Three-Dimensional Macromolecular Organization of Cryofixed Myxococcus xanthus Biofilms as Revealed by Electron Microscopic Tomography

Despite the fact that most bacteria grow in biofilms in naturaland pathogenic ecosystems, very little is known about the ultrastructureof their component cells or about the details of their communityarchitecture. We used high-pressure freezing and freeze-substitutionto minimize the artifacts of chemical fixation, sample aggregation,and sample extraction. As a further innovation we have, forthe first time in biofilm research, used electron tomographyand three-dimensional (3D) visualization to better resolve themacromolecular 3D ultrastructure of a biofilm. This combinationof superb specimen preparation and greatly improved resolutionin the z axis has opened a window in studies of Myxococcus xanthuscell ultrastructure and biofilm community architecture. Newstructural information on the chromatin body, cytoplasmic organization,membrane apposition between adjacent cells, and structure anddistribution of pili and vesicles in the biofilm matrix is presented.

INTRODUCTION

Top

INTRODUCTION

Bacteria are usually found concentrated at solid-liquid interfaces,where they colonize natural and man-made surfaces in community-likestructures termed bacterial biofilms (6). Bacteria in biofilmsexhibit protein expression patterns different from those ofplanktonic bacteria (41). Within a biofilm, bacteria are typicallyembedded in an extracellular polymeric substance (EPS) matrix(7, 17, 43) that protects the microbial community members fromdesiccation, from phagocytosis, and, in the case of human pathogens,from the host immune system (8).

While a variety of biofilms have been studied in depth by opticalimaging approaches, only a few have been faithfully preservedand visualized at high resolution by transmission electron microscopy(3, 22, 25, 26, 45, 48, 49). A recent ultrastructural studyof Pseudomonas aeruginosa biofilms shows that high-pressurefreezing/freeze-substitution results in superior preservation(22). Unlike conventional sample preparation, millisecond cryofixationand low-temperature dehydration protocols minimize artifactsfrom chemical fixation, extraction, and aggregation (15, 16,19, 20, 22).

Conventional two-dimensional (2D) imaging provides x,y positionalinformation but cannot resolve features along the z direction.Conventional 2D projection imaging of thin ( $~$ 70- to 100-nm) sectionsresults in the superposition of $~$ 5 to 20 layers of protein alongthe electron path, obscuring molecular details. True 3D visualizationrequires electron tomographic data acquisition and 3D reconstruction,upon which both intracellular and extracellular features areresolved. In this way, cellular organelles, molecular machines,and macromolecular complexes can be visualized in their nativemicroenvironment. To date, electron tomographic 3D reconstructionof a bacterial biofilm has not been reported.

Myxococcus xanthus is a well-studied gram-negative soil scavengerand predator (21), and it serves as a model system for biofilmformation. M. xanthus moves over surfaces either by adventurous(A) motility or social (S) motility (for a review, see reference18). It forms spore-bearing fruiting bodies upon starvation,in a process that requires the coordinated movement of manythousands of cells. S motility relies on retraction of typeIV pili (T4P) upon interaction with amine-containing polysaccharidesin the extracellular matrix (11, 27, 38, 46). T4P are also requiredfor cell surface adhesion and "clumping." It is unclear howprominent T4P are in stable microbial communities and what theirpossible role might be. Also, it is unknown how bacterial communitymembers interact with one another and what mechanisms they employfor effective communication and signal and/or material transferin such dense communities.

Here, we report faithful preservation of M. xanthus biofilmsand the first macromolecular resolution-based insights intotheir 3D architecture, obtained by electron tomographic imagingof rapidly frozen, freeze-substituted, and resin-embedded biofilmsections. The tomograms reveal the 3D organization of the bacterialchromosome, direct cell-cell and cell-substrate interactions,the presence of extracellular filaments (including but not restrictedto T4P), and abundant extracellular vesicles that, interestingly,we found to be tethered to each other and to the bacterial outermembrane.

