Determining mutational fingerprints at the human p53 locus ... - Nature

9 downloads 137 Views 916KB Size Report
p53 cDNA was treated in vitro with the antineoplastic drug ... alkylating agent CCNU, an antineoplastic drug used ... Table 3 reports the complete list of mutations ...
Oncogene (1997) 14, 1307 ± 1313  1997 Stockton Press All rights reserved 0950 ± 9232/97 $12.00

Determining mutational ®ngerprints at the human p53 locus with a yeast functional assay: a new tool for molecular epidemiology Alberto Inga1, Ra€aella Iannone1, Paola Monti1, Francesco Molina2, Martino Bolognesi2, Angelo Abbondandolo1,3, Richard Iggo4 and Gilberto Fronza1 1

CSTA-Mutagenesis Laboratory, National Institute for Research on Cancer (IST), and 2Biostructure Unit, Advanced Biotechnology Centre, IST, Largo R. Benzi, 10, 16132-Genova; 3Chair of Genetics, University of Genova, Italy; 4Swiss Institute for Experimental Cancer Research (ISREC), 1066 Epalinges, Switzerland

In order to isolate experimentally induced p53 mutations, a yeast expression vector harbouring a human wild-type p53 cDNA was treated in vitro with the antineoplastic drug chloroethyl-cyclohexyl-nitroso-urea (CCNU) and transfected into a yeast strain containing the ADE2 gene regulated by a p53-responsive promoter. p53 mutations were identi®ed in 32 out of 39 plasmids rescued from independent ade- transformants. Ninety-two percent of CCNU induced mutations were GC-targeted single base pair substitutions, and GC4AT transitions represented 73% of all single base pair substitutions. In 70% of the cases the mutated G was preceded 5' by a purine. The distribution of the mutations along the p53 cDNA was not random: positions 734 and 785 appeared as CCNU mutational hotspots (n=3, P50.0003) and CCNU induced only GC4AT transitions at those positions. The features of these CCNU-induced mutations are consistent with the hypothesis that O6-alkylguanine is the major causative lesion. One third of the CCNUinduced mutants were absent from a huge collection of 4496 p53 mutations in human tumours and cell lines, thus demonstrating that CCNU has a mutational spectrum which is uniquely di€erent from that of naturally selected mutations. This strategy allows direct comparison of observed natural mutation spectra with experimentally induced mutation spectra and opens the way to a more rigorous approach in the ®eld of molecular epidemiology. Keywords: CCNU; mutation spectrum; p53; ADE2

Introduction Alteration of the p53 tumour suppressor gene is the most common genetic defect known to occur in human tumours (Friend, 1994). Current models suggest that wild-type p53 plays a key role in the maintenance of genomic stability (Livingstone et al., 1992) and is required for the apoptotic response to radiotherapy and chemotherapy (Lowe et al., 1993, 1994). The critical biochemical function associated with p53

Correspondence: G Fronza, CSTA-Mutagenesis Laboratory, National Institute for Research on Cancer (IST), Largo R. Benzi, 10, 16132-Genova, Italy Received 12 September 1996; revised 8 November 1996; accepted 12 November 1996

tumour suppressor activity is the ability to activate transcription (Vogelstein and Kinzler, 1992a; Ory et al., 1994). Known p53 transcriptional targets include an inhibitor of cyclin dependent kinases, p21Waf1/ Cip1, a gene inducing DNA damage-dependent growth arrest (gadd45), and an inducer of apoptosis (bax) (Hainaut, 1995). More than 90% of the p53 mutations observed in tumours are missense mutations. This makes p53 an ideal candidate to test the mutational spectrometry approach to molecular epidemiology (Vogelstein and Kinzler, 1992b; Harris, 1993), and there is already some p53 mutation data supporting the hypothesis that carcinogens and mutagens leave ®ngerprints on DNA (Vogelstein and Kinzler, 1992b; Greenblatt et al., 1994; Ziegler et al., 1994). A limitation inherent in these studies is the failure to experimentally reproduce the observed p53 mutational spectra. Ideally, one should determine the p53 mutational spectrum induced by the putative aetiological factor in an experimental system and show that the same spectrum is observed in tumours of patients exposed to that factor. Previous studies have relied on comparison of observed mutation spectra in one gene with induced mutations spectra in another. Such heterologous comparisons can highlight basic similarities, but ignore di€erences in primary nucleotide sequence and intrinsic susceptibility to mutagenesis of di€erent reporter genes. The determination of mutational ®ngerprints at the same locus should reduce the confounding e€ect of intergenic variation, and thereby increase the power of molecular epidemiological studies. We present here the application of a p53 functional assay (Flaman et al., 1995) to the ®eld of mutagenesis. In this system, human p53 expressed in S. cerevisiae controls transcription of the ADE2 gene. Hence, yeast colonies containing wild-type p53 are white and normal sized on limiting adenine plates, while those containing mutant p53 are red and small. The assay was developed to functionally characterize the status of p53 in tumours, blood and cell lines. In order to validate the assay for mutagenesis studies, we used it to select mutants induced by the bifunctional alkylating agent CCNU, an antineoplastic drug used for the treatment of many types of cancer, and compared it with the CCNU mutational ®ngerprint recently determined at the bacterial supF locus in CV1 cells using a shuttle vector approach (Inga et al., 1995).

