Three-Dimensional Structure of the Ultraoligotrophic Marine Bacterium “Candidatus Pelagibacter ubique”

ABSTRACT

SAR11 bacteria are small, heterotrophic, marine alphaproteobacteria found throughout the oceans. They thrive at the low nutrient concentrations typical of open ocean conditions, although the adaptations required for life under those conditions are not well understood. To illuminate this issue, we used cryo-electron tomography to study “Candidatus Pelagibacter ubique” strain HTCC1062, a member of the SAR11 clade. Our results revealed its cellular dimensions and details of its intracellular organization. Frozen-hydrated cells, which were preserved in a life-like state, had an average cell volume (enclosed by the outer membrane) of 0.037 ± 0.011 μm³. Strikingly, the periplasmic space occupied ∼20% to 50% of the total cell volume in log-phase cells and ∼50% to 70% in stationary-phase cells. The nucleoid occupied the convex side of the crescent-shaped cells and the ribosomes predominantly occupied the concave side, at a relatively high concentration of 10,000 to 12,000 ribosomes/μm³. Outer membrane pore complexes, likely composed of PilQ, were frequently observed in both log-phase and stationary-phase cells. Long filaments, most likely type IV pili, were found on dividing cells. The physical dimensions, intracellular organization, and morphological changes throughout the life cycle of “Ca. Pelagibacter ubique” provide structural insights into the functional adaptions of these oligotrophic ultramicrobacteria to their habitat.

IMPORTANCE Bacterioplankton of the SAR11 clade (Pelagibacterales) are of interest because of their global biogeochemical significance and because they appear to have been molded by unusual evolutionary circumstances that favor simplicity and efficiency. They have adapted to an ecosystem in which nutrient concentrations are near the extreme limits at which transport systems can function adequately, and they have evolved streamlined genomes to execute only functions essential for life. However, little is known about the actual size limitations and cellular features of living oligotrophic ultramicrobacteria. In this study, we have used cryo-electron tomography to obtain accurate physical information about the cellular architecture of “Candidatus Pelagibacter ubique,” the first cultivated member of the SAR11 clade. These results provide foundational information for answering questions about the cell architecture and functions of these ultrasmall oligotrophic bacteria.

INTRODUCTION

The SAR11 clade of heterotrophic marine alphaproteobacteria, which is now widely accepted as the most successful clade of organisms on Earth, includes the bacterium “Candidatus Pelagibacter ubique” (herein referred to as Pelagibacter). SAR11 members are distributed throughout the global oceans and account for about 25% of all plankton cells (1, 2). These organisms survive at levels of nutrition at which most would starve, and they are thought to play a major role in carbon cycling on Earth (3). They are among the smallest free-living cells known (2) and have among the smallest genomes found in autonomously replicating cells (4, 5), probably as a result of genomic streamlining (6). Nonetheless, Pelagibacter genomes appear to encode nearly all of the basic functions of alphaproteobacterial cells, which enables the organisms to grow in low-nutrient ocean systems. The abilities of SAR11 cells to compete for nutrients and to replicate efficiently probably explain their successful colonization of the oceans. The Pelagibacter strain used in this study, HTCC1062, is typical of SAR11 organisms that occupy the surface waters of temperate and polar oceans.

Oligotrophic marine bacteria occupy a special place in the literature regarding small cells because natural populations of marine bacteria, particularly those from ocean gyres where nutrients are present in extremely low abundance, often approach the theoretical limits for the smallest cells that can be imagined, based on fundamental biological considerations. The term “ultramicrobacteria” was coined to describe the size of such bacterial cells (<0.1 μm³), which are abundant (0.2 × 10⁶ to 2 × 10⁶ cells/ml) in aquatic ecosystems (7). Theories explaining the small size of these organisms usually argue either that the organisms are small because they are starved or that small size is selectively favored in very-low-nutrient ecosystems. Accumulating data have demonstrated that bacterioplankton spend much of their time growing and are active in geochemical cycles (8), although undoubtedly such cells endure periods when nutrient limitation shuts down replication (9). The alternative explanation, i.e., that small size can make cells more competitive, assumes either that the reduced material costs of replicating drive minimization (6) or that high surface-to-volume ratios make cells more competitive for nutrients (10). For example, in some conceptual models, surface-to-volume relationships are recognized as potentially important factors influencing the balance between nutrient transport capacity and biomass (11). In one model, large cells are not competitive in dilute ecosystems because they waste energy producing metabolic enzymes that will never be pushed to process substrates at close to their turnover capacity (10). Other studies have examined the relationships between the number of periplasmic substrate binding proteins and whole-cell kinetics, yielding the conclusion that a higher ratio of substrate binding proteins to cytoplasmic proteins provides cells with kinetic properties that are favorable at very low nutrient concentrations (12). Pelagibacterales are the most abundant planktonic organisms in the oceans and were reported as the smallest free-living cells (0.01 μm³) when they were first cultivated in a laboratory (2). Subsequent fluorescence microscopy measurements indicated that SAR11 bacteria from different oceanic regions were typically as large as, if not larger than, other prokaryotes (3). Modeling efforts aimed at a better understanding of the functional adaptions of oligotrophic ultramicrobacteria will benefit from accurate data on cell dimensions.

