Strains were grown at room temperature (25°C) with shaking at 200 rpm in Lysogeny Broth (LB) (Bertani, 2004), DSM (Schaeffer et al., 1965b), or S7₅₀ minimal medium containing 50 mM MOPS (adjusted to pH 7.0 with KOH), 10 mM (NH₄)₂SO₄, 5 mM potassium phosphate (pH 7.0), 2 mM MgCl₂, 0.7 mM CaCl₂, 50 μM MnCl₂, 1 μM ZnCl₂, 1 μg/ml thiamine-HCl, 20 μM HCl, 5 μM FeCl₃ and 1% glucose supplemented with 0.1% glutamate and 0.004% casamino acids (Grossman and Losick, 1988; Harwood and Cutting, 1990). Spectinomycin (100 μg/ml), chloramphenicol (5 μg/ml), macrolide-lincosamide-streptogramin (MLS) (1 μg/ml erythromycin and 25 μg/ml lincomycin), or kanamycin (10 μg/ml) were added to the growth medium where appropriate. The growth medium was supplemented with 1 μM IPTG when required. For comparative analysis of growth, small round cells formation and sporulation, samples were collected at regular time intervals for OD₆₀₀ measurement and evaluation of the % of small round cells and sporulation. The % of small round cells was determined by cell counts on acquired images. The evaluation of the % of sporulation is described below (see the “heat kill assay” method). Unless stated otherwise, samples were collected after 90–100 h of growth for microscopy. To investigate the viability of small round cells, cultures of B. subtilis generating small round cells were diluted 1:10 into fresh growth medium and grown for 60 min prior to the time-lapse experiments. Table 1 lists all the strains used in this work.

Samples were collected in regular time intervals and serially diluted with sterile double distilled water (ddH₂O). The number of spores was measured as heat-resistant (20 min at 80°C) colony-forming units (CFU) on LB plates. In parallel, viable cells were measured as total colony-forming units (CFU) on LB plates. In this case, serial dilutions were performed in sterile 0.9% NaCl. % sporulation = [spores/ml)/(cells/ml)] × 100% (Grossman and Losick, 1988). However, because the numbers of live vegetative cells decline after the start of the stationary phase, the percentage of sporulation during stationary phase was calculated relative to the number of viable cells reaching the stationary phase.

Cells were mounted on agarose gel pads containing S7₅₀ medium (Harwood and Cutting, 1990) on microscope slides. Fluorescence microscopy was performed on a Zeiss Axio Imager A1 or a Zeiss Axio Observer Z1 microscope equipped with a CoolSnap HQ CCD camera (Photometrics) or pco.edge camera (scientific CMOS camera) respectively. The electronic data processing was conducted with Metamorph 6.3-Software (Meta Imaging Software) or VisiView Test-Version 2.1.1 (Visitron Systems GmbH). Fluorescence intensities were measured using the ROI manager tools of ImageJ (v1.48; Rasband,W.S., ImageJ, U.S. National Institutes of Health, Bethesda, Maryland, USA, http://imagej.nih.gov/ij/, 1997-2014) (Schneider et al., 2012).

Endospore formation is the preferred response of B. subtilis cells to different nutrient-limited conditions in minimal medium (Chubukov and Sauer, 2014). However, the extent to which this bacterium can adjust its structure and physiology during extended stationary phase in optimum growth condition is less known.

To follow sporulation behavior of wild-type B. subtilis cells during prolonged stationary growth under laboratory conditions, minimal S7₅₀ medium (Grossman and Losick, 1988; Harwood and Cutting, 1990) was chosen for this study, which is frequently used to grow B. subtilis. For comparative analysis Lysogeny Broth (LB) (Bertani, 2004) and sporulation medium DSM (Schaeffer et al., 1965b) were also used. Cells were grown at room temperature (25°C) with shaking at 200 rpm and samples were collected at regular time intervals for OD₆₀₀ measurement, microscopy analysis and evaluation of the % of sporulation. Whenever B. subtilis was grown in S7₅₀ minimal medium, LB broth or DSM medium, sporulation occurred earlier in S7₅₀ and DSM media (~ 12 h) compared to LB broth (~ 21 h). In all media, the percentage of spores in relation to the total viable population increased with time and yielded 0.94 and 0.21% in S7₅₀ medium and LB broth, respectively, after 126 h (Figures 1A,B; the sporulation data are shown in Tables S1,S2). Meanwhile, spores readily accumulated in DSM medium and reached 14.5% of total viable counts after 126 h (Figure 1C; the sporulation data are shown in Table S3). Spores were formed already during exponential growth phase in S7₅₀ medium (Figure 1A), in contrast to DSM medium and LB broth, in which spores appeared only when cell density declined (Figures 1B,C). These observations indicate that spore development in DSM medium and LB broth arose upon nutrient exhaustion, while in S7₅₀ medium the heterogeneity of the cell population with respect to sporogenesis started earlier during growth.

