Aging of a Bacterial Colony Enforces the Evolvement of Nondifferentiating Mutants

Until now, bacterial cells facing nutrient deprivation were shown to enter dormancy as a strategy to survive prolonged stress, with the most established examples being sporulation, stationary phase, and persistence. Here, we uncovered an opposing strategy for long-term bacterial survival, in which mutant subpopulations cope with a challenging niche by proliferating rather than by stalling division. We show that this feature stems from mutations in genes disturbing the capability of the cells to differentiate into a quiescent state, enabling them to divide under restrictive conditions. Our study challenges the dogma of bacterial aging by highlighting an additional survival strategy resembling that of cancerous cells in animal organs.

IMPORTANCE Until now, bacterial cells facing nutrient deprivation were shown to enter dormancy as a strategy to survive prolonged stress, with the most established examples being sporulation, stationary phase, and persistence. Here, we uncovered an opposing strategy for long-term bacterial survival, in which mutant subpopulations cope with a challenging niche by proliferating rather than by stalling division. We show that this feature stems from mutations in genes disturbing the capability of the cells to differentiate into a quiescent state, enabling them to divide under restrictive conditions. Our study challenges the dogma of bacterial aging by highlighting an additional survival strategy resembling that of cancerous cells in animal organs. KEYWORDS Bacillus subtilis, colony, long-term stationary phase, sporulation, spo0A B acteria constantly compete for vital nutrients and yet can persist for long periods once nutrients have been exhausted (1). Various approaches were developed by bacteria to cope with challenging surroundings, with the most common one being residing in stationary phase, characterized by decreased metabolic activity and increased resistance to various stresses (2)(3)(4). After prolonged exposure to starvation, bacteria can enter into long-term stationary phase, which is typically associated with global modulation of gene expression and metabolic activity (5)(6)(7)(8)(9). Furthermore, phenotypes such as growth arrest, increased population heterogeneity, tolerance to antibiotics and oxidative stress, as well as an elevated mutation rate, were attributed to this phase (10)(11)(12)(13)(14)(15). In some bacteria, mainly in Escherichia coli, phenotypic changes during this phase were found to be associated with genetic mutations, conferring a growth advantage in stationary phase phenotype (GASP) (16)(17)(18)(19). Only a few genes have been implicated in the GASP phenotype, including mutations located in genes involved in metabolic pathways, as well as in the stress response transcriptional regulator rpoS, and yet the growth advantage due to rpoS alterations is not fully understood (18,20). A more global analysis of mutations acquired to enable rejuvenation of aging populations is missing.
Upon starvation, the Gram-positive soil bacterium Bacillus subtilis can differentiate into various forms corresponding to stationary phase, competence, or biofilm formation (21)(22)(23)(24)(25). Moreover, B. subtilis and its relatives developed an alternative strategy to resist nutrient deprivation by forming a metabolically dormant cell type known as a spore, allowing the organism to endure harsh environments and to revive when conditions become favorable (26,27). However, the spore rarely acquires new features or undergoes genetic alterations, whereas bacterial cells facing stress can adapt to new conditions and evolve. Intriguingly, all the alternative cellular differentiation pathways in B. subtilis are under the regulation of the same transition state regulator Spo0A. The activity of Spo0A is controlled by a phosphorelay, consisting of the major sporulation kinase KinA that autophosphorylates and transfers the phosphoryl group to Spo0A via the intermediate proteins Spo0F and Spo0B. Each alternative cellular fate is determined by the different phosphorylation level of this transcriptional regulator (28)(29)(30)(31). The perseverance of various strategies to cope with nutrient limitation suggests that coexistence of a variety of differentiated cell types is needed for long-term species durability.
Here, we examined the aging process of cells within B. subtilis colonies, a preferential form of multicellular community, enabling the accommodation of phenotypic and genetic diversity and the independent emergence of mutant clones (see, for example, references 32, 33, 34, 35, 36, and 37). We revealed the appearance of redividing subpopulations harboring mutations in an array of genes, crucial for cellular differentiation, including sporulation, competence, and/or stationary phase. These mutants were found to adopt features, allowing their propagation despite nutrient shortage. We propose that the emergence of rejuvenating populations, appearing during colony aging, emanates from mutants locked in a nondifferentiating state.

