Mutational Analysis of Thrombopoietin for Identification of Receptor and Neutralizing Antibody Sites*

, Thrombopoietin (TPO) is a hematopoietin important for megakaryocyte proliferation and production of blood platelets. We sought to characterize how TPO binds and activates its receptor, myeloproliferative leu-kemia virus receptor. The erythropoietin-like domain of TPO (TPO 1–153 ) has been fused to the gIII coat protein of M13 bacteriophage. Forty residues were chosen for mutation to alanine using the criteria that they were charged residues or predicted to be solvent-exposed, based on a homology model. Phage enzyme-linked im- munosorbent assay was used to determine affinities for binding to both the TPO receptor and five anti-TPO 1–153 monoclonal antibodies. Mutations at mostly positively charged residues (Asp 8 , Lys 14 , Lys 52 , Lys 59 , Lys 136 , Lys 138 , Arg 140 ) caused the greatest reduction in recep-tor-binding affinity. Most of these residues mapped to helices-1 and -4 and a loop region between helix-1 and helix-2. Two of the monoclonal antibodies that blocked TPO binding and bioactivity had determinants in he-lix-4. In contrast, the other three monoclonal antibodies, which were effective at blocking TPO activity but did not block initial binding of TPO to its receptor, had epitopes predominantly on helix or 3. These results suggest that TPO has two distinct receptor-binding sites that function to dimerize TPO receptors in a sequential fashion.

Thrombopoietin (TPO) 1 is the hematopoietic cytokine responsible for stimulation of megakaryocyte formation and regulation of platelet release (1)(2)(3). The gene for TPO encodes a 38-kDa protein (332 amino acids) that can be divided into two distinct domains (1,4). The N-terminal region (153 amino acids) is predicted to be a four-helix bundle and has considerable sequence similarity to erythropoietin (EPO; 23% amino acid identity), whereas the C-terminal domain has no homology to known proteins and contains several asparagine-linked gly-cosylation sites. A number of studies have indicated that the N-terminal domain, TPO  , is responsible for binding and activation of the TPO receptor (MPL) (1,5). The extracellular domain of the receptor for TPO has considerable sequence similarity to the other receptors in the class I hematopoietin receptor superfamily (6).
The four-helix bundle cytokines typically activate their receptors by homo-or hetero-oligomerization in which two distinct binding sites on a single hormone sequentially bind two of the same receptor molecules or two or more different receptor molecules, respectively (for recent review, see Ref. 7). The receptor binding sites for a number of cytokines (hGH, human prolactin, IL-4, IL-6, and EPO) map to similar regions of the hormone: one of these sites (site 1) has primary determinants in helix-4 and the loop connecting helix-1 and helix-2, whereas the other site (site 2) has primary determinants in helix-1 and helix-3. In the case of hGH, site 1 binds first followed by site 2 (8); for IL-4, site 2 appears to bind first followed by site 1 (9). We wondered if TPO contains two distinct receptor binding sites, and if so, do these map to regions seen for the other members of this class.
Alanine-scanning mutagenesis has been used to probe binding determinants in protein-protein complexes (10) (for review, see Ref. 11). Here, the sequence of TPO  was aligned to that of IL-4 and mapped upon the known structure of IL-4 (12). Forty residues predicted to be solvent-exposed (mostly charged residues) were mutated to alanine and displayed as a single copy on M13 phagemid particles (13). The alanine mutants were analyzed by phage ELISA (14) for binding to the TPO receptor and to five monoclonal antibodies (mAbs) that block TPO bioactivity. Interestingly, mAbs having binding determinants in helix-4 blocked receptor binding and bioactivity, whereas those having determinants in helix-1 or helix-3 blocked bioactivity but not receptor binding. These data plus the receptor epitope mapping suggest that TPO 1-153 has two receptor-binding sites that bind and activate its receptor by a sequential dimerization-type mechanism similar to that for hGH.

