Identification of Interaction Partners and Substrates of the Cyclin A1-CDK2 Complex*□S

The CDK2-associated cyclin A1 is essential for spermatogenesis and contributes to leukemogenesis. The detailed molecular functions of cyclin A1 remain unclear, since the molecular networks involving cyclin A1-CDK2 have not been elucidated. Here, we identified novel cyclin A1/CDK2 interaction partners in a yeast triple-hybrid approach. Several novel proteins (INCA1, KARCA1, and PROCA1) as well as the known proteins GPS2 (Gprotein pathway suppressor 2), Ku70, receptor for activated protein kinase C1/guanine nucleotide-binding protein -2-like-1, and mRNA-binding motif protein 4 were identified as interaction partners. These proteins link the cyclin A1-CDK2 complex to diverse cellular processes such as DNA repair, signaling, and splicing. Interactions were confirmed by GST pull-down assays and co-immunoprecipitation. We cloned and characterized the most frequently isolated unknown gene, which we named INCA1 (inhibitor of CDK interacting with cyclin A1). The nuclear INCA1 protein is evolutionarily conserved and lacks homology to any known gene. This novel protein and two other interacting partners served as substrates for the cyclin A1-CDK2 kinase complex. Cyclin A1 and all interaction partners were highly expressed in testis with varying degrees of tissue specificity. The highest expression levels were observed at different time points during testis maturation, whereas expression levels in germ cell cancers and infertile testes decreased. Taken together, we identified testicular interaction partners of the cyclin A1-CDK2 complex and studied their expression pattern in normal organs, testis development, and testicular malignancies. Thereby, we establish a new basis for future functional analyses of cyclin A1. We provide evidence that the cyclin A1-CDK2 complex plays a role in several signaling pathways important for cell cycle control and meiosis.

establish a new basis for future functional analyses of cyclin A1. We provide evidence that the cyclin A1-CDK2 complex plays a role in several signaling pathways important for cell cycle control and meiosis.
Regulation of the cell cycle is one of the most complex features of eukaryotic cells. Cyclins are regarded as the major regulators of the cell cycle (1)(2)(3). Cyclin-dependent kinases provide the catalytic subunit of the active cyclin-CDK 1 complex (4 -6). Cyclins are involved in the regulation of the mitotic as well as the meiotic cell cycle (7,8), although the distinct regulatory mechanisms differ (9,10). Dysregulation of the cell cycle leads to abnormal cell growth and thus contributes to tumorigenesis (11)(12)(13)(14).
Two A-type cyclins are known until now to be involved in cell cycle regulation: cyclin A1 and cyclin A2. Cyclin A2, also known as cyclin A, is a key regulator of the cell cycle in mammalian cells. It is ubiquitously expressed and essential for progression through the cell cycle. Cyclin A2 is involved in both S phase and G 2 /M transition through its association with distinct CDKs (15,16). Several lines of evidence indicate its oncogenic potential (17)(18)(19). Cyclin A2 associates with CDK2 (20) at the onset of DNA replication in S phase (21) and with CDK1 mainly at the G 2 /M transition (15).
The second type of human cyclin A (22), named cyclin A1 (in homology to findings earlier in mouse and Xenopus), associates with CDK2 in vitro and in vivo but not with CDK1, in contrast to the interaction of cyclin A1 with CDK1 in mice. The CDK2cyclin A1 complex shows kinase activity on histone H1 (22).
Cyclin A1 expression is tissue-specific, and high levels of expression are restricted to testis in the healthy organism in humans (22), to eggs and early embryos in Xenopus (23), and to the germ line in mice (24). Cyclin A1 is expressed shortly before or during the first meiotic division in spermatogenesis (25), and male cyclin A1 knockout mice are infertile (26). Spermatogenesis is arrested prior to entry into metaphase I associated with inactive cyclin B-CDK1 complexes and therefore loss of Mphase factor activity (27). Cyclin A1 expression is also diminished in patients suffering from infertility (28). Cyclin A1 expression increases at the entry into S phase of previously synchronized leukemic cells (22). In G 2 /M phase, cyclin A1 expression and cyclin A1-CDK2 kinase activity reach their maximum levels, but cyclin A1 is detectable throughout the cell cycle in contrast to cyclin A2 (29). It interacts with the cell cycle regulators E2F and pRb and neutralizes the cell cycle inhibition by pRb in SAOS-2 cells (29), which indicates a tissuespecific role in mitosis. However, the expression throughout the cell cycle rules out a major regulatory role for cyclin A1 in the mitotic cell cycle. The promoter of cyclin A1 is dependent on four Sp1 transactivation sites in a CpG island upstream of the transcriptional start site (30).
Cyclin A1 is supposed to play a role in the pathogenesis of myeloid leukemia, since it is highly expressed in leukemias of myeloid origin (31). Upon induction of myeloid differentiation, cyclin A1 expression decreases (31). Overexpression of murine cyclin A1 in transgenic mice leads to abnormal myelopoiesis in the first months after birth as well as to the development of myeloid leukemia at a low frequency. This indicates that cyclin A1 alone is not sufficient to induce transformation but contributes to leukemogenesis (32).
The molecular functions of cyclin A1 in pathological and physiological settings have not been investigated in detail. The distinct functions of cyclin A1 in mitosis, meiosis, and malignant diseases remain currently unknown. In addition, potential roles of cyclin A1 in other cellular processes have not been characterized. The major drawback of functional analyses of cyclin A1 on the molecular level results from the lack of knowledge about interaction partners and substrates of cyclin A1.
The aim of this study was the identification of interaction partners of the cyclin A1-CDK2 complex in the testis. In a yeast triple-hybrid approach, we identified several potential interacting proteins of cyclin A1-CDK2 from a testis cDNA library. We confirmed the interaction with cyclin A1 by GST pull-down assays and analyzed the expression of the interaction partners in normal organs, in testis maturation, and in testicular malignancies. We have cloned a novel gene, INCA1 (inhibitor of CDK interacting with cyclin A1). INCA1 interacted with cyclin A1, was phosphorylated by the cyclin A1-CDK2 complex, and was tissue-specifically expressed in the testis. In summary, we characterized novel interaction partners of the cyclin A1-CDK2 complex that will guide the future functional analysis of cyclin A1.

