Mechanisms of p16INK4A inactivation in non small-cell lung ... - Nature

Oncogene (1998) 16, 497 ± 504  1998 Stockton Press All rights reserved 0950 ± 9232/98 $12.00

Mechanisms of p16INK4A inactivation in non small-cell lung cancers Sylvie Gazzeri1, Valerie Gouyer1, Clair Vour'ch2, Christian Brambilla1 and Elisabeth Brambilla1 1

Groupe de recherche sur le cancer du poumon, 2INSERM U309, DYOGEN, Institut Albert Bonniot, 38706 La Tronche Cedex, France

The cyclin-dependent kinase inhibitor p16 (p16INK4A/ CDKN2/MTS1) is a potent inhibitor of the cyclin Ddependent phosphorylation of the retinoblastoma gene (Rb) product, the inactivation of which induces loss of Rb-dependent G1 arrest through inappropriate phosphorylation of the Rb protein. To analyse the role of p16INK4A as a tumor suppressor in the genesis of non small cell lung cancers (NSCLC) and correlate loss of p16INK4A protein expression to genetic or epigenetic mechanisms, we have performed a comprehensive study of p16 status in a series of 43 NSCLC. To this end, we have investigated p16INK4A protein expression with immunohistochemistry, deletions of the gene by FISH, and determined the methylation status of exon 1a using a PCR-based methylation assay. Finally, possible mutations were studied by SSCP and subsequent sequencing. Twenty one of the 43 (49%) NSCLC studied exhibited an absence of p16INK4A nuclear staining. Of these, three (14%) had frameshift or missense mutations, seven (33%) displayed methylation of exon1a and 10 (48%) displayed homozygous deletions. In total, 95% of the tumors with p16INK4A negative staining carried one of these three alternative genetic or epigenetic alterations. Furthermore, a high degree of chromosome 9 polysomy was found (58%) in those tumors with p16INK4A inactivation. Taken together these results suggest that deregulation of the p16 gene locus is a frequently occurring event in NSCLC through distinct mechanisms including rare point mutations, promotor methylation and frequent homozygous deletions. Furthermore, our data show that immunohistochemistry is a rapid and an accurate technique for screening of p16INK4A gene inactivation events that result in loss of protein expression. Keywords: p16INK4A; lung cancer; protein expression; mutation; methylation; homozygous deletion

Introduction The cyclin-dependent kinase (CDK) inhibitor p16 (p16INK4A/CDKN2/MTS1) (Serrano et al., 1993) is an important component of cell cycle regulation. The p16INK4A gene encodes a cell cycle protein, which is a potent inhibitor of CDK4 and CDK6 (Serrano et al., 1993, 1995, for a review, see Sherr, 1996) and negatively regulates the cyclin D-dependent phosphorylation of the retinoblastoma gene product, thereby inhibiting cell cycle progression from G1 to S-phase by sequestration of E2F (Lukas et al., 1995). Other Correspondence: S Gazzeri Received 15 May 1997; revised 15 September 1997; accepted 15 September 1997

reports have provided evidence that the p16INK4A levels may be regulated by Rb protein (Li et al., 1994; Tam et al., 1994), suggesting that Rb participates in a feedback loop to limit the levels of p16INK4A. Genetic alterations of p16INK4A, leading to its inactivation in tumors, results in the deregulation of cell proliferation through loss of G1 arrest control. Accumulating evidence implicates p16INK4A as a tumor suppressor gene in tumorigenesis. p16INK4A maps to 9p21, a chromosomal locus frequently altered in many primary tumors and cell lines (Hussussian et al., 1994; Nobori et al., 1994; Kamb et al., 1994; Okamoto et al., 1994). Homozygous deletions, inactiving mutations and, more recently, methylation at the 5' CpG island associated with transcriptional silencing have been identi®ed as alternative mechanisms of p16INK4A inactivation (Cairns et al., 1994, 1995; Merlo et al., 1995). Furthermore, ectopic expression of wild-type p16INK4A in some tumor cell lines leads to G1 arrest and growth inhibition (Jin et al., 1995; Liggett et al., 1996; Spillare et al., 1996). Moreover the e€ects of p16INK4A require a functional Rb, suggesting the existence of a common p16INK4A/Rb growth suppressor pathway (Koh et al., 1995; Guan et al., 1994; Liggett et al., 1996; Lukas et al., 1995). The redundancy of Rb and p16INK4A inactivation on this common pathway is supported by the observation of an inverse correlation between alterations in the expression of both proteins in several tumors types (Okamoto et al., 1994; Otterson et al., 1994; Geradts et al., 1995; Tam et al., 1994; Shapiro et al., 1995; Kelley et al., 1995). The notion of a feedback regulation of p16INK4A by Rb is consistent with the accumulation of high levels of p16INK4A mRNA in cells lacking Rb function (Li et al., 1994). In human lung cancer, p16INK4A aberrations are frequent (Shapiro et al., 1995; Okamoto et al., 1994; Kratzke et al., 1996; Sakaguchi et al., 1996). Homozygous deletions have been described in lung tumors, with a highly variable frequency (9 ± 25%). They were more frequently detected in cell lines (Kamb et al., 1994; Nobori et al., 1994) than in fresh tumor tissue (Cairns et al., 1995; Packenham et al., 1995; De Vos et al., 1995; Marchetti et al., 1997; Wiest et al., 1997) probably due to stromal contamination of the tumoral samples. Methylation of the 5' CpG island associated with transcriptional silencing has been investigated in only one study of primary lung tumors where its incidence was 26% (Merlo et al., 1995). p16INK4A mutations are less common (0 ± 8%) (Nakagawa et al., 1995; Cairns et al., 1994; Rusin et al., 1996; Okamoto et al., 1995) except for one study reporting 30% of intragenic mutations (Hayashi et al., 1994). To date, the real prevalence of these three known mechanisms (i.e. homozygous deletion, methylation and mutation) in p16INK4A deregulation, and their correlation with the protein expression status, have