Phagemid Construction, Alanine-scanning Mutagenesis, and Phage Preparation
The vector, pMP11 containing the TPO 1-155 gene (gift of Dan Yansura, Genentech, Inc.), was tailored by site-directed mutagenesis (15) to create EcoRI and AvrII restriction sites that facilitated excision of the TPO gene. This was inserted into an EcoRI/XbaI-digested derivative of phGHam-g3 (16). This construct places the TPO gene (codons 1-155) at the N-terminal end of two Gly-Gly-Gly-Ser repeats that is followed by the C-terminal domain (codons 249 -406) of M13 gene III. Deletion mutagenesis (15) was used to remove codons 154, 155, and the amber stop codon (TAG) to produce the phagemid vector pML0433 to display TPO  . Alanine mutations in the TPO 1-153 gene were made by sitedirected mutagenesis (15). Monovalent phage particles were prepared * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  as described previously (13,16) with the exception that phagemid and M13K07-infected XL-1-Blue cells (Strategene, La Jolla, CA) were cultured at 30°C. Additionally, monovalent phage particles were generated using the VCS helper phage (Strategene). Typically, phage particles obtained from 30 ml of cell culture were resuspended into 0.5 ml of 10 mM sodium phosphate, 137 mM NaCl, 2.7 mM KCl, pH 7.2 (PBS) for preparation of the final phage stock.

Preparation of TPO Receptor (MPL-IgG)
Recombinant MPL-IgG was prepared essentially as described by de Sauvage et al. (1). Cell culture supernatant from calcium phosphate transfected 293 cells was passed over a protein A-agarose (Repligen, Cambridge, MA) column at 1 ml/min then washed extensively with 10 mM tris-HCl, 150 mM NaCl, pH 7.5 (TBS). Material was obtained from the column with a gradient elution using ImmunoPure Gentle Ag/Ab Elution Buffer (Pierce, Rockford, IL). Fractions containing MPL-IgG were immediately desalted using PD-10 columns (Pharmacia Biotech, Uppsala, Sweden), concentrated, and dialyzed against TBS. Concentration of MPL-IgG was estimated using ⑀ 280 ϭ 1.9 ml mg Ϫ1 cm Ϫ1 .
Generation of Anti-TPO 1-153 mAbs mAb 1831-Balb/c mice were immunized with approximately 10 g per dose of TPO expressed from 293 cells. TPO was formulated in QS21/alum and RIBI adjuvant (RIBI ImmunoChem Research Inc., Hamilton, MT) and injected intraperitoneally on weeks 1, 2, 5, 6, 9, 10, 13, 23, and 26. The immunized mice were tested for antibody response by ELISA. The mouse with the highest titer was given an additional boost of 10 g of TPO in RIBI during week 27, and 3 days later the spleen was fused with the mouse myeloma line X63-Ag8.653 (17) using 50% polyethylene glycol 4000 (Boehringer Mannheim) by the procedure of Kohler and Milstein (18).
mAbs 1643, 1645, 1646, and 1653-Balb/c mice were immunized with approximately 20 g of TPO in RIBI adjuvant on days 0, 3, and 21. The mice were given an additional boost of 10 g of TPO in RIBI on day 31. Three days later the popliteal nodes were fused with the mouse myeloma line X63-Ag8.653. The fused cells were plated in 96 well tissue culture plates at a density of 200,000 cells per well. Hybridoma selection, using HAT media supplement (Sigma) was begun 1 day postfusion. Beginning on day 10, the hybridoma supernatants were screened for the presence of TPO 1-153 -specific antibodies using a solid phase ELISA. Stable antibody producing clones were obtained by limiting dilution and large quantities of specific mAbs were produced in ascites fluid. The antibodies were purified on protein A-Sepharose columns (Fementech, Inc., Edinburgh, Scotland) and stored sterile in PBS at 4°C. Isotyping of the monoclonal antibodies was done using a Mouse MonoAb ID/SP isotyping kit from Zymed (Zymed, South San Francisco, CA) according to the manufacturer's instructions.

