Bacterial strains used in this study are presented in Table 1. V. vulnificus strains were grown in LB media consisting of 0.5% (w/v) yeast extract, 1% (w/v) tryptone, and 1% (w/v) NaCl, pH 7.0. Plating medium contained 1.5% (wt/vol) Bacto agar. V. vulnificus strains were stored as 20% glycerol stocks at -80°C. Cultures were first streaked on LB plates and incubated overnight at 30°C. Individual colonies were used to inoculate 20 mL LB broth and were grown overnight at 30°C with vigorous shaking (250 rpm). Growth kinetics experiments were performed in three independent experiments with all V. vulnificus strains to determine the mid-log growth point for optimal OMV production. Briefly, liquid starter cultures were grown statically overnight, then a 1:2000 dilution of the starter culture was inoculated into fresh media and grown under standard conditions. Optical density (OD₆₀₀) readings were taken from flasks at 30 min intervals for 7–12 h post-inoculation. To simulate growth phase synchronization, 100 μL of overnight culture was inoculated into 100 mL of fresh LB broth and grown to mid-log phase (OD₆₀₀ ≅ 0.5, ∼3–4 h).

Neisseria meningitidis B and its CPS mutant, m7, were streaked on GC base agar (Difco) plates supplemented with glucose and iron and grown overnight at 37°C with 5% carbon dioxide. Liquid cultures were aerated in GC broth supplemented with glucose, iron, and 4.3% sodium bicarbonate overnight at 30°C. The following morning the culture was diluted into fresh media and grown to approximately mid-log phase.

To determine how specific mutations affected the motility of V. vulnificus strains, we carried out motility assays. Briefly, LB plates containing 0.3% (wt/vol) agar were spot-inoculated with 1 μL of culture normalized to OD₆₀₀ = 1 and incubated at 30°C. Motility zones were measured after 4 and 6 h. Cultures were spotted in triplicate, and the experiment was repeated three times. The same was repeated for complemented and reverted strains, however, these were photographed after 18 h due to slower growth.

The process of OMV release, the structure of discharged OMVs, and the morphology and integrity of V. vulnificus cells were characterized by cryo-electron microscopy (cryo-EM) and cryo-ET. Non-centrifuged 4 μL aliquots of bacterial cultures in exponential growth phase were combined with 1 μL of 10 nm gold fiducials (Sigma-Aldrich) and plunge frozen on glow-discharged, 200 mesh, copper Quantifoil grids (Quantifoil, Germany) in liquid ethane using a Vitrobot Mark III system (FEI, Hillsboro, OR, United States). The grids were observed in a JEOL JEM-2200FS field emission TEM (FEG-TEM) (JEOL Ltd., Tokyo, Japan) operating at 200 kV and equipped with an in-column energy filter, a DE-20 direct detection device (Direct Electron LP, San Diego, CA, United States), and a 4k × 4k UltraScan 4000 CCD camera (Gatan, Pleasanton, CA, United States). Single axis tilt series were acquired with SerialEM (Mastronarde, 2005), with an increment of 2° covering -60° to +60°. The cumulative dose was under 150 electrons/Ų and the defocus was between -4 and -10 μm. The tomograms were reconstructed by r-weighted back-projection using IMOD (Kremer et al., 1996; Mastronarde, 1997). Volume-rendered segmentations were performed manually using the Amira package (FEI Visualization Sciences Group, Hillsboro, OR, United States).

OMV diameters and distances from the center of each OMV to the outer membrane of the closest V. vulnificus cell were measured using IMOD’s drawing tools. Graphs were made in Prism software (GraphPad Software, Inc., La Jolla, CA, United States). Delaunay triangulations to determine distance to nearest neighbor were performed in ImageJ/Fiji (Plugins>Analyze>Delaunay Voronoi) (Schindelin et al., 2012).

Vibrio vulnificus cells were dispersed on the cryo-EM grids as 1.5–3 μm long bacteria. Cell morphology was consistent with Gram-negative species in which the cell cytoplasm was surrounded by the cell wall, which was composed of an inner membrane, peptidoglycan layer, and an outer membrane. Most cells maintained a single, polar membrane-sheathed flagellum, and occasionally pili were observed. OMVs were seen in abundance surrounding the capsulated strains and were arranged in a single plane as concentric rings (Figures 1, 2A,B). Also observed were the presence of tubes of outer membrane that pinched off into individual vesicles at the distal tips, forming a “beads-on-a-string” appearance (Figure 2A). These tubes were seen at the cell poles as well as at the division plane of dividing cells (Figure 2C).

