Internalization of the thrombopoietin receptor is regulated by 2 cytoplasmic motifs

Thrombopoietin (TPO) and its receptor, Mpl, are the primary regulators of megakaryocytopoiesis and play a critical role in hematopoietic stem cell biology.^1-6  Upon ligand binding, Mpl facilitates tyrosine phosphorylation of cytoplasmic signaling proteins and activation of several signaling pathways, including Janus kinase (Jak)/signal transducer and activator of transcription (STAT), Shc/Ras/Raf/mitogen-activated protein or extracellular signal—related kinase (MEK)/mitogen-activated protein kinase (MAPK), phosphatidylinositol (PI)—3-kinase, and protein kinase-C.^7-9  As a result of this complex network of signal transduction, TPO stimulation augments cellular proliferation, lineage-specific megakaryocyte development, and antiapoptosis.^10,11  We and others have used truncated and mutated Mpl receptors (121 amino acids in murine Mpl; 122 amino acids in the human form) to identify several functional subdomains and protein docking sites within the cytoplasmic domain of Mpl.^12-16 

A central principle of receptor signaling is that activated (ie, dimerized or oligomerized) receptors recruit signaling proteins to the inner surface of the cell membrane. The cytoplasmic domain of the receptor functions as a molecular scaffold to assemble the intact signaling complex. Internalization of receptors is believed to result in turning off signal transduction by (1) disassociating multiprotein networks, (2) targeting active proteins for destruction in the proteasomes or lysosomes, and (3) reducing the amount of receptor on the cell surface that can bind additional ligand.^17,18  For Mpl, internalization of the active receptor is believed to play a critical role in the regulation of circulating TPO, which is produced constitutively by the liver^19,20  and appears to be cleared from plasma by the aggregate mass of megakaryocytes and platelets (ie, Mpl-expressing cells).^21-23  It has been convincingly demonstrated that platelets can remove TPO from plasma both in vitro and in vivo.^24-26 

Because of this important role for ligand-dependent internalization of Mpl, we sought to identify specific elements of the receptor cytoplasmic domain that regulate this process. By engineering the murine hematopoietic cell line BaF3 to express wild-type (WT) and mutant forms of Mpl, we followed the time course and extent of TPO-dependent receptor internalization. This led to the identification of 2 distinct regions of interest, one involving the primary site of Mpl tyrosine phosphorylation and a second involving the box2 motif,²⁷  which contains the only 2 dileucine (LL or IL) motifs in the cytoplasmic domain of Mpl. Although the molecular mechanism is not clearly understood, dileucine motifs have been identified in other receptors (ie, growth hormone, interleukin-6, and epidermal growth factor receptors) as important promoters of internalization.^28-30  Of interest, mutations that diminish Mpl internalization markedly reduce the TPO-dependent proliferative potential of the cell lines. These data suggest that intracellular trafficking of Mpl signaling complexes may play an important and necessary role in cytokine signaling as well as regulating TPO plasma levels and potentially down-regulating TPO responsiveness.

Parental BaF3 cells were originally provided by Alan D'Andrea (Dana Farber Cancer Institute, Boston, MA) and were maintained in RPMI 1640 (BioWhittaker, Walkersville, MD) supplemented with 10% heat-inactivated fetal calf serum (FCS; BioWhittaker), 2 mM l-glutamine (BioWhittaker), 1% penicillin-streptomycin-fungizone (PSF; BioWhittaker), and 0.1% conditioned medium containing recombinant murine interleukin-3 (mIL-3). BaF3 cells were transfected by electroporation with c-mpl cDNA as previously described.¹³  Plasmid-containing cells were selected with G418 (1 mg/mL; Calbiochem, LaJolla, CA). Individual clones were isolated through limiting dilution in 96-well plates (Corning, Corning, NY) and were analyzed by flow cytometry to ensure comparable cell surface expression of Mpl. Additionally, polyclonal cell lines were generated for each construct and tested for comparable Mpl expression as detailed below.

