Cyclin G1 and Cyclin G2 Comprise a New Family of Cyclins with Contrasting Tissue-specific and Cell Cycle-regulated Expression*

We describe the isolation and characterization of cDNAs encoding full-length human and murine cyclin G1 and a novel human homologue of this cyclin desig- nated cyclin G2. Cyclin G1 is expressed at high levels in skeletal muscle, ovary, and kidney. Following an initial up-regulation from early G 1 to G 1 /S phase, cyclin G1 mRNA is constitutively expressed throughout the cell cycle in T and B cell lines. In contrast, in stimulated peripheral T cells, cyclin G1 mRNA is maximal in early G 1 phase and declines in cell cycle progression. Cyclin G1 levels parallel p53 expression in murine B lymphocytes; however, in several human Burkitt’s lymphomas, murine lymphocytes treated with transforming growth factor- (cid:98) , early murine embryos, and several tissues of p53 null mice, cyclin G1 levels are either inverse of p53 levels or expressed independent of p53. The cyclin G1 homologue, cyclin G2, exhibits 60% nucleotide sequence identity and 53% amino acid sequence identity with cyclin G1, and like cyclin G1, exhibits closest sequence identity to the cyclin A family. Distinct from cyclin G1, the amino acid sequence for cyclin G2 shows a PEST- rich sequence and a potential Shc PTB binding site. Cyclin differentially in cerebel-lum, and in

Transitions through the eukaryotic cell division cycle are primarily coordinated by the sequential activation of cyclin-dependent kinases (CDKs) 1 which are, in turn, regulated by subunit associations and phosphorylation (reviewed in Refs. [1][2][3]. The cyclins represent a group of closely related molecules which primarily function at specific stages of the cell cycle as regulators of CDK activity by binding and forming active com-plexes with specific partner CDKs. This cyclin-CDK association is in part determined by the conserved cyclin region of ϳ110 amino acids referred to as the cyclin box (4 -6). The cyclin box exhibits ϳ30 -50% identity between the different cyclin types, the consensus sequence varying depending on the class and subclass of cyclin (5,7,8). Cyclins have been classified into different groups on the basis of their structural similarity, functional period in the cell division cycle and regulated expression. In addition to providing positive growth control, CDKs and cyclin-CDK pairs may participate in metabolism and signal transduction unrelated to cell cycle as evidenced by pho80-pho85 cyclin-CDK complex participation in yeast phosphate metabolism (9,10), the expression of CDK5 in nonproliferating brain tissue (11)(12)(13), and the SRB10/11 cyclin-CDK regulator of RNA polymerase II (14).
To date at least 12 different cyclins in budding yeast, 4 in fission yeast, and 10 in mammalian cells (cyclins A-H with multiple family members for some types) are known, all primarily displaying sequence homology within the cyclin box region (1)(2)(3)(4)(5). Two broad classes of cyclins, based on genetic complementation of cell cycle mutations, have been identified in yeast: the G 1 or "START" cyclins regulating G 0 to G 1 to S phase transition, and the G 2 /M phase class necessary for mitosis. In higher eukaryotes, the G 1 cyclins (reviewed in Refs. 2, 15, and 16) are represented by the D-type (D1-D3) (12,(17)(18)(19)(20)(21) and cyclins C (12,22) and E (12,23); the mitotic cyclins by cyclin A (13,24) and the B-type (B1-B3) (25)(26)(27), with cyclin A acting earlier than the B-type. Cyclin F (28), as well as cyclin A (when complexed with CDK2), are believed to function during S phase. Cyclin H is reported to act indirectly at S to G 2 transition by regulating, in a complex with the CDK activating kinase MO15, the phosphorylation of other cyclin-CDK complexes in a cyclin/kinase signal transduction cascade (29,30). Although partner CDKs have been identified for many cyclins, a number of CDK related molecules have been discovered for which no cognate cyclin is known (reviewed in Refs. 2, 3, and 31). The activity of some cyclin-CDK complexes is also subject to regulation through interaction with CDK inhibitors (p15 and p16 INK4 , p21 WAF1/CIP1 , and p27 KIP1 ) produced in response to negative stimuli to prevent cell cycle progression (reviewed in Refs. [32][33][34]. Cyclin G was first identified serendipitously in screens for src family kinases in rat fibroblasts (35), and later by differential screen for transcriptional targets of p53 (36). The function of cyclin G has not been determined, however, identification of multiple p53 binding sites in the genomic DNA (37), transactivation by wild-type p53, and induction following ␥-irradiation (responses associated with checkpoint and cell cycle arrest (38,39)) implicate cyclin G in negative growth control or DNA damage repair (36,37). The absence of either prototypic protein destabilizing (PEST) sequences of G 1 class cyclins (40,41), or the "destruction box" sequence controlling the ubiquitin-de-* The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM 1 The abbreviations used are: CDK, cyclin-dependent kinases; EGF, epidermal growth factor; PCR, polymerase chain reaction; ORF, open reading frame; bp, base pair(s); EST, expressed sequence tagged; RACE, rapid amplification of cDNA ends; UTR, untranslated region; kb, kilobase pair(s); GAPDH, glyceraldehyde-3-phosphate dehydrogenase; TGF-␤, transforming growth factor ␤; FACS, fluorescence-activated cell sorter. pendent degradation of mitotic cyclins (42), indicate alternate regulation of cyclin G protein expression. Additionally, the presence of an EGF-R/ErbB-like autophosphorylation motif has suggested a role for cyclin G in signal transduction (35). We have isolated cDNAs encoding full-length human and murine cyclin G which we have designated cyclin G1. Here we describe the expression pattern for cyclin G1 mRNA throughout the cell cycle in B and T cells, and provide correlative evidence for its regulation by p53 in murine B lymphocytes, yet contrast this with mRNA expression inverse to p53 levels during development and independent of p53 in various cell lines and differentiated tissues. In parallel, we describe the cloning of a novel human cyclin G1 homologue, designated cyclin G2, and compare its tissue and cell cycle position-specific transcript expression to that of cyclin G1. The presence of two potentially phosphorylated tyrosines in the G2 protein sequence distal to CDK interaction may implicate cyclin G2 as a component of signal transduction pathways regulating cell growth.

