Binary or Nonbinary Fission? Reproductive Mode of a Predatory Bacterium Depends on Prey Size

ABSTRACT Most bacteria, including model organisms such as Escherichia coli, Bacillus subtilis, and Caulobacter crescentus, reproduce by binary fission. However, some bacteria belonging to various lineages, including antibiotic-producing Streptomyces and predatory Bdellovibrio, proliferate by nonbinary fission, wherein three or more chromosome copies are synthesized and the resulting multinucleoid filamentous cell subdivides into progeny cells. Here, we demonstrate for the first time that the predatory bacterium Bdellovibrio bacteriovorus reproduces through both binary and nonbinary fission inside different prey bacteria. Switching between the two modes correlates with the prey size. In relatively small prey cells, B. bacteriovorus undergoes binary fission; the FtsZ ring assembles in the midcell, and the mother cell splits into two daughter cells. In larger prey cells, B. bacteriovorus switches to nonbinary fission and creates multiple asynchronously assembled FtsZ rings to produce three or more daughter cells. Completion of bacterial cell cycle critically depends on precise spatiotemporal coordination of chromosome replication with other cell cycle events, including cell division. We show that B. bacteriovorus always initiates chromosome replication at the invasive pole of the cell, but the spatiotemporal choreography of subsequent steps depends on the fission mode and/or the number of progeny cells. In nonbinary dividing filaments producing five or more progeny cells, the last round(s) of replication may also be initiated at the noninvasive pole. Altogether, we find that B. bacteriovorus reproduces through bimodal fission and that extracellular factors, such as the prey size, can shape replication choreography, providing new insights about bacterial life cycles.

the curves are different among the analyzed prey bacteria. The fastest decrease in prey cell optical density was observed for S. flexneri cells. Similar dynamics of predation was observed for S. enterica. While the dynamics of B. bacteriovorus predation on the third tested pathogen P. mirabilis cells was substantially different: there was a slow decrease in optical density at the beginning of measurements followed by a rapid decrease in optical density (Fig. S1A).
To further elucidate the differences in predation dynamics (mirrored by the killing curves), we measured different parameters of the prey and predator cells, including the prey cell size before infection, the diameter of the formed bdelloplasts (Fig. S1B), and the number of predatory cells released after prey cell lysis ( Fig. 1) using single cell microscopic data. The average size of the prey cell varied and was reflected by differences in the bdelloplast diameter (Fig. S1B). As expected, the diameter of the bdelloplast was positively correlated with the length (and the area) of the prey cell (calculated Pearson correlation coefficient, R = 0.87). The number of predatory cells formed inside the bdelloplast depends on the prey cell size (Fig. 1). In contrast to E. coli and other larger prey cells, predation on a tiny P. mirabilis cell resulted in the generation of two progeny cells (100%, n = 100 prey cells analyzed) (Fig. 1). More progeny cells were formed in larger-sized prey cells as follows: two to five (usually 3 or 4) and four to eight (usually 5 to 7) progeny cells were produced from S. enterica and S. flexneri, respectively (Fig. 1). To prove that the positive correlation between number of predator offspring and prey size is independent of the bacterial species, we analyzed E. coli S17-1 cells of different sizes (see Table S2 in the supplemental material). Our measurements confirmed that the number of predator offspring scales with the E. coli size (a similar relationship was also observed for S. flexneri as a prey) (Table S2). Taken together, the results indicate that the infection dynamics of B. bacteriovorus predation is related to the number of progeny cells released from a single prey cell. The predatory killing curves illustrated this correlation (Fig. S1A); at the beginning of infection, when more progeny predators were released (e.g., from S. flexneri cells), more prey cells were killed as reflected by a faster decrease in the prey cell optical density (Fig. S1A).
In summary, our results suggest for the first time that B. bacteriovorus divides not only by Bdellovibrio bacteriovorus Exhibits Bimodal Fission mBio nonbinary (more than two, even or odd daughter cells) but also by binary (two daughter cells in P. mirabilis) fission, and the mode of division appears to depend on the prey size.
Life cycle of B. bacteriovorus in different pathogens. As we observed prey-dependent differences in predation dynamics, we next decided to closely investigate the main stages of the B. bacteriovorus reproductive phase (i.e., chromosome replication, segregation, and filament septation) in P. mirabilis, S. enterica, and S. flexneri at the single-cell level using time-lapse fluorescence microscopy (TLFM). For this purpose, we constructed a set of B. bacteriovorus reporter strains that allowed us to monitor the localization of DnaN, ParB, and FtsZ proteins involved in the chromosome replication, segregation, and filament septation, respectively (for details, see Table 1). All B. bacteriovorus reporter strains exhibited a killing curve for E. coli (see Fig. S2A in the supplemental material) similar to that of the wild-type strain. Production of the fusion proteins was confirmed using Western blotting (Fig. S2B).