Native murine c-mpl cDNA was originally provided by Zymogenetics (Seattle, WA). Truncated receptors were generated by polymerase chain reaction (PCR) mutagenesis as previously described.¹³  Point mutations were created using the high-fidelity Pfu DNA polymerase (Stratagene, LaJolla, CA) to introduce new site-directed mutations. Each plasmid was sequenced to confirm PCR fidelity and accurate mutagenesis (Prism sequencing kit; Applied Biosystems, Foster City, CA) and verified by sequence analysis using the BLAST (basic local alignment search tool) sequencing program (National Center for Biotechnology Information [NCBI], National Institutes of Health [NIH], Bethesda, MD). The serial carboxyl truncation Mpl mutants were named according to the number of membrane-proximal cytoplasmic residues remaining: T53; T69; T83; T98; T111; T116; and wild-type (WT) Mpl. Mutant variants of these Mpl constructs were named according to the introduced substitution: Mpl/Y112F; T69/LL-AA; T69/EIL-AAA; T69/box1(φ) (Figure 1).

For each Mpl construct, multiple G418-resistant clones and polyclonal cell populations were tested for cell surface expression of Mpl by flow cytometry (FACScaliber; Becton Dickinson, Franklin Lakes, NJ). Approximately 500 000 cells were resuspended in 250 μL fixation buffer for 30 minutes: phosphate-buffered saline (PBS; Sigma, St Louis, MO) containing 1% paraformaldehyde (Sigma), 0.1% bovine serum albumin (BSA; Sigma), and 0.02% sodium azide (Sigma). Cells were collected by centrifugation, resuspended in 50 μL binding buffer (PBS containing 0.1% BSA and 0.02% sodium azide), and then incubated with polyclonal rabbit antibody raised against the extracellular domain of Mpl for 30 minutes at 4°C (1:500 dilution; antibody provided by Zymogenetics, and Amgen, Thousand Oaks, CA). Cells were extensively washed and then resuspended in 20 μL binding buffer containing the secondary antibody, phycoerythrin (PE)—conjugated goat anti—rabbit IgG antibody for 20 minutes at 4°C (1:30 dilution; Jackson Immuno-Research, West Grove, PA). Cells were washed once to remove unbound antibody, resuspended in fixation buffer, then analyzed by flow cytometry. Ten thousand gated events were collected for each sample and were plotted on a semilogarithmic histogram. At least 2 independent clones in addition to a polyclonal cell line expressing each receptor construct were chosen with similar Mpl cell surface expression (less than one log variability in median fluorescence) for comparison in subsequent studies. The cell surface display of Mpl as measured by fluorescence intensity varied, on average, by no more than 20% among all the constructs studied in these experiments.

BaF3 cells expressing murine WT Mpl were incubated for 2 hours without interleukin-3 (IL-3) in RPMI supplemented with 10% FBS, 1% l-glutamine, and 1% PSF, then incubated for varying lengths of time with TPO at a concentration previously determined to induce maximal cellular proliferation (16.7 ng/mL) at 37°C.³¹  Internalization was stopped by washing the cells in ice-cold binding buffer. The cells were transferred to fixation buffer (at 4°C), then analyzed by flow cytometry as described above. The mean and median fluorescence channels were both recorded, but because of normal distribution of the cells, the 2 values were very similar and substitution of one for the other did not change our calculations. We used the median fluorescence of parental BaF3 cells to correct for nonspecific staining and autofluorescence. Changes in cell surface expression of Mpl were determined as follows: baseline Mpl expression (FL^BEFORE) = median fluorescence of Mpl-expressing cells — median fluorescence of parental BaF3 cells; Mpl expression after TPO stimulation (FL^AFTER) = median fluorescence of TPO-stimulated cells — median fluorescence of parental BaF3 cells; percent of Mpl remaining on the cell surface after TPO incubation = [(FL^BEFORE - FL^AFTER)/FL^BEFORE] × 100. Conversely, the percent of Mpl internalizes after TPO stimulation = (1 - [(FL^BEFORE - FL^AFTER)/FL^BEFORE)]) × 100. Because these calculations are based on relative, not absolute, differences, raw data were converted to “percent of Mpl internalized” in order to combine results from distinct cell lines and from experiments performed on different days. The data from multiple cell lines for each mutant were averaged to compensate for clone-to-clone variability. The Student 2-tailed t test for variance of means (Excel, Microsoft, Redmond, WA) was applied to the compiled data to analyze the differences in internalization among all the constructs. Unadjusted P values from pairwise comparisons are expressed in the text. A P value of less than .01 was considered statistically significant.

