hypermethylation of the cell cycle inhibitor p15ink4b 3′-untranslated region interferes with its...

Hypermethylation of the cell cycle inhibitor p15INK4b 3'-untranslated regioninterferes with its transcriptional regulation in primary lymphomas

Marcos Malumbres1, Ignacio PeÂ rez de Castro2, Javier Santos2, JoseÂ FernaÂ ndez Piqueras2 andAngel Pellicer1

1Department of Pathology and Kaplan Comprehensive Cancer Center, New York University Medical Center, 550 First Avenue,New York, NY 10016, USA; and 2Laboratorio de GeneÂtica Molecular Humana, Dep. de BiologõÂa, Universidad AutoÂnoma deMadrid, Cantoblanco, 28049 Madrid, Spain

The cyclin-dependent kinase inhibitor p15INK4b has beenshown to be involved in human and rodent tumors andseems to act as a tumor suppressor gene in hematologicalmalignancies. Alterations of this gene in tumors includemainly homozygous deletions and hypermethylation ofthe CpG island in the promoter region. In this work, wedescribe a new area sensitive to methylation in the 3'untranslated region (UTR) of the murine p15INK4b gene.This region shows di�erent levels of methylationdepending on the tissues, being relatively highlymethylated in brain and gut, and weakly methylated inliver, spleen or thymus. DNA methylation and expressionis similar in both maternal and paternal alleles indicatingno imprinting e�ect. Although methylation of the p15INK4b

3'-UTR is low in normal thymus, increased levels (up to100%) of speci®c methylation in this region are found inup to 30% of radiation- or carcinogen-induced thymiclymphomas, correlating with decreased gene expression.Hypermethylation of the p15INK4b 3'-UTR frequentlyoccurs in tumors with loss of heterozygosity (LOH)but without methylation of the promoter CpG island orintragenic mutations. Furthermore, in vitro CpG methy-lation of the 3'-UTR produces reduced levels of aluciferase reporter in cultured cells. Methylation of twoCpG sites in a 120 bp region is su�cient to interferewith transcription of the reporter gene. These datasuggest that although the levels of p15INK4b in normaltissues can be mainly determined by promoter regulatoryelements, strong hypermethylation of the 3'-UTR caninterfere with transcription. Thus, hypermethylation ofthe 3'-UTR may explain the lack of p15INK4b geneexpression in a subset of tumors with no promotermethylation and could be a new alternative mechanismfor tumor suppressor gene inactivation in tumorigenesis.

Keywords: DNA methylation; p15INK4b CDK inhibitor;3'-UTR; T-cell; lymphoma

Introduction

The p15INK4b gene (Hannon and Beach, 1994) belongsto the INK4 family of cyclin-dependent kinaseinhibitors (CDKIs), involved in the regulation of themammalian cell cycle progression (Sherr and Robert,1995). The members of the INK4 family (p16INK4a,

p15INK4b, p18INK4c and p19INK4d) speci®cally inhibit thecomplexes formed by cyclin D (D1, D2 or D3) and thecyclin-dependent kinases (CDKs) 4 or 6. These CDKIsshare multiple ankyrin repeat motifs and modulate thephosphorylation of the retinoblastoma (Rb) protein bycyclin D-CDK4/6, eliciting a G1 block of the cell cycle(Sherr and Robert, 1995; GranÄ a and Reddy, 1995;Harper and Elledge, 1996). A di�erent family ofCDKIs includes p21CIP1,WAF1, p27KIP1 and p57KIP2. Themembers of this family inhibit all G1/S phase CDKs(Harper and Elledge, 1996; Sherr, 1996). p21CIP1 isinduced directly by p53 and mediates p53-dependentcell cycle arrest and inhibition of DNA replication.p27KIP1 tends to accumulate in quiescent cells anddeclines in response to growth factor stimulation (Sherrand Robert, 1995; Harper and Elledge, 1996).The p15INK4b protein is an important mediator of the

antiproliferative e�ect of TGF-b (Hannon and Beach,1994, ReynisdoÂ ttir et al., 1995). TGF-b regulatesp15INK4b at least at two levels: producing a mRNAaccumulation through several Sp1 sites in its promoter(Hannon and Beach, 1994; Li et al., 1995) andincreasing the protein stability (Shandu et al., 1997).The antiproliferative e�ect of TGF-b, however, is notrestricted to p15INK4b, since TGF-b can produce cellcycle arrest in p15INK4b-de®cient cells through repressionof the CDK tyrosine phosphatase Cdc25A (Iavaroneand MassagueÂ , 1997). Formation of p15INK4b-CDK4complexes at the expense of cyclin D increases the poolof free p27KIP1, producing a inhibition of the cyclin E-CDK2 complexes as well (Sherr and Robert, 1995;Peters, 1994).Di�erent evidences have shown the involvement of

some CDKIs in carcinogenesis. The p16INK4a andp15INK4b genes, which lie very close in the human andmurine chromosomes (Kamb et al., 1994; Quelle et al.,1995), are frequently deleted in a wide range of tumors(Harper and Elledge, 1996; Hirama and Koe�er,1995). Deletions most frequently involve both p16INK4a

and p15INK4b. Although p16INK4a is thought as the maintarget for these deletions (Cairns et al., 1995), speci®cdeletion of p15INK4b has been found in some cases (Jenet al., 1994; Glendeling et al., 1995; Rasool et al.,1995). Inactivation of these genes occurs also by pointmutations (Pollock et al., 1996) or by hypermethylationof their promoter region (Merlo et al., 1995; Herman etal., 1996). Both p16INK4a and p15INK4b genes, in humanand rodents, show a CpG island in their promoterregion and de novo methylation of this region has beenfound in a signi®cant percentage of tumors, beingcorrelated with the lack or decreased gene expression(Merlo et al., 1995; Herman et al., 1997; Malumbres et

Correspondence: A PellicerReceived 11 March 1998; revised 22 July 1998; accepted 22 July 1998

Oncogene (1999) 18, 385 ± 396ã 1999 Stockton Press All rights reserved 0950 ± 9232/99 $12.00

http://www.stockton-press.co.uk/onc

al., 1997). Although p16INK4a is inactivated byhypermethylation in a wide spectrum of tumors, thehypermethylation of the p15INK4b CpG island seems tobe more restricted to hematological malignancies(Herman et al., 1996, 1997). None of the othermembers of the INK4 family, p18INK4c or p19INK4d,have been de®nitively involved in tumorigenesis.Since the importance of p15INK4b as a tumor

suppressor gene and its inactivation by promoterhypermethylation in hematological malignancies (Her-man et al., 1997; Batova et al., 1997; Malumbres et al.,1997), we have further investigated the pattern ofmethylation of p15INK4b in normal and tumoral tissues.We found some thymic lymphomas in which the lackof p15INK4b expression cannot be explained by promotermethylation. We report in this article that this subsetof tumors have a high level of hypermethylation in the3'-untranslated region (UTR) and that hypermethyla-tion of this area produces decreased gene expression ina in vitro reporter system. Although no methylation ofthe p15INK4b promoter region is found in normal tissues,the 3'-UTR is di�erentially methylated in normal cells,varying the level of methylation in the di�erent tissues.This methylation is not allele dependent, showing noimprinting e�ect. Surprisingly, increased speci®chypermethylation of this area is found in p15INK4b-de®cient thymic lymphomas that do not havemethylation in the CpG island of the promoter

region. Using a luciferase reporter system, we showhere that in vitro hypermethylation of the 3'-UTRproduces decreased gene expression in NIH3T3 cellsand this e�ect is mediated by at least two CpG sites inthe p15INK4b 3'-UTR. Thus, both types of hypermethyla-tion, in the promoter or in the 3'-UTR, may act asalternative or cooperating mechanisms for p15INK4b

inactivation in T-cells.

