Defects in the central metabolism prevent thymineless death in Escherichia coli, while still allowing significant protein synthesis

Abstract

Starvation of Escherichia coli thyA auxotrophs for the required thymine or thymidine leads to the cessation of DNA synthesis and, unexpectedly, to thymineless death (TLD). Previously, TLD-alleviating defects were identified by the candidate gene approach, for their contribution to replication initiation, fork repair, or SOS induction. However, no TLD-blocking mutations were ever found, suggesting a multifactorial nature of TLD. Since (until recently) no unbiased isolation of TLD suppressors was reported, we used enrichment after insertional mutagenesis to systematically isolate TLD suppressors. Our approach was validated by isolation of known TLD-alleviating mutants in recombinational repair. At the same time, and unexpectedly for the current TLD models, most of the isolated suppressors affected general metabolism, while the strongest suppressors impacted the central metabolism. Several temperature-sensitive (Ts) mutants in important/essential functions, like nadA, ribB, or coaA, almost completely suppressed TLD at 42°C. Since blocking protein synthesis completely by chloramphenicol prevents TLD, while reducing protein synthesis to 10% alleviates TLD only slightly, we measured the level of protein synthesis in these mutants at 42°C and found it to be 20–70% of the WT, not enough reduction to explain TLD prevention. We conclude that the isolated central metabolism mutants prevent TLD by affecting specific TLD-promoting functions.

Introduction

Thymineless death (TLD) is a mysterious phenomenon in which cultures of thyA thymidine auxotrophs, which can synthesize DNA precursor dTTP only when supplied with exogenous thymine or thymidine (collectively referred to as “T”), not only stop DNA synthesis upon removal of T-supplementation (T-starvation), but unexpectedly suffer a deep loss of titer (Ahmad et al. 1998; Khodursky et al. 2015). Discovered 70 years ago in Escherichia coli (Barner and Cohen 1954; Cohen and Barner 1954), TLD is observed in other bacteria, in eukaryotic organisms including fungi and human cells, and can also be induced chemically by inhibiting dTMP synthesis (Ahmad et al. 1998). In fact, induction of TLD-like phenomena is the modus operandi of several antibacterial, anticancer, and immunomodulating agents (Cohen 1971; Schilsky 1992; Van Triest and Peters 1999; Genestier et al. 2000; Ladner 2001; Khodursky et al. 2015). However, in spite of its biomedical significance and years of research, the molecular mechanisms of TLD remain unclear, potentially hampering development of improved therapeutics.

The natural suspicion that TLD targets the chromosomal DNA was confirmed early on (Fuerst and Stent 1956; Gallant and Suskind 1961; Maaløe and Hanawalt 1961). Perplexingly, at first, the physical studies detected no significant changes in the chromosome during T-starvation (Freifelder and Maaloe 1964; Smith and Burton 1965)—although later this was qualified by reports of a few ss-breaks (Freifelder 1969) and a minor DNA loss at the replication points (Reiter and Ramareddy 1970). More recent measurements confirm the overall stability of the chromosome in T-starved cells (Kuong and Kuzminov 2012; Rao and Kuzminov 2019). Because of the unclear nature of DNA problems, the initial TLD concept had to be broad, postulating that T-starvation “unbalances cell growth” by blocking DNA synthesis while allowing normal RNA and protein synthesis (Cohen 1971). This idea is consistent with the observation that slowing down cell growth during T-starvation invariably reduces TLD depth (Barner and Cohen 1957; Freifelder and Maaloe 1964; Khan and Kuzminov 2019). In fact, “balancing” the overall biopolymer production by blocking either protein synthesis (Okagaki et al. 1960) or RNA synthesis (Barner and Cohen 1954, 1958) does prevent TLD. Unfortunately, the idea of unbalanced growth offered little mechanistic insights about how unspecified (and undetectable) chromosomal problems in T-starved cells caused TLD—and was eventually superseded by various models, discussed below, linking TLD with specific (testable) phenomena of the DNA metabolism. Although a timely and welcome development, testing specific models unexpectedly led into the current dead-end.

One of the reasons that 7 decades of research failed to unravel the mechanisms of TLD was the reliance on the natural assumption that TLD was due to disruption in the DNA replication and repair (Ahmad et al. 1998; Khodursky et al. 2015). Most of the genetic studies, therefore, sought TLD-alleviation in defects of the DNA metabolism, employing candidate gene approaches. For example, blocking initiation of DNA synthesis would prevent or significantly reduce TLD (Bouvier and Sicard 1975; Martín and Guzmán 2011; Khan and Kuzminov 2019). Also, long-lived single-strand gaps were expected to be the initial DNA irregularity accumulating in T-starved cells (Ahmad et al. 1998; Kuong and Kuzminov 2010, 2012). Indeed, TLD is alleviated by inactivation of the RecAFOR pathway of recombinational repair of blocked single-strand gaps (Ahmad 1980; Nakayama et al. 1982, 1988; Fonville et al. 2010; Kuong and Kuzminov 2010). However, while inactivation of the recAFJOQR pathway indeed alleviates TLD, the death was never prevented completely even by the combinations of several different alleviating mutations (Fonville et al. 2010).

In the last 20 years, several new models of TLD have been proposed based on additional TLD-alleviating conditions or defects, although, again, none of them could explain the lethality completely. To make matters worse, some of the promising leads failed to confirm in subsequent studies, such as the report implicating a toxin-antitoxin module in TLD (Sat et al. 2003; Fonville et al. 2010). Another interesting TLD explanation was the loss by T-starved cells of the capacity to scavenge endogenous hydrogen peroxide—which was even backed by an unexpected enhancement of TLD by exogenous hydrogen peroxide (Hong et al. 2017). However, subsequent examination showed that various deficiencies in hydrogen peroxide scavenging have no effect on TLD, while sublethal hydrogen peroxide treatments (because they transiently knock out metabolism) completely block TLD (Rao and Kuzminov 2022).

The most dramatic TLD phenotype discovered in the last 15 years—a striking (though again partial) loss of the origin-centered chromosome macrodomain (Sangurdekar et al. 2010; Kuong and Kuzminov 2012; Rao and Kuzminov 2021)—gave hopes of a long-awaited breakthrough, but failed to explain complete chromosome inactivation or to account for TLD in mutants, like recA or recBCD, with no or limited origin macrodomain loss (Kuong and Kuzminov 2012). Another exciting possibility that T-starvation caused cell envelope damage also turned out to be specific to certain mutants (rffC and rffT), yielding no cell lysis if those functions were intact (Rao and Kuzminov 2020). Finally, we have recently tested the longest-standing basic TLD “facts” and demonstrated (by painstaking experimentation, because the results ran contrary to common sense expectations) that the futile incorporation–excision cycles, active DNA replication, futile replication fork breakage–repair cycles, or T-starvation–induced chromosome fragmentation could all be eliminated without abolishing TLD (Khan and Kuzminov 2019, 2022; Rao and Kuzminov 2019).

As the phenomenon of T-starvation–caused lethality remains a mystery, while the number of testable ideas about the possible nature of TLD keeps shrinking (Rao and Kuzminov 2021, 2022), fresh insights into the TLD enigma are desperately needed. We realized that there was a lack of systematic efforts to isolate TLD-resistant mutants; indeed, of the dozens of previous studies on TLD genetics, only 1 reported an unbiased enrichment for TLD-resistant mutants—yielding the recQ defect (Nakayama et al. 1984). Since the general protein synthesis inhibition blocks TLD (Okagaki et al. 1960; Cummings and Kusy 1969; Khan and Kuzminov 2019), suggesting the possibility of induction of still-unknown TLD-causing functions, we sought to systematically isolate additional TLD-resistant mutants.