TPO was iodinated using the Bolton-Hunter reagent as previously described.³²  BaF3 cells expressing WT Mpl maintained in IL-3 were resuspended at a concentration of 1 × 10⁶ per 100 μL in binding buffer consisting of RPMI 1640 supplemented with 50 mMol/L HEPES (N-2-hydroxyethylpiperazine-N′-2-ethanesulfonic acid) (pH 7.4) and 1% bovine serum albumin containing ¹²⁵I-TPO (0.2 ng/mL) in the presence or absence of unlabeled TPO (200 ng/mL). Triplicate samples were used for each point. The cells were incubated with ¹²⁵I-TPO for one hour at 37°C. The cell suspension was then sedimented through Percoll (density, 1.03 g/mL) to remove unbound radioactivity, and the cells were resuspended in 200 μL phosphate-buffered saline. Surface-bound ¹²⁵I-TPO was “stripped” by exposure to barbitol sodium acetate, pH 3.0, for 6 minutes at 4°C. After centrifugation, the radioactivity associated with the cell pellet (representing internalized ¹²⁵I-TPO), and the radioactivity in the supernatant (ie, removable, surface-bound ¹²⁵I-TPO) was quantified in a Packard 5330 gamma counter (Packard, Downers Grove, IL).

Cellular proliferation was measured using the 3-[4,5-dimethylthiazol-2-yl]-2,5-diphenyltetrazolium bromide; thiazolyl blue (MTT; Sigma) assay as previously described.¹³  Cells were washed twice with cytokine-free media and then resuspended in RPMI 1640 with 10% calf serum, 1% l-glutamine, 1% PSF, and dilutions of murine TPO-conditioned supernatant (0.01%, 0.1%, 1%, and 10%; 1% = 3.3 ng/mL TPO), which is sufficient for maximal cellular proliferation with wild-type Mpl³¹  or conditioned media containing murine IL-3 at a dose sufficient to support maximal proliferation. After growth for 36 hours, MTT (final concentration, 1 mg/mL) was added. Incubation at 37°C was continued for another 5 hours. The cells were then lysed, and absorbance at 570 nm minus 630 nm was determined using an enzyme-linked immunosorbent assay (ELISA) plate reader. Samples were run in triplicate, and proliferation is expressed as a percentage of maximal IL-3—stimulated growth.

Mpl internalization was examined using BaF3 cells engineered to express WT Mpl. Cells growing in log-phase in IL-3—containing media were washed twice and resuspended in growth media without IL-3. Recombinant TPO (16.7 ng/mL) was added at a concentration previously determined to induce maximal cellular proliferation.³¹  At baseline and at various time points after TPO exposure, aliquots were withdrawn for measurement of Mpl cell surface expression by fluorescent immunostaining and FACS analysis. These data (Figure 2) demonstrate that 65% to 75% of cell surface Mpl disappears after TPO exposure and that internalization proceeds rapidly; more than half of the receptors are removed from the cell surface within 10 minutes. After prolonged incubation with TPO (2-3 hours) the amount of cell surface Mpl is not significantly different from that seen after 60 minutes.

Although flow cytometry is a sensitive and reproducible technique for assessing changes in receptor cell surface expression, we sought to confirm our results using another established method by using radiolabeled TPO and acid stripping. These results, shown in Table 1, indicate that after a 60-minute incubation at 37°C, 67% of the cell-associated ¹²⁵I-TPO is internalized, and 33% of the cell-associated ¹²⁵I-TPO is surface-bound. These results are indistinguishable from the flow cytometry results (ie, 70% ± 2.4% of Mpl internalized after a one-hour stimulation with TPO at 37°C).

Flow cytometry has an advantage over measurement of ¹²⁵I-TPO in that presence of the receptor on the cell surface is assessed directly rather than via measurement of the ligand. However, it is possible that the ligand-bound (ie, activated) Mpl and the inactive receptor will have distinct affinities for the polyclonal Mpl antibody. To test this directly, we preincubated BaF3/Mpl cells with 3 known inhibitors of receptor internalization: the Src kinase inhibitor PP2 (50 μM); sodium azide (0.1%); and cytochalasin B (10 μg/mL). Cells were then incubated with or without TPO (16.7 ng/mL) for one hour, immunolabeled with the polyclonal Mpl and secondary detecting antibodies, and analyzed by flow cytometry. The mean fluorescence of BaF3 Mpl cells without TPO (14.5) and with TPO (15.7) was statistically indistinguishable. These data demonstrate that the binding of TPO to Mpl does not block subsequent binding of the polyclonal antibody to Mpl.