TPO-phage ELISA for Determination of TPO Receptor and mAb EC 50 Values
Typically, 96-well Maxisorp immunoplates (NUNC, Roskilde, Denmark) were coated with either the respective anti-TPO 1-153 mAb or anti-hFc F(abЈ) 2 (Jackson ImmunoResearch, West Grove, PA) (100 l at 1 and 2 g/ml, respectively, in 50 mM carbonate buffer, pH 9.6) overnight. Plates were washed using PBS, 0.05% Tween 20 (Sigma); for TPO receptor experiments, MPL-IgG was added to anti-hFc F(abЈ) 2 -coated wells (100 l at 1 g/mL in PBS) and incubated at room temperature for 2 h. Plates were again washed using PBS, 0.05% Tween 20. Phagemid particles were diluted serially using either PBS, 0.05% BSA, 0.05% Tween 20 (BSA blocking buffer) or 25 mM Na 2 HPO 4 /200 mM NaCl, 1% Hammarsten grade casein (BDH Lab Supplies, Poole, UK) (casein blocking buffer) and then transferred (100 l) to coated wells. After 2 h, plates were washed extensively with PBS, 0.05% Tween 20, incubated with 100 l of 1:5000 HRP/anti-M13 conjugate (Pharmacia Biotech) in BSA blocking buffer for 15 min then washed with PBS, 0.05% Tween 20, and PBS. Plates were developed using an o-phenylenediamine dihydrochloride/H 2 O 2 (Sigma) solution, stopped with H 2 SO 4 , and read spectrophotometrically at 492 nm. Additionally, phage binding to polyclonal anti-TPO antisera was performed as described above to confirm TPO and TPO mutant expression. For the competition phase of the assay, each phage stock was diluted into either BSA or casein blocking buffer to give a subsaturating phage concentration in the final assay volume (100 l). Typically, an amount of phage giving approximately 0.3-0.4 OD units after development was chosen. The diluted phage (50 l) was combined with serially diluted MPL-IgG or mAb (50 l) and then rapidly transferred to either MPL-IgG-or mAb-coated wells, respectively. Samples were incubated for 2 h before development as described above.

BIAcore Determination of anti-TPO 1-153 mAb Affinity
Apparent equilibrium dissociation constants (K d ) for anti-TPO 1-153 mAb binding to TPO were determined using surface plasmon resonance (19) on a Pharmacia BIAcore instrument. Typically, full-length TPO (Genentech, Inc.) was reacted with the activated sensor chip by random lysines to yield either 500 or 1500 resonance units. Dissociation rates were measured by passing 35 l of either 1.3 M, 0.65 M, or 0.33 M mAb in PBS with 0.05% Tween 20 at a flow rate of 20 l/min over the sensor chip; after the injection, the decrease in resonance units was monitored with time. Binding profiles were analyzed by nonlinear regression using a two-component, bivalent binding model (BIAevaluation version 2.0; Pharmacia Kinetics). Average dissociation rates were determined using f (1) (2) , where f is the fractional response value derived from the nonlinear curve fits. Association rates were determined from the concentration dependence of the binding profiles for various mAb solutions. Typically, 325 nM solutions of ligand were serially diluted (2-fold) four times, and 35 l were injected with flow rates of 20 l/min. Sensor chips were regenerated with injection of 5 l of 10 mM HCl at a flow rate of 5 l/min. Equilibrium dissociation constants were calculated by K d ϭ k off /k on .

TPO Receptor Binding Inhibition Assay
MPL-IgG was captured on 96-well immunoplates (NUNC) as described above. Wells were blocked with 150 l of PBS, 0.5% Tween and washed. Each mAb stock was serially diluted 2-fold in PBS, 0.5% BSA, 0.05% Tween 20, pH 7.4, then combined with a solution of biotinylated full-length TPO to give a final TPO concentration of 35 pM. Mixtures were incubated for 1 h before transferring 100 l to TPO receptorcoated wells. Samples were incubated for 1 h at room temperature with gentle agitation. Plates were washed, incubated with 1:5000 streptavidin-HRP (Sigma; 100 l), then washed again. Wells were developed using TMB substrate (Kirkegaard and Perry Labs, Inc., Gaithersburg, MD), quenched with H 3 PO 4 , and read by spectrophotometer at 450 nm.

