Stable Form of JAB1 Enhances Proliferation and Maintenance of Hematopoietic Progenitors*

Overexpression of JAB1 is observed in a variety of human cancers, but how JAB1 is involved in tumor development remained to be investigated. Here we analyzed mice with modified Jab1 expression. Mice ectopically expressing a more stable form of JAB1 protein under the control of a constitutive promoter were rescued from the embryonic lethality caused by the Jab1–/– allele and developed a myeloproliferative disorder in a gene dosage-dependent manner. Hematopoietic cells from the bone marrow of Jab1 transgenic mice had a significantly larger stem cell population and exhibited higher and transplantable proliferative potential. In contrast, Jab1+/– mice, which express ∼70% as much JAB1 protein as their wild-type littermates, showed inefficient hematopoiesis. Expression of the tumor suppressor p16INK4a was inversely correlated with that of JAB1, and the oncoprotein SMYD3, a newly identified JAB1 interactor, suppressed transcription of p16 in cooperation with JAB1. Thus, the expression and function of JAB1 are critical for the proliferation and maintenance of hematopoietic progenitors.

Overexpression of JAB1 is observed in a variety of human cancers, but how JAB1 is involved in tumor development remained to be investigated. Here we analyzed mice with modified Jab1 expression. Mice ectopically expressing a more stable form of JAB1 protein under the control of a constitutive promoter were rescued from the embryonic lethality caused by the Jab1 ؊/؊ allele and developed a myeloproliferative disorder in a gene dosage-dependent manner. Hematopoietic cells from the bone marrow of Jab1 transgenic mice had a significantly larger stem cell population and exhibited higher and transplantable proliferative potential. In contrast, Jab1 ؉/؊ mice, which express ϳ70% as much JAB1 protein as their wild-type littermates, showed inefficient hematopoiesis. Expression of the tumor suppressor p16 INK4a was inversely correlated with that of JAB1, and the oncoprotein SMYD3, a newly identified JAB1 interactor, suppressed transcription of p16 in cooperation with JAB1. Thus, the expression and function of JAB1 are critical for the proliferation and maintenance of hematopoietic progenitors.
JAB1 was identified originally as a protein binding to the transcription factors c-JUN and JUN D (1) and later as the fifth (CSN5) of eight (CSN1-8) subunits of the COP9 signalosome (CSN) complex (2), which is located in the nucleus, well conserved from yeast to humans, and involved in a variety of biological responses (3)(4)(5)(6). JAB1 is a multifunctional protein that (i) determines the specificity of transcription factors, such as c-JUN, JUN D, and E2F-1 (1, 7); (ii) mediates the phosphorylation of c-JUN, NFB, and p53 by CSN-associated kinases (2,8); (iii) catalyzes the reaction to remove a ubiquitin-like polypeptide, NEDD8, from the Cullin subunit (deneddylation) (9,10), thereby regulating the activity of Cullin-based ubiquitin ligases, such as the SKP1-Cullin-F box protein complex; and (iv) controls the stability and intracellular distribution of many signaling molecules, including the CDK inhibitor p27 (11), by ill defined mechanisms. However, it remains to be determined whether these different activities were regulated simultaneously or sepa-rately, and were specifically associated with the JAB1 subunit or regulated by the entire CSN complex. JAB1 seems to play an essential role in the control of mammalian cell proliferation and survival. In mice, the level of JAB1 expression is higher in embryos, where vigorous cell proliferation occurs, than in adults (12). Jab1-null embryos die soon after implantation, and Jab1 Ϫ/Ϫ embryonic cells exhibit impaired proliferation and accelerated apoptosis. Jab1 ϩ/Ϫ mice, which express only ϳ70% as much Jab1 protein as their wild-type littermates, were viable and fertile but smaller in size, and Jab1 ϩ/Ϫ mouse embryo fibroblast cells exhibited an inefficient down-regulation of the CDK inhibitor p27 during G 1 and delayed cell cycle progression (13).
Recently, high levels of JAB1 have been found in human cancers. Although a certain level of JAB1 is expressed in normally proliferating and in nonproliferating cells, an increase in JAB1 expression of severalfold is observed in a variety of human cancer cells, such as breast cancer (14,15), malignant lymphoma (16), ovarian tumor (17), melanoma (18), non-small cell lung cancer (19,20), laryngeal squamous cell carcinoma (21), and thyroid medullary carcinoma (22). In hepatocellular carcinoma (23), a high level of JAB1 is associated with an increase in the genomic DNA copy number of the Jab1 locus. In chronic myeloid leukemia (24), JAB1 functions downstream of BCR-ABL kinase through the mitogenactivated protein kinase and phosphatidylinositol 3-kinase pathways and facilitates cell cycle progression. A high level of JAB1 is often correlated with a poor prognosis and sometimes related to a low level of the CDK inhibitor p27. Furthermore, in cases of breast cancer, coordinated amplification of the MYC and CSN5 (JAB1) genes was observed (25). However, the physiological significance of JAB1 overexpression in cancer cells and the molecular mechanism by which JAB1 is involved in the proliferation of cancer cells remain to be examined. In this study, we analyzed mice with a modified JAB1 expression. We demonstrated that the level of JAB1 expression is a critical determinant of the proliferation and maintenance of hematopoietic progenitor cells in vivo, which may explain how the overexpression of JAB1 contributes to tumor development.