p16INK4A inactivation in lung cancer S Gazzeri et al

498

not been documented in the same series of human lung tumors. Accordingly, in order to analyse the role of p16INK4A as a tumor suppressor in the genesis of NSCLC, we have examined 43 human NSCLC for p16INK4A protein expression using immunohistochemistry (IHC) and correlated the data with those of a genetic analysis of the p16INK4A locus aimed at detecting homozygous deletions, methylations and mutations. We found that p16INK4A was frequently altered in NSCLC and that homozygous deletion and DNA hypermethylation are major events in its deregulation. Furthermore IHC is an accurate method for detection of p16INK4A gene inactivation events that result in loss of protein expression.

Results

both antibodies was detected in seven of the 43 cases analysed. In three cases immunostaining with p16 C20 antibody was intense on a high percentage of cells (20 ± 60%) whereas p16 INK4 (Pharmingen) immunostaining was negative. However, according to our criteria for positivity, these cases were classed as positive (average score of positive nuclei 410%). Four other cases showing low intensity of staining for p16 INK4 (10% positive cells) but a complete absence of p16 C20 were classi®ed as p16INK4A negative. Mutations in exons 1 and 2 of p16INK4A gene SSCP of exons 1 and 2 of p16INK4A was performed on all 43 tumors. Mobility shifts were found in two tumors re¯ecting a mutation in exon 1, and in four tumors re¯ecting a mutation in exon 2. DNA sequencing of these abnormal cases revealed frameshift mutations in two tumors (T30, T41) and missense

p16INK4A protein expression p16INK4A protein expression was studied by immunohistochemistry (IHC) using two polyclonal antibodies on a panel of 43 tumor tissue samples (Table 1). The results of this immunohistochemical analysis were recorded independently by two investigators (SG, EB), who assessed the percentage of positive cells and the intensity of staining. Type II pneumocytes, suprabasal bronchial cells and bronchioli epithelial cells, which were present in the vicinity of the tumor on the same slide, were used as positive internal controls. Normal lung tissue served as external control, using the same target cells as above. Lymphocytes were used as a negative internal control on slides. For both antibodies used, p16INK4A immunostaining was considered as positive when at least 10% of tumor nuclei were stained, with or without cytoplasmic staining. Cytoplasmic staining alone was not considered as positive p16INK4A staining. When immunoreactivity appeared discordant between the two antibodies, the mean frequency of positive cell nuclei was used, and cases showing an average score greater than or equal to 10% positivity were considered positive. Negative p16INK4A staining was observed in 7/19 (37%) adenocarcinoma, 9/16 (56%) squamous cell carcinoma and 5/8 (62%) basal cell carcinoma (Table 1; Figure 1) independently of tumor stage and tumor site (primary or metastatic). In total, 21 of the 43 NSCLC analysed (49%) exhibited a complete absence of p16INK4A staining. Furthermore both antibodies p16 C20 and p16INK4 (pharmingen) appear suitable for IHC based p16INK4a protein expression analysis in NSCLC. Overall the concordance between both antibodies reactivities (positive or negative) was of 84%. However, discordant immunoreactivities between

Figure 1 P16INK4A immunostaining. (A) P16INK4A positive immunostaining in an adenocarcinoma (bronchiolo alveolar carcinoma) showing positively stained nuclei (immunoperoxidase6200). (B) P16INK4A negative immunostaining in a poorly di€erentiated squamous carcinoma. Note some positively stained nuclei in stromal cells (immunoperoxidase 6200)

Table 1 Negative immunostaining of p16INK4A according to histology, stage and site of the 43 NSCLC studied

Adenocarcinoma n=19 Squamous carcinoma n=16 Basal cell carcinoma n=8 Total n=43

Total Nb and % of p16INK4A negative immunostaining

Stage I ± II

7 (37%) 9 (56%) 5 (62%) 21 (49%)

3/11 3/8 4/5 10/24 (42%)

Nb of p16INK4A negative IHC immunostaining Stage III ± IV Primary site 4/8 6/8 1/3 11/19 (58%)

5/15 6/13 5/8 16/36 (44%)

Metastasis 2/4 3/3 0 5/7 (71%)

p16INK4A inactivation in lung cancer S Gazzeri et al

mutations in four tumors (T3, T24, T21, T29) (Table 2 and Figure 2). The two tumors with frameshift mutations displayed a complete absence of p16INK4A staining using IHC. Of the four tumors with missense mutations, three exhibited positive p16INK4A staining. Normal lung tissue DNA was available in two (T21, T24) of these three patients and p16INK4A sequencing did not reveal any mutation thus excluding a polymorphism in these two cases. The last tumor carrying a missense mutation (T29) did not express p16INK4A protein on the basis of IHC analysis but displayed de novo methylation of exon1a as described below. We could not exclude a polymorphism in this last case since normal lung was not available. p16INK4A methylation status A PCR-based methylation assay was performed in the same series of tumors using methylation-sensitive endonucleases for DNA cleavage in exon 1 of p16INK4A (CfoI, HpaII, SmaI, SacII), followed by PCR amplifica-

