Transcription Preferentially Inhibits Nucleotide Excision Repair of the Template DNA Strand in Vitro*

experiments with a superhelical plasmid no enhanced repair of the transcribed gene was observed compared to nontranscribed regions. We conclude that RNA polymerase stalled at a photodimer does not direct the (A)BC excinuclease to the damaged template strand and therefore cannot account for the strand-specific repair observed

It has been reported that pyrimidine dimers (pyrimidineopyrimidine) are removed preferentially from actively transcribing genes. Furthermore, the preferential repair is restricted to the transcribed strand of these genes. Currently there is no mechanistic explanation for these phenomena.
In this study we investigated the effect of transcription on nucleotide excision repair using defined  Bohr et al. (1985) discovered that pyrimidine dimers in the dihydrofolate reductase gene of Chinese hamster ovary cells were removed four to five times more rapidly compared to nontranscribed regions of the chromosome. Mellon et al. (1987) demonstrated that this enhanced repair was due almost exclusively to the preferential repair of the transcribed strand. More recently Terleth et al. (1989) and Mellon and Hanawalt (1989) showed preferential repair in transcribed genes of S. cerevisiae and Escherichia coli, respectively. The latter authors measured the rate of repair of pyrimidine dimers in the two strands of the luc operon following UV (254 nm) irradiation. They found that when the operon was actively transcribed, the photodimers were removed five to 10 times more rapidly from the transcribed strand compared to the nontranscribed strand, no difference was found in the rate of repair of the two strands in uninduced cells.
Initially, preferential repair of transcribed genes was ascribed to a more open conformation of the transcribed regions of chromatin (see Bohr et al., 1988). However, the discovery that preferential repair is limited to the transcribed strand suggests a more specific coupling mechanism between transcription and repair. A simple model for this coupling proposed by Mellon and Hanawalt (1989) is as follows. RNA polymerase transcribes a gene until it reaches a UV photoproduct; transcription then halts, the RNA polymerase makes a stable elongation complex at the transcriptional stopsite, and this complex constitutes a high affinity site for the nucleotide excision repair proteins which remove the photodimer. A precedent for the targeting of UV photoproducts in DNA for repair by a lesion-bound protein exists. Photolyase, which binds to the pyrimidine dimer-containing strand of the duplex (Husain et al., 1987), stimulates the excision of the photodimer by (A)BC excinuclease, the enzyme that initiates nucleotide excision repair in E. coli (Sancar et al., 1984;Sancar and Sancar, 1988;Orren and Sancar, 1989).
The above model is an oversimplification of what takes place in an E. coli cell as it neglects the topological effects of transcription (Gamper and Hearst, 1982;Tsao et al., 1989), transcription factors, the effect of transcription-coupled translation and other phenomena as possible contributors in directing repair enzymes to the transcribed strand. Nevertheless, the simple model outlined above is testable in uitro. In this study we have used either a 137-mer duplex containing a single thymine photodimer located on the nontranscribed or the transcribed strand downstream from a tat promoter, or a UV-irradiated superhelical plasmid containing a tat promoter in a coupled transcription/repair system which includes E. coli RNAP,' (A)BC excinuclease, and DNA polymerase I and ligase when necessary to study the effect of transcription upon repair. We find that under these in vitro conditions, the template strand is repaired less efficiently than the nontranscribed strand.  Sancar et al. (1964)  were purified by the method of Thomas et al. (1985). Two types of substrates were used. One was a 137-mer duplex which contained a tuc promoter and a thymine photodimer (T<>T) in either the transcribed or the nontranscribed strand. The 137-mers labeled with "P at the 5' end of the strand with TOT were constructed by ligating the eight individual oligomers shown in Fig. 1 as described previously for a psoralen-adducted substrate (Shi et al., 1987 andVan Houten et al., 1987). The central 11-base oligomer with the T<:>T was prepared and purified by the method of Banerjee et al. (1988). The second type of substrate for transcription-repair experiments was superhelical pDR3274 plasmid irradiated with 75 Jm-' of 254-nm UV light from a Quantacount monochrometer (Photon Technology, Inc., Princeton, NJ). This UV dose produces about 2 UV photoproducts (cyclobutane dimers and 6-4 photoproducts) per kilobase pair (estimated by incision of cyclobutane dimers and assuming 6-4 photoproducts are 10% of total photoproducts). pDR3274 is a pBR328 derivative which contains, in addition to a functional tet gene, the uurC gene under the control of a tuc promoter (Thomas et al., 1985).