Cryo-electron tomography (cryo-ET) has been successfully used for in situ structural analysis of both prokaryotic and eukaryotic cells, because of its ability to reveal a life-like, frozen-hydrated state with resolution suitable for determining the macromolecular organization of intact cells (13–17). As a specimen is tilted in the electron beam, a set of two-dimensional (2D) projection images can be collected and used to reconstruct a three-dimensional (3D) image of the object, allowing 3D modeling. In addition, rapid freezing (which occurs within ∼1 ms) preserves the cells in a nearly native state without chemical fixation or staining, which allows accurate measurement and examination of the cell dimensions and internal structures. In the past few years, cryo-ET has yielded great insights into the internal organization of many commonly studied Gram-negative bacteria (18–20), archaebacteria (21), cyanobacteria (22), mycobacteria and mycoplasmae (23), viruses (24, 25), and bacterial phages (26) and, more recently, the cellular architecture of eukaryotes (27, 28).

In this study, we used cryo-ET to visualize the 3D structure of frozen-hydrated Pelagibacter cells. The cryopreserved cells were considerably larger than the glutaraldehyde-fixed specimens of the same strain that were studied previously (2). However, our images revealed very small cells with a proportionately large periplasmic space, an eccentrically located nucleoid occupying nearly one-half of the cytoplasm, and a high ribosome concentration in relation to the Pelagibacter growth rate, which is less than 1 division per day. We determined the cell length distribution in late-log-phase and stationary-phase cell cultures and found hints of asymmetric cell division. In addition, we found an outer membrane (OM) pore complex that resembled PilQ secretin and cell-cycle-specific formation of extracellular filaments, most likely type IV pili.

RESULTS

Cell morphology and dimensions of Pelagibacter.

Tomographic reconstructions of cells frozen from culture medium clearly revealed the overall morphology and subcellular organization of Pelagibacter. Most cells frozen at the log phase were crescent-shaped (Fig. 1A and B, 2A, and 3A to E), and only the shortest were relatively straight or oval (Fig. 3F). The lengths of cells from log-phase cultures varied from 400 to 1,300 nm (Fig. 1E). The cell diameters were less variable, however, ranging from 190 to 280 nm (among a total of 36 reconstructed cells from log-phase cultures). These data are consistent with the model that cells approximately double in length and then divide by fission (29). Cryo-ET also revealed a distinct subcellular organization in Pelagibacter. The cell interior appeared as two major zones, i.e., the nucleoid and a ribosome-containing area (Fig. 1A). Cells from stationary-phase cultures tended to be more spherical (Fig. 1C and D), with the cell lengths ranging from 600 to 900 nm with only one very long exception (1,750 nm) (see Fig. 5E), which appeared to be a cell that failed to complete division but continued to grow. The relative volume and disposition of the subcellular compartments were different in stationary-phase cells; the cells were wider, the cytoplasm was smaller, and the volume of the periplasm was increased (Fig. 1C to E).

FIG 1.

Cryo-electron tomography of intact frozen-hydrated Pelagibacter cells. (A) A 15-nm-thick tomographic slice of a Pelagibacter cell grown in the log phase. Note the uniform periplasmic space between the outer membrane (OM) and the inner membrane (IM). The region occupied by the nucleoid (N) shows a filamentous structure and lacks larger ribosome-like complexes (R). (B) Surface-rendered 3D model of the Pelagibacter cell shown in panel A. Note that the model was sliced open to allow viewing of the cell inside. Modeled structures include the OM (dark blue), IM (cyan), peptidoglycan layer (white), and cytoplasm (orange). (C) A 15-nm-thick tomographic slice of a Pelagibacter cell grown in the stationary phase. The cell morphology is more spherical and the periplasmic space is dramatically increased, compared to log-phase cells. (D) Surface-rendered 3D model of the Pelagibacter cell shown in panel C. Note that the model was sliced open to allow viewing of the cell inside. (E) Distribution of cell length and cell diameter values among all 50 reconstructed Pelagibacter cells. Of the studied cells, 36 were from cultures in the log phase and 14 from cultures in the stationary phase. Scale bar, 100 nm (A; also valid for panel C).