EXPERIMENTAL PROCEDURES
Cell Culture, Transfection, and Colony Formation Assay-NIH3T3 mouse fibroblasts and 293T human embryonic kidney cells were cultured in Dulbecco's modified Eagle's medium supplemented with 2 mM glutamine, 100 units/ml penicillin, 100 g/ml streptomycin, and 10% fetal bovine serum. GFP 2 fusion expression vectors were generated based on pMSCV-IRES-puro (Clontech), in which cDNAs were inserted in frame with a GFP sequence (26). The vectors for RNA interference were constructed with the pSilencer expression vector system (Ambion) according to the manufacturer's instructions. Transfection was carried out using a modified calcium phosphate-DNA precipitation method (27). Recombinant retroviruses were produced by transfecting 293T cells with MSCV-based expression vectors, together with a plasmid encoding a 2 helper virus. The supernatants were harvested 48 -72 h post-transfection and immediately used for infection. Bone marrow cells were cultured in Dulbecco's modified Eagle's medium supplemented with 15% fetal bovine serum, 5% WEHI conditioned medium, mouse stem cell factor (50 ng/ml), mouse IL-3 (6 ng/ml), and human IL-6 (10 ng/ml) and infected with culture supernatants twice according to the spin infection procedure in the presence of Polybrene (4 g/ml). For colony formation assays, cells isolated from bone marrow, spleen, and peripheral blood were plated (1.5-2.0 ϫ 10 4 cells/35-mm dish) on a methylcellulose medium (MethoCult GF M3434; StemCell Technologies Inc.). After 7-12 days, colonies containing more than 100 cells were enumerated. For the second and third plating, cells were collected from the methylcellulose medium and replated (1.0 ϫ 10 4 cells/ 35-mm dish) in fresh medium. Some colonies were cytocentrifuged onto glass slides, stained with a May-Grunwald Giemsa solution, and viewed by phase-contrast microscopy in order to evaluate lineage.
Mice-Jab1 ϩ/Ϫ mice and ES cells were generated as described (13). Mouse JAB1 cDNA was combined with a tandem affinity purification (TAP)-tagged sequence, cloned into a pCAG-IRES-puro expression vector (26), introduced into Jab1 ϩ/Ϫ RF8 mouse ES cells by electroporation, and selected in 5 g/ml puromycin. Two TAP-JAB1-positive Jab1 ϩ/Ϫ ES cell lines were microinjected into C57BL/6 embryos, and chimeric males were crossed with C57BL/6 females. The transgene was detected by standard genomic PCR using a pair of primers: 5Ј-AGT GGT ATG GCC CAG AAA ACC TGG GAA TTG-3Ј and 5Ј-CTG CTG GTT CAG CAT CTG TGT ACT AAC ATC AAT CC-3Ј. The genotype of the endogenous Jab1 gene locus was determined as described (13). All mice were maintained in the Nara Institute of Science and Technology Animal Facility in accordance with Nara Institute of Science and Technology guidelines. Blood samples were routinely analyzed with an automated cell counter (F-820 analyzer; Sysmex) and also by inspecting blood smears after May-Grunwald Giemsa staining (Merck). For myelosuppressive treatment, mice were injected with 5-fluorouracil (5-FU) (300 mg/kg of body weight; Kyowa Hakko Kogyo) intravenously, and hematopoietic recovery was monitored in peripheral blood. At specific points in time, mice were sacrificed, femurs were fixed in 4% formaldehyde, and BM sections were stained with hematoxylin and eosin. For the bone marrow transplantation experiment, cells were injected into the veins of NOD.CB17-Prkdc scid /J (NOD-SCID) mice (Jackson Laboratory), and peripheral blood was routinely monitored.
Histological Methods-Tissues were fixed in 4% formaldehyde, embedded in paraffin, cut into 4-m sections, and stained with hematoxylin and eosin. For blood smears, blood was obtained from a tail vein onto glass slides and stained with May-Grunwald Giemsa solution. Blood counts were determined using an automated cell counter. For cytospins, 1.0 ϫ 10 5 cells were cytocentrifuged (500 rpm for 3 min) onto glass slides and stained with May-Grunwald Giemsa solution. Myeloperoxidase activity was detected with the DAB staining kit (Muto Pure Chemicals) according to the manufacturer's instructions. Chloroacetate esterase staining was performed with assistance from Narabyouri Research.
Protein Analyses-Cell lysis, immunoprecipitation, SDS-PAGE, and immunoblotting were performed as described (11,13,24,26). Rabbit polyclonal antibodies to JAB1 and SMYD3 were generated against bacterially produced recombinant polypeptides. A rabbit polyclonal antibody to GFP (BD Living Colors) and mouse monoclonal antibodies to ␥-tubulin (GYU-88) and an HA epitope (clone 12CA5) were obtained from BD Biosciences, Sigma, and Roche Applied Science, respectively.
