensis were grown in LB media until they reached OD₆₀₀ of 3. Subsequently, the cells were centrifuged at 5000 × g for 20 min at 4°C; the pellet contained whole cells while the supernatant contained the OMVs. To remove any cells present in the supernatant, it was filtered through a 0.45 µm filter. Subsequently, the supernatant was centrifuged at 38,400 × g for 1 hr at 4°C; the OMVs were in the resultant pellet. The pellet was resuspended in 20 mL of 50 mM HEPES pH 6.8 buffer, filtered through a 0.22 µm filter, spun again as described above, and ultimately resuspended in 50 mM HEPES pH 6.8.

For cellular samples, 10 nm gold beads were first coated with BSA (bovine serum albumin) and then mixed with the cells. Subsequently, 4 µL of this mixture was applied to a glow-discharged, thick carbon-coated, R2/2, 200 mesh copper Quantifoil grid (Quantifoil Micro Tools) in an FEI Vitrobot chamber with 100% humidity. Excess fluid was blotted away with filter paper and the grid was plunge-frozen in a mixture of ethane/propane. For the purified OMVs of S. oneidensis, the sample was first diluted to a 0.4 mg/mL concentration before it was applied to the grid (Phillips et al., 2020). Cryo-ET imaging of the samples was done either on an FEI Polara 300 keV field emission gun transmission electron microscope equipped with a Gatan imaging filter and a K2 Summit direct electron detector in counting mode, or a Thermo Fisher Titan Krios 300 keV field emission gun transmission electron microscope equipped with a Gatan imaging filter and a K2 Summit counting electron detector. For data collection, either the UCSF Tomography (Zheng et al., 2007) or SerialEM (Mastronarde, 2005) software was used. For OMVs, tilt series spanned –60° to 60° with an increment of 3°, an underfocus of 1–5 µm, and a cumulative electron dose of 121 e/Ų. For F. johnsoniae, tilt series spanned –55° to 55° with 1° increment, an underfocus of 4 µm, a cumulative electron dose of 100 e/Ų, and a 3.9 Å pixel size. For M. xanthus, tilt-series spanned –60° to 60° with an increment of 1°, an underfocus of 6 µm, and a cumulative electron dose of 180 e/Ų. For B. burgdorferi, tilt series spanned –60° to 60° with 1° increment, an underfocus of 10 µm, and a cumulative electron dose of 160 e/Ų. For H. hepaticus, tilt series spanned –60° to 60° with increments of 1°, an underfocus of 12 µm, and a cumulative electron dose of 165 e/Ų.

F. anhuiense and C. pinensis images were recorded with a Gatan K3 Summit direct electron detector equipped with a Gatan GIF Quantum energy filter with a slit width of 20 eV. Images were taken at magnification corresponding to a pixel size of 3.28 Å (C.pinensis) and 4.4 Å (F. anhuiense). Tilt series were collected using SerialEM with a bidirectional dose-symmetric tilt scheme (–60° to 60°, starting from 0°) with a 2° increment. The defocus was set to – 8 to 10 μm and the cumulative exposure per tilt series was 100 e⁻/A². Images were reconstructed with the IMOD software package.

Reconstruction of tomograms of cellular samples was done using the automatic RAPTOR pipeline implemented in the Jensen lab at Caltech (Ding et al., 2015). Tomograms of purified S. oneidensis OMVs were reconstructed using a combination of ctffind4 (Rohou and Grigorieff, 2015) and the IMOD software package (Kremer et al., 1996). Subtomogram averaging was done using the PEET program (Nicastro, 2006), with twofold symmetry applied along the particle Y-axis.