FIG 2.

Characterization of the periplasmic space of Pelagibacter. (A) Intact Pelagibacter cell grown in the log phase. (B and C) Magnified images of the two periplasmic regions outlined in panel A. Along the long axis of the cells, the thickness of the periplasmic space is about 18 nm at the cell body (B). Note that two or three layers of peptidoglycan (PG) can be observed in many regions. The diameter of the periplasmic space is increased to 39 nm at the cell poles, and the PG layers become less organized (C). (D) Protuberance of the OM, resembling a vesicle. Note the tether-like structure (white arrowhead) in the periplasmic space. (E) Enlarged periplasmic space with many particles at the cell pole. The black arrowheads in panels B, C, and E indicate the measured width of the periplasmic space. Scale bars, 100 nm (A) and 50 nm (B, C, D, and E).

FIG 3.

Characterization of the eccentrically located nucleoid, suggesting polarity along the short cell axis. (A to F) Representative Pelagibacter cells cultured in the log phase. The nucleoid (N) regions (outlined by red dashed lines) are always on the convex side of the cells, whereas the cytoplasm, containing ribosome-sized particles (R), is predominantly on the concave side of the cells. (G and H) 3D modeled OM (dark blue), IM (cyan), nucleoid (red), and ribosome-like particles (yellow) in the cells shown in panels E and F, respectively. Scale bars, 100 nm (A; also valid for B, C, and D) and 100 nm (E; also valid for F, G, and H).

FIG 5.

Dividing Pelagibacter cells and pili. (A) Tomographic slice showing asymmetric cell division. (B) Magnification of the region outlined in panel A. Note the still-present intercellular connection (arrow). (C) Tomographic slice of a dividing cell with two type IV pili that attached to the carbon support film. The blue arrowhead indicates constriction of the membrane. (D) Magnification of the region outlined in panel C. The inner membrane (IM) (cyan) and outer membrane (OM) (dark blue) are indicated by transparent lines. The pili appear to originate from the IM. The black arrowheads indicate the measured pilus diameter. (E) Tomographic slice of a dividing cell with many pili originating from the dividing site. (F) Magnification of the division site with pili that was outlined in panel E. (G) 3D model of the dividing cell shown in panel E. The pili (magenta), OM (dark blue), IM (cyan), nucleoid (pink), and ribosome-like particles (yellow) are shown. Scale bars, 100 nm (A, C, and F), 20 nm (B and D), and 200 nm (E).

The membranes of these Gram-negative cells had a smooth appearance and were well preserved in our frozen-hydrated specimens (Fig. 1A to D). The reconstructed cells from log-phase cultures contained total cell volumes (enclosed by the OM) ranging from 0.015 μm³ to 0.058 μm³, depending on the cell length, and an average volume of 0.037 ± 0.011 μm³ (Table 1; see also Table S1 in the supplemental material). The cell volume of Pelagibacter was reported previously as about 0.01 μm³, based on measurements of negatively stained and glutaraldehyde-fixed specimens (2); the difference between our measurements and the published results was likely due to shrinkage of samples previously prepared by chemical fixation. The cell volumes of SAR11 bacteria collected from different regions of the ocean ranged between 0.031 and 0.051 μm³ when viewed by DNA fluorescence (3), which was consistent in part with our measurements and confirmed that SAR11 cells can grow somewhat larger than thought previously.

TABLE 1.

Volume characteristics of log-phase and stationary-phase cells

Parameter	Log phase (30 cells)			Stationary phase (5 cells)
	Whole-cell volume (μm3)	Cytoplasm volume (μm3)	Periplasm proportion (%)	Whole-cell volume (μm3)	Cytoplasm volume (μm3)	Periplasm proportion (%)
Minimum	0.015	0.007	21	0.026	0.011	47
Maximum	0.058	0.039	50	0.040	0.019	73
Mean	0.037	0.024	36.7	0.035	0.015	56.2
Standard deviation	0.011	0.008	7.0	0.007	0.003	10.8

Pronounced periplasm.