Quantitative RT-PCR Analysis-Total RNA was isolated and converted to cDNA using ISOGEN reagent (Nippon Gene) and RNase-free Superscript reverse transcriptase (Invitrogen) according to the manufacturer's instructions. The polymerase chain reaction was performed within a linear range described previously (26,28), and the data were normalized to the expression level of ␤-actin for each sample. We confirmed that the reactions were quantitatively performed within a linear range in several control experiments. The following oligonucleotide primer pairs were used: Jab1, 5Ј-GAA GCG CAG AGT ATC GAT GAA A-3Ј and 5Ј-GCC AAT GCT GAG ATT TTG CA-3Ј; ␥-tubulin, 5Ј-ATG AGT GAC GTG GTG GTC CAA-3Ј and 5Ј-CGG TCT GTG GCT ATC AGG TTC A-3Ј; ␤-actin, 5Ј-TCA GCA AGC AGG AGT ACG ATG A-3Ј and 5Ј-CTG CGC AAG TTA GGT TTT GTC A-3Ј. Primer pairs for Ink4a, Arf, and Cip1 were reported previously. Real time PCR was performed with an ABI7700 sequence detection system using a quantitative RT-PCR kit with SYBR Green (Invitrogen) according to the manufacturer's instructions. PCR products were confirmed by agarose gel electrophoresis. All experiments were performed in triplicate.
GST Pulldown Assay-cDNA fragments containing the CSN6 and SMYD3 coding sequences were inserted into the pGEX vector (Amersham Biosciences) in frame with GST. GST-fused proteins were expressed in bacteria and purified as described (11). Crude cell extracts containing HA-JAB1 protein were prepared from COS7 cells transfected with the HA-JAB1 expression vector in an EBC buffer (50 mM Tris-HCl, pH 7.5, 120 mM NaCl, 0.5% Nonidet P-40, and 1 mM EDTA) containing 5 mg/ml aprotinin, 1 mM phenylmethylsulfonyl fluoride, 0.1 mM NaF, 0.1 mM NaVO 4 , and 1 mM dithiothreitol. Binding was performed in the EBC buffer, the protein complexes were washed in the same buffer, and the bound protein was detected by immunoblotting. The same amounts of beads used for the binding assay were subjected to immunoblotting with antibodies to GST to quantitatively examine the GST-fused proteins.
Transactivation Assay-Fragments containing ϳ2 kb of DNA upstream of exon 1␣ of the Ink4a locus were inserted into the pGL3-Basic and pGL3-Enhancer vectors (Promega) to generate luciferase reporter constructs. Cells were transfected with the p16 reporter and the cytomegalovirus-␤-galactosidase plasmid, together with additional SMYD3 and JAB1 expression plasmids. Luciferase and ␤-galactosidase activities were determined using reporter gene assay kits (Roche Applied Science) according to the manufacturer's instructions. The luciferase activity was normalized to that of ␤-galactosidase for each sample.
Chromatin Immunoprecipitation (ChIP) Assay-Cells were cross-linked in 1% formaldehyde for 10 min, and chromosomal DNA was fragmented by sonication. The fixed chromatin samples were subjected to immunoprecipitation using antibodies specific to SMYD3 and JAB1. The sets of primers used for the ChIP assay were mouse Ink4a

Ectopic Overexpression of Stable Form of JAB1 Rescues Jab1
Knock-out Phenotype-In order to analyze the effect of the ectopic expression of JAB1 in a wide range of cell lineages, we constitutively expressed ectopic JAB1 protein in the whole body of mice. In a previous experiment, we were successful in transiently overexpressing JAB1 protein in a variety of cell types (11). However, after selection with a drug marker, few stable cell lines that overexpressed ectopic JAB1 protein were established. Following exhaustive efforts using a variety of expression con-structs (HA-and FLAG-tagged JAB1 as well as an intact JAB1), cell lines, and transfection methods, we finally ended up with a TAP-tagged JAB1 (TAP-JAB1) vector, with which we established a number of human and mouse cell lines overexpressing ectopic JAB1 protein. This is partially due to the stabilization of JAB1 protein by fusion with the TAP tag. Pulse-chase experiments in the presence of cycloheximide revealed that the halflife of endogenous JAB1 was ϳ5.1 h, whereas that of TAP-JAB1 was about 46 h. Encouraged by this result, we introduced TAP-JAB1 into mouse ES cells and successfully established 11 independent ES clones, which stably express ectopic JAB1 protein (Fig. 1A). We picked two independent lines and used them in assays giving basically the same results. The level of TAP-JAB1 in ES cells was ϳ4-fold more than that of endogenous JAB1 protein normally expressed in ES cells. The morphology of ES cells expressing TAP-JAB1 was indistinguishable from that of parental cells, suggesting that TAP-JAB1 expression did not alter the multipotency of ES cells. In fact, after the injection of ES cells expressing TAP-JAB1 into mouse embryos, mice expressing TAP-JAB1 were born. TAP-JAB1 transgenic (Tg) mice appeared normal and healthy at birth and were fertile. The expression of TAP-JAB1 was stably maintained in all organs and cell types tested, which included kidney, lung, spleen, bone marrow, and embryonic fibroblasts (examples of kidney and lung are shown in Fig. 1B). Ectopic TAP-JAB1 expression increased the amount of the COP9 signalosome complex without increasing the amount of endogenous JAB1 protein, suggesting that TAP-JAB1 was incorporated into the COP9 complex in addition to enhancing the small complex ( Fig. 1C), which makes a clear distinction from that of HA-tagged JAB1 protein. HA-JAB1 was preferably incorporated into the small complex (24).
