substrates in Xenopus laevis egg extracts - BioMedSearch

Download PDF
1 downloads 0 Views 2MB Size Report
Comment
Processes of DNA end joining are assumed to play a ... conditions to faithfully restore the original sequence at the break ... by complex reactions involving fill-in DNA synthesis, nucleotide ... contain longer nonpalindromic PSS-tails, damaged bases, abasic ... this case, the last incorporated nucleotide was (32P)-labelled to.
.. 1994 Oxford University Press

Nucleic Acids Research, 1994, Vol. 22, No. 9 1643-1650

Nonhomologous DNA end joining of synthetic hairpin substrates in Xenopus laevis egg extracts Nicole Beyert+, Susanne Reichenberger, Michael Peters§, Markus Hartung, Bernd Gottlich, Wolfgang Goedecke, Walter Vielmetter and Petra Pfeiffer* Institut fur Genetik der Universitat zu K6ln, Zulpicher Strasse 47, D-50674 K6ln, Germany Received January 21, 1994; Revised and Accepted March 24, 1994

ABSTRACT Processes of DNA end joining are assumed to play a major role in the elimination of DNA double-strand breaks (DSB) in higher eucaryotic cells. Linear plasmid molecules terminated by nonhomologous restriction ends are the typical substrates used in the analysis of joining mechanisms. However, due to their limited structural variability, DSB ends generated by restriction cleavage cover probably only part of the total spectrum of naturally occurring DSB termini. We therefore devised novel DNA substrates consisting of synthetic hairpin-shaped oligonucleotides which permit the construction of blunt ends and 5'- or 3'-protruding single-strands (PSS) of arbitrary sequence and length. These substrates were tested in extracts of Xenopus laevis eggs known to efficiently join linear plasmids bearing nonhomologous restriction termini (Pfeiffer and Vielmetter, 1988). Sequences of hairpin junctions indicate that the short hairpins are joined by the same mechanisms as the plasmid substrates. However, the bimolecular DNA end joining reaction was only detectable when both hairpin partners had a minimal duplex stem length of 27bp and their PSS-tails did not exceed 1Ont. INTRODUCTION Double strand breaks (DSB) in chromosomal DNA are potentially lethal lesions because they may cause irrevocable loss of essential genetic information (1-4). They may arise by a variety of cellular processes involved in the general DNA metabolism (5) or may be induced by exogenous DNA damaging agents like ionizing radiation (6,7). Elimination of this perilous damage may be achieved by mechanisms of homologous and illegitimate recombination. While in procaryotes and lower eucaryotes a balance between both recombination processes apparently exists and homologous recombination may predominate under certain conditions to faithfully restore the original sequence at the break (8,9) broken molecules in higher eucaryotes are mainly joined

end-to-end (5). These end joining mechanisms are able to convert nonhomologous DNA termini into covalently closed junctions by complex reactions involving fill-in DNA synthesis, nucleotide loss, and ligation. This often leads to sequence changes at the break sites which may include basepair substitutions, deletions and insertions. Such illegitimate joining events are potentially mutagenic and may lead to genomic rearrangements, genetic disease and cancer (11). However, the probability of such serious consequences is lower in multicellular organisms because the genomes of somatic cells are diploid and contain high proportions of nonessential DNA. Therefore, mechanisms of DNA end joining may be regarded as effective tools for the elimination of DSB and probably represent a major DSB repair pathway in higher eucaryotic cells as suggested by increasing experimental evidence (12,13). Mechanisms of DNA end joining have been investigated in detail in a variety of eucarytic in vivo (cultured mammalian cells, 13-16; Xenopus laevis eggs, 17; yeast, 9,18,19) and in vitro systems (extracts from Xenopus laevis eggs 20-22; nuclear extracts from human cells, 23-27) all of which were shown to be able to join DNA termini containing either blunt ends or short protruding single strands (PSS). Junctional sequence analysis revealed the existence of various different junction types with distinct features which led to the postulation of different pathways of DNA end joining (21,22,25,26,27). Since the molecular structures of naturally occurring DSB are largely unknown, joining assays employ plasmids linearized by restriction endonucleases (RE) to mimic broken DNA molecules with DSB ends of well defined sequence and structure. However, due to the dependence of RE cleavage on specific, mostly palindromic, recognition sequences the spectrum of RE generated DSB ends used for the analysis of joining mechanisms is necessarily limited with respect to structure and sequence. It is likely, that naturally occurring DSB, generated e.g. by radiation or chemical exposure, are more complicated in that termini would contain longer nonpalindromic PSS-tails, damaged bases, abasic deoxyribosyl units, damaged sugars, and bulky adducts. To overcome the restrictions of RE generated DSB we have designed

*To whom correspondence should be addressed Present addresses: +Institut fir Biotechnologie, Forschungszentrum Julich GmbH, D-52425 Julich and §Minden Phanna, Karlstrasse 42-44, D-32423 Minden,

Germany

1644 Nucleic Acids Research, 1994, Vol. 22, No. 9 novel artificial substrate types based on synthetic hairpin-shaped oligonucleotides carrying PSS tails of freely variable polarity, length and sequence. Using a modified joining assay, these substrates were employed to explore the capacities of nonhomologous DNA end joining in X. laevis egg extracts (20). We found that for successful DSB rejoining a minimal DNA duplex length of 27bp is required and that PSS-tails longer than lOnt are not joined any more.

MATERIALS AND METHODS Xenopus laevis egg extracts Extracts from X. 1aevis

eggs were

prepared

as

described

previously (28). Hairpin oligonucleotide substrates Oligonucleotides used for substrate preparation were synthesized on a Gene Assembler (Pharmacia) in our lab, desalted by elution over NAP-10 columns (Pharmacia) and purified by electrophoresis on preparative 10% sequencing gels (7M urea; 19:1 acryamide: bisacrylamide; Serva) to remove shorter chain termination products. All hairpin shaped oligonucleotides were assembled from (i) a constant 'stem oligo' (SO) and a variable 'linker oligo' (LO) (Fig. 1). For 5'-PSS-hairpin substrates the recessed 5 '-OH end of the SO was (32P)-phosphorylated and its ss 3 '-overhang was hybridized to an appropriate LO. The remaining nick was ligated with T4 DNA ligase to yield a 5'-hairpin with an internal phosphate label and a 5'-PSS tail which is defined by the 5'-extension of the LO (Fig. lA). Ligation products were purified from unreacted input oligos by electrophoresis on preparative 10% sequencing gels. Thereafter, 5'-PSS-tails were phosphorylated with cold ATP. Hairpin substrates with blunt ends were generated from unlabelled 5'-hairpins by fill-in of the 5'-PSS with Klenow polymerase. In this case, the last incorporated nucleotide was (32P)-labelled to verify that fill-in DNA synthesis was complete (only these hairpins are blunt and thus visible by autoradiography) (Fig. 1B). For preparation of hairpin substrates with 3'-PSS the LO were 5'(32P)-phosphorylated and ligated to the recessed 3'-end of the approriate SO (Fig. IC). The recessed 5'-OH ends of the resulting gel purified 3'-hairpin substrate were also phosphorylated with cold ATP.