MATERIALS AND METHODS
Cell Lines and Culture-Subclones of the murine B cell lines WEHI-231 and Bal-17 were courtesy of A. DeFranco, University of California, San Francisco. The murine T cell lymphoma EL-4 and C1498 myeloid leukemia cell lines; human Jurkat T lymphoblastoid, NALM-6 pre-B acute lymphoblastic leukemia; IM-9 B-lymphoblastoid cell line and the human Burkitt's lymphoma B cell lines Ramos, Daudi, Raji, and DG-75 were obtained from J. Ledbetter (this institution). The above cell lines and subclones were grown in RPMI 1640 (Life Technologies, Inc., Grand Island, NY) supplemented with 10% heat-inactivated fetal bovine serum, 2 mM L-glutamine, 1 mM sodium pyruvate (and 50 M 2-mercaptoethanol for WEHI-231 and Bal-17) at 37°C in 6% CO 2 . All human cell lines were grown in this complete RPMI media supplemented with 100 units/ml of penicillin G sodium and 100 g/ml streptomycin sulfate. Human peripheral T cells were isolated from peripheral blood as described (43), and activated to proliferation by treatment with 1 g/ml phytohemagglutinin for 24 h followed by culture in T flasks in complete RPMI supplemented with 5 ng/ml recombinant interleukin-2 (Boehringer Mannheim). All cultured cells were maintained in and sampled from exponential phase of growth at a density between 2-5 ϫ 10 5 cells per ml for cell lines and 1-2 ϫ 10 6 cells per ml for the stimulated human peripheral T cells.
PCR Cloning-Human cyclin G1 and cyclin G2 ORF cDNA fragments were cloned by PCR from total RNA isolated from Jurkat, NALM-6, and IM-9 cells. cDNA was generated from RNA using the SuperScript cDNA synthesis kit (Life Technologies, Inc.) and the PCR from these templates was done using PCR Optimizer TM kit (Invitrogen Corp., San Diego, CA). Oligonucleotide primers were designed from regions in expressed sequence tagged (EST) and/or partial cDNA sequences obtained by a BLAST search of the GenEMBL data bases for sequences showing significant identity at the nucleic (human cyclin G1) or deduced amino acid (human cyclin G2) level to the rat cyclin G1 (EMBL X70871).
For human cyclin G1, three fragments overlapping different regions of the GeneBank sequence for rat cyclin G1 (EMBL Z24820, X77794, and GB T16966) were used to generate a composite, putative full-length human cyclin G1 sequence and design corresponding oligonucleotide primers (forward primers: 5Ј-AAGATGATAGAGGTACTG-3Ј, 5Ј-AA-GATGATAGAGGTACTGACAACAACTGACTC-3Ј, and 5Ј-CAGGAGT-CTAGATGTCAGCC-3Ј; reverse primers: 5Ј-CAGTTAAGGGACCATT-TCAGGAATTGTTGG-3Ј, and 5Ј-GCAATGATAGACAATGCCAA-3Ј) to generate PCR fragments with partial and full ORFs. These fragments were cloned into the pCR TM II cloning vector following the TA Cloning R kit protocol (Invitrogen Corp.), their sequence, determined as described below, was subsequently utilized to generate probes for human and murine cDNA libraries. For murine cyclin G1, original forward primers (5Ј-GGCGGATCCAAGATGATAGAAGTACTG-3Ј, 5Ј-GGCGGATCCCT-GAGTCTAACTCAGTTCTTTGGC-3Ј) and reverse primers: (5Ј-GGC-GAATTCATTCTACTCCTTCTGTTAACTCCAC-3Ј, and 5Ј-GGCCTTA-AGCTAACCCATGGTTTCGGGAATTGTTGGG-3Ј) were designed (with engineered restriction enzyme recognition sites at the termini) from regions showing nearly 100% identity in nucleotide sequence alignment between the GeneBank rat cyclin G1 sequence (EMBL X70871) and our composite putative human cyclin G1 sequence. When differences or ambiguities were present between the sequences in the alignment, the nucleotides of the rat sequence were used to construct the primers for the murine cyclin G1. These primers were then utilized to generate PCR fragments from murine splenocyte and peripheral blood leukocyte cDNA libraries (Clontech, Palo Alto, CA) that were then cloned and sequenced as described below. Sequences of PCR-derived fragments were compared to those of cDNA library clones obtained by hybridization screening (described below).
For human cyclin G2 an automated data base search using BLASTX was instituted to search the translation products from all six reading frames of the GenEMBL sequence updates on a weekly basis and compare the deduced amino acid sequences to various regions of rat cyclin G1, in particular the cyclin box. One such search identified a 361-bp EST sequence obtained from the cDNA derived from brain tissue of a human infant (EMBL Z46163). This EST showed a BLAST probability (p) value of 2.1e Ϫ11 to the first 44 amino acids of the rat cyclin G1 box used for the search. Subsequent BLASTX translation of the entire EST sequence yielded p values ϭ 3.3e Ϫ32 , 5.6e Ϫ22 , and 1.1e-21 to the respective deduced amino acid sequence of human, rat, and mouse cyclin G1 sequences now present in the GenEMBL data base. The deduced amino acid sequence identities ranged between 66 and 44% for two respective contiguous regions in two different reading frames, which together encompassed the entire EST sequence. Oligonucleotides were designed to the 5Ј and 3Ј termini of this EST (forward primers: 5Ј-GGGGTCCAACTTCTCGGGTTGTTGAACG-3Ј and 5Ј-CCAACTTC-TCGGGTTGTTGAACGTCTACC; reverse primers: 5Ј-GTACATTTAC-ACTGACTAATCCGG-3Ј and 5Ј-CTAATCCGGATCACATCATGAGTG-3Ј) and used to PCR clone this fragment from a human brain cDNA library and cDNAs prepared from human B and T cell lines. These fragments were isolated, cloned, and sequenced as described above and below. The sequence-confirmed and corrected clones encompassed a single, contiguous ORF encoding a peptide with 53% amino acid identity to human cyclin G1. This cDNA fragment was then labeled with digoxigenin and used to probe a Jurkat cDNA library as described below.
To determine the sequence for the 3Ј end of the cyclin G2 ORF not encompassed by the cDNA library clones, overlapping cDNAs from both Jurkat and Daudi total RNA were obtained using the PCR based 3Ј-RACE method and reagents (Life Technologies, Inc.). Following the manufacturer's suggested protocol, first strand synthesis of the cDNA template was obtained with a supplied tagged oligo(dT) primer. Second strand synthesis was done with two different G2 specific forward primers in the 3Ј exons on either side of the 3Ј most splice-site junction (5Ј-GGACAACAGCTACTATAGTGTTCC-3Ј and 5Ј-GAAAGTGAGGA-CTCTTGTG-3Ј) and a supplied 3Ј primer annealed to the template cDNA at 65°C in a 20-cycle PCR. A second round of PCR with the nested 5Ј primers was subsequently performed for 15 cycles on an aliquot of cDNA obtained from the above 2nd strand synthesis. The products were TA cloned and multiple independent clones were isolated and sequenced (described above and below).
