Electron Microscopy Reveals Unexpected Cytoplasm and Envelope Changes during Thymineless Death in Escherichia coli

ABSTRACT

Bacterial rod-shaped cells experiencing irreparable chromosome damage should filament without other morphological changes. Thymineless death (TLD) strikes thymidine auxotrophs denied external thymine/thymidine (T) supplementation. Such T-starved cells cannot produce the DNA precursor dTTP and therefore stop DNA replication. Stalled replication forks in T-starved cells were always assumed to experience mysterious chromosome lesions, but TLD was recently found to happen even without origin-dependent DNA replication, with the chromosome still remaining the main TLD target. T starvation also induces morphological changes, as if thymidine prevents cell envelope or cytoplasm problems that otherwise translate into chromosome damage. Here, we used transmission electron microscopy (TEM) to examine cytoplasm and envelope changes in T-starved Escherichia coli cells, using treatment with a DNA gyrase inhibitor as a control for “pure” chromosome death. Besides the expected cell filamentation in response to both treatments, we see the following morphological changes specific for T starvation and which might lead to chromosome damage: (i) significant cell widening, (ii) nucleoid diffusion, (iii) cell pole damage, and (iv) formation of numerous cytoplasmic bubbles. We conclude that T starvation does impact both the cytoplasm and the cell envelope in ways that could potentially affect the chromosome.

IMPORTANCE Thymineless death is a dramatic and medically important phenomenon, the mechanisms of which remain a mystery. Unlike most other auxotrophs in the absence of the required supplement, thymidine-requiring E. coli mutants not only go static in the absence of thymidine, but rapidly die of chromosomal damage of unclear nature. Since this chromosomal damage is independent of replication, we examined fine morphological changes in cells undergoing thymineless death in order to identify what could potentially affect the chromosome. Here, we report several cytoplasm and cell envelope changes that develop in thymidine-starved cells but not in gyrase inhibitor-treated cells (negative control) that could be linked to subsequent irreparable chromosome damage. This is the first electron microscopy study of cells undergoing “genetic death” due to irreparable chromosome lesions.

INTRODUCTION

Thymineless death (TLD) is a universal phenomenon in which cells deficient in conversion of deoxyuridine into thymidine—and therefore unable to synthesize the DNA precursor dTTP without exogenous thymine or thymidine (collectively abbreviated “T”)—die upon growth without T supplementation (1, 2). Why the resulting T starvation is so toxic is still unclear, even after more than 65 years of studies in various organisms, especially in the best-known model system, Escherichia coli, in which T starvation is induced by switching thyA mutants from T-supplemented growth medium to the same medium without T (3). Upon T removal, E. coli thyA mutants first enter the resistance phase, during which they maintain viability for an equivalent of at least one generation, continuing DNA synthesis at a lower rate (4, 5) until they replicate ∼1/4 of the chromosome (bringing the total DNA to 1.5× the initial amount) (6) (Fig. 1A). At this point, the starved cultures suddenly switch to the rapid exponential death (RED) phase (6), which is characterized by a viability drop of ∼3 orders of magnitude and loss of a chromosomal DNA amount equivalent to the amount accumulated during the resistance phase (Fig. 1A) (6). Finally, after ∼6 h of T starvation under standard conditions, the cultures emerge from the RED phase to enter the survival phase (4, 6), during which the few remaining viable cells are likely supported by thymine/thymidine from the surrounding decomposing dead cells (Fig. 1A).

FIG 1.

Major phenotypes of thymineless death. (A) The three distinct phases of TLD. Green shading indicates the resistance phase; red shading, the RED phase; and purple shading, the survival phase. Data are adapted from the work of Rao and Kuzminov (6). All of the data points are means (n ≥ 3) ± standard error of the mean (SEM). (Top) Changes in the thyA mutant cell culture titer during the first ∼9 h of T starvation. The titer of growing (Thy⁺) culture is shown for comparison. Also shown are the titer changes in the wild-type (WT) culture treated with a DNA gyrase poison, 20 μg/ml nalidixic acid (WT + NAL). (Bottom) Changes in the chromosomal DNA amount in the thyA mutant cells during T starvation. (B) The chromosome profile of marker frequencies during T starvation of a thyA mutant. x axes, chromosome coordinates (bp); y axes, normalized number of sequencing reads. Profiles, taken at time zero (just before dT removal), as well as at 5 min, 1 h, or 3 h after dT removal, are shown in pairs of two consecutive times to emphasize changes from the previous time point. Changes in the chromosome profiles past 3 h of T starvation are minimal.