EXPERIMENTAL PROCEDURES
Yeast Triple-hybrid System-To identify cyclin A1-interacting proteins, a yeast triple-hybrid (Y3H) screen was performed using the Matchmaker™ Gal4 two-hybrid system 3 (Clontech). All experiments were carried out according to the recommendations of the supplier (Clontech, Yeast Protocols Handbook, Matchmaker™ two-hybrid system 3).
In brief, a human testis cDNA library was cloned into the pACT2-AD vector (Clontech). Cyclin A1 served as bait for the library translation products. Cyclin A1 and CDK2 were both cloned into pBridge. The plasmids were co-transformed into the yeast strain AH109 (Clontech). The number of screened clones was calculated to be 1 ϫ 10 7 clones. Transformed yeast cells were grown on high stringency selection plates (ϪMet, ϪLeu, ϪTrp, ϪHis, ϪAde, ϩ5-bromo-4-chloro-3-indolyl-␤-D-galactopyranoside (ϩX-gal)). The cDNA inserts in the positive yeast colonies were amplified by nested PCR. Sequences were analyzed by alignment to the NCBI data bases. For control purposes, we performed a conventional yeast two-hybrid screen in parallel with cyclin A1 as bait but without additional co-expression of human CDK2. This approach helped to exclude the identification of proteins binding to cyclin A1 only and not to the active cyclin A1-CDK2 complex. Only sequences identified in more than one yeast colony in the Y3H and not present in the yeast two-hybrid control screen are presented here and were used for further investigations.
GST Fusion Proteins and GST Pull-down Assays-The interaction partners (INCA1 and GPS2 (G-protein pathway suppressor 2), fulllength coding sequence; all others, longest clone isolated from the Y3H) were cloned in frame into the pGEX-5X-2 plasmid (Amersham Bio-sciences) for expression of GST fusion proteins. GST fusion protein was expressed in Escherichia coli BL21-DE3, and purification was carried out according to the manufacturer's recommendations (GST Gene Fusion System; Amersham Biosciences) using glutathione-agarose beads (Sigma). To control the preparations for equal concentrations and protein degradation, 10 l of the slurry were run on an SDS-PAGE gel and stained with Coomassie Brilliant Blue.
For GST pull-down assays, 1 g of GST protein on glutathione beads was washed once in 1ϫ binding buffer (50 mM Tris-HCl (pH 7.5), 1.0% Nonidet P-40, 400 mM NaCl, 1 mM dithiothreitol) and resuspended in 100 l of binding buffer in siliconized tubes. 4 l of in vitro transcribed and translated cyclin A1, which was radioactively labeled with [ 35 S]methionine with the TNT QuickCoupled Transcription/Translation System (Promega, Madison, WI) or lysates of baculovirus-infected Sf9 cells expressing cyclin A1 were added, and the reaction mix was incubated for 1 h at 4°C. After washing with binding buffer and SDS-PAGE, the gel was dried and analyzed by autoradiography or Western blotting for cyclin A1, respectively.
In Vitro Kinase Reactions-For in vitro kinase assays, GST fusion proteins were incubated with cell lysates of baculovirus-infected Sf9 cells. 5 Ci of [␥-32 P]ATP (PerkinElmer Life Sciences) were added to 15 l of GST fusion beads (50% slurry, equal concentrations controlled by Coomassie staining) and 6 g of insect cell lysate and incubated for 30 min in 1ϫ kinase buffer (10 M ATP, 50 mM Hepes (pH 7.5), 1 mM dithiothreitol, 10 mM MgCl 2 , 0.1 mM Na 3 VO 4 , 1 mM NaF). After washing and SDS-PAGE, phosphorylation of INCA1 was detected by autoradiography. Site-directed mutagenesis of potential phosphorylation sites to alanine was carried out using the QuikChange TM strategy (Stratagene, La Jolla, CA) following standard protocols (for primer, see Supplementary Table I).
Real Time Quantitative RT-PCR and Northern Blot-For expression analysis in normal organs, a commercially available cDNA library containing pooled human cDNA from various organs was used (Clontech). For expression analysis in malignant tissues, total RNA was isolated from fresh frozen tissue samples using TRIzol (Invitrogen) according to the manufacturer's recommendations. Twenty g of total RNA were used for Northern blot hybridization (31) and probed with the 32 Plabeled cDNA of human INCA1.
For RT-PCR, 1 g of total RNA was reverse transcribed using random primers and Moloney murine leukemia virus reverse transcriptase (Promega) following the manufacturer's protocol. cDNA samples were diluted to 100 l, and 2.5 l of cDNA were used for each PCR. PCR amplification of the housekeeping gene glyceraldehyde-3-phosphate dehydrogenase was used to confirm the quality of cDNA and standardize the total amount of cDNA between different samples. The quantitation of mRNA expression levels was carried out using a real time fluorescence detection method based on TaqMan technology (PerkinElmer Life Sciences) (33,34). All primer and probe combinations (see Supplementary Table I) were designed to span an exon-exon junction to avoid amplification of genomic DNA. The probes were labeled at the 5Ј-end with the fluorescent dye 6-carboxy-fluorescein (cyclin A1 and INCA1) or 6-carboxy-rhodamine 6G (glyceraldehyde-3-phosphate dehydrogenase) and at the 3Ј-end with the quencher 6-carboxytetramethylrhodamine. For the other interacting clones, specific primer pairs were designed, and the SYBRGreen detection method was used. At least two independent analyses were performed for each sample. The expression data regarding testis maturation were extracted from published primary microarray data (35).
Statistical analyses were carried out with SPSS 11.0 for Windows. For correlation analyses, the nonparametric Kendall Tau correlation analysis was used. To identify significant differences in gene expression, the nonparametric Mann-Whitney U test and Kruskal-Wallis test were performed. All p values indicate two-sided comparisons, and p Ͻ 0.05 was considered as significant.
