No Endospore Formation Confirmed in Members of the Phylum Proteobacteria

Bacterial endospore formation is an important process that allows the formation of dormant life forms called spores. Organisms able to sporulate can survive harsh environmental conditions for hundreds of years.

different types of sporulation, the underlying transformations are morphologically distinct and encoded by nonhomologous pathways (10).
Endospore formation begins with asymmetric cell division, with the septum placed near one pole of the cell, and produces two cells with different fates (11)(12)(13). Upon septation, the smaller compartment becomes engulfed through a phagocytosis-like mechanism, yielding a prespore bound by two lipid membranes in the cytoplasm of the mother cell. Subsequent endospore maturation involves the synthesis of protective layers, such as the peptidoglycan-based cortex and proteinaceous coat. Metabolic inactivation is achieved by gradual dehydration of the core through replacement of water with dipicolinic acid (DPA) and calcium ions and compaction of DNA with DNA-binding proteins. Together, these modifications account for the resistance properties of endospores (14). Ultimately, the spore is released upon lysis of the mother cell (15). In contrast, other modes of sporulation, such as those observed in Actinobacteria and Myxococcus, produce spores through morphological differentiation and cell division without engulfment.
Several studies in the past decade have reported, but not proven, formation of endospores in Proteobacteria (16,17). While endosporulation has recently been confirmed in some Gram-negative bacteria, all of the identified organisms still belong to the phylum Firmicutes, highlighting the question of evolutionary origins of the bacterial outer membrane (18). Additionally, sporulation involves tight cooperation of hundreds of genes distributed across the chromosome, hindering acquisition of this pathway through horizontal gene transfer (10,19,20). Therefore, if confirmed, the ability to form endospores across distantly related bacterial phyla suggests an ancient nature of the process and can provide clues to the characteristics of the last bacterial common ancestor (10). In this study, we investigated two articles attributing sporulation to members of Proteobacteria (16,17). Briefly, Girija et al. (17) described endospore production in the purple, nonsulfur bacterium Rhodobacter johrii, strain JA192(T), a close relative of the model organism for bacterial photosynthesis Rhodobacter sphaeroides. The second study, by Ajithkumar et al. (16), reported endospore formation in Serratia marcescens subsp. sakuensis (strain no. 9; KRED T ), a pathogenic bacterium that infects humans and causes bacteremia, urinary tract infection, and wound infections (21). Thus, confirmation and further characterization of endospore formation in these organisms can bring valuable insight into the physiology of these species and the role of endospore formation in diversification and speciation of modern phyla.
In this study, we employed cutting-edge structural biology techniques, such as cryo-electron tomography (cryo-ET), correlative light and electron microscopy (CLEM), and energy-dispersive X-ray spectroscopy (EDX), as well as biochemical and microbiological approaches, to characterize endospore formation in R. johrii and S. marcescens. Our results showed that R. johrii and S. marcescens were unable to form endospores as previously reported (16,17). Further analyses indicated that the putative spores in R. johrii were lipid storage granules (SGs) rich in triacyclglycerols (TAGs) and that the phase-bright objects in S. marcescens were aggregates of cellular debris. Overall, our observations contradict the previously published studies by Girija et al. and Ajithkumar et al. and support the observation that these members of Proteobacteria are unable to form endospores.

RESULTS AND DISCUSSION
Prolonged incubation induces formation of phase-bright objects in R. johrii and S. marcescens. For initial assessment of the previously reported endospore formation, we cultivated R. johrii and S. marcescens according to the published conditions and examined the cultures with phase-contrast light microscopy (LM) (Fig. 1). Vegetative R. johrii cells appeared phase-dark (Fig. 1A); however, after 7 s incubation, phase-bright objects were observed either at one pole or mid-cell (Fig. 1B, black arrows). Vegetative S. marcescens cells also appeared phase-dark (Fig. 1C). Although phase-bright objects were occasionally visible at mid-cell following a 65-day incubation (Fig. 1D, black arrows), the majority of culture was dead and appeared as "ghost" cells ( Fig. 1D, white arrows).
Altogether, our results recapitulate reports of formation of phase-bright objects in R. johrii and S. marcescens following extended incubation under nutrient-limited conditions.
