The kinase domain of Jak2 mediates induction of bcl-2 and delays cell death in hematopoietic cells.

Granulocyte-macrophage colony-stimulating factor (GM-CSF), interleukin (IL)-3, and IL-5 stimulate DNA synthesis and proliferation and inhibit apoptosis in hematopoietic cells. Multiple signal pathways are activated by binding of these ligands to their receptors, which share a common beta subunit. Janus protein kinase 2 (Jak2) binds to the membrane proximal domain of the beta chain and is phosphorylated on receptor ligation. To explore the role of Jak2 in the regulation of specific signal transduction pathways, we constructed fusion proteins with a CD16 external domain, a CD7 transmembrane region, and a Jak2 cytoplasmic domain. This cytoplasmic domain consisted either of wild type Jak2 (CD16/Jak2-W) or Jak2 mutations with deletions of (a) the amino terminus (CD16/Jak2-N), (b) kinase-like domain (CD16/Jak2-B), (c) kinase domain (CD16/Jak2-C), or (d) amino-terminal and kinase-like domains, leaving the kinase domain (CD16/Jak-K) intact. In contrast to the CD16/Jak2-W fusion protein, which requires cross-linking for activation, CD16/Jak2-N, CD16/Jak2-B, and CD16/Jak2-K were constitutively phosphorylated, and they stimulated Shc phosphorylation and increased binding of STAT to DNA in Ba/F3 cells. Cell lines derived from IL-3-dependent Ba/F3 cells stably transfected with CD16/Jak2-W, CD16/Jak2-N, or CD16/Jak2-B mammalian expression vectors died at a rate similar to that of the parental cells on IL-3 deprivation. In contrast, CD16/Jak2-K cell lines exhibited increased expression of bcl-2 and pim-1 mRNA and maintained their viability when compared with control cell lines. Thus, activation of tyrosine phosphorylation by creating a CD16/Jak2-K fusion is sufficient to activate pathways that prevent cell death.

Signal transduction and growth stimulation by GM-CSF, IL-3, 1 and IL-5 are mediated by a heterodimeric receptor composed of ␣ and ␤ subunits. The ␣ subunit is specific for each cytokine (1), whereas the ␤ subunit is shared in common (2). Binding of the growth factors dimerizes the receptor and induces signal transduction pathways that, in turn, cooperate to induce cell growth. However, the mechanisms by which dimerization of the ␣ and ␤ subunits transmit these signals remain unclear. Various pathways, involving Jak-2, Pim-1, c-Myc, c-Fos, c-Jun, and the Ras/Raf/MAP kinase pathway, have been implicated (3)(4)(5), however, and their regulation and relative contributions to DNA synthesis and proliferation and inhibition of apoptosis are areas of active investigation. Deletion mutagenesis studies suggest that the signal transduction pathways and, perhaps, functional outcomes are specifically associated with the membrane-proximal (1-544) and distal (544 -763) regions of the ␤ subunit (3)(4)(5). The proximal region has been associated with binding of Jak2 and activation of Jak2 kinase (4 -6), induction of c-myc transcription, and stimulation of proliferation. The membrane distal region is associated with activation of the Ras/Raf/MAP kinase pathway, induction of c-fos and c-jun genes, and inhibition of apoptosis (4,5).
Jak2 is a member of a family of protein kinases composed of Jak1, Jak2, Jak3, and Tyk2 that are characterized by a kinase domain in the carboxyl portion, a kinase-like domain (pseudokinase), and a large amino-terminal domain characterized by seven highly conserved regions (7)(8)(9)(10). The aminoterminal domain appears to provide the site of interaction between Jak2 (3,6,11,12) and the ␤ subunit of the hematopoietic growth factor receptors at a region proximal to the membrane that contains a "box 1" sequence and 14 additional amino acids (6). Deletion of this Jak2-binding region prevents transduction of growth factor-induced signals (13)(14)(15). However, deletions of specific portions of intracytoplasmic domains of the ␣ subunit can also block GM-CSF, IL-3, and IL-5 signaling and can inhibit the growth factor-induced activation of Jak2, Pim-1, and c-Myc (16 -19). Transfection of cells with dominant negative forms of the Jak2 kinase blocks both erythropoietin-induced mitogenesis and inhibition of apoptosis, suggesting that Jak2 is necessary for both proliferation and prevention of cell death (20,21).
Although removal of the distal domain of the ␤ subunit does not block IL-3-or GM-CSF-mediated cell growth, the cells undergo apoptosis at a more rapid rate (22). Transfection of Ras into such cells restores their normal growth pattern. If only a small portion of the ␤ subunit is removed (residues 626 -763), then this cell growth defect can also be restored by adding fetal calf serum, which presumably acts through regulation of the Ras pathway (23). Mutation of tyrosine 750 in the ␤ chain reduces the sensitivity of cells to GM-CSF and decreases the tyrosine phosphorylation of Shc (24), suggesting that this is a docking site for signal transduction molecules.
