The Replication Intermediates in Escherichia coli Are Not the Product of DNA Processing or Uracil Excision*

The current model of DNA replication in Escherichia coli postulates continuous synthesis of the leading strand, based on in vitro experiments with purified enzymes. In contrast, in vivo experiments in E. coli and its bacteriophages, in which maturation of replication intermediates was blocked, report discontinuous DNA synthesis of both the lagging and the leading strands. To address this discrepancy, we analyzed nascent DNA species from ThyA+ E. coli cells replicating their DNA in ligase-deficient conditions to block maturation of replication intermediates. We report here that the bulk of the newly synthesized DNA isolated from ligase-deficient cells have a length between 0.3 and 3 kb, with a minor fraction being longer that 11 kb but shorter than the chromosome. The low molecular weight of the replication intermediates is unchanged by blocking linear DNA processing with a recBCD mutation or by blocking uracil excision with an ung mutation. These results are consistent with the previously proposed discontinuous replication of the leading strand in E. coli.

The current model of DNA replication in Escherichia coli postulates continuous synthesis of the leading strand, based on in vitro experiments with purified enzymes. In contrast, in vivo experiments in E. coli and its bacteriophages, in which maturation of replication intermediates was blocked, report discontinuous DNA synthesis of both the lagging and the leading strands. To address this discrepancy, we analyzed nascent DNA species from ThyA ؉ E. coli cells replicating their DNA in ligase-deficient conditions to block maturation of replication intermediates. We report here that the bulk of the newly synthesized DNA isolated from ligase-deficient cells have a length between 0.3 and 3 kb, with a minor fraction being longer that 11 kb but shorter than the chromosome. The low molecular weight of the replication intermediates is unchanged by blocking linear DNA processing with a recBCD mutation or by blocking uracil excision with an ung mutation. These results are consistent with the previously proposed discontinuous replication of the leading strand in E. coli.
Before the polarity of DNA strand extension was established, both DNA strands at the replication fork were thought to be synthesized continuously, being extended by two distinct DNA polymerases with opposite polarities (1, 2) (Fig. 1A). The finding that DNA polymerases extend DNA strands only in the 5Ј to 3Ј direction (reviewed in Ref. 3) made it clear that at least the DNA strand synthesized in the direction opposite to the one of the replication fork movement has to be replicated in pieces and then assembled (maturated) into a full-length molecule (4 -6). Since this strand could not be synthesized continuously, its synthesis would lag behind the apparently continuous synthesis of the opposite DNA strand, which would thus lead the replication fork progress. The two strands were eventually called the "lagging" and the "leading" strands, respectively, and the hypothetical difference in their synthesis formed the basis of the semidiscontinuous model of DNA replication (Fig. 1B) (7). However, the semidiscontinuous paradigm is based on in vitro results and ignores a large body of in vivo data, which is reviewed below.
Okazaki and colleagues (1,8) used pulses of [ 3 H]thymidine to characterize the newly synthesized DNA of wild type Escherichia coli and Bacillus subtilis strains as well as E. coli bacteriophage T4. The cells were grown at 20°C to slow down the maturation of replication intermediates into the full-length DNA molecules, and short (10 -60-s) pulses of label were used (1, 8 -11). Using alkaline (denaturing) sucrose gradients to separate replication intermediates from their template strands, Okazaki and colleagues observed that the newly synthesized DNA after 2-10 s of labeling migrated mostly as low molecular weight (LMW) 2 species, with a mean length between 1 and 2 kb (1,8). A chase with nonradioactive thymidine shifted the label into high molecular weight (HMW) DNA, proving that the LMW species are true replication intermediates (5,8). These experimental results were confirmed by others (6,12), and the LMW replication intermediates became known as "Okazaki fragments." The main result of these early studies was that all new DNA, on both the lagging and the leading strands, appeared first in small pieces (1,5,8). This interpretation was confirmed in polA mutants, deficient in DNA polymerase I (13)(14)(15) and with DNA ligase mutants (11,16,17). Not only was all DNA synthesized in LMW pieces in polA and ligA mutants, but it then stayed short for some time, indicating that both DNA polymerase I and DNA ligase are the maturation functions. Later it was demonstrated that in polA mutants (18), and especially in rnhA polA double mutants (19), some Okazaki fragments retain short RNA primers at their 5Ј ends, which were predicted to provide the original 3Ј-OH for DNA polymerization (20). At the same time, Yudelevich and colleagues showed that 1) in nondenaturing conditions, the bulk of the low molecular weight replication intermediates stay with the high molecular weight chromosomal DNA; 2) treatment with DNA polymerase I and DNA ligase in vitro turns the short pieces into high molecular weight DNA (6). With this evidence, Okazaki and colleagues proposed a discontinuous model of DNA replication (Fig. 1C) in which both the lagging and the leading strands are synthesized in short fragments and maturated later into HMW species by the combined action of DNA polymerase I (plus RNase H) and DNA ligase (15,19).