While our efforts were ongoing, a new study linked TLD with intracellular acidification by tracking changes in random insertion libraries after 3 h of T-starvation (Ketcham et al. 2022). Even though most of the identified TLD-alleviated mutants in that study were metabolism-defective slow-growers (see the Discussion), this report demonstrated (i) productivity of enrichment for TLD-alleviated mutants and (ii) unappreciated metabolic dimensions to TLD.

Here, we describe a general enrichment strategy, based on better survival in the classic T-starvation protocol, for isolation of TLD-alleviating mutants (Fig. 1a). We further enhanced it by performing some enrichments at 42°C with subsequent outgrowth at 28°C or 22°C—designed to isolate TLD-alleviating Ts mutants. Using this strategy, we have isolated dozens of novel mutants exhibiting various levels of TLD resistance. Interestingly, completely unpredicted by the current TLD models (Fonville et al. 2010; Kuong and Kuzminov 2012; Khodursky et al. 2015), but in agreement with the recent report (Ketcham et al. 2022), most of the isolated mutants affect general metabolism, rather than DNA replication or repair. The diversity of TLD-relieving defects highlights a close link between the central metabolism and TLD; investigation of this metabolic lead may eventually provide new insights into the TLD mystery.

Fig. 1.

Our strategy, protocols, and isolation of recombinational repair mutants as a confirmation that enrichment for TLD alleviation works. a) The scheme of enrichment for TLD-resistant mutants. b) Comparison of the 2 T-starvation protocols. The difference is in the pre-starvation phase: in the standard protocol, the culture is growing, while in the awakening protocol, the culture is stationary. There is no difference between the 2 protocols in either the starvation phase or the outgrowth phase. c) Position and orientation of pRL27 inserts in the dnaA-gyrB region. d) TLD effects of select suppressors under the standard conditions of T-starvation at 42°C (the default assay). All mutants, except recF, were isolated only once. Statistical significance was tested by Student's test, one-tail, assuming equal variance. *P < 0.01; **P < 0.001; ***P < 0.0001. The P-values are given only for P > 0.01. e) The position and orientation of the 2 inserts around priA. f) TLD time course of the 2 insertions around priA.

Materials and methods

Bacterial strains, plasmids, and growth conditions

All strains used in this study were E. coli K12 derivatives in the AB1157 background (Supplementary Table 1). Strains were maintained at room temperature on LB agar plates supplemented with 10 μg/mL thymidine. As before, all strains used for TLD assays, besides the ΔthyA mutation, also carry a deletion of the entire deoCABD operon, to reduce their thymidine requirements (Kuong and Kuzminov 2010; Khan and Kuzminov 2019); for clarity, we refer to this AB1157 ΔthyA ΔdeoCABD parental strain as simply “thyA” throughout this study. Antibiotics, when required, were used at the following final concentrations: ampicillin 100 μg/mL and kanamycin 10 or 50 μg/mL. MOPS minimal phosphate medium (Bochner and Ames 1982) supplemented with 0.2% glucose, 0.2% casamino acids (CAAs), and 20 μg/mL thiamine (called MOPS-CAA in this study) was used for T-starvation of thyA mutants, whereas the same medium supplemented with 10 μg/mL thymidine (called MOPS-CAA + dT) was used to pre-grow strains for TLD assays (Khan and Kuzminov 2019). MOPS-CAA + dT agar contained 15 g agar in 1 L of MOPS-CAA + dT medium and was used for plating electroporants during preparation of pRL27 insertion libraries. Electroporation was performed in 1 mm gap disposable cuvettes (Fisher Scientific) using BioRad Electro Pulser (Bio-Rad). Drug-resistant markers were moved among strains by P1 transduction (Miller 1972). Plasmid constructs are listed in Supplementary Table 2. Plasmid DNA was prepared by alkaline lysis procedure (Birnboim 1983), while genomic DNA was prepared using the standard phenol extraction (Kouzminova and Kuzminov 2006).

The 10 bp deletion from the extreme 3′ end of the coaA gene was made by 1-step gene replacement protocol (Datsenko and Wanner 2000) and confirmed by PCR and by the inability to form colonies at 42°C. The ribB::pRL27, coaA::pRL27, and coaA(Δ10 bp) strains were further confirmed by whole-genome sequencing. For the ribB and coaA complementation studies, genes with their corresponding promoters were PCR-amplified and cloned in pBR322 (Bolivar et al. 1977) or pGB2 (Churchward et al. 1984) vectors. For extra-chromosomal expression of GhxP and BglB proteins, the CDS of the corresponding genes were PCR-amplified and cloned under the lac promoter in pLAC22 (Warren et al. 2000). For complementation of rfbA mutants, various lengths of the rfbA gene were PCR-amplified and cloned in pLAC22. To further characterize the thyA rfbA::pRL27 mutant, the rfbA::pRL27 fragment with upstream and downstream sequences was excised from the genome as an EcoRV fragment and cloned into pBR322. Precise rfbA::pRL27 fragment without any flanking DNA was generated by PCR, using the genome of thyA rfbA::pRL27 mutant as template, and also cloned in pBR322. Plasmids expressing rfbC were generated by PCR amplifying and cloning rfbC CDS or rfbA::pRL27-rfbC fragment of thyA rfbA::pRL27 mutant under the lac promoter in pLAC22.

T-starvation assays

The 2 types of T-starvation protocols were as described (Khan and Kuzminov 2019) (Fig. 1b). In the standard T-starvation assays, thyA mutants and their derivatives were grown at 28°C in MOPS-CAA + dT medium for 15–18 h (O/N), diluted in the morning 50 times into the same medium and further grown at the desired temperatures until OD₆₀₀ ∼ 0.1–0.15. At this point, cells were harvested by centrifugation, and cell pellets were resuspended and re-harvested 3 times using 1 mL of sterile 1% NaCl to remove traces of thymidine. After the final wash, T-starvation was initiated by suspending the cells in the volume of MOPS-CAA to yield an OD of 0.1–0.15 and continued for 5 h at temperatures described in the assays.

In the awakening T-starvation assay (Fig. 1b), strains were grown at 37°C in MOPS-CAA + dT medium for 18 h or at 28°C in MOPS-CAA + dT medium for 48 h (to saturation in both cases), harvested, and cell pellets were washed twice in 1% sterile NaCl, as above (1 fewer wash than in the standard protocol, because dT was exhausted in saturated cultures). After the second wash, cells were suspended in the same volume of 1% NaCl and diluted 20 times in MOPS-CAA to initiate T-starvation, which in this case was performed at 37°C or 42°C, for 6–8 h (Khan and Kuzminov 2019).

To quantify TLD in the standard or awakening conditions, serial dilutions of cultures (in 1% NaCl) were spotted on LB plates and developed at room temperature for 15–20 h. At this point, pin-prick colonies were counted under a stereo microscope, and the titer of T-starved cultures was determined using spots containing between 10 and 100 colonies. Fewer than 10 counts per spot result in a significant titer uncertainty, while more than 100 counts per spot underrepresent the actual titer, because of the colony overlap (double colonies).

Creation of transposon libraries

For insertional mutagenesis in the thyA mutant, we used plasmid pRL27 that carries a Tn5-based insertion cassette conferring kanamycin resistance, with a cognate transposase gene positioned outside the insertion cassette (Larsen et al. 2002). Multiple libraries were created by electroporating 10 ng of pRL27 into thyA mutant and selecting for kanamycin-resistant electroporants on LB + dT or MOPS-CAA + dT plates at various temperatures (Supplementary Table 3). After O/N incubations, plates containing ≥10,000 colonies were overlaid with 1–2 mL of MOPS-CAA + dT, and the colonies were scraped into the medium to make suspensions of insertion libraries, which were then collected into sterile vials, made 15% for glycerol, and stored at −80°C. These typically 30,000–50,000 candidate-strong libraries were subsequently used for isolation of TLD-resistant mutants.

