Absence of p16INK4a and truncation of ARF tumor suppressors in chickens

Soo-Hyun Kim*, Michael Mitchell†, Hideta Fujii‡, Susana Llanos*§, and Gordon Peters*¶

*Molecular Oncology, †Computational Genome Analysis, and ‡Developmental Laboratories, Cancer Research U.K., London Research Institute, Lincoln’s Inn Fields, London WC2A 3PX, United Kingdom

Edited by Peter K. Vogt, The Scripps Research Institute, La Jolla, CA, and approved November 14, 2002 (received for review September 13, 2002) The INK4b-ARF-INK4a locus on human chromosome 9p21 (Human Whereas INK4a operates upstream of pRb (3, 4), the ARF Genome Organization designation CDKN2B-CDKN2A), and the cor- protein functions upstream of p53 by binding directly to MDM2 responding locus on mouse chromosome 4, encodes three distinct and protecting p53 from MDM2-mediated degradation (1, 2). products: two members of the INK4 cyclin-dependent kinase in- Current thinking is that the INK4a͞ARF locus plays a key role hibitor family and a completely unrelated protein, ARF, whose in cellular defenses against hyperproliferative signals and stress. carboxyl-terminal half is specified by the second exon of INK4a but For example, INK4a accumulates in human diploid fibroblasts in an alternative reading frame. As INK4 proteins block the phos- (HDFs) that undergo replicative senescence, either as a conse- phorylation of the retinoblastoma product and ARF protects quence of telomere attrition or in response to oncogenic Ras p53 from degradation, the locus plays a key role in tumor suppres- (9–11). Similarly, ARF accumulates as mouse embryo fibro- sion and the control of cell proliferation. To gain further insights blasts (MEFs) approach their replicative limits and in response into the relative importance of INK4a and ARF in different settings, to a variety of oncogenes (12–14). However, there are clear we have isolated and characterized the equivalent locus in chick- differences in the relative importance of INK4a and ARF in cells ens. Surprisingly, although we identified orthologues of INK4b and from different lineages or species (14–17) and in the way they are ARF, chickens do not encode an equivalent of INK4a. Moreover, the regulated. For example, Ras induces ARF in MEFs but not in reading frame for chicken ARF does not extend into exon 2, HDFs (15, 18, 19), whereas pRb represses INK4a in HDFs but because splicing occurs in a different register to that used in not in MEFs (10, 20). There have also been suggestions that the mammals. The resultant 60-aa product nevertheless shares func- sequences encoded by exon 2 make different contributions to the tional attributes with its mammalian counterparts. As well as intracellular localization and function of ARF in the two species indicating that the locus has been subject to dynamic evolutionary (21, 22). pressures, these unexpected findings suggest that in chickens, the To gain further insight into these questions, we sought to tumor-suppressor functions of INK4a have been compensated for isolate the equivalent locus from chicken, both because it by other . represents an intermediate between fish and man in evolutionary terms and because of the relative resistance of chicken cells to ͞ immortalization in tissue culture, similar to HDFs. After char- he INK4a ARF locus has an important role in the control of acterizing 18 kb of genomic DNA and two groups of cDNA Tcell proliferation and in tumor suppression (1, 2) and is clones from late-passage chicken embryo fibroblasts (CEFs), we incapacitated in a variety of familial and sporadic cancers (3, 4). conclude that the chicken INK4b-ARF-INK4a locus is able to INK4a The INK4a product, p16 , functions as an inhibitor of encode an equivalent of p15INK4b and a truncated yet functional cyclin-dependent kinases (Cdks) 4 and 6 (5), hence the official version of ARF specified only by exon 1␤. Surprisingly, a partial designation CDKN2A. These Cdks, along with D-type cyclins, duplication of exon 1␤ has replaced exon 1␣, and we find no regulate the phosphorylation of the retinoblastoma protein evidence that chicken cells contain a p16INK4a orthologue. (pRb) in the late G1 phase of the cell cycle (6). Interestingly, p16INK4a is the prototype of a family of INK4 proteins, each Materials and Methods comprising between 3 and 5 ankyrin repeats (3, 4), orthologues Cells. Primary CEFs were grown at 37°C in DMEM supple- of which have been identified in a variety of mammals as well as mented with heat-treated 10% (vol͞vol) FCS and 2% chicken in Fugu and Xiphophorus fish (7, 8). All members of the gene serum (GIBCO͞BRL). The U20S human osteosarcoma family isolated thus far show the same exon 1–exon 2 splice cell line and the NARF-2 derivative line in which human junction and, in mammals, INK4b and INK4a occur in a con- p14ARF is expressed from an isopropyl-␤-D-thiogalactoside served tandem arrangement on human chromosome 9p21 and (IPTG)-regulated promoter were cultured as described (23). syntenic regions on mouse chromosome 4 and rat chromosome Cells were transiently transfected by calcium phosphate precip- 5, suggesting that they evolved by a gene-duplication event (3). itation and harvested after 48 h. Retroviral infection of the TIG3 Because there is no evidence for such a duplication in Fugu (8), strain of HDFs expressing the ecotropic virus receptor was as it appears that INK4b was the primordial gene at this locus. described (24). However, the INK4a locus has the highly unusual capacity to encode two structurally and functionally different proteins. Two Bacterial Artificial Chromosome (BAC) and cDNA Libraries. The transcripts, designated ␣ and ␤, are produced; they initiate at chicken BAC library was obtained from the UK separate promoters and incorporate different first exons (1␣ and

