Identification and Characterization of a Cell Cycle and Apoptosis Regulatory Protein-1 as a Novel Mediator of Apoptosis Signaling by Retinoid CD437*

CD437, a novel retinoid, causes cell cycle arrest and apoptosis in a number of cancer cells including human breast carcinoma (HBC) by utilizing an undefined retinoic acid receptor/retinoid X receptor-independent mechanism. To delineate mediators of CD437 signaling, we utilized a random antisense-dependent functional knockout genetic approach. We identified a cDNA that encodes ∼130-kDa HBC cell perinuclear protein (termed CARP-1). Treatments with CD437 or chemotherapeutic agent adriamycin, as well as serum deprivation of HBC cells, stimulate CARP-1 expression. Reduced levels of CARP-1 result in inhibition of apoptosis by CD437 or adriamycin, whereas increased expression of CARP-1 causes elevated levels of cyclin-dependent kinase inhibitor p21WAF1/CIP1 and apoptosis. CARP-1 interacts with 14-3-3 protein as well as causes reduced expression of cell cycle regulatory genes including c-Myc and cyclin B1. Loss of c-Myc sensitizes cells to apoptosis by CARP-1, whereas expression of c-Myc or 14-3-3 inhibits CARP-1-dependent apoptosis. Thus, apoptosis induction by CARP-1 involves sequestration of 14-3-3 and CARP-1-mediated altered expression of multiple cell cycle regulatory genes. Identification of CARP-1 as a key mediator of signaling by CD437 or adriamycin allows for delineation of pathways that, in turn, may prove beneficial for design and targeting of novel antitumor agents.

The natural and synthetic derivatives of vitamin A, collectively known as retinoids, are involved in many biological processes including development, cellular proliferation, and differentiation (1,2). Retinoids exert their actions through modulation of gene expression (3). Ligand-bound receptors heterodimerize and bind to specific DNA consensus sequences called RAR 1 elements or RXR elements located in the regula-tory regions to modulate ligand-dependent transcription of the target gene (3). Unfortunately, the therapeutic anti-cancer efficacy of both natural and synthetic retinoids, which function through this classical pathway, has thus far been limited to the treatment of acute promyelocytic leukemia, CTCL, AIDS-related Kaposi's sarcoma, and premalignant processes (4,5).
We recently described the abilities of a novel retinoid [3-(1adamantyl)-4-hydroxyphenyl]-2-naphthalene carboxylic acid (CD437) to induce growth arrest and apoptosis in a wide variety of malignant cell types, including breast, prostate cancer, and leukemia by an RAR/RXR-independent mechanism (6 -8). In addition, [3-(1-adamantyl)-4-hydroxyphenyl]-2-naphthalene carboxylic acid has been reported to inhibit the growth of human cancer xenografts in nude mice (9). The mechanism(s) by which CD437 triggers cell cycle arrest and apoptosis is as yet not completely defined. On the basis of our research, CD437 was found to enhance expression of the cyclin-dependent kinase inhibitor (CDKI) p21 WAF1/CIP1 and DNA damage-inducible GADD45 genes (10, 11) through a unique mechanism in human breast carcinoma (HBC) cells. We also found that CD437 potently induced c-Jun N-terminal kinase/stress-activated protein kinase and that induction occurred upstream of caspase activation and apoptosis (12). Inactivation of c-Jun, a downstream effector of activated c-Jun N-terminal kinase/ stress-activated protein kinase, was also found to impair CD437dependent apoptosis but not the induction of the expression of the CDKI p21 WAF1/CIP1 in A549 lung cancer cells (13).
The roles of apoptosis mediators have been ascertained by expressing dominant negative peptides directed to specific cellular proteins involved in apoptotic signaling cascades. This powerful approach allows us to inactivate a known mediator, but not novel mediators that may either be key signaling molecules or co-operate with known mediators. Given that CD437 utilizes unique pathways for growth arrest and apoptosis, we hypothesized that novel mediators of CD437-dependent growth arrest and apoptosis existed. We then utilized a functional knockout genetic approach by overexpressing antisense cDNAs (14) to obtain novel molecules involved in CD437-dependent apoptotic signal cascades. The antisense knockout approach involves random inactivation of genes via a cDNA library cloned in an antisense orientation in an expression vector plasmid (14,15). This approach is based on the premise that spe-cific inactivation or knockout of a growth-inhibitory and/or apoptosis-promoting gene would result in a growth advantage to the transfected cells in a specific restrictive environment. This growth advantage constitutes a powerful forward selection that is helpful in isolating the desired inactivation event from random inactivation. Although nascent protein synthesis may be unnecessary for CD437-dependent apoptosis in certain cell types (such as leukemia), stable overexpression of a particular antisense RNA should over time deplete the target protein, thereby causing a partial or complete block in the CD437-dependent apoptotic signal. This methodology allows, in principle, the identification of important genes, the products of which may be modulated at the transcriptional, post-transcriptional, translational, or post-translational level during the CD437-dependent cell cycle arrest or apoptosis. Thus, if a set of novel mediators of growth arrest and/or apoptosis is activated in the presence of CD437, their specific inactivation via overexpression of a transfected antisense cDNA library should lead to cellular resistance to CD437 apoptosis.
The functional knockout genetic approach yielded several CD437-resistant MDA-MB-468 HBC sublines. Plasmids expressing antisense cDNAs were rescued, and one of the plasmids (AS-6) harbored a cDNA insert that belongs to a novel ϳ130-kDa protein (referred to as cell cycle and apoptosis regulatory protein (CARP)-1). CARP-1 expression in HBC cells is regulated by serum as well as by agents such as CD437 and adriamycin. This report details the role(s) of CARP-1 in regulating cell cycle and apoptosis pathways.

EXPERIMENTAL PROCEDURES
Materials-Dulbecco's modified Eagle's medium, Ham's F-12 medium, and fetal bovine serum (FBS) were purchased from Invitrogen. The reagent kit for probe labeling using [ 32 P]dATP (3000 Ci/mmol) was purchased from PerkinElmer Life Sciences. Oligonucleotides for PCR amplification (see below) were purchased from Integrated DNA Technologies, Inc., Corvallis, IA. The novel retinoid CD437 was provided by Dr. U. Reichert, CIRD-Galderma, Sophia-Antipolis, France. The chemotherapeutic drugs taxol and etoposide were obtained from Bristol-Myers Squibb Co., adriamycin from Sigma, and cisplatin from American Pharmaceutical Partners Inc. (Los Angeles, CA). The antibiotics such as gentamycin, neomycin (G418), and hygromycin were purchased from Invitrogen. The ELISA reagent kit for apoptosis was purchased from Roche Applied Science (Pleasanton, CA).