CCNU induced p53 mutation fingerprint in yeast A Inga et al

1308

Results Induction of p53 mutations by CCNU treatment A functional assay was used to identify p53 mutations in a yeast expression plasmid damaged with CCNU according to the in vitro mutagenesis protocol shown in Figure 1. Transformants were selected on plates lacking leucine but containing sucient adenine for adenine auxotrophs to grow and turn red (MM +ade5). Survival, estimated from the number of transformants, showed a CCNU concentration-dependent decrease (Table 1). The spontaneous mutation frequency was 261074 and increased in a CCNU concentration-dependent way (Table 1). Molecular analysis was limited to mutants clones induced at 6 mM and 12 mM CCNU, in order to avoid the confounding contribution of spontaneous mutations at lower CCNU concentrations, where a lower increase of mutation frequency was observed. At 6 mM and 12 mM CCNU, at least 94% of mutations are expected to result from the CCNU treatment. Identi®cation of p53 mutations in red yeast colonies To determine whether adenine auxotrophy was due to p53 mutations rather than, for example, mutations in the promoter, the p53 expression plasmid was rescued from 39 ade7leu+ colonies and the p53 open reading frame from nucleotides 125 to 1122 was PCR-ampli®ed and tested by gap repair (Figure 1, and Flaman et al., 1995). Thirty-two PCR fragments gave 100% red colonies after gap repair, and p53 mutations were found by DNA sequencing in every case (Tables 2 and 3). The remaining clones gave a low percentage of red colonies, suggesting that the mutation causing adenine auxotrophy lays outside the region tested by gap repair, an assumption con®rmed by DNA sequencing in three cases. Table 3 reports the complete list of mutations identi®ed, with information on the type of mutation, the amino acid change, and its location in the protein structure (Cho et al., 1994). In total, 32 mutant clones contained 37 mutations (Tables 3 and 4). Only four clones contained multiple mutations; the remaining 28 (87%) had single point mutations.

Figure 1 Mutagenesis protocol to isolate and to characterize CCNU induced p53 mutants

Table 1

Survival

and

mutation

induction

in

CCNU damaged pLS76 after passage through

Survival (%)

CCNU (mM) 0

n + n + n+ + n+ n

p53 mutant frequency 6 6 6 6 2.0

100 b

73

24

16

b

12 a

F=mutant

34

29

(

3)

18

(

3)

Table 2 CCNU

b

Fa

±4

10

1

±4

10

8

±4

10

17

(19/5386)

113

b

frequency of

undamaged DNA.

35

b

18

and

strain

(11/6819)

( =3) 6

undamaged

(3/15045)

( =5) 3

yIG397

±4

10

56

(22/1953)

CCNU damaged/mutant

frequency of

=number of independent transfections

Characterization of CCNU-induced p53 mutants after passage of pLS76 in yIG397 strain 6 mM