To PCR clone a mature cDNA encompassing the cyclin G2 ORF, oligonucleotides were designed (with appended restriction enzyme sites) to sites 5Ј to initiation codon and 3Ј to the last intron/exon junction near the termination codon (forward primer: 5Ј-GGCGGATC-CCTCTGTGTGGTGTCTTTACTG-3Ј; reverse primer: 5Ј-GCCGAAT-TCGGTGCACTCTTGATCACTGGG-3Ј) and used to amplify DNA fragments from Jurkat cDNA as described above.
Isolation of Library cDNA Clones-Probes for human cyclin G1 and G2 genes were synthesized by PCR using the Genius TM system digoxigenin labeled dNTP Mix (Boehringer Mannheim). The resulting PCR fragments were purified by agarose gel electrophoresis using the Gene-Clean II R DNA purification kit (BIO 101 Inc., La Jolla, CA) and used to screen both human T cell (Jurkat Zap R II, Stratagene, La Jolla, CA) and murine leukocyte (peripheral blood leukocyte in gt11, Clontech) cDNA libraries. For high-stringency, intra-species screening of the Jurkat cDNA library, Southern hybridizations of plaque lifts on Hybond TM -N ϩ nylon filters (Amersham) were done with the digoxigeninlabeled DNA fragment at ϳ5-20 ng/ml in hybridization buffer containing 50% formamide, 50 mM sodium phosphate, pH 6.5, 50% formamide, 5 ϫ SSC hybridization buffer (44) at 42°C and washed as described below for Northern blots. For low-stringency, cross-species screening of the murine peripheral blood leukocyte cDNA library, the hybridization buffer contained 30% formamide and filters were hybridized and washed at 37°C. Development of the washed filters with alkaline phosphatase-conjugated anti-digoxigenin antibodies and the Lumi-Phos TM 530 reagent (Boehringer Mannheim) was done according to manufacturer's protocol. Secondary and tertiary screening to purify positive plaques was done as described above. The murine tertiary screens were also done at high stringency, but with murine cyclin G1 cDNA probes (see above). The phage DNA was amplified, extracted, and purified for sequence analysis following the manufacturer's recommended and standard molecular biology techniques (44).
Nucleotide Sequence Determination and Analysis-DNA sequence determination (44) was done utilizing the Sequenase TM (version 2.0) system following procedures recommended by the manufacturer. cDNA fragments were cloned into either the pCR TM II cloning vector (Invitrogen), gt11 phage (Clontech), Zap R II phagemid or pBluescript II (Stratagene) cloning vectors and the sequence reaction was primed with either a "universal," vector-specific oligonucleotide or synthetic oligonucleotides homologous to the cloned fragment's internal sequences.
[␣-33P ]dATP or [␣-32P ]dATP (at 800 Ci/mmol) was used to radioactively label the DNA fragments. Nucleotide sequences were read from scanned gels with the aid of BioImage R sequence analysis software. The computer-aided editing and alignment of DNA sequences was accomplished using Genetics Computer Group (GCG) (Madison, WI) sequence analysis software. Additional nucleotide and cDNA-derived peptide sequence comparisons were performed using the BLAST program. Final alignments were performed using the GCG Pileup and Pretty programs.
Counterflow Centrifugal Elutriation and Analysis of Cell Cycle Position-Human and murine lymphocytes were separated into progressive stages of the cell cycle according to mass and cell volume by centrifugal elutriation. Exponentially growing cells were maintained at densities of 2 ϫ 10 5 to 5 ϫ 10 5 cells per ml in roller bottles, and fresh medium was added 12 h prior to harvest to enhance logarithmic growth. Cultures were concentrated by centrifugation and 2 ϫ 10 8 cells were gently resuspended in 10 ml of elutriation buffer (RPMI plus 5% fetal bovine serum, and for murine cell lines, 1 mM EDTA), and then passed through an 18-gauge needle to insure monodispersion. Cells were separated into populations of progressively increasing cell size in a centrifuge (Beckman J-6B) equipped with an JE-6B elutriation rotor at 21°C. Cells were loaded at a rotor speed of 1,800 rpm and an initial flow rate of 11 ml/min maintained by a Master Flex model 7550 -60 digital pump (Cole-Parmer Instrument Co., Chicago, IL). Fractions were obtained by increasing the flow rate of the elutriation buffer at 2 or 3 ml/min increments, and immediately collecting the first 125 ml of dispensed media following each increase. PHA and interleukin-2 stimulated human peripheral T cells were elutriated in a similar manner except 4 ϫ 10 8 cells were loaded with an initial pump rate of 11 ml/min into the chamber held at 2030 rpm; 100-ml fractions were collected at 3 ml/min increments. To verify cell cycle position, aliquots of each fraction were fixed with 70% methanol, stained with 50 g/ml propidium iodide (Calbiochem, San Diego, CA) in phosphate-buffered saline containing 100 units/ml RNase A, and analyzed for DNA content using flow cytometry (FACScan; Becton-Dickinson Instruments; San Jose, CA). Relative DNA synthesis was determined via [ 3 H]TdR incorporation as described (45).
The glyoxal denaturation of total RNA and electrophoresis in glyoxalagarose was done following standard protocol (44). After electrophoresis a control check of the relative amount and quality of the RNA was done by short wave UV fluorescent shadowing of the ribosomal RNAs on a F-254 TLC plate. The fractionated RNAs were transferred and fixed to Pall Biodyne A nylon membranes (Life Technologies, Inc.) followed by removal of residual glyoxal as described (44,47).
The [␣-32 P]dTTP and [␣-32 P]dCTP radioactive labeling of DNA fragments was done using PCR generated and GeneClean II R isolated DNA fragments as templates and reagents obtained from the Boehringer Mannheim Random Priming kit. Hybridization of DNA probes to Northern blots was done essentially following the method "B" protocol described by Pall BioSupport. PhosphorImaging (Molecular Dynamics) exposure of the washed Northern blot filters was routinely done immediately prior to autoradiography.
Examination of p53 Functional Activity by Growth Arrest-To evaluate p53 functionality, logarithmic cultures in complete growth media were exposed to drug or radiation treatments and assessed for G 1 arrest in response to DNA damage. For human cells, radiation treatments of 5 gray were delivered by a 137 Cs ␥-irradiator of an incident dose rate of 1 gray/min followed by 16 h culture. For murine B cells, cultures were grown in media containing actinomycin D (Sigma) at 0.5 nm for 24 h. Immediately following the above treatment course, cell cycle analysis was performed by pulsing cells with 10 mM 5-bromodeoxyuridine for 4 h and staining cells with fluorescein isothiocyanate-conjugated anti-5bromodeoxyuridine (Beckton Dickinson) to detect DNA synthesis and propidium iodide for total DNA content (Becton Dickinson protocol), followed by flow cytometry as described above.