Table 2 Mutations of p16INK4A in human NSCLC Samples

p16INK4Aa Codon

Nucleotide change

T21 T41

+ 7

22 33 ± 36

CGG?CCG GAG . . . . GCC?GACG

T29 T3 T24 T30

7 + + 7

57 74 83 105

GCC?GTC GAC?TAC CAC?CGC CTGGAC?CTGAC

tion of the digested DNAs. Two tumors (T29, T39) showed methylation with all four restriction enzymes used, one tumor (T42) with three enzymes (CfoI, SacII, HpaII) and the four others (T14, T18, T28, T32), with only one enzyme (SmaI for T14, SacII for T18 and T32 and CfoI for T28) (Table 3). Normal lung tissue was available in ®ve of these seven patients (T18, T29, T32, T39, T42) and DNA hypermethylation was never observed in these normal tissues. Overall, patterns consistent with partial or total methylation were observed in seven of the 21 (33%) lung tumors displaying absence of p16INK4A staining and were never found in the tumors with positive p16INK4A immunostaining (Table 3). It should be noted that all the histological types of NSCLC examined were a€ected by methylation: 33% of squamous carcinoma, 25% of adenocarcinoma and 40% of basal cell carcinoma. Figure 3 shows representative examples of the 43 lung cancer cases examined, from which normal lung and the corresponding lung tumor were available. The methylation status at SmaI and SacII restriction sites was also analysed by Southern blotting in the seven tumors with abnormal methylation status and the results were concordant with those of the PCR methylation assay.

Coding change arg?pro 8-base pair deletion ala?val asp?Tyr his?arg 1-base pair deletion

a

Proteic expression on the basis of immunohistochemistry, Absent (7), Present (+)

Table 3 De novo methylation of p16INK4A in human NSCLC

Partial methylationa Total methylationb Total

p16INK4A negative IHC

p16INK4A positive IHC

5/21 2/21 7/21 (33%)

0/22 0/22 0/22

a Methylation with one, two or three or the four methylation-sensitive enzymes used. bMethylation with the four methylation-sensitive enzymes used

M

H

S

C

s

U

M

H

S

C

s

U

N T30 T29 Normal A TGC

T30 AT GC Normal lung

1 bp deletion

T18

Normal lung

Figure 2 Examples of mutations in exon 2 of the P16INK4A gene. SSCP analysis showed altered migration pattern for T30 when compared to normal tissue (N). DNA sequence analysis revealed a 1 bp deletion in codon 105 indicated by an arrow. (Sequences read 5' to 3', bottom to top of gel)

Figure 3 PCR methylation analysis of exon 1 of the P16INK4A gene in normal and tumor tissues. Results of some of the cases analysed are shown. DNAs were digested with MspI (M), HpaII (H), SmaI (S), CfoI (C), or SacII (s). Fifty ng of digested or undigested (U) DNA were ampli®ed with speci®c primers as described in the Materials and methods section. Tumor T29 shows methylation with all four methylation-sensitive restriction enzymes used whereas tumor T18 reveals methylation with SacII enzyme only. Corresponding normal tissues are not methylated

499

p16INK4A inactivation in lung cancer S Gazzeri et al

500

FISH analysis To address the molecular status of the p16INK4A-protein negative tumors (by IHC) for which mutations or de novo methylation of p16INK4A were not detected (12 cases), a FISH analysis was performed on touch preparations of tumor material. Initial examination of normal bronchial epithelia showed an hybridization eciency of more than 95%. Hybridization was successful on all 12 malignant neoplasms studied. One hundred nuclei were analysed for each touch preparation. Unexpectedly, seven of the 12 tumors studied (58%) carried three or more copies of chromosome 9, revealing a high degree of chromosome 9 instability in the tumors analysed in this study (Table 4). This instability was not related to the stage of the tumors. Homozygous deletion of the p16INK4A locus was assessed when no visible signal could be detected with the p16INK4A probe whereas one, two or more positive signals were revealed by the centromeric probe. Absence of any p16INK4A signal was found at a high frequency in the tumor cells of 10 of the 12 tumors analysed, indicating an homozygous deletion of the p16INK4A chromosomal locus (Table 4). A particular pattern was observed in the two remaining tumors. For tumor T43, no p16INK4A signal was observed in only 10% of the nuclei and we did not consider this case as homozygously deleted. However, the persistence of only one signal for p16INK4A in the majority of the nuclei analysed was typical of a loss of heterozygosity at the p16INK4A locus in this tumor. On the other hand, in tumor T19 a very high degree of

Table 4 FISH analysis of homozygous deletion of p16INK4A and chromosome 9 polysomy in NSCLC with p16INK4A negative immunostaining Tumor T13 T15 T16 T17 T19 T27 T31 T33 T34 T35 T40 T43

