Proc. Natl. Acad. Sci. USAVol. 81, pp. 5623-5627, September 1984Biochemistry

Site-specific carcinogen binding to DNA(benzo[a]pyrene diol epoxide/chicken ,B-globin gene)

T. CHRISTIAN BOLES AND MICHAEL E. HOGANDepartment of Molecular Biology, Princeton University, Princeton, NJ 08544

Communicated by Walter Kauzmann, May 4, 1984

ABSTRACT Benzo[a]pyrene diol epoxide (BPDE) is a well-studied environmental carcinogen that binds covalently toDNA. Here we describe a photochemical technique that allowsus to map BPDE-binding sites within cloned gene sequences.The technique is based upon our observation that, when ir-radiated with laser light at 355 nm, one single-strand DNA cutis produced at each BPDE binding site. In initial experimentswe have studied the distribution of such cuts in cloned DNAfrom the chicken adult /8-globin gene. We rind that BPDEbinding in this gene sequence is distinctly nonrandom. Whileseveral prominent BPDE-binding sites are evident, a 300-base-pair sequence immediately 5' to the RNA cap site is moststrongly attacked by the carcinogen. This region is believed tocontain important transcriptional control sequences. We dis-cuss the possibility that sequence-specific binding to such regu-latory elements may be an important feature of the mechanismof the carcinogen.

Benzo[a]pyrene (BP) is one of the most thoroughly studiedchemical carcinogens. However, its detailed mode of actionis still largely unknown despite 50 years of work (for a recentbrief overview of the history of this field, see ref. 1). Inrecent years the DNA-binding forms of the carcinogen andthe metabolic pathways leading to their formation have beenidentified. The consensus of this work is that (+ )-(7R,8S,9R, 10R)-7,8,8a-dihydroxy-9a,10a-epoxy-7,8,9,10-tetrahydroxybenzo[a]pyrene [(+)-BPDE] is the metabolitethat forms the majority of BP adducts. The major adductformed in vitro (2, 3), and in vivo (3), results from nucleo-philic attack of the N-2 of guanine upon the C-10 position ofBPDE to yield a trans opening of the 9,10-epoxide ring. Thisguanine derivative constitutes 80-90% of all stable adducts(2, 3) and has been proposed to be responsible for themutagenic (4) and carcinogenic (5, 6) effects of BP. Theadduct appears to distort DNA at its binding site; however,the secondary structure of the adduct is still disputed (7-10).Here, we examine the photochemistry of BPDE when

bound covalently to a DNA helix. We find that, when irradi-ated, the carcinogen cuts the DNA strand to which it isbound. We then show that the photochemical cutting processcan be developed into a powerful tool for mapping carcino-gen-binding sites in a eukaryotic gene.

MATERIALS AND METHODSBPDE Modification. DNA fragments 146 ± 2 base pairs

(bp) long were prepared from chicken erythrocyte nucleo-somes (11), then modified with (± )-BPDE as describedelsewhere (9), with the exception that unreacted BPDE andBP tetrol were removed by three phenol/chloroform extrac-tions. The extent of modification was measured optically,using 6346 = 2.95 x 104 cm-' M` for the BPDE-DNAadduct and E258 = 1.3 x 104 cm- I M - I (base pairs) for DNA.

Poly(dG)'poly(dC) (Boehringer) was sonicated (00C),treated with S1 nuclease (Boehringer) to remove single-strand regions, then digested with proteinase K, extractedwith phenol, and fractionated over a Sepharose 6B column asdescribed (12). The DNA fraction used for this experimentmigrated on a denaturing gel as a 330 + 50 base species.DNA was modified with [3H]BPDE and bound BPDE wasdetermined by liquid scintillation counting.