12 mM

Total

Mutants analysed 18 21 39 (100%) Mutants showing: GAP repair± 3 (17%) 4 (19%) 7 (18%) GAP repair+ 15 (3%) 17 (81%) 32 (82%) Mutants sequenced GAP repair± 3 3a GAP repair+ 32 15 17 Mutants with: single mutations b 13 28 15 multiple mutations 4 2 2 Mutated sitesc 32 Mutated proteins with: 15 32 17 single substitutiond 28 13 15 other mutations 4 2 2 Mutated codons 25 aNo mutation in the 125±1122 nucleotide of the cDNA region. bMutant showing at least two mutations. cNumber of di€erent nucleotides found mutated. dMutations giving rise to a truncated form of the p53 protein Analysis of p53 mutations at the DNA level With only one exception, all mutations were targeted at GC base pairs, and apart from three single G deletions, all mutations were base pair substitutions. GC4AT transitions were the major class of mutations observed. If one excludes deletions occurring in runs of Gs, whose location is uncertain, mutations involved 28 di€erent nucleotides. Compared with mutations found in human cell lines and tumours (Hollstein et al., 1996; Cariello et al., 1996), CCNU induced mutations at three novel sites (Table 3), and it induced base pair substitutions at two other sites (nucleotides 214 and 215) previously shown to have been a€ected only by deletions (Hollstein et al., 1996; Cariello et al., 1996). Thirty-two percent of the CCNU-induced base pair substitutions represented mutations not seen before. The distribution of the mutations along the p53 cDNA was not random: for example, some sites were mutated three times. It was recently deduced that for PCR-induced mutagenesis there are about 542 sites where mutations can inactivate p53 as a transcription factor in yeast. Excluding frameshifts, there are 457 mutable sites in the region coding for the p53 DNA binding domain (Flaman et al., 1994), but only 372 mutated sites have been reported so far in the literature (Cariello et al., 1996). Using the latter

CCNU induced p53 mutation fingerprint in yeast A Inga et al

Table 3

Mutations induced by CCNU (6 mM or 12 mM) at the

Site Mutant Mutationa 6 mM CCNU induced mutants with single mutations 1

-1G

2

G4A

3

G4A

464

4

G4c

467

5

G4c

509

6

G4A

656

7

G4A

734

8

G4A

774

9

G4A

785

g

Sequence

211±216

G4A

785

11

G4A

796

12

G4A

827

13

G4A

836

locus (pLS76) in

b

CCCCCC4CCCCC TGgG4TGaGe CAcC4CAtCe CCgG4CCcC GAcG4GAgGe GCcC4GCtC CGgG4CGaC AgAC4AaAC TGgT4TGaTe TGgT4TGaTe GgGA4GaGA TGcC4TGtC TGgG4TGaG

d

330

10

(5'-43')

p53

yIG367

1309

strain

Aminoacid change

Domainc

71±72 frameshift

out

f

G117E

L1

f

T155I

S3±S4

R156P

S4

f

T170R

L2

P219L

S7

G245D

L3

D259N

S9±S10

f

G262D

S9±S10

f

G262D

S9±S10

G266R

S10

A276V

S10±H2

G279E

H2

6 mM CCNU induced mutants with multiple mutations 14 15

G4A

321

-1G

898±902

G4A

640

G4A

474

Y107Y silent

TAcG4TAtG CCCCCA4CCCCA AcAT4AtAT CGcG4CGtG

300±301 frameshift

ND

H214Y

S7

R158R silent

±

mutants 16, 17 and 18 were gap repair

12 mM CCNU induced mutants with single mutations d

19

G4A

20

G4A

473

21

G4T

530

22

G4A

637

23

A4T

683

24

G4A

725

25

G4A

730

26

G4A

733

27

G4A

734

28

G4A

734

29

G4A

771

30

G4A

785

31

G4A

796

32

G4A

817

33

G4A

836

g

fh

CCcT4CCtTe GCgC4CCaC CCcC4CCaC TcGA4TtGA TGaC4TGtCe CTgC4CTaC GgGC4GaGC CgGC4CaGC CGgC4CGaC CGgC4CGaC GgAA4GaAA TGgT4TGaTe GgGA4GaGA GcGT4GtGT TGgG4TGaG

293

P98L

preceds S1

R158H

S4

P177H

H1

R213STOP

S6±S7

f

D228V

S7±S8

C242Y

L3

G244S

L3

G245S

L3

G245D

L3

G245D

L3

E258K

S9±S10

f

G262D

S9±S10

G266R

S10

R273C

S10

G279E

H2

P71P silent

out

12 mM CCNU induced mutants with multiple mutations 34

G4T

213 214

d

215

G4A 35

CCcCCC4CCaCCC CCCccC4CCCatCe CCCccC4CCCatCe CCCC4CCC GGcG4GGgGe

d

G4T

250±253

±1G

d

247

G4C

fi

P72I

out

84±85 frameshift

out

f

A83G

out

±j

mutants 36, 37, 38, 39 were gap repair a

Mutation at the GC base pair are reported as induced by G-targeted lesions.

b

Sequence of the not transcribed (coding) strand; lower

c

case mutated base; underlined p53 codon. Localisation of the mutation according to the topological diagram of the core domain of p53 (Cho

et al.,

1994) S:

b

strand, H:

a

helix; L: loop; S4-S5: mutation localized between

determinated; out: outside the cristallized portion of the never found in human tumours (Hollstein

et al.,

p53

1996).

core domain.

mutated in human tumours (4496 mutations) (Hollstein

b

strand 4 and

b

strand 5; ND: structure not

Base in the cDNA of p53 where base pair substitutions were

e

Base pair substitutions never observed (Hollstein

change never found in human tumours (4496 mutations) (Hollstein j

d

et al.,

et al.,

1996).

g

et al.,

Temperature-sensitive mutant.

h

1996).

f

Aminoacid

Codon never found

i

1996). Never found mutated, but this is a site of a known polymorphism.