Results

Characterization of a new methylation region in the 3'-UTR of p15INK4b

The murine p15INK4b gene, as the human counterpart,consists of two exons and one single intron betweenthem. In RF/J mice, the exons 1 and 2 are ¯anked bytwo XhoI sites in the 5' and 3' untranslated regions(UTR), giving rise to a 4.5 kb XhoI band (Figure 1).During the screening of CpG island methylation inthymus DNA, an extra band of about 6 kb wasdetected in XhoI6EcoRI digestions that could not beexplained by methylation in the promoter region(Figure 2a). This band was present in all the samples,including normal and tumoral tissues. In order tocharacterize this DNA fragment, genomic DNA from aRF/J thymus was digested with XhoI and EcoRI and

Figure 1 Physical map of the murine p15INK4b gene from RF/J mice. The position of some of the primers used in this work isindicated (F1, Mp15E1F; F, Mp15F; 1R, Mp15-3'-1R; 3R, Mp15-3'-3R). The recognition sites and the expected size of thefragments produced after the digestion is also shown for the restriction enzymes more relevant to this work. In addition, theexpected fragments after RT±PCR ampli®cation and the position of the probes is indicated. Asterisks show restriction sites presentin RF/J or DBA/2 DNA but absent in the C57BL/6J strain

Tumor-associated hypermethylation of the p15 3'-UTRM Malumbres et al

386

the fragments of about 6 kb were separated by gelelectrophoresis and subcloned in pBLUESCRIPT KS+

digested with EcoRI6SalI or SalI alone. This partialgene library was used to transform E. coli and thespeci®c clones with the 6 kb fragment were detected byplate hybridization using the exon 1 of p15INK4b as aprobe. After selection and sequencing, these clonesshowed the presence of 6.0 kb XhoI ±EcoRI insertcontaining the genomic p15INK4b gene and downstreamsequences (Figure 1). There were no signi®cantnucleotide di�erences with the previously knownp15INK4b sequence. Interestingly, the internal XhoI sitein the 3'-UTR sequence was conserved in the clonedDNA fragment, indicating that it was protected fromdigestion in the genomic DNA. A similar cloningprocedure was performed using genomic DNA from aC57BL/6J mammary gland. The cloned insert containsthe same p15INK4b gene with some nucleotide di�erences.Some of them eliminate the XhoI/TaqI and NcoI sitesas indicated in Figure 1.Thus, the XhoI site in the 3'-UTR is partially

prevented from digestion in the normal thymusDNA from RF/J mice. Further analysis showedthat the intensity of the 6 kb band was even higherin kidney DNA (Figure 2a), indicating a strongerprevention of digestion in this site. To checkwhether this lack of digestion was due to CpGmethylation or other factors like conformationaldi�erences in the DNA, two di�erent non methyla-tion-sensitive restriction enzymes in the same regionwere used. Total kidney DNA was digested withNcoI to detect digestion in a NcoI site placed 13 bpdownstream the XhoI site (Figure 1), or with TaqI,

whose recognition sequence (TCGA) is internal tothe XhoI site (CTCGAG) and is not sensitive tomethylation. Hybridization of these DNAs with ap15INK4b 3'-UTR speci®c probe (Figure 1) showedtotal digestion (Figure 2b and c) indicating thatCpG methylation is probably the cause of the lackof digestion at the XhoI site. Other methylation-sensitive restriction enzyme, HpaII, was checked formethylation. There is one HpaII site in the codingsequence of exon 2, and other HpaII site 317 bpdownstream, in the 3'-UTR of p15INK4b (Figure 1).After digestion with HpaII6EcoRI and hybridizationwith the 3'-UTR probe, two bands of 317 bp andabout 1.5 kb are expected. The presence of a thirdband of 1.8 kb indicates partial digestion of theHpaII site in the 3'-UTR and, in concordance withthe results obtained with XhoI, the intensity of thisextra band is stronger in kidney than in thymus.The control digestions with MspI (a restrictionenzyme which recognizes the same sequence thanHpaII, but is methylation-insensitive) showed com-plete digestion. No bands longer than 1.8 kb weredetected in the HpaII digestions, indicating that themethylation is restricted to the HpaII site in the 3'-UTR and it does not a�ect the HpaII site in thecoding region of exon 2. The p15INK4b gene fromother inbred strains as 129/J or DBA/2 wasampli®ed in a long-template PCR using primersMp15E1F and Mp15-3'1R. These two mouse strainsalso have the XhoI site in the 3'-UTR, as detectedby restriction mapping and sequencing, and similarpattern of methylation was found in the 3'-UTR ofthese strains (data not shown).

Figure 2 Methylation of the XhoI and HpaII sites in the p15INK4b 3'-UTR. (a) Genomic DNA from normal RF/J thymus or kidneywas digested with XhoI6EcoRI. The 4.5 kb band corresponds to the expected size for p15INK4b (Figure 1). An additional band of6 kb is present in both tissues. (b) Digestion with NcoI6EcoRI showing total digestion of the 3'-UTR NcoI site. This NcoI is notpresent in C57BL/6J. The 10.5 and 12 kb are produced by an EcoRI site upstream the p15INK4b gene. (c) Digestion of genomicDNA with TaqI. Only the expected bands (785 and 342 bp) are present indicating total digestion in thymus and kidney. The strainC57BL/6J lacks the internal XhoI/TaqI site showing a unique band of 1124 bp. (d) Genomic RF/J DNA from thymus or kidney wasdigested with EcoRI and HpaII (methylation-sensitive) or MspI (methylation-resistant). The 1.8 kb band shows lack of digestion inthe HpaII site located in the 3'-UTR, but not in the HpaII site located on the exon 2. The intensity of this 1.8 kb band is stronger inkidney than in thymus. The exon 1 probe was used in a, whereas the 3'-UTR probe (Figure 1) was used in b, c and d


387

Tissue-speci®c levels of 3'-UTR methylation

The pattern of p15INK4b 3'-UTR methylation was thenstudied in di�erent normal tissues. Genomic DNAwas isolated from the brain, gut, heart, kidney, liver,lung, mammary gland, salivary gland, spleen, thymusand testis of DBA/2 mice. Since these mice onlycarry the 4.5 kb allele, the level of methylation wasquanti®ed as the percentage of the 6.0 kb-banddetected in XhoI6EcoRI digestions. The level ofmethylation in the p15INK4b-UTR ranged from 20 ±75% in these tissues (Figure 3a). Whereas liver,spleen and thymus show low methylation at theXhoI site (about 20 ± 25%), the 6.0-kb bandrepresents up to 60 ± 75% in brain, testis or gut.No signi®cant variation was found in thymus,kidney or lung tissues obtained from 1.5-month or6-month mice (data not shown).

p15INK4b expression and 3'-UTR methylation do notcorrelate in physiological conditions

