Exopolysaccharide defects cause hyper-thymineless in Escherichia coli via massive loss of chromosomal DNA and cell lysis

T. V. Pritha Raoa and Andrei Kuzminova,1

aDepartment of Microbiology, University of Illinois at Urbana–Champaign, Urbana, IL 61801

Edited by Thomas J. Silhavy, Princeton University, Princeton, NJ, and approved November 10, 2020 (received for review June 17, 2020) Thymineless death in Escherichia coli thyA mutants growing in the magnitude during T-starvation (8), indicating some other absence of thymidine (dT) is preceded by a substantial resistance yet-to-be-identified major lethality factors. phase, during which the culture titer remains static, as if the chro- Since actively growing cells continuously require a lot of dT to mosome has to accumulate damage before ultimately failing. Sig- replicate chromosomal DNA, existing replication forks were nificant chromosomal replication and fragmentation during the inferred to be the points of TLD pathology (7, 8, 13–15). Indeed, resistance phase could provide appropriate sources of this dam- T-starvation severely inhibits chromosomal DNA replication age. Alternatively, the initial chromosomal replication in thymine (15) and is associated with accumulation of single-stranded (T)-starved cells could reflect a considerable endogenous dT DNA, suggesting generation of single-strand (ss) gaps by source, making the resistance phase a delay of acute starvation, attempted replication in the absence of dT (7, 16). These ss-gaps rather than an integral part of thymineless death. Here we identify such a low-molecular-weight (LMW)-dT source as mostly dTDP- induce the SOS response (7, 8, 17), which contributes to the glucose and its derivatives, used to synthesize enterobacterial pathology of TLD by induction of the SulA cell division inhibitor common antigen (ECA). The thyA mutant, in which dTDP-glucose (8). Also, replication initiation spike in the T-starved cells trig- production is blocked by the rfbA rffH mutations, lacks a LMW-dT gers the destruction of the origin-centered chromosomal sub- pool, the initial DNA synthesis during T-starvation and the resis- domain during TLD, suggesting that it is the demise of the

tance phase. Remarkably, the thyA mutant that makes dTDP- nascent replication bubbles, rather than the existing replication MICROBIOLOGY glucose and initiates ECA synthesis normally yet cannot complete forks, that eventually kills the chromosome (15, 17). it due to the rffC defect, maintains a regular LMW-dT pool, but Although the thyA mutants cannot synthesize dT, they grow cannot recover dTTP from it, and thus suffers T-hyperstarvation, normally if supplemented with exogenous dT/T. Upon removal dying precipitously, completely losing chromosomal DNA and of dT from the growth medium, the E. coli thyA strain has a eventually lysing, even without chromosomal replication. At the two-generation-long resistance phase (also called the lag phase) + thyA same time, its ECA parent does not lyse during T-starvation, (1), when the colony-forming unit (CFU) titer of the culture while both the dramatic killing and chromosomal DNA loss in the stays constant (Fig. 1 A, Top). This is followed by the rapid ex- ECA-deficient thyA mutants precede cell lysis. We conclude that: 1) the significant pool of dTDP-hexoses delays acute T-starvation; 2) ponential death (RED) phase, when the CFU titer falls by ap- T-starvation destabilizes even nonreplicating chromosomes, while proximately three orders of magnitude within several hours T-hyperstarvation destroys them; and 3) beyond the chromosome, (Fig. 1 A, Top). T-hyperstarvation also destabilizes the cell envelope. Significance enterobacterial common antigen | chromosome fragmentation | chromosome replication | dTDP-glucose | cell lysis The thyA mutants of Escherichia coli cannot synthesize their own thymidine (dT) and rapidly die without dT supplementa- cute starvation for thymidine triphosphate (dTTP), one of tion. Confusingly, this thymineless death has an integral re- Athe four precursors for DNA synthesis, is lethal in both sistance phase, as if chromosomal damage has to accumulate bacterial and eukaryotic cells (1). Following a short resistance before eventually becoming irreparable. Here we show that phase, the rapid death of thyA auxotrophs in media lacking during the resistance phase, the thyA mutants survive by de- thymine or thymidine (“T-starvation”) known as thymineless riving dT from a significant pool of endogenous dTDP-sugar death (TLD) was first described in Escherichia coli (2, 3) and conjugates. Inability to synthesize dTDP-sugars shortens the since then was extensively studied to identify the cause of le- resistance phase, while inability to recover dTTP from this pool thality (1, 4, 5). Because the bulk of thymidine (dT) in any cell is eliminates the resistance phase altogether. Moreover, the resulting dT-hyperstarvation causes catastrophic chromosome used for chromosomal DNA synthesis, lack of dT was always loss and cell lysis. We conclude that the resistance phase, rather assumed to cause some form of chromosomal damage, and than being a part of thymineless death, delays acute starva- hence the role of DNA repair pathways during T-starvation was – tion, while dT-hyperstarvation has the second, cell envelope the focus of intense investigation (6 9). These studies revealed dimension. that certain pathways, like double-strand break repair initiated by the RecBCD helicase/nuclease, Holliday junction resolution Author contributions: T.V.P.R. and A.K. designed research; T.V.P.R. performed research; by RuvABC, and antirecombination activity of the UvrD heli- T.V.P.R. and A.K. analyzed data; and T.V.P.R. and A.K. wrote the paper. case, keep cells alive during the resistance phase of T-starvation. The authors declare no competing interest. Other events, like attempted single-strand gap repair initiated by This article is a PNAS Direct Submission. the RecFOR complex, the function of the RecQ helicase and Published under the PNAS license. RecJ exonuclease, and SOS induction of the cell division inhib- 1To whom correspondence may be addressed. Email: [email protected]. – itor SulA, are detrimental for T-starved E. coli cells (8, 10 12). This article contains supporting information online at https://www.pnas.org/lookup/suppl/ However, the thyA mutants of E. coli inactivated for all of the doi:10.1073/pnas.2012254117/-/DCSupplemental. latter “toxic DNA repair pathways” still die by two orders of First published December 14, 2020.

www.pnas.org/cgi/doi/10.1073/pnas.2012254117 PNAS | December 29, 2020 | vol. 117 | no. 52 | 33549–33560 Downloaded by guest on September 29, 2021 Fig. 1. A significant endogenous pool of LMW dT decreases during T-starvation. (A) The phenomenon of TLD in E. coli suggests accumulation of chro- mosomal damage during the resistance phase (green) that would later kill cells during the RED phase (red). The data are adapted from ref. 16. Henceforth, the data are means (n ≥ 3) ± SEM. Cultures were grown at 37 °C in the presence of dT, which was removed at time = 0, while incubation in the growth medium continued. Top, begins after 1-h-long resistance phase, during which the culture titer is stable. Bottom, during the same first hour without dT, cells manage to synthesize the amount of DNA equal to half of what they already had before dT removal. However, during the RED phase genomic DNA is gradually lost. (B) Scheme of 50% methanol fractionation of the intracellular thymidine into HMW dT (the dT content of the chromosomal DNA) and LMW dT. (C) A 0.7% agarose gel analysis of the HMW and LMW fractions of the 50% methanol-treated cells, as well as pure LB treated the same way, for DNA and RNA content (staining with ethidium bromide). Inverse images of stained gels are shown, in which the indicated samples were either incubated in the bufferor with the indicated enzyme (DNase I for the top gel, RNase A for the bottom gel). (D) The size of the LMW-dT pool, normalized to the HMW-dT content of the chromosome, either during normal growth in dT-supplemented medium or during T-starvation. Thymidine was removed at time = 0. The strain is KKW58.

An obvious explanation for the resistance phase is existence of exhausted. In other words, the resistance phase could simply post- an intracellular source of dT to support slow replication; how- pone TLD, rather than being an integral part of it. Previously, we ever, chromosomal DNA amount was consistently reported to have tested the two obvious high-molecular-weight (HMW) dT remain flat during TLD (15, 18–20). Besides, the recent sys- sources, namely, the stable RNAs and the chromosomal DNA, but tematic test of potential candidates for a source of dT or its found that incapacitation of neither one reduced the resistance analogs supporting the resistance phase returned empty-handed phase or precluded the early DNA synthesis during T-starvation (16). Thus, the mechanisms behind the initial resistance to (16). Thus, the question of whether the resistance phase is a part T-starvation, followed by the sudden shift to the RED phase of the TLD phenomenon remains unresolved. remain unclear, leading to a reasonable assumption that the In the current study, we investigated a seemingly remote resistance phase is an integral part of the TLD phenomenon, possibility of a substantial low-molecular-weight (LMW)-dT pool during which chromosomal damage accumulates until it becomes supporting the resistance phase of T-starvation in E. coli. While irreparable, ushering the RED phase (1, 5). Specific early events the bulk of dTTP in E. coli immediately incorporates into the during the resistance phase of TLD that would later turn poi- chromosomal DNA, a fraction of dTTP is recruited into the sonous during the RED phase were proposed to be futile dTDP-hexose pool (23), to participate in the synthesis of the incorporation–excision cycles (1, 21), ss-gap accumulation caus- exopolysaccharide (EPS) capsule, made of core lipopolysaccha- ing the SOS induction (1, 8, 16), futile fork breakage-repair cy- ride (LPS) (24), O-antigen (OA) (25), and enterobacterial cles (16, 22), and overinitiation from the origin (5, 15). common antigen (ECA) in E. coli (26). The first step of this Two recent observations, in combination with an old popular recruitment is to conjugate dTTP with glucose; the hexose TLD explanation, further support the idea of the resistance moiety of the resulting dTDP-glucose then undergoes several phase as the TLD period during which chromosomal damage ac- modifications, before eventually incorporating into oligosaccha- cumulates without affecting viability for the time being. First, the ride precursors of the outer antigens, while the activating dTDP resistance phase coincides with accumulation of double-strand handle is released back into the DNA precursor pools (26). We breaks in the chromosome, which then paradoxically disappear ignored LMW dT before because, if the total dT content of the during the RED phase (7, 16). Second (and in contrast to the re- chromosomal DNA is taken for 100%, the pool of dTTP con- ports mentioned above of constant chromosomal DNA amount stitutes ∼0.7% of it, while dTDP-glucose (unresolved from other during T-starvation) (15, 18–20), we have found that during the dTDP-hexoses?) adds only another 2.4% (27). No more LMW- resistance phase the amount of the chromosomal DNA actually dT species are known in the cell, so the total expected LMW-dT increases ∼1.5 times over the prestarvation level, but then the pool comes to ∼3% of the total dT content of the chromosomal chromosomal DNA is apparently destabilized during the RED DNA, not nearly enough to support the resistance phase with its phase, since it is slowly reduced to the original level (16) (Fig. 1 A, ∼50% increase in the chromosomal DNA mass (Fig. 1A). Bottom). Therefore, both the apparent chromosomal replication To investigate the role of the LMW-dT pool in TLD, we and the significant chromosome fragmentation during the resistance started by developing a simple protocol to extract the LMW-dT phase could lead to accumulation of chromosomal damage (SOS pool from growing cells and to compare it to the (HMW) induction is an indicator of this accumulation) (7). chromosomal dT content. Here we show that early on during On the other hand, the early DNA synthesis and the resistance T-starvation the pool of dTDP-sugars becomes the major source phase in T-starved cells could reflect the existence of a source of of dTTP for the chromosomal DNA replication. This unexpected dT, available early on during T-starvation, that fuels the initial rebalancing of the dTTP pool with the help of cell envelope DNA accumulation and delays viability loss until this pool is delays TLD and prevents T-hyperstarvation, a