Isolation of Cyclin G1 cDNA Clones-
The NH 2 -terminal peptide region of most cyclins contains the destruction box or in some cases PEST protein destabilization sequences involved in the regulation of their cyclic expression pattern. Unlike the primary structure of these cyclins, the previously reported rat and human cyclin G protein sequences initiated from an ATG at the start of the cyclin box region and lacked flanking NH 2terminal regions (35,36,48). In addition, the predicted cyclin G protein was reported to contain a potential tyrosine phosphorylation site in the carboxyl terminus, and the transcript appeared to be up-regulated both by p53 (36) and following transition from G 0 to G 1 (35). We have cloned human and murine cDNAs encoding cyclin G. Alignment of gene bank rat cyclin G cDNA with the GenEMBL data base (described under "Materials and Methods") identified several partial human sequences with high nucleic acid and deduced amino acid sequence identity extending to either side of the cyclin box region of rat cyclin G. Comparison of the rat nucleotide sequence reported (35) with that submitted to the EMBL data base (EMBL X70871) differed by one nucleotide, the sequence in the literature apparently lacking an additional adenosine at position Ϫ60. Our predicted ORF extended an additional 135 nucleotides 5Ј to the first cyclin box ATG. Based on the above alignments we PCR cloned cDNA fragments encompassing human cyclin G and sequence of these cDNA fragments verified our previous alignment. These fragments were used for screening both human and murine cDNA libraries. In parallel, PCR primers based on our alignments between the rat and the putative human cyclin G ORF were used to PCR clone murine cyclin G cDNA from a murine splenic cDNA library.
Hybridization screening of a Jurkat cDNA library with the PCR-isolated cDNA probes identified 5 independent human clones comprising two overlapping cDNAs encompassing full and partial human cyclin G cDNA. Screens of the murine peripheral blood leukocyte library identified 10 clones comprising four overlapping partial cDNAs encompassing the full cyclin G ORF. Sequence was verified by double stranded analysis of all human and murine PCR and cDNA library cyclin G clones obtained from lymphoid tissues and cell lines. The cyclin G ORF in both human and murine extended another 132-135 nucleotides, encoding a protein with an NH 2 -terminal region of 45 (murine) or 46 (human) amino acids beyond the cyclin box with a revised molecular mass of 34 kDa (Fig. 1). This finding is consistent with the NH 2 -terminal flanking region and revised molecular weight recently reported in the full-length sequence of rat cyclin G (37). Here we designate this cyclin G1 to distinguish it from its homologue. In the human 3Ј-UTR we find base substitutions of a thymidine for a guanosine, an adenosine for a guanosine, and a guanosine for an adenosine at 208, 263, and 346 nucleotides 3Ј to the TAA STOP codon, respectively (48). Comparative analysis of the cyclin G1 3Ј-UTRs determined that the human sequence lacked a block of ϳ800 -1050 nucleotides found in the rodent sequences, yet contains an island of conserved sequence shared by all three 3Ј-UTRs. A concentration of multiple copies of the sequence AUUUA, a mRNA destabilization motif (49 -51), is found within this island of conserved sequence (data not shown).
Distribution of Cyclin G1 Expression in Murine Tissues and Regulation of Cyclin G1 mRNA during Cell Cycle-To determine the size and distribution of cyclin G1 in murine tissues and cell lines, Northern blot analysis was performed. Murine cyclin G1 mRNA was ϳ3.4 kb in size and differentially expressed in murine tissues. Message was most prominent in kidney, heart, and skeletal muscle relative to ␤-actin (not shown) and GAPDH levels (Fig. 2, left panel). As we are interested in the role cyclin G1 might play in leukocyte proliferation, we compared cyclin G1 expression in several murine cell lines. The highest levels of cyclin G1 were seen in the B lymphoblastoid cell line BCL-1, and the lowest in the myeloid cell line C1498 (Fig. 2, right panel).
Expression of murine cyclin G1 through the cell cycle in B lymphocytes was examined in cell lines from two distinct stages of developmental progression: WEHI-231, representing an IgM ϩ immature B cell which undergoes growth arrest upon B cell receptor cross-linking, and Bal-17 representing an IgM ϩ IgD ϩ mature B cell phenotype which continues to proliferate upon IgM ϩ cross-linking (52). Logarithmically growing cultures of these two B cell lines were separated into populations  1-3), T-cell (lanes 4 -7), thymic epithelium (lanes 8 -9), monocyte and melanoma (lanes 10 and 11, respectively) cell lines. at progressive stages of the cell cycle by centrifugal elutriation, and Northern blot analysis was performed on total RNA isolated from each population (Fig. 3, A and B). WEHI-231 cyclin G1 expression increased ϳ2.5-fold in cells transiting early G 1 phase and remained at this level throughout the rest of the cell cycle. In contrast, cyclin G1 message in corresponding fractions from Bal-17 was markedly reduced from that seen in WEHI-231 (ϳ5-6-fold less than the maximum WEHI-231 transcript level). The constitutively low level of cyclin G1 throughout the cell cycle exhibited no increase upon early G 1 phase transition (Fig. 3C).
Our earlier survey of wild-type p53 in various cell lines had indicated that no detectable amount of wild-type p53 protein was precipitable (using either of two different monoclonal antibodies to wild-type p53) from metabolically labeled protein lysates of Bal-17, in contrast to that seen for WEHI-231 and EL-4 (Fig. 3D, top panel). The report of the cyclin G gene as a target of p53 activation (36) prompted us to further investigate the status of p53 expression in these two cell lines. Lack of a detectable wild-type p53 protein in Bal-17 was verified by Northern blot analysis of both Bal-17 and WEHI-231. Probing the same membrane used for the cell cycle Northern analysis of cyclin G1, shown in Fig. 3B, for p53 message clearly indicated that no p53 transcript, and thus no p53 protein, is present in Bal-17 cells (Fig. 3D, bottom panel). p53 function was confirmed in WEHI-231 by actinomycin D induced G 1 arrest (53). Predictably, Bal-17 cells under the same conditions were nonresponsive to DNA damage and showed no inhibition of S phase entry (data not shown). Taken together these results are in agreement with those of Okamoto and Beach (36) that the cyclin G1 gene in murine lymphocytes is a target for transcrip-tional activation by wild-type p53.