Cell Lines and Cell Culture-Estrogen receptor-negative, p53-negative MDA-MB-468 HBC cells were cultured and maintained essentially as described (16). The stable sublines were generated by transfecting the MDA-MB-468 HBC cells with the vector plasmids pcDNA3, pcDNA3/Hygro, or pMBJF-1, or plasmids expressing CARP-1 antisense (clone 1.6; see below) or c-Myc antisense, followed by selection in the presence of 800 g/ml neomycin or 400 g/ml hygromycin using described methodology (11). The expression plasmid pcDNA3-Myc clone 1 for c-Myc protein has been described before (16). Additional plasmid expressing c-Myc antisense was derived from plasmid pcDNA3-Myc clone 1 by positioning entire c-Myc protein encoding open reading frame (ORF) cDNA in antisense orientation downstream of the CMV promoter. Individual sublines harboring either the vector plasmid (vector sublines), plasmid expressing c-Myc antisense (c-Myc antisense sublines), or plasmid expressing CARP-1 antisense (clone 1.6 sublines) were derived and propagated in the continuous presence of the appropriate antibiotic. In addition, HBC cells were transfected with vector plasmid pcDNA3/myc-HisA or plasmid 6.1 expressing CARP-1 protein fused with Myc and polyhistidine tags at the C terminus (see below). Multiple neomycin-resistant stable sublines were generated and propagated as described (11).
Cloning of CARP-1 cDNAs-MDA-MB-468 HBC cells were either untreated or treated separately with 1 M CD437 for 1, 3, 6, 12, 24, and 48 h. Total RNAs were extracted and pooled. Poly(A) ϩ RNAs were prepared following standard methods and then utilized to synthesize double-stranded cDNAs, followed by generation of antisense cDNA library as described (14,15,(17)(18)(19). The cDNAs were cloned in antisense orientation in episomal plasmid pMBJF-1 that is a derivative of pTK01 (14), where SV40 promoter has been replaced by CMV promoter elements. RT-PCR analysis of Escherichia coli DH10B transformed with the above library showed greater than 80% colonies having recombinant plasmids with a range of cDNA insert sizes (0.1-2.5 kb), suggesting excellent recombination efficiency and complexity of the library. The library was then transfected into MDA-MB-468 HBC cells and stable sublines selected in the presence of hygromycin (400 g/ml) and CD437 (1 M), followed by plasmid DNA extraction from each subline and subsequent sequencing of the inserts as described (11,20).
The CARP-1 cDNA (GenBank TM accession number AY249140) having the entire CARP-1 protein encoding ORF was cloned by RT-PCR methodology utilizing total RNA derived from MDA-MB-468 HBC cells. The sequence of the primers for PCR amplification (see Table I) were derived from the GenBank TM data base having mouse and human cDNAs with homologies to the cDNA insert of plasmid AS-6 (see below). Multiple subfragments of the CARP-1 cDNA were independently PCRamplified by using combinations of oligos AS-6.1 plus AS-6.9 (fragment 1), AS-6.8 plus AS-6.11 (fragment 2), AS-6.10 plus AS-6.13 (fragment 3), AS-6.12 plus AS-6.15 (fragment 4), or AS-6.14 plus AS-6.17 (fragment 5). The PCR-amplified cDNA fragments were digested with either BamHI plus DraI (fragment 1), DraI plus NheI (fragment 2), NheI plus SacI (fragment 3), SacI plus EcoRV (fragment 4), or EcoRV plus XhoI (fragment 5) followed by their sequential ligation into BamHI plus XhoI cut vector plasmid pcDNA3/myc-HisA (Invitrogen) to obtain a recombinant plasmid clone 6.1 (Fig. 1). This recombinant plasmid has CARP-1 cDNA with the entire protein encoding ORF, the translation start codon ATG but lacks translation termination codon TGA, whereas Myc and polyhistidine epitope tags are fused in-frame at the C terminus of the CARP-1 protein. In addition, ϳ1.4-kb CARP-1 cDNA fragment was PCR-amplified using oligos AS-6.1 and AS 6.11, followed by restriction digestion of the cDNA with BamHI plus NheI. This cDNA subfragment was then cloned in antisense orientation in BamHI-and NheI-digested plasmid vector pcDNA3/Hygro (Invitrogen) to obtain plasmid clone 1.6 ( Fig. 1).
MTT and Apoptosis Assays-Sublines transfected with vector plasmid pCDNA3/Hygro or clone 1.6 were either untreated or separately treated with 1 M CD437 for 72 h, followed by determination of their viability by MTT assay (18). The parental MDA-MB-468 as well as vector-or clone 1.6-transfected HBC sublines were also treated either with adriamycin (34 M), taxol (30 M), cisplatin (10 g/ml), or etoposide (10 g/ml) over a range of time points, and their viabilities were determined by MTT assays as described (18,22). The assessment of the live cells was derived by measuring the absorbance of the converted dye at a wavelength of 570 nm.
In addition, the parental MDA-MB-468 as well as vector plasmidtransfected subline or two independent sublines derived from stable transfection of clone 1.6 were separately treated with either 1 M CD437 or 34 M adriamycin for 72 h. The parental MDA-MB-468 HBC cells were also independently transduced with retroviruses expressing either CARP-1-myc-His fusion protein or alkaline phosphatase (see below). Cell lysates were prepared from untreated as well as treated cells to determine apoptosis by utilizing cell death detection ELISA kit (Roche Diagnostics). The net absorbances at the suggested wavelengths (A 405 nm minus A 490 nm ) were obtained for the lysates derived from treated as well as untreated cells. The histograms indicating levels of apoptosis were generated either by plotting net absorbances or by calculating the ''enrichment factor'' as suggested by the manufacturer's protocols. The enrichment factor indicating release of mono-and oligonucleosomes following apoptosis was calculated by dividing the net absorbance values from the treated samples by those obtained from untreated controls.