The whole-cell volumes were similar for log-phase cells (n = 30) and stationary-phase cells (n = 5), but the volume of the cytoplasm was reduced in stationary-phase cells, relative to an increased periplasmic space (Table 1). The periplasmic space of Pelagibacter was relatively large, occupying 20% to 50% of the total cell volume for log-phase cells and 50% to 70% for stationary-phase cells (Table 1; see also Table S1). Especially in log-phase cells, the periplasmic width was relatively uniform along the side of the cell body, with an average diameter of 17 ± 3 nm (Fig. 2A and Table S2), which was similar to that of Escherichia coli (30). At the cell poles, however, the periplasmic width became larger and more variable. The average periplasmic width was 48 ± 36 nm at the poles, but the width increased to >100 nm in some cells (Fig. 2E and Table S2). Although we observed one or two layers of density in the periplasm, most likely representing peptidoglycan (Fig. 2B and C) and scattered particles (Fig. 2E), the matrix of the periplasm was more translucent than the cytoplasm, suggesting a lower density of macromolecules. Occasionally, we found small vesicles attached to the extracellular surface of the OM (Fig. 2D). Sometimes, a tether-like structure originated from the inner membrane (IM) and connected to the vesicle (white arrowhead in Fig. 2D).

Eccentrically localized nucleoid.

Although the cytoplasm was considerably more crowded than the periplasm, cryo-ET revealed a distinct subcellular organization. The cell interior contained many electron-dense particles with diameters of ∼20 nm; their shapes and sizes were consistent with their identification as ribosomes (Fig. 3A to F) (31). The nucleoid, which was identified by its fibrous texture and lack of ribosome-like particles, was located eccentrically at the convex side of the cell (Fig. 3A to E). Even in very short cells that displayed little curvature, presumably the products of recent cell division, the nucleoid and ribosomes were spatially segregated at opposite sides of the cell, revealing an asymmetry along one of the short cell axes (Fig. 3F). This is in contrast to the centrally located nucleoid and peripherally located ribosomes observed in many other bacterial species (18, 19, 32). Three-dimensional modeling of nucleoids and ribosome-like particles indicated that these subcellular structures were compartmentalized into two distinct zones, each taking up roughly one-half of the volume enclosed by the IM (Fig. 3G and H).

Nucleoid volume.

Some bacterioplankton have evolved small genomes, making them some of the simplest free- living cells known (33, 34). One of the fundamental constraints on cell size is the physical volume occupied by the genome. Therefore, we closely examined the nucleoid of Pelagibacter to better understand how space is partitioned within the cytoplasm. We compared the nucleoid measurements obtained from cryo-ET (∼0.012 μm³, estimated as one-half of the cytoplasm volume) (Table 1) to the estimates of the minimal hydrated volume of the 1.3-Mb genome of Pelagibacter as follows. The diameter of B-DNA (with some water bound) is ∼25 Å; if we treat DNA as a cylinder, then its volume (V) is V = πr²L, where r is 12.5 Å and the length of the DNA (L) is 3.4 Å × 1.3 × 10⁶ bp (the Pelagibacter genome has 1.3 × 10⁶ base pairs, and the distance between base pairs in the helix direction is 3.4 Å), which is equivalent to a volume of 0.0022 μm³. This estimated DNA volume is about one-fifth of the measured nucleoid volume of Pelagibacter. Apparently, the packing of DNA in this organism is not tight, as would be expected for transcriptionally active DNA.

Abundance of ribosome-like particles.

The fraction of the interior cell volume (surrounded by the IM) that was not occupied by the nucleoid was richly filled with electron-dense particles, which were previously interpreted as ribosomes on the basis of their size (∼20 nm), shape, and contrast (31). Based on this interpretation and a template matching method (35), we detected cellular concentrations of 10,000 to 12,000 ribosome-like particles/μm³. This observation suggests that ribosomes are relatively abundant in the Pelagibacter cytoplasm, considering its small volume and slow growth. The ribosome concentration of Pelagibacter is about twice as high as the concentration in slowly growing E. coli cells, although Pelagibacter grows significantly (30-fold) more slowly than E. coli (Table 2) (2, 36); however, our results are consistent with those reported in other cryo-ET studies of intact bacteria (19, 37).

TABLE 2.