33550 | www.pnas.org/cgi/doi/10.1073/pnas.2012254117 Rao and Kuzminov Downloaded by guest on September 29, 2021 significantly more lethal phenomenon accompanied by complete 3H-dT labeling (28). The LMW 3H-dT is separated from HMW chromosome destruction and cell lysis. 3H-dT by permeabilizing cells with cold 50% methanol to facil- itate efficient metabolite extraction (a standard metabolome Results isolation protocol) (29), while trapping HMW species (proteins, E. coli Possesses a Substantial LMW-dT Pool. As explained in the nucleic acids) inside the cell envelope (Fig. 1B and SI Appendix, introduction, we expected the overall LMW-dT pools in E. coli Fig. S1A). Upon subsequent filtration, the HMW-dT of the to be ∼3% of the total dT content of the chromosomal DNA. To chromosomal DNA is retained on the filter within cells, while the facilitate comparison of the overall LMW-dT pool inside the LMW-dT pool is collected as the flow-through (subsequently cells to the chromosomal DNA dT content of the same cells, we also dried on the filter, to equalize the counting conditions). utilized the standard “3H-dT incorporation into the chromo- Agarose gel electrophoresis and DNase/RNase digestion reveal somal DNA” technique, but in addition to collecting labeled cells chromosomal DNA and long (mostly ribosomal) RNA in the on the filter (HMW dT), we also collected and quantified the HMW (filter) fraction, while revealing no DNA and some short released LMW flow-through and then related its 3H label to the RNA in the LMW (flow-through) fractions, most of this short 3H-DNA label retained on the filter. RNA material coming from the growth medium itself (Fig. 1C). To chronically label thyA mutants with 3H-dT, we grew these Taking the chromosomal dT content (HMW dT) for 100%, we cells in lysogeny broth (LB), as LB contains enough dT to sup- found that in the exponentially growing thyA mutant cultures just port limited growth of thyA mutant, but not enough to prevent before T-starvation the size of the LMW-dT pool equals ∼60% MICROBIOLOGY

Fig. 2. The T-hyperstarvation in the thyA rfbA rffHC and thyA rffHC mutants leads to cell lysis. (A) Scheme of dTTP pool partitioning into the chromosomal DNA (by DNA synthesis) vs. the pool of dTDP-glucose (by RfbA and RffH dTDP-glucose pyrophosphporylases). dTDP-glucose is converted into precursors for the EPS-capsule synthesis, while dTDP is returned to the DNA precursor pool by RffT. The rffG, wecE, rffC, and rffT genes are specific to the ECA-synthesis pathway. (B) Spotting of serial dilutions of thyA (KKW58), thyA rfbA (RA48), thyA rffH* (RA49), and thyA rfbA rffH* (RA51) mutants on LB ± dT ± 1.5% SDS. The seqA mutant (ER15) spotting is shown as a control for SDS sensitivity. (C) The T-starvation hypersensitivity phenotype of the thyA rffHC mutant is due to the rffC defect, rather than rffH defect. The thyA mutant strains are: Rff+, KKW58; rffC, RA57; rffH, RA60; and rffHC, RA50. (D) Cell lysis, measured as leakage of the cytoplasmic enzyme beta-galactosidase out of the cell, in liquid M9CAA cultures in the absence of dT. Thymidine was removed at time = 0. Strains are as in B.(E) DAPI staining of the strains in B undergoing T-starvation for 0, 1, 3, or 20 h.

Rao and Kuzminov PNAS | December 29, 2020 | vol. 117 | no. 52 | 33551 Downloaded by guest on September 29, 2021 of the chromosomal DNA dT content instead of the expected We want to stress that the thyA rfbA, the thyA rffH*, and the ∼3% (Fig. 1D and SI Appendix, Fig. S1B). The parental ThyA+ thyA rfbA rffH* mutants all grew without problem and obvious strain has the same size of the LMW-dT pool and starts shrinking phenotypes on LB +dT (Fig. 2B). ОnLB−dT, thyA and thyA it as the cultures grow slower at higher densities (SI Appendix, rfbA mutants also grew normally (LB has enough of its own dT), Fig. S1C). and thyA rfbA rffH* mutant formed smaller colonies, as expected In order to see whether this 20-fold higher-than-expected pool from its supposedly defective envelope. However, the (RfbA+) of endogenous LMW dT is accessible to support chromosomal thyA rffH* mutant surprisingly struggled (Fig. 2B), as if RffH was replication during T-starvation, we followed its changes in the more than just a redundant paralog of RfbA, or the mutant had T-starved cells. At the end of the resistance phase of an additional defect causing hypersensitivity to T-starvation. T-starvation, the LMW-dT fraction falls to ∼20% of the chro- mosomal dT content (Fig. 1D), suggesting that at least two-thirds The thyA rffC Mutant Is Hypersensitive to T-Starvation. The extreme of this LMW dT is presumably utilized for chromosomal repli- sensitivity of the thyA rffH* and thyA rfbA rffH* mutants to de- cation during this period, which may explain the DNA accu- tergents, further confirmed by their poor plating on MacConkey mulation and constant viability during the resistance phase of agar ±dT (SI Appendix, Fig. S2) (MacConkey instead of SDS T-starvation (Fig. 1A). This fraction of LMW dT remains at contains bile salts), and the hypersensitivity of the thyA rffH* similar levels thereafter (Fig. 1D). In the ThyA+ parent, the mutant to T-starvation made us look for possible downstream 3 LMW H-dT is similarly incompletely chased with unlabeled effects of our precise rffH deletion. It turned out that the endogenous dT (SI Appendix, Fig. S1D), showing that at least downstream rffC gene overlaps the upstream rffH by 22 bp, so a ∼70% (and maybe up to 95%, see below) of dT can be extracted complete rffH deletion also removes the beginning of the rffC out of the LMW-dT pool in E. coli. gene, making it effectively rffHC two-gene inactivation (SI Ap- pendix, Fig. S3A). In fact, deletion of rffC makes the thyA mutant thyA The Mutants Lacking dTDP-Glucose Develop Envelope Stress in grow poorly on LB −dT, just like the original ΔrffHC (=rffH*) the Absence of dT. Our finding of the considerable LMW-dT pool allele does (SI Appendix, Fig. S3B, the “vector” variants). in E. coli and its two-thirds reduction during the resistance phase Moreover, complementation of the rffHC and rffC mutants with of T-starvation raised the question about its nature. It could not plasmids carrying either rffH+ or rffC+ genes confirms that it is be mostly dTTP, as the DNA precursor pools are reported to be the rffC+ plasmid, rather than rffH+ plasmid, that complements around 1% of the nucleotide contents of the chromosomal DNA the T-hyperstarvation of both mutants (SI Appendix, Fig. S3B). (27). What remained was the pool of dTDP-hexoses and the Finally, in contrast to T-hyperstarvation of the original thyA unknown pool of derived oligosaccharide precursors of the outer rffHC or the new thyA rffC mutations, the shorter rffH deletion antigens (23). that does not disrupt the rffC gene (SI Appendix, Fig. S3A) allows E. coli uses dTDP-hexoses in the synthesis of OA and the thyA mutant to grow without problem on LB −dT (Fig. 2C), ECA—the two constituents of the EPS capsule of gram-negative establishing the rffC defect as the one responsible for the hy- cells (25, 26). Two supposedly redundant dTDP-glucose pyro- persensitivity to T-starvation of the rffHC mutants. phosphorylase enzymes (see for example ref. 30), encoded by The product of the rffC gene is dTDP-fucosamine acetyl- paralogous rfbA and rffH genes in E. coli (31), synthesize dTDP- glucose, which is the first intermediate for dTDP-conjugated transferase that catalyzes the penultimate step in the dTDP- hexoses to be integrated in both OA and ECA (Fig. 2A) (32, hexose modification (Fig. 2A) before the modified hexose joins 33). E. coli K12 background that we use lacks OA and synthe- the trisaccharide repeats (ECA lipid III), from which ECA itself sizes only ECA (34); after several modifications the hexose part is assembled, and so the rffC mutant accumulates its precursor, of the conjugate is deposited on the growing ECA-sugar chain in ECA lipid II (35) (see Fig. 5A below). Our further analysis has the inner membrane, while dTDP is phosphorylated to dTTP revealed that the rffH mutation has no T-starvation phenotypes, and returns to the DNA precursor pools (Fig. 2A) (35). The unless combined with the rfbA defect; for example, the thyA rfbA − EPS-capsule synthesis is a significant endeavor in growing cells rffH combination grows slowly on LB dT and is additionally (the dry weight of the combined lipopolysaccharide exceeds the inhibited by SDS (SI Appendix, Fig. S4). At the same time, the one of the chromosome) (36) and apparently demands an ade- complete inability of the double rfbA rffH mutant to make dTDP- quate pool of dTDP-hexoses, potentially explaining the sub- glucose blocks the pathway in which RffC later acts (Fig. 2A), stantial size of the LMW-dT pool we detect. making the double rfbA rffH defect epistatic to (masking) the The rfbA and rffH mutations that we P1-transduced into our thyA rffC mutant hypersensitivity to T-starvation (compare thyA mutant were precise deletions from the Keio collection Fig. 2B and SI Appendix, Fig. S4). Since by the time we have (37). We later discovered that the rffH deletion allele had an established it is the rffC defect that causes hypersensitivity to inadvertent second mutation; therefore, until its nature is T-starvation of the original thyA rffH* mutant, the bulk of our revealed in the next section, we will refer to this allele as rffH*. data were already collected with the thyA rffHC and thyA rfbA The complete defect in dTDP-glucose production should de- rffHC strains, we continue to present these results, verifying the stabilize the EPS capsule, as even (ThyA+) rfbA and rffH single key findings in the thyA rffC (RffH+) and thyA rfbA rffH (RffC+) mutants already exhibit sensitivity to surfactants, like sodium mutants. In interpreting these results, it helps to remember that dodecyl sulfate (SDS) and bile salts (32, 38). However, on LB whenever RfbA is functional, the thyA rffHC mutant displays +SDS +dT, all four combinations in the ΔthyA background grew TLD phenotypes of the thyA rffC mutant, whereas whenever almost equally well (Fig. 2B), indicating no obvious envelope RfbA is inactivated, the resulting thyA rfbA rffHC mutant dis- problems (while the seqA mutant, our positive control for SDS plays TLD phenotypes of the thyA rfbA rffH mutant. sensitivity, was characteristically inhibited) (39). In contrast, on LB +SDS −dT, the thyA and especially thyA rfbA mutants were The thyA rffHC and thyA rfbA rffHC Mutants Lyse during T-Starvation. visibly inhibited, as if the limited dT affected their envelope The thyA mutants undergo one cell division during T-starvation, stability, while both thyA rffH* and thyA rfbA rffH* could not revealed by direct cell count under a light microscope (16). In form colonies (Fig. 2B), indicating that exogenous dT was critical contrast, the thyA rfbA, thyA rffHC, and thyA rfbA rffHC mutants for their envelope stability. We conclude that, in the absence of failed to divide even once, when incubated in liquid cultures dT, all thyA mutants become sensitive to SDS, suggesting cell without dT (SI Appendix,Fig.S5). Also, we noticed that the envelope vulnerability, while the defect in dTDP-glucose pyro- thyA rffHC and thyA rfbA rffHC mutant cells appeared phosphorylases further increases this sensitivity. “lighter” in phase contrast at later time points, and some of