According to the current model of semidiscontinuous replication in E. coli, the mobile primosome DnaB-DnaG unwinds the template DNA duplex and lays RNA primers on the tem-plate for the lagging strand, whereas the DNA polymerase III utilizes these primers to synthesize nascent DNA strands (21)(22)(23). On the other hand, the leading strand is thought to be synthesized continuously, primed only once at the replication origin (23). When DNA polymerase III reaches the RNA primer of a downstream Okazaki fragment, it cannot remove the RNA portion because it does not have a 5Ј-3Ј exonuclease or RNase H activities. RNase H helps to remove RNA primers in vivo (24,25), but the bulk of RNA primers in E. coli are removed by DNA polymerase I, which elongates Okazaki fragments in the process, creating DNA-DNA nicks, suitable for ligation (22). DNA ligase then forms a phosphodiester bond between adjacent DNA fragments, thus completing the maturation of Okazaki fragments (22). Therefore, one requirement for a reliable detection of replication intermediates in E. coli is inactivation of either DNA ligase or DNA polymerase I, which is achieved by shifting temperature-sensitive mutants in the corresponding genes to the nonpermissive temperature of 42°C.
The second and equally important requirement for detection of replication intermediates is an instant termination of the DNA metabolism, to prevent residual maturation of the replication intermediates during processing of the cultures before lysis. In fact, KCN/ice, the most popular early method of metabolism termination (4,6,8,26) allows maturation of the replication intermediates to continue for several seconds after the arrest of strand elongation (the latter one being monitored by incorporation of thymidine label in DNA). This imprecise stop in early experiments resulted in accumulation of Ն50% of all label in the intermediate molecular weight (IMW) and HMW fractions even after very short 10-s pulses, which was at first ignored (1,6,12). Eventually, Iyer and Lark investigated these IMW and HMW replication intermediates and found them to be composed of the newly synthesized DNA, attached to the 3Ј-ends of the previously synthesized material, suggesting a continuous synthesis (4). Such an asymmetry in the replication of DNA supported the semidiscontinuous model of DNA replication ( Fig. 1B) and was also consistent with a hybrid model, according to which the leading strand is synthesized in much longer pieces than is the lagging strand (Fig. 1D). However, the realization that these IMW and HMW replication intermediates could be the result of maturation of some Okazaki fragments into longer species led to the search for better termination procedures. Two new stopping techniques, KCN ϩ pyridine/ice (27) and phenol-ethanol-acetate mix (28,29), were soon described. They minimized formation of IMW and HMW replication intermediates at the earliest time points, revealing their origin as a result of maturation of existing LMW intermediates.
A radical solution to the maturation problem is offered by in vitro systems. Olivera, Bonhoeffer, and co-workers (30,31) studied replication intermediates in whole cell lysates on cellophane disks, providing dNTPs and inactivating DNA ligase with a by-product of the ligation reaction, nicotinamide mononucleotide. Although they did not show the evidence of a complete inactivation of DNA ligase, Olivera and Bonhoeffer reported a bimodal distribution of the sizes of replication intermediates, with about 50% of the replication intermediates sedimenting as LMW fragments, the other 50% of the pulse-labeled DNA fragments being of intermediate molecular weight (30,31). Interestingly, 5-fold dilution of provided dNTPs reduced the size of IMW species almost to the size of LMW species, suggesting that 1) both are formed by discontinuous synthesis, and 2) the size of replication intermediates is determined by competition between the rates of propagation versus the rates of priming of the new chains (31).
On the genetic front, two types of E. coli mutants were found with changes in the size of Okazaki fragments: the dnaG(Ts) mutants synthesized longer Okazaki fragments at semipermissive temperatures (32), whereas sof mutants synthesized shorter Okazaki fragments (33). Eventually, the DnaG protein was shown to be the hypothesized primase that lays short RNAs to initiate Okazaki fragments (34), whereas sof mutants were shown to be the same as dut mutants, deficient in hydrolysis of a noncanonical DNA precursor dUTP (35). The dut mutants feature extremely short Okazaki fragments due to uracil incorporation into nascent DNA with its subsequent rapid excision by uracil DNA glycosylase (the product of the ung gene) (36). On the basis of these results, Tye et al. (35) proposed that a fraction of Okazaki fragments, even in dut ϩ cells, forms as a result of uracil excision. However, the subsequent testing of this idea showed that uracil excision in vitro does not change the sedimentation pattern of replication intermediates in wild type E. coli, indicating very little uracil incorporation in dut ϩ cells (37). Eventually, Tye et al. (37) ruled out a significant role for uracil incorporation and excision in the generation of Okazaki fragments in E. coli by demonstrating that their size in both wild type and polA mutant cells does not change if the uracil excision is inactivated in vivo by an ung mutation.
Although the original idea of Okazaki, that both strands at the replication fork are synthesized in pieces (Fig. 1C), thus survived the experimental challenge of uracil incorporation and excision, it could not overcome the practical appeal of the simpler alternative, in which the leading strand is synthesized continuously (Fig. 1B). Using whole cell lysates on cellophane disks, Olivera demonstrated that he can regulate the size of "Okazaki fragments" in vitro by varying the concentration of dUTP in the feeding buffer (38). This "regulation" was suppressed by the inactivation of ung (39). These in vitro results allowed Olivera to speculate that 1) the effective concentration of dUTP in dut ϩ cells is higher than assumed, and 2) the leading strand is synthesized continuously in vivo but is quickly fragmented due to uracil excision, creating an appearance of a discontinuous synthesis (38). However, these ideas were never tested in vivo and were in disagreement with the experimental results of others (37).
