A bacterial tubulovesicular network

The cellular space of a eukaryotic cell is highly organized and is divided into functionally differentiated compartments by membrane-bound structures. The origin of such complex membranous organization is unknown and is an important issue in cellular, molecular and evolutionary biology. Although not as developed, bacterial intracellular organization has also proved to be surprisingly complex. In the past few decades, membrane-defined compartments have been observed in various prokaryotes, demonstrating that cellular subfunctionalization and differential localization also occurs in tiny organisms (Murat et al., 2010). In this context, the planctomycete bacterium Gemmata obscuriglobus is of considerable interest because of its complex and dynamic endomembrane system (Lee et al., 2009). This endomembrane system is associated with proteins that show structural and functional similarities to the membrane coat proteins, such as clathrin, which sustain the eukaryotic endomembrane system (Santarella-Mellwig et al., 2010). Furthermore, membrane internalization vesicles in G. obscuriglobus bacteria enable the uptake and degradation of external proteins in a process that is reminiscent of eukaryotic endocytosis (Lonhienne et al., 2010). Other bacteria with non-classical membrane organization, such as magnetotactic or photosynthetic bacteria, do not display such similarities with eukaryotes.

Historically, the cell plan of planctomycetes has been interpreted as being different from a classical Gram-negative (G−) one (Fuerst, 2005; Lindsay et al., 2001). However, recent genomic and electron-microscopy data have considerably weakened this interpretation and instead support the hypothesis that it is a variation to the G− cell plan (Santarella-Mellwig et al., 2013; Speth et al., 2012; Devos, 2013). Here, we stick to the more recent, G−-anchored, interpretation (Devos, 2013). In this interpretation, the outermost and internal membranes are equivalent to the G− outer membrane and inner membrane, respectively, defining the space between them as the periplasm.

Two cell types that are likely to represent major stages of the cell cycle have been reported in G. obscuriglobus (Lee et al., 2009; Santarella-Mellwig et al., 2010). The first cell type is characterized by extensive invaginations of the inner membrane inside the cytoplasm (Santarella-Mellwig et al., 2010; Santarella-Mellwig et al., 2013). The second cell type has increased periplasmic volume, which is populated by vesicle-like structures. In the course of our analysis of the G. obscuriglobus membrane organization, we investigated the organization of vesicles in this second cell type using electron microscopy methods.

We previously reported the detection of proteins showing structural and architectural similarity with the eukaryotic membrane coat proteins sustaining the eukaryotic endomembrane system, such as clathrin or the nucleoporins. We have shown that these proteins are in contact with the membrane of the vesicles in the periplasm of G. obscuriglobus cells (Santarella-Mellwig et al., 2010). Here, we show that these vesicles sustained by the membrane coat proteins are for the most part interconnected and form a network of tubules and vesicles – a tubulovesicular network (TVN) – linking the outer membrane to the inner membrane in this bacterium.

Serial electron tomography of G. obscuriglobus cells resulted in volumes where vesicle-like structures were observed in the periplasm of the bacterial cells (Fig. 1A). We observed that some of these vesicle-like structures were connected (Fig. 1B; supplementary material Movie 1). A more careful investigation revealed that most of the structures were connected to each other, forming a continuous membrane organization within the periplasm. This suggested the presence of a TVN, inside the periplasm of the bacterial cells. In addition, some vesicles were connected to the outer membrane, whereas some others were connected to the inner membrane (Fig. 2A). Thus, the majority of the periplasmic vesicles were connected to one another, or to one of the major cell membranes (Fig. 2A; supplementary material Movie 1).

Volume segmentation revealed a network of connected vesicles linked to the inner membrane and outer membrane inside the periplasm (Fig. 2B). Thus, in these cells, the outer membrane was connected to the inner membrane through a continuum of membranes, forming a TVN. Such a network was found in the periplasm of most of the cells of this type observed, representing roughly one-third of the cells in a typical population. In addition, these vesicles contain ribosomes, suggesting continuity with the cytoplasm. The content of some vesicles appeared to be darker in comparison to others or to the cytoplasm (Fig. 1B), implying regulation of material exchange at the connections between vesicles and subfunctionalization of the cellular space. Cellular subfunctionalization is supported by the previous report that G. obscuriglobus internalizes and degrades external proteins in the periplasm (Lonhienne et al., 2010).

Immunolocalization experiments with our previously generated antibody against one of the G. obscuriglobus membrane coat proteins (g4978) revealed colocalization with the bacterial TVN (Santarella-Mellwig et al., 2010). Most of the gold particles were localized in the periplasm and a significant proportion was found in contact with the TVN membranes (Fig. 2C). We counted 237 gold particles in 13 cells, 74 gold particles were found associated to the inner membrane (gold within a 15 nm distance of the membrane center) and 125 were found in the periplasm. The ratio of the inner membrane area to the cell area in the cross sections was found to be 0.17, giving a statistically significant P-value of 0.001 for the 31% of gold particles associated with the inner membrane; in agreement with our previous analysis (Santarella-Mellwig et al., 2010). Therefore, it is likely that membrane coat-like proteins are involved in the formation or maintenance of the bacterial TVN, similar to their counterparts in eukaryotes.