Phosphorylation and ligation protocols for hairpin substrate preparation 25pMoles of SO (or LO) resuspended in a volume of 241tl containing 85mM imidazole-Cl pH6.7; 30.6mM MgCl2; 8.5mM DTT; 10% (w/v) PEG 6,000 (Sigma) were boiled for 3min and immediately cooled on ice. After addition of 10/iCi (32P)-'y-ATP (5,00OCi/mMole; Amersham) the solution was adjusted to 40Il with water and incubated with 15U T4 polynucleotide kinase (Epicentre) for 35min at 37°C. The reaction was stopped by addition of 241 of O.5M EDTA and 6O0,u of water and incubation for 10min at 65 °C. After extraction with chloroform the solution was adjusted to 4M LiCl and precipitated with 2.5 vol. ethanol for lh on ice. Pellets were resuspended in 30Ou water, mixed with 150pMoles LO (or lOOpMoles SO) and adjusted to 75y1 with water. Annealing of the oligos was performed in the thermocycler (10min 65°C; 10min 50°C; 10min 40°C; 10min 18°C). Subsequently, samples were cooled on ice for 5min, adjusted to ligation conditions (5OmM Tris HCI pH7.6; -

10mM MgC12; 0.2mg/ml BSA; 10mM DTT; 70MM ATP) and incubated with 5 Weiss-U of T4 DNA ligase (US-Biochemicals) in a total volume of 100t1 at 16°C over night. After phenol extraction, total ligation samples were electrophoresed in preparative 10% sequencing gels. Ligation products were visualized by autoradiography, excised and eluted by the diffusion method. After ethanol precipitation, hairpin substrates were subjected to another kinase reaction with cold ATP (same reaction conditions as above) to phosphorylate the remaining 5 '-OH ends. Chloroform extracted ethanol precipitated hairpin substrates were resuspended in water at 50,000 cpm/,ul (approx. 5fMoles/pi). Since the internal (32P)-phosphate label tends to damage DNA molecules hairpin substrates were employed in extract joining reactions immediately after preparation and were usually stored for less than two weeks. Thus the possible occurrence of artifact bands resulting from damaged DNA was kept minimal. Extract joining and T4-DNA ligase assays and analysis of reaction products 1fMole (corresponding to 10,000Ci) of each of two different hairpin substrates or 2fMoles of one hairpin substrate in 241 water were incubated with l0,ll of Xenopus laevis egg extract at 13°C for 90min. For preparative reactions, 1OfMoles of hairpin substrates in 21 water were incubated with 10,ul egg extract under the same conditions. For T4-ligation assays, the same amounts of hairpin substrates were incubated in a total volume of IOtd under ligation conditions (see above) with lWeiss-U of T4-DNA ligase for 90min at 13°C. Reactions were stopped by addition of 300td stop mix (20mM Tris.HCl pH 7.6; 300mM NaCl; 10mM EDTA; 1% (w/v) SDS; lmg/ml proteinase K, Sigma) and incubated at 65°C for 30min. After phenol extraction and ethanol precipitation, pellets were resuspended in water. Joined products were either directly separated in 10% sequencing gels or first digested with appropriate restriction enzymes which cleave the dumbell shaped joined products to yield diagnostic junctional restriction fragments which were analysed in 12% sequencing gels.

PCR amplification and sequencing of joined products Total purified preparative joining samples (approx. 1OfMoles of 5'-hairpin substrates) which had been treated with RNaseA (US Biochemicals) were digested with SmaI (Boehringer) to cleave off the hairpin loops from the dumbell-shaped joined products to generate linear double stranded DNA fragments suitable for PCR amplification (compare Fig.2A). Since these fragments were too short to be amplified efficiently (the longest SmaI fragment derived from the 5'-hairpins combination B-72/S-82 is only 69bp), primers with nonhybridizing 5'-extensions were used in the PCR reaction to generate 141bp PCR fragments. The forward primer (PF) used for the amplification of the 5'-3' strand carried a 35nt 5'-extension (5'-CCAGTCGGGA AACCTGATCC GCTCAAGC TG TTTCCGTCGA CACAGTCTCG AGGC, the annealing portion is underlined). The reverse primer (PR) carried a 40nt 5'-extension (5'-CGCCCGCCGC GCCCCGCGCC CGTCCCGCCG CCCCCGCCCC GGAACATCTA GAA TTCC). SmaI digested joining samples were phenol extracted, precipitated with ethanol and dialysed for 2h on discfilters (0.025,u; Millipore) against water. PCR amplification was performed over 40 cycles (1.5min 94°C; 1.5min 54°C; 0.5min 59°C; 2min 72°C) with 130ng of each PCR primer (PF and PR) in a total volume of l00y1 containing 50mM KCl; 10mM Tris HCl pH8.4; 25rmM MgCl2; 0.01 % (w/v) gelatine; 0.001 %

Nucleic Acids Research, 1994, Vol. 22, No. 9 1645 (v/v) Tween 20; 0.001 % (v/v) NP40; 20,uM of each dNTP and 5U Taq DNA polymerase (Beckman) and yielded SpMoles of the 140bp junctional PCR fragment. Sequencing of the short 141bp PCR products was facilitated by preparation of ss templates. This was done by using pF in 5'-biotinylated form to yield a junctional PCR product whose 5'-3'-strand carried a 5'-biotin and thus could be coupled to streptavidin coated magnetic beads (dynabeads, Dynal). Strand separation was performed according to the supplier's protocol to release the uncoupled 3'-5'-strand into the aqueous phase to yield lpMole of ss template sufficient for sequence analysis using a T7 DNA polymerase sequencing kit (Pharmacia). To prevent formation of secondary structures the ss template was boiled for 5min and stepwise annealed (5min 65°C; 10min 37°C; 5min 22°C; 2min ice) with a 10-20 fold molar excess of 5'-(32P)-phosphorylated sequencing primer (ps: 5'-AGTCGGGAAA CCTGATCCG). The labelling mix was omitted, and instead of 4.5,u1, 3.51d of the template/primer/polymerase mixture were added to the reactions containing short run ddNTP mixes.