The proposal of Olivera might have been without significant consequences, were it not for the fact that all later in vitro experiments on reconstitution of the replication fork of E. coli and its bacteriophages with highly purified enzymes yielded strong evidence for continuous synthesis of the leading strand for as long as at least 40 kb in the complete absence of DNA ligase (40 -42). In all of these systems, the lagging strands were synthesized in a coupled reaction on the same parental DNAs that provided the template for the leading strand synthesis, as short pieces of the size range characteristic of Okazaki fragments (the omission of DNA ligase in the in vitro reactions preserved Okazaki fragments) (40 -42). Thus, the in vitro reconstitution experiments rendered strong support for the semidiscontinuous model (Fig. 1B), turning it into a paradigm and a textbook fixture.
However, the discrepancy between the in vivo demonstration of fully discontinuous replication versus the in vitro demonstration of semidiscontinuous replication remains unresolved (43,44). As a first step to address this discrepancy, we have isolated replication intermediates from one of the strongest E. coli ligA(Ts) mutants, pulse-labeled at the nonpermissive temperature. To detect Okazaki fragments, we tried three labeling/separation approaches: 1) gel electrophoresis of nonlabeled DNA followed by blot hybridization; 2) [ 3 H]thymidine in vivo labeling with subsequent agarose gel electrophoresis; and 3) [ 3 H]thymidine in vivo labeling with subsequent sucrose gradient sedimentation. We found that 1) all of the newly synthesized DNA in ligase-deficient conditions has a low molecular weight; 2) this LMW nascent DNA can be chased into chromosomal size DNA if the mutant cells are returned to permissive temperatures; and 3) these short DNA species are not the results of uracil excision or the products of DNA processing by the RecBCD helicase/nuclease.

EXPERIMENTAL PROCEDURES
Bacterial Strains-All E. coli strains are K12 derivatives. The wild type strain GR523 and its ligA251 derivative GR501 have been described (45) and were obtained from Mary Berlyn at the E. coli Genetic Stock Center. LA8 is GR501 ⌬ung::cat, whereas LA9 is GR523 ⌬ung::cat. AK147 is GR523 ⌬recBCD3::kan, whereas AK148 is GR501 ⌬recBCD3::kan. AK149 is GR523 dut-1 zic4901::Tn10, whereas AK150 is GR501 dut-1 zic4901::Tn10. LA20 is GR501 ypeB::kan, made by the Datsenko-Wanner technique (46). LA35 is a ligAϩ derivative of AK150 ⌬ung::cat, essentially AK149 ⌬ung::cat. LA33 is LA35 ligA251 ypeB::kan. All derivatives of the original GR501 and GR523 strains were constructed by P1 transduction of mutations from other laboratory strains by their association with antibiotic resistance determinants. The ligA251 defect was confirmed by the inability to plate above 37°C and by detection of LMW replication intermediates at 42°C (45); the recBCD defect was confirmed by sensitivity to T4 2 mutant phage (47); the dut defect was confirmed by the inability to plate in the presence of 10 mM uracil (48); the ung defect was confirmed by the ability to serve as a host for uracil-containing phage isolated from dut ung mutant cells (49), as described (50); the double dut ung defect was confirmed by detecting the high density of uracils in DNA from these cells (51), as described (52).
Media and Growth Conditions, Pulse-labeling, and DNA Isolation-Cells were grown at 30°C to an A 600 of 0.4 followed by incubation at the nonpermissive or permissive temperatures. For [ 3 H]thymidine labeling, cells were grown in 2-10 ml of M9 minimal medium (53) supplemented with 0.2% casamino acids (M9CAA) and pulse-labeled with 5-10 Ci/ml [ 3 H]thymidine for 60 s. In some experiments, the pulse was followed by a chase for the specified amount of time in the presence of 50 g/ml cold thymidine. For nonlabeled DNA isolation, the cells were grown similarly, but no label was added. To stop the cellular metabolism instantly, we mixed the culture with an equal volume of the phenol-ethanol-acetate mix (29, 54) (a mixture of 75 ml of ethanol, 21 ml of 100 mM sodium acetate (pH 5.3), 2 ml of 100 mM EDTA, and 2 ml of phenol). The cells were later collected by centrifugation, and DNA was extracted in several ways, depending on the subsequent analytical procedure. For gel electrophoresis, native DNA was purified by SDS lysis with subsequent phenol-chloroform extraction (52). For alkaline sucrose gradients, denatured chromosomal DNA was isolated by suspending the cells in 0.5 ml of 1% sodium N-lauroyl sarcosinate containing 40 mM EDTA, followed by lysis with 0.5 ml of 0.4 M NaOH (54). After a 15-min centrifugation at 16,000 ϫ g, the supernatant was transferred to a fresh tube, and 300 l of it were dialyzed (as detailed below) and loaded on an alkaline sucrose gradient.