Experiments with the Linear Substrate
Transcription-The transcription buffer contained 40 mM Tris-HCl, pH 7.95, 50 mM KCl, 10 mM MgCL, 2 mM ATP, 5 mM dithiothreitol, 50 pg/ml bovine serum albumin, and RNasin at 0.8 unit/pi. Reaction mixtures (12.5 ~1) containing about 0.5-5.0 fmol of end-labeled DNA, RNAP (RNA polymerase), and 1.3 pM [(u-32P]UTP (3 Ci/pmol, Du Pont-New England Nuclear) were incubated at 37 "C for 10 min to form initiation complexes. Then 12.5 ~1 of transcription buffer (at 37 "C) containing GTP, CTP, and UTP (unlabeled) at 400 pM each was added and the-reaction mixtures were incubated at 37 "C for an additional 8 min. The reactions were stopned bv adding EDTA to 10 mM, sodium dodecyl sulfate to 0.2%, LidG to lb0 mM,-and 0.1 mg of oyster glycogen. The nucleic acids were precipitated with ethanol, resuspended in a formamide dye solution, and analyzed on 8% polyacrylamide sequencing gels. The reaction mixtures were incubated at 37 "C for 10 min, CaClz was then added to 5 mM and the DNA was digested with 20 pg of DNase I for 6 min. Cold oyster glycogen (100 rg in 25 ~1) was then added as carrier and the DNA was precipitated with ethanol and analyzed on 8% polyacrylamide sequencing gels. Incision by (A)BC Excinuclease-The incision reactions were conducted in loo-p1 volumes with about 1 fmol of DNA, 4.2 nM UvrA, 63 nM UvrB, and 140 nM UvrC: CTP, GTP, and UTP each at 200 pM; and, when indicated, 1.2-236 nM i. coli photolyase or 0.3-66 nM RNAP. RNAP or nhotolvase were incubated with the DNA for 5-10 min at 37 "C before addition of (A)BC excinuclease unless otherwise indicated. The mixtures were further incubated at 37 "C for 8 or 15 min unless otherwise indicated and then the reactions were stopped and the products were analyzed as under "DNase I Footprinting of RNA Polymerase." RNA Sequencing-Purified truncated transcripts were subjected to sequence-specific degradation using ribonuclease B. cerus and ribonuclease Tl and instructions provided by the supplier. RNA ladders were generated from purified full-length transcripts by heating at 90 "C in 50 mM bicarbonate buffer (pH 9.0) for 15 min to 2 h (Donis-Keller et al., 1977). Reaction products in urea dye loading mixtures were analyzed on 11 and 20% polyacrylamide sequencing gels.

Experiments with Superhelical Substrate
Quantitative Analysis of Transcription-Transcription was carried out in transcription-rep&r buffer -(TRB) which contained 40 mM HEPES, pH 7.8, 50 mM KCI, 7.4 mM MgC12, 2 mM ATP, 3.4% glycerol, 0.9 mM dithiothreitol, and 0.8 unit/p1 RNasin and, when indicated, dNTPs at 40 YM each (extent dCTP at 10 uM) and rNTPs at 200 pM each. For quantitative analysis of transcri$ion, initiation complexes were formed by incubating RNAP with 70 fmol of plasmid and 1.3 pM of unlabeled UTP in 25 ~1 of TRB for 5 min at 37 "C. Then, 25 pl of TRB at 37 "C containing 400 pM each of CTP, GTP, and UTP plus 66 nM [w~'P]UTP was added. The reaction mixtures were incubated at 37 0c and-the reactions were stopped by adding 50 ~1 of cold 10% trichloroacetic acid. The volume was adiusted to 1 ml with 5% trichloroacetic acid, and RNA was collected by centrifugation and two washes with 5% trichloroacetic acid. The RNA was then dissolved with 1 ml of 1 M Tris-HCl, pH 8.0, containing trace amounts of RNase A and quantified by scintillation counting.

