Oncogene (1999) 18, 5423 ± 5434 ã 1999 Stockton Press All rights reserved 0950 ± 9232/99 $15.00 http://www.stockton-press.co.uk/onc Functional evaluation of tumour-speci®c variants of p16INK4a/CDKN2A: correlation with protein structure information Margarida Ruas1, Sharon Brookes1, Neil Q McDonald1 and Gordon Peters*,1 1Imperial Cancer Research Fund, PO Box 123, 44 Lincoln's Inn Fields, London WC2A 3PX Inherited mutations in the CDKN2A/INK4a/MTS1 may harbour CDKN2A mutations, a proper under- tumour suppressor gene on chromosome 9p21 are standing of the impact of these mutations is desirable associated with familial predisposition to melanoma and (Dracopoli and Fountain, 1996; Hayward, 1996). other tumour types. Nonsense and missense mutations The major challenge is to distinguish between are also found in a variety of sporadic cancers, and over innocuous polymorphisms and real germline muta- 140 sequence variants have already been recorded in the tions that impair the function of CDKN2A to dierent literature. In assessing the relevance of these variants degrees. Moreover, since the sporadic mutations and for counselling members of aected families, it is described thus far have invariably been found by important to distinguish inactivating mutations from analysing tumour cells, those regarded as polymorph- harmless polymorphisms. Existing functional assays have isms, because of their presence in adjacent unaected frequently reached con¯icting conclusions and no single tissue, could represent rare germline variants that test appears adequate. Here we evaluate a number of predispose to cancer. Conversely, it is possible that alternatives including a novel assay based on retroviral the sequence variants of CDKN2A recorded in sporadic delivery of p16INK4a cDNAs into human diploid ®bro- tumours may re¯ect a mutator phenotype caused by a blasts. Among the 17 sequence variants analysed, three lesion in DNA repair or `caretaker' genes (Kinzler and distinct categories can be distinguished: those that Vogelstein, 1997), and may not necessarily aect the abrogate the binding of p16INK4a to CDK4 and CDK6, function of the protein. The prevalent use of PCR in those that alter the properties of the protein without the analysis of CDKN2A gene sequences also raises the preventing it from interacting with CDKs, and those that possibility that some recorded mutations are PCR have no discernible eect on protein function. These artefacts. distinctions can be rationalized by considering the impact The product of the CDKN2A gene is a 16 kDa of the amino acid changes on the three-dimensional protein, p16INK4a, comprising four ankyrin-type repeat structure of the protein. motifs, that binds speci®cally to and blocks the function of the cyclin dependent kinases CDK4 and Keywords: cyclin-dependent kinase inhibitor; tumour CDK6 (Serrano et al., 1993). When associated with D- suppressor; familial melanoma; ankyrin repeats; type cyclins, CDK4 and CDK6 promote passage replicative senescence through the G1 phase of the cell cycle by contributing to the phosphorylation and functional inactivation of the retinoblastoma gene product, pRb (reviewed in Sherr, 1994; Weinberg, 1995). Thus, p16INK4a and other Introduction members of the INK4 family (p15INK4b, p18INK4c and p19INK4d) are all capable of imposing a G1 cell cycle The CDKN2A/INK4a/MTS1 tumour suppressor gene arrest that is dependent on the presence of functional on human chromosome 9p21 is a frequent target of pRb (Sherr and Roberts, 1995; Ruas and Peters, 1998). inactivating mutations in both sporadic and familial All four proteins bind directly to CDK4 and CDK6 in cancers (reviewed in Ruas and Peters, 1998). For vitro but, despite their biochemical and structural example, the gene sustains homozygous deletions in similarities, only p16INK4a has the credentials of a 40 ± 60% of bladder cancers, gliomas and T-cell acute tumour suppressor. A likely explanation is that lymphoblastic leukaemias (T-ALL), is silenced by p16INK4a accumulates as cells reach the end of their hypermethylation of the promoter region in colon, replicative lifespan (senescence) so that inactivation of breast and non-small cell lung cancers (NSCLC), and p16INK4a favours the emergence of immortal cell clones sustains missense and nonsense mutations in a (Alcorta et al., 1996; Hara et al., 1996; Loughran et al., signi®cant proportion of biliary tract, pancreatic and 1996; Rezniko et al., 1996; Yeager et al., 1998). oesophageal cancers (Ruas and Peters, 1998). Over 140 The known properties of p16INK4a have been single amino acid changes have already been noted in exploited in the development of functional assays, human tumour cells, aecting 95 of the 156 residues in which fall broadly into three types: (i) the ability of the