Immunoprecipitation and Western Blot Analysis-Logarithmically growing parental HBC cells as well as various sublines were either untreated or treated with different agents for the indicated time points. For immunoprecipitation, aliquots of cell lysates containing 1.0 mg of proteins were incubated with indicated antibodies and Sepharose G at 4°C for 3 h. Immunoprecipitates or the protein extracts were electro-phoresed on 9 -12% SDS-polyacrylamide gels essentially as described before (10,18,22). The filters containing protein extracts were independently probed with indicated antibodies per the manufacturer's guidelines and the protocols described before (10).
Flow Cytometric Analysis-Flow cytometric analysis of DNA content was performed to assess the cell cycle phase distribution as described previously (23). Normally growing parental wild type, the vector plasmid-transfected, c-Myc antisense-transfected sublines, or clone 1.6transfected HBC sublines were independently stained for DNA content using propidium iodide. After staining, cells were analyzed on a BD Biosciences FACScan cytometer (San Jose, CA). The data were analyzed using the multicycle program from Phoenix Flow Systems (San Diego, CA).
Gene Regulation by CARP-1-To study the downstream targets of CARP-1, Atlas Human 1.2 array membranes (Clontech) with ϳ1200 genes were utilized in conjunction with radiolabeled total RNAs from vector pcDNA3/Hygro subline 1 or clone 1.6 subline 9. Total RNAs (3.0 g) from each subline were separately labeled with [ 32 P]dATP and hybridized with the array membranes essentially following the manufacturer's guidelines. The membranes were washed and autoradiographed for 2 weeks, and the hybridization signals were quantitated utilizing the manufacturer's suggested program.
Retroviral Expression of CARP-1-Retroviral gene transfer and expression system (Clontech, Palo Alto, CA) was used to generate retroviruses expressing CARP-1. First, the entire CARP-1-myc-His-tagged cDNA insert of plasmid clone 6.1 was excised with BamHI and PmeI and ligated into BglII and StuI cut retroviral vector plasmid pLNCX2 to obtain plasmid clone 16.1. Next, the plasmid clone 16.1 or pLAPSN (retroviral vector expressing alkaline phosphatase) were independently transfected into the RetroPack PT67 packaging cells (Clontech), followed by selection of PT67 sublines in the presence of neomycin (500 g/ml). Multiple, independent sublines or pooled populations expressing either myc-His-tagged CARP-1 or alkaline phosphatase were obtained. The expression of myc-His-tagged CARP-1 protein in each subline was determined by Western immunoblot using anti-His (Cterminal) monoclonal antibody (Invitrogen). Two independent PT-67 sublines expressing the highest levels of myc-His-tagged CARP-1 (lines 16.4 and 16.6) and one subline expressing alkaline phosphatase were selected for determination of viral titers essentially per the suggested protocols.
Immunolocalization of CARP-1-CARP-1 antibodies ␣1 and ␣2 were generated (Sigma Genosys, The Woodlands, TX) by immunizing rabbits with keyhole limpet hemocyanin-conjugated peptides containing epitopes RERERERRR (amino acids 320 -328) and EDD-KEEEERKRQEEI (amino acids 698 -712), respectively, of the putative sequences deduced from the CARP-1 cDNA. Localization of CARP-1 in parental HBC cells or HBC subline expressing CARP-1-myc-His fusion protein was studied by immunocytochemical analyses (22) by utilizing ␣1 anti-CARP-1 (1:2000 dilution) polyclonal antibody or anti-His-tag (27E8) monoclonal antibody (Cell Signaling, Beverly, MA), respectively. In addition, ␣1 anti-CARP-1 polyclonal antibody (1:1000 dilution) in conjunction with mouse monoclonal antibody for proliferating cell nuclear antigen (PCNA; Santa Cruz Biotechnology, Santa Cruz, CA) were utilized to study expression of CARP-1 and PCNA proteins in a human breast cancer biopsy specimen. Dual labeling of CARP-1 and PCNA was carried out as above followed by confocal microscopic analysis to determine expression of CARP-1 and PCNA proteins. The antibody-stained sections were then photographed under different magnifications utilizing Zeiss microscope with attached 35-mm camera for recording the photomicrographs.

Isolation of CARP-1 cDNA by Functional Gene Knockout-
CD437 is a novel retinoid that utilizes a retinoid receptorindependent mechanism(s) to elicit anti-proliferative and apoptotic effects in the HBC cells (6 -8). We hypothesized that novel mediators of CD437 signaling exist, and their specific inactivation via overexpression of the transfected antisense cDNA library should lead to resistance to apoptosis by CD437. To accomplish inactivation of rare proteins and those encoded by low abundance mRNAs, poly(A) ϩ RNAs were used as template for generating a cDNA library such that mRNA species that are present at ϳ20 copies per cell are included in the library (14). RNAs were prepared from MDA-MB-468 HBC cells treated with CD437 over a range of time points to allow cloning of molecules involved in early, middle, and late signaling events triggered by CD437. CD437 at a concentration of 1 M was found previously (6, 10) to be optimal in eliciting cell cycle arrest as well as apoptosis. After cDNA library was transfected into HBC cells, stable sublines were selected in the combined presence of hygromycin and CD437 as described (11). Of the several HBC colonies isolated and propagated to obtain plasmid DNAs, one harbored the recombinant plasmid AS-6 having an ϳ0.15-kb insert (positions 1760 -1900; Fig. 1). A composite restriction map was created from the human and mouse partial cDNAs in the GenBank TM that were homologous to the insert of the AS-6 plasmid, followed by cloning of the full-length cDNA that encodes a novel 130-kDa protein CARP-1 ( Fig. 1).
A protein data base search yielded very high homology scores for two putative motifs in the CARP-1 peptide. A sequence of 31 amino acids (from positions 637 to 667) showed 88% alignment with a putative DNA-binding (bihelical) motif predicted to be involved in chromosomal organization. A 41-amino acid sequence (positions 151-191) showed 76% alignment with a putative cold-shock protein domain that function as an RNAbinding domain. The partial mouse cDNA-encoded peptides also contain DNA-binding (bihelical) and cold-shock protein domains that are identical to those present in the human CARP-1 protein. The plasmid expressing myc-His-tagged wildtype CARP-1 having putative cold-shock and DNA-binding motifs was then generated as detailed under ''Experimental Procedures'' (clone 6.1, Fig. 1).