5Ј-Rapid Amplification of cDNA Ends for INCA1-Transcribed fragments of the genomic clone hRPC.1050_D_4 had been isolated from six colonies. The encoded gene was cloned by 5Ј-rapid amplification of cDNA ends (Invitrogen). In addition, the murine homolog of INCA1 was identified by homology searches and 5Ј-rapid amplification of cDNA ends. The GenBank TM accession numbers for INCA1, KARCA1, and PROCA1 are given in Supplementary Table II. For the analysis of the nucleotide and the amino acid sequence of INCA1, several World Wide Web-based resources were used. Detailed information about the software and prediction results are listed in Supplementary Table III. The discovered cDNA sequence was PCR-amplified from human testis cDNA and cloned into different vectors. The vector pcDNA3.1(ϩ) (Invitrogen) was used as an expression vector in mammalian cells in transfection experiments. In addition, INCA1 was cloned into pcDNA3.1(ϩ) fused to enhanced green fluorescent protein (EGFP) (Clontech) or fused to a Myc tag. Other vectors and constructs used are described in the corresponding sections under "Experimental Procedures." Where applicable, cyclin A1 was expressed in pcDNA3.1(ϩ) (33).
Promoter Activity and Luciferase Assay-Following PCR amplification of the sequence upstream of the human INCA1 gene, it was cloned into the pGL3 basic vector in promoter position to the firefly luciferase gene. Luciferase assays for promoter activity were carried out essentially as described (30,34) using the Dual-Luciferase reporter assay system (Promega, Madison, WI). Briefly, COS-7 or S2 cells were transfected with the promoter constructs in pGL3 vectors and pRL-CMV for standardization purposes. The promoter activity was determined as the ratio of firefly luciferase luminescence divided by Renilla luciferase activity. All experiments were carried out independently three times, and data are indicated as mean with S.E.
For localization analysis of INCA1, EGFP-INCA1 or EGFP alone were transfected into COS-7 cells, stained with 1 g/ml DAPI for DNA and 2.5 g/ml TRITC-phalloidin (both from Sigma) for actin, and analyzed by conventional fluorescence or confocal microscopy.
For co-immunoprecipitation, 500 l of lysate of transfected cells were rotated with 2 g of ␣-Myc antibody at 4°C for 1 h, and then 60 l of protein A/G-agarose beads (Santa Cruz Biotechnology) were added, and the reaction mix was rotated at 4°C again for 2 h. After washing with ice-cold radioimmune precipitation, the bound proteins were subjected to SDS-PAGE and blotted for EGFP-cyclin A1.
For co-immunoprecipitation experiments with endogenous proteins, ML-1 cells, which highly express cyclin A1 (22), were starved for 24 h at 0.1% fetal calf serum to induce expression of INCA1 and lysed in Nonidet P-40 lysis buffer. A total of 300 g of protein were precleared with rabbit preimmune serum and then incubated at 4°C overnight with either preimmune serum or anti-INCA1 serum. Agarose beads coupled to protein A/G (Santa Cruz Biotechnology, Inc., Santa Cruz, CA) were used for precipitation. Beads were washed and subjected to SDS-PAGE under nonreducing conditions. The Western blot was then detected with anti-cyclin A1 antibodies (Pharmingen). A nonspecific band at about 90 kDa was used to demonstrate equal loading of both samples.

Identification of Cyclin A1/CDK2 Interaction Partners-
The molecular function of cyclin A1 remains to be elucidated due to a lack of information about interactions on the molecular level. To identify cyclin A1 interaction partners, we screened a human testis cDNA library for interacting proteins. A Y3H with CDK2 co-expression was performed (Fig. 1A). The Y3H should mimic best the physiological conditions of the active kinase complex of cyclin A1 and CDK2. To minimize the probability of false positive clones, we only report sequences here that were found in at least two independent clones. We identified eight sequences for putative cyclin A1 interaction partners in more than one yeast clone: four known genes, two expressed sequence tags (ESTs), and two unknown cDNAs (Fig. 1B).
Four Known Genes Interacted with the Cyclin A1-CDK2 Complex-The G-protein pathway suppressor 2 (GPS2; accession number U28963, identical to AMF1) was identified in four clones. GPS2 has been assigned to different functions in Gprotein, Ras, mitogen-activated protein kinase, and p53 signaling and is part of the nuclear NCOR1-HDAC3 complex that inhibits c-Jun N-terminal kinase activation. We found in two clones the nuclear protein Ku70 (accession number NM_001469), which binds double-stranded DNA and plays an essential role in DNA repair. GNB2L1 (guanine nucleotidebinding protein ␤-2-like-1; accession number NM_006098), which is identical to RACK1 (receptor for activated protein FIG. 1. Yeast triple-hybrid system for cyclin A1/CDK2 interaction partners. A, Y3H was used to identify interaction partners of the cyclin A1-CDK2 complex from a testis cDNA library. Cyclin A1 served as bait fused to the GAL4-DNA-binding domain (BD), CDK2 was coexpressed from the same vector, and the testis cDNA library was fused to the GAL4-activation domain (AD). Yeast growth only occurred upon interaction of cyclin A1, CDK2, and a library cDNA. B, eight genes were identified in more than a single yeast clone. Three of them encoded novel proteins named INCA1, KARCA1, and PROCA1. kinase C1), is a homolog of the ␤-subunit of heterotrimeric G proteins. RBM4 (mRNA-binding motif protein 4; accession number NM_002896) is the human homolog to the Drosophila lark gene that is essential for embryonic development. Human RBM4 modulates alternative pre-mRNA splicing and antagonizes SR protein function. In two clones, sequences of the mRNA DKFZ p686G052 (accession number AL832200, abbreviated here as DKFZ) were enclosed. In a data base, this sequence was found in 157 ESTs (UniGene Hs.317304) from several tissues including acute myeloid leukemia. In 69 ESTs from normal organs, 13 sequences derived from testis. The mRNA encodes a hypothetical protein of 1686 aa of unknown function.