Interestingly, in addition to spores, the formation of small round cells was observed that appeared in a way similar to the well-characterized minicells (Adler et al., 1967; Reeve et al., 1973) (Figures 1D–G). Appearance of this cell type could be detected in all media, but was most prominent in S7₅₀, in which a rapid increase was noted at the transition between logarithmic growth and stationary phase. The small round cells eventually constituted 15.2% of the total cell counts (1750 cells counted; small round cells counts data are shown in Table S4) in S7₅₀ (Figure 1D) compared to 0.45 or 0.9% in LB or DSM, respectively (Figures 1E,F, the small round cells counts data are shown in Tables S5,S6). The small round cells exhibited diameters varying from 0.7 to 1.15 μm (150 cells measured) (Figure 1G). The undomesticated B. subtilis strain (NCIB 3610) as well as other species such as Bacillus atrophaeus, pumilus, and sphaericus (Figure S1) also produced small round cells, suggesting that this phenotype is not a trait of the lab strain PY79 and may be conserved among endospore-forming bacteria.

A brief analysis of the small round cells showed that they were significantly different from minicells because of the presence of DNA in these cells visualized by DAPI staining (Figure 2A), which was absent in minicells from a mutant defective in proper placement of the cell division machinery (Figure 2B).

The question that immediately arose was to find out the origin of these small round cells. The accumulation of small round cells in the growth medium as the culture reached the stationary phase where sporulation events should have happened more frequently suggests that the occurrence of small round cells in S7₅₀ could be the result of failed spore development.

B. subtilis has different developmental states that are mostly controlled by Spo0A, the master regulator of the stationary phase. To further confirm that there is a link between sporogenesis and small round cells formation, the expression level of Spo0A using a gfp fusion to the spoIIG promoter was monitored. spoIIG is a sporulation gene which is controlled by Spo0A (Satola et al., 1992). Spo0A is the sporulation initiator protein and its intracellular level determines the entry of cells into the sporulation process. High levels of Spo0A are required for the activation of sporulation-specific genes (Fujita et al., 2005). Figures 3A,B show that the intensity of GFP (expressed from spoIIG promoter from an ectopic locus) fluorescence signals in cells producing small round cells (54171.84 ± 30261.937 arb. Units; mean ± SD of 178 measurements) was similar to that of sporulating cells (60386.68 ± 26532.14 arb. Units; mean ± SD of 193 measurements). This suggests that the Spo0A concentration in cells producing the small round cells was sufficient to trigger sporulation. Moreover, all cells producing the small round cells exhibited GFP fluorescence suggesting that small round cells formation requires the activity of Spo0A. Indeed, I also investigated a spo0A mutant which is known to arrest sporulation at stage 0 (Piggot and Coote, 1976). No single small round cells (observation from three independent experiments, 300 cells analyzed each) could be detected in this strain (Figure 3C). Cells lacking spo0H gene, which codes for σ^H that directs genes expression at an early stage of sporulation, also did not generate small round cells (observation from two independent experiments, 350 cells analyzed each). However, for unknown reasons spo0H deletion strains investigated in this work exhibited elongated shaped cells (Figure 3D). Because sporulation is the unique developmental state requiring an asymmetric septation, these findings are consistent with the idea that the small round cells can only be formed upon initiation of spore development. Thus, the small round cells evolve from the inability of cells to complete sporulation. Of the cells generating the small round cells (the small round cells still being attached to their larger siblings), 20% (Figure 2C, 238 cells counted) underwent more than one asymmetric division at the same pole or opposite poles (Figures 1G, 2A, black asterisk), indicating that a failure to develop a spore does not imply that the cell returns to the vegetative state, and thereby allowing a subsequent reinitiation of sporulation. Thus, the initial sporulation signals are rather strong and remain stable in cells producing the small round cells.