Several phenotypes of the T-starved E. coli cells implicate serious damage to the chromosomal DNA in the TLD mechanisms. Cell filamentation (7–9) and induction of the SOS response (4, 10) both indicate chromosome damage. Both physical and genetic assays during T starvation indicate that the chromosome is fragmented by double-strand DNA breaks (4, 6), while also accumulating single-stranded DNA (ssDNA) gaps (6, 11). Finally, chromosome profiling in T-starved cells reveals a brief period of initiation right after removal of dT from the growth medium, followed by a dramatic yet curiously incomplete destruction of the origin-containing chromosome macrodomain, as well as some destruction of the terminus (Fig. 1B) (5, 12).

Since dTTP is one of the four DNA precursors, inability to synthesize it during T starvation was expected to generate all these problems in the nascent DNA behind replication forks (2), and this paradigm adequately explained TLD until recently (3). However, since then, it was revealed that active origin-initiated replication is neither essential for TLD, nor required for the accompanying loss of chromosomal DNA (8, 13), suggesting that the DNA loss may have a structural, rather than functional, nature.

Besides being the only DNA-specific nucleotide, thymidine is also employed in the biosynthesis of exopolysaccharide of bacterial surface antigens, transiently serving as the activation “handle” for several dTDP-hexoses before returning to the dTTP pool (14). Recently, we confirmed an old observation that some mutants in dTDP-hexose metabolism exacerbate TLD in E. coli (15). In particular, we found that the mutants that cannot synthesize dTDP-hexoses have a shorter resistance phase with no initial DNA synthesis, experience a deeper death during the RED phase, and suffer partial lysis after 3 h of T starvation (8). Moreover, the mutants that synthesize dTDP-hexoses normally, but trap dT in the dTDP-hexose pool, have no resistance phase, experience almost complete chromosomal DNA loss (which is still independent of the origin-initiated DNA replication), and lyse uncontrollably after 1 h of T starvation (8). At the same time, both types of these mutants grow normally when supplemented with dT. Such phenotypes of “T hyperstarvation” suggest that (i) besides the chromosomal DNA, dT has a role in the cytoplasm and/or cell envelope maintenance, and (ii) the problems with envelope and/or cytoplasm somehow translate into chromosomal problems, independently of replication (8).

Because the chromosome was considered the only target of TLD, the question of how much T starvation affects the cell structure beyond the chromosome has never been adequately addressed. Only anecdotal evidence exists regarding unexpected structural changes in T-starved cells (16). Optical microscopy imaging after 4′,6-diamidino-2-phenylindole (DAPI) staining for the nucleoid showed that, during the resistance and the RED phases, only the expected changes in the T-starved cells happened (7, 8), namely, (i) cell elongation and (ii) nucleoid compaction and central positioning. The only irregularities in the elongated cells were apparent cytoplasm voids after 3 h of T starvation (8). It should be mentioned that cells dying due to starvation for another DNA precursors, dGTP, also show unusual morphologies (17).

Again, because TLD is considered a “chromosome death” phenomenon, T-starved E. coli cells were never inspected in the electron microscope. Interestingly, the only transmission electron microscopy (TEM) picture of Salmonella enterica subsp. enterica serovar Typhimurium cells after 4 h of T starvation—done as a control for a cell lysis treatment—shows multiple cytoplasm disengagements from the envelope, although no cell lysis (18). Gross cytoplasm irregularities were obvious upon overnight T starvation in the thyA mutants (8), but it was unclear how much of these changes could be attributed to the overall structural degradation of long-dead cells.

