Translesion DNA polymerases remodel the replisome INAUGURAL ARTICLE and alter the speed of the replicative helicase

Chiara Indiania, Lance D. Langstona, Olga Yurievaa, Myron F. Goodmanb, and Mike O’Donnella,1

aLaboratory of DNA Replication, Howard Hughes Medical Institute, The Rockefeller University, 1230 York Avenue, Box 228, New York, NY 10065; and bDepartments of Biological Sciences and Chemistry, University of Southern California, Los Angeles, CA 90089

This contribution is part of the special series of Inaugural Articles by members of the National Academy of Sciences elected in 2006.

Contributed by Mike O’Donnell, February 11, 2009 (sent for review January 22, 2009)

All cells contain specialized translesion DNA polymerases that very slow to rise, peaking 45 min after SOS induction (4, 9, 11, replicate past sites of DNA damage. We find that Escherichia coli 12). In vitro studies have shown that these DNA polymerases translesion DNA polymerase II (Pol II) and polymerase IV (Pol IV) function with the ␤-clamp for lesion bypass and trade places on function with DnaB helicase and regulate its rate of unwinding, ␤ with a stalled Poll III replicase, as originally demonstrated in slowing it to as little as 1 bp/s. Furthermore, Pol II and Pol IV freely bacteriophage T4 (14–17). Once the lesion is bypassed, the exchange with the polymerase III (Pol III) replicase on the ␤-clamp high-fidelity Pol III can switch back on the ␤-clamp to reestablish and function with DnaB helicase to form alternative replisomes, the fast moving E. coli replisome (650 nt per s) (14, 17). In this even before Pol III stalls at a lesion. DNA damage-induced levels of view of translesion synthesis, the TLS polymerase only occupies Pol II and Pol IV dominate the clamp, slowing the helicase and the ␤-clamp during the short time interval of lesion bypass. stably maintaining the architecture of the replication machinery It seems unlikely that TLS polymerases would be able to while keeping the fork moving. We propose that these dynamic replace Pol III in the context of a moving replisome, because actions provide additional time for normal excision repair of TLS polymerases are very slow and have low fidelity. These lesions before the replication fork reaches them and also enable characteristics are inconsistent with the rapid speed and high the appropriate translesion polymerase to sample each lesion as it fidelity required for chromosome replication. It is possible that

is encountered. TLS polymerases are excluded from replisomes by the tight BIOCHEMISTRY connection between Pol III and the DnaB helicase via the DNA repair ͉ DNA replication ͉ replication fork ͉ replisome sliding clamp ␶-subunit. Additional insurance against TLS polymerase entry into the replisome is the known action of DnaB helicase when it hromosomes are duplicated by replisome machines contain- becomes uncoupled from Pol III, whereupon it continues for Ϸ Cing helicase, primase, and DNA polymerase activites (1). only 1 kb before DnaB stops unwinding (18–20). Presumably, Replicative DNA polymerases are tethered to DNA by ring- DnaB dissociates from DNA when it is uncoupled from leading- shaped sliding clamp proteins. The Escherichia coli polymerase strand Pol III action. Once DnaB dissociates, it cannot get back III (Pol III) holoenzyme replicase contains 10 different subunits onto DNA without specialized helicase-loading factors, and consisting of a clamp loader that binds 2 molecules of Pol III, replication fork progression is halted. Therefore, even if a TLS each tethered to DNA by the ␤-sliding clamp for processive polymerase enters the replisome, the subsequent loss of Pol III leading and lagging strand synthesis (2). Pol III also binds tightly interaction with DnaB may result in DnaB dissociation, prevent- to the hexameric DnaB helicase that encircles the lagging strand ing further progression of the fork. template. The Pol III-to-DnaB interaction is mediated by the Fork breakdown upon takeover by Pol IV is supported by ␶-subunit within Pol III (3). This tight connection couples Pol III recent studies that show that overexpression of Pol IV in the motion to DnaB unwinding and increases the rate of unwinding absence of DNA damage inhibits cell growth in Bacillus subtilis by DnaB helicase from a basal level of 35 bp/s to a rate in excess and E. coli (21, 22). The authors of the E. coli study (21) note that of 500 bp/s (3). These tight connections of Pol III to DnaB and induction of Pol IV stops replication abruptly and kills the cell, the ␤-clamp provide the replisome with sufficient stability, in suggesting that Pol IV entry into the replisome is catastrophic. principle, to replicate the entire genome. In practice, however, Interestingly, these observations did not require the clamp replication of long chromosomal DNA is an uneven path with a binding residues of Pol IV, suggesting that Pol IV blocks variety of obstacles along the way, including protein blocks and replication by a process distinct from TLS polymerase switching damaged templates. on a clamp. However, Pol IV was overexpressed to abnormally In the presence of high levels of DNA damage, Ͼ40 gene high levels in that study. Thus, the mechanism of the replication products are induced by the SOS response that act to restore block by Pol IV was suggested to be separate from the physio- genomic integrity and assure cell survival (4). Among the logical action of the enzyme. Lower levels of Pol IV expression SOS-induced gene products are 3 translesion DNA polymerases, appeared to decrease replication without cell killing, suggesting Pols II, IV, and V (encoded by polB, dinB, and umuCD, Pol IV may slow a replication fork, but this hypothesis could not respectively), that perform potentially mutagenic DNA synthesis be confirmed and was not tested in vitro. across template lesions (5). Pols IV and V are members of the In the current study we examine the effects of normal and Y family of error-prone DNA polymerases; they lack a proof- induced levels of Pol II and Pol IV on moving replisomes reading 3Ј–5Ј exonuclease (6, 7). Pol II is a B-family polymerase;

it has high fidelity and contains proofreading activity (8). Pols II Author contributions: C.I., L.D.L., and M.O.D. designed research; C.I., L.D.L., and O.Y. and IV are present during normal cell growth and may be performed research; M.F.G. contributed new reagents/analytic tools; C.I., L.D.L., M.F.G., involved in repairing low levels of DNA damage that occur and M.O.D. analyzed data; and C.I., L.D.L., and M.O.D. wrote the paper. routinely in the cell (9–11). Pol V, however, is closely associated The authors declare no conflict of interest. with mutagenesis and is not detectable in cells under normal Freely available online through the PNAS open access option. conditions (7, 12, 13). 1To whom correspondence should be addressed. E-mail: [email protected]. Pol II and Pol IV are among the first genes that are up- This article contains supporting information online at www.pnas.org/cgi/content/full/ regulated in the SOS response, whereas the levels of Pol V are 0901403106/DCSupplemental.