Asymmetric cell division in B. subtilis is the second major morphological change observed during sporulation (Schaeffer et al., 1965a). My observations indicated that the development of the spore septum must be required for the small round cells formation. To test whether the ability to form proper polar septa affects the occurrence of the small round cells, the bifunctional SpoIIE protein was examined. SpoIIE is involved in asymmetric division and is required for the establishment of compartment-specific activation of transcription factors (Barák et al., 1996). SpoIIE is proposed to play a role in the switch from one Z-ring observed during vegetative growth to two polar Z-rings when cells enter sporulation (Khvorova et al., 1998). Consequently, SpoIIE also assembles into rings (E-ring) that are dependent on Z-ring formation (Arigoni et al., 1995; Levin et al., 1997; King et al., 1999). Figure 3E shows rings and helical patterns of SpoIIE-GFP localization. 92% of small round cells (146 cells analyzed) had E-rings at their septum. The remaining small round cells showed no or very faint SpoIIE-GFP signals at their septum (Figure 3E). Interestingly, an E-ring also frequently formed at the mid cell position indicating that the interaction of SpoIIE with the vegetative Z-ring precedes the polar relocalization of the Z/E-rings via the helical pattern distribution of both proteins (Ben-Yehuda and Losick, 2002) (Figure 3E).

The precise role of SpoIIE in this dynamic relocalization of FtsZ is still unclear. However, in the absence of SpoIIE, sporulating cells are still able to assemble polar Z-rings and to form asymmetrically positioned septa (Levin et al., 1997; Wu et al., 1998). These septa are rather thick and similar to those formed during vegetative growth. The literature suggests that SpoIIE may regulate cell wall synthesis at the asymmetric division sites (Piggot and Coote, 1976; Illing and Errington, 1991; Barák and Youngman, 1996) and may play a role in peptidoglycan remodeling during engulfment (Campo et al., 2008). Indeed, a recent study indicated that SpoIIE recruits the morphogenic protein RodZ at the asymmetric septum during sporulation where both proteins might be involved in peptidoglycan thinning during engulfment (Muchova et al., 2016). The second distinct function of SpoIIE during sporulation is the activation of σ^F in the forespore due to its phosphatase activity (Duncan et al., 1995; Arigoni et al., 1996; Feucht et al., 1996). Therefore, a spoIIE mutant arrests spore development at stage II (Margolis et al., 1991). The spoIIE null mutant analyzed in this work also generated the small round cells (Figure 3F), but only up to 3.0 ± 0.6% (mean ± SD of three independent experiments; 247 cells counted each); a frequency that is 5-fold less than that observed in wild-type strain. This finding is consistent with the previous report that spoIIE null mutants have a noticeably reduced frequency of polar septum formation (Barák and Youngman, 1996) and shows a correlation between the ability to form polar septa and the generation of small round cells. Taken together, these findings support the idea that the small round cells formation requires the asymmetric septa involved in sporulation.

Because E-ring signals could be also visualized at the small round cells septa (Figure 3E), I investigated whether σ^E or σ^F were indeed expressed during the formation of small round cells in the former mother cell or forespore respectively. The activity of σ^E or σ^F was visualized using strains bearing gfp fusions to spoIID or spoIIQ promoter of whose expression is controlled by σ^E or σ^F, respectively (Rong et al., 1986; Londoño-Vallejo et al., 1997). Figure 4 shows that σ^E and σ^F were active in cells undergoing sporulation, but not in those generating small round cells. Sporulating cells showed discrete curved septa indicating the engulfment of the forespore at the cell pole while a true polar cell division was often marked by a clear invagination of the cell wall (Figures 4A,B). In some cases, cells turned on σ^E or σ^F following the formation of the small round cell (Figures 4A,B, black asterisks). This observation is consistent with the above finding that cells generating the small round cells retain sporulation signals and can attempt another round of sporulation initiation. Unexpectedly, the forespores were always formed at poles where small round cells had been generated, suggesting the presence of specific markers at the cell poles. These markers would eventually drive a subsequent sporulation event at the same pole. Thus, the formation of the small round cells does not require the activities of σ^E or σ^F but σ^H controlling the early stationary phase genes involved in the initiation of sporulation.

At the onset of sporulation, as cells possess two chromosomes, the origins of replication are targeted to the opposite cell poles by RacA in a DivIVA-dependent manner (Ben-Yehuda et al., 2003). This mechanism ensures that the future sporangium gets a copy of the chromosome at both poles. When the asymmetric division is taking place, SpoIIIE, a DNA translocase assembles at the closing septum and actively pumps the remaining chromosome from the mother cell into the forespore (Wu et al., 1995; Wu and Errington, 1997; Bath et al., 2000).