Preparation of 32P-End-Labeled .-Globin DNA. A 2110-bpEcoRI/Hindll fragment from pCA(3G1 (13) was subclonedin the EcoRI/HindIII fragment of pBR322. This plasmid(pHR-16a) was used for the mapping experiments in Fig. 4(see Fig. 5 for a schematic representation of the insert).To 32P-end-label the DNA strand with the same sense as

the mRNA, referred to as the "HindIII strand," pHR-16awas cleaved with HindIII, then 3'-end-labeled at that sitewith the Klenow fragment of DNA polymerase I and [a-32P]dNTPs (14). After purification of the labeled DNA byextraction with phenol and precipitation with ethanol, it wascleaved with EcoRI. The digest was electrophoresed in a 1%agarose gel. The 2.1-kbp band was localized by autoradiog-raphy, and the DNA was electroeluted from the excised gelband (14). The DNA was purified by Elutip (Schleicher &Schuell) chromatography. DNA was then resuspended in 10mM Tris'HCl/l mM EDTA, pH 7.5 (TE), and modified withBPDE to give an average of 1 BPDE per 2110-base singlestrand. The end-labeling strategy for the "EcoRI strand" isthe same, with the difference that the plasmid was firstdigested with EcoRI and 3'-end-labeled at that site.

RESULTSBPDE-Mediated DNA Cutting. The strategy of our work is

to selectively excite BPDE with intense laser light at 355 nm,then monitor the DNA strand scission that may result. Suchcutting can be quantitated in a simple manner by denaturinggel electrophoresis if the DNA samples are homogeneouswith respect to molecular weight. The DNA used for ourinitial experiments is 146 + 2 bp chicken DNA, extractedfrom nucleosome core particles (11).

Fig. 1 Upper illustrates the assay we use to quantitateBPDE-mediated cutting. DNA to be assayed for light-in-duced cutting is divided into two identical samples. One isirradiated with 355-nm laser light, then both are subjected toelectrophoresis in a denaturing acrylamide gel. As seen inFig. 1 Upper, BPDE is responsible for light-induced cutting.No cutting is detected in the absence ofBPDE (slots a and b).The amount of cutting (the band intensity decrease) in-creases steadily with increasing carcinogen density (slots cthrough h).The extent of cutting can be quantified accurately. Let A-

be the integrated area of the densitometer trace of the photo-graphic negative of an unirradiated DNA band; A+ is theband area associated with a duplicate, irradiated sample. The

Abbreviations: BP, benzo[a]pyrene; BPDE, BP diol epoxide; bp,base pair(s).

The publication costs of this article were defrayed in part by page chargepayment. This article must therefore be hereby marked "advertisement"in accordance with 18 U.S.C. §1734 solely to indicate this fact.

5623

5624 Biochemistry: Boles and Hogan

a b c d e fg hV.. .... X

3.2

2.8

2.4 -

UL)0

L)0

cn

3)

2.0-

4.61-

4.2 -

0.81-

0.4

0.01

-0.410.0

ro

0

340 350 360

Wavelength (nm)

FIG. 1. (Upper) Gel assay for BPDE-mediated cutting. Three-microliter samples of BPDE-modified 146-bp chicken DNA (3 xl0-4 M bp of DNA) were irradiated for 5 min with 355-nm laserlight at a continuous power output of 20 mW. Irradiated samplesand nonirradiated controls were electrophoresed under denaturingconditions in 5% acrylamide/7 M urea gels. After electrophoresis,gels were equilibrated in buffer without urea, stained with ethidiumbromide, and photographed under UV light. Lanes a, c, e, and gare irradiated samples. Lanes b, d, f, and h are nonirradiated sam-ples. Lanes a and b were loaded with identical amounts of unmodi-fied DNA. Lanes c and d were loaded with identical amounts ofDNA modified with BPDE to a population average of 2.6 adductsper 146-base single strand. Similarly, lanes e and f were loaded withDNA with 1.6 adducts per strand, and lanes g and h were loadedwith DNA with 1.1 adducts per strand. (Lower) Bleaching of BPDE-DNA complexes. Samples of BPDE-modified 146-bp chicken DNA (1.9BPDE per strand of DNA) were irradiated for 5 min with 355-nmlaser light at various intensities. Nine-microliter aliquots of irradi-ated samples were diluted 1:100 and absorbance spectra were re-corded on a Hewlett-Packard 8504 spectrophotometer. From top tobottom, the four curves represent spectra obtained from identicalsamples irradiated at 0, 5, 10, and 20 mW of continuous power, re-spectively.