RESULTS
Aging bacterial colonies give rise to rejuvenating microcolonies. To initially characterize community features of the aging bacterial colony, we investigated the ratio between spores and nonsporulating cells during the course of 15 days. As expected, the number of spores increased over time; however, even after 15 days of incubation, when nutrients became limited, approximately 30% of the colony inhabitants were nonsporulating cells ( Fig. 1A; see also Fig. S1A to C in the supplemental material). This is consistent with the mixture of spores and dividing cells previously observed in developing B. subtilis colonies that were prompted to sporulate (38). To isolate exclusively the nonsporulating subpopulation, we used a sporulating but nongerminating (NG; RU124) strain, from which only cells that did not enter sporulation could be recovered. Further, assuming that spore germination continuously occurs during aging, such strain reports the total accumulation of spores throughout the colony life span. Interestingly, a portion of nonsporulating cells similar to that of the wild type (WT) was monitored in this strain at all tested time points (Fig. S1D), indicating that germination has a low impact on colony composition. We hypothesized that these nonsporulating cells provide the reservoir needed to enable the population to evolve under harsh conditions. To characterize the nonsporulating population, we followed aging in colonies derived from a nonsporulating mutant strain (NS; RU9), which is able to properly enter into sporulation but stalls at later stages ( Fig. S1C and E) (27,39). Cell division was tracked utilizing a green fluorescent protein (GFP) fusion to the major cell division protein FtsZ (40), and the colony architecture, as well as the morphology of individual cells, was monitored over time. Intriguingly, we visualized the emergence of small microcolonies arrayed over the mother colony, starting at approximately day 15, a phenomenon that was manifested as the colony aged (Fig. 1B, subpanels 1 and 2, and Fig. S2A and B). Concomitantly, along with the accumulation of lysed cells, we observed the frequent reappearance of elongated dividing cell chains harboring FtsZ rings (Fig. 1B, subpanel 3, and Fig. S2C). Two-photon microscopy detected foci of FtsZ-GFP-expressing cells within the depth of the old colony, signifying centers of highly dividing cells that correspond to the emerging microcolonies (Fig. 1C). As a comparison, an evenly distributed FtsZ-GFP signal was observed in a day-old colony (Fig. S2D). Importantly, the formation of microcolonies was also evident in aging colonies derived from NG and WT strains (Fig. S2E). Taken together, these results show that in the course of colony aging, subpopulations regain the ability to divide, leading to the formation of visible microcolonies.
Emergence of morphomutants from aging colonies. The appearance of actively dividing microcolonies suggests that adaptive populations, evolved within the aging colony, could be the source of this occurrence. To enrich and select for the aging colony survivors, we replated 25-day-old colonies from the NS parental strain harboring multiple microcolonies. The analysis uncovered the emergence of progeny colonies, displaying morphologies different from the WT (Fig. 1D and Fig. S3A). Further examination showed that these atypical colonies accumulated over time, reaching 5 to 20% of the population at day 25 ( Fig. 1E and Fig. S3B and C). These altered colonies were mostly associated with microcolony-enriched regions ( Fig. S3B and C), suggesting that they originated from the rejuvenating cells. The appearance of such a morphologically distinct population was also evident in aging colonies stemming from the WT and from NG parental strains (Fig. S3A). Examination of the modified colonies revealed that the morphology was already detected on the first day of plating, indicating the existence of genetic mutations. We therefore termed the mutants yielding these atypical colonies "morphomutants." Identification of genetic mutations acquired by the morphomutants. The identified morphomutants are likely to harbor selective advantage mutations, enabling them to propagate under nutrient deprivation. To uncover the mutated genes, wholegenome sequence (WGS) analysis was carried out on 17 independent isolates derived from the NS (RU9) parental strain, as well as from the NG (RU124) strain, shown to have a spore/nonsporulating cell ratio similar to that of the WT (Fig. S1D), and as such reflects the natural composition of aging colonies. The analysis revealed single or double mutations within the genome of each of the morphomutants, assigned to different genes ( Fig. 2A and Table S1B). Remarkably, further investigation enabled us to sort the majority (12/22 located in 11 strains) of the mutated genes to be part of the Spo0A differentiation pathway (Fig. 2). The remaining mutations (10/22 located in eight strains) were situated in genes required for basic cellular processes of replication, transcription, translation, and metabolism ( Fig. 2A and Table S1B).