1␤) spliced to a common second exon (1, 2). Whereas the MEDICAL SCIENCES ␣-transcript specifies p16INK4a, the ␤-transcript encodes p14ARF This paper was submitted directly (Track II) to the PNAS office. (p19ARF in mouse), so-called because the second exon is trans- Abbreviations: HDF, human diploid fibroblast; CEF, chicken embryo fibroblast; BAC, bac- Ϫ terial artificial chromosome; MTAP, methylthioadenosine phosphorylase; Cdk, cyclin- lated in the 1 (alternative) reading frame to that used to dependent kinase; HA, hemagglutinin. INK4a ␤ generate p16 . As exon 1 is poorly conserved and has no Data deposition: The sequences reported in this paper have been deposited in the GenBank obvious relatives in the current databases, its evolutionary database (accession nos. AY138245–AY138247). origins remain unknown. For example, did exon 1␤ originally §Present address: Ludwig Institute for Cancer Research, St. Mary’s Hospital, London W2 belong to a different gene and move to the INK4a͞b locus at 1PG, United Kingdom. some point after the INK4a͞b duplication? ¶To whom correspondence should be addressed. E-mail: [email protected].

www.pnas.org͞cgi͞doi͞10.1073͞pnas.0135557100 PNAS ͉ January 7, 2003 ͉ vol. 100 ͉ no. 1 ͉ 211–216 Downloaded by guest on October 1, 2021 Mapping Project Resource Center. The cDNA library was constructed from CEFs at Ϸ50 population doublings by using the ␭ZAP-II cDNA synthesis kit according to the manufacturer’s protocols (Stratagene). After screening of 1.35 ϫ 106 recombi- nant phages (unamplified), 170 plaques showed positive signal, of which 50 were purified and analyzed further.