www.pnas.org͞cgi͞doi͞10.1073͞pnas.0901403106 PNAS Early Edition ͉ 1of8 Downloaded by guest on September 25, 2021 Fig. 1. Pol II and Pol IV form alternative replisomes with DnaB helicase and the ␤ clamp. (A) Minicircle DNA was incubated with DnaB, then TLS polymerase (5 pmol), clamp loader, clamp and dCTP, dGTP for 5 min. Replication was initiated by addition of SSB, DnaG primase, dATP, dTTP, [␣32P]dTTP, and the 4 rNTPs. Newly-synthesized leading strand DNA is in red, and lagging strand is in blue. As controls, either DnaB or clamp/clamp loader were excluded from the reaction. (B and C) Time courses of Pol II (B) or Pol IV (C) in the presence of DnaB, the ␤-clamp, and the clamp loader (lanes 1–6), in the absence of clamp and clamp loader (lanes 7–12), and in the absence of DnaB (lanes 13–18). Products were analyzed in alkaline agarose gels. (D) Pol II (15 pmol) and Pol IV [5 pmol (Center) or 7 pmol (Right)] were added simultaneously to the reaction. (Left) The gel shows synthesis by Pol II in the absence of Pol IV for comparison.

reconstituted in vitro. We find that these TLS polymerases can addition of Pol II or Pol IV in the presence of only dCTP and gain access to the moving Pol III replisome despite the strong Pol dGTP to prevent fork movement. After 5 min, replication was III–␶–DnaB connection. Surprisingly, the fork does not break initiated upon adding dATP, dTTP, SSB, DnaG primase, and down. Instead the TLS polymerases function with DnaB to form the 4 rNTPs (see Fig. 1A). In these assays, dissociation of DnaB replisomes that move more slowly than the intrinsic rate of from DNA will halt the fork because SSB is present and the assay DnaB, with the rate of helicase unwinding adjusting to the speed does not contain DnaB-reloading factors. Replication fork pro- of the polymerase at the fork. At the levels of Pol II and Pol IV gression is followed by analysis of timed aliquots in an alkaline present in undamaged cells, the Pol III replisome is unaffected, agarose gel. The results are compared with similar reactions whereas at levels of Pol II and Pol IV that correspond to their performed in the absence of DnaB helicase or the absence of the concentrations in DNA damage-induced cells, the TLS poly- clamp and clamp loader. merases dominate the replication fork and slow it down. Ex- Surprisingly, the analysis shows that Pol II and Pol IV function pression of Pol II and Pol IV in vivo in the absence of DNA with DnaB and the ␤-clamp, because the presence of both of damage also slows DNA replication and depends on TLS these factors is required to synthesize long DNA chains (Fig. 1 polymerase interaction with the ␤-clamp. Slow replication fork B and C, lanes 1–6). DNA synthesis continues for at least 20 min, advance in the context of DNA damage makes biological sense indicating that the DnaB helicase acts processively and is main- as it buys time for normal DNA repair, decreasing collision tained the entire time. The rates of Pol II and Pol IV fork events of the fork with DNA lesions. progression are Ϸ10 nucleotides (ntd)/s and 1 ntd/s, respectively (Fig. S1), significantly slower than Pol III replisomes under the Results same conditions and consistent with the slower synthetic rates of Pol II and Pol IV Form Alternative Replisomes with the ␤-Clamp and Pol II and Pol IV compared with Pol III (24, 25). The ability of DnaB Helicase. To investigate possible alternative roles for Pol II Pol II and Pol IV to function with DnaB in the absence of Pol and Pol IV beyond translesion synthesis, we examined their III indicates that replisomes containing these TLS polymerases ability to form a functional replisome with the ␤-clamp and remain intact rather than undergo replication fork breakdown. DnaB helicase in the absence of Pol III. For these studies, we These results also suggest that the speed of a replicative helicase used a minicircle replication fork substrate consisting of a can be limited by the rate of synthesis of the polymerase traveling synthetic 100-mer circular duplex with a 5Ј T40 ssDNA tail (20, along behind it. Lagging strand products are also observed by 23). Each strand of the synthetic minicircle substrate contains using [␣32P] dATP as the radiolabel as shown in Fig. S2.We only 3 nt: dG, dC, dA on the leading strand template, and dG, presume that leading and lagging strand synthesis is uncoupled dC, dT on the lagging stand template. Hence, leading and lagging under these conditions, as further described in Discussion. strands can be specifically labeled in separate reactions using Both Pol II and Pol IV are present in normal cells and are either [␣32P] dTTP or [␣32P] dATP, respectively. greatly up-regulated during the SOS response. As shown in Fig. The replication machinery at the fork was assembled by 1D, the rate of DNA synthesis in the presence of Pol II and Pol loading DnaB on the 5Ј dT40 tail of the minicircle DNA in the IV (Center and Right) was slower than that of Pol II alone (Left), absence of ssDNA-binding protein (SSB), then the ␤-clamp was suggesting that both polymerases have access to the fork and loaded onto DNA by using the clamp loader followed by the alternate in taking control of the growing DNA ends in a