Furthermore, the CARP-1 5Ј-untranslated region subfragment having part of the protein encoding ORF was RT-PCRamplified using oligos AS-6.23 and AS-6.25 (Table I) and total RNA from MDA-MB-468 HBC cells. The PCR product was subcloned into pBSK vector plasmid, and multiple independent recombinant plasmids were sequenced to confirm the presence of a translation termination codon located 67 nucleotides upstream and in-frame of the CARP-1 ATG (not shown). Thus, CARP-1 peptide encoded by the plasmid clone 6.1 (see Fig. 1) possesses the complete N terminus of the CARP-1 protein.
HBC Cells Expressing Reduced CARP-1 Show Increased Cell Viability and Reduced Apoptosis When Treated with CD437 or Adriamycin-Whether CARP-1 plays a role in apoptosis signaling in the presence of CD437 or other chemotherapeutic drugs was investigated next. The parental MDA-MB-468 HBC cells were treated with different drugs over a range of times, and their viabilities were assessed by MTT assays. Cell viabilities of 20% or less were observed following 72 h of treatment with either adriamycin, etoposide, or cisplatin, while Ͻ20% cell viability was observed after 24 h of treatment with taxol (not shown). The reduction in viability of HBC cells treated with adriamycin, etoposide, or cisplatin was due to apoptosis as evidenced by fragmented nuclei following acridine orange staining (not shown).
We found that 1 M CD437 induced apoptosis in HBC cells within 48 -72 h. Whether CARP-1 played a role in CD437-dependent apoptosis was investigated in four independent sublines transfected with vector plasmid pCDNA3/Hygro or 15 independent sublines transfected with clone 1.6. Each subline was separately treated with 1 M CD437 for 72 h. CD437 treatment of vector-transfected sublines resulted in Ͼ80% loss of cell viability, whereas several clone 1.6 transfectants only showed a 20 -30% loss ( Fig. 2A). The variability in viability of CD437-treated clone 1.6 transfectants could be due to the varying copy number or nuclear integration site(s) of the plasmid 1.6 that, in turn, influenced the levels of antisense RNA and subsequently CARP-1 protein. The resistance to apoptosis by CD437 observed in multiple, independent sublines transfected with clone 1.6 suggested a role for antisense-mediated reduced levels of CARP-1 in this process.
The CD437-resistant HBC sublines transfected with clone 1.6 (see Fig. 2A) were then utilized to investigate whether expression of CARP-1 antisense contributed to chemotherapeutic drug resistance. Independent sublines transfected with either vector (sublines 1-3) or clone 1.6 (sublines 8 -11 and 13; see Fig. 2A) were treated with taxol for 24 h or adriamycin, etoposide, or cisplatin for 72 h. Treatment with cisplatin or taxol resulted in reduced viability of the treated sublines transfected either with the vector or clone 1.6 (not shown). However, all clone 1.6-transfected sublines had significantly increased viability when compared with vector alone transfectants after treatment with etoposide or adriamycin (Fig. 2B).
Whether differences in the loss of viabilities of vector or clone 1.6-transfected sublines treated with either CD437 or adriamycin was a result of apoptosis was ascertained by utilizing cell death detection ELISA kit (Roche Applied Science) as described under ''Experimental Procedures.'' Treatment with CD437 or adriamycin resulted in the levels of apoptosis in the wild-type or vector subline 1 cells that were approximately three times that observed for the clone 1.6-transfected subline 9 or 10 ( Fig.  3). Thus, the data suggest that increased viabilities of CD437or adriamycin-treated clone 1.6 sublines 9 or 10 ( Fig. 2) reflect alterations in the cell numbers due to reduced apoptosis.
Expression of CARP-1 Antisense Results in Reduced Levels of CARP-1 Protein-Western immunoblot analysis of clone 1.6 expressing HBC sublines 8 -11 and 13 (see Fig. 2A) using ␣1 anti-CARP-1 polyclonal antibodies showed ϳ50% reduced levels of CARP-1 when compared with the vector plasmid-transfected HBC subline (Fig. 4, A and B). Moreover, clone 1.6 sublines 1 and 4 that displayed reduced viabilities following treatment with CD437 (see Fig. 2A) did not have reduced CARP-1 when compared with clone 1.6 sublines 9 and 10 ( Fig.  4, C and D). Taken together, the data in Figs. 2-4 suggest that reduced expression of CARP-1 contributes to increased cell viability as well as lower apoptosis of HBC cells further implicating CARP-1 in apoptosis signaling pathways utilized by CD437, adriamycin, or etoposide but not by taxol or cisplatin. The chemotherapeutic agents adriamycin, etoposide, or cisplatin are known to cause DNA damage leading to apoptosis in HBC cells. Because CARP-1 contains a predicted motif for DNA (bihelical) binding, it remains to be determined whether CARP-1 binding to DNA induced by CD437, adriamycin, or etoposide leads to DNA damage followed by apoptosis.
CD437, Adriamycin, or Serum Regulate CARP-1 Expression in HBC Cells-Next we studied whether growth arrest and/or apoptosis signaling pathways target CARP-1 expression. HBC cells were treated with either 1 M CD437 or 34 M adriamycin  serum resulted in a time-dependent decrease in CARP-1 expression (Fig. 5C). The data in Fig. 5 demonstrate that CARP-1 expression is regulated by growth factors as well as agents like CD437 and adriamycin. Because CD437 treatment of HBC cells results in G 1 /G 0 cell cycle arrest (6), the increased levels of CARP-1 either in the presence of CD437 or in the absence of serum may support a role for CARP-1 in the progression of the cell cycle.