DNA and RNA Analyses. CEF genomic DNA was isolated by using the Easy-DNA kit (Invitrogen) and analyzed by Southern hy- bridization performed at either 60°C (normal stringency) or 55°C (low stringency) following standard procedures (25). Total and polyadenylated RNA was prepared from late passage CEFs by using the RNeasy kit (Qiagen, Chatsworth, CA) and the Poly(A) Pure kit (Ambion, Austin, TX), respectively. RNA blots were hybridized at 42°C in ULTRAhyb buffer (Ambion) and washed according to the manufacturer’s protocol. The PCR primer sequences used to generate the probes described in Fig. 2A and the conditions for the touch-down RT-PCR of methyl- Fig. 1. Organization of the INK4b͞ARF͞INK4a locus in different species. (A) thioadenosine phosphorylase gene (MTAP) are detailed in Schematic representations (not to scale) of the INK4b, ARF, INK4a, and MTAP Supporting Text, which is published as supporting information on genes in Fugu, humans, and chickens. INK4b exons are shown as black, ARF as the PNAS web site, www.pnas.org. cross-hatched, INK4a as stippled, and MTAP as empty boxes. The arrows indicate the direction of transcription. Dotted lines denote the splicing for the Plasmid Construction. The longest chicken ARF cDNA was used ARF transcript. (B) Restriction map of the chicken INK4͞ARF locus. Exons are as a template for PCR by using primers that introduced a BglII depicted as in A. All XbaI (X), HindIII (H), BamHI (B), SfiI (Sf), BglII (Bg), and PstI site at the presumed 5Ј end of chicken ARF (excluding the ATG) (P) sites are indicated. The locations of critical NotI and XhoI fragments and an XbaI site downstream of potential termination codons in described in the text are shown below the map. The two sets of differently all three reading frames of exon 2 (nucleotides 150–816 in the shaded boxes above the map denote regions of the genomic sequence that are directly duplicated (see Fig. 6A). sequence deposited under accession no. AY138245). The result- ant 683-bp product was cloned into the pEGFP-C3 vector (CLONTECH) to create an in-frame fusion protein with GFP Ch precipitation, we used SMP14 for MDM2, Y-11 for HA-tagged at its N terminus, designated GFP-ARF . A second plasmid proteins (sc-805, Santa Cruz Biotechnology), and H-22 for Cdk4 was generated comprising nucleotides 147–326 of chicken Ј Ј (sc-601, Santa Cruz Biotechnology). Methods used to visualize ARF cDNA flanked by a 5 BglII site and a 3 KpnI site. The human ARF and the various GFP fusion proteins by direct and 196-bp PCR product was cloned into the pEGFP-N1 vector indirect immunofluorescence, and to assay p53 stabilization by (CLONTECH) in which the initiation codon for the GFP has ARF, were described (26). been eliminated, as described (26). This procedure created a fusion protein, designated Ex1␤Ch-GFP, in which the 60 amino Results ␤ acids encoded by chicken exon 1 were fused in-frame to a Isolation and Characterization of a BAC Containing the Chicken C-terminal GFP tag. All of the recombinant constructs were INK4a͞b Locus. Repeated efforts to identify a chicken orthologue validated by DNA sequencing. Expression vectors for human of p16INK4a by low stringency hybridization of genomic libraries p53, MDM2, and a variety of human ARF-GFP fusion proteins or by PCR amplification with degenerate primers were unsuc- have been described (26). The retroviruses were constructed by cessful, even using conditions and primers that allowed the cloning the relevant coding domains into the pBABE-puro vector containing two tandem copies of the hemagglutinin (HA) isolation of INK4 orthologues from fish (7, 8). Therefore, we tag (27). exploited the close linkage between INK4a and the evolution- arily conserved MTAP. In humans, MTAP is Ϸ100–150 kb distal DNA Sequence Analysis. All sequence information was generated to INK4a in the opposite transcriptional orientation (28–30), and by the automated ABI PRISM 377 DNA sequencer using the same organization is preserved in other species (Fig. 1A) including Fugu, where there is no tandem duplication of Big Dye Terminator Cycle Sequencing (Applied Biosystems). ͞ DYEnamic ET Terminator cycle sequencing kit (Amersham INK4a b (8). By using degenerate primers, based on the con- Pharmacia) was used to overcome sequencing-compression sensus sequence of human, mouse, and Xenopus MTAP, we problems. Random and directional BAC sequencing was per- isolated a 163-bp RT-PCR product whose sequence showed 85% formed by Seqlab (Go¨ttingen, Germany). Random sequences of identity with human MTAP at the amino acid level. Further the 15-kb XbaI fragment were generated by using the Tn7 details of the chicken MTAP gene will be described elsewhere transposon-based Genome Priming System (New England Bio- (M.M., S.-H.K., G.P., and J. Sgouros, unpublished work). labs). The computational programs used for analysis and assem- When this PCR product was used to screen a BAC genomic bly of DNA sequences were MACVECTOR V.6.5 (Oxford Molec- library, prepared from white leghorn chicken blood DNA, a ular, Oxford, U.K.), SEQUENCHER V.4.1.2 (Gene Codes, Ann single positive clone was identified (91-M20). On further anal- Arbor, MI), GAP and BESTFIT (Genetics Computer Group, ysis, the BAC was shown to include specific restriction fragments Madison, WI). that hybridized, albeit at low stringency, to a human p16INK4a cDNA probe. The organization of the chicken genomic locus is Protein Analyses. Cell lysate preparation, immunoblotting, and depicted schematically in Fig. 1A and as a restriction map in Fig. immunoprecipitation were performed as described (26). The 1B. Importantly, two different NotI fragments, 0.3 and 2.3 kb in antibodies used for immunoblotting were 3E1 for GFP, SMP14 size (Fig. 1B), contained sequences with strong similarity to exon for MDM2 (MS-291, NeoMarkers, Union City, CA), DO-1 for 2 of human and mouse INK4a͞b, suggesting that the BAC p53 (sc-126, Santa Cruz Biotechnology), DCS35 for Cdk4 (MS- included the second exons of both INK4a and INK4b. We could 299, NeoMarkers), K6.83 for Cdk6, (MS-398, NeoMarkers), and not, however, identify which was which because of the common F-7 for HA (sc-7392, Santa Cruz Biotechnology). For immuno- exon 2-homology region.

212 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0135557100 Kim et al. Downloaded by guest on October 1, 2021 presumptive first exon and 3Ј UTR of chicken INK4b detected the same 2.0-kb RNA (Fig. 2B, lanes 2 and 3), whereas probes representing the presumptive exon 1␤ and 3Ј UTR of ARF detected the 1.6-kb RNA (Fig. 2B, lanes 4 and 5). With the caveat that INK4a and ARF transcripts could be similar in size, the data are more consistent with the idea that CEFs do not express an RNA that encodes INK4a.