We previously found that Jab1 knock-out mice die in the embryonic stage (they die before embryonic day 6.5-7.5) (13). To investigate whether ectopic TAP-JAB1 can rescue Jab1 Ϫ/Ϫ mice, we first expressed TAP-JAB1 in mice with the Jab1 ϩ/Ϫ genetic background. After intercrossing TAP-JAB1 Jab1 ϩ/Ϫ mice, we found that TAP-JAB1 Jab1 Ϫ/Ϫ mice were born alive (Fig. 1D), with a frequency (18.5%, n ϭ 27) close to that expected from the Mendelian rule. The appearance of these mice was normal and healthy, indicating that the TAP-JAB1 transgene can substitute for the endogenous Jab1 gene. Therefore, together with the result in Fig. 1C, we concluded that TAP tagging did not interfere with the normal function of JAB1 protein.
JAB1-Tg Mice Develop a Myeloproliferative Disorder-TAP-JAB1 transgenic (JAB1-Tg) mice appeared normal at birth and showed little difference in size, behavior, and reproductive ability from their wild-type littermates. After 6 months, however, they started to exhibit abnormal peripheral blood (PB) leukocyte numbers ( Fig. 2A) and anemia (Fig. 2, B and C) and developed a myeloproliferative disease, the population being dominated by granulocytes (Fig. 2E). Interestingly, this occurrence (more than 2 ϫ 10 4 leukocytes/mm 3 in PB) was correlated with the integrity of the endogenous Jab1 locus in a gene dosage-dependent manner (Fig. 2D), because both TAP-JAB1 Jab1 ϩ/ϩ and TAP-JAB1 Jab1 ϩ/Ϫ mice developed this disease, whereas TAP-JAB1 Jab1 Ϫ/Ϫ mice did not, and the frequency was slightly Jab1 Controls Hematopoietic Progenitors OCTOBER 24, 2008 • VOLUME 283 • NUMBER 43

JOURNAL OF BIOLOGICAL CHEMISTRY 29013
higher and the timing of occurrence was a little earlier for TAP-JAB1 Jab1 ϩ/ϩ mice than for TAP-JAB1 Jab1 ϩ/Ϫ animals. (Note that Jab1 ϩ/Ϫ alleles express ϳ70% as much JAB1 protein as Jab1 ϩ/ϩ alleles, so that the difference is subtle between Jab1 ϩ/Ϫ and Jab1 ϩ/ϩ cells, whereas Jab1 Ϫ/Ϫ cells contain no endogenous JAB1 protein.) These results suggest that the TAP-JAB1 transgene was cooperating with the endogenous JAB1 to induce the abnormality. However, once mice developed a myeloproliferative disorder, there was no significant difference in the counts of white blood cells and red blood cells (Fig. 2, A and B) and the type of cells (granulocytes) that dominated in PB (Fig.  2E) between TAP-JAB1 Jab1 ϩ/ϩ and TAP-JAB1 Jab1 ϩ/Ϫ animals.
The mice were sacrificed for autopsy when they developed anemia, less than 8 ϫ 10 6 RBC/mm 3 , or leukocytosis, over 2 ϫ 10 4 cells/mm 3 , or when they apparently became sick. JAB1-Tg mice frequently showed a massive enlargement of the spleen (Fig. 2, F and G), and JAB1-Tg BM appeared pale and hypercellular. Flow cytometric analysis revealed that the c-Kit ϩ Sca-1 ϩ Lin Ϫ (KSL) fraction, which was enriched for the hematopoietic stem cell/progenitor population, was expanded in JAB1-Tg bone marrow (5.27 Ϯ 0.53 ϫ 10 4 to 1.67 Ϯ 0.42 ϫ 10 5 cells/bone marrow, p Ͻ 0.01, n ϭ 5). Fig. 2H shows one example of such an analysis, in which the percentages of the KSL population among all bone marrow cells are shown (0.04% versus 0.25% for those of wild type and JAB1-Tg mice, respectively). Furthermore, the CD34 Ϫ/low population was enriched in the KSL fraction isolated from JAB1-Tg mice (Fig. 2H, 9.0 versus 19%), indicating that JAB1-Tg mice contained a larger pool of hematopoietic stem cell/progenitor cells than wild type animals.
FACS analysis of spleen cells revealed that the number of Gr-1 hi /Mac-1 hi granulocytes increased in spleen (2.30 Ϯ 0.96 ϫ 10 6 to 1.79 Ϯ 0.38 ϫ 10 7 , p Ͻ 0.002, n ϭ 5) of JAB1-Tg mice. Fig.  2I represents one such analysis, in which percentages of the Gr-1 hi /Mac-1 hi population among total spleen cells are shown (1.9 versus 5.7% for wild type and JAB1-Tg mice, respectively). Fig. 2I also shows that the single anti-Mac1 staining revealed a marked difference between the two (8.2 versus 23%). These data show that granulocytes and myeloid progenitor cells were accumulated in the spleen of JAB1-Tg mice. The cytological investigation of the May-Grunwald Giemsa-stained cytospins also confirmed the results, indicating that JAB1-Tg mice develop a progressive myeloid hyperplasia with overproduction of granulocytes. In contrast to this, we observed less influence on lymphopoiesis and erythropoiesis.
In 4 of 4 sick JAB1-Tg mice examined, histological and cytological analyses revealed that a massive number of leukemic cells had infiltrated nonhematopoietic organs, such as the lung, liver, and kidney (examples of liver and lung are shown in Fig. 2, J and K, respectively). Examinations of cell morphology at higher magnification (Fig. 2L for the example of the liver) revealed that the infiltrates were blood cells, some of which were positive for myeloid markers, such as chloroacetate esterase and myeloperoxidase (data not shown).