33552 | www.pnas.org/cgi/doi/10.1073/pnas.2012254117 Rao and Kuzminov Downloaded by guest on September 29, 2021 the cells exhibited what looked like segmental cytoplasm voids Fluorescent microscopy of DAPI-stained cells detects gross (see below). cytoplasm, envelope, and chromosome stability issues. When To test for possible cell lysis in T-starved cultures of the four grown in the presence of dT, thyA mutant cells were expectedly mutants, we measured the extracellular presence of the cyto- small and dividing normally (Fig. 2 E, Top row). After 1 h of plasmic enzyme beta-galactosidase (40, 41). We detected no or T-starvation, the cells became ∼1.5 times wider and 2 to 3 times little cell lysis in the thyA and thyA rfbA mutants (at least during more elongated, with centrally positioned bright nucleoids. After the first 6 h of T-starvation), but a significant cell lysis within 2 h 3 h of T-starvation, the thyA mutant cells mostly kept their new of T-starvation in the thyA rffHC mutant (Fig. 2D) (at the peak width, but became 4 to 6 times longer than nonstarved cells, with amounting to ∼50% of the total cellular beta-galactosidase centrally positioned single nucleoid and somewhat mottled cy- content, see below). As expected from the pathway configura- toplasm (developing lighter segments), but still no signs of cell tion (Fig. 2A), this lysis was partially suppressed in the thyA rfbA lysis (Fig. 2E), similar to the report before (42). Others report rffHC mutant (Fig. 2D), the incomplete nature of the suppres- that live cell imaging of E. coli strains defective in their envelope sion reflecting the sensitivity to T-starvation of the double rfbA maintenance present the retraction of cytoplasm as a highly rffH defect itself (SI Appendix, Fig. S4). As we detected no phage dynamic event, accompanied by vesicle release (43, 44). Inter- in the supernatants of the T-starved thyA rffHC mutant cultures, estingly, after 20 h of T-starvation, the thyA mutant cells devel- this cell lysis was not due to prophage induction. We conclude oped sickly appearance: they were further elongated, with that the loss of the RffC enzyme makes the thyA mutant cell segmental cytoplasm loss in at least half of the cells, with no envelope hypersensitive to T-starvation; in contrast, the loss of DNA in the majority of cells and with little DNA remaining in a both RfbA and RffH enzymes sensitizes cells to T-starvation to a few still centrally located surviving nucleoids (Fig. 2 E, Top row). lesser extent, while at the same time suppressing the rffC defect However, T-starved cultures would be long stabilized at that time by shutting down the ECA-biosynthesis pathway (Fig. 2A). in the “survival” phase (7, 16), so these dramatic changes were likely reflecting metabolic of long-dead cells. We DAPI Staining Visualizes the Second Dimension of TLD at Three conclude that, during the RED phase (2 to 5 h under our con- Severity Levels. The dramatic cytoplasm and cell envelope insta- ditions), T-starved thyA mutant cells develop some cytoplasm bility phenotypes of the thyA rfbA rffHC and thyA rffHC mutants irregularities, but no cell lysis or even periplasm leakage. How- during T-starvation made us wonder whether the thyA single ever, this was not true for the rfbA rffHC and rffHC mutants. mutant undergoes similar changes, perhaps on a lesser scale. When grown in the presence of dT, cells of all four strains Still, the test for the periplasm instability using alkaline phos- (thyA, thyA rfbA, thyA rffHC, and thyA rfbA rffHC) were similarly phatase (40, 41) showed no periplasm leaking in the T-starved small and dividing normally (Fig. 2 E, Leftmost column). The MICROBIOLOGY thyA mutants; as a positive control, a massive release of this thyA rfbA mutant progressed through T-starvation similar to the periplasmic enzyme was evident in the T-starved thyA rffHC thyA single mutant, although cells appeared more mottled at 3 h mutant cells (SI Appendix, Fig. S6). of T-starvation, and half of the cells turned to “” completely

Fig. 3. RfbA and RffH build the LMW-dT pool that supports DNA synthesis during the resistance phase of TLD. The strains are thyA (KKW58), thyA rfbA (RA48), thyA rffH (RA60), thyA rfbA rffHC (RA51), and thyA rfbA rffH (RA65) mutants. (A) The evolution of LMW-dT pools in the thyA rfbA, thyA rffH,and thyA rfbA rffH mutants during T-starvation. Thymidine was removed at time = 0, after which aliquots were taken for the “0-h” cultures. Values for thyA mutant from Fig. 1D are for comparison. (B) Time course of TLD in the thyA, thyA rfbA, thyA rfbA rffHC, and thyA rfbA rffH (RffC+) mutants. (C) Time course of TLD in the thyA and thyA rffH mutants. (D) Loss of chromosomal DNA in the thyA, thyA rfbA, and thyA rfbA rffHC mutants during T-starvation. (E) Change over time of the absolute amounts of replication origin DNA in the thyA, thyA rfbA, and thyA rfbA rffHC mutants during T-starvation. (F) Same as in D, but for the terminus. (G) The current model of RfbA + RffH vs. ECA-synthesis action in thyA mutants in two growth conditions: with dT supplementation (Top)or during the resistance phase of T-starvation (Bottom).