two complementary terminal bases to from an overlap structure whose gaps are subsequently filled-in (Fig.2A; 22). Intermolecular joining of two hairpins yields dumbell-shaped products (Fig.2A). Due to the palindromy of the chosen

A) 5-hairpin substrates (52, 62, 72, 82nt; 22, 27, 32, 37bp double-strand) 5'-SO-37

CCCGGGAGTCCTCTAGATCTGC)3 (TGGGCCCTCAGG5s 'TI

5'-0-52 5'-S-52

5-L-Bam

5'-LO-Sal

GATCTAGACGCTA5'-p 3 lAGATCTAGACGACC s5-P 3

5'-SO-45 fT CCCGGGTACCAGGCCTGCAGATCTACG3

5'-0-62 5'-S-62

TT GGGCC CATGGTCCGls 5'-L0-Bam 3 As-TCTAGATGCCTAl5-p 3 T G CA A C GACls -P 5'-L0-Sal

5'-SO-57 TTCC CGG GAACATC TAGAATTC CCGGTC6CGAT6]3 TT 5-LO-Bam 3 jCCA6CGC T5ACCTA5-P 5'-8-72 3' 6CCAi6C6CTACAG IJ5s-p 5'-LO-Sal 5'-S-72

GGGCCCCTTGTAGATCTTAA G6s

-

-

-

5'-SO-65 (T >CCCGGGTCGACACAGTCTC6A66CCT6CA6ATCTAC 3

VT

RESULTS Design of synthetic hairpins as substrates for DNA end joining DNA substrates providing DSB ends of freely variable structure and sequence were designed to have the shape of hairpins to be assembled from two synthetic oligonucleotide portions, (i) a constant 'stem oligo' (SO) and (ii) a variable 'linker oligo' (LO). The SO can fold back to form hairpins with short duplex stems carrying a loop of 4 thymidines on one side and an 11 or 13nt long ss overhang on the other side serving for hybridization with a complementary LO whose nonhybridizing 5'- or 3'-extension defines the polarity, length and sequence of the PSS tail of the final hairpin substrate (Fig. IA,C). The 5'-hydroxyl at the nick between SO and LO was (32P)-phosphorylated and incorporated into the duplex stem by ligation to provide an internal phosphatase resistant radiolabel. In this way, eight hairpins with duplex stems of different lengths (22, 27, 32, 37bp; T-loop and PSS-tail not counted) and freely variable 5'- or 3'-PSS tails were generated (Fig. 1 A,C). This type of substrate exhibits an exonuclease resistant and joining inactive hairpin loop on one side and an open reactive PSS- or blunt end on the other side. Consequently, two hairpin partners are required to be linked via their joining active ends in the bimolecular joining reaction. The short lengths of the substrates and their resulting dumbell shaped products allows separation by the sensitive denaturing PAGE technique to detect even smallest changes in size. Restriction analysis of the reaction products is facilitated by RE-sites contained within the hairpin sequence (Fig.2A). Joining of hairpin substrates with 5'- and 3'-PSS in the Xenopus laevis in vitro system To test the suitability of the hairpin oligonucleotides for the investigation of joining activity by direct comparison with the previously used RE-linearized plasmid substrates we constructed hairpin substrates whose PSS mimicked the 'sticky' ends generated by BamHI (B) and SalI (S) cleavage (B: 5'-GATC-3'; S: 5'-TCGA-3'; Fig. 1). The nonhomologous Bam/Sal terminus configuration was chosen because it had been previously shown to be efficiently joined in the Xenopus in vitro system (20,22). Junction formation was found to proceed by the 'overlap mode in which the two antiparallel PSS-tails overlap each other via the

GGGCCCAGCT6TGTCA6A6CACC6F57* 5'-L0-Bam 5'-LO-Sal

5'-0-82 5'-S-82

5'-LO for long PSS (n=2,4,7,8,9 10,13,18)

3 AC6TCTA6AT6CIA 51-P 3 l6AC6TCTA6AT6C.G ls1--p

3li AC 6 TC TAGAAT

B) fill-in of 5-hairpins yields blunt ends (56, 66, 76, 86nt; 26, 31, 36, 4lbp) 5'-S-52

5'-B-52 (T.CCCGG GAGTCCTCTAGATCTGCIDDD>

C TG CIFD> \TTGGGCCCTC AGGIA6ATCTAGAC6C TAGs-PG6ACGAGC6C-p A 3 C TGiTC GAC.CTAG 5-P a-56

CTGCTCS 3 GACGAGC 5-p

S-56

C) 3-hairpin substrates (52, 62, 72, 82nt; 22, 27, 32, 37bp dou-strnd) 3-0-52 3-S-52

3'-LO-Bam *5 GC6A 3-LO-Sal s i CC TCGC A AC T

3

TTGGGCC CTC AGCCGGAGCGCTTGs5-P 3'-SO0-37 3-0-62

3'-L0-Bam *s5CCTGCAGATCTC6AT 3

3'-S-62

3'-LO-Sal *s CCTGCAGATCTCTC 43 TTC CCCGGGTACCTCAG|3 TT GGGCCCATGGAGTCCGGACGTCTA6AGs5-P

3'-S0-45

3-0-72 3-S-72

*s5CCTCGCGAACGAR3

3-LO-Bam 3-LO-Sal sICCTC6CGAACTC613~ TT CC CGGGTTG TTCCTAGAATTCG|3 KT1iGGGCCCAACAAGATCTTAAGCCGGAGCGCTTGs5-p

3'-S0-57

3'-L0 for long PSS (n=2,4,7,8,9, *5s C C T G C A G A T C T C A

3

10,13,18) 3'-L0-Bam *s CCTGCAGATCTC6AT 3 3'-L0-Sal *5 3GCAGATCTCTC63

3-0-82 3-S-82

1T CCCGGGTCGACACAGTCTCGAGAG(3 (1..GGGCCCAGCTGTGTCAGAGCTCTCCGGACGTCTAGAGIs-P 3'-SO-65

Figure 1. Survey over the assembly of the sixteen different hairpin substrates with 5'- (A) or 3'-PSS (C). Bold letters in the LO indicate the sequence of the resulting Bam (B)- or Sal (S)-PSS. Hairpins with long PSS-tails [(A)nCG] were prepared from the longest SO (65nt) (bottom of panels A and C). Asterisks at the recessed 5'-ends of SO (A) or 5'-ends of LO (C) mark the positions of 5'-(32P)-phosphate labels which are incorporated into the hairpin stem by ligation of the two oligo portions. 5'-P indicates the cold phosphate which is transferred to the 5'-OH-end of the finished hairpin substrate in the last kinase reaction. Preparation of hairpin substrates with blunt ends is exemplified in panel (B) for the shortest hairpins. Fill-in of the 5'-PSS of appropriate unlabelled hairpins is symbolized by open triangles. Open letters indicate the newly synthesized sequence in the resulting blunt ends. The asterisk marks the site of the 5 '-2P)-phosphate label resulting from incorporation of a labelled nucleotide at the terminal position.