Repair
Synthesis-Repair synthesis reactions were conducted in TRB containing 0.4 unit/p1 RNasin. Initiation complexes were formed by incubating DNA (about 140 fmol) with RNA polymerase (13 pmol) and UTP (5 pM) in 50 ~1 at 37 "C! for 5 min. Transcription and repair were started by adding 50 ~1 of a prewarmed solution in TRB to bring final concentrations of CTP, GTP, and UTP to 200 uM each, dATP, dGTP, and TTP to 40 fiM each, dCTP to 10 pM including 2-8 PCi of [a-32P]dCTP, UvrA to 1.4 nM, UvrB to 16 nM, UvrC to 30 nM. PolI to 200 nM. and T4 DNA liease to 20 units/ml. Under these co&l&ions we found the UvrA and piotoproduct concentrations to be limiting factors, i.e. there was an increase in repair synthesis with higher UvrA concentrations or UV doses (data not shown). Reactions were stopped after 8 min at 37 "C by mixing each with a 100 ~1 volume of phenol. The DNA was extracted with phenol and ether and precipitated with ethanol. Following digestion with EcoRI, the DNA was further purified on 0.8% agarose gels, electroeluted from gel slices, precipitated with ethanol, resuspended, dialyzed, digested with BglII and Hind111 and the resulting fragments were separated on 1.2% agarose gels. The gels were stained with ethidium bromide, and the DNA was visualized and photographed on a UV light box. Gels were then dried and autoradiographed. Following autoradiography, the gels were rehydrated, the DNA bands were excised, and the agarose was digested with 1 ml of 1 N perchloric acid at 65 "C for 1-3 h, 0.1 ml of 10 N NaOH was added to neutralize the samples, and the radioactivity was measured by scintillation counting. The relative amounts of DNA recovered were determined by scanning (using a Zeineh Softlaser Scanning Densitometer model SLR-2D/ 1D) photographs of ethidium bromide-stained gels and the amounts of radiolabel incorporation into DNA were normalized for the amount of DNA recovered in each lane. RESULTS We used two types of substrate to investigate the effect of transcription on nucleotide excision repair. One of the substrates was a 137-mer duplex containing a strong tuc promoter and a single thymine photodimer in either the nontranscribed or the transcribed strand ( Fig. 1). Using this substrate we were able to examine effects of template and nontemplate lesions on transcription and consequences for repair. The second substrate was a UV-irradiated superhelical plasmid which contained the tat promoter from which a gene (uruC) constituting approximately one third of the plasmid is transcribed (Thomas et al., 1985). This substrate was used to examine, by the repair synthesis assay, whether the transcribed sequences were repaired more efficiently than nontranscribed sequences and whether transcription had any effect on the rate and extent of repair in the transcribed and nontranscribed regions of the plasmid.
Transcription and Repair of a Linear Substrate Containing a Thymine Dimer at a Defined Site Transcription-It has been known for a long time that pyrimidine dimers block transcription in both pro-and eucaryotic cells (Sauerbier and Hercules, 1978). We constructed a synthetic DNA fragment ("T" in Fig. 1) with a strong promoter (tuc) and a thymine dimer in the transcribed strand to evaluate the efficiency of T<>T as a transcriptional block in uitro. Fig. 2A shows the results of transcription experiments conducted using 137-mers with or without a thymine photodimer. With the dimer-free template transcription continues to the end of the fragment generating transcripts 89 or 90 nucleotides long (see below). When the dimer is present in the transcribed strand transcription terminates at the dimer producing a doublet (see below). Within the sensitivity of our assay (which would detect 0.5-1.0% of the radioactivity of the full length transcripts) we conclude that RNAP did not transcribe past the T<>T.
To learn where transcription actually begins and ends with respect to the photodimer, the bands in the doublet which represent truncated transcripts were indi- vidually excised from the gel, purified, subjected to sequencespecific degradation reactions, and analyzed on an 11% polyacrylamide sequencing gel (Fig. 2B). Both bands from the doublet had the same sequences up to their 3' ends which terminated with a base (see Fig. 2, legend) inserted opposite the 3' T of the photodimer. The length heterogeneity of the truncated (and full length) transcripts must arise from differences at their 5' ends, specifically from initiation at either the 9th or the 10th nucleotide 3' to the -10 sequence. The latter is deduced from the DNA sequence (Fig. l), the fact that the transcripts were end-labeled with UTP (and could not be end-labeled with ATP), and analysis of sequencing reactions on 20% gels (not shown) to obtain the sequence of the 5' ends.
