Sequence specificity of 125I-labelled Hoechst 33258 in intact human cells

?I. Mol. Riol. (1988) 201. 437442

Sequence Specificity of ’ 251-labelled Hoechst 33258 in Intact Human Cells Vincent Murray and Roger F. Martin

481 Little

Molecular Science Group Peter MoxCallum Cancer Institute Lonsdale Street, Melbourne, Victoria 3000, Australia

(Received 23 October 1987, and in revised form 30 December 1987) LJsing polyacrylamide/urea DNA sequencing gels, the DNA sequence selectivity of ‘251-labelled Hoechst 33258 damage has been determined in intact human cells to the exact base-pair. This was accomplished using a novel procedure with human aRI-DNA as the target DNA sequence. In this procedure, after size fractionation, the aRI-DNA is selectively purified by hybridization to a single-stranded Ml3 clone containing an aRI-DNA insert. in intact cells from The sequence specificity of [ ‘251]Hoechst 33258 was indistinguishable purified high molecular weight DNA; and this is surprising considering the more complex environment of DNA in the nucleus where DNA is bound to nucleosomes and other DNA binding proteins. The ligand preferentially binds to DNA sequences which have four or more consecutive A. T base-pairs. The extent of damage was measured with a densitometer and, relative to the damage hotspot at base-pair 94, the extent of damage was similar in both purified high molecular weight DNA and intact cells. [ “‘I]Hoechst 33258 causes only double-strand breaks, since single-strand breaks or base damage were not detected. These experiments represent the first occasion that the sequence specificity of a DNA damaging agent. which causes only double-strand breaks, has been determined to the exact base-pair in intact cells.

1. Introduction

Martin & Holmes (1983) have demonstrated, using defined plasmid sequences in vitro, that four consecutive A. T base-pairs are necessary for strong binding. Recently an X-ray crystallographic structure of Hoechst 33258 bound to the oligonucleotide CGCGAATTCGCG has been determined (Pjura et al., 1987). This structure revealed that, Hoechst 33258 bound to the sequence ATTC and not the expected AATT. There are two methods for determining the sequence specificity of DNA binding agents. The first type is usually called DNA “footprinting” and utilizes a DNA damaging agent which cleaves DNA (preferably) at every base-pair, but DNA cleavage is inhibited where the ligand binds t,o DNA. In the second method a DNA cleaving agent is attached (or is already attached) to the DNA ligand and where the DNA ligand binds to DNA. the DNA cleaving agent is brought into close proximity to DNA and “marks” the area of DNA ligand binding. We wished to elucidate the sequence specificity of Hoechst 33258 in intact human cells. The first method cannot be used for intact cells and so the 1251 second method was employed utilizing (covalently attached to Hoechst 33258) as the DNA cleaving agent. “‘I decays bv electron capture

Various clinically important drugs bind to and damage DNA, and this is thought to be related to their biological effectiveness. The sequence specificity of DNA damaging agents in intact cells is an important parameter of its effectiveness as a drug. Techniques exist, for determining the sequence specificity (to the exact base-pair) of agents that cause single-strand breaks in intact cells (Lippke et al., 1981; Grunberg & Haseltine, 1980; Murray & Martin, 1985a); up till now no technique allowed the analogous determination for agents that cause mainly double-strand breaks. Previously the estimation of the exact DNA sequences cleaved by agents causing double-strand breaks was limited by the resolution of agarose gels (tens to hundreds of base-pairs depending on the position of the doublestrand breaks in the gel) after Southern blotting. In the method described in this paper, the final products are analysed on polyacrylamide/urea DNA sequencing gels and hence damage can be located to the exact base-pair. Hoechst, 33258 is a his-benzimidazole that binds to the minor groove of DNA (Mikhailov et aZ., 1981). It binds preferentially to A +T-rich regions of DNA 437