MATERIALS AND METHODS

Top

MATERIALS AND METHODS

Ten milliliters of CYE broth, consisting of 5 mM MOPS (morpholinepropanesulfonicacid) (pH 7.6), 2 mM MgSO₄, 0.5% (wt/wt) Bacto Casitone, and0.25% (wt/wt) Bacto yeast extract (5), was inoculated with Myxococcusxanthus DK1622 (wild type) and grown for 24 h at 32°C and100 rpm. Four milliliters of this culture was used to inoculate200 ml of CYE broth, which was incubated for 18 h at 32°Cat 100 rpm. The culture was concentrated to a final opticaldensity at 600 nm of 10. Cellulose microdialysis hollow fibers(Spectra/Por; Spectrumlabs, CA) were cut into 1-mm-long pieces,autoclaved together with 1- by 1-mm² squares of Isopore membranefilters (HTTP, 0.4 µm; Millipore, MA), and subsequentlyplaced on CYE agar. Ten microliters of concentrated culture(optical density at 600 nm of 10) was placed on top of eachSpectra/Por fiber/Isopore filter or Isopore membranes and incubatedfor 6 h at 32°C followed by 20 h at room temperature. Inthe case of cellulose tubes, biofilms were grown by placingthe $~$ 200-µm-thick cellulose microdialysis tube with theMyxococcus culture on top of a moist agar plate prior to overnightshipment. Using these culturing conditions, the biofilms growingon Isopore filters were noticeably thicker, typically 4 to 10cell layers thick, whereas the biofilms grown on cellulose tubeswere thin, sometimes as thin as 2 cell layers.

High-pressure freezing and freeze-substitution. Bacterial biofilms on microdialysis tubing or Isopore membranefilters were briefly immersed in 10% glycerol or 20% bovineserum albumin in CYE medium. Filters or tubing was transferredinto 200-µm-deep type A aluminum planchettes that weresandwiched against the flat sides of type B planchettes. Thespecimens were cryofixed in a Bal-Tec HPM010 high-pressure freezer(2,100 bars, 5 to 7 milliseconds) (Bal-Tec, Inc., Carlsbad,CA). Using the Leica automated freeze-substitution system AFS(Leica Microsystems, Vienna, Austria), cryofixed specimens werefreeze-substituted in anhydrous acetone containing 1% osmiumtetroxide and 0.1% uranyl acetate and infiltrated with Epon-Aralditefollowing established protocols (34). Specimens were flat-embeddedbetween two microscopy slides and polymerized at 60°C over1 to 2 days (35). Resin-embedded biofilm samples were remountedunder a dissecting microscope for precise orientation.

2D projection transmission electron microscopy for sample surveying. Thin (70- to 100-nm) sections were collected on Formvar-coatedslot grids and poststained with 2% uranyl acetate in 70% methanolfollowed by either Reynold's or Sato's lead citrate. The sectionswere imaged in an FEI Tecnai 12 transmission electron microscope(FEI, Eindhoven, The Netherlands) or a Zeiss 10 instrument,operated at 120 kV or 100 kV, respectively. These samples wereused for ultrastructural evaluation as well as to assess thequality of fixation for the tomography studies described below.

Electron microscopy tomography. For tomography, thin sections (<100 nm) were imaged in aTecnai T20 LaB6 instrument operated at 200 kV (UCSF), and semithick(200- to 300-nm) sections were imaged in an FEI Tecnai F30 microscopeoperating at 300 kV (Boulder Laboratory for 3D Electron Microscopyof Cells, University of Colorado). Binned 2k by 2k tilt serieswere collected on 4k by 4k charge-coupled-device Gatan camera(Gatan, Inc., Pleasanton, CA) every 1° from –70°to +70°, using the UCSF tomography or SerialEM softwarepackage (32). The nominal setting of defocus was –0.2µm to –1 µm, and the pixel size of the datacorresponded to 1 nm. Series were aligned with the help of either10-nm or 15-nm gold fiducial markers (BBI Research, Inc., WI),and 3D reconstructed using the IMOD software package (24). IMODwas also used for data inspection and analysis. Where appropriate,3D volume data were inspected interactively using the volumeand surface rendering tools VOLUME ROVER (1) and CHIMERA (39;http://www.cgl.ucsf.edu/chimera).