To examine the cell morphology changes during T starvation in detail, we visualized the thyA mutants by TEM after three different cell preparation techniques. Our results are summarized below.

RESULTS AND DISCUSSION

DAPI staining of T-starved cells.

Fluorescence microscopy of DAPI-stained cells reveals the overall cell shape and dimensions, the number and the position of nucleoids within cells, and the nucleoid shape and condensation state. Cells grown in the presence of dT are small and have a regular (short) rod shape (cell length = 2 to 4× cell width) accommodating either one or two compact nucleoids (Fig. 2, 0 h). After 1 h of T starvation, the still-alive cells become both wider (∼1.5×) and longer (∼3×), and all feature a single bright (= diffuse) nucleoid in the middle of the cell. Thickening of thyA mutant cells grown in limited thymidine has been reported before (19, 20). A notable fraction of cells is trying to divide, with the (apparently blocked) division septum pinching the centrally positioned nucleoid (Fig. 2, 1 h, red arrows). After 3 h of T starvation, at which point only 1 out of 100 cells remains alive (Fig. 1A), the cells become longer (length = 10× width), but they are still mostly straight and still feature a centrally located single nucleoid (Fig. 2, 3 h). They apparently stop their attempts at cell division, but some of them instead develop nucleoid-sized cytoplasm voids, typically near the nucleoid itself (Fig. 2, 3 h, violet arrows). Finally, after 20 h of T starvation, cells become 10× to 30× longer than their width, kinky, and with segmental cytoplasm, sometimes resembling beans in a pod. About one quarter of the cells still retain a single dim nucleoid. However, T-starved cultures would be long stabilized at that time in the “survival” phase (4, 6) (Fig. 1A), so these dramatic changes likely reflect structural decomposition of long-dead cells. Since the cytoplasmic effects become dramatic after 20 h of T starvation and do look like continuation of the cytoplasmic problems already seen at 3 h, we next examined the effect of T starvation on the cytoplasm in the electron microscope.

FIG 2.

4′,6-Diamidino-2-phenylindole (DAPI) staining of a thyA mutant T starved for 0, 1, 3, or 20 h. In the 1-h panel, red arrows point to cells stuck during septation, with the unsegregated nucleoid blocking septum formation. In the 3-h panel, purple arrows point to the cytoplasm voids.

TEM of T-starved cells.

Although transmission electron microscopy (TEM) examination of TLD in E. coli was never reported, a single TEM picture of thyA mutant Salmonella after 4 h of T starvation showed local cytoplasm recession from the cell envelope, yet no cell envelope problems (18). This one picture seemed to corroborate our earlier observations of cytoplasm irregularities in the DAPI-stained thyA mutant cells after 3 h of T starvation and widespread cytoplasmic problems after overnight starvation (8) (Fig. 2). To reveal any subtle yet potentially consequential cytoplasm or envelope changes during T starvation, we used TEM to analyze thyA mutant cells undergoing TLD.

TEM of the thyA mutant grown in the presence of dT (no T starvation) showed cells of regular shape and dimensions (2 μm × 1 μm), some of them dividing, with nucleoids clearly visible as irregular lighter areas in the otherwise homogeneous cytoplasm (Fig. 3, 0 h). One hour into T starvation, the cells became elongated and thicker (Fig. 3,1 h). In contrast to the bright and compact nucleoids in the DAPI-stained T-starved cells (Fig. 2), after 1 h of T starvation, the diffuse nucleoids become barely discernible in the center of elongated and wider cells. Note that these cells are still alive at this point of T starvation (Fig. 1A).

FIG 3.

Transmission electron micrograph of the thyA mutant cells in cultures undergoing T starvation for 0, 1, 3, or 5 h. Note branching cells in the 1-h and 5-h panels (arrows).