Enrichment of TLD-resistant mutants

Frozen libraries were thawed and diluted 100–1,000-fold in MOPS-CAA + dT to start the cultures, which were grown at various temperatures (28, 37, or 42°C, as described below and in the Results) to an OD ∼0.1 to 0.2 and subjected to standard T-starvation at the designated temperatures, as described above. At the end of starvation, the cultures were harvested, diluted, and plated, in all but 1 assay, on MOPS-CAA + dT + kan plates. In various assays, the plates were incubated at RT, 28°C, 37°C or 42°C (Supplementary Table 4) for 15–18 h, and the growing lawns of survivors were scraped and suspended into 1–2 mL of MOPS-CAA, diluted to an OD of 0.03–0.04 in MOPS-CAA + dT, and subjected to another cycle of the above enrichment. The cycles were repeated until resistance to TLD was observed (Fig. 1a), which took an average of 6 rounds of enrichment.

To isolate candidates under the awakening conditions, libraries were grown at 37°C in MOPS-CAA + dT O/N, and the saturated cultures were diluted 20-fold in MOPS-CAA. The awakened cultures were then starved for 8 h at 37°C, after which the starved cells were supplemented with 0.2% glucose, 0.2% CAA, 4 μg/mL thiamine, 10 μg/mL thymidine, and 50 μg/mL kanamycin. These recovered cultures were grown overnight and subjected to the same cycle of awakening, starvation, and recovery in liquid culture for 8 rounds.

After the final enrichment round under both the standard and the awakening conditions, the survivors were streaked on LB + dT + kan for isolation, and individual clones were picked and tested for their TLD resistance. To verify the link of resistance with the pRL27 inserts, P1 lysates were made from promising candidates and Kan resistance was transduced back into the thyA mutant. Candidates isolated after cycles of recovery at RT following T-starvations at 42°C were also tested for temperature sensitivity by streaking at 42°C on LB + dT plates. In these candidates, pRL27-linkage of both temperature sensitivity and TLD resistance was confirmed by back transduction of both traits into the original thyA mutant.

Finally, the chromosomal location of pRL27 integration sites was determined by digesting genomic DNA with MluI (which is absent in the insertion cassette), self-ligating the digested fragments, and transforming the ligation reactions into DH5α λ::pir⁺ cells. The plasmids from kanamycin-resistant transformants were screened and sequenced as described before (Bradshaw and Kuzminov 2003).

Protein synthesis assays

Measurement of β-galactosidase activity was performed as described (Miller 1972; Kouzminova et al. 2004). For the total protein synthesis quantification, [4,5-³H] L-leucine (Moravek, Inc.) was directly added to the cultures to the final concentration of 3.3 μCi/mL. At indicated times, 300 μL aliquots were removed and transferred to glass tubes containing 5 mL of chilled 5% TCA. The TCA-precipitated material was then collected on a 1.6 μm glass fiber filter (Fisher Scientific) using vacuum manifold and washed by 5 mL of chilled 5% TCA, followed by 5 mL of ethanol. After the ethanol wash, the filters were air-dried, spotted with 100 μL of 100 mM KOH, and air-dried again. Finally, the filters were transferred to the scintillation tubes and incubated in scintillation cocktail overnight before counting in LS 6500 multipurpose scintillation counter (Beckman Coulter).

Replication and transcription assays

DNA or RNA synthesis was quantified by incorporation, respectively, of [³H] thymidine (Perkin Elmer) or [³H] uridine (Moravek, Inc.), using conditions described above for the measurements of [4,5-³H] L leucine incorporation.

Results

Protocol development and mutant statistics

Starting with the thyA mutant, we enriched for TLD-alleviating mutants (TLD suppressors) (Fig. 1a;Supplementary Fig. 1) under a variety of conditions (Supplementary Table 4). As a mutagen, we used pRL27 insertional cassette (Larsen et al. 2002), which inserts randomly, with occasional hotspots (Kouzminova et al. 2004; Lukas and Kuzminov 2006; Ting et al. 2008; Budke and Kuzminov 2010; Das et al. 2023). We made 10 mutant libraries in total: the first 2 libraries were plated on rich medium (LB + dT), the other 8 were generated on MOPS-CAA + dT plates, to match the T-starvation protocol (Supplementary Table 3). Since we perform T-starvation in the MOPS-CAA medium, the TLD suppressors from LB-outgrown libraries were trivial: they alleviated TLD because of their slow growth in the MOPS-CAA medium. The temperature of library generation (Supplementary Table 3) also affected subsequent enrichments—for example, we failed to isolate temperature-sensitive (Ts) mutants from libraries plated at 42°C. We varied the temperature during the enrichment rounds (Supplementary Table 4) because of the known temperature effects on TLD (Khan and Kuzminov 2019). Besides, by using 42°C during the enrichment, but 28°C or lower during the outgrowth, we specifically went after potential Ts mutants in essential functions.

Currently, there are 2 established protocols of T-starvation (Fig. 1b): the standard assay switches exponentially growing cultures to the same medium lacking dT, while the recently developed awakening protocol dilutes stationary cultures into the fresh medium lacking dT; both result in a similar, about 3 orders of magnitude, death within several hours (Khan and Kuzminov 2019). While most (80 out of 88) of our enrichments were performed using the standard protocol, 8 were done under the awakening conditions. Enrichments under the standard conditions of T-starvation were further differentiated through recovery of TLD survivors either at 42°C or at RT/28°C (Supplementary Table 4). We did not directly compare the effect of recovery temperatures, by plating TLD survivors at 2 different temperatures in the same enrichment round.

Within the 88 individual enrichment rounds, we performed a total of 514 cycles of T-starvation and isolated 63 TLD-alleviating mutants (some of the pRL27 inserts are shown in Supplementary Fig. 2). A total of 61 inserts landed in the coding sequence of 23 genes, whereas 2 pRL27 inserts were recovered in intergenic regions. Of the 23 genes interrupted by pRL27, 16 were hit only once, whereas 7 genes were hit more than once, with 1 gene receiving the record 19 inserts (Table 1).

Table 1.

The genes, the number of hits, and the functions/processes affected by pRL27 inserts.

Gene	# of hits	Protein/enzyme	The affected function/process
coaA	19	Pantothenate kinase	CoaA biosynthesis
recF	8	RecA-loading on ssDNA gaps	Recombinational repair
rfbA	5	Glucose-1-phosphate thymidyltransferase I	Rhamnose biosynthesis
ribB	4	3,4-Dihydroxy-2-butanone-4-phosphate synthase	Riboflavin biosynthesis
ghxPa	4	Guanine/hypoxanthine permease	Transport of purines
nadA	3	Quinolinate synthase	NAD biosynthesis
priA	2	Primosomal protein N	Replication fork restart
bglB	2	6-Phospho-β-glucosidase B	β-Glucosides catabolism
nrdE	1	Ribonucleoside-diphosphate reductase-2-subunit alpha	Ribonucleotide synthesis
pgk	1	Phosphoglycerate kinase	Glycolysis and gluconeogenesis
gltB	1	Glutamate synthase large subunit	Glutamate biosynthesis
yjcS	1	Linear primary-alkylsulfatase	Putative sulfatase
ygiQ	1	Radical SAM superfamily protein	Uncharacterized function
rfbB	1	dTDP-glucose 4,6-dehydratase 1	Rhamnose biosynthesis
rfbD	1	dTDP-4-dehydrorhamnose reductase	Rhamnose biosynthesis
recJ	1	ssDNA-specific exonuclease	DNA repair and recombination
recR	1	RecA-loading on ssDNA gaps	Recombinational repair
hsrA	1	Putative transporter	Uncharacterized
ptsG	1	Glucose-specific PTS enzyme IIBC component	Glucose uptake
ptsN	1	PTS IIA–like regulatory protein	Potassium homeostasis
soxS	1	Superoxide response, transcriptional regulator	Oxidative stress
zur-csbD	1	intergenic region	Unknown
gadEa	1	Transcriptional activator	Acid resistance
ybdMa	1	ParB-like nuclease domain-containing protein	Uncharacterized

^aIsolated in the awakening conditions only.