0 1988 Academic Press Limited

V. Murray

438

and R. F. Martin

where an intense shower of low energy electrons are produced at every decay (average is 22: Charlton & Booz, 1981) which causes a double-strand break in DNA (Krisch & Sauri, 1975). The tandemly repeated human @RI-DNA was used as target DNA to examine the sequence selectivity of [ “‘1 ] Hoechst 33258. There are approximately 50,000 copies of 340 bpt aRI-DNA per haploid genome (Darling et al., 1982). The 340 bp uRI-DNA is sufficiently homogeneous to give a unique sequence by Maxam-Gilbert DNA sequencing (Wu & Manuelidis, 1980). In order to detect double-strand breaks a novel experimental system was developed. DNA sequences of less than 340 bp were collected from agarose gels by a modification of the method of Dretzen et al. (1981). The aRI-DNA sequences were purified by hybridization to a single-stranded Ml3 clone containing an aRI-DNA insert (Murray & Martin, 1985b).

2. Materials and Methods (a) Materials All chemicals used were of reagent grade. All enzymes were obtained from Bethesda Research Laboratories or BRESA (Australia). [cr-32P]dATP (5000 Ci/mmol) was from New England Nuclear. Hoechst 33258 was labelled to a specific activity of 2100 Ci/mmol with carrier-free lz51 (Amersham International) by the method of Martin & Pardee (1985).

collect aRI-DNA. which had double-strand brraks. DNA of less than 340 bp was reverse electrophoresed into DEAE-cellulose membranes for .5h at high currclnt (340 bp DNA was reverse electrophoresed for 0.5 h). The EcoRI 5’ overhang was end-labelled with [32PJdATP (Maniatis et al., 1982) and then subjected to the Ml3 hrbridization procedure as described (Murray & Martin, 19856). Briefly 32P-labelled aRT-DNA is hybridized to single-stranded Ml3 DNA containing 21 cloned copv of crRI-DNA. Only one strand of the ctRIDNA hybiidizes to t,he Ml3 DNA and t,hr hybridized ctRT-DNA was then purified by l~ol~aerylan~idr gel electrophoresis. The use of Ml3 clones alth either (+) or (-) strand aRI-DNA inserts enables isolation of rit heI purified (-) or ( +) strand aRI-DNA. respectively. An important aspect of this techniyue is that non-aR,T-l)KA sequences are removed from the DNA preparation. (Cl) Densitwmrter

A densitometer analysis of [ 1251]Hoerhst X)258induced damage hotspots was made using an LKB 2202 laser densitometer scanner. The peak heights were determined and were normalized to correct for difierent sample loads in each lane. This normalizat,ion was carried out using the aRI-DNA-related bands at bp 112. 123. 172 (the mean is defined as 1.0). These bands are present in both control and t,est, samples and can he used as a constant reference point. Tn Table 1 (column B for each sample) the densitometer readings are ~alculat~rd with reference to the densitometer value at bp 94. and therefore indicate the relative intensity of the hotspots

(b) CIell culture K562 cells (Lozzio & Lozzio. 1975) were grown in suspension culture by conventional methods in medium containing 10% (v/v) foetal calf serum. (c) Incubations

with [““IIHoechst

33258

Approximately 2 x lo6 or lo6 of phosphate-buffered saline (PBS)-washed K562 cells, 10 pg or 5 pg of purified high molecular weight DNA from K562 cells were incubated in a volume of 50~1 with approximately 6x IO7 cts/min of [ “‘T]Hoechst 33258 (0.48 PM) in 20 mM-Tris (pH 7.9), 110 mM-NaCl. 5 mM-EDTA, 5 mMKI, 10% (V/V) glycerol at room temperature for 1 h (to allow uptake of the ligand) and for 31 days at -70°C. The molar binding ratio of DNA (base-pair) to [‘251]Hoechst 33258 DNA was approximately 750 for 5 pg and 1500 for 10 pg of DNA. The cells were washed twice with PBS and then the high molecular weight DNA was extracted as described (Murray 6 Martin, 1985a). Briefly. this method involved cell l*ysis with SDS. drproteinization with proteinasr K. extraction with phenol. precipitation with ethanol and RNase t’reatment. The incubations with purified high molecular weight DNA were precipitated with ethanol. Both cellular-derived and purified DNA were cut with EcoRI after a sample had been taken to assess damage on a 1.5:/o (w/v) agarose/ethidium bromide gel. The EcoRIcut samples were run on a 1.5% agarose/ethidium bromide gel and 340 bp aRI-DNA was eluted into DEAEcellulose membranes (Dretzen et al., 1981). However, to t Abbreviations used: bp. base-pair(s); PBS, phosphate-buffered saline.