To define the growth phases for V. vulnificus under our lab conditions, growth curves were performed. Overnight cultures were diluted into fresh LB media, and samples were taken at 30 min intervals for 12 h. A plot of OD₆₀₀ versus time (Figure 3A) revealed an approximate mid-log time point of 3.5–4 h and a late-log time point of 7–8 h. No growth defects were observed for the V. vulnificus strains examined. This was important to define because basic cell morphology is known to be dependent on growth phase of the cells, whether they are actively dividing or are in stationary phase. In addition, the expression of cell factors such as CPS is reported to peak during log-phase growth and decline during stationary phase (Wright et al., 1999). Differences in basic cell morphology can be seen in Figure 2. Figures 2A–C are from cells vitrified during mid-log growth phase. Cells have compact spacing between inner and outer membranes, with tubes and OMVs exiting the membrane via narrow, defined constrictions. Figures 2D–F represent morphologies seen at late-log growth phase. There is greater and more irregular spacing between the inner and outer membranes, suggesting a loss of peptidoglycan adhesiveness which results in a less controlled blebbing or sloughing of the outer membrane. Figure 2E shows another phenomenon associated with older cultures, which is flagellar sheath shedding to produce vesicles. This is a first reporting of this phenomenon in V. vulnificus, but it has been recently proposed that flagellar sheath shedding, although different in appearance, is a principal mechanism for LPS shedding via the outer membrane in Vibrio fischeri, and this shedding is attributed to rotation of the flagellum (Aschtgen et al., 2016).

To avoid complications in OMV quantification due to purification of vesicles and membrane fragments from bulk culture (Kulp and Kuehn, 2010), we relied on quantification of OMVs directly from images of individual cells and their immediate surrounding area. We imaged vitrified V. vulnificus cells at mid-log growth by acquiring 3 × 3 image montages at 10,000× magnification to include an area of ∼7.5 μm². The general appearance and OMV distribution for each strain are shown in Figure 4. Within this area, all OMVs were counted and their distance from the nearest outer membrane measured. Individual OMV diameters were also recorded. We performed this analysis for three cell strains: wild-type (CMCP6), an acapsular mutant (wza::TnPhoA; FLA1009), and a non-motile mutant (ΔmotAB; FLA674). These measurements are plotted in Figure 5. The total number of cells measured was for wild-type CMCP6 = 108, wza::TnPhoA acapsular mutant (FLA1009) = 167, ΔmotAB flagellar motor mutant (FLA674) = 148. The average number of OMVs per cell for wild-type CMCP6 = 63, wza::TnPhoA (FLA1009) = 54, ΔmotAB (FLA674) = 17. For wild-type V. vulnificus, the distance from the outer membrane fell largely within a defined ∼1 μm radius (mean = 695 nm). The unencapsulated mutant wza::TnPhoA, in distinct contrast, showed an equal number of OMVs but a very broad range of OMV distribution, from just on the outer membrane up to ∼3.5 μm away (mean = 1.5 μm). For the non-motile ΔmotAB mutant, there were far fewer OMVs per individual cell compared to wild-type or unencapsulated mutants. Their distribution was over a much narrower range than wild-type, only up to ∼1 μm away (mean = 570 nm). Perhaps the most significant observation was the distinct lack of OMVs close to the outer membrane in both capsulated strains, which was in contrast to the close proximity of the OMVs for the unencapsulated mutant, suggesting that the presence of CPS may contribute to the zone of OMV clearing within ∼200 nm of the bacterial outer membrane.

To assess the nearest neighbor distances between OMVs, we used Delaunay triangulation to calculate the mean center-to-center distances of convex arrangements of OMVs (Figure 6). For wild-type CMCP6, the mean distance was 202 ± 45 nm. For unencapsulated mutant the mean distance was 471 ± 214 nm, a much greater variability. And the non-motile mutant had a mean distance of 212 ± 52 nm, similar to wild-type. The diameters of OMVs differ by mutation as well. The mean diameter for wild-type was 47 nm, for the unencapsulated mutant it was 31 nm with greater variability, and 53 nm for the non-motile mutant.