ratio A+/A - is then the fractional change in the number offull-length strands due to irradiation. In terms of Poissonstatistics, A + /A - is a measure of the probability, P0, that nocuts occurred, given a population average of N cuts perstrand, as calculated from the Poisson distribution function

Po = At/A- = exp(-N)

-0

V)

0--

2.0

4.5

0.54

0

I1

0

0.5 1.0 1.5 2.0 2.5 3.0 3.5BPDE /Molecule

1 2 3 4 5 6

BPDE /Helix

FIG. 2. (Upper) Relationship between BPDE dose and DNAcutting. BPDE-modified 146-bp chicken DNA was irradiated for 10min at 30 mW (e, *) or 20 mW for 5 min (o, o). Irradiated sampleswere electrophoresed with an identical, unirradiated control in a 5%acrylamide gel under denaturing conditions (squares): c, 7 Murea/83 mM Tris borate/2 mM EDTA, pH 8.3; *, 30 mM NaOH/2mM EDTA, pH 12.5. Other BPDE complexes were assayed undernondenaturing conditions (circles): 83 mM Tris borate/2 mMEDTA, pH 8.3. The extent of BPDE-mediated cutting was quan-titated from photographic negatives of the gels. The units of theaxes correspond to cuts or BPDE per 146-base single strand for thesquares and cuts or BPDE per 146-bp helix for the circles. The datawere fit to a linear function by the method of least squares. For thedenaturing gel conditions the best-fit slope is 0.925 ± 0.030 with anintercept of 0.013 ± 0.004. For nondenaturing conditions the slopeis 0.048 ± 0.009; intercept 0.030 ± 0.020. (Lower) Strand speci-ficity of BPDE cutting. BPDE-modified 300-bp poly(dG)-poly(dC)was 32P-end-labeled on either the dG strand or the dC strand. Sam-ples were irradiated and electrophoresed with controls on 2%agarose/NaOH denaturing gels. Gels were fixed with 10%1 trichloro-acetic acid and autoradiographed. BPDE-mediated cutting of thedG strand (e) or the dC strand (o) was quantitated from the auto-radiograms.

The quantity N can be decomposed further in terms of theaverage number of BPDE per strand (n) times the probability(p) that a nick occurs at a BPDE-binding site (the cuttingefficiency).

In Fig. 2 Upper we display the relationship between light-induced cutting and BPDE-binding density. When the prod-uct is assayed under denaturing conditions, the number of

0~~~~~~~

0~~~~

0~~~~~~~~1

Proc. Natl. Acad. Sci. USA 81 (1984)

4.0~

Proc. Natl. Acad. Sci. USA 81 (1984) 5625

cuts (the log of band area change) varies linearly with in-creasing BPDE per strand, with a slope p = 0.92 + 0.03.This indicates that approximately one cut is made for eachBPDE adduct. In contrast, when the product is assayedunder nondenaturing conditions it is evident that the prob-ability of double-strand cutting is about 5% that for singlestrands (lower curve in Fig. 2 Upper). The low double-strandcutting probability confirms the single-strand character ofBPDE-induced cuts.

It should be emphasized that increasing the irradiationdose 2-fold does not increase the efficiency of BPDE-me-diated cutting and that the same cutting efficiency is ob-served in two different denaturing gel systems (Fig. 2 Up-per). From these data, we conclude that under our irradiationconditions the cutting reaction has proceeded to completion.To determine the strand specificity of cutting we have

examined BPDE-mediated cutting of poly(dG)-poly(dC).Since BPDE binds readily to G, but has little affinity for Cresidues (2, 3), the poly(dG)-poly(dC) helix will be modifiedby BPDE only on its G strand. Poly(dG)-poly(dC) was modi-fied with BPDE, then 3'-end-labeled with the Klenow frag-ment of DNA polymerase I and either [a-32P]dGTP or [a-32P]dCTP (14) so that cutting of each strand could be exam-ined independently.As seen in Fig. 2 Lower, the data clearly show that only the

dG strand is cut; cutting of the dC strand is not detectableeven when six BPDE are bound per helix. Therefore, whenirradiated, BPDE cuts only theDNA strand to which it iscovalently bound.