1646 Nucleic Acids Research, 1994, Vol. 22, No. 9 A)

Nonhomologous DNA end joining of two 3-hairpins 3'-Sal-72

3'-Bam-62 CCCGG

rT GGGCICC6iGI6 #

6

M3-P-5§I

I6IA1i3P 27bp -

~-

C6

T,

66CC CTTT

3

32bp

L

-

heterodimeric dumbell-shaped joined product (138nt) mV AT S

a

#^ hAA ~~~65bp^ 59bp

B) Expected fito and jig pvoxts of 5 - aW 3-hirpin uwh he- u Sal-PSS uW tI Smul-hfagnt 52/52 L J

52/62 L J

52/72 L J

52/82 L J

62/62

LJ

62/72 L

62/82

L

72/82 L J

72/72 L J

82/82 L J

52/52 104 105 52/62 104

52/72 104

(114) 118

124

(124)125

52/82 104

144

(1341135

164

62/62

124 12

62/72

124

62/82

124

72/72

52 5 52

52

(134)13

144

575

62

(144"1

164

625"I

72

o5aoSlr !':rOPf?fii ng

-

-t

,. -

(144)145 5 62

72182

144

(154"15 1764

72/82 82/82

144

(154)151

{

1164 6~ ~ ~ 72 41" 1

Figure 2. (A) Schematic drawing of the joining reaction of two hairpin substrates with nonhomologous 3'-PSS (B-62/S-72). The total lengths of the hairpins are given in nt and the lengths of the duplex portions in bp. The positions of the internal (32P)-phosphates are marked by asterisks and the SmaI-site used in analytical studies to cleave off the hairpin loops is also indicated. White on black letters in the PSS-tails symbolize the basematches used for overlap formation between the Bam- and the Sal-PSS as described previously (20). Open letters in the junction of the final dumbell-shaped product indicate the sequences resulting from fill-in of the small gaps of the overlap intermediate. (B) Expected sizes of dumbell-shaped reaction products (in nt) and their corresponding SmaI-fragments (in bp, small numerals) resulting from sticky end-ligation (L, normally typed numerals without or in brackets) and nonhomologous DNA end joining (J, bold italic numerals) of all possible hairpin combinations (left column) in the Xenopus egg extract. The SmaI-fragments expected to result from reactions involving 52-mers are not given because these hairpins did not yield any reaction products. As an example how the table is to be read, see combination 62/72 which represents four possible combinations (B-62/B-72, B-62/S-72, S-62/S-72, S-62/B-72). In all four combinations, the two 62-mers can form short sticky-end ligated homodimers of 124nt whose SmaI-fragments are 52bp. In the two combinations containing hairpins with equal PSS (B-62/B-72, S-62/S-72), the 62-mers can be sticky-end ligated with the longer hairpins to yield heterodimers of 134nt and a 57bp SmaI-fragment. In the other two combinations involving hairpins with nonhomologous PSS (B-62/S-72, S-62/B-72), heterodimers of 138nt resulting from nonhomologous DNA end joining are formed whose SmaI-fragments are 59bp. Note that these heterodimers are longer by 4nt than the heterodimeric sticky endligated products which is due to gap filling of the overlap intermediate as depicted in panel A. Ligation converts the 72-mers of all four combinations into long homodimers of 144nt yielding 62bp SmaI-fragments.

restriction ends, products resulting not only from nonhomologous DNA end joining (B/S) but also from sticky end ligation (B/B; S/S) are generated. To distinguish between these possibilities hairpin pairs of different sizes were employed to form short and long homodimers indicative of sticky end ligation and heterodimers of intermediate size indicative of nonhomologous DNA end joining. As a typical example for the joining of hairpins in X. laevis egg extracts, denaturing PAGE analysis of the reaction products derived from all possible combinations of the eight hairpins carrying 3'-Bam and Sal-PSS are shown in Fig.3A,B. In agreement with previous results obtained with plasmid substrates

Figure 3. (A,B) Autoradiographs of denaturing 10% PAGE separations of the products formed by hairpin substrates with 3'-Bam (B)- and 3-Sal (S)- PSS in extract joining reactions. The combinations of the hairpins (sizes and PSS-type) present in each lane are given at the bottom of the gels. Due to the denaturing gel conditions all DNA molecules run as single-strands. The sizes of the hairpins (HP) and their products which exist as open and closed dumbells and run as linear (LSS) and circular single-strands (CSS) are given in nt on the left and right side. Bold italic numerals with an asterisk mark nonhomologously joined LSS products. The control lane (+) on the left side of each panel displays the unreacted hairpin input substrates. In most lanes, additional bands are seen which occurred only after incubation with egg extract (or T4-DNA ligase, data not shown) but whose origin is not clear. Since these bands are completely absent in the unreacted input substrate it is unlikely that they are related to secondary structures with different gel migration properties or to products of DNA damage resulting from the (32p)_ phosphate label. Obviously, they vary in size in dependence of the hairpin length and its PSS-sequence and therefore could be possibly attributed to band shifts caused by residues of DNA binding proteins remaining on the DNA after phenol extraction. (C) Summary of the data obtained from extract joining reactions of all sixteen hairpins carrying PSS (see Fig. I A,C). Since the results were equivalent for 5'- and 3'-PSS no discrimination between the two types is made in this table. Plus-signs indicate that formation of ligation (L) or joining (J) products was detected whereas the minus-signs represent absence of product formation in the indicated hairpin combinations. Increasing efficiency of product formation is indicated by two plus-signs and hardly detectable amounts of product by a plus in brackets.