In contrast, the wild-type or TAP-GFP control mice, which constitutively express ectopic TAP-tagged GFP protein throughout the body, showed little evidence of increased myelopoiesis for up to 13 months (Fig. 2D). Thus, these results indicate that JAB1-Tg mice develop a progressive myeloid hyperplasia with overproduction of granulocytes, which resem-   OCTOBER 24, 2008 • VOLUME 283 • NUMBER 43 bles human CML in the chronic phase, and this phenotype is due to the enhanced activity or function of JAB1 protein.

Jab1 Controls Hematopoietic Progenitors
Increased Proliferation and Extended Survival of JAB1-Tg Hematopoietic Progenitors-To investigate the potential for growth and differentiation of the hematopoietic progenitors expanded in JAB1-Tg mice, cells isolated from bone marrow, spleen, and peripheral blood of wild-type and JAB1-Tg mice were quantitatively analyzed by a colony formation assay in a methylcellulose medium supplemented with a combination of cytokines (IL-6, IL-3, mouse stem cell factor, and erythropoietin) ( Table 1). The cells from BM formed colonies of erythroids, granulocytes, macrophages, granulocyte/macrophages, and all mixed lineages, and colony-forming frequencies of all lineages were increased for cells from JAB1-Tg mice compared with those of wild-type mice (ϳ1.8-fold increase in total) ( Table 1, first and second rows). Cells from control spleen and PB barely formed any colonies, whereas JAB1-Tg spleen and PB cells formed colonies of all lineages (Table 1, fifth and sixth rows). The formation of colonies predominantly consisting of either erythroids, granulocytes, or macrophages in addition to the mixed type suggests that, besides stem cells (or immature progenitors), more committed progenitor cells were also contained in spleen and PB.
We next collected BM-derived cells from all colonies formed in a methylcellulose medium and replated the same number of cells/plate in a new methylcellulose medium to form secondary colonies (Table 1, third row). Again, plating frequencies of all lineages were higher for JAB1-Tg cells, although the overall efficiency was reduced compared with that for the primary plating. In the third round, wild-type cells barely formed colonies, but the cells that originated from JAB1-Tg BMs efficiently formed colonies mainly of the granulocyte/macrophage lineages (Table 1, fourth row). In all cases, the colonies were markedly larger than those of control cells (data not shown).
Because we observed a marked increase in the pool of hematopoietic stem/progenitor cells in JAB1-Tg bone marrow (Fig.  2H), we next examined whether the enhanced potential for proliferation and survival of JAB1-Tg bone marrow cells (Table 1) was attributable to stem/progenitor cells contained in the bone marrow. For this purpose, we purified KSL (both CD34 ϩ/high and CD34 Ϫ/low ) populations by FACS and directly plated them in a semisolid medium in the presence of the combination of cytokines (Table 2). Purified KSL stem/progenitor populations exhibited an extremely high plating efficiency (note that the number of cells plated in Table 2 was one-tenth of that in Table   1), and CD34 Ϫ/low cells always formed more colonies than CD34 ϩ/high cells. All colonies formed were the mixed type containing erythroids, granulocytes, and macrophages in this assay, indicating that the sorted cells were the stem/progenitor cells. Insertion of TAP-JAB1 cDNA consistently enhanced colonyforming activity regardless of the status (ϩ/ϩ, ϩ/Ϫ) of the endogenous Jab1 allele. Thus, ectopic expression of TAP-JAB1 enhanced the potential of proliferation and survival of hematopoietic stem/progenitor cells in vivo.
Enhanced Potential of Hematopoietic Progenitors in JAB1-Tg Mice Is Cell-autonomous-To examine whether the enhanced proliferation and survival is a direct result of the introduction of JAB1 into hematopoietic progenitor cells, we infected primary BM cells freshly isolated from wild-type mice with the retroviruses containing GFP and GFP-JAB1 fusion genes and cultured them in a liquid medium in the presence of combined cytokines. We found that the introduction of JAB1 markedly improved the proliferation and survival of hematopoietic cells compared with the empty vector-infected control cells in an in vitro culture system (data not shown). Encouraged with this result, we further analyzed the potential of the hematopoietic progenitors by conducting a transplantation experiment (Fig.  3). We isolated bone marrow cells from the wild-type and JAB1-Tg mice and injected them into the veins of recipient mice. For this purpose, we used NOD-severe combined immune deficiency (SCID) mice as the host, because they are deficient in B and T cell function and an ideal model for cell transfer experiments. The injection of wild-type BM cells increased white blood cell counts in peripheral blood, and JAB1-Tg BM cells had an even greater effect (Fig. 3A, top). Notably, the CD3-positive T cells, which were virtually absent in NOD-SCID PB, increased in proportion from 0.4 to 2.5% and 5.3% for wild type and JAB1-Tg BM, respectively, indicating that BM cells were successfully transplanted and that the ability

TABLE 2 Increased colony formation of JAB1-Tg KSL cells in vitro
All colonies formed were the mixed type containing erythroids, granulocytes, and macrophages. Note that the number of cells plated was one-tenth of that in Table 1

Jab1 Controls Hematopoietic Progenitors
to produce blood cells was greater for JAB1-Tg BM cells than WT BM cells. At 6 months post-transplantation, the mice were sacrificed and analyzed. Consistent with the results for PB, the spleen was enlarged, and the degree of change was again much more evident for JAB1-Tg BM cells than wild-type cells (Fig.  3A, bottom). Unexpectedly, the KSL population of BM cells from mice transplanted with WT BM cells was only marginally changed, whereas the transplantation of JAB1-Tg BM cells markedly increased the number of KSL cells in BM (Fig. 3, A (middle) and B), consistent with the results of the donor mouse analysis (Fig. 2). Further analysis showed that the majority of JAB1-Tg BM KSL cells expressed low levels of CD34 antigen (Fig. 3C). Genomic PCR analysis confirmed that JAB1-Tg cells were maintained in spleen and bone marrow of transplanted NOD-SCID mice (Fig. 3D). The average lifespan of typical NOD-SCID mice is about 257-269 days, largely due to the development of lymphomas. Wild-type BM transplantation, although it improved the development of lymphoid lineage cells at least in a temporal manner, in the end did not rescue the hematopoietic deficiency of NOD-SCID animals (average life span was ϳ250 days). In contrast, NOD-SCID mice that successfully received a transplantation of JAB1-Tg BM cells showed a recovery of the lymphoid lineage in PB and survived much longer (ϳ400 days) than average NOD-SCID animals, indicating that wild-type BM can temporarily reconstitute blood cells but JAB1-Tg BM has a more profound effect, presumably due to the reconstitution of HSCs, which was absent for WT BM transplantation. Thus, these results imply that the enhanced proliferation and prolonged survival of BM cells in JAB1-Tg mice are due to the ectopic expression of JAB1 in hematopoietic progenitor cells and probably not the result of altered properties in the niche or other types of cells in transgenic animals.