By 3 h into T starvation, when 99% of cells are dead, their nucleoids become completely invisible by TEM, but now the cytoplasm also starts changing in most cells. A fraction of the cells stays unchanged. About half of all cells develop multiple small clearings, as if their cytoplasm becomes ready to boil (Fig. 3, 3 h). A third group of cells develop singular large cytoplasm voids occupying the whole width of the cell (Fig. 3, 3 h). Polar voids of this type may have a similar nature to the polar cytoplasm retractions in envelope-stressed mutants (21, 22). If located in the middle of the cell, these voids could be similar to the cellular voids reported in the ftsK mutants, in which cytoplasm division becomes uncoupled from envelope septation (23). Importantly, we have not observed cells featuring both small clearing and large voids, suggesting either two alternative pathways of cytoplasm instability or rapid coalescence of small clearings into one large void.

By 5 h into T starvation, when 99.9% of cells are dead (Fig. 1A), there is the first sign of cell structure deterioration, as cells with large voids start accumulating envelope-associated clearings (Fig. 3, 5 h). Cells with small multiple clearings look the same. One more change associated with T starvation is an occasional formation of what looks like cell branching (Fig. 3, arrows), which was noticed previously during T-limited growth (16). We conclude that the cytoplasm becomes unexpectedly dynamic during T starvation.

High-resolution TEM of the cytoplasmic voids.

In order to see if the small and the large cytoplasmic voids have a similar nature, we performed high-resolution TEM of cells at 4 h of T starvation, using a high-pressure freezing and freeze substitution protocol (Fig. 4). The resulting images revealed that small voids are not actually empty but have some (sometimes quite dense) content (Fig. 4), which was apparently lost during the previous fixation protocol, so we renamed them “bubbles.” We also found that the bubbles are rather uniform in size, never growing past a certain size of ∼0.2 μm. Last, we failed to observe any intermediates between bubbles and large voids; again, the cells that had one of the two features did not have the other. We conclude that (i) two cytoplasmic features form in T-starved cells, small bubbles and significant voids, and (ii) these two features seemed to represent separate formation pathways, as we did not observe them together in the same cell (Fig. 4B to D).

FIG 4.

High-resolution transmission electron micrograph of thyA mutant cells T starved for 4 h. Bars, 1 μm.

A modified cell fixation protocol.

To make sure our general observations were not biased by our cell fixation protocol, we used a significantly different protocol (see Materials and Methods, embedding no. 2), which gave similar pictures (cf. Fig. 5 and Fig. S1 in the supplemental material versus Fig. 3). In dT-supplemented (growing) cultures, cells are uniformly small, often dividing, with dispersed yet clearly visible nucleoids (Fig. 5 and Fig. S1, 0 h). One hour into T starvation, the cells become elongated and thicker, occasionally with some cell poles appearing damaged, suggesting defective cell division (Fig. 5 and Fig. S1, 1 h). In contrast to the clearly visible nucleoids in growing cells, at 1 h, T-starved cells feature a barely discernible diffuse nucleoid in the middle of the cell, which apparently interferes with cell division.

FIG 5.

A different technique of cell fixation for TEM. The pattern of cell deterioration during T-starvation stays the same.

By 3 h into T starvation, the nucleoid-free cytoplasm of the further elongated cells accumulates numerous small round clearings (Fig. 5 and Fig. S1, 3 h). By 5 h of T starvation, some cells start losing their exopolysaccharide (EPS) capsule (the outermost layers of the cell envelope) (Fig. 5 and Fig. S1, 5 h); in fact, we also observed accumulation of what looked like cell-separated EPS capsule material (Fig. S1). Yet, in the majority of cells, the cytoplasm stays within the cell envelope (or what remains of it), with occasional segmental voids developing within the envelope. Finally, after overnight T starvation, we observed ∼10× elongated cells with a lighter cytoplasm (Fig. S1, 20 h), as well as cell fragments and cell ghosts (Fig. 5, 19 h, and Fig. S1, 20 h). Thus, our main observations about the cytoplasm dynamics during T starvation were confirmed with a different cell fixation protocol. We also observed what looks like cell envelope instability with the new fixation protocol.

