Identification of short 'eukaryotic' Okazaki fragments synthesized from a prokaryotic replication origin

Although archaeal genomes encode proteins similar to eukaryotic replication factors, the hyperthermophilic archaeon Pyrococcus abyssi replicates its circular chromosome at a high rate from a single origin (oriC) as in Bacteria. In further elucidating the mechanism of archaeal DNA replication, we have studied the elongation step of DNA replication in vivo. We have detected, in two main archaeal phyla, short RNA-primed replication intermediates whose structure and length are very similar to those of eukaryotic Okazaki fragments. Mapping of replication initiation points further showed that discontinuous DNA replication in P. abyssi starts at a well-defined site within the oriC recently identified in this hyperthermophile. Short Okazaki fragments and a high replication speed imply a very efficient turnover of Okazaki fragments in Archaea. Archaea therefore have a unique replication system showing mechanistic similarities to both Bacteria and Eukarya.     

DNA replication in dnaB mutants of Escherichia coli: Gene product interaction and synthesis of 4 S pieces

The DNA produced by several dnaB mutants has been examined both in vivo and in vitro. The alleles chosen for study had previously been shown to differ over a wide range in the apparent severity of their effects on DNA replication.Comparison of DNA replication between dnaB heteroallelic diploids and the constituent haploid strains indicates interaction between the dnaB products in the heteroallelic diploids. The data are consistent with a functional multimeric aggregate of dnaB gene products that is at least a tetramer.Alkaline sucrose gradient profiles of pulse-labeled DNA, synthesized by some of the mutants in vivo and by mutant lysates in vitro, exhibit a peak at about 4 S. The 4 S DNA is most apparent in those mutants in which replication is most severely restricted by temperature.This 4 S material can be chased in vitro into DNA larger than Okazaki pieces, and density transfer experiments indicate that these pieces are formed at the replication fork. Conversion of the 4 S material to large DNA is not altered by inhibition of polynucleotide ligase either by the presence of the lig-4 polA1 mutations in vivo or by the addition of nicotinamide mononucleotide in vitro. The in vitro observations suggest that 4 S pieces are formed on only one side of the replication fork.     

Endo R DpnI restriction of Escherichia coli DNA synthesized in vitro. Evidence that the ends of Okazaki pieces are determined by template deoxynucleotide sequence.

A restriction enzyme from Diplococcus pneumoniae, Endo R DpnI, cuts methylated DNA on cellophane discs into pieces which are about the same size as Okazaki pieces.DNA was synthesized in vitro on cellophane discs in the presence of β-nicotinamide mononucleotide to prevent joining of Okazaki pieces. This DNA was methylated by the addition of S-adenosyl methionine to the reaction mixture. When Endo R DpnI was used to cut methylated DNA made in vitro in the presence of S-adenosyl methionine and β-nicotinamide mononucleotide, no decrease in sedimentation of the Okazaki pieces was observed. Control experiments demonstrated that Okazaki pieces were methylated in vitro and that Endo R DpnI was capable of cutting double-stranded DNA containing methylated Okazaki pieces, that is, synthesized in β-nicotinamide mononucleotide and S-adenosyl methionine.These results are interpreted to mean that the ends of Okazaki pieces are non-randomly distributed with respect to 6-methyl adenine residues.     

The Roles of Family B and D DNA Polymerases in Thermococcus Species 9°N Okazaki Fragment Maturation

During replication, Okazaki fragment maturation is a fundamental process that joins discontinuously synthesized DNA fragments into a contiguous lagging strand. Efficient maturation prevents repeat sequence expansions, small duplications, and generation of double-stranded DNA breaks. To address the components required for the process in Thermococcus, Okazaki fragment maturation was reconstituted in vitro using purified proteins from Thermococcus species 9°N or cell extracts. A dual color fluorescence assay was developed to monitor reaction substrates, intermediates, and products. DNA polymerase D (polD) was proposed to function as the replicative polymerase in Thermococcus replicating both the leading and the lagging strands. It is shown here, however, that it stops before the previous Okazaki fragments, failing to rapidly process them. Instead, Family B DNA polymerase (polB) was observed to rapidly fill the gaps left by polD and displaces the downstream Okazaki fragment to create a flap structure. This flap structure was cleaved by flap endonuclease 1 (Fen1) and the resultant nick was ligated by DNA ligase to form a mature lagging strand. The similarities to both bacterial and eukaryotic systems and evolutionary implications of archaeal Okazaki fragment maturation are discussed.     