% of cells with 53 copies chrom.9

% of cells without p16INK4A signal

94 64 34 9 97 9 39 45 9 46 4 6

80 84 43 71 0 82 46 82 73 79 87 10

chromosome 9 polysomy was detected as 97% of the nuclei had more than two copies of chromosome 9 with more than 50% of them containing four, ®ve or six copies of chromosome 9. Despite the absence of homozygous deletions of p16INK4A in those two cases, we considered them as genetically altered. In total homozygous deletions were found in 4/7 adenocarcinomas (57%), 5/9 squamous carcinomas (56%) and in 1/5 basal cell carcinoma (20%). Overall, 48% (10/21) of the tumors with p16INK4A negative staining had an homozygous deletion. Examples of the FISH analysis for homozygous deletion and chromosome 9 polysomy are presented in Figure 4. Correlation between IHC and molecular analyses Table 5 summarizes the results obtained in all 43 tumors in correlation with IHC staining. Of the 21 tumos with p16INK4A negative staining, we observed that 18 harbored homozygous deletion or hypermethylation or mutation and one case had both mutation and methylation. In total, 90% of the tumors with p16INK4A protein negative expression displayed one of these three alternative genetic alterations. Inactivation by either of these mechanisms was not observed in one adenocarcinoma and one basal cell carcinoma with negative staining whereas missense mutations were found in three tumors which retained p16INK4A nuclear staining. As previously stated, discordancies between the p16INK4A antibody reactivities were detected in seven tumors. Of these, p16INK4A genetic or epigenetic alterations were detected in none of the three tumors stated as p16INK4A positive (T22, T23, T37) but strikingly, in all four tumors showing only low p16 INK4 staining (T30, T33, T41, T42) and that we considered as negative for p16INK4A expression. Correlation between p16INK4A genetic alterations and tumor stage and metastasis No correlation was found between IHC status and any speci®c abnormality of p16INK4A gene and cancer extension, including node invasion (Table 6). p16INK4A genetic and epigenetic alterations were obviously evenly distributed according to primary or metastatic site of the studied samples, and according to stage of the disease at the time of diagnosis.

Figure 4 Dual-color FISH evaluation of the 2230 P1 clone containing the P16INK4A gene (biotin-labeled, FITC detected, green) and the chromosome 9 probe (directly labeled with cyanine-3-dCTP, red). (A) Example of a normal pattern indicated by two signals for the centromere probe of chromosome 9 (red) and two signals for the 2230 P1 probe (green). (B) Homozygous deletion of 2230 P1 in tumor T17 indicated by the presence of two signals for the centromere of chromosome 9 (red) and no detectable signals for 2230 P1 (green). (C) Polysomy 9 and homozygous deletion of 2230 P1 in tumor T15 indicated by more than two signals for the centromere probe of chromosome 9 (red) and none for 2230 P1 (green)

p16INK4A inactivation in lung cancer S Gazzeri et al

Table 5 Tumor T1 T2 T3 T4 T5 T6 T7 T8 T9 T10 T11 T12 T13 T14 T15 T16 T17 T18 T19 T20 T21 T22 T23 T24 T25 T26 T27 T28 T29 T30 T31 T32 T33 T34 T35 T36 T37 T38 T39 T40 T41 T42 T43

501

p16INK4A genetic alterations in human NSCLC versus p16 staining

Histological type

RB staining

p16 staining

Mutation

DNA methylation

Homozygous deletion

Adk Adk Adk Adk Adk Adk Adk Adk Adk Adk Adk Adk Adk Adk Adk Adk Adk Adk Adk Sq Sq Sq Sq Sq Sq Sq Sq Sq Sq Sq Sq Sq Sq Sq Sq Bas. Cell Bas. Cell Bas. Cell Bas. Cell Bas. Cell Bas. Cell Bas. Cell Bas. Cell

7 + + + + + + 7 + 7 + + + + + + + + + + + + + + 7 + + + 7 + + + + + + + + + + + + + 7

+ + + + + + + + + + + + 7 7 7 7 7 7 7 + + + + + + + 7 7 7 7 7 7 7 7 7 + + + 7 7 7 7 7

7 7 missense 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 missense 7 7 missense 7 7 7 7 missense frame-shift 7 7 7 7 7 7 7 7 7 7 frame-shift 7 7

7 7 7 7 7 7 7 7 7 7 7 7 7 methylated 7 7 7 methylated 7 7 7 7 7 7 7 7 7 methylated methylated 7 7 methylated 7 7 7 7 7 7 methylated 7 7 methylated 7

nd nd nd nd nd nd nd nd nd nd nd nd deleted nd deleted deleted deleted nd 7 nd nd nd nd nd nd nd deleted nd nd nd deleted nd deleted deleted deleted nd nd nd nd deleted nd nd 7

ADK, adenocarcinoma, Sq, squamous carcinoma, Bas. Cell, basal cell carcinoma, nd, not down, Absent (7), Present (+). RB staining was previously performed using F8, PMG3 245 and RBA1 (C36) RB antibodies as described in Gouyer et al. (1994)

Table 6 p16INK4A genetic alterations according to tumoral status Stage I Mutationa n=6 Methylation n=7 Homo. Deletion n=10 TOTAL n=23

2 3 2 7

Primary site Stage Stage II III 2 0 3 5

2 3 1 6

LN Stage metastatic IV site 0 0 0 0

0 1 4 5

a

Three cases showed p16INK4A positive staining and one case withINK4A negative staining carried both mutation and methylation of P16INK4A. LN, lymph node