The identified mutations belonging to the Spo0A differentiation pathway were located in genes encoding the major Spo0A phosphorelay components (e.g., spo0A, kinA, and spo0F), as well as in the central transition state regulator abrB, and in downstream components required for various Spo0A regulated developmental transitions (e.g., pnp and pbpB) (Fig. 2B). Reconstruction of several of the obtained alleles (i.e., oppB, spo0F, yaaT, and spo0A) in a WT background recapitulated the morphomutant colony morphology, indicating that these alleles cause this distinctive phenotype (Fig. S3D). These results show that morphomutants, deficient in the Spo0A pathway, frequently evolve in aging colonies. Morphomutants are deficient in differentiation. The occurrence of mutations in the Spo0A pathway led to the hypothesis that the evolved morphomutants could be deficient in differentiation, an attribute that may facilitate their division under nutrient restrictive conditions. To investigate this possibility, we examined the ability of the morphomutant strains to differentiate into three major developmental fates: (i) sporulation, by following the formation of polar septa, the hallmark of sporulation initiation (27); (ii) competence, by employing a reporter harboring the promoter of the main competence transcription factor comK fused to gfp (P comK-gfp) (41); and (iii) stationary phase, by monitoring the expression of YlbP-GFP fusion, shown to be elevated during stationary phase (42) (Fig. S4A and B) and, accordingly, in cells of aging colonies (Fig. S4C). An apparent delay in polar septa formation was prominent in 10 of the 11 examined strains upon sporulation induction, with some displaying an almost complete sporulation halt ( Fig. 3A and B and Fig. S5). Consistently, aging morphomutant colonies derived from a sporulating parental strain, which was capable of forming mature spores, exhibited a strong deficiency in spore formation kinetics within the colony, showing very few spores, if any, after 15 days of incubation ( Fig. 3G and H and Fig. S6 and S7). In almost all sporulation-deficient mutants (8/10), a substantial defect in the induction of competence was observed ( Fig. 3C and D). In line with these results, deficiency in entry to stationary phase was prominent in at least 7 of the 11 examined morphomutants (Fig. 3E and F). Taken together, the acquisition of mutations perturbing the Spo0A pathway yielded a group of morphomutants harboring differentiation deficiencies, with the majority (10/11) displaying measurable defects in sporulation, competence or stationary phase destinies. These results reflect the existence of a general strategy to tolerate colony aging by the acquisition of mutations that reduce the efficiency to differentiate.
Morphomutants mutated in genes involved in basic cellular processes are deficient in differentiation. Apart from mutants in the Spo0A pathway, the morphomutants harbored mutations in genes related to basic cellular functions ( Fig. 2A and Table S1B). Interestingly, several genes involved in elementary cellular processes were found to impact bacterial differentiation (see, for example, references 43 and 44). We therefore hypothesized that this group of morphomutants could be deficient in differentiation, similarly to those with mutations in the Spo0A pathway. Testing the ability of these strains to sporulate revealed that five of six exhibited a marked delay in initiating sporulation ( Fig. 4A and B and Fig. S8). Moreover, morphomutant colonies derived from a sporulating strain showed a strong deficiency in spore formation during aging, with very few spores observed after 15 days of incubation ( Fig. 4C and D). Lastly, although no significant effect on the transition to competence was detected in these mutants (data not shown), they displayed slower kinetics of entry into stationary phase than the parental strain (Fig. 4E). These results support the view that a plethora of mutations that develop during aging lead to the same outcome of evading differentiation.