Does T starvation destabilize the cell envelope?

We took a closer look at the cell envelope problems that developed after 5 h of T starvation in the experiments shown in Fig. 5 and Fig. S1, specifically looking for examples of cell fragments with unraveling cell envelope. Surprisingly, in these decomposing cells, the cytoplasm still holds its overall shape, even though the cell envelope is missing or peeling off (Fig. 6, top left orange frame). In the same session in which the image in Fig. S1 was taken, we also observed several examples of the cell-free filamentous material, which looks like cell envelope structures (Fig. 6, right orange frame). At first, we failed to observe examples in which the filamentous material (Fig. 6, right orange frame) directly emanated from what looks like an unraveling cell (Fig. 5 and 6, left orange frame), which would indicate cellular origin of this filamentous material.

FIG 6.

Cell envelope problems late into T starvation. Examples of cells with obvious envelope problems from the experiment shown in Fig. 5 are in the left orange frame, while two examples of what looks like cell-free envelope material from the experiment shown in Fig. S1 are in the right orange frame. All of these were observed only after 5 h of T starvation (orange frames). In the green frame, there is a single sighting of a similar cell-free material after 1 h of the control nalidixic acid treatment (see Fig. 7).

We eventually spotted one possible instance, but at 20 h of T starvation (Fig. S1, 20 h). Therefore, we remain unsure that the filamentous material is derived by cell envelope spooling off decomposing cells. Also, the clear branching of this filamentous material (Fig. 6) is inconsistent with its being generated by simple envelope unwrapping off T-starved cells. Finally, we observed (although only once) layers of a similar material (Fig. 6, green frame) after 1 h of treatment with a DNA gyrase poison, nalidixic acid (NAL), a treatment that is not expected to lead to cell envelope separation, suggesting that the filamentous material accumulation might be an artifact of this fixation protocol.

Nalidixic acid-treated cultures provide a negative control.

Nalidixic acid (NAL) inhibits DNA gyrase in the middle of its catalytic cycle and therefore kills cells by targeting their chromosome (24). As a general control for cytoplasm behavior during “genetic death,” we used NAL-treated wild-type (WT) E. coli cultures that died with kinetics similar to our TLD, including a 1-h-long resistance phase (Fig. 1A). Untreated cells look similar; they are uniform in size, with a significant fraction of them dividing, and with the nucleoids clearly visible (Fig. 7 and Fig. S2, 0 h). After 1 h of the NAL treatment, cells elongate significantly without widening, with dispersed yet still visible nucleoids (Fig. 7 and Fig. S2, 1 h)—interestingly, both changes are unlike those after 1 h of T starvation. At 3 h of NAL treatment, cells elongate some more and show signs of attempted cell division (Fig. 7, 3 h, arrows). Also, the cytoplasm lightens a bit, while the nucleoids blend in, becoming barely discernible in the middle of the elongated cells (Fig. 7 and Fig. S2, 3 h). By this point, 99% of cells are dead (Fig. 1A). In contrast to T starvation, NAL-treated cells do not develop multiple bubbles, but they do develop a few modest midcell cytoplasm voids after 5 h of the NAL treatment (Fig. 7 and Fig. S2, 5 h). After an overnight NAL treatment, many cells show cytoplasm recession from the cell envelope in multiple places, perhaps reflecting a common structural degeneration of dead cells (Fig. 7, 19 h, and Fig. S2, 16 h).

FIG 7.

AB1157 cells undergoing death due to gyrase poisoning. Similarities and differences from TLD. The arrows in the 3-h panel mark sites of attempted cell divisions.

Conclusion.

To our knowledge, this is the first electron microscopy study of bacterial cells dying of ostensibly “pure” chromosomal damage that is supposed to cause no morphological changes besides cell filamentation. The study was motivated by our recent findings of oriC-initiated replication independence of TLD, of structural changes in T-starved cells, of their SDS sensitivity, and of lysis of cells undergoing T hyperstarvation (8, 13), all of which suggest a possibility that T starvation also causes inconspicuous yet critical changes in the cell envelope or cytoplasm, with these changes then causing the chromosome damage. Using DAPI staining and TEM, we now report that T starvation does have an impact on the cell structure beyond the chromosome, affecting both the cytoplasm and cell envelope.