Identification of Three Novel Proteins Interacting with Cyclin A1-In six clones, we found fragments of a novel mRNA that was not part of any human cDNA or EST sequence but was present in a genomic clone (accession number AC004771). Since this transcript was detected most often in our Y3H assay, we cloned and characterized it in more detail (see below; Figs. 7-9). The sequence was unknown before and did not contain any homologies or conserved motifs. Additional functional analyses revealed an inhibitory effect on CDK activity. 2 Therefore, we named it INCA1 (inhibitor of CDK interacting with cyclin A1).
In addition, we identified sequences from another mRNA (MGC 33338; accession number BC022077) in two Y3H clones. The published mRNA sequence of 1328 nucleotides (nt) contains a small open reading frame (ORF) encoding only 111 aa. In our sequences, we found an insertion of 19 nt in the middle of the sequence giving rise to a novel ORF coding for 288 aa using a different reading frame than the hypothetical protein of 111 aa. The larger protein (31 kDa) was in the same reading frame as the GAL4 activation domain in the pACT2 vector used for the Y3H assay, confirming this new reading frame. The novel full-length mRNA contains 1347 nt with a coding sequence of 867 nt. The predicted protein contains amino acid sequences with homologies to two kelch motifs, two ankyrin repeats, and a galactose oxidase domain (InterProScan); therefore, we named it KARCA1 (kelch/ankyrin repeat-containing cyclin A1-interacting protein).
In two other clones, we found a novel mRNA sequence that was only partially similar to an EST (EST zt86e02; accession number AA398001). This novel sequence does not contain a stop codon but comprises an ORF over its full-length of 399 nt encoding a protein of 133 aa. The gene is located on chromosome 17 and transcribed from the genomic sequence (accession number AC010761) in three exons all adhering to the GT/AG rule for intron start and end sequences. This novel protein was predicted to be localized in the nucleus and contained a prolinerich region from aa residues 4 to 60 that could be involved in protein-protein interaction. We assume that we identified a partial coding sequence of a novel protein that we called PROCA1 (proline-rich cyclin A1-interacting protein).
A Novel Gene, INCA1, Encoded a 221-aa Protein and Localized to Chromosome 17-We identified different fragments of one cDNA in six yeast clones. The sequences corresponded to a genomic clone (clone hRPC.1050_D_4, GenBank TM accession number AC004771.1). This gene was not described before and did not exhibit any homologies to other known genes (see below). Due to further functional experiments, 2 we named it INCA1. We performed 5Ј-rapid amplification of cDNA ends with the unknown cDNA sequence of INCA1 and isolated a mRNA sequence of 1383 nt including an ORF of 221 aa ( Fig.   2A). Human INCA1 consisted of eight exons with the coding sequence located from exon 3 to 8. The second exon was alternatively spliced, leading to an mRNA of only 1221 nt, which did not alter the ORF. We also identified the murine homolog of INCA1. Murine INCA1 included seven exons with 1218 nt. The coding sequence was located from exon 2 to 7 and encoded a protein sequence of 231 aa. The first murine exon could be alternatively spliced, giving rise to four splice variants of 1080, 1150, 1189, and 1218 nt, which all encode the same protein.
Human INCA1 localized to chromosome 17p13, the murine homolog mapped to chromosome 11B. The primary nucleotide and amino acid sequences of human and murine INCA1 were analyzed with several World Wide Web-based bioinformatic tools (Supplementary Table III). For the human INCA1 protein, a molecular mass of the unmodified protein of 25.2 kDa was predicted, and for the murine homolog, a mass of 26.5 kDa was predicted. The human and the murine INCA1 coding sequences showed 70.8% homology on the nucleotide and 57.4% identity on the amino acid level. The localization of both proteins was predicted to be nuclear. Both sequences contained multiple consensus sites for phosphorylation and other posttranslational modifications. Upstream of the human INCA1 gene, we found a functional promoter sequence for INCA1 (compare Fig. 9).
The nucleotide and amino acid sequences of the three novel genes INCA1, KARCA1, and PROCA1 are presented in Fig. 2 and have been deposited into GenBank TM (accession numbers in Supplementary Table II).
Interaction with Cyclin A1 and Phosphorylation by the Cyclin A1-CDK2 Complex-The eight putative interacting proteins and their fragments isolated in the Y3H are given in Fig. 3A. All of these proteins contained several SP or TP motifs, which provide the minimal consensus motif for CDK phosphorylation, and four of them contained the full consensus motif for CDK phosphorylation (S/T)PX(K/R) (Fig. 3A).
The interaction with cyclin A1 was confirmed for all eight proteins in vitro by GST pull-down assays. Recombinant radioactively labeled cyclin A1 was incubated with GST fusion proteins of the interaction partners. Cyclin A1 was not pulled down with GST alone but rather with all other GST fusion proteins (Fig. 3B).
The analysis of the interaction with cyclin A1 was hindered by the lack of available antibodies for the interacting proteins. We cloned mammalian expression vectors for three interaction partners (INCA1, GPS2, and Ku70) and confirmed also the interaction in vivo for these. In co-immunoprecipitation assays, antibodies against Myc-tagged interaction partners INCA1, GPS2, and Ku70 also precipitated cyclin A1, whereas unspecific IgG or Myc alone did not (GPS2) (Fig. 3C). The interaction with INCA1 was studied in more detail and is presented in Fig.  7. The interaction with Ku70 was further analyzed and will be published elsewhere. 3 To examine whether the cyclin A1-CDK2 complex could phosphorylate its interaction partners, we carried out in vitro kinase assays with the GST fusion proteins. The fusion protein or GST alone bound to glutathione-agarose beads was incubated in the presence of [␥-32 P]ATP with lysates of baculovirusinfected Sf9 insect cells expressing either cyclin A1 and CDK2 together (Fig. 3D) or cyclin A1 or CDK2 alone (Fig. 3E). GST alone was not phosphorylated by cyclin A1-CDK2. Cyclin A1-CDK2 phosphorylated INCA1, Ku70, and RBM4 (Fig. 3D). This  FIG. 2. Sequences of the novel proteins INCA1, KARCA1, and PROCA1. A, the mRNA of human INCA1 encoded a 221-aa protein that was not homologous to any other known protein. The mRNA consisted of eight exons of which the second one was alternatively spliced leading to mRNAs of 1383 and 1221 nt in length, respectively (nuclear localization signal (NLS)). B, KARCA1 was a novel protein of 288 aa that showed homology to kelch motifs and ankyrin repeats. C, PROCA1 encoded a novel amino acid sequence interacting with cyclin A1 in our Y3H assay. The identified clone spanned 133 aa and was in frame to the GAL4-AD. Since the protein sequence did not start with a methionine and did not contain any STOP codon, we assume the full-length protein to be larger and therefore called this sequence the partial coding sequence. The N-terminal part of PROCA1 contained a proline-rich region.