ccnalysis

3. Results The

degree of uptake of ‘251-labelled Hoe&t by the K562 cells was assessed using centrifugation. Approximately 4OY, of t’he input counts remained associat,ed with t’hr cell pellet after the initial centrifugation step and two PRS washings. rrsing a dye-exclusion ijest.. greater than 98% of the cells had intact, cellular membranes. After incubation with 12SI-labelled Hoechst 33258 t,he high molecular weight DNA was isolated from the cells. The percentage of the “‘I counts associated with DNA after t,he phenol extraction stage was onl? 0.1 %. After the a,garosr gel 1251 courrts remained in procedure no significant either the cellular DNA or purified DNA sa,rnples. A portion of high molecular weight’ DNA for each sample was analysed by agarose gel rlectrophoresis. For both t’hr cellular and purified 1)N.A samples that contained no [ ‘251]Hoechst 33258. only the high molecular weight> DNA hand was visible. For the other samples. all of which had hem incuhatjed was nppart~nt. with [‘25 I]Hoechst, DNA Imakagcb For t’he samples with less cells or less input’ I)NA. more breakage was observed. The cellular samples were less cleaved than the purified DXA samples. In contrast’ to other agent,s that eleavr DNA with sequence t~xarnple: hlVornyc,in specaificil y. fi)r (Murray 8 Martin. 1985a); or micrococc~al nucleast (Musich et d., 1982), [ “‘1 ] Hoechst, 33258 darrragc, did not result, in any observable bands at 170. 340. 510 bp, etc., which indicat’rs lack of’ preferential 33258

Sequence Specijcity 1 2 3 4 5 6

7 8 9 101112 -

340

-

250

-

200

-

150

-

120

-

100

purified DNA. The exact location of bands in the autoradiograph can be ascertained by comparison to the Maxam-Gilbert G +A sequencing track. There are prominent bands at 112, 123 and 172 bp in both the control and test lanes. (These bands are presumably CXRI-DNA sequences which have EcoRI ends but are shorter than the 340 bp sequence.) However, in the samples that were incubated with [‘251]Hoechst 33258, there are an additional series of bands with peaks at 69, 84, 94, 106, 135, 143 and 157 bp. Hotspots on the other DNA strand were also detected at 281 and 262 bp (data not shown). The position of the hotspots relative to the “consensus” crRI-DNA sequence is shown in Figure

2. These bands are characteristic

consecutive

DNA DNA. -

80

Figure 1. Autoradiograph depicting [‘251]Hoechst 33258 damage to intact K562 cells and purified DNA. Lanes 2, 5, 6 and 8 to 11 were incubated with [“‘I]Hoechst 33258. Lanes 1, 4 and 7 are control lanes and were not incubated with [1251]Hoechst 33258. Lanes 4 to 6 are from intact K562 cells. Lanes 7 to 11 are from high molecular weight purified DNA. Lane 10 is purified DNA incubated at 4°C. Lane 11 is purified DNA incubated at 4°C with no glycerol present. Lanes 4 to 11 are from DNAs smaller than 340 bp in length. Lanes 1 and 2 are derived from the 340 bp aRI-DNA band. Lanes 3 and 12 are the Maxam-Gilbert G + A sequencing tracks. An 8% polyacrylamide DNA sequencing gel was used and the ( -) strand damage is shown. Sizes are indicated in bp.