These observations have implications for our interpretation and understanding of the bacterial cell plan, and of the evolution of compartmentalization and discrete organelle function. We report the presence of a TVN in the periplasm of G. obscuriglobus, which physically connects the outer membrane to the inner membrane. This unexpected result suggests that there might be a selection mechanism, which regulates transport between the cytoplasm and the outside of the cell, through the periplasmic TVN. The implications of this observation at the level of cell biology and evolution are still to be clarified. This network physically connects the outer membrane to the inner membrane. This seems counterintuitive, because it appears to put the cytoplasm in direct communication with the outside of the cell. However, our observation of connections between vesicles of different content demonstrates the regulation of transfer between vesicles and thus, the filtering of transport between the outside and the cytoplasm of the cell, through the periplasmic TVN, most likely in both directions (export and import). This is in agreement with the previous observation of protein internalization and degradation in the periplasm of those bacteria (Lonhienne et al., 2010).

In the past few decades, we have learned a lot about the ancestral eukaryotic endomembrane system, and membrane organization in the organisms preceding the first eukaryote. Comparative genomic and phylogenetic analyses have revealed that the first eukaryotic cell possessed a complex endomembrane system, and a near-modern array of the protein families associated with it (Field and Dacks, 2009; Koumandou et al., 2013). In addition, common aspects of function and biogenesis of functionally distinct compartments of many eukaryotes suggest, among other possibilities, that the primitive eukaryotic endomembrane system might have been composed of a multifunctional TVN. This network was probably formed by distinct communicating compartments serving as the site of protein synthesis, endocytosis and degradation of internalized material (Abodeely et al., 2009). Therefore, a TVN that links the nuclear envelope to endocytic vesicles and where degradation of the internalized exogenous material takes place, has been suggested as a possible characteristic feature of a primitive eukaryotic endomembrane system (Abodeely et al., 2009). The ancestral TVN probably communicated with membrane-bound vesicles coated with clathrin-like membrane coat proteins at the periphery of the cell to receive internalized material (Devos et al., 2004).

This is very similar to what we observed in G. obscuriblobus. Here, we report the presence of a membranous TVN in this bacterium. The presence of a TVN, together with its previously reported features, reveals striking similarities between the G. obscuriglobus endomembrane system and an inferred ancestral eukaryotic one. Indeed, like the eukaryotic one, the bacterial endomembrane system is complex and dynamic to an extent unequalled so far in prokaryotes (Lee et al., 2009). The bacterial inner membrane sends invaginations towards the cytoplasm that are reminiscent of the eukaryotic endoplasmic reticulum (Santarella-Mellwig et al., 2013). However, there is no nucleus-like organization of the membranes around the DNA in the bacteria (Santarella-Mellwig et al., 2013). Vesicle-like membranous structures are also present in the bacterial periplasm (Lindsay et al., 2001). As we report here, those vesicles are connected and form a TVN. The bacterial endomembrane system is also involved in external compound internalization and degradation in the periplasm, most likely through the TVN, indicating subfunctionalization and filtering before cytoplasmic internalization (Lonhienne et al., 2010). This feature is unique in bacteria and related to endocytosis, which, until recently, was held as one of the strictly eukaryotic characteristics. In addition, this endomembrane system is in contact with proteins that are structurally similar to membrane coat proteins, such as clathrin, which are most likely involved in its maintenance or organization (Santarella-Mellwig et al., 2010). There is no proof of homology between the bacterial and eukaryotic membrane coat proteins, but structural, architectural and functional similarities support an evolutionary relationship between them (Devos, 2012). No sign of lateral gene transfer in any direction, to or from eukaryotes has, however, been detected (Santarella-Mellwig et al., 2010).

In conclusion, there is no evidence that the G. obscuriglobus and eukaryotic endomembrane systems are related. There are alternative explanations for the similarities observed between the G. obscuriglobus and a putative primitive eukaryotic endomembrane system; either planctomycetes represent an independent emergence of endomembrane organization, or the planctomycetal endomembrane system is related to a primitive eukaryotic one. The former could provide important information about the formation of complex structure and convergence, whereas the latter might provide a glimpse into the evolution of the complex endomembrane system of modern eukaryotes. On the basis of these similarities and others, various scenarios of relationships between the planctomycetes and the eukaryotes have been proposed (Fuerst and Webb, 1991; Forterre, 2011; Reynaud and Devos, 2011). In either case, planctomycetes provide an excellent opportunity to examine the endomembrane organization in a non-eukaryotic system, without the complexity found in eukaryotes.

G. obscuriglobus cells were grown as previously described (Santarella-Mellwig et al., 2010). The cells were frozen with a HPM010 high-pressure freezing machine (Abra Fluid, Switzerland) and freeze substituted in an AFS2 machine (Leica, Vienna) with either 1% osmium tetroxide, 0.1% uranyl acetate and 5% H₂O and embedded in Epon, or with 0.5% uranyl acetate and embedded in Lowicryl HM20 (Santarella-Mellwig et al., 2010). Thin (60 nm) and thick (250 nm) sections were placed on Formvar-coated grids and 15 nm fiducial gold markers were added. Sections were then stained with uranyl acetate and lead citrate. Antibody labeling was carried out as previously described (Santarella-Mellwig et al., 2010). Thin sections were imaged with a CM120 Phillips electron microscope. For tomography, dual-tilt axis acquisition was performed on thick sections with a Technai F30 300 kV microscope (FEI Company). Serial sections were reconstructed and tomograms were joined using IMOD (Kremer et al., 1996). Contours were traced on every slice within the tomogram (about 6 nm voxel size).

To determine the statistics of antibody proximity to membranes, we defined a membrane proximity area at a 15 nm distance from the membrane center and counted gold particles in the membrane proximity, in the periplasm and in the cytoplasm. We then used a one-sample Student's t-test with two-tailed P-value calculation using ratios of the cell compartment area to the total cell area as expected values for a random distribution to compare with the observed distribution of gold particles.