TPO-dependent receptor internalization was studied using BaF3 cells expressing serial truncations of the intracytoplasmic domain of Mpl. After 60 minutes of TPO stimulation samples were fixed for immunostaining. For each Mpl receptor mutation, at least 3 distinct cell lines were studied, and the experiments were performed 3 or more times to ensure internal consistency and reproducibility. The data, shown in Table 2 and graphically in Figure 3, indicate that Mpl receptors containing 116 cytoplasmic residues (T116; Figure 3A) are internalized to the same extent as full-length, or WT, Mpl (121 amino acids). However, receptors with 111 membrane-proximal cytoplasmic residues (T111; Figure 3B) are internalized significantly less efficiently (maximum internalization of 45% ± 7%; P < .0001). Truncated receptors with 98 (T98), 83 (T83), or 69 (T69; Figure 3C) residues were not significantly different from one another. In contrast, a truncated receptor with only 53 cytoplasmic residues (T53; Figure 3D) was not appreciably internalized following TPO exposure (6% ± 5%; P = .0008 compared with T69). Next, internalization of WT versus T69 receptors was studied at several time points (Figure 4). These data demonstrate that at any given time, the truncated receptors are removed from the cell surface at a slower rate and eventually achieve a different equilibrium point.

Because analysis of truncated receptors indicate that the 5 residues between 111 and 116 are required for complete internalization, we focused our attention on Y¹¹², the principle site of TPO-induced receptor tyrosine phosphorylation.¹³  A full-length receptor with substitution of phenylalanine for tyrosine at position 112 (Y112F) was expressed in BaF3 cells, and TPO-induced internalization of Y112F was compared with WT Mpl, T116, and T111 (Table 3). We found that the Y112F substitution is equivalent to T111 for TPO-induced internalization and that both Y112F and T111 are significantly distinct from T116 (P = .0005 in each case), indicating that Y¹¹² plays an important role in Mpl internalization. A cellular MTT proliferation assay demonstrates that TPO-induced cell growth of the Y112F mutant is similar to our previously published data for the Y112F/Y117F double mutant and comparable to T111,¹³  suggesting that signals arising from Y¹¹² contribute to TPO-stimulated internalization as well as proliferation (Figure 5).

The data from the truncated receptor experiments (Table 2) indicate that Mpl cytoplasmic residues 54 through 69 provide a second critical signal for TPO-induced receptor internalization. The primary sequence for this region (LLEILPKSSESTPLPL) includes the only 2 dileucine motifs within the cytoplasmic domain of Mpl; dileucine motifs have been recognized in other receptors as potential signals for internalization.^28-30  Therefore, simultaneous alanine substitutions of 2 or 3 amino acids within this region were generated, and the resulting receptors were expressed in BaF3 cells. TPO-induced internalization of the truncated T69 receptor with mutation of LL-54,55-AA (32% ± 2.8% after 60 minutes; Figure 6A) was unchanged when compared to the T69 receptor (29% ± 2.9% after 60 minutes). In contrast, mutation of EIL-56,57,58-AAA in T69 significantly diminished ligand-dependent internalization of T69 (15% ± 3.8% after 60 minutes, P = .003; Figure 6A). Our data suggests that in the truncated T69 receptor, the residues E⁵⁶I⁵⁷L⁵⁸, including the distal dileucine motif, promote receptor internalization.

Previously published data have shown that the same interval (cytoplasmic residues 54-69, which include the box2 core motif) is also critical for TPO-induced proliferation.¹³  We hypothesized that the signals regulating proliferation and internalization might be linked, and we tested BaF3 cells expressing T69 with the dileucine mutations in a proliferation assay (Figure 6B). The data show that mutation of the L⁵⁴-L⁵⁵ (LL-AA) dileucine motif modestly diminishes TPO-induced cell growth, while substitution of alanine for residues E⁵⁶, I⁵⁷, and L⁵⁸ (EIL-AAA) has a significantly greater effect on TPO-stimulated cell growth.