Absence of INK4a Exon 1␣ in the Chicken Genome. In the anticipation that we would locate the elusive first exon of INK4a, we determined the sequence of the entire genomic locus. Over 1,500 random sequences were generated from two independent sources. The 15-kb XbaI fragment that was known to encompass both second exons (Fig. 1B) was subjected to random transposon insertions (see Materials and Methods). We also contracted Seqlab (Go¨ttingen, Germany) to conduct random sequencing of Fig. 2. Northern and Southern blotting of chicken INK4͞ARF.(A) The the entire BAC. The locus contains extensive regions of repet- structure of full-length chicken INK4b and ARF cDNAs are shown schematically itive sequence as well as duplications of unique sequence but, by with relevant nucleotide numbers. The arrows indicate splice junctions. The shaded regions depict the respective protein-coding domains. Black bars combining the information from the cDNAs and further direc- below each cDNA show the locations of the different probes used. (B) North- tional sequencing, we were able to assemble a single contig of 18 ern blot analyses of polyadenylated RNA from late passage CEFs hybridized kb, representing the entire INK4b-ARF-INK4a locus (accession with the probes described in A. Lane numbers correspond to probe numbers. no. AY138246). A restriction map of this region is depicted The sizes of transcripts were estimated relative to RNA molecular weight in Fig. 1B, together with the positions of the exon–intron marker I (Roche Molecular Biochemicals). (C) Comparative Southern blot boundaries. analysis of BAC and CEF genomic DNA digested with BamHI (B), HindIII (H), PstI Two surprising features emerged from these data. First, we (P), PstI plus BglII (P͞Bg), and SfiI (Sf) and hybridized with the exon 1␤ of ARF only found a first exon for chicken INK4b, not INK4a. This result (probe 4 in A). would explain the absence of a recognizable INK4a transcript. The second feature was a duplication of the genomic sequences ␤ Isolation and Characterization of cDNAs from the Chicken INK4a͞b within and adjacent to the putative exon 1 of ARF. As depicted B A Locus. To facilitate the characterization of the genomic locus, a in Fig. 1 (and Fig. 6 , which is published as supporting information on the PNAS web site), there is a repeat of two 1.6-kb XhoI fragment encompassing one of the presumptive blocks of almost identical sequence. Interestingly, this duplica- INK4 exons (Fig. 1B) was used to probe a cDNA library tion of exon 1␤-like sequences occurs in the region in which we constructed from late passage CEFs. Two major groups of expected to find exon 1␣ of INK4a. cDNAs (with 13 and 22 members, respectively) were identified As the characterization of the genomic locus relied on a single based on the unique sequences flanking the common exon BAC clone, we performed comparative Southern blot analysis of 2-homology region. The longest cDNA from the first group the BAC and CEF DNA by using probes described in Fig. 2A. (accession no. AY138247) was 1,899 bp, with an ORF capable of With most probes, the BAC and cellular DNA gave identical encoding a 139-aa protein (discussed in more detail below). results (data not shown). Importantly, probe 4, encompassing BLASTX searches indicated strong similarity to human INK4a and exon 1␤, detected two sets of restriction fragments, one corre- INK4b (E values 8e-32 and 4e-33, respectively). The longest sponding to the bona fide exon and the other derived from the cDNA from the second group (accession no. AY138245) was duplicated region (Fig. 2C). The different hybridization intensity 1,523 bp, and notional translation showed no evidence for reflects the fact that the distal copy has less extensive homology sequences resembling the amino terminal region (encoded by with the probe. We suspect, but have not proved, that odd exon 1) of either INK4a or INK4b. Although BLASTX searches additional bands in the genomic DNA (e.g., the SfiI fragments also failed to reveal similarity to ARF, there was a presumptive in Fig. 2C) or slight differences in size reflect polymorphisms initiation codon and reading frame that had some of the ␤ between the BAC (haploid) and the CEF (diploid) . hallmarks of exon 1 (discussed in more detail later). Therefore, Importantly, we have no reason to suspect gross rearrangements we suspected and subsequently confirmed that most of the or deletions in the BAC that could account for the absence of cDNAs in this second group corresponded to the chicken exon 1␣. The chicken exon 1␤ probe also detected two sets of ␤ equivalent of the -transcript. restriction fragments in quail cell DNA (data not shown), ␣ ␤ Based on the mammalian model, the - and -transcripts indicating that the duplication of exon 1␤Ϫlike sequences and Ј should have the same 3 UTR, yet the two families of chicken concomitant loss of exon 1␣ may pertain in other avian species. cDNAs had different 3Ј UTRs, implying that the first group must correspond to INK4b. As the cDNA library yielded multiple The Chicken INK4b Protein Binds Cdk4 and Cdk6 and Causes Cell-Cycle full-length clones, it was puzzling that we did not find a cDNA Arrest. The 2.0-kb cDNA is capable of encoding a 139-aa protein, capable of encoding INK4a. A trivial explanation would be that with a predicted molecular weight of 14,504 Da and a pI of 11.3. the reverse transcripts simply failed to read into exon 1␣, The alignment of the amino acid sequence with those of human although 12 of the 22 cDNAs in this group did read into exon 1␤. and mouse INK4b and INK4a is shown in Fig. 3A. A phylogenetic Alternatively, the relevant RNA may not have been present in tree of all known INK4 proteins places chicken INK4b between MEDICAL SCIENCES late-passage CEFs. To distinguish between these possibilities, the mammalian INK4a and INK4b proteins and the more selected regions of the cDNAs were used to probe a Northern primitive INK4b orthologue in fishes, but clearly distinct from blot of CEF RNA. A probe containing the exon 2 sequences INK4c and INK4d (Fig. 3B). common to both cDNAs (the 1.6-kb XhoI fragment shown in Fig. To confirm that the chicken INK4b product is functional, the 1B) identified two prominent transcripts of Ϸ2.0 and 1.6 kb (Fig. coding domain was transferred into a retrovirus vector contain- 2B, lane 1). Allowing for polyadenylation, these sizes would be ing an amino-terminal 2xHA tag and transduced into HDFs. A in agreement with the longest cDNAs, implying that the latter similarly tagged version of human p16INK4a and the empty vector were full-length clones. Significantly, probes representing the provided positive and negative controls. After drug selection,