When the T<>T was in the nontranscribed strand it did not cause pausing by RNAP or premature termination of the transcript (Fig. 2A, lane 8). This result is consistent with the finding of Shi et al. (1987) that a psoralen-thymine monoadduct in the nontranscribed strand did not block transcription.
Stalling of RNAP at the Thymine Photodiner-To investigate the formation of stable RNAP-DNA-T-T complexes at the lesion site we conducted DNase I footprinting experiments using the substrate with a T-G-T in the templatestrand. Incubation of RNAP with the substrate in the absence of rNTPs led to the formation of a stable initiation complex as evidenced by a standard RNAP footprint over the promoter region (Fig. 3, lanes 2-5). When rNTPs were included in the reaction mixture the RNAP footprint was no longer at the promoter, rather it was displaced in the 3' direction, covering an area which includes the thymine photodimer (Fig. 3, lanes  6-9). Thus, as was found with a psoralen cross-link in transcribed DNA (Shi et al., 1987), RNAP makes a stable elongation complex at the site of a T-T in the template.
Effect of Transcription on Excision of TOT from the Template and Nontemplate Strands by (A)BC Excinucleose-The data presented so far indicate that a ToT in the transcribed strand blocks transcription resulting in a truncated transcript and a stable elongation complex. Does this complex facilitate the excision of TOT by (A)BC excinuclease? (A)BC excinuclease hydrolyzes primarily the 8th phosphodiester bond 5' and the 5th phosphodiester bond 3' to a T<>T (Sancar and Rupp, 1983). The 137-mers N and T in Fig. 1 (dimer in the nontemplate and template strands, respectively) were labeled with 32P at the 5' end of the damaged strand and digested with ABC excinuclease. The end-labeled digestion product of N runs as a 67-mer on a denaturing gel and digestion of T gives a 56-mer (Fig. 4A). The T substrate is digested with somewhat higher efficiency than N, due to a moderate effect of neighboring sequences on enzyme activity. Incision of both substrates is inhibited at high concentrations of RNAP (Fig.  4A, lanes 8 and 9) due to nonspecific binding of RNAP to DNA (Fig. 3, lanes 4 and 5). When the reaction mixtures contained RNAP plus all four rNTPs, the incision of the transcribed strand was inhibited to a greater extent than incision of the nontranscribed strand. The optimal RNAP concentration (13 nM) for producing an elongation complex footprint (Fig. 3, lanes 4 and 8) drastically inhibited the excision of T<>T from the transcribed strand (Fig. 4A,  compare lanes 8 and 18). The selective inhibition of incision of T<>T in the transcribed strand under transcription conditions was also observed when we measured the kinetics of incision (Fig. 4B). Thus, the data shows that RNAP stalled at a T-T does not stimulate but actually inhibits the excision of the photoproduct by the excision nuclease.
The selective inhibition of incision of the template lesion was unexpected, yet it was consistently observed under conditions in which stimulation of incision by DNA photolyase was readily demonstrated (Fig. 4). Nevertheless we further examined the possible stimulation of incision of the template dimer by transcription. We reasoned that a low concentration of RNA polymerase might have a subtle enhancing effect on the kinetics of incision that we had not detected. However, the results obtained with the T substrate shown in Fig. 5 indicate that 0.7 nM RNAP did not stimulate incision kinetics while photolyase clearly did. Similarly, varied orders of addition of RNAP and (A)BC excinuclease failed to produce stimulation of incision ( Fig. 2A and data not shown). Photolyase as a Possible Coupling Factor-It has been suggested (Patterson and Chu, 1989) that photolyases function as auxiliary proteins which target lesions for repair by nucleotide excision. Therefore, we conducted further experiments to examine whether more stimulation could be achieved when both RNAP and E. coli DNA photolyase were present in the reaction mixture compared to photolyase alone. Addition of photolyase to a preformed elongation complex fails to relieve the inhibitory effect of RNAP and addition of RNAP to photolyase-TOT complexes interferes with the stimulatory In lone 1 the 5' terminally labeled DNA substrate was treated with (A)BC excinuclease. The other lanes contain DNA that was incubated with RNAP for 10 min and then digested with DNase I. Samples in lanes 2-5 contained ATP and UTP whereas the samples in lanes 6-9 contained all four rNTPs. The footprints of RNAP initiation and elongation complexes are bracketed, and the region of possible overlap of the two footprints is indicated with a broken line. No discernible footprint is evident on the nondimer (T-T) template (lanes 10-13); however, there is an overall protection of the area 3' to the promoter presumably because of transcription. The 5' incision site of (A)BC excinuclease is indicated by (A)BC. Transcription and Repair of Supercoiled, UV-irradiated DNA Transcription causes drastic topological changes in DNA (Gamper and Hearst, 1982;Tsao et al., 1989). It is conceivable that these topological features of transcribed DNA would be associated with the enhanced rate of repair observed in vivo. Along these lines Pu et al. (1989) recently reported that the rate of incision by (A)BC excinuclease of N-methylmitomycin cross-linked DNA was about 209fold higher in supercoiled molecules than with relaxed molecules.