we found intermediate reaction products in the form of nicked or gapped dumbells which correspond to the monomeric open circular (oc) plasmid intermediates (21) and were further reacted to covalently closed dumbells as the final products corresponding to the covalently closed circular (ccc) plasmid monomers. The denaturing gel conditions convert these products into faster migrating linear single-strands (LSS) which form sharp distinct bands and more retarded circular single-strands (CSS) which form a diffuse smear, probably due to incomplete denaturation of the closed dumbells. Evidence that the CSS bands in fact represent the final reaction products was derived from the size correlation between the LSS and CSS-bands which is clearly seen in Fig.3A: for instance, the CSS-material produced by the 62-mer hairpins

Nucleic Acids Research, 1994, Vol. 22, No. 9 1647 A)

B)

I

css

css ,4

LSS ' 1;4;

",4

LS. ".4;'l

"5

I

.13J ';'4

4

'4 -0

0

HP ,p, .t

HP

;'2 h72

'I.) ,.t

69

62

64 59

"I

:2

5,' IH

8

8?

F

R

.R.

0.

;t

4w-

R.

72

f,;'

;Li4 T4

9 .E

U%NA

p q 4 s;. e,

r:

#lw Om?__m~ BP /2

162

+ -4-

B

S RS R 3S

+B 3S3

S B

3BS

Rc

95

Figure 4. If not otherwise indicated, the labelling of both gels corresponds to that described in the legend of Fig.3A,B. (A) Denaturing 10% PAGE separation of the products obtained from sticky end-ligation of hairpins carrying 5'-BamPSS with T4-DNA ligase. (B) Restriction analysis of the products formed by sticky end-ligation and nonhomologous end joining of the 62, 72, 82nt hairpins with 5'-Bam (B)- and -Sal (S)- PSS in the Xenopus egg extract separated in a denaturing 12 % PAGE. The left half of the gel displays the bands of the extract joining reaction in undigested form, the right half the SmaI-restriction pattern. Unreacted hairpin substrates (labelled with '+') in undigested (left lane) and digested form (lanes in the middle) serve as size markers. SmaI-cleavage was never complete and left uncut and partially cut material. Restriction fragment sizes and their origin (HP for hairpin input, L for ligation product, asterisk for nonhomologous joining product) are indicated on the right side. The SmaIfragments of some ligation products coincide with the bands of uncut hairpin input material (L/HP).

migrates faster than the CSS-material of 72-mer and 82-mer hairpins. Furthermore, kinetic studies (data not shown) confirmed that the LSS-bands always occurred as the first products and gradually yielded increasing amounts of CSS-material indicating that the joining reaction is a two step reaction in which intermediate LSS-products containing one closed strand are presumably further reacted in a separate second step in which the second strand of the fmal CSS-product is closed. Equivalent results were obtained previously with plasmid substrates (29). In contrast to the plasmid substrates, however, the yield of the final CSS-products mostly remained below the yield of the LSSproducts which might be due to incomplete cold phosphorylation (see Materials and Methods) and/or possible structural contstraints of the hairpin substrates. Due to the fact that CSS-products were not formed efficiently in all experiments and could not be resolved to yield distinct bands the following data were mainly derived from the analysis of the partially rejoined LSS-products. Successful joining of hairpins depends on the length of the duplex stem Except for the 52mers which completely failed to form neither the expected 104nt homodimer nor any of the three possible heterodimers (118, 128, 138nt compare Fig.2B) all other hairpins readily formed LSS- and CSS-products (Fig.3A,B). If nonhomologous joining of the Bam/Sal terminus configuration

occurred by overlap formation as schematically drawn in Fig.2A, the joined product should be longer by 4nt than the corresponding ligation product (e.g. B-62/S-72 or S-62/B-72 = 138nt whereas B-62/B-72 or S-62/S-72 = 134nt; compare Fig.2B). This size difference is seen in the LSS-fractions of Fig.3B: the heterodimers formed for instance between the 62nt- and 72nt-hairpins bearing nonhomologous PSS (B/S or S/B) migrate slightly behind the heterodimers formed between hairpins with complementary PSS (B/B or S/S). Equivalent results were obtained in joining reactions with 5'-hairpins (data not shown). This indicates that the Xenopus in vitro system is able to form genuine joining products from combinations of hairpins with nonhomologous PSS and ligation products from hairpins with complementary PSS (Fig.3C). These results were confirmed by experiments in which hairpins were joined to linear 3kb plasmid substrates (data not shown). While the 62-, 72- and 82-mers were readily joined to plasmid molecules carrying nonhomologous ends, the 52-mers did not react at all suggesting that these shortest substrates might represent a size limit for DNA end joining reactions in the Xenopus in vitro system. To verify that the remarkable failure of the 52nt haipins to be ligated or joined was due to special features of the extract joining system and not to a defect in the hairpins we subjected these substrates to ligation with T4-DNA ligase. The resulting band patterns are exemplified in Fig.4A for hairpins with 5'-Bam-PSS. Ligation of two B-52 with T4-DNA ligase yielded the expected homodimers of 104nt while ligation of the 52mer with longer hairpins yielded additional heterodimers (114, 124, 134nt). Although the low overall yield of the 104nt homodimer LSSproduct and the absence of CSS-products in the reactions containing 52-mer and 62-mer hairpins indicate that structure and/or sequence properties of these hairpins may influence ligation efficiency these results suggest that the observed complete failure of the 52mer hairpins to be joined in the egg extract is related to intrinsic properties of the joining system. We therefore conclude from these experiments that successful ligation of complementary ends as well as joining of nonhomologous ends in the X. laevis extract system requires that both hairpin partners are at least 62nt long. This size limit corresponds to a duplex stem length of 27bp. Analysis of Bam/Sal joining products by restriction cleavage and dideoxy-sequencing Using plasmid substrates with the Bam/Sal terminus configuration we had previously shown that joining in the Xenopus in vitro system proceeds by overlap formation via the two terminal basematches (20,22; see Fig.2A). To analyse whether nonhomologous DNA end joining of Barn- and Sal-hairpins would create the same junctions as the corresponding plasmid substrates we subjected the joined products to restriction cleavage and dideoxy sequencing. Fig.4B shows the results of restriction analysis using SmaI which cleaves off the hairpin loops and is expected to generate fragments of 59, 64 and 69nt from nonhomologously joined heterodimers and 52, 57, 62, 67 and 72nt from ligated homodimers (see Fig.2B). Although hairpins and dumbells were always incompletely digested (probably due to a destabilization of the DNA duplex stem at the loop structure) all bands expected to be generated by nonhomologous DNA end joining can be unambiguously identified. The junction resulting from joining of the 5'-hairpin combination B-72/S-82 was sujete %t diex-sqec ngAn verirfied to cornt-ain the sequence (5'-GATCGA) which is expected to arise by overlap

1648 Nucleic Acids Research, 1994, Vol. 22, No. 9 A)

nfl SUqsof mf.pwatWdbyflmnf 5sPSS

A

O r nF JJI tI (I ...f.Iwc

.&

a51 prr.