Kim et al. PNAS ͉ January 7, 2003 ͉ vol. 100 ͉ no. 1 ͉ 213 Downloaded by guest on October 1, 2021 Fig. 3. Sequence and functional evaluation of chicken INK4b. (A) Amino acid sequence alignment of chicken p15INK4b with mouse and human INK4a and INK4b. The vertical lines delineate ankyrin repeats, and the asterisks identify conserved residues. (B) Phylogenetic tree for INK4 proteins generated from multiple alignment of published INK4 sequences with the CLUSTALX V.1.8 program to perform neighbor-joining, excluding gaps, and correcting for multiple substitutions. (Bar ϭ 0.05 substitutions per site.) GenBank accession nos. are AAA92554, XP࿝027626, AAC39783, AAC27450, AAC08963, NP࿝031696, AAC52193, AAC52194, AAB35360, S77734, AAC23670, AAG44950, AAB09560, AAD21313, AJ250231, and AJ250232. (C and D) TIG3 HDFs were infected with pBabePuro retroviruses encoding HA-tagged chicken p15INK4b (p15Ch), human p16INK4a (p16Hu), or the empty vector (Vec). (C) The drug-resistant cells were lysed, immunoprecipitated with an HA antibody, and immunoblotted for HA, human Cdk4, and Cdk6. (D) Percentage of cells incorporating BrdUrd 6 days after infection. Results represent the average of three separate analyses of between 100 and 200 cells.

cell lysates were subjected to immunoprecipitation with an in independent derivations of the genomic sequence, a single anti-HA antibody. As shown in Fig. 3C human Cdk4 and Cdk6 error in the nucleotide sequence could alter the reading frame. both coprecipitated with the HA-tagged chicken p15INK4b as well Therefore, we constructed two types of expression vector: one in as with human p16INK4a. In reciprocal immunoprecipitations, which a GFP tag was inserted at the amino terminus of the Cdk4 antiserum coprecipitated chicken p15INK4b (not shown). putative ARF coding domain, and the other in which a GFP tag To assess the effect of chicken INK4b on the cell cycle, the was inserted in-frame at the end of exon 1␤ (Fig. 5A). Equivalent drug-resistant cell pools were labeled for2hwith5mMBrdUrd constructs for human ARF have been described (26). After ϩ at day 6 after infection, and the proportion of BrdUrd cells was transfection of U20S human osteosarcoma cells, the respective determined by immunohistochemistry. Both chicken INK4b and fusion proteins were detected by immunoblotting for GFP. human INK4a caused a substantial inhibition of cell prolifera- Whereas the human ARF-GFP fusion proteins had the expected tion, as judged by the reduced incorporation of BrdUrd (Fig. 3D). In preliminary experiments, we have confirmed that chicken p15INK4a will also arrest CEFs (S.-H.K. and G.P., unpublished work).

The Chicken ARF Protein Does Not Have an Alternative Reading Frame. The second cDNA, corresponding to the 1.6-kb transcript, has a methionine codon at nucleotide 147 that could serve as an initiation site for a protein of 60 amino acids and a predicted pI of 13.8 (Fig. 4A). With a molecular weight of 7,234 Da, this product would be less than half the size of mammalian ARF. In mammals, the splice from exon 1␤ to exon 2 enables ARF translation to continue in the Ϫ1 reading frame relative to that of p16INK4a, whereas the corresponding splice in chickens puts exon 1␤ in register with the ϩ1 reading frame (Fig. 4B). As this frame specifies a stop codon at the beginning of exon 2, the chicken ARF protein is encoded entirely by exon 1␤. As exon 1␤ sequences do not show a high degree of conservation (only 45% identity between the human and mouse), alignment of the chicken sequence with other species is not straightforward (Fig. 4A). Although there is Ϸ35% identity between the chicken protein and the first 64 residues of human ARF encoded by exon 1␤, computer alignments identify additional matches between chicken exon 1␤ and human exon 2 (not shown), the significance of which is unclear. As illustrated in Fig. 4 B and C, the second exon of the chicken ARF transcript retains an ORF capable of encoding 121 residues Fig. 4. (A) Amino acid sequence alignment of chicken ARF with the equiv- ␤ that could have formed the carboxyl terminus of an INK4a-like alent exon 1 regions of human, mouse, rat, and opossum proteins. (B) ␣ Ϫ Schematic representation of the human and chicken INK4a͞ARF transcripts product, if exon 1 had existed. Also, the 1 reading frame in with exons shown as boxes and coding domains as gray (INK4a) or cross- this exon is capable of encoding 162 amino acids that could have hatched (ARF). The three possible reading frames (Ϫ1, 0, ϩ1) are included for formed the carboxyl terminus of ARF, if splicing had occurred exons 2 and 3. (C and D) Coding potential of chicken INK4a exon 2 sequences. in the register used in mammals (Fig. 4D). Although we have The alignment of exon 2 encoded sequences in (C) ‘‘0’’ reading frame with confirmed the relevant sequences in multiple cDNA clones and human p16INK4a and (D) ‘‘Ϫ1’’ reading frame with human p14ARF.