2of8 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0901403106 Indiani et al. Downloaded by guest on September 25, 2021 ␤-clamp (referred to as Pol III*) was added. If Pol III* takes control of the replication fork, the radiolabel incorporated by Pol II or Pol IV during the pulse period will quickly shift to a higher INAUGURAL ARTICLE molecular mass during the chase period because of the rapid rate of synthesis by Pol III*-␤. As shown in Fig. 2B, Pol II (Left) synthesized Ϸ3 kb during the 7-min pulse of [␣32P] dTTP, and Pol IV (Right) synthesized Ϸ1 kb during a 15-min pulse (lanes 1, 5, and 9), consistent with the results of Fig. 1. Addition of a large excess of cold dNTPs had no effect on the rate of synthesis by Pol II or Pol IV in the absence of added Pol III* (Fig. 2B, lanes 2, 6, and 10), but when Pol III* was added immediately after the chase, the pulse-labeled band shifted dramatically higher, in a time-dependent manner (Fig. 2B, lanes 3, 7, and 11). As a control for the effectiveness of the chase, Pol III* was added to replication forks assembled without Pol II or Pol IV and showed no detectable incorporation of [␣32P] dTTP (Fig. 2B, lanes 4, 8, and 12). This pulse–chase experiment confirms that the DnaB helicase is fully functional within the alternative Pol II and Pol IV replisomes and that the basic architecture of the machinery at the replication fork is maintained by the specialized DNA polymerases despite their slow speed compared with Pol III*-␤. Hence, DnaB is stable on DNA in a functional state, and its rate of DNA unwinding is determined by the speed of the DNA polymerase that functions with it. We also conclude that Pol III* Fig. 2. Pol III regains control from a moving Pol II- or Pol IV-based replisome. takeover of Pol II and Pol IV replisomes is rapid and efficient. (A) The Pol II- or Pol IV-based replisome was assembled as in Fig. 1A. After a 7-min pulse (Pol II) or 15-min pulse (Pol IV), a 100-fold excess of unlabeled Pol II and Pol IV Recruit the Clamp and Helicase from a Moving Pol III BIOCHEMISTRY dNTPs (chase) was added 10 s before Pol III (1 pmol), and reactions were Replisome. Our previous studies with singly-primed ssDNA cir- quenched at different times after the addition of Pol III. Radiolabeled DNA cles showed that when Pol III holoenzyme was stalled at a primed synthesized by the TLS polymerase during the pulse period is shown in red, and site by nucleotide omission Pol IV takes control of the primer DNA synthesized by Pol III during the chase period is in blue. (B) DNA synthesis terminus even at low concentrations of Pol IV, but when Pol III by Pol II (Left) or Pol IV (Right) without Pol III and cold dNTPs (lanes 1, 5, and is engaged in active DNA synthesis, much higher levels of Pol IV 9), with cold dNTPs (lanes 2, 6, and 10), and with Pol III and cold dNTPs (lanes are required to take over the primed site (14). However, at a 3, 7, and 11). Control experiments where Pol III was added to an unoccupied fork after addition of cold dNTPs are shown in lanes 4, 8, and 12. replication fork the Pol III holoenzyme has several additional connections that may anchor it to DNA and block entry of other DNA polymerases. Specifically, at a replication fork Pol III dynamic fashion. If the 2 polymerases did not repeatedly trade connects to 2 ␤-clamps on both the leading and lagging strands places at the fork, 2 distinct products would be observed that through its twin polymerase structure. Furthermore, Pol III ␶ correspond to the rates of Pol II- and Pol IV-driven replication holoenzyme also connects to the DnaB helicase via the -subunit forks. We confirmed that Pol II acts in a distributive manner at (26). These Pol III-specific connections may prevent takeover of the replication fork by showing that the speed of Pol II repli- a Pol III replisome by Pol II or Pol IV. somes depends on the Pol II concentration (Fig. S3). Based on the results of Figs. 1 and 2, one would predict that if Pol II or Pol IV can recruit the ␤-clamp from a Pol III Pol III Switches with Pol II and Pol IV to Produce a Rapid Replisome. replisome, the result would be a functional, but slow, alternative The experiments of Fig. 1 indicate that DnaB is a component of replisome. To examine the ability of TLS polymerases to take the Pol II and Pol IV replisomes and presumably stimulates over a moving replication fork, increasing concentrations of Pol II and Pol IV were added to a moving Pol III-based replisome synthesis by unwinding DNA. However, DnaB is not known to in the minicircle replication fork assay along with [␣32P]dTTP function with other DNA polymerases, nor is DnaB known to (leading strand) or [␣32P]dATP (lagging strand), as illustrated in unwind DNA as slowly as observed in these alternative repli- Fig. 3A. Timed aliquots were collected and analyzed in an somes. Hence, it remained possible that DnaB helicase is acting alkaline agarose gel. The first 4 lanes in Fig. 3B show the long in some other capacity to stimulate synthesis on the minicircle leading-strand products typically formed by the fast-moving Pol replication fork substrate. If DnaB is truly functional and III replisome, as reported (20). Reactions were then analyzed in encircles DNA in the alternative replisomes, then Pol III should the presence of increasing amounts of Pol II (Fig. 3B Left, lanes be capable of displacing the TLS polymerase with resumption of 5–20) or Pol IV (Fig. 3B Right, lanes 5–20). The results indicate the rapid unwinding characteristic of the coupled DnaB–Pol III that both Pol II and Pol IV interfere with the accumulation of replisome. To examine this issue and provide further evidence long leading-strand products generated by Pol III in a dose- that replisomes containing Pol II or Pol IV are fully functional dependent manner. Inspection of the gel clearly shows that Pol and have not broken down, we performed a series of pulse–chase IV is more efficient at inhibiting the replisome than Pol II; experiments by adding Pol III to a prelabeled replication fork rolling circle replication was reduced Ϸ50% upon addition of 0.5 containing a moving Pol II or Pol IV (see Fig. 2A). Pol II or Pol pmol of Pol IV (Fig. 3B Right, lanes 9–12), whereas 2 pmol of Pol IV was first assembled with DnaB, the sliding clamp, and clamp II was required for a 40% decrease of DNA synthesis (Fig. 3B loader in a 5-min preincubation, and then DNA synthesis was Left, lanes 13–16). Lagging-strand replication was reduced to the initiated for 7 min (Pol II) or 15 min (Pol IV) in the presence of same extent as leading-strand synthesis (Fig. S4). [␣32P] dTTP (pulse). After this labeling reaction, a large excess All 5 E. coli polymerases function with the clamp and bind ␤ of unlabeled dNTPs (chase) was added to prevent incorporation through a conserved consensus sequence (27). Pol II and Pol IV of additional [␣32P] dTTP. Then, Pol III holoenzyme lacking the bind to ␤ via the C-terminal 5 or 6 amino acid residues,

Indiani et al. PNAS Early Edition ͉ 3of8 Downloaded by guest on September 25, 2021 Fig. 4. Pol II and Pol IV slow down the replication fork. (A) The Pol III-based replisome was assembled as described in Fig. 3A. DNA synthesized by Pol III is in blue, and DNA synthesized by the TLS polymerase is in red. (B) After 25 s (0 time point, lanes 1 and 9) a high concentration of Pol II (27 pmol; Left)orPol IV (35 pmol; Right) was added, and aliquots were collected (lanes 2–8). Time courses of Pol III alone (lanes 9–16) are shown for comparison. The dotted red line highlights the difference in DNA chain length at 120 s with and without added Pol II.