CARP-1 Is a Perinuclear Protein and Is Expressed in Different Epithelial Cancer Cell
Types-To study CARP-1 localization, HBC cells were transfected with plasmid clone 6.1, followed by selection of multiple, independent sublines expressing myc-His-tagged CARP-1 (Fig. 6A). Immunocytochemical localization of CARP-1 was investigated as described under ''Experimental Procedures.'' As shown in Fig. 6B, CARP-1 is present in the perinuclear compartment. Perinuclear localization of CARP-1 was further corroborated by confocal microscopic analysis of a breast cancer biopsy specimen following dual labeling utilizing antibodies for CARP-1 (␣1 antibody) and PCNA. Expression of PCNA (labeled with green fluorescent marker) is exclusive to the nuclear compartment, whereas sporadic expression of CARP-1 (labeled with red fluorescent marker) is noted in the surrounding perinuclear regions of the cells (Fig.  6B). In the absence of yellow color (observed where the red and green colors overlap) fluorescence in the confocal micrograph in either the nuclear or perinuclear compartments together with the data from immunocytochemical analyses strongly suggest that CARP-1 is a perinuclear protein. In addition, CARP-1 is expressed ubiquitously in various cancer cell lines of human breast, colon, and prostate and pancreatic and leukemia origins as indicated by RT-PCR amplification of CARP-1 and ribosomal phosphoprotein 36B4 (24) cDNAs, as well as Western immunoblots (not shown).

CARP-1 Regulates Expression of Key Cell Cycle Regulatory
Proteins-To elucidate the CARP-1 role in signaling pathways regulating cellular growth and/or apoptosis, we investigated the CARP-1 target genes by gene array technology as detailed under ''Experimental Procedures.'' A subset of genes showing alterations in their expression in clone 1.6 subline 9 when compared with their expression in the vector pcDNA3/Hygro subline 1 are listed in Table II. Because clone 1.6-transfected subline 9 expresses 50% reduced levels of CARP-1 (see Fig. 4), it is noteworthy that several genes associated with cell cycle progression and/or cell proliferation such as c-myc, cyclin B1, DNA topoisomerase II␣, p21 rac1 , and histone deacetylase (HDAC) 3C were significantly up-regulated (Table II). On the other hand, the array data did not show altered expression of genes such as proliferating cell nuclear antigen (PCNA) and stress-activated protein kinase 4 (p38␦) in clone 1.6-transfected subline 9 (not shown). Together, the data suggest that CARP-1 targets expression of specific genes including those that are regulators of cell cycle progression and/or cell proliferation.
Whether reduced levels of CARP-1 result in altered expression of the proteins corresponding to the above noted genes was studied by Western immunoblotting. Furthermore, because CD437 treatment induces CARP-1 expression in HBC cells (Fig. 5), we hypothesized that CD437 treatment will also alter expression of above CARP-1 target genes in HBC cells. Expression of oncogene c-myc as well as CDKI p21 WAF1/CIP1 proteins was analyzed in HBC sublines expressing reduced CARP-1 (Fig. 4) as well as CD437-treated or untreated parental MDA-MB-468 HBC cells. Indeed, reduced expression of CARP-1 results in increased levels of c-Myc (Fig. 7A) and reduced levels of CDKI p21 WAF1/CIP1 (Fig. 7C) in multiple independent sublines expressing reduced CARP-1. CD437 treatment of parental MDA-MB-468 HBC cells, on the other hand, resulted in reduced expression of c-Myc (Fig. 7B) and elevated expression of p21 WAF1/CIP1 protein (Fig. 7D). CD437 treatment has been shown to cause elevated expression of p21 WAF1/CIP1 mRNA in HBC cells (10), and together with the data in Fig. 7, C and D, strongly suggest that CARP-1 targets p21 WAF1/CIP1 expression in the presence of CD437. This conclusion is further supported by the observation where clone 1.6 sublines expressing reduced CARP-1 did not elicit enhanced p21 WAF1/CIP1 expression following treatment with CD437. CD437 treatment, on the other hand, resulted in increased levels of p21 WAF1/CIP1 in clone 1.6 sublines that did not express reduced CARP-1 (Fig. 7E). Thus, CD437-dependent signaling involves increased expression of CARP-1 that, in turn, down-regulates c-Myc and up-regulates CDKI p21 WAF1/CIP1 proteins leading to cell cycle arrest. This property of CARP-1 may then serve as one of the contributing factor(s) in inducing apoptosis in the presence of CD437.
In addition to regulation of c-Myc and p21 WAF1/CIP1 by CARP-1, expression of cyclin B1, p21Rac1, HDAC3, and DNA topoisomerase II␣ in multiple HBC sublines expressing clone 1.6 was investigated by Western immunoblotting. As expected, reduced levels of CARP-1 results in increased expression of cyclin B1, p21Rac1, HDAC3, and DNA topoisomerase II␣ (Fig.  7F), suggesting that CARP-1 not only targets c-Myc and p21 WAF1/CIP1 expression, but it also interferes with expression of other genes involved in the pathways regulating cell proliferation and/or cell cycle progression. Because chemotherapeu- tic agents such as adriamycin and etoposide are known to target DNA topoisomerases, the increased expression of DNA topoisomerase II␣ in clone 1.6-expressing sublines may also be a contributing factor in their resistance to apoptosis in the presence of etoposide or adriamycin.
Because depletion of CARP-1 results in elevated expression of genes that are positive regulators of cell cycle and/or cell growth, we wished to determine whether there were alterations in any of the phases of the cell cycle in the case of the sublines expressing reduced levels of CARP-1 that may consequently have resulted in their increased viabilities and reduced apoptosis in the presence of CD437. Flow cytometric analyses of parental MDA-MB-468 HBC cells, HBC subline transfected with the vector plasmid only, two sublines expressing reduced CARP-1 (clone 1.6 sublines 9 and 10), or two sublines expressing reduced c-Myc (c-Myc antisense sublines 3, 6; see below) showed no significant alterations in their cell cycle phase distribution or DNA contents (not shown).