Chromosome replication and segregation. Chromosome replication and segregation dynamics of B. bacteriovorus were studied in P. mirabilis, being the simplest model as only two daughter cells are produced (Fig. 1). Firstly, we intended to observe the localization of the replisome(s) during the reproductive phase. We therefore replace the native DnaN of B. bacteriovorus with a DnaN-mNeonGreen. The appearance and disappearance of DnaN-mNeonGreen foci indicate the assembly and disassembly, respectively, of replisome complexes and thus correspond to the initiation and termination of DNA replication (22). DNA synthesis started (i.e., a DnaN-mNeonGreen focus appeared) consistently at the invasive pole (3,19) in bdelloplasts (n = 100, 96%) ( Fig. 2C; see also Video S1 at https://figshare.com/ authors/Jolanta_Zakrzewska-Czerwinska/14854612) at 46 6 4 min after bdelloplast formation ( Table 2). The second focus appeared at 31 6 4 min (n = 100) after the first initiation of DNA replication and subsequently migrated to the opposite pole (the pole that contained  the flagellum before prey attack, hereinafter called the "former flagellar pole"). The pole-topole distance was covered in 11 6 2 min (n = 100). Next, the replisomes relocated from the cell poles to a midcell position ( Fig. 2; see also Video S1 at the URL mentioned above). Both replisomes remained at this position until the end of chromosome replication, which was indicated by diffusion of the DnaN-mNeonGreen fluorescent foci. It is worth noting that we never observed more than two foci during the reproductive phase in P. mirabilis-derivative bdelloplasts (n = 100). Thus, the DnaN-mNeonGreen fluorescent foci presumably represent the two replication forks that, during DNA synthesis, transiently split at the invasive pole and then merge in the midcell position. The average duration of chromosome replication was 144 6 7 min (n = 100). We next investigated the dynamics of chromosome segregation in relation to DNA replication for B. bacteriovorus growing in P. mirabilis. For this purpose, we created allelic replacements of ParB and DnaN with fluorescent fusion proteins in B. bacteriovorus (see Table 1; see also Table S1), which allowed us to simultaneously monitor both processes in a single cell. The ParB focus (i.e., oriC region) was observed as previously described (3,23) at the invasive pole (see Fig. 3A). Notably, B. bacteriovorus mNeonGreen-ParB/DnaN-mCherry a Average times were obtained after statistical analysis performed on 100 predator filaments for each prey. For all measured values, 95% confidence interval was calculated, assuming a normal distribution of results. The appearance of fluorescent foci was measured from the bdelloplast formation (time zero). The time of newly synthesized oriC (i.e., ParB focus) migration to the opposite pole was measured from the duplication of ParB focus. b In B. bacteriovorus filament growing inside S. flexneri, more than three DnaN and four ParB foci were observed, but data are not presented. strain exhibited a proliferation-phase fluorescent pattern for DnaN-mCherry ( Fig. 3; see also Video S2 at https://figshare.com/authors/Jolanta_Zakrzewska-Czerwinska/14854612) consistent with that observed in the B. bacteriovorus DnaN-mNeonGreen strain ( Fig. 2; see also Video S1 at the URL mentioned above). In this system, DNA replication was followed by chromosome segregation (Fig. 3; see also Video S2 at the URL mentioned above). Shortly after the initiation of DNA synthesis (at 16 6 3 min and 69 6 4 min after bdelloplast formation) (see Table 2), one of the newly replicated oriC regions visible as a ParB focus (ParB complexes colocalize with the oriC region) (3) started to migrate to the opposite cell pole, reaching its destination after 19 6 2 min (Table 2). In contrast to replisomes, the segrosomes did not relocate to midcell but instead remained at the filament poles ( Fig. 3; see also Video S2 at the URL mentioned above). Careful examination of dozens of proliferating B. bacteriovorus cells in P. mirabilis (n = 45; as an example, see Fig. 4; see also Video S3 at the URL mentioned above) using an agarose pad in combination with ibidi cell-imaging dishes (predatory cells are less motile in this system) (19) revealed that the former flagellar pole (i.e., the one opposite to the invasive pole marked by a distinct ParB complex) (3,23) during the attack phase later became an invasive pole with a visible ParB complex. Thus, the two poles arising after division become the flagellar poles. The newly created invasive pole is fully functional as the daughter cell attacks the next prey cell through that pole ( Fig. 4; see also Video S3 at the URL mentioned above). To understand how chromosome replication and segregation dynamics are affected by the production of multiple progeny cells, we fed B. bacteriovorus on larger than P. mirabilis cells, S. enterica (usually 3 or 4 daughter cells) and S. flexneri (4 to 8 daughter cells) (differences in sizes among analyzed preys are shown in Fig. 