RESULTS AND DISCUSSION

Top

RESULTS AND DISCUSSION

Structural organization of Myxococcus xanthus biofilms. To facilitate specimen handling for freezing and ultrastructuralanalysis, Myxococcus xanthus biofilms were grown on two typesof surfaces: Isopore membranes and cellulose tubes. In bothcases, the bacteria assumed a preferred orientation along thesurface, allowing us to select for either longitudinal (Fig.1A) or cross-section (Fig. 1B and C) views. The preferred alignmentof the bacterial cells along the surface that they adhered tomay be explained by elasticotaxis (44). The biofilms growingon Isopore filters were noticeably thicker, typically 4 to 10cell layers thick, whereas the biofilms grown on cellulose tubeswere thin, sometimes as thin as 2 cell layers, and more denselypacked. Biofilms on both supports were exquisitely preservedby the high-pressure freezing and freeze-substitution procedures.

View larger version (94K):
[in this window]
[in a new window]

FIG. 1. Ultrastructural characterization of high-pressure-frozen, freeze-substituted Myxococcus xanthus biofilms. (A) 2D projection images of a longitudinal section of thick and sparsely populated biofilm grown on an Isopore filter, showing intracellular features and vesicles, some of which appear to enclose cargo. (B and C) Cross-sections through a thin, tightly packed biofilm resting on a cellulose substrate. Notice how the bacteria accommodate each other by adjusting their shapes. Excellent preservation is manifest in the absence of extraction artifacts in both cellular contents and extracellular material. The biofilm boundary at the air-extracellular material interface is denoted by multiple arrows. Bar, 500 nm.

We imaged the same areas of M. xanthus biofilms by 2D projectionand by 3D tomography, and the side-by-side comparison seen inFig. 2 illustrates the radical improvement of resolution inthe z axis that is provided by 3D tomography. The biofilms grownon Isopore filters were sectioned longitudinally (Fig. 2A and B),and thus the structures of both the cytoplasm and the chromatinbody are well defined. In this case, the bacteria are more sparselypopulated than on the cellulose tubes, and therefore vesiclesin the extracellular space were easily visualized. The denselypacked biofilm grown on cellulose was visualized in cross-section(Fig. 2C and D); cytoplasmic elements were well resolved, andcell-cell junctions between closely apposed cells were observed.

View larger version (209K):
[in this window]
[in a new window]

FIG. 2. 3D electron tomographic reconstruction of Myxococcus xanthus biofilms. (A and B) A 250-nm thick section of multicell layered biofilm grown on an Isopore filter. (C and D) A 100-nm thin section of 4- to 10-cell layered biofilm grown on cellulose. Shown here are 2D projection views (A and C) and corresponding slices through tomograms (B and D) at $~$ 1-nm thickness. Cellular features obscured by superimposition in two dimensions are resolved in the third dimension in the tomogram, allowing 3D visualization and feature extraction. Asterisks denote chromosomal DNA, and arrows point out the biofilm boundary. Bar, 500 nm.

Organization of bacterial chromosomes. At the center of the bacteria we detected high-contrast multistrandedfilamentous densities resembling a twisted closed circle thatoften showed prominent clefts along the longitudinal axis (Fig.3A to C). We interpret the entity of these filamentous densitiesas the bacterial chromosomal DNA. Interestingly, and particularlyobvious when viewed in cross-section, we typically found thechromosomal DNA at the center of the cell, independent of whetheror not the cell is in intimate lateral contact with its neighboringcell. Therefore, even in such close contacts (see below), thereis no structural evidence of chromosomal DNA exchange, althoughthe exchange of small portions cannot be excluded. In the reconstructedvolumes, strands that we consider to be DNA were $~$ 3 to 5 nm thick.

View larger version (202K):
[in this window]
[in a new window]

FIG. 3. 3D reconstruction of intracellular features. (A to C) Chromosomal DNA is of high contrast and easily detectable. It was typically extended along the cellular axis and coiled along the central axis. (D) Crescent-shaped membranous structure. (E) Concentric-ring-shaped organelle. In both panels D and E, the intermembrane space is threaded by links of unknown nature. Bar, 100 nm.