Fig. 3. Pol II and Pol IV takeover of Pol III-based replisomes. (A) The replisome III*. Because Pol III is poorly processive in the absence of ␶,we is assembled on DNA as in Fig. 1, except Pol III* is used instead of Pol II or Pol ␶ IV. After 10 s, different amounts of TLS polymerase (pink) are added together were not able to determine whether plays a role in the takeover with [␣32P]dTTP to label leading-strand DNA. Newly synthesized DNA is in blue of a Pol III replisome by Pol II or Pol IV. We did, however, (lagging) and red (leading). (B) Time courses of Pol III replication in the determine that ␶ is not required for rolling circle replication by presence of 0 pmol (lanes 1–4), 0.1 pmol (lanes 5–8), 0.5 pmol (lanes 9–12), 2 Pol II (Fig. S5), suggesting that the role played by ␶ in stimulating pmol (lanes 13–16) and 5 pmol (lanes 17–20) of Pol II (Left) or Pol IV (Right). DnaB-dependent replication might be unique to Pol III, perhaps Timed aliquots of leading strands were analyzed in alkaline agarose gels by making Pol III more processive in the context of the repli- (lagging strand analysis is in Fig. S4). (C) Five picomoles of WT Pol II or Pol IV ⌬ cation fork rather than by directly stimulating unwinding by (lanes 7–12) or Pol II/IV C-term (lanes 13–18) were added to a moving Pol III DnaB. on a minicircle substrate. Leading-strand synthesis is shown (Pol III only is in lanes 1–6). Pol II and Pol IV Remodel the Pol III-Based Replisome and Slow Replication Fork Advance. The results of Fig. 3 indicate that when respectively (24, 28). To determine whether Pol II and Pol IV either Pol II or Pol IV is added to a Pol III-based replisome, the function with ␤ in takeover of Pol III-based replisomes, we replisome is remodeled to include the TLS polymerase, leading expressed and purified mutant versions of Pol II and Pol IV that to a slower moving fork that does not break down but remains lack their respective C-terminal ␤-binding residues and tested intact. Nonetheless, it remains possible that Pol II and Pol IV them in the minicircle replication assay outlined in Fig. 3A. For simply inhibit the replication reaction without taking control of these experiments, 5 pmol of either WT or mutant Pol II or Pol the replication fork or stop the fork and cause it to break down. IV was added 10 s after initiation of replication by Pol III. As Hence, we designed the experiments illustrated in the scheme of expected for ␤-dependent function, the C-terminal mutant Pol Fig. 4A to determine whether the replication fork continues to II and Pol IV were no longer capable of inhibiting Pol III-based slowly advance after the addition of Pol II or Pol IV. Because Pol replisomes in the minicircle replication assay (compare WT III-based replisomes quickly produce replication products that lanes 7–12 vs. ⌬C-term lanes 13–18 for Pol II, Fig. 3C Left, and approach the resolution limit of alkaline gels, we examined very Pol IV, Fig. 3C Right). These results support the conclusion that short time points after addition of the TLS polymerase. In the Pol II and Pol IV take control of the replisome from Pol III by case of Pol II, it is clear that the fork does not break down; fork competitive binding to the clamp and slow the fork. As a control, progression continues but at a slower rate (Fig. 4B Left). Pol IV we determined that the mutant Pol II and Pol IV retained Ͼ70% appears to stop the replication fork (Fig. 4B Right), but consid- of the activity of the corresponding WT polymerase in gap-filling ering that the Pol IV fork only moves 1 ntd/s it would only extend assays that do not require ␤. DNA by 120 nt at the final time point of this experiment. This Pol III is connected to the DnaB helicase at the replication short extension cannot be detected considering that the Pol III fork by the ␶-subunit of the clamp loader, and this connection is replisome has already moved nearly 10 kb by the time Pol IV is essential for rapid, processive rolling circle replication by Pol added. Based on the results of Fig. 1C, showing that Pol IV

4of8 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0901403106 Indiani et al. Downloaded by guest on September 25, 2021 of the 2 specialized polymerases with the ␤-clamp, genes en- coding Pol II and Pol IV mutants that lack the C-terminal ␤-interacting residues were cloned into the arabinose inducible INAUGURAL ARTICLE plasmid and tested for [3H] thymidine incorporation in the presence and absence of arabinose. The mutant polymerases were found to have no measurable effect on DNA synthesis. The results in Fig. 5 D and E indicate that Pol II and Pol IV interact with ␤ to reduce intracellular replication, consistent with the in vitro results of Fig. 3C. Although we were unable to distinguish whether DNA syn- thesis in vivo was slowed or stopped by overexpression of Pol II and Pol IV, the results of Fig. 5 are remarkably consistent with the in vitro experiments of Fig. 3, suggesting that the reconsti- tuted rolling circle replication system can be mimicked by Pol II and Pol IV overexpression in living cells. Despite the good agreement of the in vivo and in vitro data, we cannot eliminate the possibility that differences in sulA inhibition of cell growth could reflect a differential induction of the SOS response when WT Pol II or Pol IV are overexpressed, and thereby might account, at least in part, for the differences in DNA synthesis (Fig. 5 B and C). It is interesting to note that UV irradiation of Pol II mutant cells delays recovery of replication restart (29), which may be caused, in part, by more efficient takeover by Pol IV in the absence of Pol II (i.e., Pol IV is the more potent of the 2 polymerases in slowing the fork). However, a subsequent study did not observe a defect in the rate of recovery after UV irradiation of either Pol II or Pol IV mutant cells (30), suggesting