Kinetics of DNA Synthesis or Degradation-Cells were grown in M9CAA at 30°C to A 600 of 0.4. For DNA synthesis, before adding 1 Ci/ml of [ 3 H]thymidine, the medium was supplemented with 0.8 -1.0 mM deoxyadenosine (54,55). At the desired time points, 250-l aliquots of cultures were mixed with 5 ml of 10% trichloroacetic acid on ice. After the third time point at 30°C, half of the culture was switched to the nonpermissive temperature (42°C), and several more time points were similarly taken. The precipitate was collected on glass fiber filters (Fisher G6), washed twice with 5 ml of 5% trichloroacetic acid and twice with 5 ml of ethanol. The filters were air-dried, and the radioactivity was counted in the LS 6500 Beckman scintillation system. To monitor stability of the newly synthesized DNA, cells were grown in 2.5 ml of M9CAA at 30°C to an A 600 of 0.4 and switched to a 42°C for 5 min before labeling with [ 3 H]thymidine for 60 s at 42°C. The labeling was ended by adding 250 l of 5 mg/ml nonradioactive thymidine. Time points were taken by mixing 250 l of the culture with 5 ml of 5% ice-cold trichloroacetic acid. The precipitates were collected by filtration and processed as above.
Alkaline Gel Electrophoresis-1% agarose was prepared in water, cooled to 60°C, and, before pouring, supplemented with alkaline buffer by adding NaOH to 50 mM and EDTA (pH 8.0) to 1 mM. DNA was ethanol-precipitated, dissolved in 10 l of a 50 mM NaOH, 1 mM EDTA solution, and mixed with 10 l of 1ϫ alkaline loading dye (56). Gels were run for 710 min in a 1.5 V/cm electric field, and DNA was vacuum-transferred and UVcross-linked to a positively charged nylon membrane (Hybond-Nϩ; Amersham Biosciences). Before transfer, gels were soaked in 0.2 M HCl for 40 min, then in 0.5 M NaOH for 45 min, and, finally, in 1 M Tris-HCl, pH 8.0, for 40 min. When [ 3 H]thymidine-labeled DNA was loaded, the lanes of the membrane were cut into 10 1-cm pieces, and the radioactivity in each piece was scintillation-counted. For blot hybridization of nonradioactive DNA, the membranes were prehybridized in 5% SDS, 0.5 M sodium phosphate, pH 7.4, 1 mM EDTA at 65°C for 1 h, and hybridization was performed overnight in the same buffer with a 32 P-random hexamer-labeled DNA probe.
Alkaline Sucrose Gradients-For sucrose gradients, 200 -300 l of chromosomal DNA, extracted by the denaturing protocol above, were dialyzed in Slide-A-lyzer 3500 molecular weight cut-off minidialysis cups (Pierce) overnight against 1 liter of 0.5% sodium N-lauroyl sarcosinate, 20 mM EDTA, 0.2 M NaOH to remove unincorporated label. 5-20% linear sucrose gradients containing 0.1 M NaOH, 0.9 M NaCl, and 1 mM EDTA were assembled with a gradient maker in polyallomer centrifuge tubes. A 1-ml shelf of 70% sucrose was placed at the bottom of the tube to prevent pelleting of the fast sedimenting material. Centrifugation was at 22,500 rpm and 4°C for 21.5-30 h in a SW28 rotor, Beckman L5-75B ultracentrifuge. 500-l frac-tions were collected from the bottom of the tube, and 50 l were scintillation-counted after the pH was brought to neutral with 7.0 l of glacial acetic acid.

RESULTS
The Ligase Mutant Phenotype-To test whether the leading DNA strand is synthesized continuously or in pieces, we sought to isolate Okazaki fragments from a ligA251(Ts) mutant grown at 30°C and shifted to the nonpermissive temperature of 42°C (45). It is reported that the rate of DNA synthesis in ligA251 cells reaches a low residual level after 20 min of incubation at 42°C, perhaps being partly affected by the DNA degradation, which becomes detectable after 30 min at this temperature (45). This DNA degradation, if it initiates at replication forks and generates long single-stranded DNA fragments, has a potential to create artifactual low molecular weight "replication intermediates." Therefore, for the optimal detection of true replication intermediates with the ligA251 mutant, we first defined the time window where the DNA ligase activity of the mutant was already inhibited, yet DNA synthesis would still continue with little or no DNA degradation. We found that, when a ligA251 culture, growing at 30°C, is switched to 42°C, the incorporation of label into DNA generally stops after 10 min, whereas it continues normally in the culture that remains at the permissive temperature ( Fig. 2A). On the other hand, we detect chromosomal DNA degradation in ligA251 mutant cells after 30 min at 42°C (Fig. 2B). Therefore, in order to study DNA replication in ligase-deficient conditions without the interference from DNA degradation, in our experiments, we preincubated cells at 42°C for 5-15 min and then pulsed them for 1 min before stopping the cellular metabolism and isolating DNA for analysis.
Visualization of Okazaki Fragments by Southern Blot Hybridization-One of the possible reasons for the discrepancy between the in vivo and the in vitro results could have been the detection methods; replication intermediates purified from cells were always detected with sucrose gradients, whereas replication intermediates from in vitro reactions with purified enzymes were always analyzed with gel electrophoresis (see the Introduction). To eliminate this difference, we decided to analyze in vivo replication intermediates with gel electrophoresis. To visualize Okazaki fragments, DNA from ligA251 cells switched from 30 to 42°C was purified and run under denaturing conditions in an alkaline agarose gel, transferred to a membrane, and hybridized with a 19.4-kbp probe derived from a chromosomal region near oriC. This probe produced a strong signal with chromosomal DNA near the top of the well and a much weaker signal with DNA of low molecular weight observed in ligase-deficient conditions (Fig. 3). The size of the low molecular weight DNA was between 0.5 and 2.5 kb, which is typical of Okazaki fragments (1).