Although undermethylation in the promoter regionsusually correlate with gene expression, this is not afrequent ®nding in non-promoter regions as codingsequences, introns or 3'-UTR (Mullins et al., 1987;Borrello et al., 1992). Since no methylation in thepromoter CpG island of p15INK4b is found in normaltissues, we analysed whether the 3'-UTR methylationwas related to physiological tissue-speci®c geneexpression. The abundance of p15INK4b transcripts wasevaluated by RT±PCR using RNA from di�erentDBA/2 tissues. As shown in Figure 3b, p15INK4b

expression was detected in most of tissues, being highin testis and lung, and low in liver or brain. Mosttissues with a low or intermediate level of methylation(thymus, spleen, lung or kidney) express p15INK4b,however, a comparison of these results in Figure 3aand b indicates that there is no signi®cant correlationbetween 3'-UTR methylation and expression. Forinstance, the level of expression in thymus, kidneyand gut are similar, although they are methylated todi�erent extent (20%, 55%, 70%). In addition, lowexpression is found in liver (520% methylation) andbrain (65% methylation).We also checked whether 3'-UTR methylation is

sensitive to cell cycle-dependent variation or changeswith the alteration of p15INK4b expression produced byexternal signals. Elimination of p15INK4b expressionduring human lymphocyte mitogenesis was previouslydemonstrated and correlated to increased Rb phos-phorylation (Lois et al., 1995). Thus, p15INK4b mRNAlevels and DNA methylation were monitored in theactivation of lymphocytes by mitogenic signals. Spleenlymphocytes from DBA/2 mice were activated withanti-CD3e, phorbol myristate acetate (PMA), ionomy-cin and concanavalin A. Cell samples were taken atdi�erent times (0, 20 and 50 h) and p15INK4b DNAmethylation and expression was analysed. No methyla-tion changes were observed neither in the promoter norin the 3'-UTR. However, p15INK4b expression wasdecreased in the activated lymphocytes as comparedwith non-activated cells (Figure 4). Thus, p15INK4b

expression was inhibited after 20 h of lymphocyteactivation without correlating with DNA methylationchanges.

Lack of imprinting in the murine p15INK4b gene

Since methylation is a frequent phenomenon involvedin parental imprinting, the restriction patterns in thep15INK4b 3'-UTR could suggest an allele-speci®cmethylation. To test this possibility, DNA and RNAwere extracted from the inbred strains DBA/2 andC57BL/6J and from their F1 o�springs: B6D2F1(maternal C57BL/6J and paternal DBA/2) andD2B6F1 (maternal DBA/2 and paternal C57BL/6J).As described above, both strains, DBA/2 and C57BL/6J, di�er in some nucleotides in their p15INK4b sequencewhich produce a polymorphism in the XhoI site of the3'-UTR. This site is present in DBA/2 but not inC57BL/6J, allowing the recognition of the parentalorigin of each allele. DNA from these strains wasdigested with XhoI6EcoRI, gel electrophoresed andtransferred to nitrocellulose membranes.

Figure 3 Tissue-speci®c patterns of p15INK4b methylation andexpression. (a) Tissue-speci®c methylation in the 3'-UTR ofp15INK4b. Genomic DNA from eleven di�erent DBA/2 tissues wasdigested with XhoI6EcoRI, gel electrophoresed, transferred tonitrocellulose membranes and hybridized with a p15INK4b speci®cprobe. The 6.0-kb band produced by partial digestion of the XhoIsite in the 3'-UTR shows di�erent relative intensities in thedi�erent tissues. The percentage of signal in the 6.0-kb band intwo or three di�erent hybridizations was plotted as an indicationof the level of 3'-UTR methylation. (b) p15INK4b expression insome mouse tissues. The histogram shows the relative levels ofexpression after normalization with the b-actin transcript andusing the testis expression as relative value 1. p15INK4b cDNA wasdetected after hybridization with a speci®c probe. b-actinexpression was quanti®ed after ethidium bromide staining


388

In normal kidney, about 50% of the DNA ismethylated (Figures 2 and 3) due either to partialmethylation of both alleles or complete methylation ofone allele and lack of methylation in the other. Forinstance, if the methylation was speci®c of the maternalallele, this would mean that the maternal allele istotally methylated and the paternal allele is notmethylated at all. In that case, the analysis of the F1o�spring would show di�erent levels of methylationdepending on the parental origin of the allele. Asdetected after hybridization with the p15INK4b 3'-UTRprobe, the DBA/2 allele (4.5 kb band) from each of theB6D2F1 or D2B6F1 o�spring is methylated to thesame extent, with independence of its paternal(B6D2F1) or maternal origin (D2B6F1). About 50%of the 4.5-kb band is methylated in both types of F1mice, indicating no parental e�ect on the methylationlevels (Figure 5a).According to this, the level of gene expression of

both alleles was found to be similar with independenceof the parental origin (Figure 5b). RT ±PCR wasperformed using primers speci®c for the exon 1 and the3'-UTR region of the murine p15INK4b gene, to avoidDNA contamination as template for the PCR (thestrategy for this RT ±PCR and the size of the expectedbands is shown in Figure 1). The parental origin of theallele was detected by digestion of the PCR productswith XhoI, since the XhoI site in the 3'-UTR of DBA/2is missing in the C57BL/6J cDNA (Figures 1 and 5b).

De novo methylation of the p15INK4b 3'-UTR in thymiclymphomas

To analyse if this newly discovered methylationsensitive area could be involved in the proliferationof neoplasias, as we had previously shown for theCpG island of the promoter, the pattern of

methylation in the 3'-UTR of p15INK4b was analysedin an extensive panel of thymic lymphomas inducedby g-radiation or neutron-radiation or by treatmentwith MNU. Since there is a XhoI site in both thepromoter and the 3'-UTR, the digestion withXhoI6EcoRI allows to detect methylation in both5'- and 3'-regions, as graphically explained in Figure6. Some tumors induced in RF/J or 129/J mice showthe pattern expected for the normal thymus, in whichabout 20% of the DNA is present as a 6-kb band dueto partial methylation of the 3'-UTR. About 80% of

Figure 4 p15INK4b methylation and expression in lymphocyteactivation. Spleen cells were analysed 0, 20, or 50 h afteractivation or 50 h without activation as a control. No changesin the methylation pattern are found in the 5' or 3' regions ofp15INK4b, whereas the level of expression decreases 20 h afterlymphocyte activation. RT±PCR was performed as described inMaterials and methods using 30 cycles of ampli®cation. Detectionof ampli®ed products as in Figure 3b

Figure 5 Lack of imprinting on the murine p15INK4b 3'-UTRmethylation. (a) Southern hybridization of the B6D2F1, D2B6F1and parental alleles, showing equally-distributed methylation inthe 3'-UTR with independence of the allele origin. C57BL/6Jshows a 6-kb band, since no internal XhoI is present in the 3'-UTR; DBA/2, which only carry the 4.5-kb band is 50%methylated giving rise to the 6-kb band. In F1 mice (D2B6F1and B6D2F1), 75% of the signal is in the 6.0-kb band; 50%corresponds to the C57BL/6J allele (which is a `true' 6.0 kb DNAfragment) and 25% comes from the methylated 4.5-kb band(DBA/2 allele). About 25% of the DBA/2 allele remainsunmethylated in all the F1 samples with independence of theparental origin. This membrane was hybridized with the 3'-UTRprobe (Figure 1). (b) RT±PCR products from C57BL/6J, DBA/2,B6D2F1 and D2B6F1 kidney RNA, showing equal levels ofexpression of the paternal and maternal alleles. cDNA from thesetissues was ampli®ed by PCR using primers Mp15F and Mp15-3'-3R (Figure 1). PCR products were digested with XhoI and theampli®ed fragments were detected by hybridization with a 3'-UTR probe after gel electrophoresis and blotting onto anitrocellulose membrane. The expected sizes of the RT±PCRproducts after XhoI digestion is explained in Figure 1