While most of the isolated TLD suppressors were reported for the first time, a significant number of known TLD suppressors inactivating the RecAFOR pathway of recombinational repair were also isolated (see below). The fact that more than 10% of the isolated TLD suppressors have inactivated the known TLD-promoting pathway validated our enrichment strategy. Below, we report the TLD effects of the inactivation of genes with more than 1 insert, as well as groups of single-insert genes that work in the same pathway. Since TLD is reduced in slow-growing conditions or mutants (Khan and Kuzminov 2019; Ketcham et al. 2022), for the mutants with the strongest TLD effects, we also compared their growth rates to that of the WT.

Surprisingly, beyond the known/suspected suppressors in recombinational repair (which we will discuss first), the bulk of our isolated suppressors were in the central metabolism (Table 1), with no apparent linkage to chromosome problems. We will only discuss the genes/pathways/areas of metabolism that were hit independently more than once, presenting them in the order of the strength of TLD alleviation. With suppressors reported for the first time, we will start with the unexpected inserts in the rfb genes. We will then move to inserts that alleviated TLD significantly, for which we had to make sure that these mutants were not slow-growers (as explained in the Introduction). We finish our presentation with inserts that blocked TLD completely at 42°C; since these mutants struggled to grow at 42°C and were clearly metabolically affected, we further characterized them by comparison of their effects with those of chloramphenicol block to protein synthesis.

The recombinational repair mutants: recF, recJ, recR, and priA

As explained in the introduction, TLD was always assumed to kill cells by irreversibly damaging their replicating chromosomes. The 2 observations supporting such thinking was TLD alleviation by (i) prevention of chromosome replication via blocking its initiation with either dnaA(Ts) or dnaC(Ts) defects at 42°C or by inhibiting transcription or translation (Bouvier and Sicard 1975; Martín and Guzmán 2011; Khan and Kuzminov 2019) and (ii) inactivation of the RecAFOR pathway of recombinational repair (Ahmad 1980; Nakayama et al. 1982, 1988; Fonville et al. 2010; Kuong and Kuzminov 2010). In contrast to the RecABCD pathway of recombinational repair that mends double-strand DNA breaks, the RecAFOR pathway is responsible for mending blocked ss-gaps (Kuzminov 1999)—which are exactly the DNA lesions expected to accumulate at T-starved replication forks (Kuong and Kuzminov 2010; Rao and Kuzminov 2019). Although both types of suppressors (dnaA/dnaC or recFOR) were either partial or transient (the latter ones delaying rather than decreasing the final lethality), their isolation or confirmed effects provided a powerful argument for the replicative nature of TLD.

Our enrichment strategy was validated by frequent isolation of the recF mutants known to confer more than 1 order of magnitude TLD-alleviation after several hours of T-starvation (Nakayama et al. 1982; Fonville et al. 2010; Kuong and Kuzminov 2010). In fact, 8 of the 80 enrichment runs under the standard TLD conditions yielded recF inserts—all unique and curiously concentrated in the middle third of the ORF (Fig. 1c; Supplementary Fig. 2a). We also isolated (Supplementary Fig. 2) single insertions inactivating the genes either known (recJ; Nakayama et al. 1988; Fonville et al. 2010) or expected (recR) to cause the same partial TLD suppression, as the corresponding proteins work together in the RecAFOR pathway (Kuzminov 1999).

Their functioning in the same pathway is reflected by their similar TLD effects after 5 h of T-starvation (Fig. 1d). Taking this TLD effect of the recF/J/R inactivations as the typical scale of TLD suppression, we could assign other isolated suppressors into either the weak ones, 10-fold or less improvement (nrdE and soxS), the standard suppressors, from 1 to 2 orders of magnitude improvement (ptsG and pgk), and strong ones, more than 100-fold improvement (ptsN; Fig. 1d).

Returning to suppressors in recombinational repair, we also isolated 2 inserts apparently affecting the priA gene (Fig. 1e) of the PriABC-DnaT replication fork restart pathway (Michel and Sandler 2017). Due to the extreme growth defect of a complete priA inactivation (Nurse et al. 1991), inserts within the priA ORF itself were not expected in our protocol. Isolation of TLD suppressors around priA confirms replication forks problems during T-starvation, most likely due to the futile cycle of fork disintegration and restart (Rao and Kuzminov 2019). Unfortunately, both priA inserts suppressed TLD only weakly and only after 5 h of T-starvation (Fig. 1f). At the same time, this rather disappointing result is compatible with the current understanding that the bulk of TLD is independent of the chromosomal DNA replication (Khan and Kuzminov 2019).

Unexpected alleviation by inserts affecting low-molecular-weight dT pools

An unexpected TLD suppressor, rfbA, was isolated 5 times, with independent insertions from 5 different libraries in 4 separate enrichment runs (Table 1). Unlike recF and ghxP, with inserts at various locations and in both orientations (Fig. 1c;Supplementary Fig. 2a), all 5 rfbA inserts were co-oriented with the gene and occurred only at 2 locations, both at the 5′-end of the gene, with 4 of the 5 inserts landing in the exact same position 180 bp into the gene (Fig. 2a)—apparently a hotspot for pRL27 hops.

Fig. 2.

The mysteries of rfbA and bglB inserts. a) Position and orientation of pRL27 inserts in the rfbBDAC region. b) The TLD effect of pRL27 insertion in rfbA locus, and its (over)complementation with rfbA + from a plasmid. c) TLD effects of the pRL27 insertions into rfbB and rfbD genes, as well as of ΔrfbA. d) Thymineless death profiles of thyA mutant expressing full-length and truncated versions of rfbA genes. See Supplementary Fig. 4a for explanation of the alleles. e) The position and orientation of the two bglB inserts. f) TLD effect of a bglB insert, as well as of additional bglB + copies on a plasmid.

Repeated isolation of rfbA inserts as TLD-alleviating mutations contradicted our own recent report that a complete deletion of rfbA, which codes for the enzyme glucose-1-phosphate thymidyltransferase I, modestly aggravates TLD (Rao and Kuzminov 2020). The rfbA::pRL27 inserts isolated in this study, however, show the opposite phenotype—causing about one order of magnitude relief in TLD (Fig. 2b)—which evidently allowed to enrich for them. Moreover, expressing rfbA + from a plasmid under the control of IPTG-inducible lac promoter aggravates TLD (Fig. 2b), confirming the TLD-relieving phenotype of rfbA inserts.

The rfbA gene is part of a 4-gene operon rfbBDAC encoding enzymes for the synthesis of dTDP-rhamnose (Marolda and Valvano 1995; Fig. 2a). The TLD effect of this pathway inactivation was further confirmed by our isolation of pRL27 inserts in 2 upstream genes of the same operon, rfbB and rfbD, which code for dTDP-glucose 4,6-dehydratase 1 and dTDP-4-dehydrorhamnose reductase, respectively. When tested together, mutations in all 3 genes caused similar reduction in TLD (Fig. 2c). Since the orientation of pRL27 inserts was identical in the 3 genes (Fig. 2a), we tested whether the TLD relief comes from overexpression of the downstream rfbC⁺ gene but found no effect of rfbC⁺ plasmids, in either thyA single mutant or in the thyA rfbA double mutant (Supplementary Fig. 3Ab).