Jak2 activation is a critical early event, believed to underlie most, if not all, Mpl function. Studies with Jak2 knock-out mice and mutated forms of Mpl demonstrate that no cellular proliferation occurs in the absence of Jak2 activation,^33,34  and the transforming potential of the v-mpl oncogene is lost.³⁵  Because of the central role played by Jak2 in the function of Mpl, we sought to determine whether the internalization signal arising from the membrane-proximal region of Mpl also requires Jak2 kinase activity. Jak2 is phosphorylated by TPO stimulation of the truncated T69 receptor.³⁴  However, if 2 critical proline residues in the P¹⁷-S¹⁸-L¹⁹-P²⁰ box1 motif are replaced with either glycine (as previously published³⁴ ) or alanine (ie, A¹⁷-S¹⁸-L¹⁹-A²⁰ = T69/box1(φ) in the present work), T69/box1(φ) can neither support proliferation of BaF3 cells nor activate Jak2. When expressed in BaF3 cells, this receptor can still be internalized in response to TPO, similar to the fully functional T69 receptor (26% ± 5.2%; P = .6). These results indicate that the truncated receptor T69 is removed from the cell surface by a Jak2-independent mechanism.

Because Mpl internalization appears to be the primary mechanism for regulation of plasma TPO levels, we postulated 2 possible fates for the internalized ligand-receptor complex. First, it might be targeted for destruction in the proteasomes. Second, the receptor might be recycled to the cell surface. To determine which of these hypotheses is correct in our model system, we studied the recovery rate of Mpl on the cell surface following withdrawal of TPO from the media (Figure 7). Cycloheximide (10 μg/mL) was added to half the samples both before and after TPO stimulation to block new protein synthesis. Aliquots for duplicate samples were removed at each time point. The results indicate that approximately half the internalized Mpl protein reappears on the cell surface within 15 minutes of TPO washout, suggesting that Mpl is rapidly transported to the cell surface from an intracellular pool. Although the recovery curves diverge beyond 15 minutes, the early reappearance of Mpl on the cell surface is not significantly altered in the presence of cycloheximide, demonstrating that de novo protein synthesis is not responsible for the early reappearance of receptor on the cell surface.

The cell membrane is a dynamic environment in which transmembrane proteins, including receptors, are constantly being removed from the cell surface by internalization and inserted into the lipid bilayer through synthesis and/or recycling. In the case of the thrombopoietin receptor Mpl, internalization is believed to be the primary method for regulating plasma TPO levels.^21-26  When platelet and megakaryocytic mass is reduced, the plasma TPO concentration rises and drives the proliferation of megakaryocytic precursors. Conversely, when platelet levels rise, the TPO level is reduced through receptor-mediated internalization. Although this model has been studied for several years, the receptor-based signals that direct internalization of Mpl have not been identified. In this paper, we identify 2 distinct motifs that promote Mpl internalization in hematopoietic cells, a phosphotyrosine motif at the carboxyl terminus and a dileucine motif within the box2 region of the cytoplasmic domain. Furthermore, we demonstrate that internalization directed by the motif within box2 is not dependent on Jak2 signaling, and, finally, we show that after TPO stimulation is withdrawn, a substantial amount of Mpl can be recycled to the cell surface from intracellular pools independent of de novo protein synthesis.

Mutation or deletion of the fourth cytoplasmic tyrosine residue (Y¹¹²) reduces the extent of ligand-induced removal of Mpl from the cell surface by half. Several laboratories have demonstrated that Y¹¹² is the principle site of TPO receptor tyrosine phosphorylation and serves as a docking site for several signaling proteins.^7-9  Phosphotyrosine residues have been previously identified as contributing to receptor internalization of such diverse receptors as those for vascular endothelial growth factor (VEGF, or knock-down resistance [KDR]),³⁶  epidermal growth factor,³⁷  δ-opioids,³⁸  low-density lipoproteins,^39-41  and the insulin-like growth factor.⁴²  In each of these instances, substitution of phenylalanine for a carboxy-terminal tyrosine attenuates ligand-dependent internalization as well as receptor activation. Furthermore, a common motif, NPXY, has been recognized, in which X may be any amino acid (or, occasionally, 2 residues). It is believed that the reverse beta turn induced by this motif contributes to recognition by proteins, including those with phosphotyrosine binding (PTB) domains⁴²  such as Dab1³⁹  and Dab2.^40,41  Mpl includes a similar motif at the carboxyl terminus (NHSY¹¹²), which, when phosphorylated, serves as the docking site for Shc, a PTB-containing adapter molecule, and Src homology 2 domain (SH2)—containing proteins such as STAT3 and STAT5.¹³  There is evidence that the phosphorylated form of the NPXY motif also may provide a binding site for the adapter protein AP-2 that is necessary for interaction with clathrin and coated-pit—mediated endocytosis.^43,44  Whether other signaling proteins are involved in phosphotyrosine-mediated receptor internalization is not currently known. Despite the importance of the NPXY motif, partial internalization of Mpl can occur even when it is deleted or mutated. The ability of internalization to proceed in the absence of receptor tyrosine phosphorylation also has been demonstrated for the erythropoietin receptor.⁴⁵ 