FIG. 3. Association of cyclin A1 with its novel interaction partners and substrates.
A, the open reading frames of the identified cyclin A1-interacting proteins and the parts of these proteins identified in the Y3H system that constitute potential interaction domains. In addition, minimal and full consensus motifs for CDK phosphorylation are displayed. B, in GST pull-down assays, we proved interaction of all identified potential interaction partners with cyclin A1 in vitro. GST alone or GST fused to one of the interacting proteins was incubated with in vitro transcribed/translated radioactively labeled cyclin A1. C, co-immunoprecipitation of Myc-GPS2 with cyclin A1 in lysates from transfected COS-7 cells confirmed their interaction in vivo. D, in in vitro kinase assays, GST fusion proteins were incubated with lysates from baculovirally infected Sf9 cells expressing cyclin A1 and CDK2 in the presence of [␥-32 P]ATP. E, INCA1, Ku70, and RBM4 were phosphorylated by the cyclin A1-CDK2 complex. To control the specificity of their phosphorylation, GST fusion proteins were incubated with Sf9 lysates from cells infected with empty vector, cyclin A1 alone, or CDK2 alone. phosphorylation was specific for the cyclin A1-CDK2 complex, since it was not seen for cyclin A1 or CDK2 alone (Fig. 3E). For RACK1/GNB2L1, a faint band was observed in some experiments but was not considered significant.
Cyclin A1 Interaction Partners Are Expressed in the Testis-So far, the main identified cellular function of cyclin A1 is its involvement in meiosis (26). In normal organs, cyclin A1 is tissue-specifically expressed in the testis (22) (Fig. 4A).
The expression patterns of cyclin A1 and its interaction partners were analyzed in a panel of 16 normal organs by real time quantitative RT-PCR (Fig. 4B). The amount of cDNA in each sample was standardized using the expression level of the housekeeping gene glyceraldehyde-3-phosphate dehydrogenase. Gene expression was normalized to the expression in testis (testis expression ϭ 1) for each gene.
All genes showed high expression levels in the testis. Some genes (cyclin A1 (Fig. 4A), INCA1 (Fig. 4B), and PROCA1 (Fig.  4I)) were almost exclusively expressed in the testis with low levels of expression in other organs, e.g. in the ovary. Other genes (GPS2 (Fig. 4C), Ku70 (Fig. 4D), RBM4, (Fig. 4F), DKFZ (Fig. 4G), and KARCA1 (Fig. 4H)) were highly expressed in the testis and also showed high expression levels in some other organs. Only one gene (RACK1/GNB2L1) (Fig. 4E) was expressed at an intermediate level in the testis and higher in eight other organs. The predominant expression in the testis is obvious in a logarithmic bar graph standardized to the testicular expression levels (Fig. 4J) in which all genes including cyclin A1 demonstrate lower expression (below the x axis) in other organs than in testis except RACK1/GNB2L1. Cyclin A1 and Its Interaction Partners Are Regulated during Testis Maturation and Spermatogenesis-Recently, a microarray study analyzing gene regulation throughout postnatal testis development has been published (35), and the expression data are available through the World Wide Web. Murine homologs of cyclin A1 and seven of the eight interacting proteins were included in the oligonucleotide microarray. Our novel mRNA PROCA1 was not found on the array. We extracted primary microarray data for the other genes of interest. The expression data were normalized to the maximal expression of each gene to facilitate the direct comparison of expression regulation throughout testis development for different genes (Fig. 5A).
Cyclin A1 was strongly induced in the maturing testis with highest expression from day 26 to the adult stage. Although not all other genes reached their maximal expression level at this stage, they all were substantially expressed in adult testis, corroborating their potential functional interaction in the testis.
In the microarray studies, we detected in the testis earlier expression of INCA1 than of cyclin A1. This finding was confirmed by expression studies in ATM Ϫ/Ϫ mice that have a meiotic block at prophase I (36). These mice still express INCA1 in the testis but do not express cyclin A1 any more. 4 Decreased Expression of Cyclin A1-interacting Proteins in Testis Tumors and Infertile Tissue-In testis samples from patients suffering from infertility, cyclin A1 expression was lost in germ cell aplasia, and its expression increased with progression of spermatogenesis (28). We studied the expression levels of cyclin A1 and its interaction partners in normal testis (n ϭ 6) and in biopsies from infertile patients (n ϭ 35) with histologically different stages of maturation arrest. Compared with normal testis, the average expression of all analyzed genes was 2-4-fold decreased in the infertility testis samples (Fig. 5B). The differences between normal and infertile testis were significant for cyclin A1, GPS2, Ku70, RACK1/GNB2L1, RBM4, and KARCA1. Expression levels of all genes except RACK1/ GNB2L1 significantly correlated with the expression of cyclin A1.
Cyclin A1 was expressed in aggressive germ cell tumors (e.g. immature teratoma or embryonal cell carcinoma), whereas its expression was lost in more differentiated tumor types as mature teratoma (37). We analyzed the expression of cyclin A1 and its interaction partners in normal testis (n ϭ 7) and in testis tumors (n ϭ 29). Cyclin A1 expression was decreased in malignant testis tissue compared with normal testis. Concurrently, expression of INCA1, GPS2, Ku70, RBM4, DKFZ, KARCA1, and PROCA1 was significantly reduced (Fig. 6A). Only RACK1 was more highly expressed in testis tumors than in normal testis. In correlation analysis, including all patient samples, the expression levels of cyclin A1 correlated significantly (p Ͻ 0.05) with the expression levels of GPS2, Ku70, RBM4, DKFZ, and PROCA1. We differentiated the testis tumor samples into histological subgroups (Fig. 6, B-J; embryonal carcinoma (EC, n ϭ 6), immature teratoma (iT, n ϭ 4), yolk sac tumor (YS, n ϭ 3), seminoma (S, n ϭ 7), and mature teratoma (mT, n ϭ 6)). Three tumor samples were excluded from this analysis due to an unclear histological classification. In Kruskal-Wallis analysis, the expression levels of cyclin A1, INCA1, GPS2, Ku70, RBM4, and KARCA1 differed significantly between the histological subtypes.