389

the signal remains as the 4.5-kb band. As previouslyshown (Malumbres et al., 1997), hypermethylation ofthe CpG island in some tumors is detected by thepresence of a larger (10.5 kb) band, indicating lack ofdigestion in the XhoI site located in the promoterregion (Figures 6 and 7). Some other tumors show noCpG island methylation; however, the percentage ofthe 6-kb band is increased to more than 60% or even100%, producing the disappearance of the 4.5-kbband and showing a strong de novo methylation in the3'-UTR of p15INK4b. In a few cases, the increasedmethylation in the 3'-UTR is coincident with themethylation of the 5' CpG island; in these samples,the lack of digestion in both the 5' and 3' XhoI sitesproduces an additional 12-kb band, corresponding tothe EcoRI fragment (Figures 6 and 7). The frequencyof methylation in the p15INK4b promoter and 3'-UTR inthe di�erent type of induced tumors is shown inFigure 6. In these data, only samples with amethylation in the CpG island higher than 30% areconsidered to be hypermethylated. However, thethreshold was increased to 60% in the 3-UTR sinceabout 20% of this area is methylated in the normalthymus. Hypermethylation of the CpG island in thepromoter region is the most frequent event, althoughhypermethylation of the 3'-UTR with independence ofthe promoter region occurs in up to 19% of radiation-induced tumors and 31% of MNU-induced tumors.Hypermethylation in both 5' and 3' areas is relativelyuncommon (Figure 6).p15INK4b expression was analysed in murine thymic

lymphomas by RT ±PCR using primers Mp15F and

Mp15R, as explained in Materials and methods.Ampli®cation of b-actin cDNA was used as a controlfor RNA quality and concentration. The correlationbetween p15INK4b gene inactivation and the methylationof the promoter CpG island has been shown previously(Herman et al., 1996; Malumbres et al., 1997). In oursamples, although most of the tumors (92%) withmethylation in the promoter region have decreasedp15INK4b expression (Malumbres et al., 1997), somelymphomas showed lack of p15INK4b expression withouthypermethylation of the CpG island. The analysis ofthe 3'-UTR methylation showed that most of thesetumors have a strongly increased methylation in thatregion (Figure 7). A representative sample of thetumors analysed is shown in Table 1. These tumorswere induced by g-radiation in B6RFF1 mice. Loss ofheterozygosity (LOH) is a frequent event in these cells(Santos et al., 1996; Malumbres et al., 1997). LOH inTable 1 was analysed by SSCP (Malumbres et al.,1997) and by using a new microsatellite adjacent to thep15INK4b gene (Malumbres et al., 1998). Seventeentumors in Table 1 show lack of p15INK4b expression.In one of them (R13B) the p15INK4b gene inactivationcan be attributed to a small deletion (data not shown).Ten out of the 16 remaining tumors (62.5%) havespeci®c methylation of the CpG island in the promoterand four of them (25%) show speci®c hypermethyla-tion of the 3'-UTR without promoter involvement.Interestingly, the only sample (6.3%) with stronghypermethylation in both regions (R6E) is the onlytumor with decreased p15INK4b expression without LOH.One additional sample (R6C) with p15INK4b inactivation

Figure 6 Patterns of methylation in the p15INK4b promoter or 3'-UTR in di�erent types of induced thymic lymphomas.Hypermethylation in the CpG island or in the 3'-UTR were considered to occur when the level of methylation in the XhoI or HpaIIsites was higher than 30 or 60%, respectively


390

showed neither CpG island methylation nor stronghypermethylation in the 3'-UTR; however, methylationin the 3'-UTR was increased slightly. Taken together,these results indicate that both methylations in the 5'or 3' regions can be alternative and cooperatingmechanisms for p15INK4b gene inactivation.

Methylation of the 3'-UTR produces decreased geneexpression in vitro

In order to analyse the e�ect of the 3'-UTRmethylation on gene expression, two luciferasereporter plasmids were used in which the luc gene is

driven by the thymidine kinase promoter without(ptkLUC+) or with the CMV enhancer(pCMVtkLUC+). Di�erent fragments of the p15INK4b

3'-UTR (Figure 8a) were ampli®ed by PCR asdescribed in Materials and methods and CpGmethylated in vitro using the SssI methylase. CpGmethylation was demonstrated by the lack of HpaIIdigestion after treating with SssI. Methylated and non-methylated fragments were inserted between the luccoding region and the SV40 late polyadenylation signal(pA). These recombinant constructs, without anybacterial passage, were used to transfect NIH3T3cells. Luciferase activities were measured and normal-ized to b-galactosidase activities, used as an internalcontrol. The amount of DNA and its identity wereassesed by transformation of bacterial cells with analiquot of the recombinant molecules. As shown inFigure 8B, methylation of the whole p15INK4b 3'-UTR(fragment 1) produces decreased luciferase activity inboth constructs containing the tk (Ptk) or the CMVtk(PCMV) promoters. This decrease in the luciferaseactivity is specially high in the CMV plasmid, whereonly 30% of the activity remains after methylation ofthe 3'-UTR.Since hypermethylation of the 3'-UTR in tumors

ranges from 60 ± 100%, in vitro methylation offragment 1 was performed at increasing times toobtain di�erent levels of methylation. Thus, 4 min ofincubation with SssI produced about 50% ofmethylated molecules whereas the methylation isalmost complete in 15 min (90%) and totally completein 60 min. When these partially methylated moleculeswere analysed for luciferase activity, a linear e�ect wasobserved between the level of methylation and lightunits produced (Figure 8c).Only ten CpG sites are present in the p15INK4b 3'-

UTR, a region with a very low percentage of CpGwhen compared with other fragments of the p15INK4b

genomic DNA (Malumbres et al., 1997). In order tofurther analyse the CpG sites responsible for this e�ect,several aditional fragments of the p15INK4b 3'-UTR wereampli®ed. Thus, fragments 1, 2, 3, 4 and 5, whichcontain ten, eight, six, four or two CpG sites (Figure8a) were analysed. In vitro methylation of all thesefragments showed a similar e�ect on luciferase activity(Figure 8c). The decrease of luciferase activity is similarusing fragment 1, containing all the 10 CpG sites, andfragment 5, a 120-bp DNA fragment which containsthe CpG sites present in the sequence recognized by theXhoI and HpaII restriction enzymes.

Discussion

The p53- and Rb-pathways are among the mostfrequently disrupted in cancer cells. In some type oftumors, as lung or esophageal carcinomas, almost100% of cases have lesions in either p16INK4a, cyclin D1or Rb, being the relative frequencies p16INK4a 4 CyclinD14Rb (Sherr, 1996). For instance, deletion ofp16INK4a (and, in most of cases, p15INK4b) has beenobserved in 67% of non-small cell lung carcinomas and68% glioblastomas (Hirama and Koe�er, 1995). Inhematological malignancies, selective absence ofp15INK4b is found, virtually always due to hypermethyla-tion of the CpG island in the promoter-exon 1 region

Figure 7 De novo methylation of p15INK4b in murine thymiclymphomas induced by g-irradiation (a), neutron-irradiation (b)or MNU treatment (c). (a) g-irradiation-induced tumor DNA wasdigested with HpaII6EcoRI and hybridized with the 3'-UTRprobe. 1, normal thymus; 2 ± 9, tumor DNAs. Samples 2 and 3and 5 ± 9 show increased methylation in the 3'-UTR. Thismethylation is almost 100% in tumors 2, 3, 5 and 6. RNA fromthese tumors was used for RT±PCR analysis, showing decreasedp15INK4b expression in the tumors with 3'-UTR hypermethylation.(b) Thymic lymphomas induced by neutron-radiation. This panelshows the di�erent patterns observed in the XhoI6EcoRIdigestions (see Figure 6). 1, normal thymus DNA digested withEcoRI; 2, normal thymus DNA digested with XhoI x EcoRI; 3 ±11, tumor DNA digested with XhoI6EcoRI. Samples 3, 4, 6, 7,10 and 11 show hypermethylation in the 3'-UTR (6.0-kb and),reaching almost 100% in samples 7 and 9. Sample 4 shows partialhypermethylation in the CpG island and in the 3'-UTR, givingrise to the 6.0-, 10.5- and 12.0-kb bands. Sample 5 has stronghypermethylation in the promoter region (10.5-kb band) but notin the 3'-UTR. The presence of the 12.0-kb band in sample 9indicates however hypermethylation in both 5' and 3' regions. (c)Tumors induced with MNU treatment. All the samplescorrespond to tumor DNA digested with XhoI6EcoRI. Samples1, 2, 3, 4 and 10 show increased methylation of the 3'-UTR. Nohypermethylation of the CpG island is found in these tumors