MPL/Rse KIRA
A vector containing the chimeric receptor, the extracellular domain of MPL epitope-tagged with glycoprotein D (gD-MPL) and the transmembrane/intracellular domain of the receptor-type tyrosine kinase Rse (20), was transfected into CHO cells (21). 2 The cells expressing this chimeric receptor have been shown to be responsive to TPO, as measured by intracellular tyrosine phosphorylation. Intracellular tyrosine phosphorylation was detected by incubation of the wells with biotinylated 4G10 antibody (Upstate Biotechnology Inc., Lake Placid, NY; 0.05 g/ml), streptavidin-HRP (Zymed; 1:40,000), and then developed using TMB substrate (Kirkegaard and Perry). Reactions were stopped and samples were read by spectrophotometer as described above.

Structure-based Sequence Alignment and Model of TPO-
Members of the four-helix bundle cytokine family fall into two general classes depending on their length (for review, see Ref. 22). The long-chain class that includes hGH have lengths that range from 175 to 200 residues, whereas the short-chain class that includes IL-4 have lengths that range from 105 to 160 residues. Despite differences in length and sequence, alignments from known structures show that the helical cores of both classes of molecules are very similar (12,(23)(24)(25). In fact, predictions using structural pattern-matching methods (26) have been successful for predicting the basic four-␣-helical cores (12,25,27), and these can be useful for guiding mutational analysis.
To identify the helical core in TPO 1-153 we chose to align its sequence with that of IL-4, because the structure of IL-4 is known to high resolution (12) and the four cysteines in TPO 1-153 match most closely to four of the six cysteines in IL-4 (Fig.  1A). Hydrophobic residues at the i to i ϩ 3 and/or i to i ϩ 4 positions in the predicted helical segments of TPO 1-153 align with buried residues on the IL-4 helices consistent with their having similar helical cores. Although two hydrophilic residues of TPO 1-153 are predicted to be partially buried in the core (Asp 18 and Lys 138 ), IL-4 also contains hydrophilic residues at these analogous positions (Thr 13 and Lys 123 ). The segment connecting ␣-helices 1 and 2 is much longer in TPO 1-153 than in IL-4, and in this regard TPO resembles members of the long-chain ␣-helical family. The alignment shows that ␣-helices-2, -3, and -4 in TPO have one to three glycine residues per helix, but glycines also occur within helices of other cytokines (28). The alignment was used as a guide for alanine-scanning (Fig. 1B).
Expression of TPO 1-153 on Phage-TPO 1-153 was displayed monovalently on the surface of bacteriophage as a gene III fusion (13,16). Expression of active protein was increased by growing infected Escherichia coli at 30°C instead of 37°C (data not shown). The TPO 1-153 displayed on phage bound tightly to five anti-TPO 1-153 mAbs with EC 50 values in the phage ELISA that ranged between 1 and 10 nM (data not shown). Phage particles displaying TPO 1-153 bound specifically to immobilized TPO receptor (MPL-IgG) and the interaction was competed by soluble TPO receptor (Fig. 2). Because the amount of phage added in the competition phase of the ELISA is subsaturating, analysis of the competition curve yields an EC 50 value that is related to the dissociation constant (K d ) for the soluble receptor. The EC 50 value for TPO 1-153 -phage (44 Ϯ 15 nM; Fig. 2) was almost 100-fold higher than the apparent K d (Ϸ0.5 nM; data not shown) for the soluble ligand-receptor interaction. This apparent reduction in affinity is likely due to linkage to the gene III protein and others have reported similar reductions in affinity when proteins were displayed on phage. Nonetheless, the relative changes in affinity caused by alanine mutations when measured on the phage have been shown to parallel those measured on soluble proteins (14). 3 Thus, it seems reasonable that the TPO phage ELISA can be used to compare relative affinities for various site-directed mutants of TPO.