Intracellular structures in M. xanthus biofilms. Granular and smooth cytoplasmic inclusions were observed incross-sections (Fig. 1B and C and 2C and D) and presumably representpoly-β-hydroxybutyric acid storage, although other formsof energy or lipid storage cannot be ruled out. Interestingly,in addition to these putative poly-β-hydroxybutyric storagegranules, we observed two types of intracellular organelle-likefeatures. While detectable in 2D projections, the characteristicshapes of those intracellular structures became obvious onlyby 3D tomographic reconstruction. These membrane-delineatedfeatures resemble organelles and are either of crescent (Fig.3D) or concentric spherical (Fig. 3E) shape. The intermembranespace in the crescent-shaped compartment, as well as the sphere-likeorganelle, is bridged by filaments that, given their stainingproperties, are likely to be proteinaceous in nature. Interestingly,ribosomes or similarly sized macromolecules are excluded fromthe cytoplasmic space enclosed by the rings. The relationshipbetween the sphere-like and the crescent-like features remainselusive, but the threaded intermembrane space is a common feature.

Cell-substrate and cell-cell cohesion. In densely populated thin-layer biofilms, cells were in intimatecontact with one another. Large areas of cell-cell contactswere already visible in 2D projections (Fig. 1B and C and 2C)but were studied in greater detail by electron tomography (Fig.2D). 2D projection views suggest a zipper-like structure thatconnects the outer membranes at short segments or along almostthe entire cell-cell contact surface (Fig. 4A and B). Betweenadjacent bacterial outer membranes we noticed a small but noticeablegap throughout the adhesion zone bridged by macromolecules witha size and stain distribution suggestive of a proteinaceousnature, indicating that this direct cell-cell interaction ismediated by outer membrane protein contacts. We often founddistinct deformations of the cell shape to allow for extensivecell-cell contact (Fig. 1B and C), suggesting that these contactsmark distinct cell-cell interactions and are not simply a resultof close packing of bacteria (for a review of bacterial shapes,see reference 50). In the tightly packed biofilms, EPS and fimbriaedo not appear to be involved in these cell-cell contacts. Instead,these cell-cell contacts show ultrastructural similarities tothe previously described "conjugative" (40) and "conjugational"(10) junctions described for Escherichia coli. Intimate bacterialinteractions have also been seen in M. xanthus cell pairs andmonolayers studied by atomic force microscopy, although thesestudies did not allow the characterization of such interactions(38). Conceivably, these cell-cell junctions play a role inthe swarming phenomenon central to its predatory characteristics.It is noteworthy that despite the fact that the outer membranesare in close proximity, we have not observed direct outer membranefusion, which was postulated by Nudleman et al. to account forthe direct cell-to-cell transfer of outer membrane proteins(36). While we cannot exclude that such fusion events can occur,we propose that other means of outer membrane protein transfermay exist, e.g., via vesicles, as will be discussed in moredetail below.

View larger version (220K):
[in this window]
[in a new window]

FIG. 4. 3D reconstruction of extracellular elements in biofilm organization. (A and B) In the thin and highly packed biofilms (Fig. 1B and C), cell-cell contacts are characterized by the membranes in close contact stitched to one another by extracellular linkages, in a bond strong enough to overcome the electrostatic repulsion of like surfaces. Interbacterial outer membrane fusion events were not observed. (C) At the air-extracellular material interface, there was a highly osmicated, single-track film along the entire surface of the biofilm (denoted by arrows), presumably protecting it from desiccation. (D) The polar T4P (asterisks) were easily visualized in tomograms, whereas they were not obvious from projection images. Bar, 100 nm.

Sharp biofilm boundary. 2D projection images of the cross-section views of biofilmson cellulose tubes revealed a sharp discontinuity at the edgeof the biofilm. When visualized by electron microscopic tomography,this biofilm border is a continuous high-contrast, sheet-likefeature that appears to demarcate the entire biofilm-air interface(Fig. 1B and C and 4C). Virtually every available space withinthis biofilm border is occupied by vesicles that are 30 to 60nm in diameter (Fig. 1B and C), while the space outside theborder is devoid of these vesicles. The boundary structure isa thin ( $~$ 5-nm) sheet of osmophilic material (Fig. 4C) that resemblesin dimension and appearance a lipid bilayer. The affinity forosmium and its resulting electron density make it unlikely thatthis boundary layer consists of secreted carbohydrates. Whilewe cannot exclude that this boundary represents aggregationof denatured extracellular proteins, it is uniform in thicknessand shows a striking resemblance to the lipid bilayer of thebacterial membranes in this preparation. Given the number ofsecreted vesicles (discussed below) and the fact that lipidsnaturally accumulate at an air-water interface (13, 14, 47),we propose that the sharp biofilm boundary consists of a lipidlayer that protects the biofilm from desiccation.