Several DNA-binding dyes are known to mediate the pro-duction of singlet oxygen (102), which in turn is capable ofdamaging the DNA helix (15-17). To test for the involvementof such a mechanism, we have performed BPDE cuttingassays in the presence of the '02 quenchers triethylamine(18) and sodium azide (19). The data of Fig. 3 Lower showthat the quenchers have no detectable effect on the cuttingreaction. Thus, it seems unlikely that the BPDE cuttingmechanism involves 102 production.We have also examined the fate of the pyrene moiety

during the DNA cutting reaction. In Fig. 1 Lower we presentabsorbance spectra obtained from irradiated and controlDNA samples containing 1.9 ± 0.2 BPDE molecules perstrand. A marked decrease in absorbance at 346 nm is seenafter irradiation. The data of Fig. 1 Lower have been plottedin Fig. 3 Upper (broken line) to show the dependence of thatbleaching reaction on irradiation intensity.For comparison, we have performed cutting assays on the

DNA samples used for the absorbance measurements. Theintensity dependence of the cutting reaction (Fig. 3 Upper,solid line) is also first order. Both reactions appear to besingle-photon processes with a quantum yield near 4 x 10-at 355nm. Such close correspondence suggests that the tworeactions may be coupled.To summarize, we propose a working hypothesis to ex-

plain DNA cutting by BPDE. (i) The pyrene moiety ofBPDEabsorbs a single photon of light (the reaction is first orderwith respect to intensity, as seen in Fig. 3 Upper). (i) Theexcited pyrene engages in direct photochemistry with the Gstrand to which it is fixed (Fig. 2 Lower). As a result of thereaction, pyrene is bleached and the backbone of the helix iscut (Fig. 3 Upper). Because cutting and bleaching are stoi-chiometric (Fig. 2 Upper), and because diffusible singletoxygen is not required for the cutting process (Fig. 3 Lower),it is very likely that the single DNA cut that occurs islocalized to the carcinogen-binding site.Sequence Specificity ofBPDE Binding. We have used the

cutting assay described above to examine BPDE-bindingsites in a cloned 2.1-kbp DNA fragment containing the 5' halfof the chicken adult fl-globin gene (Figs. 1-3). As describedin Materials and Methods, we have labeled the ends of that

C"It

3.20

2.80 - a

2.40 -

2.00

1.60-

1.20 -

a0.80 a

0.40 - le

0.00-

-0.40 13 1 I I0.0 4.0 8.0 12.0 16.0 20.0 24.0 28.0

I (mW)

3

'DCOCU

(0

21

10 20

Quencher (mM)

FIG. 3. (Upper) Light intensity dependence of BPDE-mediatedcutting and BPDE adduct bleaching. BPDE-modified 146-bpchicken DNA (1.9 BPDE molecules per strand) was irradiated for5 min with 355-nm laser light of various intensities. o, Intensity de-pendence of BPDE adduct bleaching at 346 nm. These data weretaken from the absorbance spectra of Fig. 1 Lower.*, Intensity de-pendence of BPDE-mediated single-strand DNA cutting. Cuttingwas assayed under denaturing conditions as in Fig. 1 Upper. Thecurves represent computer fits to the data by a generalized nonlin-ear least-squares method. Data were fit to the function - ln(A/A0)- Cl', in which A/AO is the fraction of material remaining uncut orunbleached after being irradiated for 5 min at the indicated inten-sity. C is a constant, I is the intensity of the laser in mW, and i isthe order of the photochemical reaction with respect to intensity.For the bleaching reaction (---), the fit values are C = 0.056 +0.020, and i = 1.287 + 0.127. For the cutting reaction ( ), thefit values are C = 0.112 + 0.018 and i = 1.050 0.058. (Lower)Effect of singlet oxygen quenchers on BPDE-mediated DNA cut-ting. BPDE-modified 146-bp DNA (1.9 BPDE per strand) wasmixed with various concentrations of triethylamine (e) or sodiumazide (o) then assayed for light-induced cutting as described. In allcases the supporting buffer was TE, pH 7.5. Irradiation conditionswere 20 mW, 5 min at 355 nm. The broken line corresponds to 1.9cuts per strand (100% cutting efficiency).