391

rather than homozygous deletion (Herman et al., 1996,1997; Batova et al., 1997). Up to 88% of adult acutemyelogenous leukemias or acute lymphocytic leukemiashave speci®c methylation of the p15INK4b CpG island,whereas there is neither methylation in p16INK4a norhomozygous deletions in any of these INK4 genes.Similarly, hypermethylation of the murine p15INK4b

promoter CpG island occur in 88% of thymiclymphomas, whereas only 36% of these tumors showhypermethylation in p16INK4a (Malumbres et al., 1997).However, in the analysis of murine thymic lymphomas,we have detected some additional tumors in which theabsence of p15INK4b cannot be explained by promotermethylation. This fact led us to look for additionalmechanisms of p15INK4b inactivation.The presence of unexpected DNA bands which

hybridized with a p15INK4b-speci®c probe allowed us todetect methylation in the XhoI and HpaII sites of thep15INK4b 3'-UTR in normal tissues (Figures 1 and 2).This area does not present the typical features for theCpG islands and it only contains a few CpG sites(Malumbres et al., 1997). The detected methylation isrestricted to the 3'-UTR and does not a�ect the HpaIIsite in the translated region of exon 2. The level ofmethylation varies depending on the tissue. Brain andgut show high methylation in the p15INK4b 3'-UTR,whereas the lower levels of methylation are found inthe hematopoietic tissues, thymus and spleen (Figure3), where it seems that p15INK4b plays an important

regulatory role in controlling growth (Lois et al., 1995;Herman et al., 1997; Malumbres et al., 1997).Heritable, tissue-speci®c methylation patterns have

been proposed to form the basis of a stable regulatorysystem (Riggs, 1975; Laird and Jaenisch, 1996). Methylgroup addition to DNA is proven to be necessary fornormal development, apparently because it helps toshut down genes. Since both cell cycle inhibitorsp16INK4a and p15INK4b are not expressed in embryogen-esis (Zindy et al., 1997), methylation of CpG sites otherthan the promoter CpG islands could play a role in thesilencing of these genes in the embryonic development.Thus, most of tissue-speci®c genes are fully methylatedin the germinal cells and become partially unmethy-lated in speci®c tissues (Razin and Cedar, 1991). In ourresults, the 3'-UTR of p15INK4b is highly methylated intestis and di�erentially methylated in other tissues. Nomethylation is ever found in the CpG island of thepromoter region in normal tissues.It has been proposed that the methylation of the

coding regions, introns or untranslated regions isrelatively unimportant in comparison to methylationof the promoter regions, and there may be two levels ofcontrol operating independently. Methylation of thepromoter region could be the primary switchingmechanism; methylation of the other regions couldprovide only ®ne tuning (Riggs and Jones, 1983). Inthe p15INK4b gene, methylation of the 5' promoter regionproduces gene inactivation with independence of the

Table 1 p15INK4b expression and methylation of the 5' and 3' regions in g-radiation-induced thymic lymphomas in B6RFF1 mice

Sample LOHa CpG island methylationb 3'UTR hypermethylation p15INK4b expression

R1BR1CR1DR1ER2BR2CR2DR3AR3BR4AR4BR4CR4DR4ER5BR5CR6BR6CR6ER7AR7BR8AR8DR8ER9AR9BR9CR10AR12AR13AR13BR13CR13ER15B

+7+777+777+7++7+++7++7+7+77+++++7+

77+777777777777++++7++7+7+++777++77+++++

++77777++777++7+++777+++++777++77+7++7777

7+7+++7+++7+++NAd

7777+7+7+7+++777e

777

a + or 7 indicate the presence or absence of LOH. b The level of methylation was quanti®ed as the percentage of methylatedband: 7, no increased methylation (0 ± 30%); +, weak methylation (30 ± 60%); ++, strong methylation (60 ± 100%). c + or 7indicate normal or decreased expression of p15INK4b, respectively. d NA, This tumor could not be analysed for p15INK4b expressiondue to lack of RNA sample. e This tumor showed a small deletion a�ecting the exon 1 of p15INK4b


392

methylation status in the other areas, as can beobserved in Tables 1 and 2. Strong hypermethylationof the promoter always produces gene inactivation withindependence of the 3'-UTR methylation status(correlation highly signi®cant, P50.001). Methylationof the 5' CpG island correlates also with LOH in thep15INK4b locus (Table 2).Although some correlation between expression and

methylation in the non-promoter regions has beenfound (Battistizzi et al., 1985), the lack of correlation isa frequent ®nding in tissue-speci®c genes (Mullins etal., 1987; Borrello et al., 1992). In our results,methylation of the p15INK4b 3'-UTR shows no correla-

tion with tissue-speci®c expression or physiologicalinactivation of this gene in lymphocytes (Figures 3 and4). Thus, DNA methylation in the XhoI or HpaII sitesof the 3'-UTR does not seem to have a major role inthe control of p15INK4b gene expression in several tissuesand in physiological conditions. This fact does notmean that methylation in the 3'-UTR cannot play arole in p15INK4b expression but probably there are otherfactors more important to determine the level of geneexpression. However, we show a strong hypermethyla-tion of the 3'-UTR associated with decreased expres-sion in thymic lymphomas. This correlation is highlysigni®cant (P50.001) in the samples which do not havemethylation of the promoter area (Table 2). Thesedi�erent results may be due to the fact thatmethylation in the p15INK4b 3'-UTR may have asigni®cant role on p15INK4b transcription in a speci®ccell type when a threshold is reached in pathologicalconditions (neoplasia). For instance, high levels ofmethylation could be e�ective in inhibiting transcrip-tion in lymphocytes, since these normal cells have lowmethylation, but could have no e�ect in other celltypes. As can be observed by comparing Figures 3 and4, 3'-UTR methylation is even lower (almost undetect-able) in the isolated lymphocytes than in the wholethymus. Thus, a lower methylation threshold may beneeded in thymus than in other tissues, as gut or testis,to be e�ective in down-regulating p15INK4b expression.Although promoter methylation of p15INK4b is

frequent in radiation-induced lymphomas, no hyper-methylation of the CpG island has been found inMNU-induced tumors (Malumbres et al., 1997).However, hypermethylation in the 3'-UTR is relativelyfrequent in these samples. These di�erences are likelyto be due to the di�erent mutagenic mechanisms ofMNU and ionizing radiation. It is worth noting thathypermethylation of the 3'-UTR cannot be explainedas a simple extension of the hypermethylation in theCpG island located in the promoter region. There aremainly two reasons. First, hypermethylation of the 3'-UTR does not a�ect the HpaII site located in thecoding region of exon 2, between the promoter and the