Requirement of JAB1 for the Proliferation and Maintenance of Normal Hematopoietic Progenitors-To examine whether JAB1 functions in normal hematopoiesis, we analyzed JAB1 heterozygous (Jab1 ϩ/Ϫ ) mice, which express ϳ70% of the JAB1 protein found in their wild-type littermates (13). An examination of bone marrow cells by flow cytometric analysis revealed that Jab1 ϩ/Ϫ mice contained a reduced percentage of KSL cells compared with wild-type littermates (0.12 Ϯ 0.027% versus 0.082 Ϯ 0.023%, p Ͻ 0.05) (Fig. 4, A and B). A colony formation assay of BM cells in a methylcellulose suspension culture (Table  1) indicated Jab1 ϩ/Ϫ BM cells to be less efficient at forming colonies, and this is at least partly attributed to hematopoietic stem/progenitor (KSL) populations (Table 2). These results suggest that Jab1 ϩ/Ϫ BM is less capable of hematopoiesis due to a poor maintenance of HSCs.
To further explore the potential of HSCs in Jab1 ϩ/Ϫ mice, we analyzed the hematopoietic recovery after treatment with 5-FU, which induces the apoptosis of actively cycling cells but not quiescent stem cells. To measure the kinetics of recovery, a single dose of 5-FU was administered, and peripheral blood counts were monitored. Jab1 ϩ/Ϫ mice exhibited a more severe leukopenia and slower recovery than wild-type mice (Fig. 4C). Consistent with the white blood cell count, wild-type bone marrow was depleted at day 7 after the injection of 5-FU and almost fully recovered at day 10, whereas the effect of myelosuppression on Jab1 ϩ/Ϫ mice was more severe, and the recovery was hampered (Fig. 4D). Thus, the reduction of JAB1 expression in the heterozygous mice impaired hematopoiesis, which resulted in the death of 33% (2 of 6) of the 5-FU-treated Jab1 ϩ/Ϫ mice, whereas all of the wild-type mice (5 of 5) survived under these conditions. JAB1 Regulates p16 INK4a Transcription in Cooperation with a Novel JAB1-interacting Protein, SMYD3-To obtain an insight into how JAB1 promotes the expansion and maintenance of the hematopoietic progenitor cell population, we analyzed the expression and function of various cell cycle and signal transduction regulators, finding the decreased expression or the functional inactivation of several tumor suppressor proteins in JAB1-Tg cells. Among these, we focused our attention on p16 INK4a , because recent reports show that p16 INK4a plays a critical role in the regulation of the hematopoietic stem cell pool's size and in tumorigenesis (29 -32). A quantitative RT-PCR analysis of RNA isolated from BM cells of wild-type, Jab1-Tg, and Jab1 ϩ/Ϫ mice (Fig. 5, A and B) showed that the level of the p16 INK4a transcript inversely correlated with that of the Jab1 transcript, whereas the levels of other transcripts exhibited marginal change (p15 INK4b ) and little difference (ARF and p21). Similar results were obtained when we used RNA isolated from KSL cells (Fig. 5C). A quantitative genomic PCR analysis proved that the INK4a gene locus in JAB1-Tg cells did not contain extensive deletions (data not shown), suggesting that p16 INK4a is transcriptionally regulated by JAB1.