Alanine-scanning Mutagenesis for Location of Receptor Binding Determinants-Using the alignment of TPO 1-153 on to the structure of IL-4, 40 charged and surface residues were mutated to alanine (Fig. 1B). As a control for functional display, all the alanine mutants were tested for binding to anti-TPO poly-FIG. 1. A, sequence alignment of human TPO with human IL-4. The sequences were aligned to best match the cysteines and regions predicted to be helical using the crystal structure of IL-4 as a guide (12). Those residues in IL-4 that are classified as consensus inner core side chains by comparison with other four-␣-helical cytokines (24)  clonal antisera (data not shown) and to a panel of five anti-TPO 1-153 mAbs (see below). All the variants bound the polyclonal antisera, except H23A which did not bind to either the polyclonal or to the five monoclonal antibodies suggesting it was either not expressed on the phage or grossly misfolded. The R78A variant bound the polyclonal antibody but did not bind with high affinity to three of the five mAbs tested. Thus, of the forty alanine mutants, we suspect that only two variants (H23A and R78A) are grossly disruptive to the structure or are not displayed on the phage.
A phage ELISA was used to test binding of the remaining thirty-eight alanine mutants to an immobilized and dimeric form of the TPO receptor (MPL-IgG). All but two variants (D8A and K138A) retained sufficient affinity for the TPO receptor to allow for competition in the assay (Table I; Fig. 3A). Very little signal was detected for D8A and K138A in the phage titration, and we estimate that their EC 50 values were at least 20-fold greater than the wild-type TPO 1-153 . The second class of residues, including K14A, K52A, R136A, and R140A, had EC 50 values that ranged from 5-10-fold greater than TPO 1-153 . Nine FIG. 3. Alanine scan of TPO displayed on phage for binding to the TPO receptor. A, the EC 50 value for each alanine mutant was determined and plotted as an EC 50 ratio to wild type TPO 1-153 . A value above unity indicates the alanine mutation disrupted affinity by the value shown. H23A did not bind to anti-TPO 1-153 mAbs or to polyclonal antibodies suggesting that this mutant does not express (DNE). Additionally, R78A is a mutation that disrupted binding to three mAbs and therefore may play a structural role in TPO. Errors represent standard deviations from two trials using independent TPO phage and TPO receptor preparations. For all mutants with EC 50 ratios greater than two, values of p obtained using a two-tailed t test (29) were less than 0.01. For some mutants (R10A, E31A, E50A, D62A, E72A, and Q105A) with EC 50 ratios between 1.5 and 2.0, the t test yields 0.02 Ͻ p Ͻ 0.17; therefore, these differences from wild-type are not statistically significant. B, TPO residues that had a significant effect on TPO receptorbinding are shown plotted on the structure of IL-4 (12). Red indicates a Ͼ20-fold disruption in affinity and these residues are labeled; dark blue indicates a 5-10-fold drop in affinity and light blue indicates a 2-5-fold loss in affinity. Because of gaps in the sequence alignment (Fig. 1A), only approximate locations are shown for some residues (Asp 45 , Glu 50 , Lys 52 , Lys 59 , and Leu 144 ). Because TPO is larger than IL-4 residues 1-6, 47-59, and 145-153 are not included in the model. Figs. were made using Insight II (Biosym).  (29).
c Competition of phage with soluble TPO receptor was not performed due to poor signal in the phage titration phase of the assay. d H23A did not bind to polyclonal or monoclonal anti-TPO 1-153 antibodies. R78A did not bind with high affinity to three of the five mAbs. We could not detect binding to the TPO receptor but these mutants may have been poorly expressed so we denote them as ND, not determined. e Only a single trial was performed.
Alanine-scanning Mutagenesis for Epitope Mapping Anti-TPO 1-153 mAbs-Five monoclonal antibodies that bind TPO   (mAbs 1643, 1645, 1646, 1653, and 1831) were tested for binding to phage expressing the alanine mutations (Fig. 4A). As mentioned above, only H23A did not bind any of the mAbs, and R78A bound poorly to mAbs 1643, 1645, and 1653. The remaining 38 mutants bound to at least three of the five mAbs with near wild-type affinity suggesting that any disruptive binding effects were not the result of gross misfolding or poor expression. The dynamic range in the phage ELISA for the mAbs allows us to measure increases in EC 50 (decreases in affinity) of up to 20-fold. In general, only a small set of alanine mutants (one to eight) caused a Ͼ20-fold increase in EC 50 to any particular mAb. That only a few critical residues are important for governing mAb binding has been previously reported for a number of antibodies against several other proteins (30 -33).