To investigate the role of Jak2 kinase in the regulation of signal transduction pathways, in the absence of other signals generated by the growth factor receptors, we previously fused Jak2 to the external domain of CD16 (the low affinity FcIII IgG receptor) and the transmembrane domain of CD7 (T cell differentiation gp40 protein) (CD16/Jak2) (25,26). This cDNA was encoded in a vaccinia virus that was used to infect growth factor-dependent murine Ba/F3 cells to generate high levels of fusion protein. Cross-linking with anti-CD16 antibody stimu-lated the kinase activity of the fusion protein but did not cause activation of endogenous Jak2 or phosphorylation of the ␤ chain. Activation of the CD16/Jak2 fusion protein stimulated increased phosphorylation of Shc and p145 (SHIP) (27) and increased expression of c-fos, c-jun, and jun-B mRNAs, caused only minor changes in MAP kinase activity, and did not increase cellular levels of c-myc mRNA. We concluded that activation of Jak2 is sufficient to generate some, but not all, of the signals that arise upon binding of growth factors to their receptors in factor-dependent hematopoietic cells.
In the present study, we further examined the role of specific domains of the Jak2 protein in individual signal transduction pathways and, by creating deletion mutants of the Jak2 protein and cloning them in frame with CD16/CD7, determined whether activation of Jak2 is sufficient to sustain the growth of hematopoietic cells. Recombinant vaccinia viruses were prepared for rapid, high level expression of the fusion proteins, whereas mammalian expression vectors were used to establish stably transfected cell lines for analysis of biologic effects. Deletion of the amino-terminal two-thirds of the Jak2 protein, leaving only the protein kinase domain, did not change the pattern of signal transduction. However, in contrast to wild type fusion protein, which requires cross-linking for activity, the chimeric protein encoding only the kinase domain was constitutively fully active. In growth factor-deprived stably transfected cell lines, expression of the chimeric protein encoding only the kinase domain resulted in maintenance of cell viability, delayed cell death, and increased expression of pim-1 and bcl-2 mRNA.

MATERIALS AND METHODS
Construction of Plasmid-The construction of the cDNA for CD16/ Jak2-W (wild type Jak2) and its cloning into a recombinant vaccinia virus transfer vector have been described previously (26). The cDNA for CD16/Jak2-N (N-terminal deletion) was created by replacing the MluI-NcoI fragment (the region coding for residues 1-717) with a polymerase chain reaction product encoding residues 536 -717 with a MluI site in the appropriate reading frame at the 5Ј end. The cDNA for CD16/ Jak2-B (pseudokinase deletion) was created by removing the BglII-BglII fragment (the region coding for residues 523-746) and religation. The cDNA for CD16/Jak2-C (kinase domain deletion) was made by replacing the NdeI-NotI fragment (the region coding for residues 932-1129) with a polymerase chain reaction product encoding residues 932-999 with a terminal codon and the NotI site at the 3Ј end. The cDNA for CD16/Jak2-K (both N-terminal homology domain and kinase-like domain deletion) was created by replacing the MluI-NotI fragment (the entire Jak2 coding region) with a polymerase chain reaction product encoding residues 825-1129 with a MluI site in the appropriate reading frame at the 5Ј end and a NotI site at the 3Ј end. Correct polymerase chain reaction and cloning were confirmed by dideoxy-DNA sequencing. For establishment of stably transfected cell lines, the HindIII-MluI fragment (the region coding for CD16/CD7) and the MluI-NotI fragment (the region coding for wild type Jak2 or Jak2 deletion mutants) were ligated into a HindIII-NotI site of the pcDNA3 expression plasmid (Invitrogen, San Diego, CA).
Construction of Recombinant Vaccinia Virus-Recombinant vaccinia viruses were constructed as described previously (26). Recombinant vaccinia virus transfer vector plasmids were introduced into BSC-40 cells by electroporation, followed by infection with wild type vaccinia virus. After a 24-h culture, viruses were harvested. Cells containing recombinant viruses were selected with guanine phosphoribosyltransferase and thymidine kinase selection. The presence of the recombinant vaccinia viruses was confirmed by polymerase chain reaction.
Infection with Recombinant Vaccinia Virus and Cross-linking-Ba/F3 cells were infected for 1 h in serum-free Dulbecco's modified Eagle's medium with recombinant vaccinia virus at a multiplicity of infection of 100. After a 12-h culture, infected cells were deprived of conditioned medium and incubated for 4 h in RPMI 1640 medium containing 0.5% albumin. Cells were incubated with 5 g/ml of anti-CD16 antibody F(abЈ) 2 for 2.5 min, followed by incubation at 37°C with 25 g/ml anti-mouse IgG F(ab)Ј 2 for the indicated times.