Size distribution of DNA replicative intermediates in bacteriophage P4 and in Escherichia coli.

We have tested the hypothesis that Okazaki fragment replicative intermediates have defined termini using as a model system the in vivo DNA replication of the tiny bacteriophage P4. The kinetics of formation of intermediates in P4 DNA replication have been investigated. P4 DNA replication in DNA polymerase I-deficient mutants generates Okazaki fragments with a size distribution similar to that in uninfected cells. When P4-derived Okazaki fragments are resolved by agarose gel electrophoresis, no discrete size classes appear. This finding is incompatible with sequence-specific models of Okazaki fragment formation but supports the view that these replication intermediates are initiated and terminated at random locations on the P4 chromosome.     

DNA replication in Escherichia coli: physical and kinetic studies of the replication point

DNA replication in Escherichia coli was followed using analytical composite agarose-acrylamide gel electrophoresis of native and denatured DNA samples from cells that had been pulse-labeled in vivo. The results obtained with samples of native DNA, combined with the different sensitivities of the various electrophoretic fractions of native DNA to enzymatic attack, led us to conclude that a significant amount of single-strandedness exists in the region of the replication site. Data obtained from kinetic analyses of the labeling of both the high molecular weight DNA and the fragments formed on denaturation of native DNA samples were consistent with the Okazaki model (Okazaki et al., 1968a,b) of DNA replication. Furthermore, the data indicated that eight to ten precursor fragments are present per fork during bacterial growth at both 20 and 37 °C, under the conditions used in this study.     

Evidence for the presence of ribonucleotides at the 5' termini of some DNA molecules isolated from Escherichia coli polAex2.

We have purified a set of small DNA molecules from various strains of exponentially growing Escherichia coli, including E. coli polAex2. This material included very short molecules (2 S), the nascent DNA (“Okazaki fragments”) and some longer molecules. Most of the [³H]thymidine incorporated during a brief period of labeling was found in the 5 S to 15 S Okazaki fragments. There was a large number of the 2 S molecules in the cell. The properties of the 5′ ends of these molecules were investigated using three procedures. (1) The DNA preparation, pulse-labeled with [³H]thymidine, was reacted with polynucleotide kinase and ATP to insure that all 5′ ends were phosphorylated. After subjection of the DNA to alkaline hydrolysis, the proportion of incorporated ³H pulse-label that became susceptible to digestion by spleen exonuclease was determined. In different experiments there was an increment of up to 20% in the amount of pulse-labeled E. coli polAex2 DNA that could be hydrolyzed by the exonuclease after treatment with alkali. (2) As in the preceding protocol, phosphorylation of the 5′ ends was assured by reaction with kinase and ATP; the preparation was then treated with alkali and the number of 5′-OH ends generated that could be labeled with ³²P using [γ-³²P]ATP and kinase in a second reaction was determined. The data indicated that 3 to 30% of the molecules could be labeled after alkali digestion, but not before. (3) The DNA molecules were reacted with kinase and [γ-³²P]ATP after having been exposed previously to alkaline phosphatase. The end-labeled molecules were then subjected to an alkaline hydrolysis and the resulting hydrolysate chromatographed on a polyethyleneimine-cellulose thinlayer plate. Alkali treatment was found to release 2′(3′),5′-ribonucleoside diphosphates from 1 to 30% of the molecules; pAp and pGp predominated. Control experiments showed that these ribonucleotides were covalently linked to the 5′ ends of polydeoxyribonucleotides. Curiously, the smaller the DNA molecule the less likely it was to possess a 5′-terminal ribonucleotide. Very few apparent RNA/DNA molecules were observed in the non-polAex2 strains tested. These observations are in part in agreement with previous reports, and we infer that at least some of the nascent E. coli polAex2 DNA molecules are initiated in vivo with a ribonucleotide primer. The relatively smaller proportion of molecules with apparent 5′-terminal ribonucleotides among the smaller DNA molecules and in strains other than E. coli polAex2 suggests to us that there may exist a mechanism for initiating DNA molecules that does not require an RNA primer.     