While we are currently not aware of any cytoplasmic process that could visibly react to the disappearance of dTTP, the metabolic link between the cell envelope and dTTP levels is known and comprises the substantial pool of dTDP-hexoses, which E. coli uses to synthesize enterobacterial common antigen (25). Thus, in the T-starved cells, there is a competition between the chromosome and the dTDP-hexose pool for dwindling dTTP, which may lead to inadequate synthesis of exopolysaccharide capsule (8). However, the current study finds little evidence for the gross envelope instability during the early RED phase (3 h of T starvation), when the majority of the cells die, and only some evidence of cell structure instability at later time points (5 h of T starvation and beyond) (Fig. 6), when the T-starved cultures are already transitioning into the survival phase (Fig. 1A). Thus, our earlier speculations that dTTP has a direct (and still unknown) role in building or maintaining the bacterial envelope (8) were not clarified in the current study.

We found several changes in the cell structure during the first 5 h of T starvation (Table 1), the time period during which most of the cell death occurs (Fig. 1A). By 1 h, we observed the expected cell elongation (∼3×), accompanied by cell widening (∼1.5×); nucleoid blending with the cytoplasm, perhaps driven by its diffusion; and frequent cell pole damage and occasional cell branching (perhaps as a consequence of pole damage). By 3 h of T starvation, we observed formation of numerous cytoplasmic bubbles in about half of the cells, single large cytoplasm voids in another quarter of the cells, and conspicuous absence of cells with both features. Finally, by 5 h, we observed elongation of all cells, so that cell length becomes ∼10× cell diameter, as well as cell envelope unraveling in a few cells (but no massive cell lysis).

TABLE 1.

Changes detected in T-starved or NAL-treated cultures

Treatment duration and change	Culture condition
	T starvation	Nalidixic acid treatment
1 hour
Cell widening	1.5×	None
Cell length	∼3× cell diameter	∼10× cell diameter (nucleoid still visible)
Nucleoid	Becomes diffuse and blends in with dark cytoplasm	No change
Pole damage	Frequent	None
Cell branching	Occasional	None
3 hours
Nucleoid	Blending into cytoplasm already observed at 1 hour	Blends in due to cytoplasm lightening
Cytoplasm bubbles	Numerous in ∼50% of the cells	None
Cytoplasm voids	Single large voids in ∼30% of the cells	None
5 hours
Cytoplasm voids	Already observed at 3 hours	Occasional moderate voids
Cell envelope	Unravels in a few cells without massive lysis	None
Cell length	∼10× cell diameter	∼20× cell diameter

These changes were compared to those during the first 5 h of NAL treatment that poisons DNA gyrase, which is supposed to affect only the nucleoid and therefore should serve as a negative control (Table 1). After 1 h of NAL treatment, we observed only cell lengthening (∼10×) without cell widening. After 3 h of NAL treatment, we observed gradual nucleoid blending, perhaps driven by cytoplasm lightening. Finally, by 5 h of NAL treatment, occasional midcell cytoplasm voids were observed in dramatically elongated cells (cell length became ∼20× cell diameter).

Obviously, gross morphology changes (cell filamentation and cytoplasm voids), especially at later time points, coincide between the two treatments, but there are a few obvious differences (Table 1). At the same time, the most dramatically visual cytoplasm irregularities, similar to those reported in Salmonella TLD TEM (18), were observed when both treatments were continued overnight, so these likely reflect general cytoplasm degradation in long-dead cells and thus are the consequences, rather than the causes, of chromosome degradation.