To investigate the effect of transcription on repair of superhelical DNA we used pDR3274, which was modified by irradiation with 254-nm light. Repair was measured as the UV-dependent formation of [cr-32P]dGTP-labeled repair patches synthesized by the combined action of ABC excinuclease, DNA PoZI, and DNA ligase. pDR3274 was selected because it contains a very strongly transcribed UvrC gene (Mulligan et al., 1985;Thomas et al., 1985) and the other genes are either inactive (cam) or weakly transcribed (tet, 5' end of the strand containing the central 11-mer) without (lanes 1 and 2) or with TOT (lanes 3-7) was used as substrate for RNAP and/or (A)BC excinuclease as indicated and the reaction products were analyzed on an 8% polyacrylamide sequencing gel. Transcripts were labeled at the 5' end with [LY-~*P]UTP. The positions of the full length DNA fragment (137), the DNA band generated by incision 7 bases 5' to T-T (56) and the full length transcripts (89 and 90) and truncated transcripts (25 and 26) are marked. In lanes 4-6 the DNA was digested for 8 min with (A)BC excinuclease in the presence of rNTPs. In lane 5 the DNA was preincubated with RNAP for 5 min before addition of (A)BC excision nuclease. In lone 6, the DNA was preincubated with UvrA and UvrB for 5 min before adding RNAP and UvrC. Lane 8 is from a separate experiment in which the 137mer containing T<>T in the nontemplate strand ("W' in Fig. 1) was transcribed. B, the transcription stopsite in relation to the dimer. 5'end-labeled 25-and 26-base transcripts (lnnes l-2 and 5-6, respectively) formed in the presence of TOT (as in A) were excised separately from gels, purified, and sequenced by using the enzymatic method of Donis-Keller et al. (1977). The RNA ladders in lanes 3 and 4 were generated from alkaline hydrolysis of full length transcripts. Note that even though our data shows, unambiguously, that a nucleotide is inserted across from the 3' T residue, we have not actually demonstrated that the inserted nucleotide is dAMP.  (Fig. l), was preincubated in transcription buffer for 10 min with the indicated concentrations of RNAP to form either the RNAPpromoter (-riVZ'P) or the elongation complex (+rNTP), or photolyase to form the stable photolyase enzyme-substrate complex. Then (A)BC excinuclease was added and incubation continued for another 15 min. The reaction products were analyzed on 8% polyacrylamide sequencing gels. The location of the fragments generated by incision 5' to TOT in the nontranscribed (N) or transcribed (T) strands by (A)BC excinuclease are indicated. In lanes 10 and 11, the N and T substrates were individually digested with (A)BC excinuclease, and in lane 1 the mixture of N and T was not digested. The different panels are from separate experiments which employed slightly clifferent ratios of the N and T substrates. B, kinetics of incision of the transcribed and nontranscribed strands by (A)BC excinuclease. The reaction mixture contained 6.6 nM RNAP. Repair reactions were carried out for the indicated times, reaction products were analyzed on 11% polyacrylamide sequencing gels, and the relative amounts of incision were quantified by scanning densitometry of autoradiograms. The level of incision for each substrate is expressed relative to the level of incision obtained in the absence of rNTPs, which was given a value of 100. Triangles, TOT in the transcribed strand; circles, T<T in the nontranscribed strand; closed symbols, nontranscribing conditions (-rNTPs); open symbols, plus rNTPs. Data points are the mean of three separate experiments, standard errors ranged from 2 to 8 units of relative incision. Bertrand-Burggraff et al., 1984). By separating the UvrC gene from silent regions of the plasmid after the repair synthesis reactions, we could compare repair in the UrvC gene with repair in transcriptionally inactive regions. We first examined transcription of pDR3274 by examining transcripts that were internally labeled with [(r-32P]UTP on