Expression of CARP-1 Alters Cellular Levels of CARP-1 Target Genes and Induces Apoptosis-Because decreased levels of CARP-1 resulted in reduced expression of CDKI p21 WAF1/CIP1 , and elevated levels of c-Myc and DNA topoisomerase II␣, we investigated whether overexpression of CARP-1 will result in altered levels of the above genes and induction of apoptosis. To this end, we utilized retroviral expression system to generate CARP-1-expressing retroviruses. Interestingly, multiple transfection attempts yielded a significantly lower number of sublines expressing retroviruses encoding CARP-1-myc-His fusion protein when compared with transfections with alkaline phosphatase encoding the retroviral vector (not shown). Among a total of 11 PT-67 sublines, CARP-1 expression was noted to be highest in 3 sublines (lines 16. sion protein resulted in isolation of greatly reduced numbers of sublines when compared with their alkaline phosphatasetransfected counterparts would suggest that CARP-1 expression beyond a threshold is detrimental to cell survival. Next, parental MDA-MB-468 HBC cells were cultured in the presence of 5 ml of supernatants derived from either parental non-transfected PT-67 cells, PT-67 pLAPSN subline 1, PT-67 subline 16.4, or subline 16.6 for a period of 72 h, followed by determination of apoptosis. Transduction of HBC cells with retroviruses expressing CARP-1-myc-His fusion protein resulted in apoptosis 4 -5 times higher than those transduced with supernatants derived from either untransfected PT-67 cells or PT-67 pLAPSN subline 1 (Fig. 8B). That transduction of HBC cells with retroviruses from PT-67 sublines 16.4 or 16.6 indeed resulted in apoptosis of HBC cells was corroborated by the presence of reduced levels of procaspase 3 noted in Western immunoblots of the respective lysates (Fig. 8C). The data in Fig. 8, B and C, support our hypothesis that overexpression of CARP-1 indeed results in apoptosis of the target cells and that this property of CARP-1 may have contributed to the lower retroviral titers of the supernatants derived from PT-67 sublines 16.4 and 16.6.
Whether CARP-1 expression targeted genes regulating cell cycle progression and/or cell proliferation as noted in Fig. 7 was examined next. Indeed, Western immunoblots revealed altered expression of c-Myc, CDKI p21 WAF1/CIP1 , and DNA topoisomerase II␣ in the parental MDA-MB-468 HBC cells that were transduced with retroviruses expressing CARP-1 (Fig. 8D). The data in Fig. 8D also underscore the fact that altered expression of the genes noted in HBC sublines expressing reduced CARP-1 (Fig. 7) was independent of the effects of position and/or site(s) of integration of plasmid clone 1.6.
CARP-1 Interacts with 14-3-3 Protein-To further define mechanisms of CARP-1-dependent apoptosis, we utilized HBC cells as well as PT-67 cells expressing myc-His-tagged CARP-1 (Figs. 6 and 8) in conjunction with immunoprecipitation experiments. Because CARP-1 is a perinuclear protein, and in light of the fact that CARP-1 overexpression triggers apoptosis, we hypothesized that CARP-1 interaction with protein(s) located at the plasma membrane or in the perinuclear compartment plays an important role in apoptosis induction. HBC cell lysates were incubated with either anti-epidermal growth factor receptor (EGFR) antibody or anti-14-3-3 antibody, followed by analysis of immunoprecipitates for the presence of CARP-1 by Western blots as described under ''Experimental Procedures.'' As shown in Fig. 9, CARP-1 presence is noted in the lanes containing immunoprecipitates derived from 14-3-3 antibody, whereas anti-EGFR antibodies failed to immunoprecipitate CARP-1.

Depletion of c-Myc Sensitizes Cells to Apoptosis by CARP-1, whereas Expression of c-Myc or 14-3-3 Interferes with CARP-1
Apoptosis-Whether c-Myc or 14-3-3 proteins play roles in CARP-1-mediated apoptosis was investigated as follows. Independent HBC sublines expressing ϳ50% reduced levels of c-Myc (Fig. 10, A and B) were generated by transfecting c-Myc antisense cDNA as described under "Experimental Proce- dures." As noted in Fig. 8, treatment of wild-type HBC cells with retroviruses expressing myc-His-tagged CARP-1 resulted in apoptosis over a period of 72 h. Transduction of wild-type or the vector-transfected HBC subline with retroviruses expressing myc-His-tagged CARP-1 over a period of 24 h resulted in marginally elevated apoptosis when compared with apoptosis noted in cells transduced with retroviruses expressing alkaline phosphatase (Fig. 10C). On the other hand, transduction of sublines harboring reduced c-Myc with retroviruses expressing myc-His-tagged CARP-1 protein over a period of 24 h resulted in apoptosis ϳ3 times higher than that noted in cells transduced with retroviruses expressing alkaline phosphatase (Fig.  10C). Thus, data in Fig. 10C suggest that loss of c-Myc sensitizes HBC cells to apoptosis by CARP-1. Next, wild-type MDA-MB-468 HBC cells were transfected with vector plasmid pcDNA3, pcDNA3-c-Myc clone 1 (16), or pEF-1/14-3-3 clone 1, followed by incubation of cells with retroviruses expressing alkaline phosphatase or myc-His-tagged CARP-1 for a period of 48 h. CARP-1 transduction of cells transfected with vector plasmid resulted in ϳ2.5-fold higher apoptosis when compared with their alkaline phosphatase transduced counterparts, whereas expression of c-Myc or 14-3-3 inhibited apoptosis induced by CARP-1 transduction (Fig.  10D). Together, the data in Fig. 10 underscore the fact that c-Myc and 14-3-3 proteins play important roles in CARP-1-dependent apoptosis.

DISCUSSION
CD437 is a novel, RAR␥-selective adamantyl retinoid that causes cell cycle arrest and apoptosis in a number of cell types. CD437 displays poor binding and trans-activation of RAR␤ and RAR␣, and no binding to RXRs (25). CD437 also causes apoptosis of human leukemia cell line HL-60R that does not express any functional RARs (7), indicating that CD437 functions independently of direct interactions with retinoid receptors. Although CD437-mediated growth arrest and apoptosis has been FIG. 7. CARP-1 regulation of gene expression. CARP-1 regulation of c-Myc (A), CDKI p21 WAF1/CIP1 (C and E), or topoisomerase II␣, cyclin B1, Rac1, and HDAC3 proteins (F) studied by Western immunoblot in indicated HBC sublines. Protein lysate from each subline (100 g/lane for A and F or 200 g/lane for C) was analyzed by 10% SDS-PAGE and immunoblotting with antibodies for c-Myc (A), p21 WAF1/CIP1 (C), topoisomerase II␣, cyclin B1, Rac1, and HDAC3 proteins (F). E, wild-type HBC cells or indicated sublines were either untreated (Ϫ) or treated (ϩ) with noted dose of CD437 for 6 h. Protein lysate (100 g) from each subline was analyzed on 10% SDS-PAGE followed by immunoblotting with anti-p21 WAF1/CIP1 antibody. B and D, wild-type HBC cells were treated with indicated dose of CD437 for the noted periods, and protein lysates were analyzed as in A using anti-c-Myc (B) and anti-p21 WAF1/CIP1 (D) antibodies. All the membranes were then probed with anti-actin antibody to assess loading. The presence of various proteins analyzed in Western blots of A-F is highlighted by Ͼ on the left side of each blot.