at the linker region of nucleosomes. After elution from the agarose gel the samples

cleavage

were end-labelled with [32P]dATP, the strands were separated by the Ml3 hybridization method, and the samples were then loaded on to thin polyacrylamide/urea DNA sequencing gels. The autoradiograph in Figure 1 shows the comparison between

cells and in

of hotspots

where ]‘251]Hoechst 33258 has damaged the DNA at the site of DNA binding (Martin & Holmes, 1983). These hotspots are in exactly the same position as [‘251]Hoechst 33258cleaved 340 bp aRI-DNA (V. Murray & R. F. Martin, unpublished results) and correspond to [‘25T]Hoechst 33258 binding to A + T-rich regions in the DNA. As found by Martin & Holmes (1983) the site of the damage hotspots are displaced from the centre of runs of ATs

towards

DNA. It is apparent hotspots are in exactly

[125]Hoechst 33258 damage in intact

439

of [‘251]Hoechst 33258 in Intact Human Cells

sequence

for both

the

labelled

end of the

from Figure 1 that the the same posit’ion on the intact

cells and

purified

The presence of single-strand breaks and base damage in the cleavage caused by 1‘251]Hoechst 33258 was investigated by looking for damage in the 340 bp aRI-DNA band (from the agarose gel) since single-strand breaks are present in this fraction. Figure 1 shows the cellular 340 bp aRIDNA fractions with lane 1 (control) and lane 2 (incubated with [ ‘251]Hoechst 33258). DNA from equal numbers of cells was loaded in lanes 1, 2, 4 and 6. No damage can be detected above background even at 94 bp, which is the most intense hotspot for double-strand breaks. A densitometer analysis of [‘251]Hoechst 33258induced damage hotspots was made. Table 1 shows that

the

densitometer

measurements

relative

to

bp 94 (columns B) are similar at each hotspot for all samples. Each value is an average of two experiments and the duplicates ranged from 63% to 137% of the average values shown in Table 1. However, the densitometer measurements (columns A) are different in the various samples. The densitometer

value is a measure

of the extent

of damage caused by [ ‘251]Hoechst’ 33258. With only lo6 cells the densitometer measurements (columns A), and hence the extent) of damage, are much greater than with 2 x lo6 cells by a factor of 3.28 (kO.21 standard deviation), range 2.97 to 3.50. (Equal quantities of radioactivity are present in all samples.) However, there is no consistent relationship in the densitometer measurements (columns A) between

samples

with

different’

quantities

of

purified DNA present in the incubations. Comparing the densitometer measurements (columns A) for 2 x lo6 cells (which contain about

440

V. Murray

and R. F. Martin

AATTCTCAGTAACTTCCTTGTGTTGTGTGTATTCAACTCACAGAGTTGAA TTAAGAGTCATTGAAGGAACACAACACACATAAGTTGAGTGTCTCAACTT

50

CGATCCTTTACACAGAGCAGACTTGAAACACTCTTTTTGTGGAATTTGCA GCTAGGAAATGTGTCTCGTCTGAACTTTGTGAGAAAAACACCTTAAACGT

100

A

A

A

AGTGGAGATTTCAGCCGCTTTGAGGTCAATGGTAGAATAGGAAATATCT~ TCACCTCTAAAGTCGGCGAAACTCCAGTTACCATCTTATCCTTTATAGAA A

A

150

A

CCTATAGAAACTAGACAGAATGATTCTCAGAAACTCCTTTGTGATGTGTG GGATATCTTTGATCTGTCTTACTAAGAGTCTTTGAGGAAACACTACACAC

200

A

CGTTCAACTCACAGAGTTTAACCTTTCTTTTCATAGAGCAGTTAGGAAAC GCAAGTTGAGTGTCTCAAATTGGAAAGAAAAGTATCTCGTCAATCCTTTG

250

ACTCTGTTTGTAAAGTCTGCAAGTGGATATTCAGACCTCTTTGAGGCCTT TGAGACAAACATTTCAGACGTTCACCTATAAGTCTGGAGAAACTCCGGAA

300

CGTTGGAAACGGGATTTCTTCATATTATGCTAGACAGAAGAATT GCAACCTTTGCCCTAAAGAAGTATAATACGATCTGTCTTCTTAA

344

Figure 2. The sequence specificity of [ 1251]Hoechst 33258 damage to aRI-DNA in. intact cells and purified DXA. The “consensus” sequence of 340 bp aRI-DNA is depicted. Above the DNA sequence the damage on the ( + ) strand is shown with the (-) strand damage below. The peak of damage intensity is represented by a filled triangle.