3-globin sequence in a strand-specific manner. Those labeledfragments were modified with BPDE, irradiated, and elec-trophoresed on denaturing agarose gels. Autoradiograms ofsuch gels are shown in Fig. 4. The lanes containing modified,irradiated DNA (lanes c and e in Fig. 4 Upper; lanes c, f, andi in Fig. 4 Lower) show a series of discrete bands super-imposed over low-level background radioactivity. The lanescontaining unmodified DNA (irradiated or nonirradiated)and the lanes containing nonirradiated, modified samplesshow negligible amounts of radioactive species smaller than2.1 kilobases. Such discrete DNA cutting shows that BPDE

00

0

p.0

Biochemistry: Boles and Hogan

iS

5626 Biochemistry: Boles and Hogan

(1 c (I '

me

f/ g 11 I >.t3c 2

Jr c

0.80.7

. 0.6c 0.5a 0.40 0.3

0.2O.1

c d r /. I1 i jI

-750 -500 -250 0 +250 ±500 +75Distance From Cap Site (bp)

FIG. 4. Mapping BPDE-binding sites in the chicken adult f-globingene. (Upper) HindI11 strand. The 2110-bp EcoRI/HindIII frag-ment front pHR-16a was 3'-end-labeled with 32P at the findIII siteand modified with BPDE to give an average of 1 BODE per 2110-base single strand. Samples were irradiated at 20 mW for 10 min.DNA samples were electrophoresed on a 1.7% agarose/NaOH gel;gels were fixed with 109% trichloroacetic acid and autoradio-graphed. Lanes b-e are BPDE-modified DNA; lanes c and e areirradiated samples; lanes b and d are identical nonirradiated sam-ples. Lanes g-j are unmodified DNA; lanes h and j, irradiated sam-ples; lanes g and i, nonirradiated samples. Lanes a and f are end-labeled Taq I digests of pBR322; the fragments are 1444, 1307, 616,368, 315, and 312 bases long (13). (Lower) EcoRI strand. The 2110-bp /3-globin gene fragment from pHR-16a was 3'-end-tabeled with32P at its EcoRI terminus and subsequently modified with BPDE togive an average of 1 BPDE per 2110-base single strand (see Materi-als and Methods). All Other steps were performed as for Fig. 4 Up-per. Lane a is an end-labeled HindIll digest of phage X DNA. Lanesb and are end-labeled Taq I digests of pBR322. Lanes c, f, and iare BPDE-modified irradiated DNA; lanes d, g, and j are BPDE-modified nonirradiated DNA; lanes e, h, and k are nonmodifiednonirradiated DNA.

FIG. 5. Summary of BPDE-binding sites in the pHR-16a insert.(Upper) BPDE-cutting (BPDE-binding) sites on the strand labeledat the HindIII end. (Lower) BPDE cutting sites on the other strand,labeled at its EcoRI end. Vei-tical bars in the upper section of eachpanel correspond to the positions of cutting sites. For consistency,position has been converted to separation, in bp, from the RNA capsite (20). The height of the bars is proportional to the optical den-sity of bands in Fig. 4 (measured by microdengitometry). In an end-labeling experiment such heights are directly proportional to theprobability that BPDE cuts (binds) at that position. The center fig-ure in each panel is a diagram of 8-globin gene structure (20): openboxes, transcribed noncodhng sequences; shaded boxes, coding se-quences. The position of the DNase I-hypersensitive site (21) isunderlined. At the bottom of each panel is a computer calculationof G residue density on each strand, at 30-base resolution. The datafor these calculations are the /B-globih se4uence of ref. 20. The se-quence data start at - 388 and extend through the HindIII end ofthe fragment. In these calculations the nearest neighbors G-A, G-G, G-T, and G-C are assigned a value 1 and all other pairing is as-signed a value of 0.

binds in a distinctly nohrandom fashion to chicken j3-globinDNA.