shown to involve elevated expression of CDKI p21 WAF1/CIP1 , GADD45, activation of caspases, and release of mitochondrial cytochrome c (6-13), the mechanism(s) involved remain to be fully delineated. We undertook a functional gene knockout approach to further elucidate signaling pathway(s) utilized by CD437. This approach has been used previously (14,15,19) to identify genes involved in apoptosis signaling in the presence of interferons and/or retinoic acid. By transfecting episomal plasmids containing antisense cDNAs derived from the cDNA library generated from untreated as well as CD437-treated HBC cells, we obtained multiple independent sublines that exhibited resistance to the combined presence of hygromycin (an antibiotic resistance marker expressed by the vector plasmid) and CD437 (not shown). One subline harbored a recombinant plasmid, termed AS-6, which contained an insert that belongs to an mRNA of ϳ3.8 kb size. This mRNA encodes a novel ϳ130-kDa protein termed CARP-1. CARP-1 is an important target of apoptosis signaling induced by agents like CD437 or adriamycin. CARP-1 interacts with cytoplasmic 14-3-3 protein(s) and also plays a role in cell cycle progression. Increased expression of CARP-1 by CD437 results in sequestration of 14-3-3 protein(s) and altered expression of many key cell cycle regulatory proteins that together lead to cell growth inhibition and apoptosis.
CARP-1 serves as a key mediator of apoptosis signaling as supported by data in Fig. 8. Expression of CARP-1, following transduction of HBC cells with retroviruses expressing CARP-1, results in apoptosis and reduced levels of procaspase 3. It is possible that expression of CARP-1 beyond a threshold induces apoptosis, since transfection of plasmids expressing CARP-1 under the control of CMV immediate early promoter (such as plasmid clone 6.1 or 16.1) into HBC, HCT-116 colon cancer cells (not shown), and NIH 3T3-derived PT-67 cells also resulted in their apoptosis. Whether the HBC or the PT-67 sublines (Figs. 6 and 8) express CARP-1 below a threshold that allowed for successful selection and propagation of these sublines or CARP-1 overexpression resulted in their altered genotype that subsequently contributed toward their selection and propagation remains to be determined.
The precise mechanism(s) utilized by CARP-1 to induce apoptosis are yet to be clarified. Whether CARP-1 directly binds 14-3-3 is not clear. Nevertheless, CARP-1 interaction with cy-  (Fig. 9) suggests that CD437-dependent elevated levels of CARP-1 sequester 14-3-3 protein and, in turn, may result in inhibition of 14-3-3 binding with the pro-apoptotic protein Bad. 14-3-3 proteins are a family of small, evolutionarily conserved proteins that are found in all vertebrates, invertebrates, plants, and fungi. 14-3-3 proteins display wide ranging functions. Mammalian 14-3-3 proteins have been classified as typical (␣/␤, /␦, ␥, , and ) or atypical ( and ⑀) isoforms based on their distribution and functions. 14-3-3 proteins are involved in signal transduction in the mitogen-activated protein kinase cascade (21,26). The mitogenic signals activate AKT/PKB kinases that, in turn, phosphorylate the pro-apoptotic protein Bad. The 14-3-3 proteins then bind with phosphorylated Bad and prevent it from inhibiting BclX L , thereby inhibiting apoptosis and promoting cell proliferation. CARP-1 sequestration of 14-3-3 thus likely contributes to apoptosis signaling by promoting Bad inhibition of BclX L followed by release of cytochrome c from mitochondria and subsequent apoptosis. Thus, as summarized in Fig. 11, the data presented in this report suggest that CARP-1-dependent apoptosis utilizes multiple pathways including CARP-1 interaction with 14-3-3 and CARP-1-mediated altered levels of cell cycle regulators such as c-Myc, cyclin B, and CDKI p21 WAF1/CIP1 . Loss of c-Myc sensitizes cells to CARP-1-mediated apoptosis (Fig. 10C). Overexpression of c-Myc or 14-3-3 proteins, on the other hand, inhibits CARP-1-dependent apoptosis in HBC cells (Fig. 10D) as well as activation of caspase 3 in HBC and monkey kidney COS-7 cells (not shown).
Whether CARP-1 regulation of expression of the above noted cell cycle regulatory genes utilizes 14-3-3-dependent and/or -independent pathways remains to be delineated. CARP-1 regulation of CDKI p21 WAF1/CIP1 , nonetheless, underscores its growth inhibitory/anti-proliferative property (Figs. 7 and 8). Although reduction p21 WAF1/CIP1 mRNA levels was not noted in the array experiment (Table II), it is likely that reduced levels of p21 WAF1/CIP1 mRNA in clone 1.6 subline 9 resulted in signals that were below the threshold of detection set for densitometric quantitation of the autoradiographs. In light of the fact that CD437 regulates CDKI p21 WAF1/CIP1 mRNA stability (10), whether CARP-1 regulates stability of p21 WAF1/CIP1 mRNA utilizing its RNA-binding cold-shock motif or other novel domain(s) is not known. Alternatively, due to the existence of reciprocal regulation between c-Myc and p21 WAF1/CIP1 (27), and given that c-Myc down-regulates p21 WAF1/CIP1 (28), it has yet to be determined whether CARP-1 regulation of p21 WAF1/CIP1 involves c-Myc. In the absence of direct interactions of CARP-1 with c-Myc or p21 WAF1/CIP1 , as demonstrated by co-immunoprecipitation and Western immunoblot experiments (not shown), and the fact that CARP-1 is present in the perinuclear region indicate that CARP-1 regulation of c-Myc and p21 WAF1/CIP1 may utilize additional factor(s)/mediator(s). Moreover, that CARP-1 may utilize post-transcriptional and/or post-translational mechanisms to regulate levels of its target genes is supported by the fact that CARP-1-dependent alterations in levels of p21 Rac1 and HDAC3 mRNAs (Table II) were quite different when compared with the respective proteins in HBC sublines expressing reduced CARP-1 (Fig. 7F).