1 and Fig. S1B). As expected, in these preys, chromosome replication was also initiated at the invasive cell pole of B. bacteriovorus  (Table 2). In S. enterica and S. flexneri, one of the newly replicated B. bacteriovorus oriC regions (i.e., an mNeonGreen-ParB focus) also migrated to the opposite cell pole, reaching its destination at 18 6 2 and 17 6 2 min, respectively, after oriC duplication (Videos S4 to S6 at the URL mentioned above; Table 2). Meanwhile, the DnaN-mCherry focus started to move from the invasive pole and transiently followed the ParB focus, indicating the progression of mother chromosome replication ( Fig. 5D and Fig. 6D; see also Fig. S3D; see Videos S4 to S6 at the URL mentioned above). The next steps in chromosome multiplication were related to the number of released daughter cells. When three progeny cells were produced during predation on S. enterica, DNA replication was reinitiated from the chromosome located at the invasive pole. The newly assembled replisomes (visible as a large DnaN-mCherry focus) (see Fig. 5E; see also Video S4 at the URL mentioned above) moved away from the invasive pole ( Fig. 5F and G; see also Video S4 at the URL mentioned above), and meanwhile, the first round of DNA replication was completed (indicated by disappearance of the first DnaN-mCherry signal) ( Fig. 5G; see also Video S4 at the URL mentioned above). It was challenging to analyze the later stages of chromosome replication because in most cases it was unclear whether the observed foci were from transiently splitting sister replication forks and/or from temporally overlapping replication rounds. When four daughter cells were formed in an S. enterica bdelloplast, we observed two asynchronous reinitiation events; in both cases, new replisomes were assembled at the invasive pole (see Fig. S3H and I; see also Video S5 at the URL mentioned above). Notably, the first reinitiation of DNA replication  (Fig. S3H) after the disappearance of the first DnaN-mCherry focus, which was located in the middle of the filament and originated from the first initiation.
In S. flexneri bdelloplasts, we observed up to five DnaN-mCherry foci, indicating that three (or more) pairs of replication forks were simultaneously active in a single filament. Unexpectedly, in an S. flexneri bdelloplast producing five (or more) progeny cells, at later stages of chromosome multiplication (185 min after bdelloplast formation) ( Fig. 6F; see also Video S6 at https://figshare.com/authors/Jolanta_Zakrzewska-Czerwinska/14854612), we noticed the appearance of a DnaN-mCherry focus at the former flagellar pole, suggesting that replication may also initiate at the opposite pole.
In all bacteria, replication initiation is synchronized with the segregation of newly synthesized oriC region(s) (24). In all of the preys analyzed in our study, shortly after initiation of the first replication round, one of the newly synthesized oriC regions visible as a ParB focus (2nd ParB focus) moved toward the former flagellar pole (Fig. 3, 5, and 6 and Table 2; see also Fig. S3; see Videos S2 and S4 to S6 at https://figshare.com/authors/Jolanta_Zakrzewska -Czerwinska/14854612). In S. enterica or S. flexneri, similar to E. coli (3), the number of ParB foci gradually increased as new copies of the oriC region were synthesized ( Table 2). The second round of chromosome segregation started at 74 6 5 min after the first round in S. enterica and at 69 6 5 min in S. flexneri (n = 100, data not shown). The newly assembled segrosome (3rd ParB focus) also migrated toward the opposite pole but did not reach the end of the filament, which was already occupied by the 2nd ParB complex; the 3rd ParB focus remained at the midcell ( In conclusion, the choreography of B. bacteriovorus chromosome replication occurring in S. flexneri differs from that occurring in P. mirabilis and S. enterica; in S. flexneri, DNA replication may be reinitiated from the chromosome located at the former flagellar pole. Septation. In all living cells, chromosome replication must be synchronized with cell division. Since chromosome replication is not followed immediately by cell division in B. bacteriovorus except in cases of binary fission (see below), we decided to elucidate how chromosome replication is synchronized with filament septation during the growth of B. bacteriovorus in preys that produce different numbers of daughter cells. For this purpose, we used a B. bacteriovorus strain producing DnaN-mCherry and FtsZ-mNeonGreen fusion proteins (Table 1). FtsZ filaments assemble into a ring (called the Z-ring) at the future site of the septum, and the appearance of an FtsZ focus is commonly regarded as an early marker of bacterial cell division (25). FtsZ from B. bacteriovorus is highly similar to other FtsZ homologs (53% identity with E. coli FtsZ) within the N-terminal domain, which contains conserved residues responsible for GTP binding, but possesses a unique C-terminal addition of ;160 amino acids (see Fig. S4 in the supplemental material).