Table 1 Intensity

of (12”I]Eloechst

33258 damaye at various

hotxyots

Position of hotspot (bp)

69 84

0.074

0.131

0.111

94 106 135 143 157

0.570 0.306 0.131 0.251 0.235

0.195 1.0

B. relative

densitometer

0.539 O-230 0.439 0.412

0.242 0.387

0.138 0.212

1.87

1.o

0.922 0.459 0.870 0.699

0.,502 0.247 0.465 0.374

0.389 0.702 2.70 1.42 0.801

1.09 om2

0.147 0.263

0.278 0.624

0.107 0.238

1.0

2.61

14

0.524 0.296 0.403 0.336

W845 0.600 0.686 0602

0,340 0.236 0.270 0.2”s-

o-277 0.388 2.16 0.873 0.446 0.502 0.289

measurement. the value in column A is divided by t,he densitometer

0.128 0.190

0~298 o.sf, I

O-143 (I.264

14 0.403 0.220 0.247 0.139

‘49 0.930 0439 0.397 0.295

I.0 0445 0.210 0.190 Will

measurement at bp 94.

Sequence Spec$city

of [‘251]Hoechst 33258 in Intact Human Cells

IO pg of DNA) and 10 pg of purified DNA, it is apparent that more damage occurred in the purified DNA sample by a factor of 5.04&O*%, range 3.84 to 6.32. However, a comparison between lo6 cells and 5 pg of purified DNA does not reveal any significant difference between the densitometer measurements. This is because the lo6 cells value is several-fold higher than the 2 x lo6 cells value, and the 5 and 10 pg values are approximately the same. As a control to determine the effect of -70°C and glycerol on the ligand damage, samples were incubated at 4°C with glycerol (Fig. 1, lane 10) and at 4°C without glycerol (lane 11). No significant differences in any parameter could be detected between these samples and that at -70°C with glycerol (lane 8).

4. Discussion In this paper we have shown that [‘251]Hoechst 33258 damages DNA in a similar manner in intact cells and purified DNA; there were no observable differences in position or extent of the damage for both systems. Using a novel method, for the first time the sequence specificity to the exact base-pair of a DNA damaging agent that causes doublestrand breaks was determined. This shows the advantages of using a method where a DNA damaging agent has been attached to a ligand of interest. For instance, the DNase I “footprinting” method cannot be used in intact cells. Use of “‘1 is also advantageous over other systems such as EDTA-Fe*+ conjugates, since it is very stable and does not adversely affect the cell permeability of Hoechst 33258. [‘251]Hoechst 33258 apparently causes only double-strand breaks, since we were unable to detect any significant single-strand breaks or base damage. Thus, it appears that the intense shower of low energy electrons emitted during “‘1 decay causes a double-strand cleavage and no detectable single-strand breaks or base damage at the binding site. Theoretical calculations by Charlton & Humm (1988) indicate that 2.7 to 12.8% (depending on assumptions) of ‘*‘I decays should only produce single-strand breaks. However, the limit of detection in our experiments was less than 2%, and hence is below the theoretical lower estimate. Charlton & Humm (1988) used “‘1 covalently bound to DNA in their calculations, whereas in our experiments [ ‘251]Hoechst 33258 is attached to DNA through non-covalent binding. The sites cleaved by [ ’ 251]Hoechst 33258, except at bp 69, are associated with at least four consecutive A. T base-pairs as found by Martin & Holmes (1983). At bp 69 there is no normal [1251]Hoechst, 33258 binding site. However, crRIDNA is not a completely homogeneous sequence but contains a small number of base alterations, an average of 9% probability at each base-pair. An analysis of 38 340.bp aRI-DNA sequences (,Jorgensen et al.. 1986; Murray & Martin. 1987)