Although this approach allowed us to detect what looked like genuine Okazaki fragments, the signal was much weaker than expected. From the doubling time of ϳ32 min at 30°C and from the kinetics of inhibition of DNA synthesis in ligA251 mutant at 42°C ( Fig. 2A; also see Fig. 6C), we anticipated that up to onesixth of the chromosome would be replicated in ligA(Ts) cells incubated for 5 min at 42°C. Therefore, we expected the amount of Okazaki fragments to be up to one-seventh of the overall chromosomal DNA (if the chromosome is divided into six equal parts and one of them is replicated, there are now seven equal parts). To the contrary, our results with regionspecific probes suggested that the signal from Okazaki fragments in ligA mutant conditions comprises no more than onehundredth of the total chromosomal DNA signal for a particular region. Extended incubation times at the nonpermissive temperature did not yield a significantly stronger signal for Okazaki fragments (Fig. 3), suggesting that the amount of these fragments was very low, corroborating earlier observations (27). Wild type cells under similar conditions (not shown) and ligA251 mutant cells at 28°C showed no signal below the chromosomal DNA band (Fig. 3).
Analysis of Tritium-labeled DNA Replication Intermediates-In order to enhance the detection of the replication intermediates, we pulse-labeled DNA of cells growing in ligase-deficient conditions. Classically, replication intermediates are labeled with [ 3 H]thymidine, a DNA-specific label that incorporates within seconds of addition to the growth medium (27,57). To make thymidine incorporation in bacterial cells linear during prolonged incubations, [ 3 H]thymidine labeling is usually done in thyA mutants growing in the presence of limited concentra-tions of thymine or thymidine (1,6,12). Unfortunately, thymine limitation exposes thyA mutant cells to the poorly understood phenomenon of "thymineless death," associated with fragmentation of the newly synthesized DNA (58). Indeed, thy-deficient B. subtilis cells are known to incorporate uracil in the newly replicated DNA, subsequent excision of which produces artifactual "LMW replication intermediates" (59).
We found that ThyA ϩ cells do incorporate thymidine linearly for up to 2 min (not shown), so they can be used for pulselabeling experiments without the need to introduce a thyA mutation (or deoxyadenosine supplementation) (54). We isolated [ 3 H]thymidine pulse-labeled DNA from ligase-deficient and ligase-proficient cells by phenol and chloroform extraction and ran it in alkaline-agarose gels. After the DNA was transferred to a nylon membrane, each lane in the membrane was cut into 10 1-cm-long pieces, which were then scintillationcounted to generate lane profiles (Fig. 4). As expected, pulselabeled species from wild type cells were found exclusively in HMW fractions near the well (Fig. 4A). Pulse-labeled species from ligA251 mutant cells grown at the permissive temperature of 30°C were also of HMW (Fig. 4B). In contrast, pulse-labeled species from ligA mutants grown at 42°C ran as a two-peak distribution (Fig. 4, A and B). A minor part of the signal was found near the well (HMW species), whereas the major part of the pulse-labeled species migrated between fractions 2 and 6, which corresponds to 0.3-3 kb of linear DNA, with a mode around 1 kb (Fig. 4C).
One way to test that the LMW species isolated from ligA mutants grown at 42°C are indeed Okazaki fragments is to use mutants with known changes in Okazaki fragment length. One such mutant is dut, which frequently incorporates uracils in its DNA, whereas the subsequent excision of uracils from the newly synthesized strands by uracil-DNA-glycosylase leads to fragmentation of Okazaki fragments (hence the alternative name for dut mutants, sof, for "short Okazaki fragments") (33,35). We compared the molecular weight of the pulse-labeled species synthesized in a ligA251 dut-1 double mutant at 42°C to the one in ligA251 and found a further reduction in size of the LMW species to 0.2-1 kb (Fig. 4D), suggesting that the LMW species in ligA mutants represent the newly synthesized DNA. Since the 3 H signal completely disappeared when the samples were treated with DNase prior to electrophoresis (Fig. 5, A and  B), the pulse-labeling detects authentic DNA in both wild type and ligA mutant cells. Moreover, the LMW species are true replication intermediates, because they can be chased into HMW DNA after the addition of nonradioactive thymidine and return to the permissive temperature (Fig. 5C).
During the original characterization of the LMW replication intermediates by Okazaki and others, it was reported that the nascent DNA, even if purified and analyzed under nondenaturing conditions, was still partially single-stranded (1,6,60). In confirmation of this unexpected and unexplained observation, digestion of our pulse-labeled DNA from ligA mutant cells (Fig.  5A), but not from wild type cells (Fig. 5B), with a singlestranded DNA-specific exonuclease ExoI prior to denaturation removed the bulk of the LMW tritium signal. At this point, it is unclear whether this LMW pulse-labeled DNA becomes single- The ligA251 mutant cells (GR501) were grown at 28°C, the cultures were split in two parts, and incubation continued for the indicated time at the indicated temperatures. Total DNA was isolated, and 1 g of it was run on an alkalineagarose gel, transferred to a membrane, and hybridized with a ubiE-centered 19.4-kbp chromosomal probe. Two separate experiments with different autoradiography exposure times are shown. A, a shorter exposure to emphasize the chromosomal DNA. In this experiment only, the cell's metabolism was stopped differently, by placing the cultures for 10 min at 75°C (stopping with phenol-ethanol-acetate mix produces the same results (not shown)). "Plasmid protocol" signifies a simpler variation of the SDS-phenol extraction procedure of DNA isolation (described in Ref. 68). B, a longer exposure in an independent experiment to highlight Okazaki fragments. stranded in vivo, before DNA isolation, or is an artifact of the DNA isolation procedure (see below).