that the rate of recovery depends on the kinetics of the repair BIOCHEMISTRY processes themselves and not on whether replication proceeds slowly while repair is occurring behind the fork. Fig. 5. Expression of Pol II and Pol IV in vivo reduces chromosomal DNA synthesis. The rate of DNA synthesis is expressed as incorporation of [3H]thy- Discussion midine over a 3-min interval starting 45 min after overexpression of the TLS This article demonstrates that E. coli translesion DNA poly- polymerase by the addition of arabinose. (A) Scheme of the experiment as detailed in Experimental Procedures.(B) Arabinose-induced expression of Pol merases Pol II and Pol IV form alternative replisomes with II WT. (C) Pol IV WT. (D) Pol II ⌬C-term. (E) Pol IV ⌬C-term. Pol II ⌬C-term and DnaB helicase (Fig. 1) and maintain the architecture of the Pol IV ⌬C-term are impaired in their ability to bind the clamp. For B–E, 3H replication machinery for eventual resumption of synthesis by incorporation was adjusted to account for different cell densities of unin- Pol III (Fig. 2). Whether overexpressed in the cell or studied in duced and induced cultures. vitro, high levels of Pol II and Pol IV clearly interfere with ongoing processive synthesis by Pol III (Figs. 3 and 5), by slowing rather than stopping the replisome (Fig. 4), and this inhibition of functions with DnaB, and Fig. 2B, showing that Pol IV repli- Pol III replication depends on the interaction of Pol II and Pol somes are competent for resumption of synthesis by Pol III*, we IV with the ␤-clamp both in vitro (Fig. 3) and in vivo (Fig. 5). presume that takeover of the fork by Pol IV causes the fork to On the basis of these results, we conclude that high levels of Pol slow dramatically rather than break down. II and Pol IV recruit both the ␤-clamp and DnaB helicase from Pol III at a moving replication fork during the SOS response. Induction of Pol II and Pol IV in the Cell Reduces Chromosome Surprisingly, these TLS polymerases modulate the rate of the Replication in the Absence of DNA Damage. The in vitro reactions DnaB helicase and slow the replication fork without causing it to of Figs. 3 and 4 predict that high levels of Pol II and/or Pol IV, break down. as found in the early stages of the SOS response, take over the replisome and slow down the rate of chromosome replication in DnaB as a Regulated Motor. DnaB is not known to function with vivo. To test this prediction, the genes encoding Pol II and Pol other polymerases besides Pol III, yet Pol II and Pol IV slow the IV were cloned into a plasmid under control of the inducible rate of DnaB unwinding, suggesting they may connect to DnaB arabinose promoter, and the incorporation of [3H] thymidine and regulate helicase speed. A reinterpretation of previous during replication was measured before and after induction (see studies suggests another explanation. Specifically, the rate of Fig. 5A). The amounts of arabinose used for induction yielded DnaB unwinding may simply be determined by the rate of any 1,200 molecules of Pol II per cell and 2,500 molecules of Pol IV DNA polymerase that acts on the unwound ssDNA product as per cell, within 2- to 3-fold the levels present in the cell during illustrated in bacteriophage T4 and T7 systems (43, 44). The 35 the SOS response (9, 10). The results of Fig. 5B indicate that bp/s rate of DnaB unwinding observed in an earlier study was induction of Pol II is accompanied by a 55% decrease in DNA determined by monitoring chain extension by Pol III in the replication, and induction of Pol IV decreases DNA synthesis by absence of the ␶-subunit (3). Because ␶, which binds DnaB, was Ϸ99% (Fig. 5C). The more severe decrease in DNA synthesis not present it was thought that the observed rate of unwinding upon overexpression of Pol IV compared with Pol II correlates was the intrinsic rate of DnaB. Instead, the observed unwinding well with the results of the in vitro assays of Fig. 3. Control rate may represent the rate of DnaB unwinding when function- experiments using the same plasmid, but without the Pol II or Pol ing with Pol III core. The same study also observed a 35 bp/s rate IV genes, showed that the vector does not interfere with by the primosome, which contains DnaB and other proteins, chromosomal replication. including DnA helicase (3). The current study shows that Pol II To test whether the replication inhibition observed upon and Pol IV, which are slower than Pol III, function with DnaB expression of Pol II and Pol IV in vivo depends on interaction to move replication forks at rates of 10 and 1 bp/s, respectively.