Mapping the disruptive alanine mutations on to the structural model of TPO shows that the epitopes for the five mAbs fall into three groups: mAbs 1643 and 1645 bind similarly, mAbs 1653 and 1831 bind similarly, and mAb 1646 is rather distinct from either of these two groups (Fig. 4B). Excluding R78A as a binding determinant, binding of mAbs 1643 and 1645 to TPO is dominated by residues on helix-3. The mutations, S87A and S94A were most disruptive to binding of mAb 1643, and S88A and Q92A and were most disruptive to binding of mAb 1645. mAbs 1653 and 1831 had primary determinants in helix-4 as well as a short segment at the N-terminal region of helix-2 (Fig. 4B). Binding of mAb 1653 was significantly effected by the mutations K52A, D62A, N125A, L129A, Q132A, H133A, R136A, K138A, and R140A. For mAb 1831, residues K59A, Q61A, D62A, and K138A (on helix-4) were found to be important for binding TPO.
Effects of alanine mutations on binding mAb 1646 were somewhat less disruptive (between 2-and 3-fold) and generally larger in number (Fig. 4A). Mutations that caused a greater than two-fold increase in the EC 50 for mAb 1646 were R10A, R17A, D18A, S24A, Q92A, and R98A. Generally, these mutations map to predicted helix-1 and helix-3 (Fig. 4B). Binding of phage expressing R78A could not be detected for mAb 1646. Because this residue is predicted to be located distal to the other binding determinants, it is presumed that this mutation is important for maintaining global structural integrity.
Receptor Binding and Neutralizing Activities for Anti-TPO    (Fig. 5A). These values were reasonably close to their respective K d values for binding to TPO. In contrast, the IC 50 value for mAb 1646 inhibiting binding of the TPO to its receptor was approximately 10-fold higher (13.5 nM) than its K d for binding to TPO (13.5 nM versus 1.1 nM). mAbs 1643 and 1645 only FIG. 4. Alanine scan for TPO 1-153 -phage versus five monoclonal antibodies. A, each alanine mutant was tested for binding to five anti-TPO 1-153 mAbs using the TPO phage ELISA. The effects of the alanine mutations are expressed as EC 50 ratios relative to TPO  . For mutants with bars denoted as Ͼ20-fold, the ELISA signal was too weak for the competition phase of the experiment. Data represent the average of two trials with the exception of mAb 1645. B, residues that were determined to be important for mAb binding are shown labeled on the IL-4 structure. Color coding for the residues is described in Fig. 3B. weakly inhibited TPO receptor binding and showed only Ϸ30% inhibition of TPO receptor binding even at a concentration of 500 nM (Fig. 5A), yet they bind TPO with K d values of 4.0 and 4.8 nM, respectively (Table II). Thus, while all five the mAbs bound with similar affinities only two effectively blocked initial binding to receptor.
A MPL/Rse KIRA assay was used to monitor TPO-induced intracellular domain phosphorylation. A receptor fusion, consisting of the extracellular domain of the TPO receptor linked to the transmembrane and intracellular domains of the Rse kinase, was transfected into CHO cells; TPO-specific tyrosine phosphorylation was measured in these cells using an ELISAbased format. All five of the mAbs were able to neutralize nearly all of the TPO-induced phosphorylation with similar IC 50 values (Fig. 5B). As a control, mAbs that recognize the C-terminal domain of full-length TPO have no antagonistic effect (Fig. 5B). To confirm that the mAbs could also neutralize TPO activity for the full-length TPO receptor, a cell proliferation assay was performed using the cell line HU-03 which normally expresses the TPO receptor. Using 70 nM of each mAb, inhibition of TPO activity (at 1 pM TPO) was found to be 83.6, 80.1, 84.8, 90.6, and 87.3% for mAb 1643, 1645, 1646, 1653, and 1831, respectively (data not shown). Furthermore, all antibodies were fully capable of neutralizing TPO activity for BAF-3 cells that were transfected with the full-length TPO receptor (1) (data not shown). Thus, all the mAbs, including the three mAbs that do not block initial binding of TPO (1643, 1645, and 1646), effectively neutralize bioactivity of TPO in three cell-based assays.