INCA1 Interacted with Cyclin A1 in Vitro and in Vivo-Following the identification of INCA1 as a novel protein, we confirmed the interaction between cyclin A1 and INCA1. To analyze the interaction in vitro, GST pull-down assays with cell lysates or in vitro synthesized cyclin A1 were performed. INCA1 was cloned as a GST fusion protein, expressed in Escherichia coli, and purified using glutathione-agarose beads. Cyclin A1 was either pulled down from lysates of baculovirusinfected Sf9 cells or transcribed and translated in vitro and labeled with [ 35 S]methionine. In GST pull-down assays, cyclin A1 interacted with GST-INCA1, but not with GST alone, indicating a specific interaction between cyclin A1 and INCA1 in vitro (Fig. 7A).
For analysis in vivo, COS-7 cells were transfected with expression plasmids for EGFP-cyclin A1 and Myc-INCA1. INCA1 was immunoprecipitated from whole cell lysates with ␣-Myc antibodies. The subsequent Western blot for EGFP-cyclin A1 demonstrated the specific interaction of INCA1 and cyclin A1 in vivo by their co-immunoprecipitation. Cyclin A1 was not precipitated from the cell lysate by nonspecific antibodies (Fig.  7B). In addition, we confirmed interaction of endogenous cyclin A1 and INCA1 in vivo. ML-1 myeloid leukemia cells express high levels of cyclin A1 (22). Starvation of these cells led to induction of INCA1 (data not shown). In protein lysates of these cells, we found in vivo interaction of endogenous cyclin A1 and endogenous INCA1 by co-immunoprecipitation. Cyclin A1 was not pulled down by preimmune serum but specifically with immune sera against INCA1 (Fig. 7B).
Cyclin A1 and Cyclin A2 in Complex with CDK2 Phosphorylated INCA1 in Vitro Predominantly at Ser 176 -INCA1 interacted in vitro and in vivo with cyclin A1 and was found in the yeast triple-hybrid screen for interacting partners of cyclin A1 in complex with CDK2. These findings and consensus sequences for potential phosphorylation by cyclin-dependent kinases in the INCA1 sequence led us to the hypothesis that  (1.00). The diagram bars below the x axis indicate lower gene expression than in testis; the bars above the x axis indicate higher expression than in testis. The graph clearly shows high expression in testis for all genes except RACK1/GNB2L1 and therefore illustrates the testis-specific expression pattern of most genes identified in our Y3H system. INCA1 was phosphorylated by the cyclin A1-CDK2 complex (Fig. 3, D and E). To examine whether the cyclin A2-CDK2 complex also could phosphorylate INCA1, we carried out in vitro kinase assays with the fusion protein GST-INCA1. Indeed, GST-INCA1 was strongly phosphorylated by the cyclin A1-CDK2 or by the cyclin A2-CDK2 complex but not by either protein alone (Fig. 7C).
Cloning and analysis of different INCA1 fragments indicated strong phosphorylation of the C terminus (aa 149 -221) (Fig.  7D). The N terminus (aa 1-74) was weakly phosphorylated by cyclin A1-CDK2, whereas the intermediate fragment was also phosphorylated by CDK2 alone (data not shown).
Human and murine INCA1 contain four potential phosphorylation sites that were conserved between species; 167 TPGR matched exactly the consensus site for CDK phosphorylation, whereas 23 SP, 176 SP, and 179 SP only showed a conserved SP motif necessary for CDK phosphorylation in both species. Sitedirected mutagenesis of the potential phosphorylation sites to nonphosphorylatable alanine uncovered predominant phosphorylation of Ser 23 in the N terminus and Ser 176 and Thr 167 in the FIG. 5. Expression of cyclin A1 and its interaction partners during testis maturation and spermatogenesis and in testes with decreased fertility. A, we analyzed previously published microarray data (35) for the expression of the murine homologs of cyclin A1 and its interacting proteins throughout testis development. Consecutive testis cDNA samples from day 1 after birth to adult testis were hybridized to oligonucleotide microarrays. To allow easier comparison of expression levels between different genes and to illustrate the gene regulation throughout testicular maturation, gene expression levels were normalized to the maximal expression level of each gene. Identical patterns of induction were found for cyclin A1, KARCA1, and GPS2. Substantial expression of all genes was confirmed in later stages of testis development when highest cyclin A1 expression occurred. B, gene expression levels were analyzed in cDNA isolated from normal testis (n ϭ 6) and testis infertility patient samples ( C terminus. For a double-mutant including T167A and S176A, the strong phosphorylation of the C terminus was diminished (Fig. 7E).
INCA1 mRNA and Protein Were Expressed in Vivo-Since INCA1 was the interaction partner identified most often in our Y3H assay, we studied this novel gene in greater detail. To determine whether the INCA1 gene was transcribed and translated in vivo, we performed Northern blot analysis with total RNA from different organs of two monkey species (Macaca fascicularis (M.f.) and Callithrix jacchus (C.j.)), Homo sapiens (H.s.), and Mus musculus (M.m.) (Fig. 8A). High expression of INCA1 mRNA was detected in monkey and human testis. In murine testis, only a faint signal was seen, most probably because of the human INCA1 probe used. The two bands for INCA1 in monkeys and humans correspond to the two splice variants with 1.4 and 1.2 kb, respectively.