Roles of DNA polymerase I in leading and lagging-strand replication defined by a high-resolution mutation footprint of ColE1 plasmid replication

DNA polymerase I (pol I) processes RNA primers during lagging-strand synthesis and fills small gaps during DNA repair reactions. However, it is unclear how pol I and pol III work together during replication and repair or how extensive pol I processing of Okazaki fragments is in vivo. Here, we address these questions by analyzing pol I mutations generated through error-prone replication of ColE1 plasmids. The data were obtained by direct sequencing, allowing an accurate determination of the mutation spectrum and distribution. Pol I’s mutational footprint suggests: (i) during leading-strand replication pol I is gradually replaced by pol III over at least 1.3 kb; (ii) pol I processing of Okazaki fragments is limited to ∼20 nt and (iii) the size of Okazaki fragments is short (∼250 nt). While based on ColE1 plasmid replication, our findings are likely relevant to other pol I replicative processes such as chromosomal replication and DNA repair, which differ from ColE1 replication mostly at the recruitment steps. This mutation footprinting approach should help establish the role of other prokaryotic or eukaryotic polymerases in vivo, and provides a tool to investigate how sequence topology, DNA damage, or interactions with protein partners may affect the function of individual DNA polymerases.     

The endonuclease activity of the yeast Dna2 enzyme is essential in vivo

Dna2 is a multifunctional enzyme in yeast that possesses endonuclease activity well suited to remove RNA–DNA primers of Okazaki fragments, raising the question of whether endonuclease activity is essential for in vivo Dna2 function. Systematic site-directed mutations of amino acid residues in Saccharomyces cerevisiae DNA2 conserved in the central region of many eukaryotic DNA2 homologs allowed us to identify mutant dna2 alleles that were divided into three groups based on the viability of the mutant cells: (i) viable; (ii) inviable only when expression was repressed; (iii) inviable. Biochemical analyses of recombinant mutant Dna2 proteins isolated from the latter two groups revealed that they possessed normal ATPase/helicase activity, but were impaired in their endonuclease activity. Cells expressing mutant Dna2 enzymes partially impaired in endonuclease activity were viable, but were unable to grow when expression of their mutant Dna2 enzymes was further reduced. Their growth was restored when the mutant Dna2 proteins decreased in nuclease activity were induced to overexpress. In contrast, mutant Dna2 proteins lacking endonuclease activity did not allow cells to grow under any conditions tested. These in vivo and in vitro results demonstrate that the endonuclease activity of Dna2 is essential for Okazaki fragment processing.     

Synthesis and degradation of termination and premature-termination fragments of beta-galactosidase in vitro and in vivo.

A number of abnormal polypeptides which are products of the lacZ gene in Escherichia coli have been characterized. A variety of experiments indicates that one class of at least nine different fragments arises from premature termination of protein synthesis. They have been detected as products of protein synthesis both in vitro and in vivo, although the percentage of fragments made in vitro is greater than that in vivo. The percentage of total Z gene-encoded protein which is found as fragments is estimated to be roughly 23% in vivo. This means that about 31% of the total number of β-galactosidase monomers synthesized are prematurely terminated molecules. The average molecular weight of the polypeptides produced is 91,000. This corresponds to a termination event occurring on the average once every 3200 codons. The sites at which termination occurs appear to be specific and are located primarily near the 3′-end of the gene.Polypeptides synthesized in vitro and in vivo from templates containing mutations in the Z gene have also been compared. Most of the mutationally generated fragments synthesized in vitro are stable, unlike their in vivo counterparts, which are often rapidly degraded. One fragment, however, generated by an early amber mutation in the Z gene, is degraded in the in vitro system. The mechanism of degradation appears to be specific for small abnormal polypeptides. Internal reinitiation polypeptides generated by nonsense mutations, which have been found in vivo, are not detected in the in vitro protein synthesis system under the conditions used here.     

In vivo and in vitro studies of Mgs1 suggest a link between genome instability and Okazaki fragment processing

The non-essential MGS1 gene of Saccharomyces cerevisiae is highly conserved in eukaryotes and encodes an enzyme containing both DNA-dependent ATPase and DNA annealing activities. MGS1 appears to function in post-replicational repair processes that contribute to genome stability. In this study, we identified MGS1 as a multicopy suppressor of the temperature-sensitive dna2Δ405N mutation, a DNA2 allele lacking the N-terminal 405 amino acid residues. Mgs1 stimulates the structure-specific nuclease activity of Rad27 (yeast Fen1 or yFen1) in an ATP-dependent manner. ATP binding but not hydrolysis was sufficient for the stimulatory effect of Mgs1, since non-hydrolyzable ATP analogs are as effective as ATP. Suppression of the temperature-sensitive growth defect of dna2Δ405N required the presence of a functional copy of RAD27, indicating that Mgs1 suppressed the dna2Δ405N mutation by increasing the activity of yFen1 (Rad27) in vivo. Our results provide in vivo and in vitro evidence that Mgs1 is involved in Okazaki fragment processing by modulating Fen1 activity. The data presented raise the possibility that the absence of MGS1 may impair the processing of Okazaki fragments, leading to genomic instability.     