As expected, during predation in P. mirabilis, only one FtsZ signal appeared before the B. bacteriovorus filament divided into two daughter cells. Recruitment of FtsZ-mNeonGreen to the division site started 165 6 4 min (n = 100) after bdelloplast formation (Table 3 and Fig. 7; see also Fig. S5A.G in the supplemental material). The FtsZ focus was localized in the middle of the mother cell ( Fig. 7; Fig. S5A and B; see also Video S7 at https://figshare .com/authors/Jolanta_Zakrzewska-Czerwinska/14854612), and was visible for 49 6 3 min (n = 100) (see Table 3). After FtsZ disassembly, the mother cells divided into two daughter cells. The duration of the B. bacteriovorus life cycle (from bdelloplast formation to progeny cell release) in P. mirabilis lasted approximately 4 h ( Table 2).
Investigation of the septation process during predation on S. enterica revealed that two or three FtsZ signals were usually visible inside a single filament (Fig. 7; see also Fig. S6 in the supplemental material; see Video S8 at https://figshare.com/authors/Jolanta _Zakrzewska-Czerwinska/14854612), and, after their disappearance, the filament septated into three (see Fig. 7; Fig. S6) or four daughter cells, respectively. Surprisingly, in filaments growing in a prey producing three or more sibling cells, the FtsZ foci appeared asynchronously (for details, see Fig. 7 and Table 3; Fig. S6 and S7 in the supplemental material). For example, in S. enterica producing four daughter cells, the first FtsZ focus was visible 217 6 5 min after bdelloplast formation while the last one (i.e., the third one) appeared approximately half an hour later (244 6 6 min after bdelloplast formation) ( Table 3). The FtsZ foci remained noticeable for 52 6 6 min and then underwent disassembly, and the filament then divided into four daughter cells. Interestingly, the FtsZ foci are disassembled almost synchronously; in most analyzed filaments (.90%) dividing into 3 or 4 daughter cells, the FtsZ foci (i.e., 2 or 3, respectively) disappeared within 5 min (see Table S3 at the URL mentioned above).
In the case of B. bacteriovorus growing in S. flexneri, it was difficult to precisely analyze the septation process, as the long, twisted filaments formed in this prey prevented in depth microscopic analysis. At the beginning of the septation process, we were able to observe the sequential appearance of multiple FtsZ signals (up to 7) and, as in other preys, the FtsZ rings never formed simultaneously (see, for example, Fig. S7; see also Video S9 at https://figshare.com/authors/Jolanta_Zakrzewska-Czerwinska/14854612).
Moreover, the TLFM analysis revealed that the recruitment of FtsZ to the Z ring started before the end of DNA synthesis. In all preys, the first FtsZ-mNeonGreen focus was visible approximately 10 to 20 min before the disassembly of the last replisome (Table 3 and Fig. 7; see also Videos S7 to S9 at https://figshare.com/authors/Jolanta_Zakrzewska-Czerwinska/ 14854612), namely, at 12 6 2 min in P. mirabilis (85%, n = 100), 16 6 2 min in S. enterica (88%, n = 100), and 19 6 2 min in S. flexneri (90%, n = 100).  At the end of DNA synthesis, in binary dividing cells, the FtsZ-mNeonGreen focus consistently colocalized with DnaN-mCherry at the midcell position (n = 100). This is simultaneously the junction point of the two replication forks (see the overlapped FtsZ-mNeonGreen and DnaN-mCherry signals in Fig. 7, Fig. S5 in the supplemental material, and Video S7 at https:// figshare.com/authors/Jolanta_Zakrzewska-Czerwinska/14854612). Interestingly, when three daughter cells were produced in an S. enterica bdelloplast, the first FtsZ focus was visible near the former flagellar pole (Fig. 7), where the first round of replication and segregation had already been completed (see Fig. 7 and Table 3; see also Fig. S6). Meanwhile, at the end of the second round of replication, the DnaN focus was located asymmetrically with respect to the filament length. It was closer to the invasive pole (65%, n = 100).
In S. flexneri, we were unable to precisely monitor the position of FtsZ rings in relation to the last visible replisome, especially in long filaments. However, we noticed that in a filament that would divide into five progeny cells, the first FtsZ-mNeonGreen focus frequently colocalized with the last visible DnaN-mCherry focus (78%, n = 100) ( Fig. 7; see also Fig. S7G), which was positioned close to the former flagellar pole (consistent with observations in S. enterica). We also observed that in S. flexneri producing five (or more) progeny cells, the initiation of replication may also take place at the former flagellar pole (at later stages of chromosome multiplication) (see Fig. 6F, 185 min after bdelloplast formation; see also Video S6 at https://figshare.com/authors/Jolanta_Zakrzewska-Czerwinska/14854612).