Survey of Cyclin G1 and p53 in Murine Tissues-Whereas p53 expression positively correlated with increased cyclin G1 transcript in proliferating murine B lymphocytes, its effect on the expression of cyclin G1 in other cells or tissue types, throughout development to terminal differentiation, was not known. The transcript level for cyclin G1 in various tissues from p53 null mice was compared to the level seen in the same tissues from heterozygous and wild-type p53 mice. When normalized to the GAPDH level in the tissues examined, only cells found in the kidney showed a positive correlation between wild-type p53 and cyclin G1 expression. Absence of wild-type p53 had little or no effect on the cumulative expression of cyclin G1 in stomach, brain, heart, or skeletal muscle cells (Fig. 4A,  top and bottom). When normalized to GAPDH levels (or total RNA loaded), cyclin G1 mRNA appeared to be increased upon the loss of wild-type p53 message in stomach and heart tissue.
To begin to investigate the relationship between p53 and cyclin G1 expression during embryogenesis, comparative Northern blot analysis was performed on RNA isolated from embryos of wild-type mice at progressive stages of development. Our initial comparison of 10-and 16-day embryos indicated cyclin G1 transcript increased during development in an inverse relationship to p53 expression (Fig. 4B, left panel). This result prompted examination of RNA from embryos covering a more extensive course of development (Fig. 4B, right panel). Day 7 is a highly proliferative stage when the duration of the cell cycle S and G 2 phases are condensed in certain regions of the mouse embryo and the differentiation commitment of some cell lineages begins with gastrulization and primitive streak formation (54, 55). Cyclin G1 message was very high relative to both GAPDH and p53 at day 7. Between day 7 and day 11 there appeared to be a switch to considerably reduced cyclin G1 levels relative to GAPDH and a coincident increased level of p53 message. At day 15, a developmental stage following a significant amount of organ system differentiation and morphogenesis (56), cyclin G1 message again increased relative to GAPDH while p53 levels remained the same or slightly decreased. Taken together, cyclin G1 and p53 mRNAs appear to be independently and differentially regulated during murine embryonic development.
To determine if the level of cyclin G1 mRNA could be modulated in cells not expressing the transcriptional activator p53, we tested the effect of several growth and stimulatory factors on cyclin G1 expression in the p53 null Bal-17 cells over time. The negative growth factor TGF-␤ is known to influence the expression of some cell cycle components (57,58) and has established kinetic effects on the cell cycle of murine B cells (59 -61). Exponentially growing cultures of Bal-17 (p53 Ϫ ) and WEHI-231 (p53 ϩ ) were treated with TGF-␤ at 1 ng/ml and aliquots were sampled for cyclin G1 mRNA expression over 24 h. PhosphorImager analysis of Northern blots of the cyclin G1 relative to GAPDH (control) mRNA (Fig. 4C) in Bal-17 indicated an 4.6-fold increase relative to untreated cells within 6 h of treatment followed by a decrease to nondetectable levels at 24 h, a point when growth inhibition is first noticed (data not shown) (61). Similar results were seen in WEHI-231 cells. Although the basal level of cyclin G1 is higher in WEHI-231, relative to the GAPDH control the initial level of cyclin G1 mRNA increased 2.5-fold at 6 h post-treatment followed by a similar decrease thereafter (Fig. 4C). Thus cyclin G1 mRNA expression can be modulated independent of p53 status in BAL-17 cells.
Isolation of a Full-length Cyclin G2 cDNA-Through automated search of GenEMBL we identified a 361-bp EST partial cDNA sequence (EMBL Z46163) with partial homology to cyclin G1. A 360-bp fragment of this sequence was amplified by PCR, cloned, and sequenced and used to probe a Jurkat cDNA library. We obtained 5 independent clones comprising two distinct cDNAs of 2.2 and ϳ1.1 kbp in length with 100% sequence identity to three contiguous regions of our probe. Analysis of these clones identified introns following exon junctions 5Ј and 3Ј to a region encoding a cyclin box motif. BLAST alignment of the assembled exon sequence with GenEMBL data base updates, identified 3 overlapping EST cDNAs sequences (Gen-EMBL: T65607, T65616, and T65464) previously identified in our automated BLASTX search for polypeptides with identity to the carboxyl terminus of human cyclin G1, with near-perfect nucleic acid identity to the 3Ј end of our sequence. This identity abruptly terminated, in-frame, at the 3Ј end of the predicted ORF, although the open reading frame of our assembled EST consensus sequence continued beyond this breakpoint. Despite mismatches and/or ambiguities among the EST cDNA sequences this alignment allowed us to conclude that the termination codon defining the end of our first predicted ORF was due to an in-frame stop at an intron/exon junction, and that the EST sequences had defined the spliced exons of the carboxyl terminus. The in-frame stop of the mature transcript was determined by RACE cloning of the 3Ј end using primers to the termini of our assembled sequences. Functional intron/exon junctions were demonstrated by PCR cloning of a mature fully spliced message. Resultant 3Ј-RACE clones of polyadenylated cDNA fragments overlapping the 3Ј end of the cyclin G2 ORF located an in-frame TAG translational stop codon 1032 bp 3Ј to the initiation codon. PCR cloning of the mature cyclin cDNA identified a fragment of the expected size and by Southern blot analysis was homologous to the exon sequences of our original FIG. 4. p53 independent cyclin G1 mRNA expression in tissues, during embryogenesis, and in p53-negative cells. A, top: Northern blot analysis of cyclin G1 transcript levels in the indicated tissues from p53 wild-type (ϩ/ϩ), heterozygous (ϩ/Ϫ), and gene knockout mice (Ϫ/Ϫ) compared to the GAPDH and wild-type p53 transcript levels (due to the apparent low level of p53 in muscle tissues only selected representative samples probed for p53 are shown). Bottom, quantitation of cyclin G1 mRNA levels relative to GAPDH levels in the tissues examined above. B, Northern blots of murine embryos examined for cyclin G1 mRNA expression relative to wild-type p53 and GAPDH. The stage, in days of embryonic development, is indicated above each lane. The RNA and blot on the left was prepared with 15 g of total RNA/lane while the blot on the right (Clontech) contains ϳ2.5 g of Poly(A) ϩ mRNA/lane. Blots were sequentially probed for cyclin G1, p53, and GAPDH. C, examination of cyclin G1 expression in Bal-17 and WEHI-231 cells treated over 24 h (time point above corresponding lane) with the growth inhibitor TGF-␤ and compared to GAPDH. PhosphorImager quantitation of cyclin G1 mRNA relative to GAPDH is shown below blots. clones. Double stranded sequence analysis of multiple, independent PCR clones demonstrated complete splicing of all predicted exons to form a 1035-bp ORF encoding a predicted 345 amino acid protein of ϳ40 kDa (Fig. 5A) with two potentially phosphorylated tyrosine residues, one in the cyclin box (62), and another in the carboxyl terminus. This new mature cDNA clone, encoding a cyclin homologue with 53% identity and 72% similarity to cyclin G1, is designated cyclin G2.