441

revealed that 25 of the 38 sequences contained a T at bp 72. This substitution generates four consecutive A. T base-pairs next to the bp 69 hotspot, and thus would result in a significant [‘251]Hoechst 33258 damage site although it would be of diminished intensity. Nuclear DNA is in a very complex environment, since it is bound to nucleosomes and other proteins as well as being in higher-order chromosome structures. In addition to being supercoiled, the binding of proteins would be expected to cause conformational changes to the DNA which would affect the sequence specificity of ligand binding. However, the sequence specificity and the relative intensity of damage caused by [‘251]Hoechst 33258 is similar in both intact cells and purified DNA. (Of course more subtle effects could be present which are not detected in our experiments.) The only detectable difference between the two environments is that for 2 x lo6 cells less damage occurs (approximately 5-fold) than for 10 pg. Even this difference is not apparent at a lower cell density, lo6 cells compared to 5 pg DNA. The [‘251]Hoechst 33258 appears to be efficiently taken up by the cell (40%) and the DNA efficiently damaged by the ligand, since it has a high affinity for DNA even in the presence of cellular DNA binding proteins. The high degree of [ ‘251]Hoechst 33258 damage in intact cells is surprising. As a comparison, with bleomycin there was no detectable damage in intact cells compared to purified DNA at similar concentrations, and to obtain equivalent degrees of damage a 30-fold higher concentration of bleomycin was necessary when used with intact cells (Murray & Martin, 1985a). The efficiency of damage by [‘25T]Hoechst 33258 is also surprising in view of the data of Drew & Travers (1985). They used “statistical sequencing” and sequence-specific DNA ligands to determine the preferred periodicities of ligand binding to DNA from nucleosomal cores. They found that runs of G . C base-pairs were preferentially on the outside of the nucleosomal core particles while runs of A *T were preferentially placed where the minor groove of the DNA is on the inside of the nucleosomal core particle. They also confirmed this observation by sequencing 177 145-bp DNA sequences that bound to nucleosomal core particles (Satchwell et al., 1986). Thus, since [ “‘1 ] Hoechst 33258 preferentially binds to the minor groove of DNA in runs of four or more A. T base-pairs, it would seem surprising that the presence of a nucleosome binding to the same sequence as [‘251]Hoechst 33258 does not drastically reduce damage to the DNA. The damage caused by [‘251]Hoechst 33258 in intact cells did not give rise to any det,ectable level of bands at 170, 340, 510 bp, etc. on an agarose gel. These bands are characteristic of agents that cleave in the linker region of nucleosomes, for example bleomycin (Murray & Martin, 1985a) or micrococcal nuclease (Musich et al., 1982). This indicates that when the DNA is bound to the histone octamer, the

442

V. Murray

and R. I? Martin

sites of binding are as fully accessible to the [‘251]Hoechst 33258 as the linker region. Studies by Waring and co-workers (Low et al., 1986; Portugal & Waring, 1986, 1987) could provide an explanation for our observed data. They studied the effect of DNA minor groove binding ligands on DNase I cleavage of core nucleosomes reconstituted on tyrT DNA fragments. These ligands rotated the DNA by approximately half a turn with respect to the nucleosome octamer. Thus, [‘251]Hoechst 33258, which is a minor groove binding agent, assuming it has properties similar to the agents tested by Waring and co-workers, can bind to DNA in the cell nucleus by rotating the position of the DNA with respect to the nucleosome. Hence, this explanation could reconcile our data with that of Drew & Travers (1985). However, the concentration of DNA minor groove binding agents employed to observe this effect on nucleosomes is 10 to 20 PM, which is much higher than the concentration of [‘251]Hoechst 33258 used in our experiments. The sequence selectivity of three other DNA damaging agents has been determined in both intact cells and purified DNA using human aRIDNA as a target sequence. Nitrogen mustard irradia(Grunberg & Haseltine, 1980), ultraviolet tion (Lippke et al., 1981) and bleomycin (Murray & Martin, 1985a) all gave similar damage in both environments. In vitro, the extent of bleomycin damage can be affected by base substitutions significantly far removed from the damage site 1985c; V. Murray & R. F. (Murray & Martin, Martin, unpublished results). This effect is probably due to base substitutions affecting the microstructure of DNA at sequences far removed from the damage site. Bleomycin cleavage has been compared in supercoiled and linear plasmid sequences (Mirabelli et al., 1983) and a number of differences were found. All the above data suggest that neighbouring DNA sequences are a more important determinant of the extent of cleavage by various DNA damaging agents than the environment of DNA in the cell nucleus. It also gives confidence t,o extrapolation from results obtained in vitro with DNA damaging agents to intact cells. This system offers several future applications. [ ‘251]Hoechst 33258 causes only double-strand breaks, which have been implicated as the main method of cell killing after radiation exposure, and hence would allow a comparison of damage at the nucleotide level with biological effects. Also, if the method can be extended to single copy genes, [‘251]Hoechst 33258 can be used as a conformational probe of DNA structure at a particular gene so that conformational changes could be monitored in relation to control of gene expression. Edited