To detect protein expression levels, we raised antibodies against two peptides of human INCA1 and affinity-purified them using the epitopes aa 26 -40 (␣-INCA1#1) or aa 156 -170 (␣-INCA1#2) (Fig. 8, B and C). Transfection of the INCA1 coding sequence into insect and mammalian cell lines led to expression of INCA1 protein (Fig. 8B). In Sf9 insect cells, the antibodies raised against human INCA1 detected a band at an apparent molecular mass of ϳ28 kDa (shown here for ␣-INCA1#1). This probably represented a posttranslationally less modified protein and was closer to the predicted mass of 25.2 kDa. In mammalian COS-7 cells, the apparent molecular weight was ϳ36 kDa. By bioinformatic sequence analysis, several putative consensus sites for posttranslational modifications were detected (e.g. serine glycosylation or tyrosine sulfation). Various putative phosphorylation sites for different kinases existed in the INCA1 sequence. These included consensus sites for CDK phosphorylation (see Fig. 7C and Supplementary Table III anchor, suggesting that it was not membrane-bound. A signal sequence for nuclear localization (RRKKRR, aa 75-80) was found within the INCA1 sequence. To analyze the localization of INCA1 in vivo, we cloned a fusion protein of INCA1 and EGFP. EGFP-INCA1 was expressed in COS-7 cells, and its localization was analyzed by conventional fluorescence microscopy (Fig. 8D, a and b) and confocal laser-scanning microscopy (Fig. 8D, c and d). In addition, cells were stained with DAPI for DNA in the nucleus and TRITC-phalloidin for actin in the cytoplasm. EGFP alone, used as negative control, led to staining throughout the cell with predominance in the cytoplasm (Fig. 8D, a and c). EGFP-INCA1 exclusively localized to the nucleus of the cells (Fig. 8D, b and d). Merging of EGFP and DAPI fluorescence demonstrated the differential localization of EGFP and EGFP-INCA1. Confocal microscopy revealed a nonhomogenous granular staining pattern in the nucleus (Fig.  8D, d).

The Expression of INCA1 Is Regulated by a Sp1-dependent
Promoter-To further analyze the genomic structure and transcriptional regulation of human INCA1, we cloned the genomic sequence upstream of the transcriptional start site of human INCA1. We found a fragment of 1620 base pairs reaching from Ϫ1523 to ϩ97 actively driving transcription in a luciferase assay (Fig. 9A). In COS-7 cells, promoter fragments of 1620 or 593 bp increased luciferase activity more than 50-fold compared with the empty pGL3 basic vector. The putative promoter sequence contained several consensus sequences for binding sites of the transcription factor Sp1. We therefore tested its promoter activity in S2 insect cells lacking endogenous Sp1. In S2 cells without Sp1, no promoter activity was detected, whereas coexpression of Sp1 strongly increased INCA1 promoter activity (Fig. 9B). Fragmentation of the INCA1 promoter revealed high promoter activities for fragments containing the sequence Ϫ254/ϩ97, which was substan- . 8Da, the staining pattern of EGFP (green) was predominantly localized in the cytosol and co-stained with actin (red) rather than with DNA (blue). 8Db, fluorescence of EGFP fused to human INCA1 was localized in the nucleus. 8Dcϩd, confocal laser-scanning microscopy gave a more detailed picture of localization and revealed a nonhomogenous nuclear staining pattern for INCA1. tially decreased when further sequences either on the 5Ј or on the 3Ј site were deleted (Fig. 9C). In addition, we tested the of promoter methylation alone by Aza did not significantly increase the INCA1 expression level (Fig. 9D). These data indicate that the genomic sequence upstream of the human INCA1 gene contained a functional promoter that was dependent on Sp1 and repressed by histone deacetylase activity. DISCUSSION Cyclin A1 is essential for spermatogenesis (26) and possesses oncogenic properties in the hematopoietic system (32). It is highly expressed in testis (22) and in myeloid leukemia (31). Despite some progress in recent years (27,29,38), its molecular and cellular function in either environment has not been characterized in detail due to a lack of knowledge about interaction partners and substrates of the cyclin A1-CDK2 heterodimer. A coherent view of its role in spermatogenesis as well as in leukemogenesis is still missing.
In this study, we used a yeast triple-hybrid approach to identify interaction partners of the cyclin A1-CDK2 complex in a testis cDNA library. The interaction was verified for all isolated proteins in GST pull-down assays. In in vitro kinase assays with cyclin A1-CDK2, phosphorylation of INCA1, Ku70, and RBM4 indicated their role as substrate of the cyclin A1-CDK2 complex. The other identified proteins might be phosphorylated only under in vivo conditions or are interaction partners of the complex that are not phosphorylated.
Analysis of mRNA expression revealed a close correlation of cyclin A1 expression and its interaction partners in normal organs and in samples from normal, malignant, and infertile testis. In general, this pattern of co-expression increases the likelihood of functionally important interactions between cyclin A1 and its co-regulated new interactors in vivo. The interacting proteins included five previously described proteins (GPS2, Ku70, RACK1/GNB2L1, RBM4, and the mRNA DKFZ p686G052) as well as three novel proteins (INCA1, KARCA1, and PROCA1).
The Interaction of Cyclin A1 with Known Proteins Hints at New Functions-GPS2 suppresses growth and controls intracellular signaling via inhibition of the Ras-mitogen-activated protein kinase-c-Jun N-terminal kinase pathway (39,40), modulates p53 transactivation (41), interacts with viral proteins, and is an integral subunit of the N-Cor-HDAC3 corepressor complex (42,43). GPS2 is a nuclear protein, a fact that fits well with the localization of cyclin A1 in testis, and its expression pattern in normal organs, testis maturation, and malignant and infertile testis correlates well with cyclin A1 expression patterns. Therefore, the interaction of GPS2 and cyclin A1-CDK2 in testis is likely to be physiologically relevant and will be a prime candidate for further investigations.