Comparison of Structural Features of Cyclin G1 and Cyclin G2-The recently resolved crystal structure of cyclin A (63) has enabled an alignment of cyclins G1 and G2 with cyclin A, revealing several features in their polypeptide sequences (Fig.  5B). The predicted NH 2 terminus of cyclin G1 from rat, human, and murine indicate they are nearly identical, with a single insertion (additional threonine residue at position 6 in the human sequence) and 4 to 12 amino acid exchanges, depending on the species, and no PEST or destruction box sequence motif. The sequence identity between cyclin A and cyclins G1 and G2 extends beyond the cyclin box to the NH 2 -terminal as well as COOH-terminal regions and remarkable stretches of amino acid identity throughout their sequences are consistent with the helical domains defined for cyclin A (63). Charged residues at position 266(Lys) and 295(Glu) in cyclin A proposed to be critical for CDK contact are conserved in cyclins G1 and G2 (positions 112 and 143 in Fig. 5B) and all other known cyclins (5), except cyclin F (28). The constrained alanine residues of proposed inter-helical crossing points in cyclin A (63) are conserved in the N-proximal ␣2 and ␣3 helices, but are replaced by the negatively charged residues glutamic and aspartic acid in the C-proximal ␣2Ј and ␣3Ј helices of cyclin G1 and G2 (Fig.  5B). These replacements, and extension of the interhelical regions in cyclin G1 and G2, suggest conformational distinction between A-type and G-type cyclins in the carboxyl region. An obvious difference between these two G cyclins is the increased length of the cyclin G2 COOH terminus. Importantly, drawing from the model of cyclin A (63) the carboxyl terminus of the G type cyclins may be spatially removed from the plane of CDK binding and subsequently available for other interaction. A 46-amino acid extension, including a PEST protein destabilization motif in the carboxyl terminus of cyclin G2, is missing in the cyclin G1 protein (Fig. 5B). A region in the carboxyl terminus, with high sequence identity between cyclin G1 and G2, shares sequence identity with a polypeptide sequence surrounding a phosphorylation site of the avian ErbB and human EGF receptor, a motif previously identified in cyclin G (35). Cyclin G2 (amino acids 272-294) exhibits higher sequence identity in this region to autophosphorylation sites in ErbB and EGF-R and also to a phosphorylation site in the polyoma virus FIG. 5. Nucleotide sequence of human cyclin G2 cDNA and alignment of the predicted G1 and G2 proteins to homologous cyclins. A, the cyclin G2 ORF and immediate 5Ј and 3Ј region and the predicted amino acid sequence shown below. The ATG (ϩ1) initiation codon is boxed, in-frame termination codons defining the ORF are shaded, and numbers to the right of the nucleotide sequence are relative to the start codon. The shaded boxed region in the protein sequence denotes the cyclin box-homologous region and the double outlined box in the carboxyl terminus indicates a PEST sequence. The potentially phosphorylated tyrosine residues are indicated by an asterisk above. B, amino acid alignment of rat human and mouse cyclin Gl and G2 proteins to partial human and mouse cyclin A, and Schizosaccharomyces pombe Cig1 and Cig2 proteins based on the crystal structure of human cyclin A (63). Identical amino acids are shaded and gaps introduced for optimal alignment are indicated by periods. Boxes define putative helical repeats and vertical boxes highlight the constrained alanine residues defining the interhelical crossing points (63). These are replaced by glutamic or aspartic acid in the C-proximal helices of cyclin G1 and G2. Asterisk (*) indicates conserved residues critical for cyclin A-CDK contact. middle T antigen. Collectively these autophosphorylation sites contain the newly described sequence motif N-X-X-Y (64) found to be functionally important for binding and interaction with the signaling protein Shc.
Cyclin G1 is a homologue of the fission yeast B-type Cig1 (35). Comparative "Bestfit" analysis (GCG) of the peptide sequences in the cyclin box region of cyclin G2 shows that while cyclin G2 has sequence identity with Cig1 (25%), it exhibits higher sequence identity to the fission yeast Cig2 and the budding yeast cyclin Clb-5. Amino acid identity and similarity of cyclin G2 to these yeast cyclins extends beyond the cyclin box region of cyclin G2. Overall identity (and similarity) of cyclin G2 to Cig2 is 25% (and 50%) and to CLb5 is 32% (and 49%), respectively, compared to 19% identity and 41% similarity to Cig1. Previous analysis indicated that Cig1 had considerable sequence identity to cyclin G1 in the carboxyl-terminal region (35), however, this analysis had not detected a region in the carboxyl terminus of both proteins with the consensus Comparison of Cyclin G1 and Cyclin G2 mRNA in Human Tissue and through the Cell Cycle-As the cyclin G1 and G2 encoded proteins exhibit such high sequence identity and similarity to each other relative to other cyclins (a situation analogous to the mammalian D-type and B-type cyclins, and the fission yeast cyclins Cig1, Cig2, and CDC13), we compared and contrasted the pattern of transcript expression of cyclin G2 to cyclin G1 in tissues and cell lines and examined their profile of expression throughout the cell cycle in human T-lymphocytes.
The same Northern blot membranes sequentially probed for human cyclin G1 (2.8 kb mRNA), cyclin G2 (2.9 kb), and a control gene, demonstrated obvious differences in their expression pattern. In contrast to cyclin G1, cyclin G2 was not prominently expressed in skeletal muscle but was most strongly expressed in the cerebellum. In addition, cyclin G2 mRNA levels are high in the spleen, thymus, and prostate relative to cyclin G1, relative to GAPDH and ␤-actin levels (Fig. 6A, left  panel). Examination of a variety of human B cell lines for expression of these two cyclins showed striking differences. While cyclin G1 was expressed at moderate to high levels (relative to GAPDH) in all of the cell lines examined, cyclin G2 mRNA was primarily elevated in Daudi cells (Fig. 6A, right  panel). Progressively lower levels of G2 transcript expression were detectable in Ramos, NALM-6, DG-75, and Raji cells and no significant level of G2 mRNA was observed for IM-9 or T-51 cells. It is important to note that the Burkitt's lymphoma cell lines Daudi, Ramos, and Raji harbor either homozygotic or heterozygotic mutant forms of the p53 allele (65)(66)(67) and that the level of p53 expression in these cell lines is independent of their Epstein-Barr virus status. To address the status of the p53 protein expressed in these cell lines we performed immunoprecipitations using antibodies recognizing either mutant or wild-type p53, and examined functional p53 activity by testing for G1 arrest after ␥-irradiation. Using this combined analysis the wild-type form of p53 was not detectable in lysates of the suspected mutants, nor were the p53 mutant cell lines sensitive to ␥-irradiation-induced G 1 arrest (data not shown). Clearly there is no correlation between the level of either cyclin G1 or cyclin G2 transcripts and this aspect of p53 function in these human cell lines.