This project was supported by the Australian Research Grants Scheme and Research funds from t!he Petei MacCallum Cancer Institute.

References Charlton, D. E. & Booz, J. (1981). Radiat. Kes. 87, 10 23. Charlton, D. E. & Humm, ?J. L. (1988). Znt. .I. Radiat. Biol. In the press. Darling, S. M., Crampton. J. M. & Williamson, R. (1982). J. Mol. Biol. 154, 51-63. Dretzen, G., Bellard, M., Sassone-Corsi, 1’. bt (‘hambon. P. (1981). ilnul. Biochem. 112, 295~-298. Drew, H. R. & Travers. A. A. (1985). J. Mol. Biol. 186. 773-790. Grunberg, S. M. & Haseltinc. W. A. (1980). Proc. .W. Acad. Sci., U.S.A. 77. 6546-6550. Jorgensen, A. L.. Bostock, C. ,J. & Bak, A. I,. (1986). J. Mol. Biol. 187. 185-196. Krisch, R. E. & Sauri. C. ,J. (1975). fn/. J. Kudiut. Biol. 27, 5533560. Lippke, J. A., Gordon. 1,. K., Brash. D. E. $ Haseltinr. W. A. (1981). Proc. Nat. Acad. Sci., r’.S.A. 78. 338% 3392. Low. C. M. L., Drew. H. R. & Waring, Jl. .I. (1986). ~V’uci. Acids Res. 14. 6785-6801. Lozzio, C. B. & Lozzio, B. (1975). Blood, 45. 361- 334. Maniatis, T.. Fritsch, E. F. & Sambrook. J. (1982). Molecular Cloning: A Laboratory Manual, 11.380. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY. Martin, R. F. & Holmes. N. (1983). :Vatu,rr (London). 302, 452-454. Martin, R. F. & Pardee. -MM.(1985). fnt. J. Appl. Kudiat. Isot. 36, 745-747. Mikhailov, M. V.. Zasedatelev. A. S., tirylov. .\. s. & Gurskii. G. V. (1981). Mlol. Kid. (English translation). 15, 541 ~-553. Mirabelli, C. K., Huang. C.-H. & (‘rookr, S. ‘l’. ( 1!183). Biochemistry, 22, 300-306. Murray, V. & Martin. R. F. (1985a). J. Hiol. (‘hem. 260. 10389-10391. Murray, V. & Martin, R. F. (19856). G’ene AnuZ. Techn. 2, 95-99. Murray. V. & Martin, R. F. (1985c). &cl. Acids Res. 13. 1467-1481. Murray. V. & Martin. R. F. (1987). Gme. 57, 255mm259. Musich, P. R., Brown, F. L, & Maio. ,I. -1. (1982). Plot. Nat. Acad. Sci., U.S.A. 79, 118~~122. Pjura, P. E.. Grzeskowiak, K. & Dickerson. R. E. (1987). J. Mot. Riot. 197, 257-271. Portugal, J. & Waring. M. J. (1986). &cl. Acids Rrs. 14. 8735-8754. Portugal, J. &, Waring, M. J. (1987). ~Vucl. Acids ties. 15, 885-903. Satchwell. S., Drew. H. R. & Travers, 11. A. (1986). J. Mol. Biol. 191, 659-675. Wu. J. C. & Manuelidis, L. (1980). J. 1Mol. H%ol. 142. 3633 386.

by A. Klug