214 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0135557100 Kim et al. Downloaded by guest on October 1, 2021 Fig. 5. Functional evaluation of chicken ARF. (A) Two types of chicken ARF–GFP fusion proteins (see text) are depicted schematically. The presumed stop codon is indicated by an asterisk. (B) The various ARF fusion proteins were transiently expressed in U20S cells and detected by immunoblotting against GFP. (C) The chicken ARF–GFP fusion proteins were transiently expressed in NARF cells. After induction, human p14ARF was detected by immunofluorescence, and GFP was visualized directly. Merge shows regions of colocalization in yellow. (D) After cotransfection of U2OS cells with vectors encoding human MDM2 and the human or chicken ARF GFP fusions, cell lysates were immunoprecipitated with an MDM2 antibody. After SDS͞PAGE, the precipitated proteins were immunoblotted for MDM2 (Upper) or GFP (Lower). (E) U20S cells were cotransfected with plasmids encoding human MDM2 and human p53 as indicated by ϩ and Ϫ, together with GFP alone, GFP-ARFHu, or GFP-ARFCh. The levels of the various proteins were then detected by using antibodies against MDM2 (Upper), p53, and GFP (Lower). Lanes 4 and 5 and lanes 6 and 7 represent duplicate pairs of transfections.

molecular weight of Ϸ43 kDa in SDS͞PAGE, both versions of containing 132 or 64 residues of human ARF (GFP-ARFHu and the chicken ARF-fusion proteins were Ϸ36 kDa (Fig. 5B). This ⌭1␤Hu-GFP) were expressed in U20S cells along with a plasmid size would be consistent with natural termination at the end of encoding human MDM2. Both versions of chicken ARF coim- exon 1␤, as predicted from the sequence. Curiously, the fusion munoprecipitated with MDM2 as effectively as human ARF protein with human exon 1␤ has a higher mobility than does the (Fig. 5D). In similar cotransfection assays, MDM2 promotes the chicken counterpart, despite having more amino acid residues. It ubiquitylation and proteasome-mediated destruction of p53 (32, is unclear whether this difference reflects posttranslational 33), as shown in Fig. 5E, lanes 2 and 3. In this system, GFP- modification or anomalous mobility caused by the unusual amino ARFCh was able to protect p53 from MDM2-mediated destruc- acid composition of the proteins. tion (Fig. 5E, lanes 6 and 7), although not as effectively as human p14ARF (Fig. 5E, lanes 4 and 5). Thus, the 60-aa chicken ARF Chicken ARF Binds MDM2 and Stabilizes p53. As the mammalian protein displays most if not all of the functional attributes of its ARF proteins show distinctive nucleolar localization, it was of mammalian counterparts. interest to determine the subcellular distribution of chicken ARF by using the two GFP fusion proteins. For these studies, we Discussion used a derivative of the U20S cell line (NARF) that expresses Our analysis of the chicken INK4͞ARF locus revealed a number human p14ARF from an IPTG-inducible promoter. As described of unexpected features. First, the chicken locus lacks the capacity (23, 26), induction of human ARF in these cells results in to encode p16INK4a, as we found no evidence for an INK4a nucleolar accumulation, here detected by immunofluorescence transcript or exon 1␣-related sequence in either BAC or CEF using the monoclonal antibody 4C6 (red signal in Fig. 5C). DNA. However, the presence of two copies of exon 2 implies that Interestingly, whereas some of the GFP-ARFCh fusion protein the tandem duplication of INK4a͞b took place before the branch was also in the nucleolus, a substantial proportion was localized between birds and terrestrial animals, and that exon 1␣ was in nonnucleolar bodies. A similar distribution was observed subsequently lost. The partial repeat of exon 1␤ and adjacent when using the chicken E1␤-GFP construct (Fig. 5C), with or sequences exactly where we would have expected to find exon 1␣ without induction of human ARF (not shown). As p53 has been suggests that these events are connected and happened quite reported to localize in so-called promyelocytic leukemia (PML) recently in evolution. Otherwise, the distal exon 2 sequence MEDICAL SCIENCES nuclear bodies in some circumstances (reviewed in ref. 31), we should not have been so well conserved. In the absence of a costained the cells with an antibody against the PML protein. functional INK4a gene, it will be interesting to determine There was no correlation with the ARF speckles (data not whether INK4b assumes some of its functions. Chicken p15INK4b shown). clearly has the ankyrin repeat, Cdk binding, and cell-cycle arrest In view of its distinctive nuclear localization, we next asked properties of a typical INK4 protein (Fig. 3), and as the whether chicken ARF could substitute for human p14ARF in primordial INK4 gene at this locus, INK4b probably performed functional assays. The two versions of chicken ARF (GFP- physiological roles that were subsequently assumed and perhaps ARFCh and ⌭1␤Ch-GFP) and the corresponding fusion proteins refined by INK4a. A two exon version of INK4a, the vestiges of