Ribosome concentrations in various microorganisms

Organism	Growth phase	Growth rate (h−1)	Ribosome abundance (no./μm3)	Cell volume (μm3)	Reference
Escherichia coli	Rapidly growing	2.5	65,500	1.1	36
		1.5	23,900
	Slowly growing	0.6	6,200
Sphingomonas sp. RB2256	Mid-log	0.16	40,000	0.05–0.09	69
	Starvation	0	4,000
Campylobacter jejuni	Mid-log	0.05	25,000a	0.2	19
Rickettsia prowazekii	Mid-log	0.07	17,000	0.09	68
Spiroplasma melliferum	Mid-log	>0.1	15,000a	0.05–0.07	37
“Candidatus Pelagibacter ubique”	Mid-log	0.02	10,000–12,000a	0.015–0.058	This study
ARMAN	Uncultivated		3,000a	0.03	21
WWE30OP11-OD1	Uncultivated		4,700a	0.009	53

^aRibosome abundance was estimated using cryo-ET of cells.

Outer membrane pore complex.

Pore-forming channels were found in the OM of Pelagibacter cells, especially in stationary-phase cells, which could be due to enhanced visibility resulting from their enlarged periplasm (Fig. 4A to C). We extracted 58 subtomogram volumes containing OM pores from the reconstructions, aligned them in three dimensions, and generated an averaged pore structure (Fig. 4D). To evaluate the subunit numbers in the pore structure, rotational symmetries ranging from 3-fold to 15-fold were applied to the subtomogram average (Fig. S1). Apparent subunit densities could be observed in most averages after the application of symmetry, but they appeared especially strong in maps with multiples of C₃ and C₆ symmetry (Fig. S1). We further performed a rotational correlation analysis of the averaged native pore complex map, which showed peaks at 4-fold symmetry (or multiples of 4) (Fig. S1). Our analysis is consistent with previous studies reporting 12-fold symmetry for secretin pore complexes (38–44).

FIG 4.

Outer membrane pore complexes in Pelagibacter. (A) Tomographic slice of a Pelagibacter cell in the stationary phase. Note the two pore complexes in cross-sectional view (outlined by the dashed line) in the extended periplasmic space between the outer membrane (OM) and the inner membrane (IM). (B) Magnification of the pore complexes outlined in panel A. (C) Tomographic slice of a representative Pelagibacter cell with the pore complex shown in side view with possible gate density (arrow). (D) Slice of a subtomogram average of 58 OM pore complexes, showing the pore in cross-sectional view. (E) Detection of subunits of the pore complex (arrowheads) after application of 12-fold symmetry to the averaged pore structure shown in panel D. (F) Side view of the 12-fold-symmetrized pore complex. (G) Surface rendering of the averaged and symmetrized pore complex (blue) embedded in the outer membrane (green). (H) Comparison of secretin pore complexes. (Left to right) In vitro reconstituted GspD secretin pore from Vibrio cholerae (Electron Microscopy Data Bank accession no. EMD-1763), native in situ PilQ complexes in Thermus thermophilus (Electron Microscopy Data Bank accession no. EMD-3022) and Myxococcus xanthus (the secretin domain is highlighted in blue in the transparent in situ map), and an individual putative PilQ pore complex (same pore as in panel C) and the averaged and 12-fold-symmetrized PilQ pore complex from “Candidatus Pelagibacter ubique.” Scale bars, 100 nm (A), 20 nm (B and C), and 10 nm (D; also valid for E and F). The maps of M. xanthus PilQ are reproduced with permission (Grant Jensen, California Institute of Technology).

The pore complex has an overall width of ∼13 nm (inner pore diameter, ∼5 nm) and a length (perpendicular to the membrane) of ∼20 nm (Fig. 4E and F), resembling the pore structures of the type II secretion system or the nonpiliated state of the type IV pilus system (Fig. 4G and H) (42, 45). In most electron microscopy (EM) reconstructions of pore-forming, nonpiliated secretins (of the type II or type III secretion system or the type IV pilus system), an apparently closed periplasmic gate was observed (42, 45, 46). Although we could see hints of a gate density in the raw tomograms of individual pores in side view (Fig. 4C and H), our averaged cryo-ET map did not show a gate. This could be caused by a “missing wedge” artifact, because our subtomogram average was dominated by top-view particles (Fig. 4B), compared to side-view ones (Fig. 4C), making densities distributed parallel to the ice layer less visible. The genome of Pelagibacter contains genes for PilC, PilD, PilE, PilF, and PilQ of the type II secretion system and for type IV pilus biogenesis (4). Given our structural data and information from the Pelagibacter genome (4), the OM pore complex could be an oligomer of PilQ secretin.

Dividing cells and pili.