Rao and Kuzminov PNAS | December 29, 2020 | vol. 117 | no. 52 | 33553 Downloaded by guest on September 29, 2021 lacking cytoplasm after 20 h of T-starvation (Fig. 2 E, Second row). of the chromosomal replication, while increase in the amount of As expected, the thyA rfbA rffHC mutant cells were more affected the terminus indicates completion of the chromosomal replica- by T-starvation, compared to the thyA rfbA mutant, showing some tion, reflecting the general progress of replication forks. As lysis at 3 h of T-starvation and almost complete cytoplasm loss after reported before, the thyA mutants show robust replication initi- 20 h of T-starvation (Fig. 2 E, Third row).Finally,theT-starvedthyA ation during the resistance phase, followed by origin-containing rffHC mutant cells, looking somewhat wider even in the presence of chromosomal domain destruction during the RED phase dT (in agreement with the reported measurements of ECA lipid II- (Fig. 3E) (8, 15, 17). The initial terminus’ 1.5-times increase in accumulating mutants) (45), became extra wide after 1 h of the thyA mutants indicates successful termination of the pre- T-starvation and developed huge central nucleoids (Fig. 2 E, Bottom starvation replication forks during the resistance phase, before row). Corroborating the cell lysis measurements (Fig. 2D), by 3 h of the RED phase ushers gradual chromosomal DNA loss (Fig. 3F) T-starvation most of the thyA rffHC mutant cells lost cytoplasm, and (15, 16, 22). We found that both the origin and the terminus copy the resulting ghosts had little DNA (Fig. 2 E, Bottom row). number show only a limited initial increase in the thyA rfbA rffHC Our investigations in the ECA-defective mutants so far and thyA rfbA mutants (Fig. 3 E and F), indicating both de- revealed a previously unknown dimension of T-starvation af- creased initiation from the replication origin and decreased fecting the cytoplasm dynamics and envelope stability within the overall progress of replication forks. At this point, we conclude standard time frame of TLD (the first 6 h of T-starvation). that the two paralogous dTDP-glucose pyrophosphorylases, Specifically, the cytoplasm and envelope effects of T-starvation RfbA with the help of RffH, are responsible for the bulk of the distinguish at least three severity levels of TLD: 1) the mildest LMW-dT pool in E. coli, which is employed in the EPS synthesis. one in the thyA and thyA rfbA mutants; 2) the severe T-starvation In the thyA mutants, this LMW-dT pool maintains cell viability associated with significant lysis in the thyA rfbA rffHC mutant, through the resistance phase of TLD by supporting the shrinking likely due to lack of dTDP-glucose synthesis; and 3) the dTTP pool and, via this, the initial chromosome replication T-hyperstarvation in the thyA rffHC mutant displaying massive (Fig. 3G), in effect, postponing the onset of acute T-starvation. lysis, likely due to accumulation of ECA lipid II. Next we com- pared the size and dynamics of the LMW-dT pools, as well as the RffC Helps to Extract dTTP out of the LMW-dT Pool during T-Starvation. TLD parameters in the thyA rfbA rffHC vs. thyA rffHC mutants. The severe growth defect of the thyA rffC and thyA rffHC mutants on LB without dT (Fig. 2 B and C) and lysis of the latter during RfbA and RffH Recruit dTTP into the LMW-dT Pool. Since both RfbA T-starvation (Fig. 2 D and E) suggested that RffC function becomes and RffH were supposed to contribute to the dTDP-glucose pool <<—>> – critical for the dTTP dTDP-hexose equilibrium during the (31 33), we expected the size of the LMW-dT pool to be smaller resistance phase of TLD. The (partial) suppression of both phe- when both enzymes are inactivated. Indeed, we found that, in notypes in the thyA rfbA rffHC mutant (Figs. 2 B and D and 3) contrast to 60% of the chromosomal dT content in the thyA meant that the rffC mutants are poisoned by functional RfbA and mutant, the LMW-dT pool is less than 1.5% of the chromosomal + RffH, consistent with RffC action facilitating the release of dTDP dT content in the thyA rfbA rffH (RffC ) mutant, increasing from dTDP-hexose conjugates (Fig. 2A). slightly during T-starvation (Fig. 3A). Interestingly, we found ∼ We found that the LMW-dT pool in the thyA rffHC mutant is only slightly higher numbers ( 3%) and similar evolution of the similar to the one in the thyA mutant during exponential growth; pool in the thyA rfbA mutant (Fig. 3A), suggesting that RfbA is however, in contrast to it shrinking in thyA, the LMW-dT pool in the major contributor to dTDP-glucose synthesis. In contrast, in the thyA rffHC mutant significantly expands during T-starvation the thyA rffH mutant the pool was still about half of the one in (Fig. 4A), perhaps reflecting dT redistribution from the degraded thyA mutant and it did not go much lower during T-starvation chromosomal DNA? The thyA rffC (RffH+) mutant showed (Fig. 3A), suggesting that the still functional RfbA was actively higher initial LMW-dT pools and even more dramatic expansion rebuilding it even without exogenous dT. We conclude that: 1) during T-starvation (SI Appendix, Fig. S8A). Thus, RffC action dTDP-glucose and its derivatives make >95% of the overall has no adverse LMW-dT effect during growth in the presence of LMW-dT pool in the thyA mutant; and 2) this pool is accessible to the cell and is used during T-starvation. dT, while during T-starvation it helps extract dTTP from the RfbA+RffH-made LMW-dT pool, returning it back to the DNA LMW-dT Pool Supports DNA Synthesis during the Resistance Phase, precursor pools (Figs. 2A and 3G). However, before securing this Delaying and Alleviating TLD. Not only the LMW-dT pool is all but conclusion, we had to address an obvious caveat about the rfbA, eliminated in the thyA rfbA and thyA rfbA rffH mutants (Fig. 3A), rffH, and rffC defects. but the thyA rfbA rffHC and thyA rfbA rffH (RffC+) mutants have Since on the one hand these mutations affect the cell envelope shorter resistance phases and die deeper during the RED phase, integrity, while on the other hand they also impact the LMW-dT while the thyA rfbA mutant only has a deeper RED phase pools, they could potentially do the latter by modifying either the (Fig. 3B). At the same time, the thyA rffH mutant is no more nucleoside transport into the cell, or the release of LMW species sensitive to T-starvation than the thyA mutant (Fig. 3C). In other from the cell. To detect these possible nonspecific effects of rfbA, words, the LMW-dT pool supports at least part of the resistance rffH, and rffC mutations, we applied our HMW/LMW separation 3 phase of TLD and prevents an even deeper RED phase. protocol (Fig. 1B) to measure the distribution of H-uracil be- Confirming the recently published results (16, 22) (Fig. 1A), tween HMW-rU content (cellular RNA) and LMW-rU pool DNA amount initially increased ∼1.5 times during the resistance (made of the RNA precursor UTP and EPS-capsule precursors phase in the thyA mutant before slowly declining during the RED UDP-sugars) in these mutants (Fig. 4B). We found that all four phase (Fig. 3D). In contrast, both thyA rfbA rffHC and thyA rfbA strains of the set have about the same ratio of LMW rU to the mutants fail to accumulate chromosomal DNA early on upon total HMW rU (∼8%), whether exponentially growing or after T-starvation, even though later their chromosomal DNA disap- 3 h of T-starvation (Fig. 4C). In other words, the rfbA and rffHC pearance parallels the one in thyA mutant cells (Fig. 3D). Thus, mutations do not affect the HMW/LMW balance of uracil pools, the LMW-dT pool indeed supports the initial chromosomal making it unlikely that the dramatic effects of these mutations on DNA accumulation in T-starved cells. the HMW/LMW balance of dT pools are due to nonspecific By monitoring the copy number of the origin and the terminus changes in permeability of the cell envelope. Thus, our overall of the chromosome, we characterize the dynamics of chromo- conclusion is that, while RfbA and RffH recruit dTTP into the somal DNA replication and loss during T-starvation (16). In- LMW-dT pools, RffC helps “extract” dTTP out of these pools, deed, increase in the amount of origin DNA registers initiations which becomes critical during T-starvation (Fig. 3G).

33554 | www.pnas.org/cgi/doi/10.1073/pnas.2012254117 Rao and Kuzminov Downloaded by guest on September 29, 2021 Fig. 4. RffC helps to extract dTTP out of the LMW-dT pool during T-starvation. The strains are thyA (KKW58), thyA rfbA (RA48), thyA rffHC (RA49), and thyA rfbA rffHC (RA51). (A) The evolution of LMW-dT pools in the thyA rffHC mutant during T-starvation (the red bars). Thymidine was removed at time = 0, after taking aliquots for the 0-h cultures. Values for thyA mutant (black bars) from Fig. 1D are for comparison. (B) Scheme of separation of the intracellular uridine into HMW rU (the rU content of the cellular RNA, mostly ribosomes) and LMW rU (nucleotides and sugar conjugates). (C) The size of the LMW-rU pool, normalized to the HMW-rU content of the cellular RNA, either during normal growth in dT-supplemented medium or after 3 h of T-starvation. (D) Time

course of TLD in the thyA rffHC mutant. In D and E,thethyA rfbA rffHC mutant curve is shown to illustrate partial suppression. (E) Loss of the chromosomal MICROBIOLOGY DNA absolute amounts in the thyA rffHC mutants during T-starvation. (F) Loss of the replication origin and terminus DNA in the thyA rffHC mutant during T-starvation. (G) Cell lysis upon T-starvation in the thyA rffHC mutant cultures at various temperatures. In contrast to Fig. 2D, where lysis was expressed in Miller units, here lysis is expressed as percentage of the total cellular content of beta-galactosidase, which undergoes its own peculiar kinetics during T-starvation (SI Appendix, Fig. S7). (H) Time course of TLD in the thyA rffHC mutant cultures at the same temperatures as in G.