LMW Replication Intermediates Are Not the Product of DNA Processing by RecBCD Helicase/Nuclease-The ligA251(Ts) mutant cells are known to degrade their chromosome at the nonpermissive temperature (45). Although we have been working within the time window when chromosomal degradation is undetectable (Fig. 2B), since we switched to detecting the newly synthesized DNA only, we decided to determine for how long the newly synthesized DNA remained stable in ligase-deficient conditions. For this, we preincubated cells at 42°C for 5 min, pulse-labeled them with [ 3 H]thymidine, and kept the cells at 42°C, measuring at various time points how much of the label remained incorporated in the new DNA. Unexpectedly, we found that, in contrast to the total chromosomal DNA in ligasedeficient conditions and to the newly synthesized DNA in wild type cells, the newly synthesized DNA in ligase-deficient conditions is unstable, the instability showing no lag (Fig. 6A). The cells were harvested, the DNA was phenol-extracted, and 1 g of this DNA was run in 1.1% alkaline-agarose gels. The gels were treated for transfer, and the DNA was vacuum-transferred to a nylon membrane. Lanes in the membrane were cut in 1-cm pieces and scintillation-counted. 32 P-end-labeled 1-kb and -HindIII markers were run in the same gel. After vacuum transfer, the lanes of the membrane with the markers were autoradiographed. A, ligA251 (GR501) and wild type (WT) (GR523) cells, both at 42°C. The gel is run from right to left. Note that the y axis is logarithmic. B, ligA251 (GR501) at 30°C versus 42°C. C, migration of the total DNA as well as two types of molecular weight markers, 1-kb DNA ladder (Invitrogen) and -HindIII digest, in the same gel. The two parts of the gel are aligned with each other and with the graph in A. Molecular size of fragments is indicated in bp. D, ligA251 dut-1 (AK150) at 30°C versus 42°C.
However, we also found that this newly synthesized DNA in ligase-deficient conditions becomes stable if the RecBCD helicase/exonuclease is inactivated (Fig. 6B), indicating that the newly synthesized DNA in the absence of DNA ligase becomes susceptible to degradation by an enzyme (RecBCD) that can attack only free ends of DNA (61). This RecBCD-catalyzed DNA degradation significantly limits the accumulation of nascent DNA in ligase-deficient conditions, since the ligA recBCD double mutant shows only weak if any inhibition of DNA replication at 42°C, compared with the single ligA mutant (Fig.  6C). The RecBCD degradation may be the reason behind the very low quantities of Okazaki fragments (Fig. 3). Last, the RecBCD-promoted degradation should be at first limited to replication forks, because there is no difference in the overall chromosomal DNA stability between ligA and ligA recBCD mutants during at least the first 15 min at 42°C (Fig. 6D).
Since the RecBCD-catalyzed degradation of duplex DNA generates a variety of single-stranded DNA species ranging from a few nucleotides to a few thousand nucleotides (62, 63), we tested the possibility that the newly synthesized LMW species in ligase-deficient conditions are derived from HMW species by processing of the newly synthesized DNA by the RecBCD helicase/exonuclease. If this idea were true, there should be fewer LMW and more HMW pulse-labeled species in the ligA recBCD double mutant conditions compared with the ligA single mutant. To the contrary, we observed either the same (Fig. 7A) or even higher (Fig. 7B) accumulation of LMW species in the ligA251 ⌬recBCD mutant relative to the ligA single mutant. At the same time, inactivation of the RecBCD enzyme did not lead to increase in the amount of the HMW DNA or the size of the LMW species. These observations clearly show that the LMW pulse-labeled species in ligase-deficient conditions are not generated by the RecBCD processing of double strand ends at broken replication forks.
In vitro treatment with ExoI of DNA samples from ligA251 or ligA251 ⌬recBCD mutants grown at 42°C shows the same level of susceptibility of the nascent DNA to single-stranded DNAspecific exonuclease (Fig. 7A). This result suggests that the single-stranded DNA is formed in vitro during DNA isolation rather than as a result of RecBCD processing in vivo; otherwise, the recBCD defect would influence the ExoI-susceptible fraction. A role of another DNA helicase in the generation of the short single-stranded nascent DNA in vivo cannot be excluded, however.
Alkaline Sucrose Gradients of Tritium-labeled DNA-By running a set of 32 P-labeled markers in our gels (Fig. 4C), we estimated that the HMW DNA generated in the ligase-deficient conditions represents DNA species Ն6 kb in size. Unfortunately, this rather low resolution of the standard agarose gel electrophoresis is not sufficient to distinguish whether these HMW species are the products of continuous leading strand  Fig. 4. B, the same as in A, but with 42°C DNA from wild type (WT) cells. Note the difference in scale. C, the pulselabeled LMW species from ligA251 mutant cells at 42°C are chased into HMW species if the cells are switched back to 30°C for even as little as 2 min. The total amount of radioactivity in the lane is indicated for each profile. synthesis or the result of partial maturation of Okazaki fragments. In the former case, their average size should be about one-fourth of the single-strand genome equivalent (ϳ1.2 Mb), since they should extend from the replication origin to random positions along the chromosome up to the replication terminus, some 2.3 Mb away. In contrast, if the HMW species are the result of residual Okazaki fragment joining activity or synthesis of longer leading strand-specific Okazaki fragments, they are expected to be in the 10 -100-kb range, which is 10 -100 times shorter than 1 Mb. To distinguish between the two possibilities, we employed alkaline sucrose gradients, the only method of separation of single-stranded DNA that can potentially cover the entire range of DNA species from 2 kb all the way to 2 Mb.