Since we found that the formation of FtsZ rings in the predatory filament is an asynchronous process and the number of released B. bacteriovorus cells depends on the prey/cell size, we decided to analyze the size of the progeny cells released from the analyzed pathogens in comparison to the model prey, E. coli. Length measurement of mature B. bacteriovorus cells after bdelloplast lysis showed that there were statistically relevant differences (see Fig. S8 at https://figshare.com/authors/Jolanta_Zakrzewska-Czerwinska/14854612). The average lengths of the generated B. bacteriovorus cells were 1.25 6 0.03 mm for P. mirabilis, 1.16 6 0.03 mm for S. enterica, 1.35 6 0.03 mm for S. flexneri, and 1.11 6 0.04 mm for E. coli. On average, the longest progeny cells were released from S. flexneri (usually 5 or 6 progeny cells), which had the largest cell size of the analyzed preys. The daughter cells born from P. mirabilis (2 progeny cells) were longer than those released from E. coli (usually 3 or 4 progeny cells) and S. enterica (usually 3 or 4 progeny cells). There is, therefore, no direct correlation between the number of progeny cells and the size of the progeny cells.
Understanding the proliferation of B. bacteriovorus in different pathogens may also be helpful for its use to efficiently eliminate a given pathogen, i.e., by calculating the optimal ratio between the predator and the prey. We developed a new application to simulate these infection dynamics using the number of predatory and prey cells, the average number of progeny cells released from a single prey cell (Fig. 1), and the duration of the reproductive phase (Table 2). For the purposes of our simulation, we assume that all released progeny cells immediately find new preys and that prey cells do not replicate (similar to the killing curves where prey cells were resuspended in Ca-HEPES buffer). The surviving number of prey cells is calculated as n_prey i11 = n_prey i 2 n_bdellovibrio i and the new number of Bdellovibrio cells is calculated as n_bdellovibrio i11 = n_bdellovibrio i Â n_progeny. This simulation, called BdelloSim, allowed us to calculate the time necessary to eliminate the entire prey population given the same starting population sizes of prey and B. bacteriovorus cells as those used for the killing curves (see Fig. S1 in the supplemental material; see also  Fig. S9 at the URL mentioned above) obtained by BdelloSim are consistent with the killing curves, for P. mirabilis, 2,500 min; for S. enterica, 1,740 min; and for S. flexneri, 1,520 min.
Taken together, the results of our TLFM analyses revealed the cell cycle parameters of B. bacteriovorus filaments growing in three different pathogens. Our work demonstrates that the mode of the B. bacteriovorus cell cycle depends on the prey cell size, with filaments proliferating in small cells, such as P. mirabilis, dividing by binary fission, while those growing in larger prey cells undergo nonbinary fission.

DISCUSSION
The use of predatory bacteria such as B. bacteriovorus is currently regarded as a promising strategy to kill antibiotic-resistant pathogens (9,26,27). Therefore, understanding the mode and dynamics of B. bacteriovorus proliferation in various dangerous pathogens may facilitate knowledge-based improvements in B. bacteriovorus strains for use as a "living antibiotic" in human and veterinary medicine, particularly to combat multidrug-resistant pathogens. Previously, B. bacteriovorus was believed to undergo a nonbinary life cycle. In this study, we demonstrate for the first time that this predatory bacterium proliferates by either a binary or a nonbinary mode of fission, depending on the prey and its size (Fig. 8A). We also show that the choreography of DNA replication in this system exhibits unique features and is related to the mode of fission giving rise to different number of progeny cells (Fig. 8A). Finally, we show that FtsZ rings are assembled asynchronously in nonbinary proliferating cells of B. bacteriovorus.
B. bacteriovorus undergoes binary fission when growing in a small prey cell. Our data reveal that, in small pathogenic bacteria such as P. mirabilis (cell length, ,2 mm), only two predatory cells are formed (see Fig. 8A and B). In this prey, the FtsZ ring is assembled in the middle of the filament and then undergoes disassembly, whereupon the filament is divided into two daughter cells. Interestingly, the assembly of the FtsZ ring begins before the termination of DNA replication. Detailed microscopic analysis of P. mirabilis undergoing predation by B. bacteriovorus reveals that the old pole that had been flagellated becomes an invasive pole with a visible ParB complex. Simultaneous analysis of both replication and segregation processes confirmed this observation: during ongoing DNA replication, one of the newly replicated oriC regions (observed as a ParB focus) migrates During the progress of DNA replication, the replication forks split (one replisome follows the segregated oriC region) and merge at the midcell and then disassembled. The FtsZ ring assembly is initiated at the midcell before the termination of DNA replication, and next the filament is divided into two progeny cells. toward the opposite pole (the one bearing a flagellum before the prey attack). Further microscopic observations demonstrated that, after B. bacteriovorus division, the released daughter cell with the invasive pole arising from the former flagellar pole of the mother cell attaches to the next cell prey using this cell pole and proliferates within this prey. Thus, after the division by binary fission, one daughter cell inherits the old invasive pole while the other inherits the former flagellar pole of the mother cell, which must be extensively rebuilt to become an invasive pole. Consequently, flagella must be formed at new poles emerging from the last division (Fig. 8).