Characterization of phase-bright objects in R. johrii. To further characterize the phase-bright objects observed in R. johrii, we performed correlative LM and cryo-ET experiments on phase-bright and phase-dark cells following extended incubation (Fig.  2). Tomograms of R. johrii cells with phase-bright objects revealed the presence of intracellular granules, which were highly sensitive to the electron beam, as represented by the sample damage ( Fig. 2A). Beam sensitivity was detected regardless of the total dose used (25 to 150 electrons [e 2 ]/A 2 ), suggesting that the granules were rich in lipids. Further, the spherical nature of the granules resembled previously characterized storage granules (SG) in bacterial cells (22). No evidence of sporulation-associated morphological changes, such as engulfing membranes or the presence of immature or mature spores in the sample (n = 40), was observed, indicating that the phase-bright objects were not endospores. Finally, cells with phase-bright objects always displayed 1 to 3 of the 100-to 250-nm-diameter granules ( Fig. 2A), whereas the phase-dark cells lacked the presence of granules (Fig. 2B). Thus, our observations suggest that the phasebright objects were likely lipid-containing SGs.
To determine the composition of the storage granules, we performed correlative LM and scanning electron microscopy (SEM) in combination with EDX compositional analysis. Cells possessing the putative storage granules were identified with phasecontrast microscopy (Fig. 3A) and examined at higher resolution with SEM (Fig. 3B). Correlative LM and SEM were then used to guide EDX analysis, so that spectra were collected from a region containing the putative storage granules and a cytoplasmic region lacking the storage granules (Fig. 3C). Elemental analysis of the storage granule (blue spectrum) revealed counts for carbon (C) of 80.24%, oxygen (O) of 13.26%, and copper (Cu) of 6.5% (due to the copper of the EM grid). Cytoplasmic analysis (red spectrum) revealed lower counts for carbon (61.7%) and oxygen (10.82%), copper at 6.59%, and elevated counts for nitrogen (N; 20.9%) (Fig. 3C).
Based on the cryo-ET and EDX data, we hypothesized that the granules observed in R. johrii were composed of lipids, as lipids are enriched in carbon and oxygen atoms. To characterize the nature of the granular composition, we performed whole-cell lipidomic analysis of R. johrii 7-day-old cultures expressing phase-bright objects (granules) against R. johrii cultures grown for 2 days and lacking phase-bright objects as the negative control. Cultures producing putative storage granules where enriched in several lipids, the most abundant of which were triacylglycerols (TAGs) and phosphatidylethanolamines (PEs) ( Table 1). Because PEs are typical membrane lipids, the increased levels observed under starvation conditions suggested that cells remodel their membrane composition to account for the environmental changes. TAGs are nonpolar, occur as insoluble inclusions in bacteria, and are considered a major source of energy (23,24). TAGs have been shown to accumulate in actinobacteria and mycobacteria either as peripheral deposits associated with the cell envelope or as inclusion bodies in the cytoplasm (25). Previously, in vitro studies showed that mycobacteria accumulated TAG and wax ester when subjected to stresses, such as low oxygen, high CO 2 , low nutrients, and low pH (25)(26)(27). Similarly, we observed an increased propensity to form phasebright objects in R. johrii cells incubated under low-nitrogen conditions in defined media. Therefore, it is likely that R. johrii utilizes TAG storage as an adaptive strategy in response to starvation, allowing cells to enter stationary phase and survive for longer periods. We thus conclude that the granules observed as phase-bright objects in R. johrii were storage granules enriched in TAGs.