-

*

l.

to' 5. -

a

ATC

TCT CIA SAT%CBA TAQCTAOCTAC ljCT Barm"n Bam"

B)

gBam" "Sal"

Sl. Sal

ULimp of Mir wks h Mm us B 56 B 66 B 76 B 86 S 56 S 66 S 76 S 86

(26bp)(31bp)(36bp)(41bp (26bp)(31bp)(36bp)(4lbp) 8 56 (26bp)

-

+

++

-

-

+

-

6 66 (31bp)

+

+

++ ++

+

+

+

+

8 76 (36bp) ++

++

++ ++

++

+

+

+

8 86 (41bp)

+

+

+

++

++ ++

++

+

+

S 56 (26bp) -

+

++ ++

-

-

(+)

-

S 66 (31bp)

-

+

++

+

-

-

(+)

(4)

S 76 (36bp) +

+

+

+

(+)

(4)

(+)

(+)

S86 (41bp)

+

+

+

-

-

(+)

(4)

-

Bi.

|-r I:r

Figure 5. (A) Illustaion of the thre combinations of blunt ends (compare Fig. IB). (B) Efficiency of blunt end ligation in the Xenopus egg extract in dependence of hairpin length and sequence of the blunt end (for meaning of plus- and minussigns see legend of Fig.3C).

formation (compare Fig.2A). From these data we conclude that the joining of hairpin substrates proceeds by the same mechanisms as previously described for the joining of plasmid molecules (22).

Joining of hairpins with blunt ends Hairpins with blunt ends were prepared from unlabelled 5'-hairpins by fill-in of the 5'-Bam or Sal-PSS with Klenow polymerase. In these cases, the nucleotide incorporated at the ultimate PSS position was radiolabelled (Fig. 1B). The resulting hairpins were 4nt longer than the original hairpins (B-56, B-66, B-76, B-86; S-56, S-66, S-76, S-86) and subjected to extract joining assays in any possible combination. The overall efficiency of the blunt end ligation was lower than that of the sticky end ligation obtained with PSS-containing hairpins and the results were more complex (Fig.5B). Again, the shortest hairpins (56nt) failed to undergo blunt end ligation with each other. However, in contrast to the 4nt shorter hairpins bearing 5'- and 3'-PSS (= 22bp), some of the 56-mers (= 26bp) were joined to 66-mers (= 31bp) and longer hairpins (Fig.5B). These data confirm the results obtained with the PSS beaing hairpins and allow to further narrow down the minimally accepted duplex length to 26bp meaning that the bimolecular joining reaction can take place if one partner is longer than 26bp. Inspection of the data in Fig.5B furthermore reveals that besides the duplex length other factors influenced the efficiency of the blunt end ligation which was highest in B/B-combinations, intermediate in B/S-combinations and lowest in S/S-combinations. Since the duplex sequences of the hairpins were identical in all combinations except for the four terminal basepairs the observed differences are probably due to different terminal duplex stabilities and/or the natures of the ultimate basepairs to be joined to each other (Fig.5A). The same differences were even more pronounced in control reactions with T4-DNA ligase (data not shown). A similar phenomenon which could also be attributed to different stabilities of the DNA termini was observed for sticky end ligation of 5'-PSS where B/B was always more efficiently ligated than S/S (data not shown). These results show that except

-

+

- - -r

Fiue 6. (A) Schematic drawing of the joining reacon of two equal 37bp hairpins carrying long 3'-PSS-tails. In analogy to the mechanisms previously established for plasmid substrates with short PSS (20) joining of these hairpins is assumed to proceed by overlap formation between the two terminal basematches and gap filling primed at the 3'-PSS of the partner hairpin (solid triangles). (B) Denabtuing 10% PAGE separation of the reaction products arising from end joining of 37bp haixpins with long 3'-PSS-tails. The sequences of the PSS [(A)nCG] do not permit simple ligation as seen by the absence of products in the T4 DNA ligase control in the left half of the gel. Control lanes (c) contain the B-72/S-82 hairpin combination which is converted into two homodimeric sticky end ligation products of 144 and 164nt and one heterodimeric nonhomologously joined product of 138nt (only in the extract joining reaction). All other lanes contain a single 37bp hairpin type whose PSS-length is indicated by a plus-sign. Sizes of the input hairpins and LSS products are indicated on the right side. (C) Summary of the results obtained from extract joining reactions with 37bp hairpins carrying long 5'- or 3'-PSS-tails. Lengths of the PSS-tails, the corresponding hairpins and expected joining products are given in nt. For meaning of plus-signs see legend of Fig.3C.

for the length of the substrates to be joined the terminal sequences markedly influence the efficiencies of ligation and joining reactions. PSS tails longer than lOnt are not accepted by the joining system The maximal PSS length accepted by the joining system was determined by use of the 37bp hairpins supplemented with 5'and 3'-PSS tails ranging in size from 4 to 20nt (Fig. 1). All PSStails contained varying numbers of A-residues (dA)2 to (dA)8

and a terminal 5'-GC-3' to provide comparable overlap conditions in all cases (Fig. 6A). Since these PSS-tails are not palindromic no ligation products can be formed which could confuse the results. Thus, the homodimers generated by each hairpin substrate

Nucleic Acids Research, 1994, Vol. 22, No. 9 1649 with itself are indicative of nonhomologous DNA end joining. Fig.6B exemplifies the reaction of 37bp hairpins bearing 3'-PSS tails of growing length. While 3'-PSS of 4 and 6nt were efficiently joined to form LSS- and CSS-products of 168 and 172nt, joining of the 9nt and lOnt tails was about 10 to 20 fold less efficient and yielded only LSS-products (178 and 180nt). Joining of the 1 int tail was only detectable after extensive overexposure of the autoradiograph. For hairpins carrying long 5'-PSS-tails the maximal accepted length was established to be 9nt (Fig.6C). We therefore conclude that a PSS length of lOnt represents another size limit for nonhomologous DNA end joining in the Xenopus in vitro system. It has, however, to be taken into account that the sequence and structural properties of the monotonous polyA-stretches may influence the joining reaction. Further experiments will show whether this size limit will also hold for PSS-tails with mixed sequences. The failure of the longer 5'-PSS to be joined was accompanied by an exonucleolytic resection which shortened all 5'-PSS-tails longer than lOnt by exactly lOnt. An equivalent degradation process was also observed in joining reactions involving 3'-PSS as seen in Fig.6B by the bands migrating faster than the input substrate molecules. These fragments only occurred after incubation with Xenopus egg extract and were reproducibly 1Ont shorter than the unreacted input. This unusual nucleolytic process which consistently removes discrete lO-mer oligonucleotides from 5 '-protruding or -recessed ends will be described elsewhere in more detail (30).