In Fig. 5, the position and relative intensity of the bindingsites mapped in the two end-labeling experiments are shown.It is especially important to recognize that a great deal ofcutting (BPDE binding) occurs within a 300-bp region on the5' side of the gene. This region (0 to - 300) is believed tocontain important transcriptional control signals for f-globin(20, 21). Evidently, the carcinogen recognizes and selec-tively attacks DNA sequences within this gene control re-gion.

In an attempt to explain the nonrandom distribution ofBPDE-binding sites in this gene, we have analyzed the se-

quence of the fragment (20) at 30-base resolution (the reso-lution of our electrophoresis system). These plots are alsoshown in Fig. 5. The plots show that the G density along thetwo strands is markedly nonrandom. On the EcoRI strand,the region from + 50 to -200 contains prominent G-richsegments that align with BPDE-binding sites (Fig. 5 Lower).Two other G-rich features are seen on the HindIII strand at+ 650 and + 700 bp (Fig. 5 Upper) which also align withBPDE-binding sites.

Ii Fl I . nF ,F n-750 -500 -250 0 +250 +500 +750

EcoRI Hindil

-750 -500 -250 0 +250 +500 +750

Proc. Mal. Acad. Sel'. USA 81 (1984)

q-w

Proc. Natl. Acad. Sci. USA 81 (1984) 5627

However, closer examination shows that a complete cor-relation between G density and BPDE binding cannot bemade. Intense cutting sites centered at - 100 and - 250 onthe HindIII strand occur in regions of moderate to low Gdensity. On the JcoRI strand, the prominent cutting site at- 260 falls in a region of moderate G density.On the basis of these data, we conclude that site-specific

BPDE binding is not related to the density of G residues in asimple fashion. It therefore seems likely that other effects,such as sequence-specific variation ofDNA secondary struc-ture can also influence BPDE binding. In this context, wehave searched the 6-globin sequence for regions capable ofundergoing the B-Z transition, using a modification of theprogram described in the legend of Fig. 5. At 10- or 30-baseresolution, there appears to be no correlation between BPDEbinding and the incidence of alternating purine-pyrimidinesequences (data not shown).We would like to emphasize that our studies were carried

out with linearized DNA fragments. Therefore, the observedspecificity of BPDE binding cannot be attributed to thepresence of alternate DNA structures stabilized by super-helical stress.

DISCUSSIONThe work presented in this paper constitutes direct demon-stration that a carcinogen can attack a gene in a site-specificmanner.

In an earlier experiment, [3H]BPDE adducts were found tobe distributed randomly among restriction fragments of sim-ian virus 40 DNA (22). We believe that, even if sequencefeatures exist on this DNA to direct BPDE binding, thoseexperiments lacked the sensitivity and spatial resolutionnecessary for mapping.

In a second class of experiment, BPDE binding has beenmapped within a cloned fragment of the Escherichia coli ladgene (24), taking advantage of depurination and strand break-age that can be induced when BPDE binds to N-7 of guanine.Those experiments showed'that BPDE can induce depuri-nation that is sensitive to local sequence effects. Unfortu-nately the N-7 adduct is rare [less than 20% of the totalformed in vitro (25)], and strand breakage by depurination isinefficient. We suggest that the distribution of BPDE-bindingsites may not be related to the data in a simple way in suchmeasurements.Less direct, though highly suggestive, data on binding

specificity have come from an analysis of the distribution ofbase-substitution mutations induced by B3PDE in the ladgene (23). That study clearly demonstrated that a few po-sitions in the lad sequence are especially mqtable. Again,however, the relationship between binding and the pattern ofmutation may be complex in those experiments.Here, we have shown in a direct way that BPDE recog-

nizes and selectively attacks the 5' flanking regions of thechicken 8-globin gene. We believe that such' site specificbinding may not be a unique property of the globin gene:many genes, including chicken 8-globin, display DNA se-quence features near the origin of transcription that make thegene sensitive to nuclease digestion in chromatin (26) and inpurified recombinant plasmids (27). If BPDE can recognize

such features in the -globin gene, it may recognize and bindspecifically to those same features in other parts of thegenome. That hypothesis can be tested by examining BPDEbinding to other genes.

The plasmid pCABG1 was graciously provided by Gary Felsen-feld. We acknowledge Johnny Chang-Ning Wang for the gift ofpHR-16a plasmid DNA and many helpful discussions. This work wassupported by American Cancer Society Grant NP352.