Inability to Release dTTP from the LMW-dT Pool Results in Does Cell Lysis Drive the T-Hyperstarvation Phenomenon? The lysis T-Hyperstarvation. If RffC helps extracting dTTP from LMW-dT of the thyA rffHC mutant cells in response to T-starvation seems pools during T-starvation, its inactivation should exacerbate the like an obvious reason for the catastrophic chromosomal DNA TLD phenotypes. Indeed, we found that the thyA rffC and thyA loss and for the absence of the resistance phase in this mutant; rffHC mutants completely lack the resistance phase and develop yet a closer look shows that these events are separated in time the dramatically deeper RED phase (Fig. 4D and SI Appendix, and therefore independent. First, at 1 h of T-starvation, the thyA Fig. S9) reminiscent of the thyA recBCD mutants (7, 16, and see rffHC cultures suffer little or no lysis (Figs. 2D and 4G and SI below). The better survival of the thyA rfbA rffHC mutant com- Appendix, Fig. S6) and no cytoplasm/envelope irregularities pared to the (RfbA+) thyA rffHC mutant (Fig. 4D) confirms (Fig. 2E), even though only 3% of them remain viable at this RfbA+RffH toxicity in the thyA rffC mutant, likely resulting point (Fig. 4 D and H), and at the population level they have from the continuous sequestration of dTTP into the LMW-dT already lost 36% of their genomic DNA (Fig. 4E) and ∼50% of pool by dTDP-glucose synthases even during T-starvation the replication origin and terminus DNA (Fig. 4F). Second, from (Fig. 4A and SI Appendix, Fig. S8 A and B). quantification of beta-galactosidase release, lysis affects less than Change in the chromosomal DNA amounts during T-starvation half of the T-starved thyA rffHC cells (Fig. 4G), whereas their (Fig. 1A) is also dramatically different in the thyA rffHC mutant. In general survival is below 10e-4 (Fig. 4 D and H). Finally, lysis of contrast to the significant initial chromosomal DNA replication of the thyA rffHC mutants is strongly suppressed during the thyA mutant or even the initially flat profile of the thyA rfbA and T-starvation at 28 °C (Fig. 4G), whereas TLD, although also thyA rfbA rffHC mutants (Fig. 3D), the chromosomal DNA of the reduced, still reaches 10e-3 at this temperature (Fig. 4H). Thus, thyA rffC and thyA rffHC mutants becomes immediately unstable lysis of the thyA rffHC mutants without dT, even though signif- upon thymidine removal, with more than one-third of it already icant at 37 °C at 2 h of T-starvation, must be a later event sep- gone after 1 h of T-starvation and only 10% of the starting amount arate from their immediate lethality and chromosomal remaining by 4 h of T-starvation (Fig. 4E and SI Appendix,Fig.S10). DNA loss. Moreover, this chromosomal DNA disappearance in the thyA rffHC mutant affects equally and dramatically both the origin and the T-Hyperstarvation in Other Mutants That Cannot Finish ECA Synthesis. terminus (Fig. 4F), suggesting catastrophic loss of the entire chro- ECA synthesis up to lipid III is a multistep three-branch process mosome. The instability of the chromosomal DNA in the thyA (Fig. 5A). As we have shown, blocking initiation of its dTDP rffHC mutant is again partially suppressed by RfbA inactivation branch with the rfbA rffH double defect exacerbates TLD via (Fig. 4E). reducing the LMW-dT pool (Fig. 3). We have also found that We conclude that during T-starvation there is a tug-of-war T-hyperstarvation in the thyA mutants, caused by blocking the between RfbA+RffH on the one hand and RffC on the other, completion of this branch with the rffC defect that prevents the the former continuously sequestering dTTP into the LMW-dT release of dTDP from the LMW-dT pool, causes hyper-TLD: pool and thus aggravating TLD, while the latter releasing dTTP rapid death due to chromosome destruction and cell lysis from this pool, therefore alleviating TLD. (Fig. 4). If the lack of dTDP release from the ECA synthesis

Rao and Kuzminov PNAS | December 29, 2020 | vol. 117 | no. 52 | 33555 Downloaded by guest on September 29, 2021 Fig. 5. Hyper-TLD is observed in other ECA-defective mutants and is independent of DNA replication. The strains are thyA (KKW58), thyA rffHC (RA49), thyA dnaA46(Ts) (KJK170), thyA dnaA601(Ts) (SRK291), thyA (SOS) (RA66), thyA dnaA46(Ts) (SOS) (RA67), thyA dnaA46(Ts) rffHC (RA53), thyA dnaA601(Ts) rffHC (RA54), thyA rfe (RA61), thyA rffM (RA63), and thyA rffT (RA62). (A) Scheme of ECA biosynthesis up to lipid III. The relevant genes are in color: rfbA and rffH are blue, rffC is red, while rfe, rffM, and rffT are purple. (B) Spotting of serial dilutions of the indicated mutants on LB2xYE ±dT. (C) Time course of TLD in the thyA rfe, thyA rffM, and thyA rffT mutants. Note that the control thyA strain was dying shallower in this set. (D) The standard T-starvation protocol compared to the awakening protocol, used in E–H.(E) Time course of TLD at 42 °C of the thyA and thyA dnaA46(Ts) mutants in the awakening protocol. The same cultures are also diluted into the fresh medium +dT, to show the dnaA(Ts) growth defect. (F) No SOS induction in the thyA dnaA(Ts) (SOS) mutants sup- plemented with dT at 42 °C. Strains were grown at 28 °C and switched to 42 °C ±dT at time = 0. The thyA (SOS) mutant +dT provides a negative control, while the robust SOS induction in the –dT conditions in the two strains provides a positive control. (G) Time course of TLD at 42 °C in the dnaA46(Ts) and dnaA601(Ts) variants of either the thyA mutant (black lines) or thyA rffHC mutant (red lines). (H) Loss of the chromosomal DNA during T-starvation at 42 °C in the dnaA46(Ts) and dnaA601(Ts) variants of either the thyA mutant (black lines) or thyA rffHC mutant (red lines).

pathway was indeed the reason for hyper-TLD of the thyA rffC the thyA dnaA(Ts) mutant cannot initiate replication of its mutants, then any other mutation in ECA synthesis that blocks chromosome even in the presence of dT and remains static dTDP release should also make thyA mutants hypersensitive to (Fig. 5E). At the same time, in accord with our previous con- T-starvation. Indeed, spotting on an “enhanced” LB reveals that, clusion about replication independence of TLD, in the absence like the rffC defect, addition of the rfe, rffM,orrffT defects makes of dT this thyA dnaA mutant undergoes exactly the same death as the thyA mutants grow poorly without dT, with the thyA rffT its DnaA+ progenitor (Fig. 5E). mutant essentially copying the behavior of the thyA rffC mutant Since SOS induction significantly contributes to TLD (8, 9), if (Fig. 5B), which makes sense from the ECA biosynthesis scheme the inability to initiate replication from the origin also induces (Fig. 5A). In the standard TLD assay, blocking the other branch of SOS, the observed TLD in the thyA dnaA(Ts) mutant at 42 °C ECA synthesis that converges with the dTDP branch with either rfe, could be due to this SOS induction even in the absence of rffM,orrffT defects makes the thyA mutants hypersensitive to chromosomal DNA replication. However, we found that the T-starvation (Fig. 5C). Thus, dTTP release from the LMW-dT pool thyA dnaA(Ts) mutants do not induce the SOS response at 42 °C, could be the common denominator of the various degrees of the as long as they are supplemented with dT, while undergoing T-hyperstarvation phenotype of all these mutants that synthesize strong SOS induction in both the standard (SI Appendix, Fig. dTDP-glucose but cannot finish the ECA synthesis. S11) and the awakening protocols of T-starvation (Fig. 5F). The latter result indicates replication independence of the SOS- Hyper-TLD of the thyA rffHC Mutant Is Independent of Replication and Endo I. Finding mutants that cannot finish their ECA syn- inducing chromosomal lesions during T-starvation. We also re- thesis and thus undergo faster death upon T-starvation—while port that in the same awakening protocol blocking chromosomal establishing that the resistance phase is not an integral part of replication, the thyA rffHC dnaA(Ts) mutants die with somewhat TLD—posed the question about the nature of the accompanying deeper kinetics compared to the thyA rffHC mutant (Fig. 5G). catastrophic chromosomal DNA loss, which we observed in the Thus, the hyper-TLD in the thyA rffHC mutants is also inde- thyA rff(H)C mutants (Fig. 4 E and F and SI Appendix, Fig. S10). pendent of DNA replication. TLD was always considered a chromosomal replication-centered As already reported (22), in contrast to the standard T-starvation phenomenon (1, 5), but we have recently demonstrated a sur- protocol (Fig. 5D), in the awakening protocol at 42 °C the thyA prising TLD independence of the chromosomal replication (22). mutant suffers significant chromosomal DNA loss, independently of For this, we used the “awakening protocol” to eliminate all of the its dnaA status (compare Fig. 3D vs.Fig.5H). Although for the thyA replication activity in the chromosome by bringing dT- rffHC mutant in the awakening protocol at 42 °C the massive supplemented cultures of the thyA dnaA(Ts) mutants to satura- chromosomal DNA loss is slightly delayed, its overall kinetics is tion at the permissive temperature and then outgrowing (awak- hardly affected (if a bit accelerated) by the dnaA(Ts) defects ening) them without dT at the nonpermissive temperature of (Fig. 5H), indicating that the catastrophic chromosomal DNA loss 42 °C (Fig. 5D) (22). Because of the nonpermissive temperature, due to the rffC defect is also completely replication independent.