Cell Growth and Transfection-The murine IL-3-dependent cell line, Ba/F3, was maintained in RPMI 1640 medium with 10% fetal calf serum, 10 mM 2-mercaptoethanol, 20 mM HEPES, pH 7.8, and 10% WEHI-3B conditioned medium as a source of IL-3. Ba/F3 cells were transfected by electroporation (960 microfarads; 350 V) with 30 g of plasmid DNA. Selection with G418 (1 mg/ml) was initiated 48 h after electroporation, and G418-resistant clones were selected by limiting dilution. The expression of the cDNA products was examined by Western blot analysis using anti-Jak2 antibody and flow cytometric analysis using anti-CD16 antibody.
Antibodies-The anti-Jak2 antibody used for immunoprecipitation was raised against residues 747-1129 (6). The anti-Jak2 antibody raised against residues 758 -776 was purchased from Upstate Biotechnology, Inc. (Lake Placid, NY), and the anti-Jak2 antibody (c-20) raised against residues 1110 -1129 was purchased from Santa Cruz Biotechnology (Santa Cruz, CA). Monoclonal anti-CD16 antibody 3G8 F(ab)Ј 2 was purchased from Medarex, Inc. (Annandale, NJ) and used for crosslinking, immunoprecipitation, and flow cytometric analysis. Anti-phosphotyrosine 4G10 was purchased from Upstate Biotechnology, Inc. Anti-STAT1 antibody and affinity-purified polyclonal antibody to Shc were obtained from Transduction Laboratories (Lexington, KY). Anti-STAT3 antibody and anti-STAT5 antibody were purchased from Santa Cruz Biotechnology. Immunoprecipitation and immunoblotting were performed as described previously (26).
Electrophoretic Migration Shift Analysis (EMSA)-Nuclear extracts were prepared, and EMSA was carried out as described previously (26). For the supershift assay, 1 g of specific antibody was added to the nuclear extract. Incubation was carried out at room temperature for 20 min prior to the addition of radiolabeled oligonucleotide.

Measurement of [ 3 H]Thymidine
Incorporation-Cells (5 ϫ 10 4 ) were cultured in RPMI 1640 medium and 10% fetal calf serum in the absence of conditioned medium for 5 h. Cells were either cross-linked with 1 g of anti-CD16 antibody F(ab)Ј 2 or 10 g of anti-mouse IgG F(ab)Ј 2 or stimulated with 10% WEHI-3B conditioned medium for 17 h. Subsequently, the cells were pulse-labeled with 1 Ci of [ 3 H]thymidine for 4 h prior to harvest.
Northern Blots-Northern blots were carried out as described (26). After hybridization, identical filters were stripped and reprobed with c-Fos, c-Jun, JunB, and ␣-tubulin. The pim-1 cDNA (1.0-kilobase XhoI/ HindIII fragment) and the bcl-2 cDNA were gifts, respectively, of Dr. T. Meeker (University of Kentucky) and Dr. Michael Ruppert (University of Alabama at Birmingham).

Expression of CD16/Jak2 Fusion Proteins in Recombinant
Vaccinia Virus-infected Ba/F3 Cells-To explore the possibility that specific signals are regulated by different domains of the Jak2 protein, we deleted specific segments of the protein and fused these deletion mutants in frame with CD16/CD7. The segments deleted were as follows ( Fig. 1): the region of the protein that interacts with the ␤ subunit of the receptor located at the amino terminus of the Jak2 protein and encompassing FIG. 1. A schematic diagram of CD16/Jak2 fusion proteins. cDNAs were constructed as described under "Materials and Methods." CD16/Jak2-W contains the external domain of CD16, the transmembrane domain of CD7, and wild type Jak2 coding sequences. CD16/ Jak2-N lacks the amino-terminal one-third of the Jak2 protein, and CD16/Jak2-B lacks the majority of the kinase-like domain, whereas CD16/Jak2-C lacks the carboxyl-terminal portion of the kinase domain. CD16/Jak2-K contains only the Jak2 kinase domain, lacking both the amino-terminal one-third of the protein and the kinase-like domain.