DISCUSSION Until recently, mechanisms of nonhomologous DNA end joining have been mainly investigated with the help of DNA substrates based on plasmid molecules containing RE induced DSB. One major disadvantage of this approach is the limited variability of the resulting DSB termini with regard to their structure and sequence. We therefore developed novel hairpin shaped substrate types which permit free variation of these parameters and therefore are expected to be versatile for many applications in cell-free joining-systems. The concept to construct the hairpin substrates from two synthetic oligonucleotide building blocks is especially advantageous: once synthesized, an SO of appropriate size and sequence can be used to build a large variety of hairpin substrates by assembly with suitable LO carrying different 5'- or 3'-extensions to yield the desired PSS-tails. Synthesis of the short LO is 'less expensive and gives higher yields than the one step synthesis of long continuous hairpins. Apart from the free variability of the sequences and lengths of the stem and DSBtermini oligonucleotide synthesis facilitates the introduction of specific chemical base modifications in the terminal structures of hairpins which could be used to study the mechanisms involved in the rejoining of more complicated breaks like the ones generated e.g. by ionizing radiation or chemical exposure. The internal phosphatase and nuclease protected radioactive label allows to follow the fate of unreacted input substrate molecules (degradation or elongation). However, to avoid DNA damage caused by the f-particle activity of the (32P)-phosphate substrates should be preferentially used immediately after preparation. The hairpin-loop proved to be resistant against exoand endonucleolytic attack and end joining in the Xenopus egg extract which reduced the number of possible reaction products to a comprehensible spectrum because substrates were only

reactive at their 'open' DSB end. It should be noted, however, that some cell types (e.g. those undergoining VDJ-recombination) are presumably able to process such hairpin loops (31) and that such an activity may result in more complex product spectra. The suitability of the hairpin substrates to detect joining activity and to characterize the minimal requirements for successful DNA end joining was tested in extracts from X. laevis eggs. In this in vitro system, mechanisms of DNA end joining had been previously studied in detail with the help of linear plasmid molecules carrying nonhomologous restriction ends. The features of junction formation were found to be very similar to the ones found in mammalian cells (13-16). Depending on the terminus configuration to be joined two different joining modes were distinguished in the Xenopus in vitro system: (i) the 'fill-in' mode which preserves PSS sequences in abutting terminus configurations (blunt/PSS; 5'PSS/3'PSS) by fill-in DNA synthesis (21) and (ii) the 'overlap' mode in which antiparallel terminus configurations (5'PSS/5'PSS; 3'PSS/3'PSS) are joined by formation of short mismatched overlap intermediates which are set by single fortuitously matching basepairs and determine the patterns of subsequent repair reactions (22,32). Both mechanisms were postulated to require DNA binding proteins which structurally support the precise alignment of partner termini during junction formation. In the present study, we used combinations of hairpins carrying antiparallel 5'- or 3'-PSS-termini. Employing hairpins with Bamand Sal-PSS we showed by restriction analysis and sequencing that the obtained junctions exhibited exactly the same features as those obtained from plasmid substrates carrying the Bam/Sal terminus configuration (20,22). In both cases, junctions had formed in which the two terminal fortuitous basematches were apparently used for formation of overlap intermediates whose gaps were filled by DNA synthesis. This indicates that the hairpin substrates were joined by the same mechanisms as the plasmid susbtrates. The use of hairpin substrates permitted for the first time to determine which DNA size dimensions are tolerated by the joining system. Although sticky end ligation of hairpins with the shortest stem lengths (22bp) was possible in reactions with T4 DNA-ligase, ligation and nonhomologous end joining in the Xenopus in vitro system required that both hairpin partners had a stem length of at least 27bp corresponding to about 2.5 doublehelical turns. The critical size of 27bp could be further narrowed down to 26bp using hairpins with blunt ends. In this case, hairpins of 26bp failed to be joined to each other but were joined to longer hairpins (3 lbp), although at very low efficiency. This suggests that joining active protein complexes bind primarily to free substrate molecules of at least 26bp but that alignment interactions between DNA termini can only take place if one (or preferendally both) partner is longer than 26bp. The fact that sticky end and blunt end ligation showed the same length requirements as nonhomologous DNA end joining suggests that protein factors promoting the precise alignment of DNA termini might also be involved in the ligation of complementary PSS and blunt ends. The possible dependence of end joining reactions on the length of the involved DNA molecules was also suggested by Roth et al. (33) who assessed the question whether extra nucleotides (filler DNA) in junctions of illegitimate genomic rearrangements in mammalian cells could arise by joining of oligonucleotide fragments to broken ends prior to end joining. They observed lack of efficient uptake of the tested oligonucleotides (l2bp and 16bp double-stranded portions) suggested to be possibly caused