This raises the question whether DNA is actively transported into the small round cells or simply pinched-off from the bulk nucleoid by the closing septum. To explore this question, I examined a strain expressing SpoIIIE fused to YFP as a sole source of protein during the formation of small round cells. SpoIIIE-YFP accumulated at the asymmetric septum, as previously reported (Wu et al., 1995; Wu and Errington, 1997; Bath et al., 2000) (Figure 5A). Interestingly, a SpoIIIE-YFP foci was also present at the asymmetric septum of most of the small round cells containing DNA (90% of 213 cells counted), but absent at the septum of all empty small round cells (empty small round cells represent about 1.67% of the small round cells) (Figure 5A). Thus, similar to the sporulation process, SpoIIIE is also required for DNA translocation during the formation of small round cells. Surprisingly, up to 60% (180 cells counted) of small round cells produced by a spoIIIE null strain contained DNA (Figure 5B). This observation suggests that another cellular factor is implicated in this process. Such factor could be the B. subtilis DNA translocase SftA (septum-associated FtsK-like translocase of DNA). SftA belongs to the SpoIIIE/FtsK-like protein family found in B. subtilis, but lacks the transmembrane domain (Biller and Burkholder, 2009; Kaimer et al., 2009). It has been proposed that SftA acts first at the cell division site by clearing the sister chromosomes and that SpoIIIE only comes into play when chromosomes get trapped at the closing septum (Biller and Burkholder, 2009; Kaimer et al., 2009; Wu, 2009). However, recent investigations using super-resolution microscopy revealed that SpoIIIE is consistently present at the invaginating septum (Fiche et al., 2013), suggesting that both proteins are synchronously recruited to the division machinery. A spoIIIE/sftA double null mutant neither produced small round cells nor sporulated (Figure 5C), a finding consistent with the correlation between spore and the small round cells formation. Thus, similar to their respective function during vegetative growth, SpoIIIE, and SftA act synergistically to ensure proper chromosome partitioning/translocation during sporulation or subsequent small round cells formation when the sporulation process fails.

It was interesting to find out if the translocation of DNA into the emerging small round cells might be affected while abortion of the sporulation occurs. In other words, do the small round cells contain full chromosomes? Assuming that 2/3 of the forespore chromosome, which remains in the mother cell (Wu and Errington, 1994), is translocated in an ordered manner (i.e., the terminus of the chromosome being translocated last), this question was examined by visualizing the presence of the terminus (terC) region of the chromosome in the small round cells. Figure 6A shows that about 50% (103 cells counted) of the small round cells contained the terC region while 98.33% of the small round cells contained DNA (Figure 2A). The discrepancy between the number of cells containing terC and the total number of cells containing DNA could be due to incomplete translocation of the chromosome into the forespore. In agreement with this, about 99% (68 cells counted) of the small round cells were found to contain oriC (Figure 6B). Therefore, the ~2% of anucleate small round cells may suggest a release of the oriC region from the forespore pole and subsequent retraction of the chromosome into the mother cell compartment. This could happen when opposing signals to commitment are sensed earlier in the developmental process. However, although there is no evidence that B. subtilis can reverse commitment to sporulation by retracting the chromosome, it should be noted that the chromosome will be retracted if the origin is not in the forespore (Becker and Pogliano, 2007). In any case, the high probability that the small round cells inherit a full copy of the chromosomal DNA increases the chances of viability for these cells.

Sporulating cells lacking σ^F are unable to complete spore development and end up with an empty mid-cell compartment flanked by two opposite forspores containing each a full copy of the chromosome. These forespores are able to resume vegetative growth when shifted into rich medium (Dworkin and Losick, 2005). A key question of this work was whether the small round cells were dead-end structures or whether they could resume normal growth upon addition of fresh medium. To test this, I monitored the growth of small round cells by time-lapse microscopy. Cultures of B. subtilis generating small round cells were diluted into fresh growth media and grown for 60 min at room temperature before time-lapse imaging. Strikingly, the small round cells were able to return to the rod shaped form (Figures 7A–C, Figure S2, and Movie S1). Different sizes of small round cells were monitored for growth. Out of 49 cells analyzed, 47 cells extended into rods. The small round cells observed did not extend at the same speed. The speed of growth was dependent on the size of the small round cells at the beginning of the experiment. The larger the diameter size of small round cells, the faster they reached their first division. Among the 47 cells that extended into rods, 16 cells that had diameter sizes ranging from 0.95 to 1.15 μm cells divided after 2.5–3 h and 21 cells that had diameter sizes ranging from 0.8 to 0.9 divided after 3.5–4 h. The rest of the cells (10 cells) with smaller diameter sizes (~0.7 μm) did not did divide during the time course of the time-lapse experiment, which was ~4 h.

Taken together, these experiments suggest that the here-described small round cells could provide an escape route for cells that are unable to continue the sporulation process.

The author declares that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.