Discussion The focus of this paper has been placed on the analysis of p16INK4A as a tumor suppressor gene in the genesis of NSCLC, with the aim of providing a comprehensive study of p16INK4A inactivation in human lung tumors, correlating loss of p16INK4A protein expression with a genetic or epigenetic mechanism of inactivation. In this

study we con®rmed that the tumor suppressor gene p16INK4A which maps on chromosome band 9p21 is frequently inactivated in NSCLC. A lack of p16INK4A protein expression was found in half of the tumors examined here which is in agreement with other series and indicates a major role of p16INK4A in the carcinogenic process of human NSCLC (Kratzke et al., 1996; Sakaguchi et al., 1996; Geradts et al., 1995). However, discrepancies between the reactivities of the two p16INK4A antibodies used were detected in 16% of the cases, indicating that care must be taken in interpreting both positive and negative immunostaining specially with p16 INK4 antibody. Our results on comparison of the molecular analysis of p16INK4A alterations with immunostaining indicate that C20 immunostaining by itself was 100% concordant with the presence or absence of genetic or epigenetic alterations in p16INK4A gene. p16INK4A C20 antibody can thus be considered as a reliable antibody with high anity for p16INK4A protein and no false positives, whereas p16 INK4A (Pharmingen) immunostaining is more dicult to interprate.

p16INK4A inactivation in lung cancer S Gazzeri et al

502

Loss of p16INK4A expression in many tumors can be accounted for by either homozygous deletions or hypermethylations or mutations (Cairns et al., 1994, 1995; Merlo et al., 1995). Although these three alternative mechanisms have been described in lung cancer, they had never been assessed simultaneously in a single tumor series and in comparison with the actual status of the p16INK4A protein expression. Previous reports have described an increased frequency of p16 alterations in late stages or at metastatic sites over primary tumor sites (Kelley et al., 1995; Okamoto et al., 1995; Nakagawa et al., 1995; Kratzke et al., 1996; Marchetti et al., 1997). In contrast, we have not observed a di€erence in the distribution of p16INK4A genetic or epigenetic abnormalities as a function of disease stage or between tumors from primary or metastatic sites. On the basis of our genetic analysis, we have shown that IHC is a straightforward method to detect p16INK4A inactivation. Indeed, our results demonstrate a high correlation between IHC and genetic analysis since homozygous deletions, methylations or frameshift mutations were observed in nearly all of immunohistochemical negative cases. By referring to our criteria of analysis and taking a 10% threshold value for positive nuclei, IHC appears to be a rapid and reliable method for studying the inactivation of p16INK4A in lung tumors. In this respect, our results are supported by a recent report on primary head and neck squamous cell carcinoma where the same approach was used. This study also showed a high correlation between IHC results and those of genetic analysis (Reed et al., 1996). Although we detected all known mechanisms of genetic or epigenetic inactivation of p16INK4A, it is important to note than in four tumors studied here, missense mutations which were not polymorphisms neither created stop codons, nor impaired C-terminal epitope recognition by antibodies. Both missense mutation and DNA methylation were detected in one of these tumors. Since methylation in exon 1 of p16INK4A leads to gene silencing (Merlo et al., 1995) one possibility is that p16INK4A negative immunostaining could be due to methylation in this case. An alternative transcript for p16 (p16b) which consists of a distinct exon 1 (exon 1b) spliced onto the remaining exons 2 and 3 of p16 has recently been reported (Mao et al., 1995; Duro et al., 1995; Stone et al., 1995). Because of a shift in the ORF, the b transcript theoretically encodes a protein that is structurally distinct from p16INK4A protein. Indeed, the presence of a similar alternative transcript encoding a p19IARF protein has been reported in mouse cells (Quelle et al., 1995). A human p19IARF protein has also been characterized in various cells including normal ®broblasts (Larsen, personal communication). Because of the existence of the alternate protein, we supposed that the mutation we found in tumor T29 had no e€ect on p16INK4A inactivation but could alter the reading frame and the function of the p19IARF protein. It should be noted that the same mutation at codon 57 has been previously described in acute lymphoblastic leukaemia (Qesnel et al., 1995). The three other missense mutations have not been previously described. However two of them are

located at codons 74 and 83 which were previously reported as targets of mutations in various tumors (Pollock et al., 1996). The e€ect of these mutations on p16INK4A function is unclear since p16INK4A was expressed. Nevertheless some missense mutations, including a C to T transition at codon 83, have been shown to encode a functionally inactive protein (Koh, 1995; Shapiro et al., 1995; Arap et al., 1995). Furthermore the fact that RB protein was expressed in all these three tumors suggests that the p16INK4A missense mutations are functionally inactivating. Thus only a functional study of the protein encoded by these mutant genes could classify them as inactivating mutations of p16INK4A gene. However, again we cannot exclude the potential e€ect of some of these mutations on the alternate reading frame of a p19IARF. Two tumors displayed absence of IHC staining for p16INK4A without any obvious genetic or epigenetic alteration. Analysis of mRNA expression by Northern blotting revealed an apparently normal mRNA expression (data not shown). Beside an unusual possibility of IHC false negative results, we cannot exclude the possibility that mutations could have been missed by SSCP, or that a yet undetermined mechanism of genetic or epigenetic inactivation occurred in these cases. Our results demonstrate a major implication of the homozygous deletion in the inactivation of p16INK4A in NSCLC. They also indicate the importance of the technique chosen for this analysis. Indeed, FISH is the most reliable method to discriminate between normal and tumoral cells. Furthermore we found an apparent clonal out-growth of cells showing three or more chromosome 9 copies which indicates a high degree of chromosome 9 polysomy in these tumors, probably re¯ecting a generalized chromosomal instability as has previously been reported in bronchial metaplasia (Hittelman et al., 1996). In conclusion, we believe that the immunohistochemical test we described here is a suitable approach to screen rapidly for p16INK4A abnormalities in fresh-frozen tissue. Methods are now available for application to formalin-®xed paran sections. However, although negative immunostaining of p16INK4A is a strong indication for the presence of one of the three genetic or epigenetic above mentioned abnormalities leading to p16INK4A inactivation, great care must be taken in the interpretation of positive nuclear reactivity which may not necessarily indicate the presence of functional p16INK4A in some tumors.