protein. Germline mutations are associated with p16INK4a to bind directly to CDK4 and CDK6; (ii) its inherited prediposition to melanoma and other cancers ability to inhibit the phosphorylation of pRb by cyclin (Ruas and Peters, 1998) and since it is estimated that D-CDK complexes; and (iii) its ability to cause cell approximately 10% of all melanomas have a familial cycle arrest when ectopically expressed. Only a component and as many as half of the familial cases minority of the known missense variants have been functionally analysed and although the severe loss-of- function mutants are easily detected, a substantial *Correspondence: G Peters Received 25 January 1999; revised 12 March 1999; accepted 15 April proportion of the variants tested thus far behave 1999 ambiguously in dierent assay systems or have given Function and structure of p16 mutants MRuaset al 5424 inconsistent results when analysed by dierent labora- in a simple in vitro association assay (Parry and Peters, tories (Koh et al., 1995; Lukas et al., 1995; Ranade et 1996). The dierent single amino acid changes were al., 1995; Reymond and Brent, 1995; Shapiro et al., introduced in the wild-type p16INK4a cDNA sequence by 1995; Wick et al., 1995; Yang et al., 1995; Enders et al., site-directed mutagenesis and the proteins were 1996; Lilischkis et al., 1996; Parry and Peters, 1996; synthesized by coupled transcription and translation Tevelev et al., 1996; Yarbrough et al., 1996; Zhang and in rabbit reticulocyte lysates. Binding assays were Peng, 1996; Arap et al., 1997; Harland et al., 1997; Sun performed by mixing the p16INK4a variants with et al., 1997). There are a number of possible reasons. CDK4 or CDK6 produced by in vitro translation in The ®rst is that any assay that requires the production the presence of [35S]methionine. The mixtures were of recombinant p16INK4a in bacteria is vulnerable to immunoprecipitated with antiserum against full length aggregation of the protein (Boice and Fairman, 1996; p16INK4a (Parry and Peters, 1996) and the amount of Tevelev et al., 1996; Zhang and Peng, 1996) and to labelled CDK that co-precipitated with p16INK4a was variability caused by having to recover the protein assessed by SDS ± PAGE (Figure 1). In preliminary from insoluble inclusion bodies. A second point is that tests it was con®rmed that the variant p16INK4a proteins any assay conducted in vitro, such as CDK binding or were synthesized and immunoprecipitated with similar kinase inhibition, may not be a true re¯ection of the in eciency (data not shown). vivo situation. For example, it is now clear that cyclin D-dependent kinase activity is associated with high molecular weight complexes in vivo and that D cyclins are rarely if ever found in simple binary associations with CDK4 or CDK6 (Della Ragione et al., 1996; Mahony et al., 1998; Musgrove et al., 1998; McConnell et al., 1999). Similarly, although p16INK4a does participate in binary complexes with CDKs, it does not associate detectably with D cyclins in vivo (Hall et al., 1995; Parry et al., 1995; Guan et al., 1996; McConnell et al., 1999). This could in part explain why the ability of p16INK4a to inhibit cyclin-CDK complexes assembled using baculovirus driven expres- sion in insect cells does not give a linear dose response (for example, see Serrano et al., 1993; Koh et al., 1995; Lukas et al., 1995; Ranade et al., 1995; Parry and Peters, 1996). Finally, virtually all assays based on the ectopic expression of p16INK4a in eukaryotic cells are vulnerable to variations in the levels of expression achieved. One of the most frequently used is based on the co-transfection of p16INK4a cDNA with a cell surface marker so that the proportion of transiently transfected cells arrested in G1 can be measured by ¯ow cytometry (Koh et al., 1995; Lukas et al., 1995; Parry and Peters, 1996). Assays of this type do not lend themselves to quantitative appraisal of variants that may be only partially or conditionally impaired. Because of these problems, we have sought to devise functional tests that are relatively simple to perform and interpret, and that can detect partially as well as fully impaired variants of p16INK4a. Here we compare the ecacy of established and novel assays by testing a series of sporadic and familial variants of p16INK4a, several of which have given ambiguous results in previous studies. Two assays emerge as the most reliable indicators of p16INK4a function: binding to CDK4 and CDK6 in vitro and measurements of cell proliferation following infection with retroviruses encoding p16INK4a. Our con®dence in these assays is supported by relating the observed properties of each variant to the predicted eects of the amino acid substitution on the three- dimensional structure of p16INK4a. Figure 1 Binding