33556 | www.pnas.org/cgi/doi/10.1073/pnas.2012254117 Rao and Kuzminov Downloaded by guest on September 29, 2021 Following the recent report of Endo I-catalyzed massive Discussion chromosomal DNA fragmentation during slow lysis of cells in In contrast to typically static responses to amino acid or RNA agarose plugs (46), we tested whether the fast chromosomal base starvation in E. coli (2, 49), T-starvation leads to cell death, DNA disappearance in the thyA rffHC mutants is due to Endo I but not directly. The nature of the two-generations-long resis- gaining access to the chromosomal DNA during lysis, but the tance phase, followed by the sudden shift to the RED phase DNA loss turned out to be Endo-I independent (SI Appendix, (Fig. 1A), has remained a long-standing puzzle of TLD (16). This Fig. S12A). Also, the endA inactivation failed to affect TLD ki- static period at the beginning of T-starvation, with its significant netics of both the thyA and thyA rffHC mutant (SI Appendix, Fig. initiation of DNA replication (15, 16), SOS induction (7, 8, 17), S12B). accumulation of ss-gaps, and high levels of chromosomal frag- thyA rffHC mentation (16), was considered an integral TLD period, when Chromosome Is the Primary Target of TLD in the Mutant. chromosomal damage accumulates to irreparable levels, usher- The most likely candidate for an enzyme capable of such a rapid ing in the RED phase (1, 5). Here we show that the resistance chromosomal DNA degradation is the RecBCD helicase/nucle- phase, rather than being a part of TLD, delays acute T-starvation ase (47), yet the catastrophic chromosomal DNA loss in the thyA and is supported by dTTP recruitment from the unexpectedly rffHC recBCD mutant is only slightly slower than in its RecBCD+ large internal LMW-dT pool. This pool in rapidly growing cells parent (Fig. 6A), suggesting that only a minor fraction of the comprises ∼60% of the dT content of the chromosomal DNA; chromosomal DNA of the T-starved thyA rffHC mutant is lost to once it is exhausted as a result of T-starvation, the RED the linear DNA degradation by RecBCD. As already mentioned, phase begins. the hyper-TLD in the thyA rffHC mutant is reminiscent of the While two paralogous thymidine glucose pyrophosphorylases hyper-TLD observed in the thyA recBCD mutant (SI Appendix, of E. coli, RfbA and RffH, recruit dTTP into the LMW-dT pool, Fig. S13), as if RecBCD and RffC work in the same pathway of we found that RffC, by facilitating ECA-synthesis completion, hyper-TLD prevention. Interestingly, the TLD curve of the combined thyA rffHC recBCD mutant suggests an even deeper releases dTTP from the LMW-dT pool, which becomes critical in defect (SI Appendix, Fig. S13); unfortunately, TLD in the thyA the absence of exogenous dT (Fig. 3G). As a result of the re- recBCD and thyA rffHC mutants is already too fast in the stan- duced LMW-dT pool, the thyA rfbA rffH mutant experiences a dard conditions (M9CAA medium) to detect further significant shorter resistance phase with no chromosomal DNA accumula- acceleration. To reveal the type of genetic interactions between tion followed by a deeper RED phase. In contrast, TLD in the the two defects, we employed a milder T-starvation in LB, as we thyA rffC mutant lacks the resistance phase altogether, leads to a

did before for the thyA recBCD uvrD combination (7). Under precipitous loss of chromosomal DNA, and ends in massive cell MICROBIOLOGY these conditions, the thyA mutant grows ∼10 times before pla- lysis. Moreover, the dramatic phenotypes of the thyA rffC mutant teauing and then essentially fails to die, the thyA recBCD mutant during T-starvation, including cell lysis, were partially suppressed dies by 2.5 orders of magnitude, while the thyA rffHC mutant dies by rfbA+rffH inactivation, confirming that dTDP-glucose syn- by 1.5 orders of magnitude (Fig. 6B). Remarkably, the deep TLD thesis poisons T-starved cells lacking RffC activity, by continu- of the thyA recBCD rffHC mutant in LB is clearly the product of ously recruiting the remaining dTTP into the LMW-dT pool. the individual TLD effects of the recBCD and rffHC mutations Hyper-TLD in the thyA rffC mutants is distinct from regular (Fig. 6B), indicating independent action of the two strongest TLD in thyA mutants, or even from hyper-TLD in the thyA TLD-accelerating defects. recBCD mutants, by causing a complete loss of chromosomal Since the recBCD defect makes cells unable to repair double- DNA, as well as eventual lysis in half of the cells, suggesting a strand DNA breaks (reviewed in ref. 48), and assuming that the still unknown role of either LMW dT or the chromosomal DNA substrate of the “rffC pathway” is also chromosomal DNA, the in cell envelope maintenance. independence of the recBCD and rffC pathways (Fig. 6B) pre- dicts a higher chromosomal fragmentation during T-starvation in The dTDP-Sugar-Utilizing Pathways and the EPS Capsule. We are not the thyA rffHC mutants. In the thyA mutants, chromosome sure how this significant pool of dTDP-sugars avoided — fragmentation is induced during the resistance phase of detection maybe because nobody directly compared its size T-starvation but then, counterintuitively, goes away during the with that of the chromosomal DNA? Its significant size is indi- RED phase (16) (Fig. 6 C and D). We found that not only thyA rectly confirmed by the similar size of LMW-rU pools (Fig. 4 B rffHC mutants suffer a higher chromosome fragmentation in and C), which likely are mostly UDP-sugars. Even though LMW-rU response to T-starvation, but also this fragmentation stays high pools are only 8% of the total rU content of stable RNA (Fig. 4C), (Fig. 6 C and D), even though these cells start lysing around 2 h the total mass of cellular RNA is six to seven times larger than that of T-starvation (Figs. 2D and 4G). In other words, the rffC defect of the chromosomal DNA (27), making the absolute sizes of the two induces more double-strand breaks during T-starvation (it could pools similar. Indeed, EPS-capsule synthesis utilizes both dTDP- have also blocked their subsequent repair, but the independence sugars and UDP-sugars (26). There is still a possibility, though, of the rffC and the recBCD mutant effects argues against this that the significant LMW-dT pool is K12 specific, because this formal possibility). Confusingly, when we do block repair of E. coli background lacks OA (34) and therefore might up-regulate double-strand breaks genetically with the recBCD mutation, its EPS-capsule production. chromosomal fragmentation in the thyA recBCD rffHC mutant is Okazaki and Okazaki were the first in the late 1950s to report suppressed and resembles the smoother fragmentation kinetics that the bulk of thymidine internalized by bacteria joins LMW of the thyA recBCD mutants (Fig. 6 C and E), as if the RecBCD- pools as a compound chemically distinct from the DNA pre- promoted linear DNA degradation or repair of double-strand cursor dTTP (50), later identified as dTDP-rhamnose (51). By breaks in the thyA rffHC mutant stimulates even more breaks analogy with the already well-known UDP- and GDP-sugar (compare Fig. 6D vs. Fig. 6E). conjugates, Okazaki suggested that dTDP-rhamnose partici- On the basis of our analysis of T-hyperstarvation phenotypes pates in EPS-capsule synthesis (52), but since the dTDP portion in the thyA rffHC mutants, we conclude that the resulting hyper- of this conjugate was (slowly) chased into DNA (23), he specu- TLD does not require replication forks, but it does massively lated that dTDP-sugars could represent common intermediates break chromosomal DNA (independently of Endo I), which for both polysaccharide and DNA synthesis (52). explains its synergy with the double-strand DNA break repair An enzyme synthesizing dTDP-glucose was soon reported defect (recBCD) and is the likely reason for the catastrophic loss (53), while the pools of dTDP-sugars under various conditions of chromosomal DNA. were consistently severalfold higher than the dTTP pool (54–56).

Rao and Kuzminov PNAS | December 29, 2020 | vol. 117 | no. 52 | 33557 Downloaded by guest on September 29, 2021 Fig. 6. Chromosome fragmentation in the T-starved thyA rffHC mutant. The strains are thyA (KKW58), thyA rffHC (RA49), thyA recBCD (KJK63), and thyA recBCD rffHC (RA52). (A) Loss of the chromosomal DNA in the thyA recBCD rffHC mutant during T-starvation. (B) Time course of TLD in the thyA, thyA recBCD, thyA rffHC, and thyA recBCD rffHC mutants in LB (compare with SI Appendix, Fig. S13, the same time course, but in the standard conditions). (C) Pulsed-field gel showing kinetics of chromosome fragmentation during T-starvation. A shorter time frame is used because no action happens in thyA and thyA recBCD mutants past 3 h without dT, while the thyA rffHC mutants start lysing after 2 h without dT. CZ, compression zone; LMW, low molecular weight fragments; ∼50 to 200 kbp in size, recB* = recBCD.(D and E) Quantitative kinetics of the chromosomal fragmentation, from several gels as in C. The higher background (time = 0) chromosome fragmentation in the thyA rffHC mutants reflects a peculiar sensitivity of this mutant to centrifugation (used to collect cells before encasing them in agarose plugs). Minimizing the speed of centrifugation minimizes this fragmentation even for thyA rffHC cells still supplemented with dT (as well as for the T-starved cells). (F) Our speculation about the chromosome and cell behavior during T-starvation vs. T-hyperstarvation. The shaded circle represents a cross-section of the cell, in which the chromosomal DNA (blue) is attached to the cell envelope (brown) at certain points (yellow). In the absence of dT, DNA attachment to the envelope is proposed to “freeze.” As a result, during T-starvation, increase in cell volume causes not only envelope thinning, but also chromosomal breakage and limited DNA loss. In contrast, during T-hyperstarvation, a larger increase in cell volume breaks DNA in many places, leading to a complete chromosome loss, while the cell envelope also eventually bursts.