Collectively, our results demonstrate for the first time that B. bacteriovorus may undergo binary fission and P. mirabilis provides a simple and ideal model prey for studying the life cycle of this predatory bacterium.
Unique choreography of chromosome replication in binary-dividing B. bacteriovorus. As in other preys, including E. coli (3,19), chromosome replication in B. bacteriovorus undergoing predation on P. mirabilis starts at the invasive pole. Then, shortly after the initiation of DNA synthesis, the replisomes split; one migrates toward the former flagellar pole while the other remains at the invasive pole. Before the end of DNA synthesis, the two replisomes move toward each other to meet at the cell center, where assembly of the FtsZ ring has begun. Thus, at the beginning of chromosome replication, the choreography of this process in B. bacteriovorus resembles that found in other asymmetrically dividing bacteria, such as C. crescentus (28) and Vibrio cholerae (chromosome I) (29), where the replisomes are also assembled at a cell pole. This is contrary to chromosome replication in E. coli (24,30) and B. subtilis (31), where replication starts at the midcell. However, the spatiotemporal choreography of further steps of B. bacteriovorus chromosome replication differs from those described in other asymmetrically dividing bacteria. For example, in C. crescentus, the replisomes migrate together from the cell pole to the midcell position, where replication is terminated (28,32).
The assembly of replisomes at the invasive pole of B. bacteriovorus (3,19) suggests that, as in other asymmetric bacterial cells, the oriC region must be anchored to the cell pole via a specific polar protein(s). To date, such polar proteins, also called "hub" proteins, have been characterized in C. crescentus (PopZ) (33), V. cholerae (HubB) (34), and Streptomyces (DivIVA) (35,36). The B. bacteriovorus genome lacks homologs of PopZ or HubP but does encode a version of DivIVA (37). It was recently speculated that DivIVA together with a recently identified homolog of the filament-forming protein, bactofilin (37), may be involved in maintaining the polarity of this predatory bacterium. However, the precise role of these proteins in anchoring the B. bacteriovorus oriC region at the invasive pole remains to be elucidated. We cannot exclude the possibility that other proteins function to anchor the chromosome at the invasive pole.
In summary, we herein present, for the first time, the choreography of chromosome replication and cell division during the binary cell cycle of B. bacteriovorus (see Fig. 8B).
FtsZ rings are asynchronously assembled across the elongated filament during nonbinary fission. In larger preys (cell length, .2 mm), the filament of B. bacteriovorus undergoes nonbinary division and odd or even numbers of cells are produced. As expected, the number of released progeny cells strictly depends on the number of FtsZ rings that are formed before filament division (Fig. 7). Although the septation process in B. bacteriovorus was previously shown to be synchronous (4), our study revealed that FtsZ is recruited asynchronously to the sites of division, with the first FtsZ signal appearing before the end of DNA synthesis (Fig. 7). Further FtsZ signals appear sequentially. For example, in a filament that will divide into four nascent cells, the third FtsZ signal is assembled ;30 min after the appearance of the first FtsZ signal (Table 3). However, the disassembly of FtsZ rings was found to be a synchronous process (see Table S3 at https://figshare.com/authors/Jolanta _Zakrzewska-Czerwinska/14854612), which is consistent with previous studies showing a synchronous septation in B. bacteriovorus (4). In Streptomyces, which is a Gram-positive filamentous bacterium, the multigenomic filaments also undergo synchronous division to form chains of unigenomic exospores, but, in contrast to B. bacteriovorus, the FtsZ rings are synchronously assembled in this case (5,38), and this occurs after the cessation of DNA replication (39). Thus, we speculate that in B. bacteriovorus, yet undiscovered replication checkpoints act to coordinate chromosome replication with cell cycle progression, such as by tuning FtsZ ring placement to prevent the formation of daughter cells with a missing/guillotined chromosome or an additional copy of the chromosome.
In nonbinary dividing long filaments, the initiation of chromosome replication may also occur at the cell pole opposite the invasive pole. Similar to the results obtained by Laloux's group (3), we observed that DnaN foci usually appeared sequentially at the invasive pole, suggesting that there is asynchronous firing of replication initiation from the chromosome attached to this pole (the newly appeared fluorescently tagged DnaN foci colocalized with the ParB attached to the invasive pole) (Fig. 5 and 6; see also Fig. S3 in the supplemental material). However, it proved hard to follow the further progression of chromosome multiplication (especially in filaments that would divide into several or more nascent cells) because it was difficult to distinguish between transiently splitting sister replication forks and temporally overlapping replication rounds. In Streptomyces vegetative hyphae, DNA replication is also (re)initiated from the chromosome that is anchored at the filament tip, and one of the newly replicated oriC regions bound by ParB protein is unidirectionally segregated (6). Thus, in both B. bacteriovorus and Streptomyces, after oriC duplication, ParB complexes must be reestablished at both daughter oriCs at the cell pole; thereafter, one of them is captured by apically localized hub proteins and becomes abandoned by the segregation machinery.