a 6.5% sequencing gel (data not shown). pDR3274 has three transcriptional units: tot-UvrC (1800 bp), tet (1200 bp), and RNA-I (108 bp The 137-mer (TC>T in the template strand) was preincubated with no additions (O), 0.7 nM RNAP (A), or 30 nM DNA photolyase (17) in transcription buffer with rNTPs, then (A)BC excinuclease was added and incubation continued at 37 "C. At the indicated times, samples were taken and the reactions were stopped. Samples were precipitated with ethanol, reaction products were separated on 8% polyacrylamide sequencing gels, and the bands corresponding to full length and incised 137-mer were quantified by scanning densitometry. The fraction incised was calculated as the amount incised divided by the total DNA (incised plus undigested) for each lane.

RNA synthesis was from the tat promoter.
This value is reasonably close to the 2.3 transcripts min-' for the lac operon in vivo under inducing conditions (Lewin, 1987).
We used UV-irradiated pDR3274 in our transcription-repair experiments. Production of full length tat-uvrC transcripts is inhibited in a dose-dependent manner, the transcription level being reduced to approximately e-i (data not shown) at a dose (30 J/m*) which is expected to produce about one dimer per strand of the transcription unit. We carried out repair synthesis with DNA irradiated with 75 Jm-', which produces about four dimers per uvrC gene. The reaction mixtures contained a full complement of transcription, repair, and resynthesis proteins in addition to the necessary cofactors and [(r-32P]dCTP as a tracer to measure repair synthesis. We used reaction-limiting amounts of the (A)BC excinuclease subunits so that repair synthesis was clearly detectable above the background yet far below saturating in order to detect any stimulatory effect of transcription. Following the transcription-repair reaction, the plasmid was digested with EcoRI, HindIII, and BglII to produce: (a) an -1.2-kb EcoRI-Hind111 fragment which is not transcribed; (b) an -2.1-kb EcoRI-BglII fragment containing the strongly transcribed UvrC gene; and (c) an -2.9-kb BglII-HindI fragment containing the tet gene and the origin of replication. The fragments were separated on agarose gels which were stained, photographed, and autoradiographed. The results ( Fig. 6 and Table I) indicate that the transcribed region of the plasmid is not repaired more efficiently than the nontranscribed regions. With similar concentrations of UV-irradiated DNA and Uvr proteins, the stimulatory effect of photolyase on repair synthesis was easily detectable (Sibghat-Ullah and Sancar, 1990). We considered the possibility that the topological consequences of transcription might be transmitted through the whole plasmid resulting in enhanced repair throughout, thus obscuring gene-specific repair. However, when a comparison is made between repair of all three fragments with or without transcription, no difference is seen ( Fig. 6 and Table I). In fact, considering the results obtained with linear DNA one might expect preferential inhibition of repair synthesis of the transcribed sequence when compared to either of the other fragments. While there appears to be a trend in that direction (Table I) the differences are not statistically significant (p < 0.005, Student's t tests). A plau-sible explanation for this lack of inhibition is that the inhibitory effect is diluted out because half of the photoproducts are in the nontranscribed strand (repair not inhibited) and stalling of RNAP at the first photoproduct in the template would not inhibit the repair of photoproducts downstream. Attempts to demonstrate an inhibitory effect with DNA containing on the average one T<>T per uurC were unsuccessful due to the small differences in repair synthesis when comparing singly adducted versus undamaged plasmids (data not shown). Also, higher RNAP concentrations than the one used in these experiments actually inhibited repair synthesis globally (data not shown).