Mutants 36, 37 and 38 showed no mutations between positions 125 and 1122 in the

value, and taking into account the fact that we only examined 84% of the cDNA sequence, we can postulate that 312 mutable sites were examined in this study. Assuming a Poisson's distribution of CCNU-induced mutations, the mean number of mutations per mutable site was 0.12 (37/312). Using the PCR-derived value for mutable sites, which applies to the same region as was tested in this study, the mean number of mutations per mutable site is 0.068 (37/542). Even by the more conservative estimate derived from studies on human tumours and cell lines, sites with three independent mutations (positions 734 and 785) should be considered as CCNU mutational hotspots (n=3, P50.0003). Remarkably, CCNU induces only GC4AT transitions at these positions.

p53

cDNA

Analysis of the distribution of the GC4AT transitions showed that 70% of mutated Gs were preceded 5' by a purine (Table 5). Analysis of p53 mutations at the protein level The mutant plasmids encoded 32 di€erent p53 mutations at the protein level (Table 3). Three clones contained additional silent mutations. Four mutations truncated the protein; the remainder encoded single amino acid substitutions. All but one of the aminoacid substitutions lie in the DNA-binding domain (Cho et al., 1994) (Table 3) and eight of them have not been reported so far in human tumours. The most frequently mutated residues were those at codons 245 (four independent mutants) and 262 (three independent

CCNU induced p53 mutation fingerprint in yeast A Inga et al

1310

Table 4

Molecular features of CCNU induced p53 mutations at the DNA level in yIG397

CCNU

6 mM

12 mM

Total

Mutationsa 17 (100%) 20 (100%) 37 (100%) Base pair substitution: 15 (88%) 19 (95%) 34 (92%) GC-targeted 15 18 33 GC4AT 13 14 27 GC4PyPu 2 4 6 AT-targeted 1 1 AT4TA 1 1 -1G frameshift 2 (12%) 1 (5%) 3 (8%) aThree silent mutations (mutants no. 14, 15, 34) are independent genomic CCNU induced events and were included in this analysis Distribution along the p53 sequence of CCNU induced GC AT transitions. Influence of DNA sequence context

Table 5

4

CCNU

GC AT 5 PuG 5 GG-3 5 AG-3 5 PyG 4

'

'

'

'

'

'

6 mM

13 (100%) 10 (77%) 9 1 3 (13%)

12 mM

14 (100%) 9 (64) 8 1 5 (36%)

Total

27 (100%) 19 (70%) 17 2 8 (30%)

Figure 2 Localization of the new p53 mutants in the monomeric p53-DNA complex

mutants). Almost half of the missense mutations a€ected glycine residues, followed by arginine and proline residues (four and three mutations each). The mutants were selected on the basis of decreased transcriptional activity, probably caused by changes which reduce the anity of the protein for DNA (Cho et al., 1994). The possible e€ect of the novel CCNUinduced missense mutations was analysed in relation to the protein-DNA crystal structure (Cho et al., 1994 and Figure 2). In the absence of quantitative biochemical data for such values as decreased DNA binding or transactivation activity, assignment of a