Differential Activities of DNA Polymerases in Processing Ribonucleotides during DNA Synthesis in Archaea

Consistent with the fact that ribonucleotides (rNTPs) are in excess over deoxyribonucleotides (dNTPs) in vivo, recent findings indicate that replicative DNA polymerases (DNA Pols) are able to insert ribonucleotides (rNMPs) during DNA synthesis, raising crucial questions about the fidelity of DNA replication in both Bacteria and Eukarya. Here, we report that the level of rNTPs is 20-fold higher than that of dNTPs in Pyrococcus abyssi cells. Using dNTP and rNTP concentrations present in vivo, we recorded rNMP incorporation in a template-specific manner during in vitro synthesis, with the family-D DNA Pol (PolD) having the highest propensity compared with the family-B DNA Pol and the p41/p46 complex. We also showed that ribonucleotides accumulate at a relatively high frequency in the genome of wild-type Thermococcales cells, and this frequency significantly increases upon deletion of RNase HII, the major enzyme responsible for the removal of RNA from DNA. Because ribonucleotides remain in genomic DNA, we then analyzed the effects on polymerization activities by the three DNA Pols. Depending on the identity of the base and the sequence context, all three DNA Pols bypass rNMP-containing DNA templates with variable efficiency and nucleotide (mis)incorporation ability. Unexpectedly, we found that PolD correctly base-paired a single ribonucleotide opposite rNMP-containing DNA templates. An evolutionary scenario is discussed concerning rNMP incorporation into DNA and genome stability.     

Dynamics of enzymatic interactions during short flap human Okazaki fragment processing by two forms of human DNA polymerase δ

Lagging strand DNA replication requires the concerted actions of DNA polymerase δ, Fen1 and DNA ligase I for the removal of the RNA/DNA primers before ligation of Okazaki fragments. To better understand this process in human cells, we have reconstituted Okazaki fragment processing by the short flap pathway in vitro with purified human proteins and oligonucleotide substrates. We systematically characterized the key events in Okazaki fragment processing: the strand displacement, Pol δ/Fen1 combined reactions for removal of the RNA/DNA primer, and the complete reaction with DNA ligase I. Two forms of human DNA polymerase δ were studied: Pol δ4 and Pol δ3, which represent the heterotetramer and the heterotrimer lacking the p12 subunit, respectively. Pol δ3 exhibits very limited strand displacement activity in contrast to Pol δ4, and stalls on encounter with a 5′-blocking oligonucleotide. Pol δ4 and Pol δ3 exhibit different characteristics in the Pol δ/Fen1 reactions. While Pol δ3 produces predominantly 1 and 2 nt cleavage products irrespective of Fen1 concentrations, Pol δ4 produces cleavage fragments of 1–10 nts at low Fen1 concentrations. Pol δ3 and Pol δ4 exhibit comparable formation of ligated products in the complete system. While both are capable of Okazaki fragment processing in vitro, Pol δ3 exhibits ideal characteristics for a role in Okazaki fragment processing. Pol δ3 readily idles and in combination with Fen1 produces primarily 1 nt cleavage products, so that nick translation predominates in the removal of the blocking strand, avoiding the production of longer flaps that require additional processing. These studies represent the first analysis of the two forms of human Pol δ in Okazaki fragment processing. The findings provide evidence for the novel concept that Pol δ3 has a role in lagging strand synthesis, and that both forms of Pol δ may participate in DNA replication in higher eukaryotic cells.     