JAB1 is known to regulate the transcription of many genes through various transcription factors (3,33), but none of the known JAB1-binding proteins have been reported to control the transcription of Ink4a. In order to explore the molecular mechanism by which JAB1 controls Ink4a gene transcription and subsequent cell proliferation and tumorigenesis, we searched for novel JAB1-interacting molecules by yeast two-hybrid screening. After screening 5 ϫ 10 5 clones of the cDNA library generated from K562 mRNA, we identified several genes whose products can interact with JAB1 in yeast cells. One such gene encoded SMYD3, a histone methyltransferase, overexpressed in human colorectal and hepatocellular carcinomas, directly binding to DNA by specifically recognizing the DNA sequence 5Ј-CCCTCC-3Ј or 5Ј-GGAGGG-3Ј, and thereby regulating transcription of the target genes (34). GST pulldown assays showed that SMYD3 is capable of directly binding to JAB1 in vitro, although much less extensively than other known JAB1 interactors, such as CSN6 (Fig.  6A). Co-immunoprecipitation assays using cell lysate revealed that endogenous SMYD3 specifically bound to endogenous JAB1 in living cells (Fig. 6B). However, we again found that only a fraction of SMYD3 was in a complex with JAB1 in vivo.
Searching for the SMYD3-binding motif in the vicinity of the Ink4a gene locus, we found the consensus sequence scattered  . The relative amount normalized to the expression level of ␤-actin in wild-type cells was set to 1.0. C, total RNA was isolated from KSL cells, which were sorted from bone marrow of WT and JAB1-Tg mice by flow cytometer and subjected to a quantitative RT-PCR analysis. Total JAB1 mRNA was measured as a control.
throughout the locus but particularly clustered in front of exon 1␣ of the p16 INK4a coding sequence (Fig. 6C), where the p16 promoter is expected to reside. To test whether SMYD3 cooperates with JAB1 to regulate the p16 promoter activity, we cloned a 2-kb genomic DNA fragment from upstream of exon 1␣ of p16 INK4a into the luciferase reporter plasmid in the presence and absence of the enhancer sequence derived from the SV40 genome. Introduction of the reporter construct into the NIH3T3 mouse fibroblastic cell line, which sustains an extensive deletion around the INK4a gene locus, showed that the 2-kb fragment upstream of exon 1␣ exhibited the promoter activity, and the additional enhancer sequence markedly increased the activity (Fig. 6D). Interestingly, co-transfection of JAB1 and SMYD3 significantly suppressed the promoter activity regardless of the enhancer sequence (Fig. 6D). SMYD3 alone had only a marginal effect, and ectopic expression of JAB1 had a moderate effect. However, the co-expression of JAB1 and SMYD3 was most effective, having a synergistic effect on the suppression of INK4a gene transcription (Fig. 6E).
To elucidate the role of SMYD3 in the JAB1-mediated effect, we designed an siRNA expression vector specific to mouse SMYD3. Introduction of the siRNA vector into mouse cells such as NIH3T3 cells, which expressed a low but detectable level of SMYD3 protein, reduced the SMYD3 level down to ϳ30% of that in the cells transfected with the control vector (Fig. 6F). Next, we introduced the siRNA vector specific to SMYD3 into BM cells isolated from JAB1-Tg mice and cultured them in a methylcellulose semisolid medium in the presence of the combination of cytokines as used in Table 1. Knockdown of SMYD3 in BM cells did not alter the lineage specificity but markedly reduced the efficiency with which colonies formed (from 13.3 Ϯ 1.8 to 5.98 Ϯ 1.1 per FIGURE 6. JAB1 and SMYD3 cooperate to regulate p16 INK4a transcription. A, immobilized recombinant GST proteins (shown at the top of the panels) were incubated with lysates from COS7 cells transfected with the HA-JAB1 expression vector. Bound proteins were detected by immunoblotting using antibody to Jab1. B, endogenous SMYD3 protein was immunoprecipitated from the 293T cell lysate and analyzed by immunoblotting with antibodies to Jab1 (top) and SMYD3 (bottom). C, the positions of putative SMYD3-binding motifs within the Ink4a locus. D and E, NIH3T3 cells were transfected with 1 g of luciferase reporter construct (P-B, pGL3-Basic; P-E, pGL3-Enhancer; P16-B, pGL3-Basic containing a 2-kb fragment upstream of exon 1␣; P16-E, pGL3-Enhancer containing a 2kb fragment upstream of exon 1␣) and 1 g of the cytomegalovirus promoter-driven expression vectors for SMYD3 and GFP-JAB1 together with the cytomegalovirus-␤-galactosidase plasmid. The luciferase activity was normalized to that of ␤-galactosidase for each sample. F, NIH3T3 cells were transfected with the siRNA expression vector specifically designed for mouse SMYD3, selected in puromycin, and analyzed by immunoblotting using antibody to SMYD3. ␥-Tubulin was used as a loading control. SMYD3 expression was reduced to ϳ30% of that in control cells. We used three independent RNA interference constructs with similar results. Representative results are shown. G, JAB1 transgenic bone marrow cells were infected with retroviruses expressing control and SMYD3-specific siRNA (siSMYD3) and cultured in a semisolid medium supplemented with cytokines as in Tables 1 and 2. Colonies containing more than 100 cells were enumerated after 7-12 days. Results are the means Ϯ S.D. of three independent experiments. H, HEK293T cells were transfected with the expression vectors for SMYD3 and Jab1 as indicated at the top, and the fixed chromatin was analyzed by a ChIP assay using an antibody specific to SMYD3 and sets of primers recognizing the 5Ј-flanking regions of the human INK4a and Nkx2.8 genes. I, fixed DNA isolated from Jab1-Tg bone marrow cells was immunoprecipitated with antibodies to SMYD3 and Jab1 and analyzed by a ChIP assay using primers specific to mouse Ink4a. 1000 cells, p Ͻ 0.005) (Fig. 6G) to the level of wild-type BM cells, suggesting that SMYD3 is required for JAB1-mediated enhancement of BM cells.