In our study, we also observed some dividing Pelagibacter cells. In one case, the cell had nearly completed the division but a small pinched-off area was still connecting the future daughter cells (Fig. 5A and B). From the different sizes of the two cells, we conclude that Pelagibacter organisms can undergo asymmetric cell division. Although the amount of information is too small to be statistically significant, asymmetric cell division is not uncommon among alphaproteobacteria (47). In two other cases, the cells appeared to have been frozen during the act of division (Fig. 5C to G), as suggested by constriction of the membrane (blue arrowhead in Fig. 5C). Interestingly, filaments ∼3 nm in diameter and up to ∼1 μm in length (Fig. 5D and E) were observed in those cells. The filaments extended predominantly from the cell division site into the extracellular space (Fig. 5C to G; see also Movie S1 in the supplemental material). Although their basal structures were not well resolved (Fig. 5D), the filaments seemed to originate from within the periplasmic space or from the IM. Based on the genomic sequence of Pelagibacter, which encodes the genes for type IV pilus biogenesis, including pilin PilA, OM pore PilQ, and ATPase PilT, we interpret these long filaments as type IV pili.

DISCUSSION

Very small cells have an almost irresistible philosophical allure because they embody all of the cellular features essential for life in what must approximate a minimal package. Various reviews and workshops have considered a theoretical lower limit for the size of autonomous cells from a physical and biochemical perspective, taking into account the lists of essential macromolecular components and their sizes, as deduced from comparative genomics (48, 49). On the basis of those studies, a minimal viable cell diameter of ∼250 nm (assuming a spherical cell) and a volume of 0.008 to 0.014 μm³ are thought to be required for life (48). The definition of free-living is important, because some Mycoplasma cells are smaller (∼0.013 μm³) (50) but their genomes encode far fewer proteins. Such cells lack cell walls and outer membranes, and they can replicate only in complex media, by importing many small molecules that they cannot synthesize (51). Thus, Mycoplasma organisms in nature are autonomously replicating but not free-living; they are always found in direct association with eukaryotic cells. In contrast, cells of the SAR11 clade grow freely suspended throughout the marine water column.

Our cryo-ET measurements determined the authentic size of Pelagibacter cells as 0.015 to 0.058 μm³, which is larger than some recently reported ultrasmall organisms; for example, marine actinobacteria have an average cell volume of ∼0.013 μm³, as analyzed by flow cytometry and microscopic fluorescence in situ hybridization (FISH) (52). Cells of a novel archaeon, namely, Archaeal Richmond Mine acidophilic nanoorganisms (ARMAN), have volumes of 0.009 to 0.04 μm³, as determined by cryo-ET (21). More recently, diverse uncultivated ultrasmall bacterial cells were found in groundwater; those cells consistently had small sizes of 0.004 to 0.013 μm³, as shown by cryo-ET, despite some morphological variations (53). However, since the genomes of those ultrasmall groundwater bacteria were smaller than the genomes typical of free-living cells, it was suggested that those organisms were at least partially dependent on other community members for basic metabolic building blocks.

By combining a small volume with a crescent shape, Pelagibacter achieves an even higher surface-to-volume ratio than would otherwise be possible. The mechanism for maintaining a clearly defined shape in bacteria is thought to depend on a bacterial cytoskeleton (54–57). In Caulobacter crescentus, two cytoskeletal proteins, MreB and CreS (crescentin), which are actin and intermediate filament (IF) homologues, respectively, are responsible for the characteristic crescent shape (58–60). MreB is required to determine a rod shape, and CreS is thought to induce cell curvature. The cytoskeleton biogenesis genes mreB, mreC, and mreD are found in the Pelagibacter genome. CreS is unique to Caulobacter and does not have homologues in other bacteria. However, bacterial IF-like proteins, including CreS, share a coiled-coil rod domain as a key structural feature. Genome comparisons among 26 arbitrarily selected bacteria, including Pelagibacter, identified at least one IF-like protein (based on the putative coiled-coil rod domain) encoded in the genomes of the overwhelming majority of the bacteria (61). The Pelagibacter genome encodes two uncharacterized proteins with putative coiled-coil rod domains, suggesting that Pelagibacter may use IF-like proteins to maintain its crescent shape. It should be noted that, to date, the expression of these genes in Pelagibacter has not been studied. Among the tomograms of Pelagibacter cells, we have seen little evidence of filamentous cytoskeleton structures; however, this does not rule out the possibility that cytoskeletal filaments exist in Pelagibacter, because recent cryo-ET characterizations of cytoskeletons in many bacteria indicated that no long filamentous structures representing MreB or CreS assemblies could be identified by cryo-ET (57, 62, 63). MreB and CreS perhaps are highly dynamic or exist in cells only as short segments, which are not recognizable in cellular cryotomograms.