Taking the NAL treatment as the baseline cell reaction to “pure” DNA damage, we therefore identify at least five cell morphology changes specific for T starvation, as follows: (i) significant widening of the cells compensating for a modest elongation, (ii) dramatic nucleoid diffusion, leading to its “disappearance,” (iii) frequent pole damage and occasional cell branching, as if due to an inappropriate cell division, (iv) formation of numerous cytoplasmic bubbles of unknown nature in about half of the cells, and (v) cell envelope unraveling from some cells at later points. Since cell envelope unraveling happens quite late in T starvation, it cannot contribute to the chromosomal damage, even though it does indicate a role of dT in the cell envelope maintenance. Which of the remaining four differences (cell widening, nucleoid diffusion, cytoplasmic bubbles, and/or damaged poles) may be linked to irreparable chromosome damage needs to be investigated. Here, we can only speculate about possible links of the observed morphology changes with irreparable chromosome lesions at the heart of TLD (Fig. 8); the current lack of mechanistic understanding of TLD still makes these speculations appropriate and useful.

FIG 8.

Speculations on the possible links of the observed morphology changes with irreparable chromosome lesions that cause TLD. In all panels, the cell envelope is brown, the cytoplasm is cream, and DNA or the nucleoid is blue. (A) Cell widening could break DNA, if the latter becomes permanently attached to the cell envelope in the absence of dTTP. (B) If the nucleoid “disappearance” reflects a dramatic change in nucleoid compaction and administration, then repair of chromosomal DNA may become defective, leading to accumulation of lesions. (C) Cell division may be blocked by an incompletely replicated nucleoid but may proceed anyway, guillotining the nucleoid in two and thus damaging the chromosome beyond repair. (D) The numerous small cytoplasmic bubbles may be in fact enzymatic factories producing reactive metabolic species that damage DNA in their vicinity.

As was already proposed (8), widening of the cell could generate enough force to break DNA if DNA somehow becomes permanently attached to the cell envelope in the absence of dTTP (Fig. 8A). The nucleoid diffusion and blending with the cytoplasm could reflect significant changes in the DNA compaction by the nucleoid-associated proteins, leading to a defect in repair of unspecified DNA damage (Fig. 8B). The obvious pole damage, which may even cause some cells to branch, likely reflects abnormal cell division events that could guillotine centrally stuck incompletely replicated nucleoids (producing irreparable chromosome damage) (Fig. 8C). Last, but not least, the cytoplasmic bubbles likely reflect some unusual metabolic activity, which could generate reactive metabolic species that attack DNA (Fig. 8D). It was recently proposed that T starvation kills by increased production of reactive oxygen species (26), but direct tests of this idea failed to find any evidence of oxidative damage during TLD (unpublished data).

In summary, our investigation of morphology changes during T starvation has generated several ideas, highlighting specific minor cytoplasm or envelope changes for a possibility of chromosome damage. One obvious test of these ideas would be to find how these specific morphology changes are affected in E. coli mutants in which TLD is alleviated (4, 6, 10) or exacerbated (4, 6, 8, 27).

MATERIALS AND METHODS

Strain KKW58 (AB1157 ΔthyA ΔdeoCABD) (28) was grown in M9CAA+dT medium to the early exponential phase and then subjected to T starvation as previously described (6). At indicated times, 1-ml aliquots were taken and harvested; the pellet was then dispersed into 1 ml of TEM-grade fixation buffer and stored at 4°C until all samples were collected. As a negative control, AB1157 (29) was grown in LB, treated with 20 μg/ml nalidixic acid for the indicated amount of time, and processed the same way.

The samples were submitted to either the Beckman Institute or the MRL facility, both at UIUC, for further preparation and TEM imaging. The protocols used for TEM sample preparation were as follows: protocol no. 1 (Beckman) is a variation of the modified Karnovsky’s fixative protocol (30), described in detail below; protocol no. 2 (MRL) is described in chapter 8 of Giberson and Demaree (31).

Protocol for TEM fixation, embedding no. 1 (Beckman).