Visualization of extracellular filaments: T4P. Extracellular filamentous materials were not easily detectablein 2D projections, but a set of smoothly curved filaments, $~$ 4to 6 nm in diameter and micrometers in length, became obviousin 3D reconstructions (Fig. 4D). We interpret these filamentsto be T4P, because their dimensions and the structure of theirarea of origin in the cell are consistent with those seen inother ultrastructural and atomic force microscopy studies (38).These filaments protrude from a ribosome-devoid polar regionof M. xanthus (Fig. 4D). In serial tomograms we could tracea set of T4P filaments more than 1 µm into the extracellularspace. TFP are known to mediate cell-substrate interactionsand are crucial for S motility in M. xanthus and other bacteriathat display gliding motility (18). The study presented heredid not allow us to assess the motility of individual bacteriain the biofilm, but it is interesting to notice that the ratioof filament bundles, which are found to extend for more than1 µm throughout the biofilm, is low compared to the numberof cells present in such volumes, suggesting that only a fewcells assemble T4P. In conclusion, the role of T4P in biofilmsremains unclear. If the T4P bundle is used for S motility, onewould expect the filament to end on a defined substrate or anothernearby cell, several of which were well within reach. Also,if its role is to propel the cell forward, one would expectthe T4P to be straightened under tension. Instead, the filamentswe observed in biofilms are smoothly curved and are even foundto assume a direction perpendicular to the original axis. Thescarcity of assembled T4P in the biofilm and the morphologyof visualized T4P suggest that the role of T4P in a biofilmmay be different from their well-documented role in S motility.

Possible roles of extracellular vesicles. The extracellular material in biofilms is traditionally thoughtof as being filled with EPS, which serves as an enveloping medium.EPS has been found to be essential in M. xanthus communities(27, 28). However, EPS is a term that is loosely used to describeall extracellular material, independent of its actual chemicalcomposition. In a proteomic study of M. xanthus biofilms, theextracellular material was found to be enriched in proteinsof novel function, but putative functions could be assignedfor only 5 of the 21 proteins identified (9). Apart from playinga structural role, the extracellular matrix is also likely tomediate interbacterial signaling and communication. Interestingly,quorum signaling involving 2-heptyl-3-hydroxy-4-quinolone waslinked to vesicle formation in P. aeruginosa (29). Extracellularmembrane vesicles have been observed in several planktonic gram-negativebacteria and their biofilm communities (2, 4, 33).

In both thin and thick Myxococcus biofilms, the extracellularspace is filled with vesicles with a diameter of 30 to 60 nm(Fig. 5). In the thin biofilms, almost every conceivable spaceis filled with vesicles, including the space underneath thesharp biofilm boundary (Fig. 1B and C). It should be emphasizedthat in our cryo-immobilized samples we did not observe signsof cellular stress and cell disintegration. The membranes aresmooth, unlike what is often found in chemically fixed samples,where membranes display blebbing and other artifacts. Wherebacteria were spread at lower density, such as in the thickbiofilms, we typically found the vesicles in direct vicinityof bacterial membranes. 3D reconstruction revealed that vesicleswere either directly in contact with bacterial membranes orlinked to cell membranes by what appear to be proteinaceoustethers (Fig. 5B, D, and E). Occasionally, a vesicle was seenin direct contact with three bacterial cell membranes (Fig.5C). Vesicles were found tethered not only to the bacterialouter membrane surface but also to each other (Fig. 5D and E).The tethers appear as short filaments of about 3 nm in diameterand tens of nanometers in length. On the surface of a numberof vesicles we find protrusions that are likely to be integralmembrane proteins. The majority of vesicles display an internalstain distribution that suggests either a protein cargo or somekind of internal organization.

View larger version (135K):
[in this window]
[in a new window]

FIG. 5. The biofilm extracellular space is filled with vesicles. The vesicles are tethered to the bacterial membrane, and to each other, and are arranged linearly in a nonrandom fashion. (A) The indentation in the bacterial membrane may suggest either an endo- or exocytosis event. (A, B, and D) The vesicle interior displays densities that are indicative of either the presence of a cargo or an underlying protein organization. The vesicle is tethered to the membrane by at least one or two links of unknown nature. (C) Occasionally, vesicles were found in simultaneous contact with three bacterial outer membranes. (E) Five vesicles shown in linear arrangement close to the bacterial cell surface. Links between the vesicles and between vesicles and the bacterial surface were observed. Bar, 100 nm.