Surprisingly, during the proliferation of B. bacteriovorus in S. flexneri, we noticed that the reinitiation of chromosome replication may also occur at the former flagellar pole ( Fig. 6; see also Video S6 at https://figshare.com/authors/Jolanta_Zakrzewska-Czerwinska/14854612). This phenomenon was observed only in the very late stages of chromosome multiplication at ;160 min after the first initiation of chromosome replication. We speculate that in the longer filaments (those that will divide into five or more progeny cells), the former flagellar pole is converted to an invasive pole while chromosome replication is still in progress. Therefore, the last round(s) of DNA replication may also be initiated from the chromosome that is already attached to the newly formed invasive pole. However, further work is needed to elucidate the mechanism(s) underlying the molecular switch that activates replication initiation from the chromosome where it attaches to the old invasive pole and/or the pole that has been converted from former flagellar to invasive.
In summary, our findings demonstrate that in a nonbinary dividing filament of B. bacteriovorus, chromosome replication may be initiated not only from the invasive pole, but also from the opposite pole, which has presumably been converted to an invasive pole.
Conclusions. We herein set the scene for B. bacteriovorus to become a novel model that enables the study of both binary and nonbinary cell division. Our findings provide new insight into the life cycle of a predatory bacterium that exhibits bimodal fission and demonstrate that the mode of division depends on the size of the prey bacterium inside which B. bacteriovorus grows. Further studies on the B. bacteriovorus life cycle should aim to shed light on the regulatory mechanisms responsible for switching between the binary and nonbinary modes of the cell cycle. Moreover, understanding the proliferation of B. bacteriovorus in different pathogens may facilitate its use to efficiently eliminate a given pathogen (i.e., by calculating the optimal ratio between the predator and the prey).

MATERIALS AND METHODS
Bacterial strains and culture conditions. The culture conditions and construction of bacterial strains were as described in Makowski et al. (19). Briefly, the plasmids used to construct B. bacteriovorus HD100 strains were propagated in E. coli DH5a grown in LB broth or on LB agar plates (supplemented with 50 mg/mL kanamycin) and then transformed into E. coli S17-1. The E. coli strains used as prey for B. bacteriovorus were liquid cultured in YT medium (0.8% Bacto tryptone, 0.5% yeast extract, 0.5% sodium chloride, pH 7.5) with (S17-1 pZMR100) or without (S17-1, ML35) kanamycin (50 mg/mL) at 37°C with shaking (180 rpm). B. bacteriovorus cells were grown by predation on E. coli (ML35, S17-1, or S17-1 pZMR100) or pathogenic preys (P. mirabilis, S. enterica, or S. flexneri) in Ca-HEPES buffer (25 mM HEPES, 2 mM calcium chloride, pH 7.6). All pathogenic bacteria were grown in LB broth at 37°C with shaking (180 rpm) or on LB agar plates at 37°C. Details regarding the strains, plasmids and oligonucleotides used in this study are presented in Table S1 in the supplemental material.
Predatory killing curves. B. bacteriovorus strain HD100 (wild type) was propagated in P. mirabilis, S. enterica, or S. flexneri. An overnight culture of prey cells was spun down at 5,000 rpm for 10 min at 20°C, and the cells were washed with Ca-HEPES buffer and back diluted to an optical density at 600 nm (OD 600 ) of 1.0 with Ca-HEPES buffer. The cultures of B. bacteriovorus cells were filtered through 0.45-mm filters, and 20 mL of filtered predator cells (2.1 Â 10 7 cells) was mixed with 280 mL of prey suspension (19). Lysis curves (40) were analyzed using Bioscreen C (Automated Growth Curves Analysis system; Growth Curves, USA). The decrease of optical density (OD 600 ) at 30°C was measured at 20-min intervals for 42 h. Experiments were done in three independent biological replicates. The Weibull four-parameters model was fitted to the data using the drc R package to calculate Effective kill time 50% of the host's cells (EKT 50 ) values and their standard errors (41).
DNA manipulation and construction of B. bacteriovorus HD100 strains. DNA manipulations in E. coli were carried out using standard molecular methods (42). Oligonucleotides were synthesized by Sigma-Aldrich. Reagents and enzymes were supplied by Thermo Scientific, Sigma-Aldrich, and NEB. The following B. bacteriovorus HD100 strains were constructed: DnaN-mNeonGreen, DnaN-mCherry, mNeonGreen-ParB/DnaN-mCherry, and FtsZ-mNeonGreen/DnaN-mCherry. In them, the replisomes were labeled green or red by DnaN fusion or the segrosomes and divisomes were labeled green by ParB and FtsZ fusions, respectively (Table S1). In all constructed strains, the gene of interest (dnaN, parB, or ftsZ) was exchanged for a fusion gene that was located in the native chromosomal locus and expressed under its own native promoter(s) (see Table S1).