Moreover, a faster and deeper TLD was reported in an uncharac- The loss of wecE, the gene responsible for the step preceding terized E. coli mutant deficient in dTDP-glucose synthesis (57). the rffC step in the ECA biosynthesis (Fig. 5A), has been pre- Nevertheless, the original idea of Okazaki and Okazaki about viously shown to lead to the accumulation of ECA lipid II in- dTDP-sugars as common intermediates for both EPS capsule and termediates, which causes membrane instability by diverting the DNA replication came full circle only some 60 y later, with this undecaprenol moiety away from use in PG synthesis (45). On the report that the pool of dTDP-sugars becomes the major source of basis of our results, we suspect that the rffC mutants have similar dT for the chromosomal replication during T-starvation. problems. The ECA-synthesis activity that acts after RffC mak- ing lipid III and releasing dTDP from dTDP-sugar conjugates is Is ECA Lipid II Accumulation Poisonous during T-Starvation? Unex- RffT (Fig. 5A), a mutation which also makes the thyA mutant pectedly, the thyA rfbA rffH (RffC+) mutant, which should be hyper-sensitive to T-starvation (Fig. 5 B and C). The various deficient for ECA synthesis (essentially leaving our O-minus K12 defects due to inability to finish ECA lipid III synthesis should be background (34) without EPS capsule), avoids the most severe suppressed by inactivation of the rfe gene that blocks the accu- TLD. Instead, the thyA rffC mutant develops the worst T-starvation mulation of all ECA lipid intermediates (Fig. 5A); we will be phenotypes. This severity is partially suppressed by the rfbA rffH testing this possibility in the future. Remarkably, our findings inactivation, suggesting that it is not the inability to synthesize ECA, indicate that accumulation of ECA lipid II becomes toxic only in which makes the thyA rffC mutant extremely vulnerable to the absence of dT, suggesting a role for dT in cell envelope T-starvation. maintenance.

33558 | www.pnas.org/cgi/doi/10.1073/pnas.2012254117 Rao and Kuzminov Downloaded by guest on September 29, 2021 The Nature of Chromosomal Damage during T-Starvation. Since TLD One possibility is based on the known association of random is only observed in growing cultures (22), and since the only pieces of the chromosomal DNA with the cell envelope (58), critical role of thymidine in the cell was thought to generate the specifically with the outer membrane (59, 60), reflected in the DNA precursor dTTP, faulty chromosomal replication during chromosomal DNA capture by the outer membrane vesicles (61, T-starvation was always considered the primary cause of TLD, 62). Though random and transient for any particular DNA seg- the significant effects of recombinational repair defects on TLD ment (with the exception of the hemimethylated oriC) (63), this kinetics supporting this thinking (6–9). Paradoxically, even association may in fact be secure in terms of DNA anchoring to though high levels of chromosome fragmentation develop during the cell envelope, if mediated by special spool-like proteins. If the resistance phase of TLD, this fragmentation does not directly such spooling is jammed in the absence of dTTP, while the cell contribute to lethality in the repair-proficient cells (7, 15, 16). circumference significantly expands due to the same T-starvation Instead, genetic studies strongly suggested that the main con- (Fig. 2E), a DNA segment trapped between adjacent envelope- tributor to the chromosome poisoning during T-starvation was association points may become overextended and eventually the repair of persistent single-strand gaps (6–8, 17), and we have snaps, causing a double-strand break (Fig. 6F). While most of indeed previously reported ss-gap accumulation in the chromo- these breaks are repaired, a few of them could be irreparable (for somal DNA during the RED phase (16). Still, the mechanism of whatever reason), causing the observed chromosomal DNA loss T-starvation-induced chromosomal lesions remained unclear. and TLD. The same scenario is magnified and accelerated dur- To explain how thymineless replication translates into irrep- ing T-hyperstarvation (Fig. 6F), leading to a complete loss of arable chromosomal lesions, futile cycles of DNA-uracil incor- DNA and to cell lysis (due to the cell envelope overextension, poration/excision were repeatedly proposed to yield irreparable combined with the defect in the EPS capsule and poisoning by DNA damage during T-starvation (1, 5, 21). However, we found accumulating ECA lipid II). The nature of this chromosomal that dUTP concentrations are too low in the Dut+ cells to sup- DNA disappearance, and its relation to T-starvation-triggered chromosome fragmentation, should become one of the main port futile DNA-uracil incorporation/excision cycles (16). Even directions of future TLD studies. preventing DNA-uracil excision after massive uracil incorpora- tion (in the thyA dut ung mutants) only makes the RED phase Materials and Methods shallower, but does not eliminate it altogether (16). Yet, the Details of all experimental procedures, including bacterial strains and plas- source of irreparable chromosomal lesions during TLD could be mids, growth and treatment conditions, genomic DNA isolation and analysis, futile fork breakage-repair cycles (16). detection of LMW-dT/U species, beta-galactosidase and alkaline phosphatase

To address the fork breakage-repair idea, we tested the ne- assays, measuring the SOS induction, fluorescent microscopy, and pulsed- MICROBIOLOGY cessity of chromosomal replication for TLD and were perplexed field gel electrophoresis to detect chromosome fragmentation are fully to find normal TLD in the complete absence of replication (22); described in SI Appendix, Methods. here we have confirmed this game-changing observation and extended it to the thyA rffHC mutants. The lack of replication Data Availability. All study data are included in the article and supporting information. requirement, while eliminating the biggest group of TLD mod- els, further highlights the mystery of the massive chromosomal ACKNOWLEDGMENTS. We thank Sanna Koskiniemi (Uppsala Universitet) for DNA loss during T-starvation, that becomes catastrophic in the reminding us about dTDP-hexoses and William W. Metcalf (University of thyA rffC mutants (Fig. 4 E and F) or in the awakening protocol Illinois at Urbana–Champaign) for challenging us to measure LMW-dT pools at 42 °C (22) (Fig. 5H). Since testing of the most obvious ideas and for hospitality in his light microscopy facility. We are grateful to all members of A.K.’s laboratory for constructive criticism and support and, in about this DNA loss failed to produce insights here, we can only addition, to Lenna Kouzminova for critically reading the manuscript. This speculate on its nature. work was supported by grant GM073115 from the NIH.

1. S. I. Ahmad, S. H. Kirk, A. Eisenstark, Thymine metabolism and thymineless death in 14. E. Guarino, I. Salguero, A. Jiménez-Sánchez, E. C. Guzmán, Double-strand break prokaryotes and eukaryotes. Annu. Rev. Microbiol. 52, 591–625 (1998). generation under deoxyribonucleotide starvation in Escherichia coli. J. Bacteriol. 189, 2. H. D. Barner, S. S. Cohen, The induction of thymine synthesis by T2 infection of a 5782–5786 (2007). thymine requiring mutant of Escherichia coli. J. Bacteriol. 68,80–88 (1954). 15. K. J. Kuong, A. Kuzminov, Disintegration of nascent replication bubbles during thy- 3. S. S. Cohen, H. D. Barner, Studies on unbalanced growth in Escherichia coli. Proc. Natl. mine starvation triggers RecA- and RecBCD-dependent replication origin destruction. – Acad. Sci. U.S.A. 40, 885–893 (1954). J. Biol. Chem. 287, 23958 23970 (2012). 4. S. S. Cohen, On the nature of thymineless death. Ann. N. Y. Acad. Sci. 186, 292–301 16. T. V. P. Rao, A. Kuzminov, Sources of thymidine and analogs fueling futile damage- repair cycles and ss-gap accumulation during thymine starvation in Escherichia coli. (1971). DNA Repair (Amst.) 75,1–17 (2019). 5. A. Khodursky, E. C. Guzmán, P. C. Hanawalt, Thymineless death lives on: New insights 17. D. P. Sangurdekar et al., Thymineless death is associated with loss of essential genetic into a classic phenomenon. Annu. Rev. Microbiol. 69, 247–263 (2015). information from the replication origin. Mol. Microbiol. 75, 1455–1467 (2010). 6. H. Nakayama, K. Nakayama, R. Nakayama, Y. Nakayama, Recombination-deficient 18. H. D. Barner, S. S. Cohen, Protein synthesis and RNA turnover in a -deficient mutations and thymineless death in Escherichia coli K12: Reciprocal effects of recBC bacterium. Biochim. Biophys. Acta 30,12–20 (1958). – and recF and indifference of recA mutations. Can. J. Microbiol. 28, 425 430 (1982). 19. J. Gallant, S. R. Suskind, Relationship between thymineless death and ultraviolet in- 7. K. J. Kuong, A. Kuzminov, Stalled replication fork repair and misrepair during thy- activation in Escherichia coli. J. Bacteriol. 82, 187–194 (1961). mineless death in Escherichia coli. Genes Cells 15, 619–634 (2010). 20. T. R. Breitman, P. B. Maury, J. N. Toal, Loss of deoxyribonucleic acid-thymine during 8. N. C. Fonville, D. Bates, P. J. Hastings, P. C. Hanawalt, S. M. Rosenberg, Role of RecA thymine starvation of Escherichia coli. J. Bacteriol. 112, 646–648 (1972). and the SOS response in thymineless death in Escherichia coli. PLoS Genet. 6, 21. M. Goulian, B. Bleile, B. Y. Tseng, -induced misincorporation of uracil e1000865 (2010). into DNA. Proc. Natl. Acad. Sci. U.S.A. 77, 1956–1960 (1980). 9. N. C. Fonville, Z. Vaksman, J. DeNapoli, P. J. Hastings, S. M. Rosenberg, Pathways of 22. S. R. Khan, A. Kuzminov, Thymineless death in Escherichia coli is unaffected by resistance to thymineless death in Escherichia coli and the function of UvrD. Genetics chromosomal replication complexity. J. Bacteriol. 201, e00797-18 (2019). 189,23–36 (2011). 23. R. Okazaki, T. Okazaki, Y. Kuriki, Incorporation of [3H] thymidine in a deoxyriboside- – 10. K. Nakayama, N. Irino, H. Nakayama, The recQ gene of Escherichia coli K12: Molecular requiring bacterium. Biochim. Biophys. Acta 33, 289 291 (1959). 24. C. A. Schnaitman, J. D. Klena, Genetics of lipopolysaccharide biosynthesis in enteric cloning and isolation of insertion mutants. Mol. Gen. Genet. 200, 266–271 (1985). bacteria. Microbiol. Rev. 57, 655–682 (1993). 11. K. Nakayama, K. Kusano, N. Irino, H. Nakayama, Thymine starvation-induced struc- 25. G. Samuel, P. Reeves, Biosynthesis of O-antigens: Genes and pathways involved in tural changes in Escherichia coli DNA. Detection by pulsed field gel electrophoresis nucleotide sugar precursor synthesis and O-antigen assembly. Carbohydr. Res. 338, and evidence for involvement of homologous recombination. J. Mol. Biol. 243, 2503–2519 (2003). – 611 620 (1994). 26. H. M. Kuhn, U. Meier-Dieter, H. Mayer, ECA, the enterobacterial common antigen. 12. K. Nakayama, S. Shiota, H. Nakayama, Thymineless death in Escherichia coli mutants FEMS Microbiol. Rev. 4, 195–222 (1988). deficient in the RecF recombination pathway. Can. J. Microbiol. 34, 905–907 (1988). 27. J. Neuhard, P. Nygaard, “Purines and ” in Escherichia coli and Salmonella 13. H. Nakayama, Escherichia coli RecQ helicase: A player in thymineless death. Mutat. typhimurium. Cellular and Molecular Biology, F. C. Neidhardt, Ed. (American Society Res. 577, 228–236 (2005). for Microbiology, Washington, D.C., 1987), pp. 445–473.