To examine whether SMYD3 interacts with the INK4a promoter, we performed a ChIP assay using an antibody specific to SMYD3 (Fig. 6, H and I). In human embryonic kidney-derived (HEK293T) cells transfected with the SMYD3 expression vector, we found SMYD3 to be associated with the INK4a promoter region containing clustered SMYD3-binding sites, and JAB1 coexpression did not enhance or reduce the binding (Fig.  6H). Furthermore, we detected the binding of endogenous SMYD3 to the INK4a promoter region in JAB1-Tg BM cells (Fig. 6I). Because SMYD3 possesses histone methyltransferase activity, we performed ChIP assays using antibodies to methylated (Lys 4 and Lys 9 ) histones H3 but detected little difference of histone methylation in the INK4a region (data not shown). This could mean that the SMYD3-JAB1 complex has substrates other than histones, but this hypothesis should be tested further. Taken together, SMYD3 interacts with JAB1, regulates transcription of the INK4a gene, and controls hematopoietic cells.

DISCUSSION
Overexpression of JAB1 in human cancers has been reported (14 -24, 35), and knockdown of JAB1 in cancer cell lines inhibited tumor cell growth (36). However, the functional significance of JAB1 overexpression during the development of tumors was largely unclear. The present study provides a model by which JAB1 participates in tumor development. JAB1 modulates the transcription of p16 INK4a , which plays a key role in cell cycle regulation during tumor development (30,31) and was recently shown to control specific age-associated HSC functions (29,32). There are several factors (polycomb proteins, Tbx2, E2Fs, JUN B, ETS2, MYC, etc.) that control transcription from the INK4a locus (32, 37), but not many are connected with the development of human cancer. Therefore, it is feasible that JAB1 plays a part in the regulation of INK4a connected with tumor development and stem cell aging. However, it is known that inactivation or deletion of the Ink4a locus alone does not result in myeloproliferative disorders (30), suggesting there to be an additional downstream target of JAB1, which, in collaboration with INK4a inactivation, facilitates the proliferation and maintenance of HSCs.
Another point is that JAB1 is involved in a variety of human cancers, but JAB1-Tg mice mostly developed a myeloproliferative disorder. We cannot yet explain this discrepancy in the tissue specificity of cancer development, although it may be that the aberration of other oncogenes or tumor suppressors is required for JAB1 to generate other types of cancers. This seems to be the case with breast cancer, in which Jab1 cooperates with the c-Myc oncogene (25). Another possibility is that because bone marrow cells turn over more rapidly than other types of cells, their susceptibility to transformation is greater. Most JAB1-Tg mice develop a myeloproliferative disorder, but investigation of the survivors may provide answers. In fact, a few mice that survived for ϳ2 years after birth occasionally developed other types of disorders. Extended observation or, alternatively, enhancing the incidence of tumors through treatment with DNA-damaging agents, such as UV and chemicals, will be needed to determine the tumor-related spectrum of JAB1-mediated abnormalities in mice.
Although a variety of human cancers overexpress JAB1 protein, few genetic abbreviations (point mutations or chromosomal translocations) have been found except for breast cancer, which is accompanied by amplification of the JAB1 gene (25), indicating that modulated expression rather than an altered protein structure is critical to the process mediated by JAB1 and that an upstream regulator of JAB1 is likely to be responsible for the overexpression of JAB1. We found in this study that the half-life of JAB1 protein is relatively short (ϳ5 h) and that JAB1 is not susceptible to overexpression in normally proliferating cells. In our experience, TAP tagging was the only way to stably overexpress the JAB1 protein in proliferating cell cultures and in living mice. TAP-tagged JAB1 exhibited a very long half-life (ϳ40 h). Although a TAP tag is relatively large, TAP-JAB1 can substitute for the loss of the endogenous Jab1 allele and, of course, is incorporated into the COP9 signalosome holocomplex. Therefore, TAP-tagged JAB1 protein and cells expressing TAP-JAB1 may be a very useful tool for investigating the posttranslational regulation of JAB1.
In this study, we described the oncogene-related action of JAB1; higher levels of JAB1 expression enhanced the proliferation and maintenance of HSCs and induced a myeloproliferative disorder in mice. Histological analyses of human cancer specimens (14 -24, 35) support the notion that JAB1 does have oncogenic properties. In contrast to the action of JAB1, which is identical to the fifth component (CSN5) of the COP9 signalosome (CSN) complex, we previously showed that the third component (CSN3) of the CSN complex forms a complex with myeloid leukemia factor-1, functions upstream of the E3 ubiquitin ligase COP1, and regulates the expression of the tumor suppressor p53 (26), thereby negatively controlling cell proliferation and tumorigenesis. This places CSN3 as a tumor-suppressing protein. Recently, Leal et al. (38) identified the second component (CSN2) of CSN as a putative tumor suppressor by cellular senescence bypass screening. Thus, different components of the same protein complex function both positively and negatively for tumorigenesis. Each subunit may function positively and negatively for the same signaling cascade. Alternatively, each subunit may have a unique function outside the CSN holocomplex. Whatever the essential function of the COP9 signalosome is, it is beyond the regulation of photomorphogenesis, and that of the mammalian CSN is critical for cell proliferation/survival and the eventual pathway to control tumorigenesis.