Advantageous features of morphomutant aging colonies. To further explore the morphomutants' advantage during aging, we compared parental and mutant survival over time. Four representative morphomutant strains (oppB, spo0F, yaaT, and spo0A) displaying growth kinetics similar to that of the parental strain (Fig. 5A) were used for  Aging Colony Yields Nondifferentiating Mutants this analysis. A sharp increase (ϳ9-fold) in CFU was detected for all tested morphomutants during aging compared to the parental strain (Fig. 5C), reinforcing the view that the morphomutants are capable of dividing efficiently under restrictive conditions. Next, we monitored growth resumption of cells originating from aging colonies of morphomutant strains. Morphomutant cells deriving from 8-day-old colonies and further displayed a shorter lag phase, rapidly resuming growth compared to parental cells of similar age (Fig. 5A and B and Fig. S9A). Notably, the extent of the lag phase was not significantly influenced by the initial cell number (Fig. S9B). To precisely pursue this phenotype at a single cell resolution, we conducted a series of time-lapse microscopy experiments, where cells of an 8-day-old morphomutant colony were mixed with differentially labeled parental cells of the same age, and their return to growth was monitored. Indeed, morphomutant cells showed an apparent earlier growth resumption compared to the equivalent parental cells (Fig. 5D and E). Thus, we conclude that the examined aging morphomutants gained the capacity to both increase survival and shorten their lag phase when reinitiating growth.

DISCUSSION
Based on our findings, we propose that cells residing in aging colonies, coping with nutrient exhaustion, accumulate mutations in genes crucial for bacterial differentiation pathways, enabling their division under unfavorable conditions and the subsequent formation of microcolonies (Fig. 6). Thus, as an alternative to entering dormancy in order to avoid stress (see, for example, references 1, 15, 27, and 45), this trajectory allows bacteria to proliferate in challenging niches. By analogy to animal cells, these mutants resemble cancerous cells that accumulate mutations resulting in uncontrolled cell division and tumor formation (46). Temporarily, such bacterial mutants have a growth advantage, since they can rapidly resume division and flourish, whereas their neighboring cells are paused in a given differentiation state. Naturally, this rejuvenating phenotype would necessitate access to nutrients, which could be provided to the survivors by their perished siblings (7,38,47).
Although the leverage of overgrowing mutants is fairly intuitive in the immediate time scale, their long-term evolutionary advantage is ambiguous, since they possess fewer crucial features required for coping with stress, including spore formation, competence, and entering stationary phase. Furthermore, the advantage of population heterogeneity, i.e., cells exhibiting various fates, a concept standing at the base of evolution (see, for example, references 21, 48, and 49), is weakened in these nondifferentiating mutants. We therefore assume that such mutated strains might be unable to maintain species continuity in natural settings, and we hypothesize that nondifferentiating mutations repetitively arise in natural populations but revert or vanish in challenging niches. In agreement with this view, occasionally replating the morphomutants resulted in the appearance of repressors, reverting the altered shape phenotype of their parental morphomutant strain, thus pointing toward a potential drawback of the original mutation.
Only a few genes have been implicated in GASP phenotypes. Apart from rpoS, mutations accumulating during bacterial aging were assigned in E. coli to genes involved mainly in amino acid metabolism. This includes mutations in the leucine responsive regulatory gene lrp; in the gene cstA, which encodes an oligopeptide permease; and in the ybeJ-gltJKL operon, predicted to encode a high-affinity aspartate and glutamate ABC-type transporter (18). Similarly, Pseudomonas fluorescens Pf0-1 colonies were shown to generate cells that acquired mutations within the rsmE locus, regulating a pathway implicated mainly in metabolism and host colonization (35). In line with this view, prolonged incubation of B. subtilis cultures yielded adaptive isolates Aging Colony Yields Nondifferentiating Mutants ® exhibiting alterations in the transcriptional profile of several metabolic pathways and suggesting that a key pathway in acquiring adaptive features is by harboring mutations in metabolic genes (8). The latter mutations were proposed to enable the cells to more efficiently rely on alternative sources of energy and biosynthetic substrates, including harboring both ftsZ-gfp and P rrnE -gfp (diffused pattern) fusions, and representative morphomutants harboring solely the ftsZ-gfp fusion were collected, mixed at a 1:1 ratio, and replated on the same LB agar pad. The cells were monitored using time-lapse microscopy. Of note, mCherry and CFP caused growth perturbation after aging and thus could not be utilized to differentiate between the mixed strains; therefore, P rrnE -gfp was added to the parental strain. (D) Phase-contrast (blue) and GFP (green) overlay images of the NS (RU23) parental strain and the spo0F (upper images) and yaaT (lower images) morphomutants. The legend shows the predicted GFP pattern from the parental (diffused GFP) and morphomutants (FtsZ-GFP rings). (E) Growing cells were counted, and their ratios in the entire cell population were calculated. The results are presented as average values and the SD from three independent experiments. their siblings' remnants and internal stores (18). Here, we revealed an alternative pathway to overcome nutrient limitation by losing differentiation capabilities.