Northern analysis of cyclin G1 expression in murine B cells had indicated expression was up-regulated in transition from early G 1 phase to late G 1 phase but remained at constitutively high levels throughout the rest of the cell cycle, similar to results described for cyclin G1 expression in fibroblasts following block and release experiments (35,36). To evaluate expression of cyclins G1 and G2 through the cell cycle of human lymphocytes, Jurkat T-lymphoblastoid and normal human peripheral T-lymphoblasts were fractionated by elutriation into populations at progressive phases of the cell cycle. Total RNA was extracted and examined by sequential Northern blot analysis (Fig. 6B). Cyclin G1 mRNA levels showed moderate fluctuation during the cell cycle in Jurkat cells, increasing slightly in early G 1 (2-fold) to peak levels in late G 1 /early S phase and declining to early G 1 levels by late G 2 /M (Fig. 6C). In normal stimulated T lymphoblasts, cyclin G1 expression was relatively constant following an initial decline from an early G 1 phase peak. This pattern of relatively constant cyclin G1 expression was contrasted by a clear oscillation in cyclin G2 transcript levels through the cell cycle. In both Jurkat T cells and normal T-lymphoblasts, cyclin G2 mRNA steadily increased to peak levels in the mid-S phase and decreased during G 2 /M phase progression. Thus cyclin G1 and G2 transcript were distinguished not only by the distribution and amount in different tissues and cell types, but also by regulation of their expression through the cell division cycle. DISCUSSION Cyclins thus far described are characterized by a conserved cyclin box region of ϳ110 amino acids surrounded by unique NH 2 -terminal and COOH-terminal sequences. Within subsets of cyclins, these flanking sequences also contain either destruction box or PEST motifs regulating protein stability. Outside of the cyclin box region, extensive sequence identity (Ͼ40%) is not present between the cyclins of different families, and within families such as the B-type mitotic cyclins or D-type family of G 1 cyclins, the overall amino acid identity is between 45 and 70%. We have described a new cyclin G1 homologue, cyclin G2, with 60% nucleotide sequence identity and 53% protein identity (72% similarity) to the cyclin G1. Cyclin G2 cDNA predicts for a protein of ϳ40 kDa as compared to 34 kDa for cyclin G1, and alignment of the deduced amino acid sequence from cyclin G2 and G1 ORFs demonstrates significant sequence identity NH 2 -terminal as well as COOH-terminal to the cyclin box. Analogous to the D-type cyclins (2,21), the murine, rat, and human cyclin G1 proteins are nearly identical (Ͼ95%), while human G1 and G2 (and murine G1 and G2 homologues) 2 exhibit a lower overall identity, suggesting that these two cyclins represent non-redundant molecules which play discrete functional roles.
Examination and comparison of our sequence for human and mouse cDNAs encoding cyclin G1 with those previously published (35,36,48) has identified a new translation initiation codon that includes a flanking region NH 2 -terminal to the cyclin box. This is consistent with the recently revised coding region for rat cyclin G (37). Moreover, analysis of cyclin A-CDK2 structure has demonstrated the ␣ helical region NH 2terminal to the cyclin box is requisite for cyclin-CDK interaction (63). Analysis of cyclins G1 and G2 in relation to the recently resolved crystal structure of cyclin A (63), indicates that alanine residues defining the crossover points in the ␣2 of the cyclin box for cyclin A are conserved in Cig1, Cig2, cyclin G1, and cyclin G2, but not among the B cyclins of higher eukaryotes (5), while the alanine in ␣3 helix, defining the cyclin fold, are conserved among all known cyclins except cyclin F. However, the cyclin fold defining alanines in the carboxyl proximal ␣2Ј and ␣3Ј helices of cyclin A are replaced by glutamic and aspartic acid residues in cyclin G1 and G2, and the putative inter-helical hinge regions are extended. Based on this alignment the NH 2 terminus of cyclin G1 and G2 are spatially close to and important for the cyclin-CDK interface and the COOH terminus is spatially removed to the CDK binding site, and may be available for either substrate or localizing interactions. Cyclins G1 and G2 have the highest sequence identity and similarity to the mammalian cyclin A proteins and the fission yeast B-type cyclins Cig1 and Cig2, respectively, while cyclin G2 exhibits high sequence identity to the S-phase cyclin Clb-5 of budding yeast.
Murine and human cyclin G1 mRNA have transcript sizes ϳ3.4 and 2.8 kb, respectively, in agreement with the work previously published (36,48). Parallel analysis of the human cyclin G2 mRNA defines an ϳ2.9-kb transcript. Our results indicate no obvious correlation between expression of these cyclins and the proliferative state of the tissue. Cyclin G1 is expressed at high levels in skeletal muscle, ovary, kidney, and colon. Cyclin G2, in contrast, is expressed at high levels in cerebellum, thymus, spleen, and prostate, and at low levels in skeletal muscle. While it is possible the level of mRNA detected may reflect a quantity of cycling cells in each tissue, several of these tissues are primarily composed of terminally differentiated cells and elevated levels of cyclin G1 in heart muscle, a tissue lacking self-renewing stem cells (68), make a proliferative component unlikely. Cyclin D3 is also increased upon terminal differentiation (69). Previous studies using serum starvation or metabolic blocks to synchronize cells suggested expression of human and rat cyclin G1 mRNA was either induced in G 1 phase to constitutively high levels throughout the cell cycle, or was unchanged as cells progressed from quiescence through the proliferative cell cycle. To clarify this we have examined cyclin G1 and G2 expression in the progressive stages of the cell cycle by partitioning proliferating cells with the less invasive method of centrifugal elutriation. Under these conditions cyclin G1 mRNA levels do not change significantly following an apparent induction in early G1. In contrast, cyclin G2 transcripts do exhibit cell cycle periodicity, reaching peak levels at mid-S phase. Similar to cyclin G1, transcripts of the mammalian cyclins C and D1 peak early in transition to the G 1 phase of the cell cycle following growth factor induction and oscillate only minimally throughout the proliferative cell cycle.