Rao and Kuzminov PNAS | December 29, 2020 | vol. 117 | no. 52 | 33559 Downloaded by guest on September 29, 2021 28. S. R. Khan, A. Kuzminov, Replication forks stalled at ultraviolet lesions are rescued via 45. M. A. Jorgenson, S. Kannan, M. E. Laubacher, K. D. Young, Dead-end intermediates in RecA and RuvABC protein-catalyzed disintegration in Escherichia coli. J. Biol. Chem. the enterobacterial common antigen pathway induce morphological defects in Es- 287, 6250–6265 (2012). cherichia coli by competing for undecaprenyl phosphate. Mol. Microbiol. 100,1–14 29. R. P. Maharjan, T. Ferenci, Global metabolite analysis: The influence of extraction (2016). methodology on metabolome profiles of Escherichia coli. Anal. Biochem. 313, 46. S. R. Khan, A. Kuzminov, Degradation of RNA during lysis of Escherichia coli cells in 145–154 (2003). agarose plugs breaks the chromosome. PLoS One 12, e0190177 (2017). 30. W. El-Kazzaz, T. Morita, H. Tagami, T. Inada, H. Aiba, Metabolic block at early stages 47. M. S. Dillingham, S. C. Kowalczykowski, RecBCD enzyme and the repair of double- of the glycolytic pathway activates the Rcs phosphorelay system via increased syn- stranded DNA breaks. Microbiol. Mol. Biol. Rev. 72, 642–671 (2008). thesis of dTDP-glucose in Escherichia coli. Mol. Microbiol. 51, 1117–1128 (2004). 48. A. Kuzminov, Recombinational repair of DNA damage in Escherichia coli and bacte- 31. C. L. Marolda, M. A. Valvano, Genetic analysis of the dTDP-rhamnose biosynthesis riophage lambda. Microbiol. Mol. Biol. Rev. 63, 751–813 (1999). region of the Escherichia coli VW187 (O7:K1) rfb gene cluster: Identification of 49. T. R. Breitman, A. Finkleman, M. Rabinovitz, Methionineless death in Escherichia coli. functional homologs of rfbB and rfbA in the rff cluster and correct location of the rffE J. Bacteriol. 108, 1168–1173 (1971). – gene. J. Bacteriol. 177, 5539 5546 (1995). 50. R. Okazaki, T. Okazaki, Studies of deoxyribonucleic acid synthesis and cell growth in 32. P. D. Rick, S. Wolski, K. Barr, S. Ward, L. Ramsay-Sharer, Accumulation of a lipid-linked the deoxyriboside-requiring bacteria, Lactobacillus acidophilus. I. Biological and intermediate involved in enterobacterial common antigen synthesis in Salmonella chemical nature of the intra-cellular acid-soluble deoxyribosidic compounds. Biochim. typhimurium mutants lacking dTDP-glucose pyrophosphorylase. J. Bacteriol. 170, Biophys. Acta 28, 470–482 (1958). – 4008 4014 (1988). 51. R. Okazaki, Isolation of a new deoxyribosidic compound, thymidine diphosphate 33. J. Sivaraman, V. Sauvé, A. Matte, M. Cygler, Crystal structure of Escherichia coli glu- rhamnose. Biochem. Biophys. Res. Commun. 1,34–38 (1959). cose-1-phosphate thymidylyltransferase (RffH) complexed with dTTP and Mg2+. 52. R. Okazaki, Studies of deoxyribonucleic acid synthesis and cell growth in the J. Biol. Chem. 277, 44214–44219 (2002). deoxyriboside-requiring bacteria, Lactobacillus acidophilus. III. Identification of thy- 34. D. Liu, P. R. Reeves, Escherichia coli K12 regains its O antigen. Microbiology (Reading) midine diphosphate rhamnose. Biochim. Biophys. Acta 44, 478–490 (1960). 140,49–57 (1994). 53. S. Kornfeld, L. Glaser, The enzymic synthesis of thymidine-linked sugars. I. Thymidine 35. A. Rahman, K. Barr, P. D. Rick, Identification of the structural gene for the TDP- diphosphate glucose. J. Biol. Chem. 236, 1791–1794 (1961). Fuc4NAc:lipid II Fuc4NAc transferase involved in synthesis of enterobacterial common 54. R. Hosono, H. Hosono, S. Kuno, Effects of growth conditions on thymidine nucleotide antigen in Escherichia coli K-12. J. Bacteriol. 183, 6509–6516 (2001). pools in Escherichia coli. J. Biochem. 78, 123–129 (1975). 36. F. C. Neidhardt, “Chemical composition of Escherichia coli” in Escherichia coli and 55. T. Ohkawa, Studies of intracellular thymidine nucleotides. Relationship between the Salmonella typhimurium. Cellular and Molecular Biology, F. C. Neidhardt, Ed. synthesis of deoxyribonucleic acid and the thymidine triphosphate pool in Escherichia (American Society for Microbiology, Washington, D.C., 1987), pp. 3–6. coli K12. Eur. J. Biochem. 61,81–91 (1976). 37. T. Baba et al., Construction of Escherichia coli K-12 in-frame, single-gene knockout 56. B. R. Bochner, B. N. Ames, Complete analysis of cellular nucleotides by two- mutants: the Keio collection. Mol. Syst. Biol. 2, 2006.0008 (2006). dimensional thin layer chromatography. J. Biol. Chem. 257, 9759–9769 (1982). 38. P. N. Danese et al., Accumulation of the enterobacterial common antigen lipid II 57. T. Ohkawa, Studies of intracellular thymidine nucleotides. Thymineless death and the biosynthetic intermediate stimulates degP in Escherichia coli. J. Bacteriol. 180, 5875–5884 (1998). recovery after re-addition of thymine in Escherichia coli K 12. Eur. J. Biochem. 60, – 39. E. Rotman, P. Bratcher, A. Kuzminov, Reduced lipopolysaccharide phosphorylation in 57 66 (1975). “ Escherichia coli lowers the elevated ori/ter ratio in seqA mutants. Mol. Microbiol. 72, 58. M. E. Bayer, The fusion sites between outer membrane and cytoplasmic membrane ” 1273–1292 (2009). of bacteria: Their role in membrane assembly and virus infection in Bacterial Outer – 40. A. Witte, W. Lubitz, Biochemical characterization of phi X174-protein-E-mediated Membranes, M. Inouye, Ed. (John Wiley & Sons, New York, 1979), pp. 167 202. ’ lysis of Escherichia coli. Eur. J. Biochem. 180, 393–398 (1989). 59. E. d Alençon, A. Taghbalout, R. Kern, M. Kohiyama, Replication cycle dependent as- 41. E. Dassa, P. L. Boquet, ExpA: A conditional mutation affecting the expression of a sociation of SeqA to the outer membrane fraction of E. coli. Biochimie 81, 841–846 group of exported proteins in Escherichia coli K-12. Mol. Gen. Genet. 181, 192–200 (1999). (1981). 60. A. A. Nicolaidis, I. B. Holland, Evidence for the specific association of the chromosomal 42. A. Zaritsky, C. L. Woldringh, M. Einav, S. Alexeeva, Use of thymine limitation and origin with outer membrane fractions isolated from Escherichia coli. J. Bacteriol. 135, thymine starvation to study bacterial physiology and cytology. J. Bacteriol. 188, 178–189 (1978). 1667–1679 (2006). 61. M. Toyofuku, N. Nomura, L. Eberl, Types and origins of bacterial membrane vesicles. 43. H. A. Sutterlin et al., Disruption of lipid homeostasis in the Gram-negative cell en- Nat. Rev. Microbiol. 17,13–24 (2019). velope activates a novel cell death pathway. Proc. Natl. Acad. Sci. U.S.A. 113, 62. N. J. Bitto et al., Bacterial membrane vesicles transport their DNA cargo into host cells. E1565–E1574 (2016). Sci. Rep. 7, 7072 (2017). 44. A. Mychack et al., A synergistic role for two predicted inner membrane proteins of 63. G. B. Ogden, M. J. Pratt, M. Schaechter, The replicative origin of the E. coli chromo- Escherichia coli in cell envelope integrity. Mol. Microbiol. 111, 317–337 (2019). some binds to cell membranes only when hemimethylated. Cell 54, 127–135 (1988).

33560 | www.pnas.org/cgi/doi/10.1073/pnas.2012254117 Rao and Kuzminov Downloaded by guest on September 29, 2021