TbPIF5 Is a Trypanosoma brucei Mitochondrial DNA Helicase Involved in Processing of Minicircle Okazaki Fragments

Trypanosoma brucei''s mitochondrial genome, kinetoplast DNA (kDNA), is a giant network of catenated DNA rings. The network consists of a few thousand 1 kb minicircles and several dozen 23 kb maxicircles. Here we report that TbPIF5, one of T. brucei''s six mitochondrial proteins related to Saccharomyces cerevisiae mitochondrial DNA helicase ScPIF1, is involved in minicircle lagging strand synthesis. Like its yeast homolog, TbPIF5 is a 5′ to 3′ DNA helicase. Together with other enzymes thought to be involved in Okazaki fragment processing, TbPIF5 localizes in vivo to the antipodal sites flanking the kDNA. Minicircles in wild type cells replicate unidirectionally as theta-structures and are unusual in that Okazaki fragments are not joined until after the progeny minicircles have segregated. We now report that overexpression of TbPIF5 causes premature removal of RNA primers and joining of Okazaki fragments on theta structures. Further elongation of the lagging strand is blocked, but the leading strand is completed and the minicircle progeny, one with a truncated H strand (ranging from 0.1 to 1 kb), are segregated. The minicircles with a truncated H strand electrophorese on an agarose gel as a smear. This replication defect is associated with kinetoplast shrinkage and eventual slowing of cell growth. We propose that TbPIF5 unwinds RNA primers after lagging strand synthesis, thus facilitating processing of Okazaki fragments.     

DNA degradation in an amber mutant of Escherichia coli K12 affecting DNA ligase and viability

Isolation of an amber mutant lig-321 (or dnaL321) if Escherichia coli K12 with a defect in DNA ligase activity was previously reported (Nagata & Horiuchi, 1974). This was the first demonstration that, in E. coli, conditionally lethal nonsense mutants can be isolated selectively. Unlike the hitherto available E. coli K12 DNA ligase-deficient (lig) mutants, the DNA of this mutant is degraded under lethal conditions. This paper describes its further characterization. The DNA degradation was found to be an energy-requiring process, in which endonuclease I did not seem to participate. Kinetic analyses of prelabeled DNA indicated that the parental strands were degraded. The sedimentation profile of prelabeled DNA in an alkaline sucrose gradient showed that the extensive degradation was preceded by a step in which the parental strands were broken into relatively large pieces. At least in the early phase of degradation, which we examined by alkaline sucrose gradient centrifugation of pulse-labeled DNA, synthesis of discontinuous daughter chains (Okazaki fragments, Okazaki et al., 1968) was confirmed. Joining of the nascent chains, however, was completely inhibited. Genetic analyses revealed that the mutant allele is recessive to the wild type. This agrees with in vitro studies in which the mutant crude extract was found not to inhibit DNA ligase activity of the wild type extract. These and other properties of the lig-321 mutant were compared with the other DNA ligase-deficient mutants of E. coli. The role of this enzyme in DNA replication, repair and recombination is discussed.     

RNA primers in Simian virus 40 DNA replication. II. Distribution of 5' terminal oligoribonucleotides in nascent DNA.

A cell-free simian virus 40 (SV40) DNA replication system served to study the role of RNA in the initiation of nascent DNA chains of less than 200 nucleotides (Okazaki pieces). RNA-DNA covalent linkages were found to copurify with SV40 replicating DNA. These linkages were identified by transfer of a fraction of the ³²P from the 5′ position of a deoxyribonucleotide to 2′(3′)rNMPs upon either alkaline hydrolysis or RNAase T₂ digestion of SV40 replicating [³²P]DNA. Alkaline hydrolysis also exposed 5′ terminal hydroxyl groups in the nascent DNA which were detected as nucleosides after digestion with P1 nuclease. The RNA-DNA covalent linkages resulted from a population of Okazaki pieces containing uniquely sized oligoribonucleotides covalently attached to their 5′ termini (RNA primers). The density of a portion of the Okazaki pieces in potassium iodide gradients corresponded to a content of 90% DNA and 10% RNA, while the remaining Okazaki pieces appeared to contain only DNA. Incubation of Okazaki pieces with a defined length in the presence of either RNAase T₂ or potassium hydroxide converted about one-third to one-half of them intto a second well defined group of DNA chains of greater electrophoretic mobili y in polyacrylamide gels. The increased mobility corresponded to the removalof at least seven-residues. Since alkaline hydrolysis of similar Okazaki pieces revealed that one-third to one-half of them contained rN-³²P-dN linkages, the oligoribonucleotides must be covalently attached to the 5′ ends of nascent DNA chains. Although the significance of two populations of Okazaki pieces, one with and one without RNA primers, is imperfectly understood, a sizable fraction of nascent DNA chains clearly contained RNA primers.Neither the length of the RNA primer nor the number of RNA primers per DNA chain changed significantly with increasing length of Okazaki pieces. Since the frequency of RNA-DNA junctions found in nascent DNA chains greater than 400 nucleotides was similar to that of Okazaki pieces, the complete excision of RNA primers appears to occur after Okazaki pieces are joined to the 5′ end of growing daughter strands.³²P-label transfer analysis of Okazaki pieces recovered from hybrids with isolated HindII + III restriction fragments of SV40 DNA revealed a uniform distribution of rN-P-dN sequences around the replicating DNA molecule. Therefore, most, if not all, RNA primers serve to initiate Okazaki pieces rather than to initiate DNA replication at the origin of the genome. Moreover, the positions of RNA primers are not determined by a specific set of nucleotide sequences.     