Ku70 is essential for DNA repair of double strand breaks by nonhomologous end joining (44) and thereby contributes to genomic stability (45). It is currently believed to function as a switch between nonhomologous end joining repair and homologous recombination (46). It heterodimerizes with Ku86 to form a DNA helicase and binds to and unwinds double-stranded DNA ends (47). Its role in V(D)J recombination (48) is well established. The expression and function of Ku70 in testis has been a matter of debate; substantial age-dependent expression of Ku70 in testis has been documented (49) as well as its absence in the earliest stages of meiosis (prophase I) (50). However, we isolated the Ku70 cDNA in our Y3H assay from a testis cDNA library, indicating expression in testis. In addition, we found expression of Ku70 in normal testis in our samples by real time quantitative RT-PCR (Fig. 4), and Ku70 expression was also detected on the microarrays of maturing testis (Fig. 5). Besides, cyclin A1 is expressed in organs other than testis at intermediate levels as we report here (Fig. 4); therefore, the interaction with Ku70 might also be important in processes other than spermatogenesis. Other proteins involved in DNA repair are essential for spermatogenesis, increasing the likelihood of a physiological role of Ku70 in the testis (51). However, we provide evidence for the first time for an involvement of a cyclin-CDK complex in Ku70-mediated DNA repair. In fact, we were able to link cyclin A1 to DNA repair mediated by Ku70 and regulation by p53 after irradiation. 3 The RACK1/GNB2L1 protein is a receptor for activated protein kinase C and a homolog of the ␤ subunit of heterotrimeric G proteins (52,53). RACK1 is a substrate (54) and an inhibitor of the Src tyrosine kinase leading to growth inhibition (55). RACK1 is detectable in the cytoplasm interacting with the interferon receptor (56). RACK1 also binds to the p53 homolog p73 and to pRb, which both localize to the nucleus (57); therefore, at least a fraction of RACK1 must be nuclear. In addition, pRb is also a substrate for the cyclin A1-CDK2 complex (29), suggesting a possible complex including pRb, cyclin A1, and RACK1. However, since cyclin A1 is localized to the cytoplasm in myelopoiesis (32), a functional interaction with RACK1 in this compartment is also possible.
The nuclear protein RBM4 is the human homolog to the Drosophila gene lark, which is essential for embryonic development (58). RBM4 binds to mRNA and is involved in the regulation of alternative splicing (59). Other RBM proteins are connected to spermatogenesis (60). RBM4 not only binds to cyclin A1; it is also in vitro phosphorylated by the cyclin A1-CDK2 complex. Hereby, we provide the first link of cyclin A1 to splicing, whereas other cyclins have already been linked to the splicing machinery (61,62), especially cyclin E (63), which associates with CDK2 similar to cyclin A1.
We also found the hypothetical protein DKFZ p686G052 as a novel cyclin A1-interacting partner. Its sequence was found in several ESTs isolated, for example, from testis or from AML samples, indicating an overlapping expression pattern with cyclin A1. In our expression analyses, we confirmed considerable expression of this mRNA in testis and in other organs, and we found significant induction during testis development.
Taken together, expression studies in normal organs revealed substantial expression of all interacting proteins in the testis. The induction of four of the seven analyzed identified interaction partners during testis maturation and the expression of all of them in the adult testis hints at their potential involvement in meiosis or spermatogenesis, which will be subject to further investigations.
Three Novel Proteins Will Provide New Insights into Cyclin A1 Biology-In addition, we isolated three novel proteins interacting with the cyclin A1-CDK2 complex. Following the properties of the novel amino acid sequences, we named these proteins INCA1, KARCA1, and PROCA1. INCA1 and PROCA1 were predicted to localize to the nucleus, and all three new genes were significantly expressed in the testis. The temporal expression profile of KARCA1 during testis development was identical to cyclin A1, indicating its potential involvement in meiosis and emphasizing the functional interaction with the active cyclin A1-CDK2 complex. For KARCA1, we provide the full-length mRNA sequence, whereas the PROCA1 sequence probably represents only a partial coding sequence. KARCA1 contains two kelch motifs and two ankyrin repeats. Kelch motifs were first identified in Drosophila egg chamber regulatory proteins and were assigned to diverse functions including galactose oxidation (galactose oxidase), sialic acid hydrolysis (neuraminidase), or actin cross-linking (scruin). Ankyrin repeats belong to the most common protein-protein interaction motifs. The usually tandemly repeated helix-loop-helix structures are also found twice in KARCA1. This repeat was also found in proteins with a broad range of functions reaching from transcriptional regulation, cell cycle control, cytoskeleton, and ion transport to signal transduction (e.g. Notch, p53-binding proteins, and epidermal growth factor-like domains contain ankyrin repeats). Therefore, the presence of neither the kelch nor the ankyrin motifs directly hints at a specific function, but it opens several opportunities for protein-protein interaction. Future functional studies will uncover the role of these proteins in physiological and pathological settings.
INCA1 was the interacting sequence isolated most often in our Y3H approach that we subsequently cloned and characterized. In the genomic sequences upstream of the INCA1 gene, we identified a functional promoter corroborating the identification of the full-length transcript and an intact genomic locus of INCA1. This promoter was dependent on the transcription factor Sp1, which also regulates the cyclin A1 promoter. In vitro, INCA1 was phosphorylated by cyclin A1 or cyclin A2 in complex with CDK2, but at least in testis, the tissue-specific expression pattern of cyclin A1 and INCA1 might indicate that cyclin A1 is physiologically the more relevant interacting partner. The nuclear localization of INCA1 reveals identical subcellular compartmentation of the active cyclin A1-CDK2 complex and its substrate INCA1. In healthy tissue, INCA1 expression was highest in testis, followed by intermediate expression levels in ovary, pancreas, lung, liver, and spleen. Therefore, expression patterns of cyclin A1 and INCA1 correlated well with regard to the testis, whereas expression in other organs was not closely correlated. The induction of INCA1 in testis maturation hints at a potential role in spermatogenesis. Since INCA1 was the predominant cyclin A1-CDK2-interacting protein identified in our studies, its functional characterization will shed light on the molecular mechanisms underlying cyclinregulated meiosis and spermatogenesis. First functional analyses indicate a growth-suppressive function through regulation of CDK activity by INCA1. 2 In summary, we identified eight interaction partners of the cyclin A1-CDK2 complex including three novel proteins and studied their expression pattern in several organs and developing, adult normal, malignant, and infertile testis. Our analyses will guide future functional studies to unravel the molecular functions of cyclin A1 in physiological and pathological settings and thereby establish a basis for further investigations.