To obtain rapid, high level expression of the fusion proteins, the cDNAs were cloned into vaccinia virus, which was then used to infect IL-3-dependent Ba/F3 murine hematopoietic cells. The expression of fusion proteins of the correct size was confirmed by immunoprecipitation and Western blotting using a Jak2-specific antibody (Fig. 2, A and B). On Western blot analysis, the reactivity of antibodies specific for the carboxyl terminus of the kinase-like domain or the Jak2 kinase domain confirmed the expression of the kinase-like domain, but not the kinase domain, in the CD16/Jak2-C fusion protein (Fig. 2, A  and B). Fluorescence-activated cell-sorting analysis using an antibody specific for CD16 demonstrated that the fusion proteins were expressed at equivalent levels on the cell surface (data not shown). Proteins of lower molecular weight, which represent nonglycosylated, intracellular forms of the fusion proteins, were also expressed ( Fig. 2C, major band below the arrow) as described previously (28). The levels of intracellular fusion protein appear to correlate with the multiplicity of viral infection. With lower protein expression, similar levels are expressed on the cell surface, but smaller amounts of protein are seen intracellularly (data not shown).
The CD16/Jak2-K Fusion Protein Stimulates Markedly Higher Levels of Constitutive Protein Phosphorylation than Other Fusion Proteins-Jak2 is autophosphorylated on tyrosines upon ligation of the hematopoietic growth factors by their receptors. To investigate the effect on autophosphorylation caused by deleting specific domains of Jak2, we stripped the Western blot shown in Fig. 2, A and B, and reprobed the filter with anti-phosphotyrosine antibodies. The low molecular weight, intracellular, unglycosylated forms of the CD16/ Jak2-N, -B, and K fusion proteins (Fig. 2C, arrows) were constitutively phosphorylated, but their phosphorylation state was unaffected by cross-linking of the fusion proteins. As previously shown, cross-linking of the CD16/Jak2-W fusion protein resulted in phosphorylation of the glycosylated protein (Fig. 2C, arrow), which was not phosphorylated in the absence of crosslinking. As expected, the CD16/Jak2-C fusion protein, which lacks the kinase domain was not phosphorylated whether cross-linked or not. Autophosphorylation of the glycosylated forms of the CD16/Jak2-N, -B, and -K fusion proteins was not markedly increased by cross-linking. Only the CD16/Jak2-B fusion protein, in which the kinase-like region is deleted, showed a slight increase in phosphorylation on cross-linking. Significantly, although the amounts of protein were equivalent ( Fig. 2A), the level of constitutive tyrosine phosphorylation was greatly enhanced in the CD16/Jak2-K fusion protein compared with the other fusion proteins.
To investigate the levels of cellular protein phosphorylation, the identical amounts of cell extracts used to investigate CD16/ Jak phosphorylation (Fig. 2, A-C) were subjected to anti-phosphotyrosine immunoblotting. As previously demonstrated, cross-linking of the CD16/Jak2-W fusion protein resulted in the phosphorylation of proteins with molecular masses of 145, 125, 97, 93, 67, and 54 kDa (Fig. 2D). Proteins of identical molecular weights have been shown to be phosphorylated after the addition of IL-3 and GM-CSF to factor-dependent cell lines (29 -31). Expression of the CD16/Jak2-C fusion protein did not result in significant phosphorylation of these substrates, either constitutively or after cross-linking. In multiple experiments, the base-line phosphorylation in the non-cross-linked CD16/ Jak2-W was variable, being similar to the low levels seen in CD16/Jak2-C containing cells.
In contrast, the expression of the CD16/Jak2-N, -B, or -K fusion proteins resulted in constitutive phosphorylation of proteins of similar molecular weights to those phosphorylated by activation of the CD16/Jak2-W fusion protein. However, because of the prominent phosphorylation of the CD16/Jak2 fusion proteins, it was not possible to identify all of these substrate proteins in the CD16/Jak2-N-and CD16/Jak2-N-Bcontaining cells. As demonstrated for the autophosphorylation of CD16/Jak2-B, cross-linking induced a slight increase in the level of substrate phosphorylation.
Although identical amounts of cell extracts were loaded on this SDS gel, the level of tyrosine phosphorylation of all substrates was greatly enhanced in the cells expressing the CD16/ Jak2-K fusion protein. Because the levels of tyrosine phosphorylation were so elevated, it was difficult to identify the molecular weight of the specific phosphorylated protein extracts. Immunoblots carried out using lesser amounts of proteins demonstrated that the identical proteins to those seen in the CD16/Jak-W Ba/F3 cells (molecular mass 145, 125, 97, 93, 67, and 54 kDa) were phosphorylated (data not shown). Therefore, the infection of Ba/F3 cells with a vaccinia virus-expressing fusion protein containing only the kinase domain of Jak2 induced the constitutive phosphorylation of proteins with molecular weights identical to those of the proteins phosphorylated after the addition of GM-CSF.