Fixation is 4 h on ice or in a fridge, in 2.0% paraformaldehyde and 2.5% glutaraldehyde (both EM grade) in 0.1 M Na-cacodylate buffer (pH 7.4). Buffer rinse is for 10 min on a shaker table in 0.1 M Na-cacodylate buffer (pH 7.4). Postfixation, incubation is for 90 min in the dark in a fume hood in 1.0% aqueous osmium tetroxide in 0.1 M Na-cacodylate buffer (pH 7.4), followed by buffer rinse for 10 min on a shaker table in 0.1 M Na-cacodylate buffer (pH 7.4).

En bloc staining (tertiary fixation) is overnight in a fridge in 2% aqueous uranyl acetate (freshly made). Dehydration is on a shaker table, as follows: (i) 37% ethanol, 10 min; (ii) 67% ethanol, 10 min; (iii) 95% ethanol, 10 min; and (iv) 100% ethanol, 3 × 10 min. Infiltration is on a shaker table, as follows: (i) 1:1 100% ethanol-propylene oxide, 10 min; (ii) 1:2 100% ethanol-propylene oxide, 10 min; (iii) 1:1 propylene oxide-Polybed 812 mixture (without DMP-30), 10 min; (iv) 1:2 propylene oxide-Polybed 812 mixture (without DMP-30), 10 min.

Embedment was as follows: 100% Polybed 812 mixture (without DMP-30), overnight in hood, followed by 100% Polybed 812 mixture with 1.5% DMP-30, in molds, and put into oven at 60°C for 24 h.

Protocol for TEM fixation, embedding no. 2 (MRL).

The cells in Karnovsky’s fixative are additionally fixed by heating in a microwave at 40°C for 8 s on, 20 s off, 8 s on, and then chilling for 20 s on ice. This is repeated 4 times before adding the cacodylate buffer.

A major difference from protocol no. 1 is that protocol no. 2 has 2 steps of dehydration, first by increasing concentration of ethyl alcohol (EtOH) and then by using 1:1 mix of 100% ethyl alcohol and acetonitrile, followed by only acetonitrile. For infiltration, a 1:1 mix of Lx 112 epoxy resin and acetonitrile is used instead of EtOH.

Protocol no. 2 also includes brief microwave pulses of 20 to 40 s at several intermediate steps.

High-pressure freezing and freeze substitution TEM.

A general outline from Han et al. (32) was followed with some modifications. Aluminum specimen carriers (“hats,” 3-mm diameter) were coated on the lower inner surface with 1-hexadecene and a layer of bacterial cells. The samples were then frozen using an HPM 010 (Abra Fluid AG) high-pressure freezing system and kept in liquid nitrogen until the freeze substitution phase of the process was started.

Freeze substitution was performed in 2% osmium tetroxide, 0.1% uranyl acetate, 2% H₂O, and 98% acetone in an FS-8500 freeze substitution system (RMC Boeckeler Co.). Samples were kept at −90°C for 110 h before being warmed up to −20°C over a 5-h time period. They were then kept at −20°C for 16 h before being warmed to 0°C over another 5-h period of time, after which they were washed four times in prechilled 100% acetone at 0°C, with the final wash running for 1 h.

The samples were then brought to room temperature and washed two times in 100% acetone. They were then infiltrated with 1:1 Polybed 812 (Polysciences) epoxy resin-acetone for 24 h, 2:1 resin-acetone for 36 h, and 100% resin for 24 h, and then changed to fresh resin for 3 days. All infiltration steps were conducted on an orbital shaker at room temperature. Samples were then placed into embedding molds with combined resin and hardener (DMP-30) and polymerized in a 60°C oven for 2 days.

Sections (90-nm) were collected onto bare 200-mesh copper TEM grids using a Leica UCT UltraCut ultramicrotome with a Diatome diamond knife. Staining was performed with aqueous uranyl acetate (30 min) and lead citrate (4 min) before rinsing with degassed deionized water. Sections were imaged using a Phillips CM200 transmission electron microscope.

Marker frequency analysis.

Marker frequency analysis was done essentially as described previously (33). For plotting, the locally estimated scatterplot smoothing (LOESS) signals were normalized to the average value calculated over a 264-kb region between genomic locations 2,481,500 and 2,745,500.