structural or functional consequence to a mutation is reasonable but hypothetical. Keeping this in mind, the G117E mutation may directly a€ect DNA binding, since it brings a negatively charged side chain with 4.5 AÊ of the DNA phosphate backbone. It may also in¯uence cooperativity, since it lies in a loop which is involved in protein-protein interactions in the tetramer. The T170R mutation may also a€ect packing of the tetramer. The remaining mutations are located on the face of the protein opposite to the p53-DNA binding region (T155I, P219L, D228V and G262D, Figure 2). Mutations in this region are assumed to a€ect the overall folding of the molecule (Cho et al., 1994). Two of these mutants (P219L and D228V) gave pink colonies during the initial screen at 32 ± 358C, and further testing revealed that both were temperaturesensitive. Discussion The use of the p53 gene for mutagenesis studies exploits the extreme intrinsic sensitivity to point mutations of the p53 protein: base pair substitutions at 46% of sites in the open reading frame are potentially detectable with the yeast p53 functional assay (Flaman et al., 1994). The present work demonstrates that this system is suitable for the isolation of experimentally induced mutants and for mutational spectrometry studies. The basic molecular features of CCNU induced mutations selected in yeast are consistent with those reported at the supF locus in CV1 cells (Inga et al., 1995). At both loci, the incidence and the location of GC4AT transitions are consistent with the hypothesis that O6-alkylguanine is the main determinant of CCNU mutagenicity. Alkylnitrosoureas preferentially cause GC4AT transitions at Gs preceded by a Pu residue in E. coli (Jurado et al., 1995; RoldaÁn-Arjona et al., 1994; Richardson et al., 1987; Burns et al., 1988) and in eukaryotes (Inga et al., 1995; Kunz and Mis, 1989; Minnick et al., 1992; Moriwaki et al., 1991; Palombo et al., 1992). The 5'-¯anking purine e€ect has been explained as the consequence of a structural in¯uence on the initial formation of O6alkylguanine, which may be further enhanced by alkylation speci®c repair (Horsfall et al., 1990). It is known that CCNU shows a sequence-dependent reactivity (Erickson et al., 1980; Briscoe et al., 1990; Hartley et al., 1986; Prakash and Gibson, 1992): the middle G in a run of three or more Gs is the most reactive. In the perspective of molecular epidemiology, the comparison of mutational spectra at the same locus obtained in humans and in experimental systems is of major interest. Mutational spectra obtained at di€erent loci should be compared with caution. Di€erences in mutational spectra are expected because of di€erences in DNA primary sequences and gene functions. For example, the expression of supF mutations is dependent only on the transcription of the mutated gene. On the contrary, the phenotypic e€ect of a p53 mutation requires loss of protein function, which may impose a so called `protein ®lter e€ect'. Indeed, the comparison of CCNU mutational spectra at the supF locus in CV1 cells and at the p53 locus in yIG397, revealed interesting general similarities (a majority of GC4AT

CCNU induced p53 mutation fingerprint in yeast A Inga et al

transitions and an in¯uence of sequence context). However, clear di€erences were found: multiple mutations (P50.004, Fisher's exact test) and GC transversions (P50.03, Fisher's exact test) were signi®cantly more frequent in CV1 cells than in yIG397. Furthermore, position 123 of the supF gene (5'-GGG-3') is the most frequently CCNU induced hotspot in an excision repair pro®cient background both in CV1 cells (Inga et al., 1995) and in E. coli (Iannone et al., in preparation). In the region of the p53 cDNA under study there are 37 sites with at least three adjacent Gs on the nontranscribed or transcribed strands, most of which involved glycine and proline residues. How can one predict which one of them is expected to be the CCNU induced hotspot in p53? We previously analysed sequence dependent perturbations of the average DNA helix parameters of di€erent CCNU hotspots (Inga et al., 1995) in order to determine whether structural analysis can reveal similarities between hotspots. Sequences encompassing di€erent CCNU hotspots shared identical helix parameters for no more than two base pair steps 5' (or three base pair steps 3') of the mutated G. Furthermore, positions in the supF sequence with identical helix parameters with respect to position 123, showed no mutations probably because they were located in a cold region (i.e. pre-tRNA sequence). By the same approach positions 734 and 785, which are hotspots in p53, shared no similarity with position 123 in the supF sequence (data not shown). We conclude that the analysis of sequence-dependent perturbations of the average DNA helix parameters is unable to predict mutational hotspots. It follows from this discussion that the comparison of mutational ®ngerprints at the same locus, by removing the in¯uence of confounding intergenic di€erence, is expected to increase the sensitivity of the comparative analysis. Nevertheless, one has to keep in mind that several factors, for example di€erent repair capacities, or di€erent drug exposures in vivo and in the in vitro system, can modulate mutation ®ngerprints. Therefore, when these factors are known, they have to be taken into account when two mutational spectra are compared. An extremely useful computer programme for the comparison of mutational spectra at the same locus has been developed (Cariello et al., 1994). This new tool performs a rigorous statistical test to provide a precise measure of the relatedness of two spectra. The use of this programme demonstrated that p53 mutations in hepatocellular carcinomas from areas of low and high a¯atoxin exposure are indeed di€erent (P50.003, Cariello et al., 1994, and references therein). There are three well known examples of exposure to a particular carcinogen inducing mutations consistent with the known mutagenic speci®city of the carcinogen (Hollstein et al., 1996; Cariello et al., 1996; Greenblatt et al., 1994): (i) u.v.-light as an inducer of p53 tandem base substitutions at dipyrimidine sites in skin cancer; (ii) a¯atoxin B1 (in association with HBV chronic infection) as an inducer of codon 249 GC4TA p53 mutations in hepatocarcinomas; (iii) tobacco smoke as an inducer of GC4TA transversions in lung tumours. The incrimination of these aetiologic factors was mainly obtained through identi®cation of basic similarities between mutational spectra obtained at