Shc Is Constitutively Phosphorylated by CD16/Jak2 Deletion Fusion Proteins-Although the patterns of phosphorylation are complicated, close examination of the anti-phosphotyrosine immunoblots does not disclose major differences in the range of proteins constitutively phosphorylated by the Jak2 deletion mutants. To determine whether an identical protein is phosphorylated by each of the mutants, Shc protein phosphorylation was examined. Shc is phosphorylated after the binding of hematopoietic growth factors to their receptors, suggesting that Shc is a natural target of the Jak2 cascade of protein kinases. Immunoprecipitation of Shc before and after crosslinking indicated that cross-linking of CD16/Jak2-W, but not the CD16/Jak2-C, fusion protein resulted in phosphorylation of Shc (Fig. 3). On the other hand, expression of the CD16/ Jak2-N, -B, and -K fusion proteins resulted in constitutive phosphorylation of Shc, which was not increased by crosslinking. The phosphorylation of Shc did not occur simply as a result of fusing a tyrosine kinase to CD16, since the CD16/ Zap70, which was expressed at the same level as the other fusion proteins (data not shown), did not highly phosphorylated Shc. The p145 protein, SHIP, which is associated with Shc (27), was similarly phosphorylated, requiring cross-linking of the CD16/Jak2 wild type but not the other fusion proteins (data not shown). Reprobing of this Western immunoblot demonstrates that the CD16/Jak2 protein was coimmunoprecipitated with Shc (Fig. 4).
To examine whether these fusion proteins could phosphorylate specific members of the STAT family, the activity of individual STAT family members was examined. Individual members of the STAT family of proteins are activated after the addition of specific growth factors to cells (32)(33)(34), and phosphorylation is mediated by specific members of the Jak family of proteins. These transcription factors require tyrosine phosphorylation for dimerization and DNA binding. The addition of GM-CSF to factor-dependent cell lines stimulates the activity of STAT5 while having no effect on STAT1. To evaluate the specificity of regulation of this family of proteins by the CD16/ Jak2 chimera, cell extracts were subjected to electrophoretic mobility shift assay (EMSA) using a interferon-␥ activation site or ␥-activated site (GAS oligomer) that is capable of binding multiple members of the STAT family (35). As we reported previously (26), cross-linking of the CD16/Jak2-W protein led to STAT protein binding to DNA (Fig. 5A). In contrast, crosslinking of either the CD16/Jak2-C or the CD16/Zap70 had no effect on these assays (Fig. 5A). Expression of the CD16/ Jak2-N, -B, and -K fusion proteins resulted in a band shift, which was not affected by cross-linking, although some slight variability was seen in this assay (Fig. 5A). To determine which of the STAT family of proteins was activated by these fusion proteins, antisera to STAT1, -3, and -5 were used to retard the EMSA complex. Using this technique, the addition of conditioned medium (IL-3) to Ba/F3 cells stimulates the activation of STAT5, whereas the addition of interferon-␥ activates STAT1 (36) (Fig. 5B). Antisera to STAT5 but not antisera to STAT1 and STAT3 supershifted the bands induced by cross-linking of the CD16/Jak2-W fusion protein (Fig. 5B), as well as the bands constitutively induced by the expression of CD16/Jak2-N, -B, and -K fusion proteins (data not shown). Therefore, each of the Jak2 fusion proteins was able to activate STAT5 and not STAT1, thus mimicking the biologic effects of binding of GM-CSF or IL-3 to its receptor. Deletions of the amino-terminal and kinase-like domains of Jak2 led to the constitutive activation of STAT DNA binding activity.
Expression of CD16/Jak2 Fusion Proteins in Stably Transfected Ba/F3 Cell Lines-In the experiments above, vaccinia virus vectors were used to express the fusion proteins in Ba/F3 cells, permit high levels of expression of these proteins, and facilitate analysis of the phosphorylated substrates. Vaccinia virus infection affects Ba/F3 cell growth; therefore, cells prepared using this technique could not be used for analysis of the effects of the fusion proteins on cell growth. Each of these fusion proteins was cloned into a mammalian expression vector containing a CMV promoter, which was used to transfect Ba/F3 cells. Clones were selected and expanded to establish stably transfected cell lines, and three or more independent isolates of each fusion protein expressing clones were characterized. The presence of the fusion proteins in these cells was confirmed by immunoprecipitation with Jak2-specific antibodies, followed by Western blotting using Jak2-specific antibodies (Fig. 6). Because background proteins block the ability to easily detect the CD16/Jak2-N3 fusion, fluorescence-activated cell-sorting analysis using anti-CD16 antibody was used to confirm the equivalent surface expression of the fusion proteins (data not shown). As with the recombinant virus-infected cells, only the CD16/Jak2-W and CD16/Jak2-B fusion proteins exhibited increased tyrosine phosphorylation upon cross-linking, and the CD16/Jak2-K, -N, and -B mutants were constitutively phosphorylated. The level of phosphorylation of the CD16/Jak2-K fusion protein was markedly enhanced compared with that of other fusion proteins. CD16/Jak2-K clones exhibited constitutive activation STAT and tyrosine phosphorylation of Shc (data not shown). Phosphorylation of STAT by the CD16/Jak2-N and -B clones was reduced when compared with the CD16/Jak2-K clones. In stably transfected cell lines, as in recombinant virusinfected cells, it is again evident that the non-kinase regions of Jak2 play an important role in regulating Jak2 protein kinase activity.