To clarify whether deletion of rfbA indeed causes aggravated TLD, whereas co-directional pRL27 inserts into the same gene result in TLD relief, we tested in parallel thyA ΔrfbA::kan vs thyA rfbA::pRL27 strains (Supplementary Fig. 3c). Under our standard T-starvation conditions (MOPS-CAA at 42°C), both mutants showed identical TLD relief of about 1 order of magnitude (Fig. 2, b vs c). Similar TLD relief was also observed when the 2 strains were T-starved in M9-CAA medium (used by Rao and Kuzminov 2020) at 42°C (Supplementary Fig. 3c, left), so the behavior of the mutants is not due to the differences between MOPS-CAA vs M9–CAA media. At the same time, temperature seems to affect the TLD-rescuing effect of the rfbA inserts, as no significant TLD change was observed in either of the 2 strains when starvation was done in M9–CAA at 37°C (Supplementary Fig. 3c, right)—which was the T-starvation conditions used in the previous work (Rao and Kuzminov 2020).

Since all the rfbA inserts were at the N-terminal part of the gene and co-directional with it (Fig. 2a), it was possible that expression of the truncated protein, rather than total loss of the RfbA function, was responsible for the reduced TLD in the rfbA::pRL27 mutants. To test this possibility for the inserts at position 180 bp from the 5′ end of rfbA, we cloned into an expression plasmid 2 downstream parts of the gene: the complete 702 bp sequence downstream of the insertion site (which would use vector's ATG codon for initiation, resulting in a 234 aa piece) and the shorter 516 bp sequence starting at the first internal ATG codon (coding for an 171 aa piece; Supplementary Fig. 4a). To compare, the full-length RfbA is 293 aa long. When expressed in the thyA mutant, truncated RfbA proteins failed to reduce TLD to the levels shown by the thyA rfbA::pRL27 mutant (Fig. 2d). Truncated RfbA also failed to further increase the TLD resistance of thyA rfbA::pRL27 mutant (not shown), whereas full-length RfbA in the thyA single mutant again caused slightly faster TLD (Fig. 2d), comparable to the effect of the full-length RfbA expression in the thyA rfbA mutants (Fig. 2b).

To test the possibility that a partially expressed RfbA, being driven by the kanamycin promoter of pRL27 insert, was responsible for the observed TLD resistance, we cloned the rfbA-centered chromosomal region from the thyA rfbA::pRL27 mutant. We made 2 constructs: a bigger one (designated “6.6”) harboring the rfbA::pRL27 inactivated gene together with rfbB, rfbD, galE, and parts of rfbC and wcaM genes and a smaller one (designated “3.3”) containing full rfbA::pRL27 fragment flanked by small portions of rfbC and rfbD genes (Supplementary Fig. 4b). However, neither of these plasmids caused TLD reduction in thyA strain or caused significant change in the TLD profile of thyA rfbA::pRL27 or thyA ΔrfbA::kan mutants (Supplementary Fig. 4c). So it was rfbA inactivation, rather than expression of specific rfbA::pRL27 inserts, that affected TLD. At the end of this lengthy and futile search for the nature of rfbA suppression, we can only conclude that TLD relief could be linked to the reduction in the dTDP-glucose synthesis, reported for the rfbA mutants (Rao and Kuzminov 2020).

Two identical, yet independent, pRL27 inserts were isolated in bglB (Fig. 2e), the gene coding for the enzyme 6-phospho-β-glucosidase B, which is a part of the cryptic bgl operon that produces proteins for uptake and hydrolysis of aryl-β, D-glucosides, such as salicin and arbutin (Schnetz et al. 1987). One of the products of BglB-catalyzed reaction is D-glucopyranose 6-phosphate, which is an immediate precursor for α-D-glucopyranose 1-phosphate, the substrate for the RfbA enzyme (Kuhn et al. 1988). Therefore, it is possible that the lack of BglB causes dTDP-glucose reduction, phenocopying the rfbA defect. However, our quantitative TLD kinetics showed no significant differences between thyA and thyA bglB mutants, while complementation with bglB + also failed to clarify the matter (Fig. 2f). Thus, the bglB inserts belong to a group of inserts (4 more examples are shown in Supplementary Fig. 5) that failed to produce statistically significant TLD alleviation in the standard assay after 3 repetitions. The fleeting TLD effects of some isolated suppressors suggest some uncharacterized differences in T-starvation conditions between the enrichment and the quantitative assay.

Defects of the central metabolism

As noted above, a majority of the sequenced candidates (about 70%), were isolated only once (Table 1 and Supplementary Fig. 2b) and some of them alleviated TLD by 1 order of magnitude or less in the standard assay (Fig. 1d, Supplementary Fig. 5). At the same time, other single hits showed a TLD relief matching or exceeding the one of the recF mutant, like the inserts in the pgk, ptsG, and ptsN genes (Figs. 1d and 3a; Supplementary Fig. 2b). Since these mutants could relieve TLD by growing slower in MOPS-CAA at 42°C (our standard TLD assay), we followed their growth in cultures. Indeed, all 3 mutants grew somewhat slower at 42°C in MOPS-CAA (+dT) medium, while pgk mutant also showed 1 h lag (Fig. 3b). The pgk mutant also grew slower at 28°C in MOPS-CAA (+dT; Fig. 3c), so its TLD alleviation could be ascribed to its slower growth in our T-starvation medium. At the same time, the growth characteristics of the ptsG and ptsN mutants were close to WT (Fig. 3b and c), suggesting that growth retardation is not the only way to relieve TLD.

Fig. 3.

The TLD and growth phenotypes of the strong metabolic suppressors. a–c) concern the inserts into ptsG, ptsN, and pgk. a) The TLD effects. b) Exponential growth at 42°C in MOPS-CAA (+dT). Strains were grown overnight at 28°C in MOPS-CAA (+dT) and subcultured in the same medium at 42°C. c) Optical densities of strains after overnight growth at 28°C or 42°C, in either MOPS-CAA (+dT) or LB (+dT). d–g) concern the 3 inserts into nadA. d) Position and orientation of the inserts. e) The TLD effects. f) Growth at 42°C (like in panel B). g) Optical densities of strains after overnight growth (like in c).

Unlike the rfb TLD-alleviating mutants, which potentially affect the DNA precursor pools, the metabolic functions of the ptsG, pgk, and ptsN genes have no apparent relation to the DNA metabolism. The essential pgk gene codes for phosphoglycerate kinase, a conserved and key enzyme of the central metabolism, catalyzing interconversion of 1,3-bisphosphoglycerate and glycerate 3-phosphate during glycolysis and gluconeogenesis (Irani and Maitra 1977). The pRL27 insert in pgk was at the very 3′ end and co-directional (Supplementary Fig. 2b), suggesting dysregulation, rather than inactivation, of the mutant protein. The ptsG gene encodes EIICB^glc, the glucose-specific enzyme IIBC component of phosphoenolpyruvate-dependent phosphotransferase system (PTS), which participates in the uptake of glucose and structurally related sugars (Zeppenfeld et al. 2000). Finally, the ptsN gene encodes a protein called EIIA^Ntr which controls the intracellular levels of potassium by regulating K⁺ transporter TrkA and K⁺ sensor KdpD (Lee et al. 2007; Lüttmann et al. 2009). The ptsN mutations are pleiotropic and affect nutrient utilization and growth (Powell et al. 1995).

The 3 pRL27 inserts in various positions of the nadA gene (Table 1 and Fig. 3d; Supplementary Fig. 2) all caused significant TLD reduction, with some variation in the degree of the effect (Fig. 3e). The nadA gene codes for the enzyme quinolinate synthase catalyzing the first committed step in the de novo biosynthesis of nicotinamide adenine dinucleotide (NAD; Foster 2004). Its 2 forms, NAD⁺ and NADH, are involved in various electron-transfer transactions. Since this makes NAD a critical metabolic cofactor (e.g. nadA mutants cannot grow in the glycerol-supplemented minimal M9; Joyce et al. 2006), it was likely that nadA::pRL27 insertions relieve TLD by severely affecting the central metabolism and therefore slowing down growth. Indeed, we found that the nadA1 mutant was growing much slower in MOPS-CAA at both temperatures, whereas the nadA3 mutant was Ts for growth in both MOPS-CAA and LB (Fig. 3f and g;Supplementary Fig. 6). However, the nadA2 mutant showed growth characteristics comparable to its nadA⁺ parent, while still exhibiting strong TLD relief (Fig. 3e–g)—thus providing another example of TLD suppression by a defect in the central metabolism. The recent report of mostly metabolic TLD-alleviating mutants also featured nadA and nadB inserts (Ketcham et al. 2022).