DISCUSSION
Gene-and strand-specific repair are fascinating concepts both biologically and biochemically. From a biological standpoint, for example, it has been suggested (Bohr et al., 1985) that preferential repair of actively transcribed genes may explain the rodent cell paradox: rodent cells which are as resistant to UV as human cells remove only 20% of cyclobutane pyrimidine dimers from bulk DNA in 24 h compared to human cells which remove more than 80% within the same period. However, if the comparison is made at the gene level both cell types remove 80% of the photodimers. The implications are that cells need only to repair genes which control Lane I 2 3 4 5 6 7 8 essential functions and that photodimers do not constitute an absolute block to replication and thus can be tolerated. Evidence exists for tolerance mechanisms at the replicational level in both pro-and eucaryotes (Piette and Hearst, 1983;Yang et al., 1982;Vos and Hanawalt, 1987) and such a model makes teleological sense. In contrast, a comprehensive biochemical model which takes into account the known properties of the transcription apparatus and the nucleotide excision repair system has been difficult to formulate. Three general concepts have been considered: (a) the open conformation of chromatin in transcribed regions may increase accessibility to the repair enzyme; (b) RNAP stalled at a photodimer facilitates the binding and rate of repair by the subunits of the excision repair enzyme in a manner analogous to that of E. coli photolyase (Sancar et al., 1984) and perhaps yeast photolyase (Sancar and Smith, 1989); (c) the unique topology created at the transcription site by overwinding of DNA in front of the transcription complex and underwinding behind it (Tsao et al., 1989) generates a structure with higher affinity for the repair enzyme. Although the open chromatin structure in both pro-and eucaryotes in transcribed regions may make DNA damage more accessible to repair enzymes, it cannot account for the strand selectivity of the excision nuclease unless the template strand is in some way made more accessible than the nontemplate strand in the open chromatin structure. Smerdon and Thoma (1990) suggested that when RNAP stalls at the site of a lesion in the transcribed strand, the transcribed region remains in the "open configuration" for longer periods and thus allows more time for the excision nuclease to repair lesions in the template strand. The work reported here indicates that in uitro, RNAP stalled at a lesion does not itself facilitate repair and may indeed impede repair. We show that global effects produced by the transcription of superhelical DNA do not change the rate of repair by (A)BC excinuclease in transcribed sequences. Admittedly, our in vitro experiments cannot faithfully reproduce all the topological changes one would expect to observe in ho; therefore this issue is not totally resolved. An alternative is that there must be a coupling factor missing in our in vitro system which links transcription to repair. A strong candidate for such a factor in E. coli is photolyase which is known to stimulate excision repair on its own. However, we found that a greater stimulatory effect by photolyase in the presence of RNAP did not occur. We must therefore conclude that neither photolyase nor its possible mammalian analog (Patterson and Chu, 1989) are responsible for coupling repair to transcription.
In this paper we present the first unambiguous evidence TABLE I Effect of transcription on repair synthesis in defined regions ofpDR3274 Values represent damage-specific cpm incorporated and were derived from each +UV/-UV treatment pair such as those shown in Fig. 6 by subtracting the cpm in the unirradiated DNA band (which ranged from 13 in tet to 2 in cam in experiment 1; 32 in tet to 0 in cam in experiment 2; and 66 in tet to 8 in cam in experiment 3) from the cpm in the UV-irradiated DNA band.
Repair synthesis experiments  (76) 1.10 + 0.26 ' Values in parentheses are cpm per kilobase pair (cam, 1.2; uvrC, 2.1; tet, 2.9 kbp). +RNAP/-RNAP represents the ratios of cpm per kilobase pair; the means and standard errors of ratios obtained in the three experiments are given. that a thymine dimer is a strong block to transcription; this has long been inferred but never proven conclusively in either pro-or eucaryotes (Sauerbier and Hercules, 1978). We show that RNAP transcribes up to and including the nucleotide across from the 3' T of the dimer in about 90% of cases (Fig.  2) and stops at the nucleotide across from the 5' T of the dimer in 10% of the molecules (data not shown). For comparison, Shi et al. (1988) found that with psoralen-cross-linked and -monoadducted templates, E. coli RNAP stopped at the last base before the adducted T residue. Transcription up to and including the 3' T of the dimer is one base further than the stopsite of DNA polymerase I at a dimer reported by Moore and Strauss (1978), although a more recent report has shown that Poll can insert dAMP across from the 3' T (Taylor and O'Day, 1990).