To test if the gel-detected "HMW" DNA in ligA mutant conditions was a continuous leading strand and had the average size of one-fourth of the E. coli chromosome, we extracted chromosomal DNA by a gentle alkaline lysis and separated it in alkaline sucrose gradients (Fig. 8). Pulse-labeled DNA from wild-type cells features a dominant HMW peak of the size of the full-length chromosome as well as a prominent shoulder of the IMW DNA (Fig. 8A). In contrast, the bulk of the pulse-labeled DNA synthesized under ligase-deficient conditions runs as LMW species with an average size of slightly below a 2.4-kb marker (Figs. 8 and 9C) and with a noticeable shoulder in the IMW range centered on the 11-kb marker (Fig. 8C). As expected, the bulk of the pulse-labeled DNA from ligA251 mutant grown at the permissive temperature of 30°C was of HMW, with an IMW tail (Fig.  8B). We conclude that no chromosomal length DNA is synthesized in ligase-deficient conditions and that all nascent DNA is produced as Okazaki fragments (a few of them quite long, however).
LMW Pulse-labeled Replication Intermediates Are Not an Artifact of Uracil Excision-We confirmed with sucrose gradients that pulsedlabeled DNA from dut ligA mutants at 42°C sediments even slower that the LMW species from the single ligA mutant (Fig. 9C), indicating extremely short Okazaki fragments in dut mutants. Based on his own in vitro experiments with direct dUTP addition (38) and experiments of others with dut mutants (37), Olivera proposed that the formation of "Okazaki fragments" in the leading strand in wild type cells is an artifact of incorporation and excision of uracils. Indeed, this is exactly what generates "short Okazaki fragments" in dut (sof) mutants (33,35), which we also observe (Figs. 4D, 7B, and 9C). However, Olivera's idea is weakened by the fact that, in dut ϩ cells of E. coli, the level of dUTP in the DNA precursor pools is undetectable (51), which ensures the extremely low level of uracil incorporation into DNA (measured in ung mutants as one per 3.3 ϫ 10 4 nucleotides (52)). This level is too infrequent to account for the formation of Okazaki-like fragments on the leading strand of dut ϩ cells. To test whether uracil excision contributes to the generation of LMW replication intermediates in dut ϩ cells, we inactivated uracil-DNA-glycosylase with an ung mutation in our ligA251 mutant cells. If the LMW intermediates on the leading strand were the artifact of uracil excision, the profile of replication intermediates from ligA ung double mutant cells should be bimodal, featuring an HMW peak and a general increase of the molecular weight of all replication intermediates. In contrast to this expectation, we found that the profiles of replication intermediates in the ligA ung double mutant and in ligA dut ung triple mutant are essentially superimposable with that of the ligA single mutant (Fig. 9C), ruling out uracil excision as a mechanism for generation of Okazaki fragments in vivo.

DISCUSSION
Okazaki fragments are transient LMW replication intermediates that mature into HMW species through enzymatic reactions performed in E. coli by DNA polymerase I and DNA ligase. In order to detect nascent LMW DNA, its maturation has to be blocked by inactivation of one of those two enzymes and by instantly stopping all DNA transactions in the cell. To begin reevaluating the evidence for discontinuous synthesis of the leading DNA strand in E. coli, we employed a combination of one of the strongest DNA ligase deficiencies known with one of the best ways to stop the DNA metabolism. In order to avoid the possible artifacts, linked to the use of thyA mutants in all previous [ 3 H]thymidine labeling studies, we first tried detecting Okazaki fragments without in vivo labeling. The absolute amount of Okazaki fragments was reported to be very low (27). We confirmed by blot hybridization of the total unlabeled DNA that the mass ratio of Okazaki fragments to the total genomic DNA is indeed very low, essentially disqualifying this approach. Since [ 3 H]thymidine pulse-labeling is still possible in ThyA ϩ cells, we ran [ 3 H]thymidine pulse-labeled DNA in alkaline agarose gel electrophoresis to accurately determine the size of Okazaki fragments, which turned out to be between 0.3 and 3 kb, with a mode of 1 kb. We then used a ligA ⌬recBCD mutant to eliminate the possibility that the LMW pulse-labeled DNA in ligase-deficient conditions is an artifact of the RecBCD processing of disintegrated replication forks.