Materials and methods Tissue samples Forty-three lung tumors were included in this study. In 36 cases, sample was obtained at surgery, and in the remaining cases at mediastinoscopy (six mediastinal lymph node metastasis and one bone metastasis). Samples were immediately frozen at surgery and kept at 7808C until study. For histological classi®cation, tumor samples were ®xed in formalin and/or alcoholic Bouin's ®xative. The study included 19 adenocarcinoma, 16 squamous carcinoma

p16INK4A inactivation in lung cancer S Gazzeri et al

according to WHO classi®cation (1981) and eight basaloid carcinoma considered as basal stem cell proliferation according to Brambilla et al. (1992). Immunohistochemistry p16INK4A immunostaining was performed on frozen sections ®xed with 4% paraformaldehyde using classical immunoperoxidase techniques. Two polyclonal antibodies were used: p16INK4A (1/500) (Pharmingen, CA) and p16 C-20 (Santa-Cruz, CA). After overnight incubation at 48C with the primary antibody, slides were washed in phosphatebu€ered saline (PBS) and then exposed to the secondary antibody, biotinylated donkey F(ab')2 anti-rabbit (1/1000; Kackson Laboratories, West Grove, PA) for 1 h at room temperature. They were then washed in PBS and incubated with the streptavidine-biotin-peroxidase complex (1/400; Dako) for 1 h at room temperature. The chromogenic substrate of peroxidase was a solution of 0.05% 3.3diaminobenzidine tetrahydrochloride/0.03% H 2O 2 / 10 mmol/l, imidazole on 0.05 mol/l tris bu€er, pH 7.6. The slides were counterstained with Harris' hematoxylin. Normal rabbit IgG at the same concentration as the primary antibodies served as negative controls.

controls and were included for every site examined. To rule out the possibility of incomplete digestion, all samples were digested twice with each of the enzymes in independent experiments. PCR ampli®cations from each of the duplicate digests were repeated at least twice to ensure reproducibility of the results. Fluorescent in situ hybridization (FISH) analysis

PCR-based methylation assay

Whole imprint preparations were performed by gentle pressing tumoral frozen tissue to silane-coated glass slides, air dried and stored at 7808C until use. For each case analyzed, a slide was stained with Giemsa and the samples used for touch prints were included in order to assess for adequate cellularity and representation of tumor cell populations. Slides were treated for 10 min at room temperature in a 48C nuclei isolation bu€er (50 mM KCl, 5 mM HEPES pH=8, 10 mM MgSO4, 3 mM DTT), then incubated in freezer during 10 min in nuclei isolation bu€er with TX-100 0.25%, washed with PBS and ®xed for 10 min in paraformaldehyde 4%. Fixation was quenched with Tris 100 nM pH=7.5 during 10 min and the slides were progressively dehydrated in an ethanol series (70%, 90%, 100%). Hybridization was performed with a probe speci®c for the (peri)centromeric region of chromosome 9 (Rocchi et al., 1991) and a 2230 P1 clone containing the entire genomic sequence of p16INK4A generously provided by Cairns et al. (1995). The centromeric probe was directly labeled by nick-translation with cyanine 3 dCTP (Amersham) and the 2230 P1 clone was labeled by random priming with biotin-14-dCTP (Life Technologies) according to the instructions of the manufacturer. One hundred and ®fty nanograms of each probe were precipitated with NaAc 0.3 M, 5 mg human cot DNA, 10mg of salmon sperm DNA and two volumes EtOH and resuspended in 10 ml of 50% formamide/10% dextran sulfate/26SSC. The mixture was immediately applied to the touch preparations which were then covered with an 18618 mm coverslip, sealed with rubber cement and left at room temperature for 2 h. Denaturation was performed at 868C for 10 min and the slides were incubated in a moist chamber overnight at 378C. Posthybridization washing was carried out at 458C in 26SSC/50% formamide (365 min) followed by three washes in 0.16SSC at 608C. Detection was performed at 378C using FITC-conjugated avidin for 30 min followed by three washes in 46SSC/0.1% tween 20 at 458C. The slides were counterstained with 4',6-diamidino-2-phenylindole (DAPI) in antifade solution and the preparations observed under an epi¯uorescent microscope (Zeiss Axiophot). Images were collected with a cooled CCD camera (C4880 Hamamatsu) mounted on the epifluorescent microscope. Imprint preparations were imaged using a 6361.25 N.A oil immersion objective and an intermediate magni®cation of 1.256. The exposure time was between 2 and 8 s.