Figure 8 E�ect of 3'-UTR CpG methylation on luciferaseactivity. (a) Scheme of the p15INK4b 3'-UTR showing the CpGsites. Fragments 1 (700 bp), 2 (360 bp), 3 (250 bp), 4 (240 bp) and5 (120 bp) were ampli®ed by PCR, methylated and analysed fortranscription interference. (b) Luciferase activities of the fragment1 constructs carrying the non-methylated (white box) ormethylated 3'-UTR (white box with rounded ¯ags). Ptk,thymidine kinase promoter; PCMV, thymidine kinase promoterand CMV enhancer; luc, luciferase gene; pA, SV40 latepolyadenylation signal. The chart represents the average of threeseparate experiments. Light units corresponding to the non-methylated forms were arbitrary set as 1. The actual valuesobtained with the constructs containing the CMV enhancer(PCMV) were 50 ± 80 times higher than those obtained with thethymidine kinase promoter (Ptk). All values were normalized withrespect to the b-galactosidase activity. (c) Methylation offragments 1 ± 5 (as described in a) and luciferase activity.Fragment 1 was methylated for 0, 4, 15 or 60 min, whereas allthe other fragments were methylated for 0 or 60 min. The level ofmethylation is shown by digestion with HpaII. About 50% or90% of the molecules are methylated with 4 or 15 min. ofincubation, whereas the reaction is always complete in 60 min.The relative luciferase activity of these constructs is shown belowwhere all the non-methylated forms are arbitrary set to 1

Table 2 Correlation between LOH, p15INK4b expression andhypermethylationa

CpG hyper- 3'-UTR hyper-LOH methylation methylation

Decreased p15INK4b

expression

LOH

CpGhypermethylation

0.70P50.001n=48

0.70P50.001n=470.9313b

P50.001n=270.35

P=0.015n=47

0.28P=0.083n=390.71c

P50.001n=270.48

P=0.002n=3970.12

P=0.459n=3970.65a

P=0.006n=16

aThe correlation values and their statistical signi®cance weredetermined using the Pearson test; n indicates the numbers of casesconsidered. bData from samples without 3'-UTR methylation. cDatafrom samples without CpG island methylation. dData from sampleswith decreased p15INK4b expression


393

3'-UTR. Second, although some samples showhypermethylation in the promoter and the 3'-UTR,strong hypermethylation of the 3'-UTR mostly occursalternatively to promoter methylation (e.g. samplesR1B, R2D, R4B, R4D and R13A in Table 1 or MNU-induced tumors). This is specially evident whenconsidering samples with lack of p15INK4b expression(Tables 1 and 2). Thus, the ®ne regulation putativelyprovided by the 3'-UTR could be overcome whenstrong methylation is present in this area. In othercases (R6E), this alteration could cooperate with theCpG island hypermethylation.The partial or complete in vitro methylation of the

p15INK4b 3'-UTR produces, in fact, a decrease of geneexpression as shown in the luciferase system (Figure 8).Plasmids carrying the totally or partially CpG-methylated 3'-UTR show signi®cantly lower luciferaseactivity than the same plasmids without any methyla-tion in the 3'-UTR. These data indicate thatmethylation in this area does interfere with somemachinary involved in the regulation of transcription.Methyl groups may interfere, for instance, with thebinding of individual proteins to DNA, and, in vivo,the variable binding of factors to DNA in di�erenttissues correlates with the methylation pattern in thosetissues (Razin and Cedar, 1991). Transcription factorswhich bind to enhancer or promoter regions areincluded among the proteins whose DNA-bindingability is interfered by methylation (Watt and Molloy,1988; Iguchi-Ariga and Scha�ner, 1989). On the otherhand, methylated DNA could be recognized by speci®cDNA binding proteins as MeCP2, which binds DNAwith a single CpG methylation and have beendemonstrated to inhibit transcription (Meehan et al.,1992; Nan et al., 1997). Thus, the e�ect of methylationof the 3'-UTR (a region without CpG islands) ontranscription should be di�erent to that of methylationof promoter regions, where multiple CpG sites of theCpG islands are involved and the density ofmethylation in¯uences the level of transcription(Baylin et al., 1998). In fact, as observed in Figure 8,methylation of only two of the CpG sites (thoserecognized by the XhoI and HpaII restriction enzymes)is su�cient to interfere with transcription.Since all INK4 genes show tissue-speci®c expression

(Zindy et al., 1997), DNA methylation of non-promoter regions could be present in the othermembers of the CDKI families. Recent studies areuncovering a high genetic complexity of the CDKIsand their mechanisms of regulation and inactivation.In addition to deletions, rearrangements, mutationsand CpG island hypermethylation, these show addi-tional complex features, including posttranscriptional(p21WAF1/CIP1) or post-translational (p27KIP1, p15INK4b)regulation or imprinting (p27KIP1) (Hatada and Mukai1995; Pagano et al., 1995; Hengst and Reed, 1996; Li etal., 1996; Matsuoka et al., 1996; Shandu et al., 1997;Cost et al., 1997; ReynisdoÂ ttir and MassagueÂ 1997).Regarding to untranslated regions, the transcriptionalregulatory element proposed in this work for thep15INK4b 3'-UTR is not unique in the CDKIs, sincep21WAF1/CIP1 has been shown to be not only posttran-scriptionally regulated (Li et al., 1996) but alsocontains cis elements in its 3'-UTR involved in theregulation of transcription (Rishi et al., 1997). Inaddition, the p21WAF1/CIP1 3'-UTR contains three ARE

motifs (5'-AUUUA-3'), involved in RNA instability(Savant-Bhonsale and Cleveland, 1992), although itsrole is not clear in the p21WAF1/CIP1 expression (Li et al.,1996). At least two ARE motifs are present also in themurine p15INK4b 3'-UTR, but their e�ect has not beenstudied yet.Recently, a new alternative splicing form of p15INK4b,

named p10, has been characterized in human (Tsubariet al., 1997). The mouse p10 protein is 69% identical tothe human sequence (MM and AP, unpublishedresults). Strikingly, p10 is also induced by TGF-band is able to restrain cell cycling. Since p10 andp15INK4b presumably use the same promoter (Tsubari etal., 1997), speci®c regulation of each transcript shouldbe provided by additional control regions. Forinstance, we still do not know whether the 3'-UTR ofp15INK4b is present in the p10 transcripts. A role forDNA methylation in the control of alternate tran-scripts expression has been previously suggested (Kayet al., 1997). Thus, UTR sequences could also providedi�erential regulation on both p10 and p15INK4b

transcripts. Further investigation about the 3'-UTRe�ect on p15INK4b and p10 expression in normal andtumor cells is required to clarify these hypotheses.Given the correlation between strong hypermethyla-

tion of the 3'-UTR and decreased p15INK4b expression inthymic lymphomas, we propose this may be a newmechanism for tumor suppressor gene inactivation.This idea is supported by the downregulation of geneexpression observed in vitro when the 3'-UTR ismethylated. Thus, strong methylation of this regioncould change the speci®ty for factors involved in thep15INK4b transcriptional regulation in only some tissues.For instance, methylation of non-promoter regions hasbeen recently associated to tissue-speci®c repression inthe keratin 18 gene (Umezawa et al., 1997). This andother putative functions of the p15INK4b 3'-UTR shouldbe studied in order to understand the complexregulation of this cell cycle inhibitor.

Materials and methods

Mice and induction of tumors

Several inbred mouse strains: C57BL/6J, DBA/2, RF/J,129/J, AK-Fr-1b and F1 hybrids: C57BL/6JxRF/J(B6RFF1), C57BL/6JxDBA/2 (B6D2F1 and D2B6F1),RF/JxAKR, AKRxLR1, AKRxNZB) were used in thiswork. Thymic lymphomas were induced using methylni-trosourea (MNU), g-radiation or neutron radiation, asdescribed previously (Santos et al., 1996; Malumbres et al.,1997). For tissue samples, mice were sacri®ced by cervicaldislocation and di�erent portion of the tissues were takenfor histological, DNA, or RNA analysis.