IL-3-deprived Ba/F3 Cells Stably Transfected with the CD16/Jak2-K Fusion Protein Incorporate [ 3 H] Thymidine and
Exhibit Delayed Cell Death-Ba/F3 cells depend on the presence of IL-3 to suppress apoptosis and stimulate DNA synthesis and cell proliferation. To determine whether Jak2 is capable of mediating all, or some, of the biologic effects induced by binding of the hematopoietic growth factors, the Ba/F3 cell lines expressing the fusion proteins were deprived of conditioned media (IL-3). Parental cells deprived of conditioned medium were unable to incorporate thymidine, but thymidine uptake was stimulated by the addition of conditioned medium (Fig. 7A). In contrast, the CD16/Jak2-K cell line (K-16) incorporated thymidine in the absence of conditioned medium, resulting in a slight increase on cross-linking (Fig. 7A). Similar results were obtained with all CD16/Jak2-K clones (K-1, K-4, and K-18). Cell lines expressing the other fusion proteins behaved like the parental Ba/F3 cells in that they were unable to incorporate [ 3 H]thymidine when deprived of conditioned medium. The possibility that the cell line expressing the CD16/ Jak2-K (K-16) produced growth factors that stimulated thymidine uptake was excluded by the absence of thymidine uptake in parental Ba/F3 cells incubated with conditioned medium obtained from the K-16 cells.
We next tested the effect of the fusion proteins on cell death as estimated by trypan blue exclusion. The CD16/Jak2-W, -B, and -N cell lines lost viability at the same rate as the IL-3deprived parental cells (Fig. 7B). However, all four of the CD16/ Jak2-K-transfected cell lines (K-1, K-4, K-16, and K-18) lost viability more slowly. Furthermore, the IL-3 deprived CD16/ Jak2-K cell lines lost viability more rapidly than parental cells treated with conditioned medium.
Expression of CD16/Jak2-K Increases the Level of pim-1 and bcl-2 mRNA-To obtain clues as to which genes might be involved in the factor-independent survival of cells containing the CD16/Jak2-K fusion protein, we examined the level of expression of various mRNAs by Northern blot analysis of clone K-16 cells after growth factor deprivation, cross-linking, and growth in 10% FCS (Fig. 8A). When compared with growth factor-deprived parental cells, deprived K-16 cells demonstrated slight elevations in jun-B, but not c-fos or c-jun mRNA. As in parental cells, cross-linking induced an increase in the level of expression of c-fos, c-jun, and jun-B mRNA, although the level of expression of jun-B mRNA was lower than that in parental cells, whereas that of c-jun was slightly higher (Fig.   8A). As in parental cells, the addition of conditioned medium (IL-3) to clone K-16 cells stimulated increases in c-jun, jun-B, and c-fos mRNA.
To approximate the conditions used to assess cell death of the CD16/Jak2-K cells, 10% FCS was added to the K-16 cells for 12 h (Fig. 8A, lanes 7-10). When compared with factordeprived K-16 cells, the addition of 10% FCS appeared to result in a slight elevation in the levels of junB mRNA, whereas the levels of c-jun and c-fos mRNA were unchanged. Thus, the CD16/Jak2-K cells contain low levels of specific mRNAs associated with growth. Importantly, although the level of tyrosine phosphorylation of protein substrates does not appear to increase after cross-linking of this protein kinase fusion protein, substantial increases in the cellular levels of mRNAs were still seen.