Mechanism of CRISPR-RNA guided recognition of DNA targets in Escherichia coli

In bacteria and archaea, short fragments of foreign DNA are integrated into Clustered Regularly Interspaced Short Palindromic Repeat (CRISPR) loci, providing a molecular memory of previous encounters with foreign genetic elements. In Escherichia coli, short CRISPR-derived RNAs are incorporated into a multi-subunit surveillance complex called Cascade (CRISPR-associated complex for antiviral defense). Recent structures of Cascade capture snapshots of this seahorse-shaped RNA-guided surveillance complex before and after binding to a DNA target. Here we determine a 3.2 Å x-ray crystal structure of Cascade in a new crystal form that provides insight into the mechanism of double-stranded DNA binding. Molecular dynamic simulations performed using available structures reveal functional roles for residues in the tail, backbone and belly subunits of Cascade that are critical for binding double-stranded DNA. Structural comparisons are used to make functional predictions and these predictions are tested in vivo and in vitro. Collectively, the results in this study reveal underlying mechanisms involved in target-induced conformational changes and highlight residues important in DNA binding and protospacer adjacent motif recognition.     

Coordinated leading- and lagging-strand synthesis at the Escherichia coli DNA replication fork. V. Primase action regulates the cycle of Okazaki fragment synthesis.

Replication forks formed during rolling-circle DNA synthesis supported by a tailed form II DNA substrate in the presence of the primosome, the single-stranded DNA binding protein, and the DNA polymerase III holoenzyme (Pol III HE) that had been reconstituted from the purified subunits, beta, tau, and the gamma.delta complex, at limiting (with respect to nucleotide incorporation) concentrations of the Pol III core (alpha, epsilon, and theta) produced aberrantly small Okazaki fragments, while the synthesis of the leading strand was unperturbed. These small Okazaki fragments were not arrayed in tandem along the lagging-strand DNA template, but were separated by large gaps. Similarly structured synthetic products were not manufactured by replication forks reconstituted with higher, saturating concentrations of the Pol III core. Replication forks producing these small fragments could respond, by modulating the size of the Okazaki fragments produced, to variations in the concentration of NTPs or the primase, conditions that affect the frequency of priming on the lagging strand, but not to variation in the concentration of dNTPs, conditions that affect the frequency of utilization of the primers. Significantly longer Okazaki fragments (greater than 7 kilobases) could be produced in the presence of a limiting amount of Pol III core at low concentrations of the primase. These observations indicated that the production of small Okazaki fragments was not a result of a debilitated lagging-strand Pol III core, but rather a function of the time available for nascent strand synthesis during the cycle of events that are required for the manufacture of an Okazaki fragment and that it was the association of primase with the replication fork that keyed this cycle.     

Synthesis of small polynucleotide chains in thymine-depleted bacteria

Bacteria that are depleted of intracellular thymidine nucleotide pools incorporate [³H]thymine at full specific activity, allowing the detection of early intermediates in DNA replication. A short pulse of [³H]thymine is incorporated almost exclusively into very small DNA chains which, during further incorporation of thymine, are converted into larger chains and high molecular weight DNA. The synthesis of these small DNA chains depends on the products of dna genes B, E and G. Analysis of the DNA by gel filtration on Sepharose 2B revealed an abundance of extremely short DNA chains while the frequency of larger chains decreased exponentially with increasing size. This size distribution of small DNA chains suggests a mechanism of DNA replication in which larger chains (Okazaki pieces, Okazaki et al., 1968a) arise through joining of extremely short polynucleotide chains in a process resembling crystallization rather than unidirectional chain elongation.