DNA manipulation, sequencing and hybridizations

Genomic DNA isolation and basic manipulation wasperformed using standard procedures. To construct genelibraries, mouse DNA was digested with XhoI and EcoRIand selected fragments were subcloned in pBLUESCRIPT(Stratagene) digested with SalI and EcoRI or with SalIalone. The ligations were used to transform Escherichia coliand the speci®c clones were selected by plate hybridization,using a probe speci®c for the exon 1 of the murine p15INK4b

(Malumbres et al., 1997). An additional probe speci®c forthe 3'-UTR was ampli®ed using primers Mp15-3'-1F and


394

Mp15-3'-3R. The position of these probes is shown inFigure 1. Sequencing of DNA was carried out with anautomatic 373 DNA Sequencer (Applied Biosystems). Forhybridizations, DNA was digested, separated on agarosegels, and transferred to nitrocellulose membranes (Schlei-cher & Schuell). Completion of enzyme digests was veri®edby hybridization with other gene probes, as the murinep16INK4a (exon 1), cyclin D1, or glyceraldehyde-3-phosphatedehydrogenase genes. Additional controls are described inthe ®gure legends. DNA probes were labeled with [a-32P]-dCTP (3000 Ci/mmol; Dupont-NEN) using a Randomprimed labeling kit (Boehringer Mannheim), according tothe manufacturer's protocol. Hybridizations were visua-lized and quanti®ed by a PhosphorImager (MolecularDynamics) or by exposing X-ray ®lms (Kodak), digitalscanning and analysis with the NIH Image software.

The sequences of the p15INK4b 3'-UTR and downstreamregions were submitted to the GenBank database and wereassigned accession numbers AF016457 (3'-UTR from DBA/2mice) and AF015460 (exon 2, 3'-UTR and adjacent regionsfrom C57BL/6J mice).

DNA ampli®cation, RNA isolation and RT±PCR

Total cellular RNA from tissues or cultured cells wasisolated using the guanidine thiocyanate method. Poly-(A)mRNA was isolated following the procedure described inthe mRNA isolation kit (Quiagen), and DNA contamina-tion was removed by digestion with DNaseI. cDNA wasgenerated by reverse transcription of 5 mg of total RNA or100 ng poly-(A) RNA using the MMLV reverse transcrip-tase (Life Technologies, Inc.) and a hexanucleotide mixtureas a primer (Boehringer Mannheim). Standard PCR wasperformed in a Perkin Elmer Cetus DNA thermal cyclerand consisted of 2 min at 958C, 20 ± 40 cycles of 30 s at958C, 30 s at 588C and 1 min at 728C and a ®nal step of5 min at 728C. Long-template PCR was performed usingthe Expand PCR kit (Boehringer Mannheim) and themanufacturer recommendations. Ampli®ed DNA was runin agarose gels and detected by ethidium bromide stainingor by blotting and subsequent hybridization with probesspeci®c for the murine p15INK4b gene. The following primerswere used in this work for PCR ampli®cation: Mp15E1F:5'-TGC CAC AGA CCG GGG ACA AGG-3'; Mp15F: 5'-ATG TTG GGC GGC AGC AGT GA-3'; Mp15E1F3: 5'-TGA ACC GCT TCG GGA GGC GCC CAA T-3';Mp15R: 5'-GTC AAT CTC CAG TGG CAG CG-3',Mp15-3'-1F: 5'-AGG TAT CTG CAC GCT GCC AGTGGA GAT-3'; Mp15-3'-1R: 5'-TAA CCA TGG AGA TCTCTC CAG GCT CCA-3'; Mp15-3'-3R: 5'-TTT GCA GGTGAA TCC CCA CAC ATG-3'; and Mp15-3'-4R: 5'-TAGAGG GCC CGG GAA CTT CAT AC-3'. When needed,PCR products were digested with XhoI by adding 10 unitsof the enzyme and the corresponding digestion bu�er overthe PCR products and incubating at least 3 h at 378C.Quantitative RT ± PCR was performed using templatedilutions or di�erent number of ampli®cation cycles andcomparing with the b-actin expression. b-actin cDNA wasampli®ed using primers b-actin-1F: 5'-GTG GGC CGCTCT AGG CAC CAA-3' and b-actin-1R: 5'-CTC TTTGAT GTC ACG CAC GAT TTC-3'.

Lymphocyte activation

Two 70-days old DBA/2 mouse were sacri®ced and spleencells were cultured in RPMI 1640 medium for 2 h.Lymphocytes were then activated with 5 mg/ml anti-CD3eantibodies (Pharmigen), 100 ng/ml phorbol myristateacetate (PMA; Sigma), 0.5 mg/ml ionomycin (Sigma) and10 mg/ml concanavalin A (Sigma). Samples were taken at20 h and 50 h after activation for DNA and RNA analysis.Cells at time of activation (time 0) and non-activated cellsat 50 h of culture were used as controls.

Luciferase assays

The p15INK4b 3'-UTR was ampli®ed by PCR using primersMp15UTR-F: 5'-CTA GGC CGG CCT CGA GCA AGGACT TCT TTC TCC C-3' and Mp15UTR-R: 5'-GCG GGCCGG CCT CGA GTT TTT CTT TAA ATA CAC TG-3'(fragment 1, Figure 8a); Mp15UTR-F and Mp15UTR2R:5'-TAT GGC CGG CCG TCG ACA GGG GAA GGTACT GAC-3' (fragment 2); Mp15UTR-F and Mp15UTR-3R: 5'-TAT GGC CGG CCT GAA AGG TAG AGG-3'(fragment 3); Mp15UTR-2F: 5'-ATT GGC CGG CCGCTG GAT CTG GTC-3' and Mp15UTR-2R (fragment 4);and Mp15UTR-2F and Mp15UTR-3R (fragment 5). Allthese primers contain an arti®cial FseI site. PCR productwas digested with FseI (New England BioLabs) andpuri®ed. Luciferase reporter plasmids ptkLUC+andpCMVtkLUC+(Altschmied and Duschl, 1997) were di-gested with FseI, treated with alkaline phosphatase(Boehringer Mannheim) and puri®ed. FseI-digested 3'-UTR was CpG methylated using the SssI methylase (NewEngland BioLabs) and puri®ed. Methylated or non-methylated 3'-UTR fragments were ligated to the luciferaseplasmids for 16 h at 148C. Ligation products were separatedin 1% agarose gel and the bands corresponding to theplasmids containing the insert were isolated. The identity ofthese bands was assesed by transformation of bacteria cellsand restriction analysis of the plasmids. About 100 ng ofpuri®ed recombinant molecules and 1 mg of the b-galactosidase-expressing plasmid pCH110 (Pharmacia)were used to cotransfect NIH3T3 cells. Transient transfec-tions were performed using the calcium phosphate method.Thirty-six hours after transfection, cells were collected andlysed using the bu�er provided in the Luciferase AssaySystem (Promega). Luciferase was assayed according themanufacturer recommendations. The same protein extractswere used then to measure the cheminoluminiscenceproduced by b-galactosidase using the Galacto-Light Plussystem (Tropix).

AcknowledgementsWe thank Dr Altschmied for providing the luciferasereporter vectors. This work was supported by NIH grantsCA 36327 and CA 50434 to AP and grants from theFundacioÂ n RamoÂ n Areces and DGES (PM96-001) to JF-P.MM received fellowships from the FundacioÂ n RamoÂ nAreces and the MEC (Madrid, Spain). IPC was afellowship of the Comunidad AutoÂ noma de Madrid(Spain).