Indiani et al. PNAS Early Edition ͉ 5of8 Downloaded by guest on September 25, 2021 bound helicase and sliding clamps. We presume the average residence time of any particular polymerase at the fork is defined by the relative amounts of the different DNA polymerases in the cell, their processivity, and their affinity for the sliding clamp and helicase. Likewise, the speed of the replication fork will be a function of the residence time of each DNA polymerase and its intrinsic rate of DNA polymerization. As the intracellular con- centration of TLS polymerases decreases, the population time of Pol III at the replication fork will increase, with a corresponding increase in replication fork speed. What would be the purpose of establishing slower replication forks in the face of DNA damage? We propose an additional role for SOS-induced Pol II and Pol IV besides their ability to bypass lesions. Specifically that TLS polymerases slow the fork, which may give more time for repair of DNA damage by the excision repair system, which can only function on dsDNA before a ␤ Fig. 6. TLS polymerases function with DnaB and clamps. (Left) Scheme of replication fork collides with a lesion. Of course, sometimes the Pol III-based replisome. The ␶-subunit of Pol III binds the DnaB helicase and enables rapid unwinding. Two ␶-subunits are contained within a clamp loader replisome will encounter a lesion, and TLS synthesis will bypass assembly (the clamp loader is not illustrated) and thereby dimerize 2 Pol III it, producing a mutation. Nevertheless, slower moving forks molecules for coupled leading- and lagging-strand synthesis. (Right) TLS should lower the frequency of these events. Pol II and Pol IV are polymerases trade places with one another on the ␤-clamp with Pol III to form induced early in the SOS damage response and therefore a TLS replisome. The TLS polymerases function with DnaB helicase, which replication fork control by these polymerases is expected to be slows to match their slow rate of synthesis. Leading and lagging strands may specific to the first phase (Ͻ45 min) of SOS induction before the be uncoupled in the TLS replisome. highly mutagenic Pol V is produced (4). If the burden of DNA damage is high, prolonging the SOS response, the cell eventually up-regulates production of Pol V to effectuate a salvage program If the intrinsic rate of DnaB unwinding is truly 35 bp/s, the to deal with unrepaired lesions. current study indicates that Pol II and Pol IV slow DnaB. The highest rate of DnaB unwinding is observed by using Pol III ␶ Ͼ SOS Response Can Be Activated by ssDNA Gaps Without Fork Break- holoenzyme containing ( 500 bp/s). down. It is widely accepted that the presence of ssDNA is the In light of these 4 different rates of unwinding, ranging Ͼ trigger for activation of the SOS response by RecA (31). 500-fold, we propose that the DnaB unwinding rate is coupled Research showing that replication is necessary for SOS induction to DNA synthetic rate, even in the absence of a direct coupling has been interpreted to suggest that one fork must stall or break ␶ factor like . This observation suggests that DnaB acts as a down, uncoupling helicase unwinding from DNA synthesis, to ‘‘regulated motor’’ that must be coupled to some other activity, generate ssDNA (19). However, cellular studies have shown that like DNA polymerization, to promote efficient unwinding (43, DNA damage leads to single-strand gaps in both leading and 44). A replicative helicase acting as a regulated motor will ensure lagging strands, and therefore replication forks skip over lesions, that unwinding is tightly coordinated with DNA synthesis, a leaving ssDNA gaps behind them (32–35). These gaps may then favorable property for replisome action. Otherwise, ssDNA will serve as substrate for RecA binding and consequent induction of be generated that, besides being vulnerable to nucleases and the SOS response. strand breaks, leads to the SOS response. Gaps in the lagging strand are made possible by the numerous priming events that occur during discontinuous synthesis. Spe- Logic of Slower Alternative Replisomes in the Face of DNA Damage. cifically, when a polymerase extending an Okazaki fragment The concept of switching between Pol III and TLS polymerases stalls at a lesion, it can transfer from the incomplete fragment to has been considered an ephemeral phenomenon, where a stalled a new RNA primer, leaving behind a ssDNA gap. Recent studies replicative polymerase briefly yields to the TLS polymerase only demonstrate repriming of DNA synthesis on the leading strand for bypass of the lesion, and Pol III immediately retakes control as well, and therefore the same process can explain leading of the growing DNA end to resume processive replication. The strand gaps (32, 36). Formation of ssDNA gaps behind the results shown here compel a reconsideration of this notion, replication fork offers an alternative means to generate ssDNA because high levels of Pol II and Pol IV, as found in the early for SOS induction that does not require fork stalling or break- stages of the SOS response, clearly take over replication by Pol down. The ability to skip over damage implies that replication III to form slower replisomes that function not only with the forks do not necessarily stall at every leading-strand lesion, but clamp but also with the replicative DnaB helicase. Unlike Pol III if Pol III were to continue rapid synthesis in the presence of replicase that contains 2 Pol III cores connected to a clamp significant DNA damage, encounters with additional lesions, loader, Pol II and Pol IV are not known to dimerize or connect and the risk of stalling or collapse would increase. By slowing the to other proteins besides ␤. Hence, we presume that leading- and fork, TLS polymerases diminish this risk while also assuring that lagging-strand synthesis in TLS replisomes are constitutively the fork does not break down. It is important to note that when uncoupled, with no need to hypothesize a DNA trombone loop the lesion is a nick, replication forks will collapse, explaining the (illustrated in Fig. 6). We propose that ␤-clamps are assembled need for primosomal replication restart proteins. on lagging strand RNA primers by the ␥-complex, and perhaps it may be the reason for production of both ␥ and ␶ in cells. How Are Mutations Prevented If Error-Prone TLS Replisomes Are Nevertheless, it remains possible that TLS polymerases dimerize Long-Lived? Pol II and Pol IV replicate DNA with lower fidelity or connect to other proteins in the context of a replication fork. than Pol III, and therefore an obvious concern about alternative The Pol II and Pol IV replisomes are highly stable and TLS replisomes is the increased risk of introducing heritable therefore are probably the active form during the entire SOS mutations, undermining rather than enhancing genome integ- response. These replisomes are also highly dynamic. We dem- rity. However, the cell contains several safeguards that prevent onstrate here that all 3 DNA polymerases (Pols II, III, and IV) production of heritable mutations under these circumstances. freely exchange with one another while preserving the DNA- First, errors produced by a TLS enzyme operating on undam-