Characterization of phase-bright objects in S. marcescens. Tomograms of S. marcescens were collected for 2-day and 65-day-old cultures ( Fig. 4). At 2 days, we observed regular morphology of vegetative cells, displaying cell envelope architecture typical for Gram-negative bacteria (Fig. 4A). S. marcescens grown for 65 days revealed the presence of two kinds of morphologies: cells packed with cellular debris (black asterisk in Fig. 4B) and cells void of any cellular material (white asterisks in Fig. 4B), likely correlating to cells containing phase-bright objects and ghost cells identified using LM, respectively. Extensive survey of the sample (n = 80) did not reveal any cells possessing intracellular membranes or morphologies suggestive of engulfing membranes or stages of sporulation. Neither of the two identified morphologies displayed any features similar to a cortex or proteinaceous spore coat characteristic of mature endospores. Additionally, we did not observe accumulation of storage granules within cells. Together, these results suggest that the appearance of phase-bright objects in S. marcescens was the result of accumulation of cellular debris and dehydration. Proteobacteria do not possess features of endospores following extended incubation. Independent of imaging-based methods, endospores have traditionally been identified in samples through heat resistance and increased concentration of intracellular DPA. To verify the results of our cryo-ET experiments, we first investigated the heat resistance properties of R. johrii and S. marcescens following prolonged incubation and subsequent exposure to high temperatures. Despite an extended recovery period (7 days), no viable R. johrii cells were observed on solid media following 15, 30, and 60 min of incubation at 80°C. Similarly, viable cells were not isolated from S. marcescens cultures incubated at 60°C for 15, 30, and 60 min. In contrast, Bacillus subtilis cultures producing endospores and treated at 80°C for 15, 30, and 60 min yielded viable growth on solid media after a 24-h recovery period. Thus, we were unable to replicate the results of Girija et al. and Ajithkumar et al., who found viable cells following heat treatment of R. johrii at 80°C for 20 min and S. marcescens at 60°C for 15 min, respectively. Additionally, we quantitatively analyzed the presence of DPA in cultures of R. johrii and S. marcescens displaying phase-bright objects using a colorimetric method. Whereas the purified endospores of B. subtilis contained 6.74 mg/ml of DPA, no detectable amounts of DPA were observed in R. johrii or S. marcescens after prolonged cultivation. Collectively, these results indicate that R. johrii and S. marcescens cells do not possess the classic phenotypic features that are associated with endospore formation.
Minimal subset of genes required for endospore formation not conserved in R. johrii and S. marcescens. Endospore formation relies on expression of hundreds of conserved genes in a highly regulated manner (20,(28)(29)(30). For example, over 500 genes have been previously implicated in sporulation in the model firmicute B. subtilis (28). However, establishment of the minimal subset of genes required for endospore formation remains elusive, as many of the identified targets carry out redundant functions, e.g., histidine kinases, or are part of general pathways loosely associated with sporulation, such as iron uptake and DNA repair proteins (31). Consistently, several homologs to genes linked to sporulation have been detected in other phyla, including Proteobacteria, but have been shown to play regulatory roles in distinct processes, such as cell division and development (32,33). Hence, possession of genes annotated as sporulative should not be considered concrete evidence to support sporulation capacity in a given species (5). Nevertheless, we investigated the genomes of R. johrii and S. marcescens for presence of genes that are conserved among all spore-forming bacilli and clostridia based on the COGs database and have been shown to play pivotal roles in endospore formation through functional studies (Table 2) (5,34,35). Our analysis showed that both R. johrii and S. marcescens completely lack the SpoIIDMP peptidoglycan remodeling complex required for spore cortex formation, the SpoIIQ-SpoIIIAA-AH channel complex involved in communication between the mother cell and the prespore and facilitating regulation of endospore maturation, and the major protein coat assembly components, such as SpoIVA (12,14,36). Further, the master regulator of sporulation encoded by all endospore formers, Spo0A, is absent in R. johrii and S. marcescens. Both strains also lack homologs to all five sporulation sigma factors, SigE, SigF, SigG, SigH, and SigK. Finally, R. johrii and S. marcescens do not possess DapB, required for production of dipicolinic acid in Bacilli, which plays a major role in dehydration of the spore core and, therefore, resistance and dormancy (14). A homolog of etfA (a gene associated with DPA synthesis in some Clostridia), however, was identified in R. johrii and Escherichia coli but not in S. marcescens ( Table 2). The role of this gene in Proteobacteria has not been characterized, and it could encode an electron transfer flavoprotein. Overall, our analysis confirms the lack of the major structural and regulatory sporulation genes in the genomes of R. johrii and S. marcescens. In addition, Ajithkumar et al. were also unable to detect genes related to endospore formation in S. marcescens (16).
Concluding remarks. Although endospore formation is considered a hallmark of the Firmicutes phylum (4,5,37), endospore production had been reported outside Firmicutes, particularly for two members of the phylum Proteobacteria (16, 17). These findings may affect our understanding of the evolutionary events surrounding outer membrane biogenesis and the significance of endospore formation in cell differentiation. In this study, using cutting-edge microscopy techniques and biochemical, microbiological, and bioinformatics approaches, we showed that the phase-bright objects observed in R. johrii and S. marcescens are storage granules and cellular debris, respectively. We did not observe mature spores or stages of endospore formation in vivo, and we failed to detect the pivotal biochemical and genomic features of endospore-producing bacteria in these organisms. Our findings thus demonstrate that R. johrii and S. marcescens are unable to form true endospores, which is in contrast to the results described by Girija et al. (17) and Ajithkumar et al. (16). Since we used the most advanced imaging techniques currently available to study whole-cell bacteria and their ultrastructure, previous results could be due to the presence of contamination with spore-forming bacteria or misinterpretation of methodology artifacts.