The notion that cells residing in long-term stationary phase are prone to accumulate mutations that contribute to strain durability is well established (see, for example, references 50 and 51). Diverse studies suggest various mechanisms to explain the source of the high mutation rate observed. Some studies support the view that this increase is caused by induced mechanisms in response to a specific selective pressure (see, for example, references 13 and 14), while others provide evidence that this elevation in adaptive mutations is solely due to advantageous features of the emerging mutants (see, for example, references 52 and 53). In this regard, we found that 5 to 20% of the progeny population of 25-day-old colonies displayed altered colony phenotypes. We further showed that all the sequenced morphomutants contained at least one mutation in their genome. It should be noted that we did not sequence the entire progeny population, and therefore we could not assess the precise mutant ratio. Nevertheless, it is reasonable to assume that the actual mutation rate could be even higher than the proportion of the morphomutant colonies in the population.
Interestingly, mutations in various genes, even in those that were not classically assigned to developmental pathways, led to the observed differentiation deficiency, indicating the existence of multiple paths to reach a similar consequence. The ease of obtaining such mutations suggests that they are frequent in nature, allowing bacterial growth under restrictive conditions. In line with this view, the prevalence of GASP mutations located within rpoS, which is involved in the transition to stationary phase in E. coli (18,20), suggests that they could act similarly, reducing entry into the stationary state. This assumption, however, should be tested experimentally.
Thus far, bacterial cells facing nutrient deprivation have been shown to enter dormancy as a strategy to avoid stress. Numerous examples have been previously described, including sporulation and stationary phase, as well as persister cells, known to survive antibiotic treatments due to temporarily residing in a nondividing state (11,15,45). Here, we uncovered an opposing strategy in which mutant cells cope with a challenging niche by proliferating rather than stalling division. This strategy could be a major factor affecting the dynamics of bacterial populations in natural environments.

MATERIALS AND METHODS
Bacterial strains and plasmids. The B. subtilis strains examined here were derivatives of the WT PY79 (54) and are listed in Table S1A and B; the primers are listed in Table S1C.

FIG 6
Model of the aging process of a B. subtilis colony. The colony is composed of different cell types, including spores, stationary-phase cells, and competent cells, all displaying a low growth rate. Over time, mutants in entering differentiation evolve and continue to divide under nutrient deprivation, giving rise to visible microcolonies. amplified using chromosomal DNA purified from PY79 as the template (Table S1A). The PCR product was digested with BamHI and SalI and cloned into the BamHI and SalI sites of pMINImad2. Site-directed mutagenesis was conducted on this plasmid to change the codon encoding the WT allele to the mutated allele using primers containing the desired mutation (Table S1C) and a QuikChange II kit (Stratagene) to create the desired plasmids. The plasmids were sequenced to ensure the presence of the mutation. Each mutated plasmid was then introduced into PY79 genome via a single-crossover integration by transformation at the restrictive temperature for plasmid replication (37°C), using mls resistance as a selective marker. To evict the plasmid, the strain was incubated in 3 ml of LB medium at a permissive temperature for plasmid replication (23°C) for 14 h, diluted 30-fold into fresh LB medium, and incubated at 23°C for another 8 h. The cells were then serially diluted and plated on LB agar at 37°C. Individual colonies were patched onto both LB plates and LB plates containing mls to identify mls-sensitive colonies that had evicted the plasmid. Chromosomal DNA from colonies that had excised the plasmid was purified and screened by PCR to determine which isolate had retained the mutated allele. Sequence analysis was used to validate allele replacement.
Plasmid construction. pET770 (ylbP-gfp-spc) was constructed by amplifying the C terminus of ylbP from gDNA of the B. subtilis strain (PY79) with the primers ylbP-U-EcoRI and ylbP-L-XhoI (Table S1C). The PCR-amplified DNA was digested with EcoRI and XhoI and was cloned into pKL147 (61) digested with the same enzymes.
Estimation of ratio variance. The ratios were defined as X / Y , and the variance (Var) was estimated using the following formula: The standard deviation (SD) was calculated to be the square root of the variance.