6of8 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0901403106 Indiani et al. Downloaded by guest on September 25, 2021 aged DNA result in simple mismatches that can be corrected by is 20 mM Tris⅐Cl (pH 7.5), 0.1 mM EDTA, 5 mM DTT, 5% glycerol, 40 ␮g/ml BSA, the postreplication mismatch repair system. Second, TLS DNA 0.5 mM ATP, and 8 mM MgCl2. polymerases are very slow, thereby ensuring that mismatches INAUGURAL ARTICLE caused by an inherently low-fidelity polymerase are not pro- Rolling Circle Replication Reactions. DnaB was assembled on the minicircle duced often. For example Pol IV, which makes a mistake DNA by incubating 100 fmol of DNA with 4 pmol of DnaB (as hexamer) in 15 ␮L of buffer A for 30 s at 37 °C, at which time Pol III* (100 fmol), ␤ (350 fmol), approximately every 1,000 nt, moving at a speed of 1 ntd/s will and 60 ␮M each of dGTP and dCTP were added for 5 min. DNA synthesis was make only about 4 mismatches in 1 h. Correcting these errors is initiated upon adding 1 ␮g of SSB, 50 ␮M of the 4 rNTPs, 60 ␮M DnaG primase, a relatively small task for the mismatch repair system. Third, Pol 60 ␮M dATP, and 20 ␮M dTTP (leading strand) or 60 ␮M dTTP and 20 ␮M dATP II is a high-fidelity enzyme because it contains a proofreading (lagging strand) in 25 ␮L. DNA synthesis was monitored by using [␣32P]dTTP for exonuclease. Finally, Pol IV is held in a high-fidelity state in the the leading strand and [␣32P] dATP for the lagging strand (3,000–5,000 early phase of the SOS response by interaction with full-length cpm/pmol). Pol II or Pol IV were added as indicated in the figure legends. UmuD and RecA (6, 37). When the SOS response is prolonged, Reactions lacking Pol III* contained 5 pmol of Pol IV (or Pol II), 50 fmol of 2 ␶ ␤ indicating the presence of significant or unrepairable damage, -complex, and 350 fmol of (Fig. 1). All reactions were quenched upon the addition of 0.5% SDS and 20 mM EDTA, and DNA synthesis was quantitated as UmuD2 is cleaved to a form that no longer modulates the fidelity described (40). Products were analyzed on 0.6% alkaline agarose gels. of Pol IV interacts with UmuC to form the low-fidelity Pol V TLS polymerase, thereby defining the later, mutagenic phase of Measurement of DNA Synthesis After Protein Overexpression. The protocol the SOS response. Regulation of the fidelity of Pol IV by UmuD used [3H]thymidine to monitor DNA synthesis after overexpression of differ- also explains why overexpression of Pol IV alone was found to ent DNA polymerases as described (41). Dose-dependent expression of Pol II or be mutagenic whereas co-overexpression of Pol IV along with an Pol IV was obtained by using the araC-pBAD system (Invitrogen). The plasmid uncleavable form of UmuD, UmuD(S60A), was not (37, 38). was transformed into E. coli strain BW27786, which overexpresses the arabi- One may presume that during a catastrophic and prolonged SOS nose permease, araE, allowing consistent and highly regulated expression of response, bypass of lesions by Pol V is of utmost importance, pBAD in a population of cells (42). A single colony was inoculated in 5 mL of even at the cost of forming heritable mutations. LB, overnight cultures were diluted 1:100 in 100 mL of LB, then grown to OD600 ϭ 0.1. 45 min after adding arabinose (0.1% for Pol II and 0.01% for Pol IV), and Experimental Procedures 0.5-ml aliquots were removed and added to a tube containing pulse-label solution {30 ␮L of [methyl-3H] thymidine (1mCi/mL) in 2.97 mL of LB}. After ␣ ␧ ␪ ␥ ␶ ␦ ␦Ј ␹ ␺ ␤ Materials. Subunits of Pol III holoenzyme ( , , , , , , , , , and ) were exactly 3 min at 37 °C, ice-cold 5% trichloroacetic acid was added to lyse cells purified as described (20). DnaB, DnaG, and SSB were purified as described and precipitate DNA. Pol II and Pol IV were quantitated by Western blot

␤ BIOCHEMISTRY (39). Pol III* (all holoenzyme subunits except ) was reconstituted as described analysis. Control samples (ϪAra) were from parallel cultures not treated with ⌬ (14). The C-terminal 5-residue deletion of Pol II (polB 779–783) and 6-residue arabinose. deletion of Pol IV (dinB ⌬346–351) containing a His6 tag were cloned into the 2ϩ pET16 expression vector and purified by using Ni affinity chromatography; ACKNOWLEDGMENTS. We thank Jay Keasling for the AraE overexpression Ͼ their activity was nearly that of WT Pol II and Pol IV ( 70%). The tailed form strain. This work was supported by National Institutes of Health Grants II duplex minicircle DNA was as described (20). The BW27786 strain was a GM38839 (to M.O.), AI065508 (to M.O.), R37GM21422 (to M.F.G.), and ESO generous gift from Jay Keasling (University of California, Berkeley). Buffer A 12259 (to M.F.G.).