MATERIALS AND METHODS
Bacterial strains and growth conditions. R. johrii and S. marcescens cells were purchased from the Leibniz-Institut DSMZ bacterial strain collection. R. johrii JA192 cells (DSMZ 18678) were cultivated as previously described by Girija et al. (17). Briefly, cells were grown aerobically at room temperature in R.  (16), for 7 days for vegetative growth or 65 days to induce formation of phase-bright objects. B. subtilis strain PY79 was chosen as the positive control for endospore formation, and cells were cultivated in LB broth at 37°C with shaking at 200 rpm overnight for vegetative growth or for 3 days to induce sporulation. LB agar was used for cultivation on plates for S. marcescens and B. subtilis.
Detection of phase-bright objects using phase-contrast light microscopy. R. johrii and S. marcescens cultures were pelleted and washed with 1Â phosphate-buffered saline (PBS; pH 7.4) composed of 137 mM NaCl, 27 mM KCl, 10 mM Na 2 HPO 4 , and 1.8 mM KH 2 PO 4 . Cells were imaged with an upright Zeiss Axio Imager M2 microscope (Carl Zeiss, Oberkochen, Germany) equipped with a 506 monochrome camera and a 100Â oil lens objective with a numerical aperture (NA) of 1.46.
Sample preparation for correlative light and cryo-electron tomography. R. johrii cells were lightly fixed using 2.5% paraformaldehyde in 30 mM phosphate buffer for 15 min, washed twice, and resuspended in 150 mM phosphate buffer. Bacterial cells were loaded onto Cu Finder R 2/2 EM grids (Electron Microscopy Sciences, Hatfield, PA), coated with 1 mg/ml of poly-L-lysine, and subsequently imaged at room temperature as described above. Following room temperature light microscopy, 20-nm colloidal gold particles (UMC Utrecht, Netherlands) were added and samples were plunge-frozen into liquid ethane-propane mix cooled at liquid nitrogen temperatures with a Mark IV Vitrobot (Thermo Fisher Scientific), maintained at room temperature and 70% humidity. Cryo-ET was conducted on cells with phase-bright signal as described below. For standalone cryo-ET experiments, samples were directly mixed with 20-nm colloidal gold particles, loaded onto glow-discharged carbon grids (R2/2; Quantifoil), and plunge-frozen as described above.
Cryo-ET data collection. For both standalone cryo-ET and CLEM experiments, tilt series were collected on a 300-kV Titan Krios transmission electron microscope (Thermo Fisher Scientific) equipped with a Falcon 2 camera. Tilt series were collected at nominal magnifications of Â14,000 to Â18,000, 1-to 3-degree oscillations, and a final dose of 30 to 150 e 2 /A 2 . Three-dimensional reconstructions were calculated with IMOD software package using the weighted back projection method (38).
Correlative LM and SEM with EDX analysis. R. johrii cells were fixed with 12.5% paraformaldehyde in 150 mM sodium phosphate buffer (73.6 mM K 2 HPO 4 , 26.4 mM KH 2 PO 4 ; pH 7.5) and then washed three times with 150 mM sodium phosphate buffer (39). Glow-discharged Cu R2/2 grids were coated with poly-L-lysine hydrobromide solution (1 mg/ml) and dried for 30 min at 60°C. Fixed cells were loaded onto the grids and immediately imaged with LM in 1Â PBS to identify phase-bright objects. The grids were subsequently air dried, and regions of interest identified with LM were examined with SEM using a JSM-7400F field emission scanning electron microscope (JEOL Ltd., Tokyo, Japan) operated at 5 kV without any coating. EDX analysis with a silicon drift detector (Octane, EDAX Inc., Mahwah, NJ) at 10 kV was used for semiquantitative elemental analysis of regions of interest.