various loci in experimental systems (eukaryotes and prokaryotes) and at the p53 locus in human tumours. In some cases, the correlation between risk factors and mutations has been substantiated also through the determination of a DNA lesion ®ngerprint on p53 cDNA (Pusieux et al., 1991). Recently, Cerutti and colleagues developed a sophisticated assay (RFLPPCR) in which mutations are detected at the genomic level without any phenotypic selection (Aguilar et al., 1994; Pouzand and Cerutti, 1993). By applying this method they found that the frequency of the AGG to AGT mutation at codon 249 paralleled the level of AFB1 exposure, which supports the hypothesis that this toxin has a causative and probably early role in hepatocarcinogenesis. Although RFLP-PCR has the advantage of depending only on DNA sequence information, and thus avoids phenotypic selection, the analysis is limited to few restriction sites and may not be representative of the mutagenic e€ect spread over whole genes. In this respect, the present work represents an important advance since it is based on a key p53 function and analyses a large target sequence. We compared the panel of p53 mutants obtained in yeast and the 4496 p53 mutations found in human tumours and cell lines (Hollstein et al., 1996; Cariello et al., 1996). This comparison is reasonable because yeast p53 mutants were isolated and selected for the loss of the most critical biological function (i.e. transactivation) which has been shown to be strictly associated with inhibition of cell proliferation (Ory et al., 1994), and is expected to characterize also tumour derived p53 mutations. The results of these comparisons should be interpreted with caution for several reasons. Indeed, yeast p53 mutants are unable to transactivate the reporter gene (ADE2) while relatively few p53 mutations in the data bank have been functionally characterized. Furthermore, most of the mutation analyses in human tumours were con®ned to exons 5 to 8 for technical reasons (Hollstein et al., 1996; Cariello et al., 1996), while we analysed a more extended sequence. Keeping these considerations in mind, it may not be surprising that some nucleotides showing new CCNU induced mutations were found outside exons 5 ± 8 region (codons 124 ± 306). However, referring only to that region, the fact that six out of 32 CCNU independent mutants showed p53 base pair substitutions never found in human tumours (Table 3), strongly suggests that CCNU has its own mutational speci®city (P50.00001, Chi square test). The presumptive CCNU speci®city is con®rmed also at the protein level: while only 7% of p53 mutations in human tumours involved glycine residues (Hollstein et al., 1996; Cariello et al., 1996) 46% of p53 single base pair substitutions in yeast involved glycine residues (Table 3, P50.0001, Chi square test). This result is best explained by the composition of the glycine codon (5'GGN-3') and the known preference of CCNU for runs of Gs (Erickson et al., 1980; Briscoe et al., 1990; Hartley et al., 1986; Prakash and Gibson, 1992). In conclusion, we have shown that the yeast p53 functional assay can be used to isolate induced mutants in the human p53 gene. The assay is simple, more rapid and less expensive than long term carcinogenic tests. Besides its role as new tool for molecular epidemiology, it also allows to obtain new p53 cDNA mutants, some of which may lead to a new insight into the

1311

CCNU induced p53 mutation fingerprint in yeast A Inga et al

1312

biological and biochemical functions of the p53 protein. Finally, it should be noted that although CCNU is used for the treatment of many types of cancer, it has been linked with the subsequent development of leukaemia (Gerson, 1993) and the International Agency for Cancer Research has classi®ed it as `probably carcinogenic to humans' (Marselos and Vainio, 1991). Comparison of the experimental CCNU ®ngerprint identi®ed using the yeast assay with the clinical p53 mutational spectrum in CCNU-related leukaemias should allow clearer conclusions to be drawn about the safety and wisdom of using this agent in humans.