Kim et al. PNAS ͉ January 7, 2003 ͉ vol. 100 ͉ no. 1 ͉ 215 Downloaded by guest on October 1, 2021 which remain apparent in chickens, presumably predates the the adjacent MTAP gene is in the opposite transcriptional three exon format adopted in mammalian genomes. orientation, and we did not find evidence for an intervening A further surprise was that the presumptive ARF protein in gene, such as NTp16 (39). Had the ancestral exon 1␤ translocated chickens terminates abruptly at the end of exon 1␤ rather than before the INK4 duplication, it is not clear that there would have exploiting sequences from exon 2. Because there is no ‘‘alter- been a suitable exon and polyadenylation signal available. There- native reading frame,’’ the ARF acronym is not really appro- fore, we suspect that exon 1␤ became transposed between INK4b priate. However, most of the published data on mammalian ARF and INK4a after the duplication. Thus, rather than chicken ARF imply that the amino terminal region encoded by exon 1␤ is losing the capacity to exploit exon 2, it seems more likely that sufficient for the known functions of protein (22, 23, 26, 34–38), mammalian ARF acquired this capacity through an alteration in and the data on chicken ARF concur. Thus, despite comprising the splicing register. Of course, this interpretation begs the only 60 amino acids, of which 22 are arginines, chicken ARF can question whether the exon 2-encoded sequences contribute interact with human MDM2 and protect human p53 from significantly to ARF function in mammalian cells. Paradoxically, MDM2-mediated degradation (Fig. 5), albeit less effectively our attempts to clarify such issues by characterizing the chicken than the human p14ARF. This difference could simply reflect locus now suggest that it would be informative to conduct similar reduced affinity between proteins from different species. Our investigations in other species. In the short term, however, the recent data indicate that chicken ARF will stabilize endogenous data described here will enable us to develop reagents to p53 in CEFs (S.-H.K. and G.P., unpublished work). investigate the relative roles of p15INK4b and ARF in chicken cell We favor the interpretation that, whether or not it was part of senescence and to draw comparisons between the regulatory a more complex ancestral gene, the original exon 1␤ was itself networks operating in different cell types and species. capable of specifying a functional protein. Splicing into exon 2 of INK4a might simply have facilitated the production of a poly- We thank John Sgouros, David Ish-Horowicz, and Mike Fried for helpful adenylated transcript. In principle, any 3Ј exon would suffice, but comments on the manuscript.