While the deduced amino acid sequence from the newly defined murine, rat, and human cyclin G1 ORFs exhibit neither a canonical destruction box nor PEST sequence, the predicted human cyclin G2 sequence contains a PEST-rich protein destabilizing sequence in its carboxyl terminus. The lack of destabilization motifs in cyclin G1, or any periodicity in its message during the cell cycle, suggests control of its activity may occur at the post-transcriptional or post-translational level. In contrast, both cyclin G2 and the yeast homologue, Cig2, show cell cycle dependent periodicity in their mRNA levels. Moreover, both of these contain protein destabilization motifs, Cig2 a destruction box, and cyclin G2 a PEST-rich sequence, making it likely that the expression of these cyclins is tightly regulated through the cell cycle.
Prior reports (36,37) identified murine and rat cyclin G1 as a transcriptional target of the p53 tumor suppressor. In agreement with these results, cyclin G1 levels in proliferating murine lymphocytes were paralleled by the presence or absence of wild-type p53 expression. In contrast to high cyclin G1 levels in WEHI-231 cells (determined to express wild-type p53), 6-fold lower levels of cyclin G1 are present in the p53 negative Bal-17 cells. p53 expression in many cell lines has been linked to the induction of cell cycle arrest and apoptosis (38,70), and it was proposed that the induction of cyclin G1 by p53 may be part of the pathway leading to arrest and or apoptosis (36). WEHI-231 cells undergo cell cycle arrest followed by apoptosis upon crosslinking of surface IgM, while Bal-17 cells continue to proliferate upon stimulation of its B cell receptor (52), thus it is tempting to link wild-type p53 expression and induction of cyclin G1 in WEHI-231 to cell cycle arrest and apoptosis. Establishment of p53 and cyclin G1 expression constructs into Bal-17 should indicate if there is a causal relationship between their expression and programmed cell death in murine B cell development.
Our further analysis shows the relationship between cyclin G1 transcript levels and expression of wild-type p53 mRNA is complex and contingent on growth conditions, tissue type, and stage of embryonic development. During embryonic development, and in tissues from p53 null mice, cyclin G1 transcript levels are often found inverse to or unrelated to p53 levels. Elevated levels of cyclin G1 and suppressed p53 levels are seen at day 7 (gastrulation) of mouse development. Cell cycle regulation during gastrulation differs from that at other stages of development (54) and accelerated cell division is punctuated by contraction of S and G 2 phases while the typically variable G 1 phase is relatively unchanged (55). Transcript levels of cyclin G1 in tissues from wild-type, heterozygous, and p53 null mice were reduced in kidney, but tissues such as brain, stomach, and testis showed no reduction in the amount of cyclin G1 mRNA upon loss of functional p53, and in some cases message was increased upon p53 loss. Lack of correlation between cyclin G1 expression and p53 is also evident in our Northern blot analysis of cyclin G1 message in Ramos, Daudi, Raji, and Jurkat cells and the report of abundant cyclin G1 mRNA in SAOS-2 fibroblasts (48), all previously determined to be p53 deficient cell lines (65-67, 71). p53 mutant alleles may be diverse in their ability to transactivate, yet in two of the above mentioned cell lines p53 is most likely not acting as the direct transcriptional 2 M. C. Horne, unpublished data. activator of cyclin G1. SAOS-2 fibroblasts completely lack endogenous p53 (72), yet maintain a high level of cyclin G1 message (48); and in Ramos cells the expression of another direct transcriptional target of p53, the CDK inhibitor p21 Waf1/ Cip1, is not induced by ␥-irradiation treatment that up-regulates p53 and p21 Waf1/Cip1 in p53 wild-type cells (70), indicating that the p53 protein expressed at high levels in this cell line is not a functional transcriptional activator. Transcriptional activators, in addition to p53, are likely to be involved in regulating cyclin G1 expression. The expression of p21 Waf1/Cip1 has been reported to be activated in some cell lines treated with growth inhibitory factors (70,73,74), independent of the p53 status of the cell. It has been recently shown that the transactivator NF-B binds and activates transcription through the p53 response element (75), raising the possibility that cyclin G1 transcription may also be activated by others factors such as NF-B.
It is not clear if cyclin G1 expression is linked to positive or negative growth regulation. Cyclin G1 is induced in both p53 negative and p53 wild-type murine B cells lines by treatment with the negative growth factor TGF-␤. This TGF-␤ response occurs 18 h prior to any inhibition of cell growth and more than 66 h before significant G 1 phase accumulation of TGF-␤ treated WEHI-231 cells (61). It is possible that factors affecting mRNA stability may also play a role in regulating cyclin G1 expression. TGF-␤ related stabilization of mRNAs has been reported and found to be affected by TGF-␤ responsive elements present in the 3Ј-UTR of these sequences (76 -78). The 3Ј-UTR of several lymphokine and proto-oncogene mRNAs are known to contain AU-rich elements with the motif AUUUA acting as instability determinants (49 -51, 79). The conserved block of nucleotide sequence containing reiterated copies of the canonical AUUUA motif present in the 3Ј-UTRs of human, rat, and murine cyclin G1 gene transcripts could reflect a region regulating cyclin G1 mRNA stability.
The function of these cyclins remains unclear. Although overexpression of cyclin G1 in human osteosarcoma cells or murine fibroblasts had no observable effect on the cell cycle progression of these cells (36), it is possible the NH 2 -terminal 45 amino acids missing in the cyclin G1 constructs are necessary for structural conformation. Based on up-regulation by p53, it has been proposed that high levels of cyclin G1 expression may inhibit the activity of one or more of the G 1 -type cyclins by competing for association with a partner catalytic CDK and thus acting as an anti-cyclin (36). In this regard, a possible competitive interaction between cyclin G2, cyclin G1, and a hypothetical CDK partner could be postulated. Cyclin G2 contains two potentially phosphorylated tyrosine residues suggesting a role for cyclin G2 in a signal transduction cascade and cyclin G1 may act as a competitive inhibitor. The presence of the N-X-X-Y recognition sequence for binding of the PTB domain of the Shc signaling protein (64,80) in the carboxyl terminus of cyclin G2 but not cyclin G1 may be biologically relevant.
These cyclins exhibit differential tissue distribution, cell cycle-regulated transcript expression, and unique structural features. Our recent detection of another homologue of cyclins G1 and G2 may expand this family to at least three members. 3 Further experiments to assess protein abundance, subcellular localization, and associations with other cell cycle component partners is required to shed further light on their function.