Materials and Methods Strains, vectors and media The pLS76 yeast expression vector harbouring a human wild-type p53 cDNA under the control of an ADH1 promoter (Flaman et al., 1995), and the LEU2 gene as selectable marker in yeast cells, was previously described (Ishioka et al., 1993). pRDI-22, the expression vector used for gap repair, is identical with pLS76 except for the presence of a linker cloned between the p53 BsmI and StuI sites. pRDI-22 cut with HindIII and StuI can be used for gap repair assays without gel puri®cation, and gives very low number of colonies containing self-ligated vector. The haploid strain yIG397 (MATa ade2-1 leu23,112 trp1-1 his3-11, 15 can1-100, ura3-1 URA3 3xRGC::pCYC1::ADE2) containing the ADE2 reporter gene under p53 control (Flaman et al., 1995) was used as recipient for p53 expression vectors. Standard yeast manipulations were performed as described (Guthrie and Fink, 1991). Complete medium supplemented with 200 mg/ l adenine (CM+ade200) was used for routine cultures, minimal medium lacking leucine and containing 5 mg/l adenine (MM+ade5) was used to test p53 status, and minimal medium lacking leucine and adenine (MM) was used to con®rm the ade phenotype. Synthetic minimal medium lacking leucine and containing 200 mg/l adenine (MM+ade200) was used to grow ade clones. DNA modi®cation, transfection and plasmid recovery pLS76 was treated in vitro with 3, 6 and 12 mM CCNU concentrations for 1 h in 50% ethanol (Inga et al., 1995). Control and damaged vectors were puri®ed by three DNA precipitations, and transfected into yIG397 cells by electroporation. Transformants were selected on MM+ade5 plates for their ability to grow in the absence of leucine (Flaman et al., 1995). Pink colonies, representing potential thermo-sensitive p53 mutants, were tested at

248C, 308C and 378C. The ratio between the number of transformants obtained from damaged with respect to mock treated vectors gives an estimate of survival, while mutant frequency was de®ned as the ratio of red colonies to the total. Small red colonies were re-isolated on MM+ade5 plates and adenine auxotrophy was con®rmed by plating on MM medium. Ade clones were grown overnight in MM+ade200, lysed with Lyticase (Fluka, Milano, Italy) and plasmids were rescued by a standard protocol (Guthrie and Fink, 1991). DNA ampli®cation, gap repair and sequencing From each rescued plasmid, the p53 open reading frame between nucleotide positions 125 and 1122, including the entire sequence coding for the DNA binding domain, was PCR ampli®ed using primers P3 and P4 and Pfu DNA polymerase (Eppendorf, Milano, Italy) as previously described (Flaman et al., 1995). Unpuri®ed PCR products and HindIII ± StuI linearized pRDI-22 were co-transformed by electroporation into yIG397. Gap repaired transformants were selected on suitable plates and the percentage of small red colonies was determined. p53 mutants, giving about 100% of red colonies, were re-ampli®ed with primers P3 and P4 and Taq DNA polymerase (Promega, Madison, WI). PCR products were puri®ed by Microcon 100 (Amicon Inc., Beverly, MA) and directly sequenced on both strands with ABI PRISMTM Dye terminator cycle sequencing Ready Reaction Kit (Perkin Elmer, Milano, Italy) on an Applied Biosystems Inc. 377 Automated Sequencer (Perkin Elmer, Milano, Italy). Besides P3 and P4, two internal primers were used for sequencing: P5 5'TggCCATCTACAAgCAgTCA-3', nucleotide position 479 to 497 and P6 5'-gggCACCACCACACTATgTC-3', nucleotide position 638 to 657, of the p53 cDNA. Molecular modelling The p53:DNA complex (Brookhaven data set 1TSR, Cho et al., 1994) was examined using the package TURBOFRODO (Bio-Graphics, Marseille, France).

Abbreviations CCNU, Chloroethyl-Cyclohexyl-Nitroso-Urea; Pu, purine; Py, pyrimidine. Acknowledgements We thank Drs M Bignami and E Dogliotti for critical reading of the manuscript and Dr P Menichini for helpful discussion. This work was partially supported by contract CHRX-CT94-0581 from the Commission of the European Communities and by the Italian Association for Cancer Research (AIRC).

References Aguilar F, Harris CC, Sun T, Hollstein M and Cerutti P. (1994). Science, 264, 1317 ± 1319. Briscoe WT, Anderson SP and May HE. (1990). Biol. Pharmacol., 40, 1201 ± 1209. Burns PA, Gordon AJE, Kunsmann K and Glickman BW. (1988). Cancer Res., 48, 4455 ± 4458. Cariello NF, Piegorsch WW, Adams WT and Skopek TR. (1994). Carcinogenesis, 15, 2281 ± 2285. Cariello NF, Douglas GR and Soussi T. (1996). Nucleic Acids Res., 24, 119 ± 120. Cho Y, Gorina S, Je€rey PD and Pavletich NP. (1994). Science, 265, 346 ± 355.