The methylation status of genomic DNA was analysed using the modi®ed PCR methylation assay (GonzalezZulueta, 1995). One mg of genomic DNA was digested overnight with 10 units of methylation-sensitive (HpaII, cfoI, SacII and SmaI) and methylation-insensitive (MspI) enzymes. For the assessment of the methylation status of exon 1 of the p16INK4A gene, 50 ng of digested DNA were ampli®ed by PCR using primers ¯anking the restriction sites as previously described (Gonzalez-Zulueta et al., 1995). Labeling of the PCR products was performed directly with [a-33P]dATP during PCR. Five ml of PCR products were electrophoresed on a 5% polyacrylamide gel. Undigested and MspI-digested DNAs were used as

Acknowledgements We are grateful to Dr CJ Larsen for his positive critical review of the manuscript and Mary Callanan for English correction. We also thank Christiane Drevet, Pascal Perron and Yann Souillet for technical assistance and J Cairns and D Sidransky for kindly providing the 2230 P1 clone. The ®nancial supports of this work were from the `Groupement des enteprises francËaises et moneÂgasques dans la lutte contre le cancer (GEFLUC)', the `Programme hospitalier de recherche clinique (PHRC)' and the `Association pour la recherche sur le cancer (ARC)'.

SSCP analysis and sequencing Genomic DNA was extracted from fresh frozen tissues using standard methods. 100 ng of genomic DNA was ampli®ed in a total volume of 20 ml in a bu€er containing 8 pmol of sense and antisense primers, 200 mM of each dNTP, 16ampli®cation bu€er, 1.5 mM MgCl2, 5% of dimethylsulfoxide, 2 mCi [a33P]dATP and 2.5 units of Taq polymerase (Promega). Primers for exon 1 of p16INK4A gene were (5'-3'): GAGAGGGGGAGAGCAGGCAG (sense); GCAAACTTCGTCCTCCAGAGT (anti-sense). Primers for exon 2 of p16INK4A gene were (5'-3'): GCTCTACACAAGCTTCCTTTC (sense); TGAGCTTTGGAAGCTCT (anti-sense). PCR conditions used to amplify exons 1 and 2 of the p16INK4A gene were: one cycle at 948C for 5 min; 30 cycles at 948C (1 min), 608C (1 min) and 728C (1 min) each, and a ®nal extension at 728C for 10 min. The PCR product (4 ml) was diluted twofold with a bu€er containing 50% formamide, heat-denatured and applied to a 6% neutral polyacrylamide gel in presence of 2% glycerol. For exon 2, the PCR products were digested overnight with SmaI (2.5 U per 10 ml) before denaturation and loading on the gel. The cases with mobility shift were cloned in a pTag vector (pUC backbone derived) using the Ligator Kit (Ingenius) and sequenced using the sequenase sequencing kit (USB).

503

p16INK4A inactivation in lung cancer S Gazzeri et al

504

References Arap W, Nishikawa R, Furnari FB, Cavenee WK and Su Huang HJ. (1995). Cancer Res., 55, 1351 ± 1354. Brambilla E, Moro D, Veale D, Brichon PY, Stoebner P, Paramelle B and Brambilla C. (1992). Human Pathol., 23, 993 ± 1003. Cairns P, Mao L, Merlo A, Lee DJ, Schwab D, Eby Y, Tokino K, Van der Riet P, Blaugrund E and Sidransky D. (1994). Science, 265, 415 ± 417. Cairns P, Polascik TJ, Eby Y, Tokino K, Califano J, Merlo A, Mao L, Herath J, Jenkins R, Westra W, Rutter JL, Buckler A, Gabrielson E, Tockman M, Cho KR, Hedrick L, Bova GS, Isaacs W, Koch W, Schwab D and Sidransky D. (1995). Nat. Genet., 11, 210 ± 212. De Vos S, Miller CW, Takeuchi S, Gombart AF, Cho SK and Koe‚er HP. (1995). Genes Chrom. Cancer, 14, 164 ± 170. Duro D, Bernard O, Della Valle V, Berger R and Larsen CJ. (1995). Oncogene, 11, 21 ± 29. Geradts J, Kratzke RA, Niehans GA and Lincoln CE. (1995). Cancer Res., 55, 6006 ± 6011. Gonzalez-Zulueta M, Bender CM, Yang AS, Nguyen T, Beart RW, Van Tornout JM and Jones PA. (1995). Cancer Res., 55, 4531 ± 4535. Gouyer G, Gazzeri S, Brambilla E, Bolon I, Moro D, Perron P, Benabid AL and Brambilla C. (1994). Int. J. Cancer, 58, 818 ± 824. Guan K, Jenkins CW, Li Y, Nichols MA, Wu X, O'Keefe CL, Matera AG, Xiong Y. (1994). Genes Dev., 8, 2939 ± 2952. Hayashi N, Sugimoto Y, Tsuchiya E, Ogawa M and Nakamura Y. (1994). Biochem. Biophys. Res. Comm., 202, 1426 ± 1430. Hittelman WN, Kim HJ, Lee JS, Shin DM, Lippman SM, Kim J, Ro JY and Hong WK. (1996). J. Cell. Biochem., 255, 57 ± 62. Hussussian C, Struewing JP, Goldstein AM, Higgins PAT, Ally DS, Sheahan MD, Clark Jr WH, Tocker MA and Dracopoli NC. (1994). Nature Genet., 8, 15 ± 21. Jin X, Nguyen D, Zhang WW, Kyritsis AP and Roth J. (1995). Cancer Res., 55, 3520 ± 3523. Kamb A, Gruis NA, Weaver-Feldhaus J, Liu J, Harshman K, Tavtigian SV, Stocjert E, Day RS, Jonhson BE and Skolnick MH. (1994). Science, 264, 436 ± 440. Kelley MJ, Nakagawa K, Steinberg SM, Mulshine JL, Kamb A and Jonhson BE. (1995). J. Natl. Cancer Inst., 87, 756 ± 760. Koh J, Enders GH, Dynlacht BD and Harlow E. (1995). Nature, 375, 506 ± 510. Kratzke RA, Greatens TM, Rubins JB, Maddaus MA, Niewoehner DE, Niehans GA and Geradts J. (1996). Cancer Res., 56, 3415 ± 3420. Li Y, Nochols MA, Shay JW and Xiong Y. (1994). Cancer Res., 54, 6078 ± 6082. Liggett WH, Sewell DA, Rocco J, Ahrendt SA, Koch W and Sidransky D. (1996). Cancer Res., 56, 4119 ± 4123. Lukas J, Parry D, Aagaard L, Mann DJ, Bartkova J, Strauss M, Peters G and Bartek J. (1995). Nature, 375, 503 ± 506. Mao L, Merlo A, Bedi G, Shapiro GI, Edwards CD, Rollins BJ and Sidransky D. (1995). Cancer Res., 55, 2995 ± 2997. Marchetti A, Buttitta F, Pellegrini S, Bertacca G, Chella A, Carnicelli V, Tognoni V, Filardo A, Angeletti CA and Bevilacqua G. (1997). J. Pathol., 181, 178 ± 182.