An unusually large fraction of the Pelagibacter total cell volume is periplasmic space with many particles in addition to the peptidoglycan layer, suggesting that these cells may devote large proportions of their resources to enzymes that modify and transport nutrients, such as ABC transporters involved in the transport of organic molecules (64). Another striking feature of Pelagibacter is that the nucleoid and the ribosome-containing volume are consistently partitioned to the convex and concave sides of cells, respectively (Fig. 3). Cell polarity, i.e., nonuniform spatial organization of subcellular components along an axis, is common for prokaryotes (65), but asymmetry along the short cell axis of bacterial cells has not been reported previously. It is unclear whether cytoskeletal elements or other physical constraints are responsible for the polar segregation of the cellular components in Pelagibacter (66). With the nucleoid being located against the inner cell membrane on one side, however, the interface between the nucleoid and the ribosome-containing volume that would allow cotranscriptional translation is significantly reduced. In this way, Pelagibacter might be less efficient in protein synthesis than other bacteria. To reduce the biomass invested in protein synthesis and to reduce the energetic expense of protein synthesis, the translation rate of ribosomes is often 3- to 4-fold lower in slowly growing bacteria than in rapidly growing bacteria (67). SAR11 bacteria were thought to be small and to have few ribosomes, but we found an abundance of ribosome-like particles, relative to the slow growth of this cell type. In Table 2, we list ribosome concentrations in several organisms, although it is not clear how comparable the estimates of ribosome concentrations determined with different methods, such as biochemical analyses and cryo-ET, are (19, 21, 36, 37, 53, 68, 69). An apparent excess of ribosomes or ribosome-like particles was found in some other slowly growing organisms, such as in Rickettsia prowazekii (68), Sphingomonas sp. RB2256 (69), Spiroplasma melliferum (37), and Campylobacter jejuni (19).

Secretins are integral membrane proteins that function as OM portals for several bacterial export pathways, including the type II and III secretin systems, type IV pilus biogenesis, and filamentous phage release (70, 71). Despite the diversity of these secretory systems, secretins have markedly similar overall structures and organizations. Secretins consist of several modular N-terminal domains that reside in the periplasm and a conserved C-terminal domain that inserts in the OM, where it forms an oligomeric, ring-shaped conduit for pilus or protein extrusion (72, 73). The pore complex seen in Pelagibacter has a structure similar to that of secretin GspD in the type II secretion system and to the nonpiliated PilQ secretin in the type IV pilus machinery (Fig. 4G and H) (42, 45). We also compared the sequence of Pelagibacter PilQ predicted from the DNA sequence with those of two other secretins, i.e., Escherichia coli type II secretion system GspD and Neisseria meningitidis PilQ, for which crystal structures have been resolved (see Fig. S2 in the supplemental material) (73, 74). Sequence alignment indicated that Pelagibacter PilQ has a conserved secretin domain at its C terminus and two modular N-terminal domains (N0 and N1), suggesting that PilQ is the secretin that Pelagibacter uses to assemble type IV pili or other translocators across the OM. We did observe pili in Pelagibacter cells but in only 2 dividing cells among the 50 reconstructed cells. PilQ pore complexes, which are required for pilus biogenesis, were found in many of our reconstructed Pelagibacter cells, however, suggesting that most PilQ pore complexes are not involved in pilus assembly under standard growth conditions (46, 75). We speculate that these idle secretins, in their nonpiliated state, may form a second class of transporters, which may be active in DNA uptake or protein secretion (46, 76).

Type IV pili have been associated with a variety of cell processes, including twitching motility, adhesion to surfaces, aggregation, intercellular communication, transformation competence, DNA uptake, and biofilm formation (77, 78). Pili were observed previously, by scanning EM, in Pelagibacter cells during starvation in darkness. Greater expression of the pilin gene (pilA) was detected under similar conditions (79). However, the role of pili in starved Pelagibacter cells is not known. Adhesion to surfaces and twitching motility have not yet been observed in SAR11 cells in culture or in nature, and efforts to demonstrate DNA uptake or the use of DNA as a growth substrate by Pelagibacter cells have thus far been unsuccessful. However, the presence of DNA uptake and competence genes (pilC, pilD, pilE, pilF, pilG, pilQ, comL, and cinA) in the Pelagibacter genome suggests that this organism has the ability to acquire foreign DNA (4). Multilocus sequence typing recently showed that recombination is frequent in native SAR11 populations, indicating that these type IV pili might play an important role in the acquisition of DNA from sources outside the cells (80). Interestingly, here we found that the pili were associated with the constriction sites of dividing cells instead of the cell poles, as shown for cells with twitching motility (45); this suggests that Pelagibacter pili may not be involved in cell motility but may play a role during cell division. For example, Pelagibacter cells may use pili to attach to substrates in the ocean (such as plankton or detritus), either to assist with completion of cell fission or to allow daughter cells to float away after division, to ensure low local concentrations of cells competing for the same resources.