That CARP-1 is an important regulator of the cell cycle is supported by the observations that treatment of cells with either CD437 or adriamycin resulted in a time-dependent increase in expression of CARP-1 (Fig. 5). We have shown previously (6,29) that exposure to CD437 causes G 1 /G 0 and S phase cell cycle arrest in HBC and prostate cancer cells, respectively, followed by their apoptosis. Rapid induction of CARP-1 expression following treatment with CD437 may indicate a critical role for CARP-1 in signaling events regulating the cell cycle. Similar induction of CARP-1 expression was also noted in the case of cells treated with adriamycin, suggesting the existence of overlap among signaling pathways induced by CD437 and adriamycin. CARP-1 involvement in the regulating cell cycle was further highlighted by the experiment where supplementation of serum resulted in a time-dependent decreased expression of CARP-1 (Fig. 5). Whether CD437, adriamycin, or serum components (such as growth factors) regulate CARP-1 expression by transcriptional, post-transcriptional, or post-translational mechanisms remains to be determined. Our report highlights, for the first time, that CD437 targets pathways converging on topoisomerase II␣ (see Figs. 3, 5, 7, and 8). Because topoisomerases are known to regulate cellular DNA topology (30), it is likely that signaling induced by CD437 interferes with the balance between DNA supercoiling forces. In this context, it is interesting to note that HBC sublines expressing CARP-1 antisense do not have similar viability when treated with adriamycin and etoposide. Because both drugs target cellular topoisomerase II␣, the data shown in Fig.  2 suggest that CARP-1 regulation of signaling by adriamycin utilizes both topoisomerase II␣-dependent as well as -independent pathways. Furthermore, it is of note that CD437 treatment of HBC sublines with reduced CARP-1 did not elicit cellular viabilities equal to their non-treated counterparts (Fig.  2), and moreover, that none of the CARP-1 antisense-expressing sublines was totally resistant to CD437-induced apoptosis (Fig. 3) may point to the existence of additional significant signal transducers. This possibility may also be supported by the observation that CD437-dependent alteration in c-Myc levels in parental HBC cells (Fig. 7B) are quite different when compared with the altered levels of c-Myc noted in the HBC sublines expressing reduced CARP-1 (Fig. 7A).
CARP-1 protein possesses putative DNA-binding (bihelical) and RNA-binding cold-shock motifs. A number of proteins including poly(ADP-ribose) polymerase (PARP), ATP-dependent DNA helicase II p70 subunit, apurinic endonuclease-redox protein, spliceosome-associated protein 145 (SAP 145), DNA repair protein RAD18, DEA(D/H) (Asp-Glu-Ala-(Asp/His)) boxbinding protein 1, and heterogeneous nuclear ribonucleoprotein U possess a similar DNA-binding (bihelical) motif. Accumulating evidence suggests that PARP plays dual roles. In the presence of its substrate NAD and DNA strand breakage, PARP functions as a repair enzyme. In the absence of NAD, PARP promotes activator-dependent transcription by interacting with RNA polymerase II-associated factors (31). PARP also binds transcription enhancer factor 1 and transcription factor AP-2 to increase transcription of muscle-specific genes and AP-2-mediated transcription, respectively (32,33). PARP has also been shown to up-regulate the E2F-1 gene promoter (34) further corroborating its role as a transcription factor. PARPdependent silencing of transcription involves poly(ADP-ribosyl)ation of specific transcription factors, which prevents their binding to the respective DNA consensus sequences and formation of active transcription complexes (35). Whether CARP-1 functions as a DNA-binding transcription factor is unclear. Because CARP-1 is present in the perinuclear compartment (Fig. 6) and also regulates expression of topoisomerase II ␣, its function(s) in regulating DNA topology are less likely to entail FIG. 11. Schematic diagram of apoptosis-promoting and cell growth inhibitory pathways utilized by CARP-1.
direct interactions with the bihelical nuclear DNA. However, given the perinuclear localization of CARP-1, it remains to be determined whether CARP-1 regulates the expression of gene(s) encoded by the bihelical DNAs of organelles such as mitochondria.
In addition, the proteins like Rho transcription termination factor, Y box factor homolog APY1, ribonuclease R, and exoribonuclease II proteins possess an RNA-binding cold-shock protein domain similar to that present in CARP-1. In this context, it is noteworthy that reduced expression of CARP-1 results in altered levels of several mRNAs (Table II). Thus, together with the data indicating perinuclear presence of CARP-1 in HBC (Fig. 6) as well as the megakaryocytic cells in the human bone marrow (not shown), it is likely that RNA-binding cold-shock protein domain of CARP-1 is involved in regulating cellular RNA levels. Thus, CARP-1 may regulate levels of target mRNAs using cis-trans interactions involving the RNA-binding cold-shock motif or other novel domain(s). Whether CARP-1 belongs to a family of proteins that bind DNA, RNA, and/or co-factor(s), and whether its interactions with nucleic acid sequences and/or co-factor(s) play a role in growth arrest and/or apoptosis signaling are subjects of our ongoing investigation.
The mechanism(s) by which CARP-1 protein is reduced by overexpression of antisense RNA (Fig. 4) is not clear. Nevertheless, depletion of CARP-1 resulted in interfering with growth arrest and apoptosis signaling and, thereby, allowed for selection of CD437-resistant sublines. Given that CARP-1 targets multiple cell cycle regulatory proteins (Fig. 7), the selection of sublines expressing reduced CARP-1 without significant alterations of chromosomal contents and/or cell cycle distribution may point to the adaptive abilities of the HBC cells expressing CARP-1 antisense. Whether increased expression of c-Myc and/or cyclin B1 in HBC sublines expressing reduced CARP-1 contributes to prolonged survival of cells in the presence of CD437 or adriamycin remains to be determined.
In conclusion, this study reports identification and characterization of a protein involved in signaling pathways utilized by the agents such as a novel retinoid CD437 and chemotherapeutic drug adriamycin. Expression of CARP-1 induces apoptosis, whereas expression of c-Myc or 14-3-3 interferes with CARP-1-mediated apoptosis. Whether CARP-1 functions as a tumor suppressor protein that regulates biological properties of the cancer cells is currently under investigation.