1. Kornberg A, Baker TA (1992) DNA replication. DNA Replication (Freeman Press, New 16. Furukohri A, Goodman MF, Maki H (2008) A dynamic polymerase exchange with York), 2nd ed, p 931. Escherichia coli DNA polymerase IV replacing DNA polymerase III on the sliding clamp. 2. Johnson A, O’Donnell M (2005) Cellular DNA replicases: Components and dynamics at J Biol Chem 283:11260–11269. the replication fork. Annu Rev Biochem 74:283–315. 17. Fujii S, Fuchs RP (2004) Defining the position of the switches between replicative and 3. Kim S, Dallmann HG, McHenry CS, Marians KJ (1996) Coupling of a replicative poly- bypass DNA polymerases. EMBO J 23:4342–4352. merase and helicase: A tau–DnaB interaction mediates rapid replication fork move- 18. Higuchi K, et al. (2003) Fate of DNA replication fork encountering a single DNA lesion ment. Cell 84:643–650. during oriC plasmid DNA replication in vitro. Genes Cells 8:437–449. 4. Courcelle J, Khodursky A, Peter B, Brown PO, Hanawalt PC (2001) Comparative gene 19. Pages V, Fuchs RP (2003) Uncoupling of leading- and lagging-strand DNA replication expression profiles following UV exposure in wild-type and SOS-deficient Escherichia during lesion bypass in vivo. Science 300:1300–1303. coli. Genetics 158:41–64. 20. McInerney P, O’Donnell M (2004) Functional uncoupling of twin polymerases: Mech- 5. Napolitano R, Janel-Bintz R, Wagner J, Fuchs RP (2000) All three SOS-inducible DNA anism of polymerase dissociation from a lagging-strand block. J Biol Chem 279:21543– polymerases (Pol II, Pol IV, and Pol V) are involved in induced mutagenesis. EMBO J 21551. 19:6259–6265. 21. Uchida K, et al. (2008) Overproduction of Escherichia coli DNA polymerase DinB (Pol IV) 6. Kobayashi S, Valentine MR, Pham P, O’Donnell M, Goodman MF (2002) Fidelity of inhibits replication fork progression and is lethal. Mol Microbiol 70:608–622. Escherichia coli DNA polymerase IV. Preferential generation of small deletion muta- 22. Duigou S, Ehrlich SD, Noirot P, Noirot-Gros MF (2004) Distinctive genetic features tions by dNTP-stabilized misalignment. J Biol Chem 277:34198–34207. exhibited by the Y-family DNA polymerases in Bacillus subtilis. Mol Microbiol 54:439– 451. 7. Tang M, et al. (2000) Roles of E. coli DNA polymerases IV and V in lesion-targeted and 23. Lee J, Chastain PD, 2nd, Kusakabe T, Griffith JD, Richardson CC (1998) Coordinated untargeted SOS mutagenesis. Nature 404:1014–1018. leading and lagging strand DNA synthesis on a minicircular template. Mol Cell 1:1001– 8. Cai H, Yu H, McEntee K, Kunkel TA, Goodman MF (1995) Purification and properties of 1010. wild-type and exonuclease-deficient DNA polymerase II from Escherichia coli. J Biol 24. Bonner CA, et al. (1992) Processive DNA synthesis by DNA polymerase II mediated by Chem 270:15327–15335. DNA polymerase III accessory proteins. J Biol Chem 267:11431–11438. 9. Bonner CA, Hays S, McEntee K, Goodman MF (1990) DNA polymerase II is encoded by 25. Wagner J, Fujii S, Gruz P, Nohmi T, Fuchs RP (2000) The ␤-clamp targets DNA polymerase the DNA damage-inducible dinA gene of Escherichia coli. Proc Natl Acad Sci USA IV to DNA and strongly increases its processivity. EMBO Rep 1:484–488. 87:7663–7667. 26. Gao D, McHenry CS (2001) tau binds and organizes Escherichia coli replication proteins 10. Kim SR, Matsui K, Yamada M, Gruz P, Nohmi T (2001) Roles of chromosomal and through distinct domains. Domain IV, located within the unique C terminus of tau, episomal dinB genes encoding DNA pol IV in targeted and untargeted mutagenesis in binds the replication fork, helicase, DnaB. J Biol Chem 276:4441–4446. Escherichia coli. Mol Genet Genomics 266:207–215. 27. Dalrymple BP, Kongsuwan K, Wijffels G, Dixon NE, Jennings PA (2001) A universal 11. Qiu Z, Goodman MF (1997) The Escherichia coli polB locus is identical to dinA, the protein–protein interaction motif in the eubacterial DNA replication and repair sys- structural gene for DNA polymerase II. Characterization of Pol II purified from a polB tems. Proc Natl Acad Sci USA 98:11627–11632. mutant. J Biol Chem 272:8611–8617. 28. Lenne-Samuel N, Wagner J, Etienne H, Fuchs RP (2002) The processivity factor ␤ controls 12. Woodgate R, Ennis DG (1991) Levels of chromosomally encoded Umu proteins and DNA polymerase IV traffic during spontaneous mutagenesis and translesion synthesis requirements for in vivo UmuD cleavage. Mol Gen Genet 229:10–16. in vivo. EMBO Rep 3:45–49. 13. Woodgate R, Levine AS (1996) Damage inducible mutagenesis: Recent insights into the 29. Rangarajan S, Woodgate R, Goodman MF (1999) A phenotype for enigmatic DNA activities of the Umu family of mutagenesis proteins. Cancer Surv 28:117–140. polymerase II: A pivotal role for pol II in replication restart in UV-irradiated Escherichia 14. Indiani C, McInerney P, Georgescu R, Goodman MF, O’Donnell M (2005) A sliding-clamp coli. Proc Natl Acad Sci USA 96:9224–9229. toolbelt binds high- and low-fidelity DNA polymerases simultaneously. Mol Cell 30. Courcelle CT, Belle JJ, Courcelle J (2005) Nucleotide excision repair or polymerase 19:805–815. V-mediated lesion bypass can act to restore UV-arrested replication forks in Escherichia 15. Yang J, Zhuang Z, Roccasecca RM, Trakselis MA, Benkovic SJ (2004) The dynamic coli. J Bacteriol 187:6953–6961. processivity of the T4 DNA polymerase during replication. Proc Natl Acad Sci USA 31. Little JW, Mount DW, Yanisch-Perron CR (1981) Purified lexA protein is a repressor of 101:8289–8294. the recA and lexA genes. Proc Natl Acad Sci USA 78:4199–4203.

Indiani et al. PNAS Early Edition ͉ 7of8 Downloaded by guest on September 25, 2021 32. Heller RC, Marians KJ (2006) Replication fork reactivation downstream of a blocked 39. Yuzhakov A, Turner J, O’Donnell M (1996) Replisome assembly reveals the basis nascent leading strand. Nature 439:557–562. for asymmetric function in leading and lagging strand replication. Cell 86:877– 33. Rupp WD, Howard-Flanders P (1968) Discontinuities in the DNA synthesized in an excision- 886. defective strain of Escherichia coli following ultraviolet irradiation. J Mol Biol 31:291–304. 40. Stukenberg PT, Studwell-Vaughan PS, O’Donnell M (1991) Mechanism of the sliding 34. Sassanfar M, Roberts JW (1990) Nature of the SOS-inducing signal in Escherichia coli. ␤-clamp of DNA polymerase III holoenzyme. J Biol Chem 266:11328–11334. The involvement of DNA replication. J Mol Biol 212:79–96. 41. Courcelle CT, Courcelle J (2006) Monitoring DNA replication following UV-induced 35. Amado L, Kuzminov A (2006) The replication intermediates in Escherichia coli are not damage in Escherichia coli. Methods Enzymol 409:425–441. the product of DNA processing or uracil excision. J Biol Chem 281:22635–22646. 42. Khlebnikov A, Datsenko KA, Skaug T, Wanner BL, Keasling JD (2001) Homogen- 36. Pomerantz RT, O’Donnell M (2008) The replisome uses mRNA as a primer after colliding eous expression of the P(BAD) promoter in Escherichia coli by constitutive expres- with RNA polymerase. Nature 456:762–766. sion of the low-affinity high-capacity AraE transporter. Microbiology 147:3241– 37. Godoy VG, et al. (2007) UmuD and RecA directly modulate the mutagenic potential of 3247. the Y family DNA polymerase DinB. Mol Cell 28:1058–1070. 43. Delagoutte E, von Hippel PH (2001) Molecular mechanisms of the functional coupling 38. Kim SR, et al. (1997) Multiple pathways for SOS-induced mutagenesis in Escherichia coli:An of the helicase (gp 41) and polymerase (gp 43) of bacteriophage T4 within the DNA overexpression of dinB/dinP results in strongly enhancing mutagenesis in the absence of replication fork. Biochemistry 40:4459–4477. any exogenous treatment to damage DNA. Proc Natl Acad Sci USA 94:13792–13797. 44. Stan NM, et al. (2005) Nature 435:370–373.

8of8 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0901403106 Indiani et al. Downloaded by guest on September 25, 2021