The coding sequences of the relevant B. bacteriovorus genes (dnaN, parB, or ftsZ) or reporter genes (mNeonGreen or mCherry) were amplified (all primers are listed in Table S1) using chromosomal DNA isolated from B. bacteriovorus HD100 or plasmids (pAKF220-mNeonGreen or p2Nil-lsr2-mCherry), respectively, as the templates. Gibson assembly was used to clone the PCR products into the pK18mobsacB plasmid. The obtained constructs (pK18dnaN-mNeonGreen, pK18dnaN-mCherry, pK18mNeonGreen-parB, and pK18ftsZ-mNeonGreen) were transformed into E. coli S17-1 and conjugated to B. bacteriovorus strain HD100 as described previously (19). Scarless allelic replacements of the wild-type chromosomal dnaN, parB, and ftsZ genes by the gene fusions were performed using two-step homologous recombination with suicide vector pK18mobsacB as described by Steyert and Pineiro (43). Double-crossover (DCO) events were performed on single-crossover (SCO) strains of B. bacteriovorus. SCO strains were grown for 2 days without antibiotics in Ca-HEPES and then for an additional 2 days in Ca-HEPES with 5% sucrose. The obtained mutants were diluted and grown on two-layer YPSC agar plates (1% Broadbeam peptone, 1% yeast extract, 0.5% sodium acetate, 0.25% magnesium sulfate [pH = 7.6]). The final concentrations of agarose in the YPSC plates were 1% in the bottom layer and 0.6% in the top layer. The plates were incubated at 30°C until plaques appeared. DCO strains were searched by plaque PCR screening. The obtained B. bacteriovorus strains harbored gene fusions under the control of the endogenous promoters. Proper construction of all fluorescent strains was verified by DNA sequencing, Western blotting (see Text S1), and fluorescence microscopic observations. Microscopic analysis. Agarose gel (1%) in Ca-HEPES buffer was poured onto a basic slide and allowed to solidify. Lysate (10 mL) of fresh prey (wild-type E. coli, P. mirabilis, S. enterica, or S. flexneri) was added in a layer of agarose gel, and the slide was sealed with a coverslip. Images were taken in phasecontrast mode using a Leica DM6 B microscope equipped with a 100Â/1.32 OIL PH3 objective. Captured images were analyzed using the ImageJ Fiji suite (http://fiji.sc/Fiji).
Time-lapse fluorescence microscopy. Time-lapse fluorescence microscopy (TLFM) was performed on B04A plates with an ONIX flow control system (Merck-Millipore) (44) or on an agarose pad (45) with ibidi cell-imaging dishes (19). For both systems, an overnight stationary-phase culture of prey cells was diluted 25 times and loaded into the flow chamber. Next, B. bacteriovorus cells were injected to the flow chamber (pressure, 4 lb/in 2 ) in Ca-HEPES buffer at 30°C for 2 min or until predatory cells were visible in the field of view. Agarose pads were prepared as described previously (17,25) using P. mirabilis as a prey cell. In both methods, images were recorded every 5 min using a Delta Vision Elite inverted microscope equipped with an Olympus 100Â/1.40 objective and a Cool SNAP HQ2-ICX285 camera. DnaN-mCherry was visualized by detection of mCherry (EX575/25; EM625/45) with an 80-ms exposure time and 50% intensity. Proteins fused with mNeonGreen were visualized by detection of green fluorescent protein (GFP) (EX475/28; EM525/48) with a 50-ms exposure time and 32% intensity. Bright-field images were taken with a 30-ms exposure time and 10% intensity. The captured images were analyzed using the ImageJ Fiji suite (http://fiji.sc/Fiji). TLFM experiments were done in three independent biological replicates.
Statistical analysis and B. bacteriovorus growth simulation. All obtained values and time points were calculated from at least one hundred B. bacteriovorus cells using three independent time-lapse microscopy experiments. We determined using CFU that an OD 600 of 1.0 corresponds to 1.5 Â 10 9 for P. mirabilis, 9.5 Â 10 8 for S. enterica, and 3 Â 10 8 for S. flexneri. For all measured values, the 95% confidence interval (CI) was calculated, assuming a normal distribution of results. Simulations of B. bacteriovorus population growth were done using the R language based on the experimental data concerning the life cycle length and number of progeny cells. The code is available at https://github.com/astrzalka/BdelloSim, and the online version of the application, which allows for simultaneous analysis of three different preys, is available at http://microbesinwroclaw .biotech.uni.wroc.pl:3838/BdelloSim/. TABLE S3, Fig. S8 and S9, and Videos S1 to S9 are available at https://figshare.com/authors/Jolanta _Zakrzewska-Czerwinska/14854612.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. TEXT S1, DOCX file, 0.01 MB.