Merlo A, Herman JG, Mao L, Lee DJ, Gabrielson E, Burger PC, Baylin SB and Sidransky D. (1995). Nature Med., 1, 686 ± 692. Nakagawa K, Conrad NK, Williams JP, Jonhson BE and Kelley MJ. (1995). Oncogene, 11, 1843 ± 1851. Nobori T, Miura K, Wu DJ, Lois A, Takabayashi K and Carson DA. (1994). Nature, 368, 753 ± 756. Okamoto A, Demetrick DJ, Spillare EA, Hagiwara K, Hussain SP, Bennett WP, Forrester K, Gerwin B, Serrano M, Beach DH and Harris CC. (1994). Proc. Natl. Acad. Sci. USA, 91, 11045 ± 11049. Okamoto A, Hussain SP, Hagiwara K, Spillare EA, Rusin MR, Demetrick DJ, Serrano M, Hannon GJ, Shiseki M, Zariwala M, Xiong Y, Beach DH, Yokota J and Harris CC. (1995). Cancer Res., 55, 1448 ± 1451. Otterson GA, Kratzke RA, Coxon A, Whan Kim Y and Kaye FJ. (1994). Oncogene, 9, 3375 ± 3378. Packenham JP, Taylor JA, White CM, Anna CH, Barrett JC and Devereux TR. (1995). Clin. Cancer Res., 1, 687 ± 690. Pollock P, Pearson JV and Hayward NK. (1996). Genes Chrom. Cancer, 15, 77 ± 88. Quelle DE, Zindy F, Ashmun RA and Sherr CJ. (1995). Cell, 83, 993 ± 1000. Quesnel B, Fenaux P, Philippe N, Fournier J, Bonneterre J, Preud-homme C and Peyrat JP. (1995). Br. J. Cancer, 72, 351 ± 353. Reed AL, Califano J, Cairns P, Westra WH, Jones RM, Koch W, Abrendt S, Eby Y, Sewell D, Nawroz H, Bartek J and Sidransky D. (1996). Cancer Res., 56, 3630 ± 3633. Rocchi M, Archidiacono N, Ward DC and Baldini A. (1991). Genomics, 9, 517 ± 523. Rusin MR, Okamoto A, Chorazy M, Czayzewski K, Harasim J, Spillare EA, Hagiwara K, Hussain SP, Xiong Y, Demetrick DJ and Harris CC. (1996). Int. J. Cancer, 65, 734 ± 739. Sakaguchi M, Fujii Y, Hirabayashi H, Yoon HE, Komoto Y, Oue T, Kusafuka T, Okada A and Matsuda H. (1996). Int. J. Cancer, 65, 442 ± 445. Serrano M, Gomez-Lahoz E, Depinho RA, Beach D and Bar-Sagi D. (1995). Science, 267, 249 ± 252. Serrano M, Hannon GJ and Beach D. (1993). Nature, 366, 704 ± 707. Shapiro GI, Park JE, Edwards CD, Mao L, Merlo A, Sidransky D, Ewen ME and Rollins BJ. (1995). Cancer Res., 55, 6200 ± 6209. Sherr CJ. (1996). Science, 274, 1672 ± 1677. Spillare EA, Okamoto A, Hagiwara K, Demetrick DJ, Serrano M, Beach D and Harris DC. (1996). Mol. Carc., 16, 53 ± 60. Stone S, Jiang P, Dayananth P, Tavtigian SV, Katcher H, Parry D, Peters G and Kamb A. (1995). Cancer Res., 55, 2988 ± 2994. Tam SW, Shay JW and Pagano M. (1994). Cancer Res., 54, 5816 ± 5820. WHO Classi®cation. (1981). Histological typing of lung tumors. International histological classi®cation of tumors. Geneva: World Health Organisation. Wiest JS, Franklin WA, Otstot JT, Forbey K, Varella-Garcia M, Rao K, Drabkin H, Gemmill R, Ahrent S, Sidransky D, Saccomanno G, Fountain JW and Anderson MW. (1997). Cancer Res., 57, 1 ± 6.