In alkaline-agarose gels, the newly synthesized DNA from ligase-deficient cells showed a bimodal distribution. Most of this DNA was of the LMW, with a small but reproducible fraction having HMW. The size of this HMW DNA could not be determined, however, due to the limit in the resolution power of agarose gels. Through a better resolution of HMW species in alkaline sucrose gradients, we clearly observed that all DNA isolated from ligase-deficient cells is synthesized in LMW and IMW fragments. The low molecular weight of replication intermediates is not the result of breakage during the DNA purification step, since new DNA purified by the same protocol from ligase-proficient cells reaches the length of the entire chromosome. Also, a series of DNA purification techniques were tested by Okazaki when isolating replication intermediates, and, regardless of the method used, all short-pulsed DNA was isolated as short fragments, whereas the bulk of long-pulsed mature DNA was in the chromosome-length species (54). The low level of intermediate molecular weight species seen in our gradients (corresponding to "HMW" DNA in our alkaline gels) is probably the result of a residual ligation due to incomplete inactivation of DNA ligase at the nonpermissive temperature. Alternative explanations include 1) a minor DNA ligase activity (e.g. the LigB protein in E. coli) (64) and 2) direct synthesis of these longer fragments (e.g. as a result of a lower density of RNA primers for Okazaki pieces in the leading strand) (Fig. 1D).
Most importantly, we found that the newly synthesized DNA in ligA ung conditions (defective in both DNA ligase and DNA-uracil removal) is no different than the nascent DNA from the single ligase mutant, in which DNA-uracils are removed. This excludes uracil excision as the cause of the nascent LMW DNA in ligase-deficient conditions. Our results are in agreement with the earlier demonstration of little or no difference in the level or size of Okazaki fragments between a polA single mutant and a polA ung double mutant (37). Other DNA repair processes, that could also contribute to artificial fragmentation of nascent DNA, were later tested by introduction of multiple mutations inactivating various DNA repair pathways in a ligA7(Ts) mutant (65). Since no difference in the alkaline sucrose gradient profiles were found between ligA7 mutant and its derivative carrying ung, uvrA, and mutH mutations, these results were consistent with the discontinuous synthesis of both the leading and lagging strands (65). In the future, it will be important to systematically test all DNA repair pathways in our ligase-deficient conditions for possible involvement in the generation of Okazaki fragments.
Our results confirm prior reports (1, 6, 60) that the pulselabeled replication intermediates are partially single-stranded even prior to denaturation. Currently, we have neither an FIGURE 8. Alkaline sucrose gradients show mostly LMW and some IMW DNA, but not HMW DNA, in ligase-deficient conditions. Cells were grown at 28°C and switched to 42°C for 5 min before 1 min of pulse-labeling with [ 3 H]thymidine. Cells were gently lysed by NaOH-sarcosyl solution, and the lysates were separated in alkaline-sucrose gradients. A, profiles of the pulselabeled DNA from wild type and ligA mutant cells. B, a different experiment; nascent DNA from ligA mutant cells, pulsed at the permissive temperature of 30°C versus the nonpermissive 42°C. The chromosomal marker features DNA of ligA251 mutant labeled overnight at 30°C. C, a separate experiment; nascent DNA from ligA mutant pulsed at 42°C, sedimented with two molecular size markers, 11.1 and 5.7 kb. Note almost no HMW signal in ligase-deficient conditions in this experiment. FIGURE 9. Inactivation of uracil excision does not influence the molecular weight of short nascent DNA synthesized in ligase-deficient conditions. A, confirmation of the ung mutant phenotype. The ung mutant cells are permissive for the uracil-containing phage, isolated from dut ung mutant cells (49,50). U-phage, uracil-containing ; T-phage, normal (thymine-containing) . B, confirmation of the dut ung double mutant phenotype. Plasmid DNA (pSP72; Promega) from dut ung double mutant cells has so much uracil (51) that it is completely relaxed and even linearized by in vitro treatment with uracil DNA glycosylase (UDG) and ExoIII, as described (52). C, profiles of the pulse-labeled DNA from ligA (GR501) versus ligA dut double (AK150), ligA ung double (LA8), and ligA dut ung triple (LA33) mutants. The molecular weight marker is a 32 P-labeled DNA of linearized pSP72 plasmid (2.4 kb; Promega). explanation for this single-strandedness of the replication intermediates nor an understanding of whether it is an in vitro artifact of DNA isolation procedures or the in vivo result of the replication fork progress in ligase-deficient conditions. The sensitivity of the nascent DNA to the RecBCD degradation in vivo argues that a significant fraction of the new DNA is in the linear form (either duplex or single-stranded, because RecBCD degrades both species with a similar efficiency (61)). On the other hand, the same sensitivity of the replication intermediates from recBCD mutants to a single-stranded DNA-specific exonuclease in vitro is consistent with generation of singlestranded LMW fragments during DNA isolation (e.g. as a result of reversal of replication forks in which one of the daughter strands is extended more than the opposite daughter strand).
In conclusion, the current semidiscontinuous model of DNA replication predicts that, in ligase-deficient conditions, the leading strand replicates continuously from oriC to the terminus, so its replication intermediates should be of high molecular weight, whereas the lagging strand replication intermediates should be of low molecular weight. In contrast, our observation that all replication intermediates are of low molecular weight under ligase-deficient conditions supports discontinuous replication on both the lagging and the leading DNA strands. In two lower eukaryotes, in which conditional ligase mutants are available, replication intermediates are also of low molecular weight in ligase-deficient conditions, prompting the same interpretation (66,67). In order to test this interpretation, the contribution from any type of DNA repair should be questioned. Finally, the possibility that only the lagging strand synthesis continues in ligase-deficient conditions needs to be experimentally addressed.