1650 Nucleic Acids Research, 1994, Vol. 22, No. 9 by intrinsic properties of the mammalian joining machinery which also might not be able to accept short oligonucleotides. Using hairpins with long PSS-tails we determined the maximal PSS-length accepted by the joining system. The PSS-tails were designed to provide equal overlap conditions in all cases but to vary in the length of their poly-A-stretches. The experiments showed that 5'- and 3'-PSS of 6nt were readily accepted by the joining system to yield high amounts of joined products whereas the joining efficiency drastically decreased with growing PSSlength approaching zero at PSS-lengths beyond 10-1 lnt. One explanation for this phenomenon could be that such long tails cannot be properly aligned and therefore fail to undergo junction formation. Another possibility might be that the fill-in length of the gaps (expected to be > 9nt) exhausts the capacities of the DNA polymerase involved (probably pol 3; 20) and a third consideration takes into account that the sequence and structure properties of the monotonous poly-A-stretches may negatively influence joining efficiency. Which of these possibilities holds true has to be tested in further experiments employing long PSStails with different match positions. A biological reason for the failure of the joining machinery to join PSS-tails > lOnt could be envisaged in a lack of the necessity for such a function. DSBtermini with PSS-tails of this length would have to be generated by staggered nicks in duplex DNA lying 10-1 lbp apart. However, a DNA duplex of this size is already relatively stable so that it is unlikely that the two single-strands drift apart at physiological temperatures to yield DSB-ends with long PSS. DNA continuity could be reconstituted by religation of the ss nicks in reactions independent of joining factors. Although nucleolytic degradation of the hairpin substrates was generally low in the Xenopus egg extracts (depending on the extract batch used) we found an unusual exonuclease activity cleaving off 10-mer oligonucleotides from protruding or recessed 5'-ends of hairpins carrying long PSS-tails (30). Interestingly, the intensity of degradation was reduced when DNA end joining was efficient indicating that protein components active in joining protect DSB ends from exonucleolytic attack. This result suggests that the fate of broken DNA molecules is apparently determined by the equilibrium between DNA end joining and degradation processes which both compete for free DNA ends. This idea agrees with previous findings (9,13,17,25 - 27,34).

ACKNOWLEDGEMENTS We would like to thank Borries Kemper for stimulating discussions and valuable advice concerning the construction of hairpin substrates. We are grateful to Elke Feldmann for excellent technical assistance and preparation of egg extracts. This work was funded by the Deutsche Forschungsgemeinschaft through SFB 274 and a grant (Vi-23/1-2) given to W.V. P.P. was supported by a fellowship of the Lise Meitner program of the Ministerium fiur Wissenschaft und Forschung des Landes Nordrhein-Westfalen.

REFERENCES 1. Bryant, P.E. (1986) In Simic, M.G., Grossman, L. and Upton, A.C. (eds), Mechanisms of DNA Damage and Repair, Plenum Press, New York, pp. 171- 180. 2. Friedberg, E.C. (1988) Microbiol. Rev., 15, 70- 102. 3. Jeggo, P.A. (1990) Mut. Res., 239, 1-16. 4. Bennett, C.B., Lewis, A.L., Baldwin, K.K. and Resnick, M.A. (1993) Proc. Natl. Acad. Sci. USA, 90, 5613 -5617.

5. Roth, D.B. and Wilson, J.H. (1988) In Kucherlapati, R. and Smith, G.R. (eds), Genetic Recombination, American Society for Microbiology, Washington D.C., pp. 621-653. 6. Bradley, M.O. and Kohn, K.W. (1979) Nucl. Acids Res., 7, 793-804. 7. Frankenberg, D., Frankenberg-Schwager, M. and Harbich, R. (1984) Br. J. Cancer, 49, 233-238. 8. Szostak, J.W., Orr-Weaver, T.L., Rothstein, R.J. and Stahl, F.W. (1983) Cell 33, 25-35. 9. Goedecke, W, Pfeiffer, P. and Vielmetter, W. (1994) Nucl. Acids Res. accepted. 10. Meuth, M. (1989) In Berg, D.E. and Howe, M.M. (eds), Mobile DNA, Washington D.C., American Society for Microbiology, pp. 833-860. 11. Folger, K.R., Wong, E.A., Wahl, G. and Capecchi, M.R. (1982) Mol. Cell. Biol., 2, 1372-1387. 12. Wake, C.T., Vernaleone, F. and Wilson, J.H. (1985) Mol. Cell. Biol., 5, 2080-2089. 13. Roth, D.B., Porter, T.N. and Wilson, J.H. (1985) Mol. Cell. Biol., 5,

2599-2607. 14. Roth, D.B. and Wilson, J.H. (1986) Mol. Cell. Biol., 6, 4295-4304. 15. Munz, P.L. and Young, C.S.H. (1991) Virology, 183, 160- 169. 16. Winegar, R.A., Lutze, L.H., Rufer, J.T. and Morgan, W.F. (1992) Mutagenesis, 7, 439-445. 17. Lehman C.W., Trautman, J.K. and Carroll, D. submitted for publication. 18. Kramer, K.M., Brock, J.A., Bloom, K., Moore, J.K. and Haber, J.E. (1994) Mol. Cell. Biol., 14, 1293-1301. 19. Schiestl, R.H., Dominska, M. and Petes, T.D. (1993) Mol. Cell. Biol., 13, 2697 -2705. 20. Pfeiffer, P. and Vielmetter, W. (1988) Nucl. Acids Res., 16, 907-924. 21. Thode, S., Schafer, A., Pfeiffer, P. and Vielmetter, W. (1990) Cell, 60, 921 -928. 22. Pfeiffer, P., Thode, S., Hancke, J. and Vielmetter, W. (1994) Mol. Cell.

Biol., 14, 888-896. 23. North, P., Ganesh, A. and Thacker, J. (1990) Nucl. Acids Res., 18, 6205 -6210. 24. Fairman, M.P., Johnson, A.P. and Thacker, J. (1992) Nucl. Acids Res., 20, 4145 -4152. 25. Thacker, J., Chalk, J., Ganesh, A. and North, P. (1992), Nucl. Acids Res., 20, 6183-6188. 26. Derbyshire, M.K., Epstein, L.H., Young, C.S.H., Munz, P.L. and Fishel, R. (1994) Mol. Cell. Biol., 14, 156-169. 27. Nicolas, A.L. and Young, C.S.H. (1994) Mol. Cell. Biol., 14, 170-180. 28. Schaal, H., Pfeiffer, P., Klein, M., Gehrmann, P. and Scheid, A. (1993) Gene, 124, 868-877. 29. Alex, R., Gottlich, B. and Pfeiffer, P. unpublished results. 30. Reichenberger, S., Beyert, N., Peters, M., Vielmetter, W. and Pfeiffer, P., manuscript in preparation. 31. Roth, D.B., Menetski, J.T., Nakajima, P.B., Bosma, M.J., Geilert, M. (1992) Cell 70, 983-991. 32. Pfeiffer, P., Thode, S., Hancke, J., Keohavong, P. and Thilly, W.G., submitted for publication. 33. Roth, D.B., Proctor, N., Stewart, L.K. and Wilson, J.H. (1991) Nucl. Acids Res., 19, 7201-7205. 34. Lehman, C.W., Clemens, M., Worthylake, D.K., Trautman, J.K. and Carroll, D. (1993) Mol. Cell. Biol.. 13, 6897-6906.