Whole-cell lipidomic analysis of R. johrii. R. johrii cells displaying phase-bright properties were cultivated as described above, harvested by centrifugation (20 min at 5,000 rpm), and washed twice in sterile H 2 O. Cell pellets were then lyophilized overnight and stored at room temperature for up to 7 days. R. johrii cultures grown in LB for 2 days and lacking phase-bright objects were chosen as the negative control. Sample normalization between samples of R. johrii expressing storage granules [R.j (1)] and R. johrii lacking storage granules [R.j (2)] was achieved by their culture weights. For R.j (1) samples, 4.0 mg of cells (wet weight) was used; for R.j (2) samples, 2.5 mg was used. The lipidomic results were normalized according to culture weights. For the whole-cell lipidomic analysis, a methyl tert-butyl ether (MTBE)-based membrane lipid extraction protocol was used, with modifications (40). Briefly, samples in 1.5-ml Eppendorf vials were first mixed with 300ml of ice-cold methanol and 10-ml internal standards. The mixture was then sonicated in icewater bath for 15 min for protein precipitation. One milliliter of MTBE was added to the mixture, followed by vortex mixing for 20 min at room temperature for thorough lipid extraction. Next, 200 ml of liquid chromatography-mass spectrometry (LC-MS)-grade water was added to induce phase separation, and the samples were further mixed for 30 s. After settling for 10 min, the upper layer, containing the lipids, was transferred to new Eppendorf vials. To dry the lipid samples, the solvent was evaporated using a vacuum concentrator at 4°C. A total of 100 ml of isopropanol/acetonitrile (1:1, vol/vol) was added to reconstitute the dried residue. The reconstituted solution was vortexed for 30 s and centrifuged at 14,000 rpm at 4°C for 15 min. The resulting supernatants were transferred to glass inserts for LC-tandem MS (LC-MS/MS) analysis. Only lipids above the noise level (1,000 average intensity) were considered in the analysis. A cutoff value of at least 2Â increase in average intensity and a P value threshold of 0.01 were used to determine significant increase in lipid species.
Heat inactivation and counting of endospores. After induction of the phase-bright objects in R. johrii, S. marcescens, and B. subtilis as described above, cells were washed with sterile, deionized water, spun at 10,000 Â g for 10 min, and resuspended in chilled water. Suspensions of R. johrii and B. subtilis were heated to 80°C, and S. marcescens to 60°C, for 15 min, 30 min, and 1 h, as described previously (16,17). After the heat treatment, the samples were centrifuged for 10 min at 10,000 Â g. The pellets were washed five times to remove cellular debris and then plated onto solid media. R. johrii was incubated at 30°C for 7 days, S. marcescens was incubated at 32°C for 7 days, and B. subtilis was incubated at 37°C overnight, and plates were subsequently examined for viable growth.
DPA detection. Following the detection of phase-bright objects, dipicolinic acid (DPA) was detected as previously described (41). Briefly, 5-ml cultures of R. johrii, S. marcescens, and B. subtilis containing ;10 mg of cells (dry weight) were autoclaved for 15 min at 15 lb/in 2 . The suspensions were cooled to ambient temperature, acidified with 0.1 ml of 1.0 N acetic acid, and incubated for 1 h to cluster the insoluble material. To remove cellular debris, the suspensions were centrifuged at 1,500 Â g for 10 min. To each 4 ml of supernatant, 1 ml of 1% Fe(NH 4 ) 2 (SO 4 ) 2 ·6H 2 O and 1% ascorbic acid in 0.5 M acetate buffer (pH 5.5) was added. Colorimetric shift at 440 nm was compared to a standard curve prepared with pure DPA (Sigma-Aldrich, Oakville, Canada).
Detection of genes required for endospore production. We assessed the genomes of S. marcescens subsp. sakuensis KRED T and Rhodobacter johrii JA192-the two strains originally described by Girija et al. and Ajithkumar et al., respectively (16, 17)-for endospore formation genes. Using the recently updated COGs database, we compiled a list of genes associated with sporulation (5,42). Corresponding KEGG and TIGRFAM hidden Markov models (HMMs) were used to assess the genomes for the presence of these genes using HMMER 3.2.1 (http://hmmer.org/) (43,44). Genes without representative HMMs were identified using local alignment against sequences from B. subtilis 168 and Clostridium perfringens SM101 using a minimum percent identity of 30% and an E value of 1E225. B. subtilis 168 (monoderm sporulator), Acetonema longum DSM 6540 (diderm sporulator), and Escherichia coli DSM 30083 (diderm nonsporulator) were included as sporulating and nonsporulating controls.