Pim-1 expression is regulated by the addition of GM-CSF (37), and the expression of this gene appears to be controlled by a domain of the ␤ chain close to the plasma membrane (5). Since Jak2 binds to a similar region of the ␤ chain, we evaluated the levels of pim-1 mRNA in CD16/Jak2 cells. The addition of conditioned medium (IL-3), but not FCS, to deprived parental Ba/F3 cells markedly elevated the level of pim-1 mRNA (Fig. 8B). CD16/Jak2-K cells growing in FCS, in the absence of conditioned medium (IL-3), maintained continuously elevated levels of pim-1 mRNA, similar to those observed in the hormone-treated parental cells. Thus, the CD16/Jak2-K fusion protein is capable of stimulating increases in pim-1 FIG. 5. EMSA analysis of virally infected Ba/F3 cells expressing fusion proteins. A, Ba/F3 cells infected with recombinant vaccinia viruses (VCV) were growth factor-deprived for 4 h and either cross-linked with anti-CD16 antibody for 15 min (ϩ) or left untreated (Ϫ). Nuclear extracts were isolated, and DNA binding complexes were assayed by EMSA using a GAS oligomer (see "Materials and Methods"). B, after growth factor deprivation for 4 h, Ba/F3 cells were stimulated for 15 min either with conditioned medium (CM) containing IL-3 or ␥-interferon. Ba/F3 cells infected with CD16/Jak2-W recombinant vaccinia virus were growth factor-deprived for 4 h and cross-linked with anti-CD16 Ab for 4 h. Nuclear extracts from these cells were analyzed by EMSA (see "Materials and Methods"). Anti-STAT antibodies were added to nuclear extracts for 20 min on ice prior to EMSA analysis. mRNA levels.
The Bcl-2 family of proteins plays a role in the regulation of apoptosis (38), and overexpression of Bcl-2 prevents hematopoietic cells from undergoing apoptosis (39). Using Northern blots, we evaluated the levels of bcl-2 mRNA in two different clones of CD16/Jak2-K cells grown in 10% FCS (Fig. 8B). Parental cells incubated in either 0.5% albumin or 10% FCS expressed low levels of bcl-2 mRNA. The addition of conditioned medium to parental cells for 12 h markedly elevated the levels of bcl-2 mRNA. Strikingly, the levels of bcl-2 mRNA were also elevated in both cell lines, K-16 and K-18 (K-1 and K-4 data not shown), containing the CD16/Jak2-K fusion protein. Therefore, cells expressing the fusion protein that encodes the Jak2 kinase domain constitutively express Bcl-2 protein, which in turn inhibits cell death.

DISCUSSION
The addition of GM-CSF and IL-3 to factor-dependent cells activates Jak2 protein kinase and stimulates specific signal transduction pathways leading to cell growth. However, the exact role of Jak2 activation in either the stimulation of specific phosphorylation events or in the control of specific aspects of cell growth, such as stimulation of DNA synthesis and proliferation or suppression of apoptosis, remains ill-defined. To clarify the function of Jak2 in these two processes, we have employed a CD16/Jak2 fusion protein that can be activated by cross-linking with CD16 antibodies. Under physiologic circumstances, Jak2 is activated when interacting with domains of hematopoietic growth factor receptors that are physically close to the plasma membrane. Similarly, in this model system, activation of Jak2 occurs physically adjacent to the plasma membrane. Previously, we demonstrated (26) that activation of this fusion protein neither stimulates ␤ subunit phosphorylation nor causes activation of endogenous Jak2, suggesting that this fusion protein functions independently of the hormone receptor.
In the present studies, we used deletion mutations of Jak2 fused to CD16 to analyze the roles of the various domains of Jak2 in regulating phosphorylation of specific substrates. In these experiments, viral expression of CD16/Jak2 fusion gave large amounts of intracellular fusion protein, which is tyrosinephosphorylated, with the level of intracellular protein being directly related to the multiplicity of infection. At lower multiplicity of infection (26), the intracellular levels of fusion proteins drop, while similar levels of protein are expressed on the plasma membrane. In contrast to results obtained with the Jak2 wild type fusion protein, deletion of the B or N region of the Jak2 protein allowed constitutive phosphorylation of these substrates, including the fusion protein itself in the absence of CD16 crosslinking. The proteins phosphorylated by these deletion chimeras (for example, Shc) are known substrates of Jak2 that are modified in response to the binding of hematopoietic growth factors. These fusion proteins activate STAT5 specifically and not STAT1, demonstrating that these fusion proteins are not phosphorylating cellular substrates randomly. One explanation for the ability of these fusion proteins to modify Shc and STAT proteins is that the fusion kinase is phosphorylating a transmembrane hormone receptor protein that acts as a docking site for the Shc and STAT. Recent data from other laboratories suggests that STAT5 directly binds Jak2, which contains multiple phosphorylation sites. 2 The direct coprecipitation of Jak2 and Shc has been demonstrated in a cell line that re-2 T. Hirano, personal communication. sponds to erythropoietin (40). As well, if the ␤ subunit of the GM-CSF receptor is deleted such that it does not contain any tyrosines, the addition of GM-CSF can still activate STAT5 activity (41), suggesting that additional proteins may be involved in STAT5 activation.