Erickson LC, Laurent G, Sharkey NA and Kohn KW. (1980). Nature, 288, 727 ± 730. Flaman JM, Frebourg T, Moreau V, Charbonnier F, Martin C, Ishioka C, Friend SH and Iggo, R. (1994). Nucleic Acid Res., 22, 3259 ± 3260. Flaman JM, Frebourg T, Moreau V, Charbonnier F, Martin C, Chappuis P, Sappino A-P, Limacher J-M, Bron L, Benhattar J, Tada M, van Meir EG, Estreicher A and Iggo R. (1995). Proc. Natl. Acad. Sci. USA, 92, 3963 ± 3967. Friend S. (1994). Science, 265, 334 ± 335. Gerson SL. (1993). Curr. Opin. Oncol., 5, 136 ± 144.

CCNU induced p53 mutation fingerprint in yeast A Inga et al

Greenblatt MS, Bennett WP, Hollstein M and Harris CC. (1994). Cancer Res., 54, 4855 ± 4878. Guthrie C and Fink GR. (1991). Methods in Enzymology p194, Academic Press: San Diego. Hainaut P. (1995). Current Opin. in Oncol., 7, 76 ± 82. Harris CC. (1993). Science, 262, 1980 ± 1981. Hartley JA, Gibson NW, Kohn KW and Mattes WB. (1986). Cancer Res., 46, 1943 ± 1947. Hollstein M, Shomer B, Greenblatt M, Soussi T, Hovig E, Montesano R and Harris CC. (1996). Nucleic Acids Res., 24, 141 ± 146. Horsfall MJ, Gordon AJE, Burns P, Zielenska M, van der Vliet GME and Glickman BW. (1990). Environ. Mol. Mutagen., 15, 107 ± 122. Inga A, Iannone R, Campomenosi P, Molina F, Menichini P, Abbondandolo A and Fronza G. (1995). Cancer Res., 55, 4658 ± 4663. Ishioka C, Frebourg T, Yan Y-X, Vidal M, Friend SH, Schmidt S and Iggo R. (1993). Nature Genetics, 5, 124 ± 129. Jurado J, Ferrezuelo F and Pueyo C. (1995). Mol. Carcinog., 14, 233 ± 239. Kunz BA and Mis JRA. (1989). Cancer Res., 49, 279 ± 283. Livingstone LR, White A, Sprouse J, Livanos E, Jacks T and Tlsty TD. (1992). Cell, 70, 923 ± 935. Lowe SW, Ruley HE, Jacks T and Housman DE. (1993). Cell, 74, 957 ± 967. Lowe SW, Bodis S, McClatchey A, Remington L, Ruley HE, Fisher D, Housman D and Jacs T. (1994). Science, 266, 807 ± 810.

Marselos M and Vainio H. (1991). Carcinogenesis, 12, 1751 ± 1766. Minnick DT, Veigl ML and Sedwick WD. (1992). Cancer Res., 52, 4688 ± 4695. Moriwaki SI, Yagi T, Nishigori C, Imamura S and Takebe H. (1991). Cancer Res., 51, 6219 ± 6223. Ory K, Legros Y, Auguin C and Soussi T. (1994). EMBO J., 13, 3496 ± 3504. Palombo F, Kohfeldt E, Calcagnile A, Nehls P and Dogliotti E. (1992). J. Mol. Biol., 223, 587 ± 594. Pouzand C and Cerutti P. (1993). Mutat. Res., 288, 113 ± 121. Prakash AS and Gibson NW. (1992). Carcinogenesis, 13, 425 ± 431. Pusieux A, Lim S, Groopman J and Ozturk M. (1991). Cancer Res., 51, 6185 ± 6189. Richardson KK, Richardson FC, Crosby RM, Swenberg JA and Skopek TR. (1987). Proc. Natl. Acad. Sci. USA, 84, 344 ± 348. RoldaÁn-Arjona T, Luque-Romero FL, Ariza RR, Jurado J and Pueyo C. (1994). Mol. Carcinog., 9, 200 ± 209. Vogelstein B and Kinzler KW. (1992a). Cell, 70, 523 ± 526. Vogelstein B and Kinzler KW. (1992b). Nature, 355, 209 ± 210. Ziegler A, Jonason AS, Le€el DJ, Simon JA, Sharma HW, Kimmelman J, Remington JT and Brash DE. (1994). Nature, 372, 773 ± 776.

1313