1. Phillips, D. H. (1983) Nature (London) 303, 468-472.2. Meehan, T. & Straub, K. (1979) Nature (London) 277, 410412.3. Jeffrey, A. M., Weinstein, I. B., Jennette, K. W., Grzes-

kowiak, K. & Nakanishi, K. (1977) Nature (London) 269,348-350.

4. Newbold, R. F. & Brookes, P. (1976) Nature (London) 261,52-54.

5. Buening, M. K., Wislocki, P. G., Levin, W., Yagi, H.,Thakker, D. R., Akagi, H., Koreeda, M., Jerina, D. M. &Conney, A. H. (1978) Proc. Natl. Acad. Sci. USA 75,5358-5361.

6. Nakanishi, K., Kasai, H., Cho, H., Harvey, R. G., Jeffrey,A. M., Jennette, K. W. & Weinstein, I. B. (1977) J. Am.Chem. Soc. 9, 258-260.

7. Prusik, T., Geacintov, N. E., Tobiasz, C., Ivanovic, V. &Weinstein, I. B. (1979) Photochem. Photobiol. 29, 223-232.

8. Geacintov, N. E., Gagliano, A., Ivanovic, V. & Weinstein,I. B. (1978) Biochemistry 17, 5256-5262.

9. Hogan, M. E., Dattagupta, N. & Whitlock, J. P. (1981) J. Biol.Chem. 256, 4504-4513.

10. Yang, N.-C., Ng, L.-K., Neoh, S. B. & Leonov, D. A. (1978)Biochem. Biophys. Res. Comnrun. 82, 929-934.

11. Wang, J., Hogan, M. & Austin, R. H. (1982) Proc. Natl. Acad.Sci. USA 79, 5896-5900.

12. Hogan, M. & Jardetzky, 0. (1980) Biochemistry 19, 2079-2085.13. Wood, W. I., Nickol, J. & Felsenfeld, G. (1981) J. Biol. Chem.

256, 1502-1505.14. Maniatis, T., Fritsch, E. F. & Sambrook, J. (1982) Molecular

Cloning: A Laboratory Manual (Cold Spring Harbor Labora-tory, Cold Spring Harbor, NY), pp. 113-121.

15. Houba-Herin, N., Calberg-Bacq, C. M., Piette, J. & van deVorst, A. (1982) Photochem. Photobiol. 36, 297-306.

16. de Mol, N. J., Beijersbergen van Henegouwen, G. M. J. & vanBeele, B. (1981) Photochem. Photobiol. 34, 661-666.

17. Piette, J., Calberg-Bacq, C. M. & van de Vorst, A. (1981)Photochem. Photobiol. 33, 325-333.

18. Ogryzlo, E. A. & Tang, C. W. (1970) J. Am. Chem. Soc. 92,5034-5036.

19. Hasty, N., Merkel, P. B., Radlick, P. & Kearns, D. R. (1972)Tetrahedron Lett. 1, 49-52.

20. Dolan, M., Dodgson, J. B. & Engel, J. D. (1983) J. Biol. Chem.258, 3983-3990.

21. McGhee, J. D., Wood, W. I., Dolan, M., Engel, J. D. &Felsenfeld, G. (1981) Cell 27, 45-55.

22. Pulkrabek, P., Leffler, S., Grunberger, D. & Weinstein, I. B.(1979) Biochemistry 18, 5128-5134.

23. Eisenstadt, E., Warren, A. J., Porter, J., Atkins, D. & Miller,J. H. (1982) Proc. Natl. Acad. Sci. USA 79, 1945-1949.

24. Haseltine, W. A., Lo, K. M. & D'Andrea, A. D. (1980) Sci-ence 209, 929-931.

25. King, H. M. S., Osborne, M. R. & Brookes, P. (1979) Chem.-Biol. Interact. 24, 345-353.

26. Wu, C. (1980) Nature (London) 286, 854-860.27. Mace, H. A. F., Pelham, H. R. B. & Travers, A. A. (1983)

Nature (London) 304, 555-557.

Biochemistry: Boles and Hogan