Deletion of neither the amino terminus nor the kinase-like region of Jak2 appears to change the general substrate specificity of the fusions, suggesting that these regions do not play a major role in defining the substrate specificity of these protein kinases within the CD16 fusion protein. The constitutive and enhanced phosphorylation induced by the CD16/Jak2 fusion suggests that the amino-terminal half and the kinase-like domain of the Jak2 protein might function as a regulatory region that inhibits Jak2 protein kinase activity. It is possible that Jak2 activity is normally inhibited by the amino-terminal half of the protein and that binding of Jak2 through the aminoterminal half of the protein to hematopoietic receptors relieves this inhibition. Alternatively, growth factor-induced phosphorylation of specific tyrosines within the amino-terminal half of the Jak2 protein might allow the docking of a phosphatase that inhibits the activity of the protein kinase. The PTP1C phosphatase binds to a tyrosine in the erythropoietin receptor tail and inhibits the activity of this receptor (42), with removal of this binding site greatly increasing the activity of the erythropoietin receptor (42). In addition, the two-hybrid yeast system has been used to show that a sequence in the kinase-like domain from the atrial natriuretic peptide receptor binds a phosphatase (43), suggesting that deletion of the kinase-like domain in Jak might remove a phosphatase binding site. The observation that a point mutation amino-terminal region of a Drosophila Jak-like protein leads to hyperactivation of this protein (44) again suggests that this portion of the molecule is important in regulating protein kinase activity. Our results support the idea that the amino terminus of Jak2 plays an important role in regulating Jak2 kinase activity.
We took advantage of the observation that CD16/Jak2 kinase domain fusion is constitutively highly active to examine the effects of this fusion protein on the growth of Ba/F3 cells. Cells expressing the CD16/Jak2-K fusion, but not the B or N deletion, demonstrated a delay in cell death after withdrawal from IL-3. This result was demonstrated in four independent clones, suggesting that blocking of cell death was not obtained because of clonal variation. These cells, but not those containing the B or N deletion, evidenced elevated levels of pim-1 and bcl-2 mRNAs that normally were found only in growth factor-treated cells.
The importance of pim-1 is suggested by the observation that bone marrow-derived mast cells purified from mice in which the pim-1 gene has been "knocked out" grow less well in IL-3 than control cells (45), suggesting an important role for pim-1 in regulating IL-3-induced growth. That Jak2 can regulate levels of Pim-1 has been suggested previously on the basis of the observation that erythropoietin receptors that cannot stimulate Jak2 kinase activation also cannot elevate the levels of pim-1 mRNA (46). The importance of pim-1 in hormone-mediated growth is demonstrated by the observation that dominant negative STAT5 inhibits the expression of Pim-1 and partially blocks the IL-3-dependent growth in Ba/F3 cells (34). Thus, the constitutive activation of STAT5 induced by the CD16/Jak2-K may mediate the induction of Pim-1.
Our results demonstrate that constitutive Jak2 activation alone (accomplished with our fusion proteins) is sufficient to activate the transcription of the bcl-2 and pim-1 mRNAs but is not sufficient to allow for continued cell growth. These findings are consistent with the previous observation that a dominantnegative Jak-2 mutation can block cell growth (20,47) and hormone-induced rescue of cells from apoptosis (21). Other researchers have suggested that three pathways are necessary to stimulate factor-independent cell growth (48,49). In these experiments, overexpression of Bcl-2 was not sufficient to induce factor-independent growth, but it was sufficient to inhibit FIG. 8. Regulation of specific mRNAs by K-16 clone with and without cross-linking. A, Northern blot demonstrating the effect of cross-linking on early gene mRNA levels. Parental or K-16 Ba/F3 cells were growth factor-depleted for 4 h in medium plus 0.5% bovine serum albumin (Ϫ) and then either treated with conditioned medium (CM) or subjected to cross-linking or treatment with FCS for the indicated time periods. mRNA was extracted and Northern blotted with the probes as described under "Materials and Methods." B, Northern blot of bcl-2 and pim-1. Cells were treated similarly to A for the times indicated, and mRNA was extracted and Northern blotted with the indicated probes. cell death (48,49). The addition of either c-Myc or p56 lck to the cells carrying the Bcl-2 protein made these cells growth factorindependent and suggested that three stimuli were necessary for cell growth: increased Bcl-2, increased c-Myc, and the presence of an activator of the MAP kinase pathway, either Ras or Lck (48,49). The mechanism by which both cell growth and cell death were delayed in our cells may be partly related to the elevation of bcl-2 mRNA. It is therefore likely that expressing additional genes, e.g. c-myc or p56 lck , in the Ba/F3 cells containing CD16/Jak2-K may provide signals necessary for growth. Using this fusion protein, we demonstrate that the activation of Jak2 was not sufficient alone to stimulate continued cell growth.