TLD-blocking defects

The strongest TLD suppressors were isolated when we started carrying out pre-growth and T-starvation at 42°C, while recovering TLD survivors at 22/28°C, aiming to encourage isolation of Ts mutants. Indeed, after 30 enrichment reactions (done months apart in 3 rounds of 10 reactions each), 200 cycles of TLD and 26 sequenced candidates, we isolated only 2 non-Ts mutants, recF and ptsN. Out of the 24 Ts mutants, 1 was isolated only once: the nadA3 insert (Fig. 3e–g;Supplementary Fig. 6). Curiously, the remaining 23 Ts suppressors had inserts in only 2 genes, coaA (19 inserts) and ribB (4 inserts; Table 1).

Although both the coaA and ribB genes are essential in E. coli, just like with the essential pgk above, the pRL27 inserts affected the very 3′ ends of the genes (Fig. 4a), and both mutants grew normally at 28°C. While the ribB::pRL27 insert grew normally even at 37°C and struggled to grow only at 42°C, coaA::pRL27 inserts already struggled at 37°C and failed to grow at 42°C (Fig. 4c and d;Supplementary Fig. 7a). Both mutants were also almost completely suppressed for TLD at 42°C (Fig. 4b), a phenotype that we have observed only with the nadA1 mutant before—which in that case was clearly because of nadA1 poor growth in MOPS-CAA (Fig. 3e–g). The ribB and coaA mutants are different in that they have no growth defect in MOPS-CAA medium at 28°C, but at 42°C slow down after 2 h of normal growth (Fig. 4c).

Fig. 4.

The TLD effects and growth phenotypes of the ribB and coaA inserts. a) The sites of pRL27 inserts into coaA and ribB and orientations of the transcription from the kan promoter. Only the nucleotide sequences of the very 3′ ends of the genes are shown. b) The TLD effects (the standard assay at 42°C). c) Exponential growth at 42°C (like in Fig. 3b). d) Growth of thyA coaA and thyA ribB mutants on LB + dT plates at 28°C or 42°C after 24 h incubations. e) Comparison of TLD depth after 5 h at 42°C of 2 coaA::pRL27 insertions with the terminal 10 bp coaA deletion.f–h) Temperature dependence of ribB and coaA TLD effects at various temperatures of pre-growth (for 3 h) and T-starvation. f) Both the pre-growth and T-starvation at 28°C. g) The pre-growth at 28°C, T-starvation at 42°C. h) Both the pre-growth and T-starvation at 42°C.

The coaA gene codes for pantothenate kinase, which converts pantothenate into phosphopantothenate, the rate-limiting step in the biosynthesis of coenzyme A (Begley et al. 2001). Several point mutations in coaA are known to confer temperature sensitivity (Vallari and Rock 1987; Chen et al. 2006). ribB also codes for an essential enzyme, 3,4-dihydroxy-2-butanone-4-phosphate synthase, which converts ribulose-5-phosphate into hydroxyl-butanone-phosphate, the precursor for the critical condensation step in the biosynthesis of riboflavin (Bacher et al. 1996). Since riboflavin is the precursor of the critical cofactors flavin mononucleotide (FMN) and flavin adenine dinucleotide (FAD; Bacher et al. 1996), ribB, like coaA, is a critical metabolic gene. A Ts Tn10 insert ∼45 bp from the (currently recognized) end of ribB was reported before (Raina et al. 1991).

ribB and coaA candidates differed from other multi-hit mutants, recF, ghxP, rfbA, and nadA, where insertions happened throughout the corresponding ORFs (Supplementary Fig. 2a), meaning a likely inactivation of the affected proteins. In contrast, all pRL27 insertions of coaA were found at 2 nearby locations, either 20 or 10 bp from the end of the coaA gene, and they were all co-directional with the gene (Fig. 4a). Similarly, all 4 ribB inserts, although isolated independently, were mapped at the same site 34 bp from the 3′ end of the gene, and again only in the co-directional orientation (Fig. 4a). The co-directional orientation of all inserts suggests that expression of the remaining portions of the affected genes is important. We compared the TLD effects of the two coaA inserts (10 and 20 bp from the 3′ end), as well as with the deletion of the last 10 bp of the coaA gene (we could not build the deletion of the last 20 bp). The 2 coaA insertions were equally effective in abolishing TLD; surprisingly, the constructed thyA ΔcoaA (10 bp) mutant relieved TLD by 1 order of magnitude only (Fig. 4e) while still showing Ts like the thyA coaA::pRL27 mutants (Supplementary Fig. 7b).

Genetics of TLD suppression by ribB and coaA inserts

In agreement with their Ts growth, ribB and coaA mutants showed complete suppression of TLD only when grown and T-starved at 42°C, but not when assayed at 28°C (Fig. 4f–h). Starving the 28°C pre-grown cultures at 42°C caused reduction in TLD relief in ribB, but not in coaA mutant (Fig. 4g)—again in line with the strengths of Ts of the 2 mutants (Fig. 4d). In fact, the coaA mutant showed normal growth at 2 hs at 42°C (Fig. 4c)—but completely suppressed TLD at this time point (Fig. 4g). At the same, time, normal TLD at 28°C in ribB and coaA mutants (Fig. 4f) shows that mutant RibB and CoaA proteins are functional at this temperature.

Arrested growth at 42°C, on the other hand, suggests protein malfunction that could be due to hyperactivity or loss of enzyme activity through misfolding, degradation, or both. However, complementation with the corresponding WT genes from plasmids restored regular TLD levels to both mutants (Fig. 5a and b), suggesting that TLD suppression was due to RibB or CoaA enzyme inactivation. Transducing the mutant alleles into the WT (thyA+) strain confirmed that the Ts in the suppressed thyA mutant was not due to some other unlinked mutations: AB1157 ribB::pRL27 and AB1157 coaA::pRL27 showed much reduced growth and no growth, respectively, when streaked at 42°C (Supplementary Fig. 7a).

Fig. 5.

Complementation with WT genes and suppression of hyper-TLD. a) Complementation of the TLD effect of the ribB insert by ribB + expressed from a low-copy number (pGB) or a medium-copy number (pBR) plasmids. b) Complementation of the TLD effect of the coaA insert by coaA+, similar to a). c) TLD after 5 h of T-starvation at 42°C of several hyper-TLD mutants and their coaA::pRL27 derivatives.

We also transferred the coaA allele into 3 hyper-TLD mutants, thyA ruv, thyA recBCD, and thyA rffHC (Kuong and Kuzminov 2010; Rao and Kuzminov 2020) to determine how the faster and deeper TLD in these strains was affected by our coaA suppressor. The coaA derivatives of hyper-TLD mutants, namely, thyA ruv coaA::pRL27, thyA rffHC coaA::pRL27 and thyA recBCD coaA::pRL27 were also temperature-sensitive for growth when streaked at 42°C (Supplementary Fig. 7c) and showed complete suppression of TLD when T-starved at this temperature (Fig. 5c). In fact, the residual TLD in the thyA recBCD coaA::pRL27 mutant was even less than that observed when thyA recBCD mutant was T-starved in the presence of chloramphenicol to block protein synthesis (Khan and Kuzminov 2022)—making it less likely that simple inhibition of protein synthesis explains the observed TLD block in these mutants (see next).