To determine if OMV production has any dependence on motility, we performed motility assays of several V. vulnificus strains (Figures 3B,C). As expected, motility efficiency was drastically reduced when genes encoding motor proteins (ΔmotAB) or flagellins (ΔflaCDE, ΔflaFBA) were deleted. Motility was slightly increased when mutations compromised capsule production (Figure 3B), which may be attributable to a reduced friction coefficient due to the loss of the polysaccharide capsule. Combining the OMV measurement data with the motilities of individual strains, the non-motile, but still-flagellated mutant produced far fewer OMVs per cell (nearly fourfold fewer). These OMVs remained close to the outer membrane while maintaining a ∼200 nm zone of clearance. Their mean diameters (53 nm) were slightly larger than wild-type (47 nm). To confirm that the OMV phenotypes were the direct result of the mutations, the ΔmotAB mutation was complemented with a plasmid encoding motAB and the ΔflaCDE::cat mutation of the double ΔflaFBA::aph, ΔflaCDE::cat mutant was reverted by allelic exchange via natural transformation. Note that the ΔflaFBA::aph mutation that remains after flaCDE reversion is motile (Gulig et al., 2009). The production and arrangement of the OMVs in the complemented and reverted strains replicated that of wild-type strain CMCP6 (Supplementary Figure S1). Because the wza::TnPhoA mutation was polar on an extensive operon involved with production of capsule, it was not possible to complement this mutation. Similarly, because there was no selectable marker for reversion, it could not be reverted.

Using cryo-ET we are able to reconstruct the bacterium with its surrounding OMVs in 3D. We were able to see that the OMVs were arranged in a plane around the relatively thick cell and did not extend over or under the cell (Figures 1A,B). Also evident was the 200-nm region around the outer membrane that contained no OMVs. This particular arrangement of vesicles was abolished in the non-capsulated strain (Figure 4B). With respect to OMV content, cryo-ET revealed that cargo packaging was not a requirement for the release of OMVs. Some vesicles appeared empty, while others contained unidentified densities both on the surface and inside (Figure 7, white arrowheads).

The use of Quantifoil holey carbon support films with uniform 2 μm holes raised the question of whether the concentric arrangement of OMVs around the cell was an artifact of the shape of the holes. To assess this, we also imaged vitrified cells on lacey carbon film supports that contain holes of various sizes and shapes. It was determined that the formation of concentric rings by released OMVs was not an artifact caused by the shape of the holes in Quantifoil grids (Figure 8). Additionally, nearest neighbor distance calculations for these OMVs were also comparable (∼200 nm) to that of wild-type CMCP6 imaged on Quantifoil grids (Supplementary Figure S2).

Because the bacterial capsule is made of electron transparent polysaccharides and water, it is undetectable by cryo-EM. To visualize the capsule, we first used conventional TEM staining techniques. We used a protocol for lysine-acetate-based ruthenium red-osmium fixation (LRR) to enhance ruthenium red adherence to the capsule. As seen in Figure 9A, the capsulated wild-type CMCP6 strain was densely stained with the lysine and ruthenium red (measuring 80–100 nm), while the wza::TnPhoA unencapsulated strain in Figure 9B was practically devoid of this coating. This method is not quantitative for measurements of capsule thickness due to the potential for specimen dehydration and shrinkage during the staining protocol. Additionally, due to several centrifugation steps in the staining process, the OMVs appear to have been lost, thus providing no information about their arrangement.

To facilitate the preservation of fine structural details such as the outer membrane, the polysaccharide capsule, and the relative positioning of the OMVs, we employed the technique of SPRF, a form of HPF. Bacterial colonies were sealed within a thin copper tube and plunge frozen. The contents of the frozen tubes were freeze-substituted, embedded in resin, and sectioned. In Figure 9C, the CPS can be seen extending away from the outer membrane. Their lengths were somewhat variable, but the approximate distances between two cells or between a cell and an OMV were ∼200 nm. This illustrated that the 200 nm zone around V. vulnificus cells that is devoid of OMVs in capsulated strains is due to the presence of polysaccharide chains. The unencapsulated wza::TnPhoA mutant did not have the polysaccharide layer (Figure 9D). Also, as a control, we performed standard HPF of the same wild-type strain, however, the staining of the polysaccharides was less evident (Supplementary Figure S3). This phenomenon was due to the cells being packed more densely in the planchettes and for the subsequent processing steps that limited infiltration of stain and substitution media. Nevertheless, there were hair-like projections that were attributed to the polysaccharide capsule.

We believe that this concentric, planar arrangement of OMVs is exclusive to V. vulnificus. We provide a direct comparison to N. meningitidis serogroup B (Figure Figure 10). Cryo-EM 2D images of N. meningitidis cells illustrated that the cells do produce OMVs of variable morphology and dimensions and that the OMVs did not originate from the cells in a periodic fashion or as well-ordered and evenly spaced objects.