References

Altschmied J and Duschl J. (1997). BioTechniques, 23, 436 ±438.

Batova A, Diccianni MB, Yu JC, Nobori T, Link MP, PullenJ and Yu AL. (1997). Cancer Res., 57, 832 ± 836.

Battistuzzi G, D'Urso M, Toniolo D, Persico GM andLuzzatto L. (1985). Proc. Natl. Acad. Sci. USA, 82, 1465 ±1469.

Baylin SB, Herman JG, Gra� JR, Vertino PM and Issa JP.(1998). Adv. Cancer Res., 72, 141 ± 196.

Borrello MG, Pierotti MA, Tamborini E, Biassoni D,Rizzetti MG, Pilotti S and Della Porta G. (1992).Oncogene, 7, 269 ± 275.


395

Cairns P, Polascik TJ, Eby Y, Tokino K, Califano J, MerloA, Mao L, Herath J, Jenkins R, Westra W, Rutter JL,Buckler A, Gabrielson E, Tockman M, Cho KR, HedrickL, Bova GS, Isaacs W, Koch W, Schwab D and SidranskyD. (1995). Nat. Genet., 11, 210 ± 212.

Cost GJ, Thompson JS, Reichard BA, Lee JY and FeinergAP. (1997). Cancer Res., 57, 926 ± 929.

Glendeling JM, Flores JF, Walker GJ, Stone S, Albino APand Fountain JW. (1995). Cancer Res., 55, 5531 ± 5535.

GranÄ a X and Reddy EP. (1995). Oncogene, 11, 211 ± 219.Hannon GJ and Beach D. (1994). Nature, 371, 257 ± 260.Harper JW and Elledge SJ. (1996). Curr. Opin. Genet. Dev.,6, 56 ± 64.

Hatada I and Mukai T. (1995). Nat. Genet., 11, 204 ± 206.Hengst L and Reed S. (1996). Science, 271, 1861 ± 1864.Herman JG, Jen J, Merlo A and Baylin SB. (1996). Cancer

Res., 56, 722 ± 727.Herman JG, Civin CI, Issa JPJ, Collector MI, Sharkis SJ andBaylin SB. (1997). Cancer Res., 57, 837 ± 841.

Hirama T and Koe�er HP. (1995). Blood, 86, 841 ± 854.Iavarone A and MassagueÂ J. (1997). Nature, 387, 417 ± 422.Iguchi-Ariga SMM and Scha�ner W. (1989). Genes Dev., 3,612 ± 619.

Jen J, Harper JW, Bigner SH, Bigner DD, Papadopoulos N,Markowitz S, Willson JKV, Kinzler KW and VogelsteinB. (1994). Cancer Res., 54, 6353 ± 6358.

Kamb A, Gruis NA, Weaver-Feldhaus J, Liu Q, HarshmanK, Tavtigian SV, Stockert E, Day III RS, Johnson BE andSkolnick MH. (1994). Science, 264, 436 ± 440.

Kay PH, Harmon D, Fletcher S, Ziman M, Jacobsen PF andPapadimitriou JM. (1997). Gene, 184, 45 ± 53.

Laird PW and Jaenisch RJ. (1996). Ann. Rev. Genet., 30,441 ± 464.

Li JM, Nichols MA, Chandrasekharan S, Xiong Y andWangXF. (1995). J. Biol. Chem., 270, 26750 ± 26753.

Li XS, Rishi AK, Shao ZM, Dawson MI, Jong L, Shroot B,Reichert U, Ordonez J and Fontana JA. (1996). CancerRes., 56, 5055 ± 5062.

Lois AF, Cooper LT, Geng Y, Nobori T and Carson D.(1995). Cancer Res., 55, 4010 ± 4013.

Malumbres M, PeÂ rez de Castro I, Santos J, MeleÂ ndez B,Mangues R, Serrano M, Pellicer A and FernaÂ ndez-Piqueras J. (1997). Oncogene, 14, 1361 ± 1370.

Malumbres M, PeÂ rez de Castro I, Santos J, PeÂ rez-OlleÂ R,FernaÂ ndez-Piqueras J and Pellicer A. (1998). Mamm.Genome, 9, 183 ± 185.

Matsuoka S, Thompson JS, Edwards MC, Barletta JM,Grundy P, Kalikin LM, Harper JW, Elledge SJ andFeinberg AP. (1996). Proc. Natl. Acad. Sci. USA, 93,3026 ± 3030.

Meehan RR, Lewis JD and Bird AP. (1992). Nucleic AcidsRes., 20, 5085 ± 5092.

Merlo A, Herman JG, Mao L, Lee DJ, Gabrielson E, BurgerPC, Baylin SB and Sidransky D. (1995). Nat. Med., 1,686 ± 692.

Mullins LJ, Veres G, Caskey CT and Chapman V. (1987).Mol. Cell. Biol., 7, 3916 ± 3922.

Nan X, Campoy FJ and Bird A. (1997). Cell, 88, 471 ± 481.Pagano M, Tam SW, Theodoras AM, Beer-Romero P, DelSal G, Chau V, Yew PR, Draetta GF and Rolfe M. (1995).Science, 269, 682 ± 685.

Peters G. (1994). Nature, 371, 204 ± 205.Pollock PM, Pearson JV and Hayward NK. (1996). Genes

Chromosom. Cancer, 15, 77 ± 88.Quelle DE, Ashmun RA, Hannon GJ, Rehberger PA, TronoD, Richter H, Walker C, Beach D, Sherr CJ and SerranoM. (1995). Oncogene, 11, 635 ± 645.

Rasool O, Heyman M, Brandter LB, Liu Y, Grander D,Soderhall S and Einborn S. (1995). Blood, 85, 3431 ± 3436.

Razin A and Cedar H. (1991).Microbiol. Rev., 55, 451 ± 458.ReynisdoÂ ttir I and MassagueÂ J. (1997). Genes Dev., 11, 492 ±503.

ReynisdoÂ ttir I, Polyak K, Iavarone A and MassagueÂ J.(1995). Genes Dev., 9, 1831 ± 1845.

Riggs AD. (1975). Cytogenet. Cell Genet., 14, 9 ± 25.Riggs AD and Jones PA. (1983). Adv. Cancer Res., 40, 1 ± 30.Rishi AK, Hsu CKA, Li XS, Hussain A, Gerald TM andFontana JA. (1997). Cancer Res., 57, 5129 ± 5136.

Santos J, PeÂ rez de Castro I, Herranz M, Pellicer A andFernaÂ ndez-Piqueras J. (1996). Oncogene, 12, 669 ± 676.

Savant-Bhonsale S and Cleveland DW. (1992). Genes Dev., 6,1927 ± 1939.

Shandu C, Garbe J, Bhattacharya N, Daksis J, Pan CH,Yaswen P, Koh J, Slingerland JM and Stampfer MR.(1997). Mol. Cell. Biol., 17, 2458 ± 2467.

Sherr CJ. (1996). Science, 274, 1672 ± 1677.Sherr CJ and Robert JM. (1995). Genes Dev., 9, 1149 ± 1163.Tsubari M, Tiihonen E and Laiho M. (1997). Cancer Res.,57, 2966 ± 2973.

Umezawa A, Yamamoto H, Rhodes K, Klemsz MJ, MakiRA and Oshima RG. (1997). Mol. Cell. Biol., 17, 4885 ±4894.

Watt F and Molloy PL. (1988). Genes Dev., 2, 1136 ± 1143.Zindy F, Quelle DE, Roussel MF and Sherr CJ. (1997).

Oncogene, 15, 203 ± 211.


396

hypermethylation of the cell cycle inhibitor p15ink4b 3′-untranslated region interferes with its...

Documents