Inhibition of protein synthesis

Since inhibition of protein synthesis with chloramphenicol blocks TLD (Okagaki et al. 1960; Cummings and Kusy 1969; Khan and Kuzminov 2019), there was a possibility that both ribB and coaA defects at 42°C simply inhibit general protein synthesis to the levels enough to block TLD. To address this possibility, we (i) measured how much ribB and coaA mutant were defective for protein synthesis at 42°C, and (ii) tested whether a similar reduction in protein synthesis by an appropriately low chloramphenicol concentration was sufficient to block TLD.

We first measured cellular protein synthesis by quantifying β-galactosidase via enzymatic assay (Miller 1972). When the WT and the 2 mutant strains were induced with IPTG at 42°C for 6H in MOPS-CAA (+dT), we found considerable β-galactosidase levels in the mutants, 80% of the WT level in ribB, 40% of the WT level in coaA mutant (Fig. 6a), demonstrating that TLD could be blocked without a complete inhibition of protein synthesis. We then tried a variety of concentrations of chloramphenicol to find the 1 that reduces the IPTG-induced β-galactosidase activity of thyA strain to less than 20% (to make it significantly lower than the protein synthesis level of the coaA mutant at 42°C). Such chloramphenicol concentration turned out to be 1 μg/mL (Supplementary Fig. 8;Fig. 6b); interestingly, this chloramphenicol concentration reduced, but failed to abolish TLD in the thyA mutant (Fig. 6c). In contrast, the coaA mutant, which by itself was more proficient for protein synthesis than 1 µg/mL chloramphenicol-treated thyA mutant, showed no TLD (Fig. 6b and c). Likewise, the ribB mutant showed little decrease in protein synthesis upon T-starvation, yet suffered less than one order of magnitude TLD (Fig. 6b and c). It appears that the isolated coaA(Ts) and ribB(Ts) defects suppress TLD by something other than severe inhibition of protein synthesis.

Fig. 6.

Protein, RNA, and DNA synthesis in the ribB and coaA mutants at 42°C. a–c), Strains were grown O/N at 28°C, subcultured in MOPS-CAA + dT +1 mM IPTG and grown at 42°C for 6–8 h before measurements. In some cases, the last 5 h at 42°C was without dT (did not affect the results, so they were pooled together). a) Protein synthesis is measured by β-galactosidase production, induced by IPTG. The background readings are from cultures treated with 40 µg/mL chloramphenicol (Cam40). b) Protein synthesis reduction upon addition of 1 μg/mL chloramphenicol during T-starvation. c) TLD magnitude in the cultures used for protein synthesis measurements in “B”. d–f) Strains were grown O/N at 28°C, subcultured in MOPS-CAA + dT and further grown at 28°C for 2–2.5 h. At this time, the thyA culture was split into 3, and either 40 μg/mL or 1 μg/mL of chloramphenicol was added to 2 subcultures. After incubation for another 30 min at 28°C, cultures were further split into three sets and mixed with either ³H-leucine (d), ³H uridine (e), or ³H thymidine (f) and transferred to 42°C. Aliquots were removed at the indicated times to measure the incorporation.

Since the high stability of β-galactosidase could have affected our protein synthesis measurements, we also measured the total protein synthesis using ³H-leucine incorporation, again in cultures growing in MOPS-CAA (+dT) at 42°C (these cultures were pre-grown without label at 28°C). We found that the total protein synthesis in the mutants was less than in the β-galactosidase assay: after 2 h incubations at 42°C, coaA and ribB mutants showed, respectively, 15 and 73% of incorporation of the thyA control (Fig. 6d). At the same time, thyA incubated with 40 μg/mL chloramphenicol (blocks TLD completely; Supplementary Fig. 9c), showed only about 0.4% incorporation, while 1 µg/mL chloramphenicol (fails to prevent TLD; Fig. 6c; and Supplementary Fig. 9c) reduced incorporation to 9% of the non-treated control (Fig. 6d). Thus, the bulk protein synthesis measurements also suggest that TLD suppression in ribB and coaA is due to a targeted, rather than general, inhibition of protein synthesis.

To make sure the results of 2 protein synthesis measurements agree within a single experiment, we did both measurements on the same cell cultures after 1, 2, or 5 h of T-starvation, also determining the level of TLD in these cultures after 5 h (Supplementary Fig. 9). Both the incorporation (Supplementary Fig. 9a) and beta-galactosidase (Supplementary Fig. 9b) measurements were consistent with each other; they were higher in this set than the previous measurements (Fig. 6a and b) because the (+dT) growth phase was again at 28°C in this assay. Because of this 28°C pre-growth, the TLD reduction in the ribB and coaA mutants was an order of magnitude lower (compared Supplementary Fig. 9c to Figs. 6c and 4g)—but still much better than after 1 µg/mL chloramphenicol treatment, which reduced protein synthesis more (Supplementary Fig. 9Ab).

To explore how the ribB and coaA mutations affected other cellular functions, we also measured their replication and transcription at 42°C (after pre-growth at 28°C), again using as controls thyA cultures pretreated for 30 min with either 1 or 40 μg/mL chloramphenicol. As expected, transcription was affected minimally in both mutants (Fig. 6e), suggesting no defects in the nucleotide pools; curiously, transcription even seemed stimulated by 1 µg/mL chloramphenicol. In the chloramphenicol-treated cultures (independently of the concentration), there was significant DNA synthesis observed, 17% of the untreated level (Fig. 6f)—likely reflecting the fact that protein synthesis is not required for continuation of the already initiated chromosomal replication in the exponentially growing cultures. Both mutants also retained significant DNA synthesis levels, 68% of the WT for ribB, 25% of the WT for coaA (Fig. 6f), concordant with protein synthesis levels in these mutants (Fig. 6d). Since starvation of 28°C pre-grown cultures at 42°C also causes significant reduction in TLD in the 2 mutants (Fig. 4g; Supplementary Fig. 9c), we conclude that ribB and coaA Ts defects prevent TLD by affecting specific cellular processes, rather than via simple reduction in protein synthesis.

Alleviation in the awakening protocol

All suppressors discussed until now were isolated by enrichment in the standard protocol (Fig. 1b). While performing enrichment under the awakening protocol (Fig. 1b), we isolated 1 insert in gadE, 1 insert in ybdM, and 4 different insertions in ghxP, from 4 different libraries (Supplementary Fig. 10a). Remarkably, these same libraries failed to yield any gadE, ybdM, or ghxP insertions (besides the possibly related soxS insert in the adjacent gene cluster (Supplementary Fig. 10a)) in multiple runs under the standard protocol, as if the 2 T-starvation protocols create distinct metabolic conditions. Perhaps a simpler explanation was that the awakening protocol had no plating after individual enrichment cycles, using liquid-to-liquid transfers only, until the final plating (see Materials and methods).

GhxP is a high-affinity guanine/hypoxanthine transporter (Papakostas et al. 2013), and its potential involvement into the DNA precursor metabolism could lead to a mechanistic TLD effect. However, after 3 independent repetitions, the ghxP::pRL27 inserts showed only a statistically insignificant effect on TLD under the standard T-starvation conditions (Supplementary Fig. 10b). Perplexingly, when assayed under the awakening conditions, ghxP inserts showed even lesser improvement in viability, although a consistent one among all four inserts (Supplementary Fig. 10c). Since our standard quantitative assays were at 42°C (because TLD is usually stronger at this temperature; Khan and Kuzminov 2019), while the ghxP mutants were isolated at 37°C, we also repeated the two TLD assays at 37°C—and again with no TLD effect of the ghxP mutant (Supplementary Fig. 10d and e). Nevertheless, isolation of 4 independent inserts in a particular gene makes this TLD effect credible and attests to the cumulative power of multiple enrichment cycles to isolate mutants with even the weakest effects. It should be noted that the gadE and ybdM inserts also failed to show a convincing TLD effect in the standard assay (Supplementary Fig. 5Ac).