Membrane vesicles have been shown before by electron microscopystudies (2, 4, 12, 23), where they were pinched off from thesurface and released into the surrounding medium. However, insome of these studies, membrane blebbing induced by chemicalfixation artifacts could not be ruled out. Our results cannotbe attributed to fixation artifacts, since high-pressure freezingoccurs instantaneously (within milliseconds). Moreover, membranevesicles were observed in high-pressure-frozen P. aeruginosabiofilms (42), although at a much lower density, and the intervesiculartethers were not described in that study. While a variety ofpossible functions for these extracellular vesicles have beenproposed (29, 31, 42), the exact role of such vesicles remainselusive. We propose that these vesicles serve as a vehicle forthe interbacterial exchange of outer membrane proteins. Exchangeof bacterial outer membrane proteins in the absence of genetransfer was shown by Nudleman et al. (36), who attributed theexchange to transient fusion of the outer membranes.

In this study we did not detect outer membrane fusion; instead,we observe a wealth of vesicles that are in intimate contactwith the outer membranes of one or several bacteria, suggestingvesicles as the most likely candidate for intercellular proteintransfer. A vesicle-mediated mechanism of protein transfer,however, implies that bacteria must have a sophisticated mechanismnot only for vesicle secretion but also for the uptake of thevesicles. We suggest that the protrusions on the vesicle surfacerepresent receptor molecules that are important in vesicle uptake.

Clearly, vesicles are abundant and are likely to play a rolein M. xanthus biofilm physiology. These observations contributeto the foundation of a new field of study, calling for compositionalanalysis and time-lapse high-resolution microscopy studies,to solve the mystery of bacterial vesicle function, formation,and fusion (30, 31).

Concluding remarks. Microbial biofilms are ubiquitous in all nutrient-sufficientecosystems (7), and functional mature communities develop byprocesses similar to those that produce multicellular eukaryoticorganisms (37). The well-established methods of chemical fixationand 2D projections of thin sections have produced a rich literatureconcerning the ultrastructure of individual bacterial cells,but transmission electron microscopy studies of intact biofilmcommunities are relatively rare. Advancements in high-pressurefreezing and freeze-substitution methods have allowed us tominimize fixation artifacts, or to avoid them altogether, anddevelopments in tomography allow improved resolution in thez axis. In this study we have combined these techniques to examine,for the first time, two different types of biofilm communitiesproduced by the same organism (M. xanthus) on two differentsurfaces. In this exploratory study we have visualized cellularstructures (e.g., chromatin and cytoplasmic organelles), andwe have resolved other structures that may enable interactionsbetween the cells that comprise biofilm communities. These includecell-cell "zippers," pili, and vesicles that appear to be tetheredto each other and to cells and to contain resolvable "cargo."

ACKNOWLEDGMENTS

We thank Reena Zalpuri of the UC Berkeley Electron MicroscopyLaboratory; the Boulder Laboratory for 3D Electron Microscopyof Cells (BL3DEMC), University of Colorado; and Michael Braunfeldand David Agard at UCSF.

This project was supported by the Director, Office of Science,of the U.S. Department of Energy under contract DE-AC03-76SF00098to M. Auer; by National Institutes of Health grant GM54666 toW. Shi; and by an Amado Foundation grant to J. W. Costerton.The BL3DEMC is supported by a grant from the National Centerfor Research Resources of the National Institutes of Health(RR-00592) to A. Hoenger.

FOOTNOTES

* Corresponding author. Mailing address: Lawrence Berkeley National Laboratory, 1 Cyclotron Road, Berkeley, CA 94720. Phone: (510) 486-7702. Fax: (510) 486-6488. E-mail: MAuer{at}lbl.gov

Published ahead of print on 23 January 2009.

This paper is dedicated to the memory of Terry Beveridge, whoserecent and untimely death robbed this field of one of its mostperceptive and competent researchers.

REFERENCES

Top