1. Sharpless, N. E. & DePinho, R. A. (1999) Curr. Opin. Genet. Dev. 9, 22–30. 21. Zhang, Y. & Xiong, Y. (1999) Mol. Cell 3, 579–591. 2. Sherr, C. J. (2001) Nat. Rev. 2, 731–737. 22. Weber, J. D., Kuo, M.-L., Bothner, B., DiGiammarino, E. L., Kriwacki, R. W., 3. Ruas, M. & Peters, G. (1998) Biochim. Biophys. Acta 1378, 115–177. Roussel, M. F. & Sherr, C. J. (2000) Mol. Cell. Biol. 20, 2517–2528. 4. Ortega, S., Malumbres, M. & Barbacid, M. (2002) Biochim. Biophys. Acta 1602, 23. Stott, F. J., Bates, S., James, M. C., McConnell, B. B., Starborg, M., Brookes, 73–87. S., Palmero, I., Hara, E., Vousden, K. H. & Peters, G. (1998) EMBO J. 17, 5. Serrano, M., Hannon, G. J. & Beach, D. (1993) Nature 366, 704–707. 5001–5014. 6. Weinberg, R. A. (1995) Cell 81, 323–330. 24. McConnell, B. B., Starborg, M., Brookes, S. & Peters, G. (1998) Curr. Biol. 8, 7. Kazianis, S., Morizot, D. C., Della Coletta, L., Johnston, D. A., Woolcock, B., 351–354. Vielkind, J. R. & Nairn, R. S. (1999) Oncogene 18, 5088–5099. 25. Sambrook, J., Fritsch, E. F. & Maniatis, T. (1989) Molecular Cloning: 8. Gilley, J. & Fried, M. (2001) Oncogene 20, 7447–7452. A Laboratory Manual (Cold Spring Harbor Lab. Press, Plainview, NY), 9. Alcorta, D. A., Xiong, Y., Phelps, D., Hannon, G., Beach, D. & Barrett, J. C. (1996) Proc. Natl. Acad. Sci. USA 93, 13742–13747. 2nd Ed. 10. Hara, E., Smith, R., Parry, D., Tahara, H., Stone, S. & Peters, G. (1996) Mol. 26. Llanos, S., Clark, P. A., Rowe, J. & Peters, G. (2001) Nat. Cell Biol. 3, 445–452. Cell. Biol. 16, 859–867. 27. Ruas, M., Brookes, S., McDonald, N. Q. & Peters, G. (1999) Oncogene 18, 11. Serrano, M., Lin, A. W., McCurrach, M. E., Beach, D. & Lowe, S. W. (1997) 5423–5434. Cell 88, 593–602. 28. Nobori, T., Takabayashi, K., Tran, P., Orvis, L., Batova, A., Yu, A. L. & Carson, 12. De Stanchina, E., McCurrach, M. E., Zindy, F., Shieh, S.-Y., Ferbeyre, G., D. A. (1996) Proc. Natl. Acad. Sci. USA 93, 6203–6208. Samuelson, A. V., Prives, C., Roussel, M. F., Sherr, C. J. & Lowe, S. W. (1998) 29. Olopade, O. I., Pomykala, H. M., Hagos, F., Sveen, L. W., Espinosa, R. I., Genes Dev. 12, 2434–2442. Dreyling, M. H., Gursky, S., Stadler, W. M., Le Beau, M. M. & Bohlander, S. K. 13. Radfar, A., Unnikrishnan, I., Lee, H.-W., DePinho, R. A. & Rosenberg, N. (1995) Proc. Natl. Acad. Sci. USA 92, 6489–6493. (1998) Proc. Natl. Acad. Sci. USA 95, 13194–13199. 30. Schmid, M., Sen, M., Rosenbach, M. D., Carrera, C. J., Friedman, H. & Carson, 14. Zindy, F., Eischen, C. M., Randle, D. H., Kamijo, T., Cleveland, J. L., Sherr, D. A. (2000) Oncogene 19, 5747–5754. C. J. & Roussel, M. F. (1998) Genes Dev. 12, 2424–2433. 31. Borden, K. L. (2002) Mol. Cell. Biol. 22, 5259–5269. 15. Wei, W., Hemmer, R. M. & Sedivy, J. M. (2001) Mol. Cell. Biol. 21, 6748–6757. 32. Haupt, Y., Maya, R., Kazaz, A. & Oren, M. (1997) Nature 387, 296–299. 16. Randle, D. H., Zindy, F., Sherr, C. J. & Roussel, M. F. (2001) Proc. Natl. Acad. 33. Kubbutat, M. H. G., Jones, S. N. & Vousden, K. H. (1997) Nature 287, 299–303. Sci. USA 17, 9654–9659. 34. Quelle, D. E., Cheng, M., Ashmun, R. A. & Sherr, C. J. (1997) Proc. Natl. Acad. 17. Sviderskaya, E. V., Hill, S. P., Evans-Whipp, T. J., Chin, L., Orlow, S. J., Easty, Sci. USA 94, 669–673. D. J., Cheong, C. C., Beach, D., DePinho, R. A. & Bennett, D. C. (2002) J. Natl. 35. Zhang, Y., Xiong, Y. & Yarbrough, W. G. (1998) Cell 92, 725–734. Cancer Inst. 94, 446–454. 18. Palmero, I., Pantoja, C. & Serrano, M. (1998) Nature 395, 125–126. 36. Lohrum, M. A. E., Ashcroft, M., Kubbutat, M. H. G. & Vousden, K. H. (2000) 19. Brookes, S., Rowe, J., Ruas, M., Llanos, S., Clark, P. A., Lomax, M., James, Curr. Biol. 10, 539–542. M. C., Vatcheva, R., Bates, S., Vousden, K. H., Parry, D., Gruis, N., et al. (2002) 37. Midgley, C. A., Desterro, J. M. P., Saville, M. K., Howard, S., Sparks, A., Hay, EMBO J. 21, 2936–2945. R. T. & Lane, D. P. (2000) Oncogene 19, 2312–2323. 20. Palmero, I., McConnell, B., Parry, D., Brookes, S., Hara, E., Bates, S., Jat, P. 38. Clark, P. A., Llanos, S. & Peters, G. (2002) Oncogene 21, 4498–4507. & Peters, G. (1997) Oncogene 15, 495–503. 39. Pantoja, C., Palmero, I. & Serrano, M. (2001) Exp. Gerentol. 36, 1289–1302.

216 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0135557100 Kim et al. Downloaded by guest on October 1, 2021