Since our isolation of the gadE, ybdM, and ghxP suppressors only under the awakening protocol suggested that the metabolic state under these conditions was distinct from the 1 under the standard conditions, there was a possibility that suppressors isolated under the standard T-starvation conditions would not alleviate TLD under the awakening conditions. Indeed, the ribB and coaA defects reduced, but did not block, TLD under the awakening conditions (Fig. 7a). The reduction in their TLD effects raises questions about the role of active metabolism right before the start of T-starvation. Alternatively, the presence of replication forks could also play a role in TLD rescue by ribB and coaA Ts defects.

Fig. 7.

TLD suppression under the awakening protocol. The isolated TLD-alleviating pRL27 insertion mutants were grown at 28°C in MOPS-CAA (+dT) for 48 h, following which they were harvested, washed, and subcultured in MOPS-CAA without thymidine. Such “awakened” cultures were then shaken at 42°C, and their viability was determined at various times. a) The ribB and coaA mutants. b) The ptsN, ptsG, and pgk mutants. c) The 3 nadA mutants. d) The recF, recJ, and recR mutants.

Retesting under the awakening conditions of the mutants showing significant reduction in TLD under the standard conditions (Fig. 1d) revealed that all these mutants, namely, the 3 nadA inserts, ptsN, ptsG, pgk, recJ, recF, and recR, showed TLD reduction under the awakening conditions similar to the levels under the standard conditions (with the notable exception of nadA3; Fig. 7b–d). Therefore, the absence of these mutants after enrichments under the awakening conditions remains a mystery.

Discussion

The objective of this study, prompted by the current exhaustion of testable models of TLD, was to find new genetic factors that contribute to the cell death during T-starvation. Instead, we discovered a major category of strong TLD suppressors that was completely overlooked by the previous TLD models (Fig. 8). Our enrichment of insertion libraries led to isolation of TLD-alleviating mutants affecting such broad categories as metabolism (32%), transport (16%), DNA damage repair (16%), cell envelope (12%), and transcriptional regulation (8%). The previous enrichment by Nakayama and colleagues, who used chemical mutagenesis, isolated a single TLD-alleviating mutant in recQ (Nakayama et al. 1984). For us, insertional mutagenesis proved to be a more productive strategy, with P1 transduction facilitating linkage of the TLD phenotype with a specific insertion. In terms of numbers, disruptions of genes affecting metabolism dominated our catch (33 out of 63 sequenced candidates), suggesting that, while the chromosome is the ultimate TLD target, T-starvation somehow works via the central metabolism. This surprising finding aligns well with our recent observations that TLD does not depend on DNA replication (Khan and Kuzminov 2019), chromosome fragmentation (Khan and Kuzminov 2022), futile misincorporation–excision cycles or futile replication fork breakage–repair cycles (Rao and Kuzminov 2019), but is completely blocked by metabolism-halting acute oxidative damage (Rao and Kuzminov 2022). It is also dovetails with the recent report of intracellular acidification during T-starvation that the authors proposed to be the cause of TLD (Ketcham et al. 2022).

Fig. 8.

Musing on the nature of TLD: insights from TLD suppression. In the second column, bacterial cell is schematically depicted as housing 2 balls: the blue ball of DNA synthesis and the red ball of total protein synthesis, their equal sizes indicating the DNA/protein synthesis balance. TLD2 and TLD3 signify the depth of TLD in orders of magnitude. Cam40 and Cam1 are chloramphenicol concentrations in micrograms per milliliters.

In that report, Ketcham and colleagues have isolated various defects in the central metabolism with either negative or positive effects during T-starvation. Interestingly, both types of mutants were revealed using deep sequencing of an insertion library to identify over-represented and underrepresented inserts after only 3 h of T-starvation, followed by a limited amplification. Thus, instead of changes in survival, this approach was rather tracking changes in replication rates under T-starvation. But Ketcham and colleagues then confirmed with the classic T-starvation assay that some of the positive-effect metabolic mutants indeed alleviated TLD—these were known slow-growers, like ubi (Aussel et al. 2014), ackA-pta (Wolfe 2005), cydAB (VanOrsdel et al. 2013), and atp mutants (Jensen and Michelsen 1992)—again demonstrating that slower growth rates are associated with shallower TLD. Interestingly, Ketcham and colleagues also attempted to evolve extreme TLD resistance—and indeed isolated such T-starvation–resistant clones after 50 enrichment liquid media cycles. These evolved strains had multiple (15–70) genome changes and eventually were succumbing to TLD—but only after 25–70 h, rather than in 3 h, like their progenitor (Ketcham et al. 2022). The mutations in these evolved strains did include the known rec TLD-alleviating defects, but, importantly, they would also have several metabolic defects. Thus, Ketcham and colleagues demonstrated that (i) enrichment for TLD-alleviated mutants is productive and (ii) there is the metabolic dimension in TLD waiting to be explored.

Although we have isolated the strongest suppressors multiple times, we did not isolate several known TLD-alleviating defects, such as recA, recO, recQ, or sulA (Ahmad 1980; Fonville et al. 2010; Kuong and Kuzminov 2010), suggesting that our enrichment is far from being saturated. Also, more systematic enrichments under the awakening conditions are required to test whether the two T-starvation conditions (Fig. 1b) indeed yield different sets of candidates. The inherent limitation of insertional mutagenesis is that the bulk of the mutants are gene inactivations, and therefore, essential genes are generally excluded. In spite of this limitation, we did isolate inserts in at least 3 essential genes: pgk, ribB, and coaA (and a conditionally essential nadA). Insertions in the very 3′ ends of these genes (nadA3 is Ts) argue that the very C-termini of the corresponding proteins have regulatory roles. Inactivation of the regulatory parts of the NadA, RibB, and CoaA enzymes is tolerated at 28°C, but not at 42°C, stalling the mutant growth. Remarkably, suppression of TLD by the ribB and coaA mutants at 42°C is almost complete—which was never observed before with DNA repair mutants, like recA, recF, recO, RecR, recJ, recQ, or the SOS response suppressor sulA (Fonville et al. 2010; Kuong and Kuzminov 2010). Another complete suppressor, the nadA1 mutant, has a trivial explanation of poor growth on MOPS-CAA medium—reflecting the fact that ΔnadA mutants fail to grow on minimal media (Joyce et al. 2006).

Оur enthusiasm about the discovery of NadA, CoaA, and RibB as apparent TLD-promoting factors is tempered by the fact that the corresponding defects in production of important enzyme cofactors are expected to affect many enzymatic reactions, making the mutants pleiotropic. Multiple important enzymes depend on NAD, FMN/FAD, or coenzyme-A for their function (Silverman 2002; Krivoruchko et al. 2015), and so far we have no idea which of these functions contribute to TLD. While we did not detect complete cessation in protein synthesis in our coaA and ribB inserts, the fact that they fail, or struggle to form, colonies at 42°C (Fig. 4d) is a clear indicator of their severely affected metabolism. Our results also confirm the irrelevance of DNA replication during TLD, as our ribB and coaA defects did not eliminate DNA replication either.

To sum up, results presented in this study reinforce the idea that the maximal TLD depends on robust metabolism. Until recently, the only role of robust metabolism in TLD was thought to maximize DNA replication—thus maximizing the resulting chromosome damage from T-starvation (Ahmad et al. 1998; Khodursky et al. 2015). Now that DNA replication is out of the TLD picture, especially because of our results with the awakening protocol (Khan and Kuzminov 2019; Rao and Kuzminov 2020) (Fig. 8), some other chromosome readouts have to be identified that both respond to T-starvation and reflect the resulting TLD. The isolated TLD suppressors will be helpful in identifying such relevant readouts.

Data availability

Strains and plasmids are available upon request. The authors affirm that all data necessary for confirming the conclusions of the article are present within the article, figures, and tables.

Supplemental material available at GENETICS online.

Funding

This work was supported by grant # GM 073115 from the National Institutes of Health.