The membrane-proximal cytoplasmic domain of the receptor that directs ligand-dependent internalization of Mpl includes the only 2 dileucine motifs in the cytoplasmic domain of the molecule. Leu-Leu and Ile-Leu have previously been studied in the process of receptor internalization and trafficking for a number of receptors including the cytokine receptor subunit gp130,²⁷  the epidermal growth factor receptor,²⁹  and the growth hormone receptor.³⁰  In our experiments, we found that E⁵⁶-I⁵⁷-L⁵⁸ (but not L⁵⁴-L⁵⁵) was essential for internalization of the truncated T69 receptor.

At present, it is not known how dileucine motifs promote receptor translocation, although it is possible that they stabilize dimers or facilitate dimerization similar to leucine zipper motifs.⁴⁶  Of interest, internalization of IL-2 requires cytoplasmic residues 45-54 of the IL-2R common γ-chain.⁴⁷  This region includes an isoleucine residue in similar context (VSEIPPK vs LLEILPK), suggesting that this may be a more generalizable finding. Our results with a T69 receptor unable to activate Jak2 (T69/box1(φ)) demonstrate that dileucine-mediated internalization of Mpl does not require Jak2 kinase activity. Similarly, it has been reported that ligand-induced internalization of the gp130 subunit and erythropoietin receptor can occur in the absence of Jak2 activation.^28,45  As was demonstrated for the Kit receptor, other tyrosine kinases such as Src family members may play a role in this process.⁴⁸  In some cases, the critical role of a membrane-proximal dileucine motif for ligand-induced internalization may be fully appreciated only when other internalization signals are missing, as has been demonstrated for the growth hormone receptor.³⁰ 

As shown using the nonfunctional receptor T69/box1(φ), Mpl internalization can proceed in the absence of proliferative signaling. However, it is not clear that the converse is true. Receptor mutations that decrease internalization also impair TPO-induced proliferation (Figures 5, 6; Table 3). This raises the intriguing hypothesis that internalization of activated Mpl not only serves to regulate plasma TPO levels, but also may be an essential step in proliferative signaling. One possibility is that the entire signaling complex remains intact within endosomes, which could transport activated signaling molecules (eg, STATs) to intracellular locations. Recently, the role of cytokines and their receptors as chaperones has been suggested for the interferon-gamma receptor, among others.^49,50  This possibility is a topic of ongoing investigation in our laboratory.

The only previous study on the ability of internalized Mpl to be delivered back to the cell surface was reported using human platelets.⁵¹  In this paper, platelets were continuously exposed to radiolabeled TPO for 24 hours, and the amount of associated radioactivity reached a plateau after 60 minutes with no further accumulation. The authors conclude that internalized receptors are not recycled to the cell surface nor are additional receptors synthesized. In contrast, our results in a hematopoietic cell line demonstrate that Mpl receptors appear to be returned to the cell surface after ligand withdrawal. Recycling would be an efficient mechanism for regulating TPO levels, as it allows a single receptor molecule to be re-used and clear multiple ligand molecules from plasma. It is possible that there are differences in the receptor trafficking of intact megakaryocytes and platelets that would explain this discrepancy. It is intriguing that the data presented by Li et al⁵¹  and the present study indicate 65% to 75% maximal Mpl internalization following TPO stimulation. This may represent a dynamic steady state that is the result of the balanced rates of internalization versus recycling and de novo synthesis of Mpl.

We wish to thank Ken Kaushansky (University of California, San Diego) for helpful discussions and advice; Don Foster (Zymogenetics) and Pamela Hunt (Amgen) for providing the Mpl cDNA and polyclonal Mpl antibody; and Sara Lowe and Linda Hibbeln for assistance with manuscript preparation.