The SAR11 clade indisputably plays an important role in ocean ecology. Pelagibacter, as a cultivatable representative of this clade, is a good model organism to elucidate structure-function relationships in oligotrophic ultramicrobacteria. The results we reported here constrain the range of surface-to-volume ratios for living SAR11 cells. We also identified a large periplasmic space, an eccentrically compartmentalized nucleoid, and an abundance of ribosome-like particles, all of which provide fundamental insights into the ultraoligotrophic life cycle of SAR11 bacteria. Moreover, the Pelagibacter OM structures that we interpret as secretin PilQ and type IV pili may play roles in the high rates of recombination that have been observed in natural SAR11 populations (80) or may have unknown roles in substrate uptake.

MATERIALS AND METHODS

Cell cultivation.

As described by Rappé et al. (2), “Candidatus Pelagibacter ubique” strain HTCC1062, was originally collected off the Oregon coast and cultured in the Giovannoni laboratory. This strain belongs to the Ia.1 ecotype of SAR11. Live cultures are available from the Oregon State University High Throughput Microbial Cultivation Laboratory, and the strains have been deposited in the Bigelow Laboratory for Ocean Sciences culture collection. The Pelagibacter cells used in this study were cultivated in autoclaved seawater on low-nutrient heterotrophic medium (LNHM) (2). Cool white light of 24 μmol photons/m²/s was supplied in a 14-h light/10-h dark cycle. Stationary-phase cells were taken from cultures after ∼4 days in the stationary phase.

Cryo-ET.

Quantifoil grids with holey carbon film (copper R2/2; Quantifoil Micro Tools GmbH, Jena, Germany) were negatively glow discharged. Three-microliter volumes of bacterial cell cultures that were grown to the late log phase or the stationary phase were added to the grid, and then the cells were mixed with 1 μl of a 10-fold-concentrated solution of 10-nm colloidal gold (Sigma-Aldrich). Excess fluid was blotted from the grid with Whatman filter paper for ∼2 s before the grids underwent plunge-freezing in liquid ethane using a homemade plunge freezer. Frozen-hydrated samples were imaged using a Tecnai F30 transmission electron microscope (FEI, Hillsboro, OR) under low-dose conditions. Cultures of Pelagibacter grow only to low cell density in the laboratory (81). Therefore, we recorded low- and medium-magnification maps of the entire EM grid and selected grid squares, respectively. This allowed us to search the grid systematically, locate the rare cells that were embedded in vitreous ice, and then record the tilt series for the targeted cells, with a tilt range of about ±70°, angular increment of 1° to 3°, and total dose of about 100 e/Ų. The acquisition of overview maps and tilt series was accomplished using SerialEM software (82). To increase image contrast, specimens were imaged with −6- to −10-μm defocus, using a postcolumn Gatan imaging filter (Gatan, Pleasanton, CA) in zero-loss mode, with a slit width of 20 eV. The images were recorded with a Gatan charge-coupled device (CCD) camera (2,000 by 2,000 pixels) with 9.46-Å pixel size.

Image processing, analysis, and 3D modeling.

The IMOD software package was used for alignment of the tilt series images, using the 10-nm gold particles as fiducial markers, before tomograms were reconstructed by weighted back-projection (83). In Fig. 5A, two large gold fiducial clusters were manually outlined and erased during the CCD Eraser step in IMOD, to avoid ray artifacts. Subtomogram averaging was performed using PEET software (84) to average the putative PilQ pore complexes in Pelagibacter cells. The UCSF Chimera software package was used for 3D visualization of PilQ pore complexes by isosurface rendering (85). Cell modeling was accomplished with Amira 3D software (FEI). Ribosome-like particles were identified by size (∼20 nm in diameter) within the tomograms, using cross-correlation with a globular template.

Accession number(s).

Four representative cell tomograms that are reported in this study were deposited in the Electron Microscopy Data Bank, with accession numbers EMD-8326 (a log-phase cell), EMD-8327 (a stationary-phase cell), EMD-8328 (a dividing cell with pili), and EMD-8329 (dividing and almost separated cells). The averaged map of the native PilQ pore complex of Pelagibacter was deposited under accession number EMD-8330.