DISCUSSION
The glycoprotein, TPO, is a key regulator of megakaryocyte and platelet production (1,2). This molecule has tremendous therapeutic potential for patients who have undergone chemotherapy and consequently suffer from thrombocytopenia. The N-terminal domain (amino acids 1-153) is sufficient for receptor-binding and megakaryocyte proliferation (1,5). The N-terminal domain of TPO is predicted to be a four-helix bundle like its most closely related homolog EPO (34) and more distantly related ones for which structures are known like hGH, granulocyte colony-stimulating factor, and interleukin-4. In the absence of the three-dimensional structure for TPO, a structural alignment was made based on the sequence and the crystal structure of IL-4; this model was used as a guide for sitedirected mutagenesis to begin to elucidate the functional features of TPO 1-153⅐ The TPO residues that were found to affect receptor binding could be divided into three classes. The first class consists of residues that caused greater than a 20-fold increase in the EC 50 (D8A and K138A). A second class of TPO mutants increased the EC 50 ranging from between five-and fifteen-fold (K14A, K52A, R136A, and R140). A third class of mutants reduced affinity between 2-and 5-fold (R17A, S24A, K59A, R98A, S106A, L129A, Q132A, H133A, and L144A). It is likely that other residues near these also play roles in receptor binding because only charged and polar residues were targeted for mutagenesis. Without structural information, we cannot rule out the possibility that some of the alanine mutations effect receptor binding by affecting the structure of the molecule. However, we believe the vast majority of these mutations cause only minor changes in structure because thirty-eight of the forty mutants tested bound to at least three of five anti-TPO mAbs.
Generally, the side chains important receptor-binding map to one side of the predicted four helix bundle (Fig. 3B). Most of the 15 functional residues reside on helix-1 and helix-4. Seven of the 15 identified residues map onto helix-4 and four of these are positively charged side chains (His 133 , Arg 136 , Lys 138 , and Arg 140 ). Most of the residues in EPO that align with the functionally important residues on helix-4 of TPO are predicted to be solvent-exposed in a model of EPO (1,34). Lys 138 in helix-4 of TPO plays a very significant role in receptor binding, yet may be involved in inner core packing of the four-␣-helix bundle as predicted by our sequence alignment with IL-4 ( Fig. 1A) and model (Fig. 3B). Tyr 145 in EPO aligns with Lys 138 of TPO (1), and a structural model of EPO (34) predicts that Tyr 145 is buried in the core of the globular structure. Thus, even though the K138A mutant in TPO bound to three of the five anti-TPO 1-153 mAbs, without a structure we cannot exclude the possibility that this causes local structural perturbations.
Most of the residues found to be important for receptor binding are completely conserved among humans, pigs, and mice (4). It is likely that the residues identified here for human TPO binding are also important for murine and pig TPO binding because TPO reacts broadly across these species (4). Slight variation does occur at minor binding determinants. For example, S24A which causes a 2.1-fold reduction in affinity is glycine in pig; S106 which causes a 2.2-fold reduction in affinity is aspartate in pig and glycine in mouse; His 133 which causes a 4.4-fold reduction in affinity is glutamine in both pig and mouse.
There are striking similarities between those regions found to be functional for TPO receptor binding and two other well characterized hematopoietic cytokine systems, hGH and EPO. For the hGH-receptor interaction, most of the binding determinants for initial receptor binding (site 1) are found on helix-4 (10,35). Most of the residues dominating receptor binding by EPO reside in regions predicted to be helix-1 and helix-4 (36). Overall, the alanine-scanning results for EPO corroborate very well with those presented here for TPO in that a number of positively charged side chains along helix-4 are important for modulating receptor binding.
Two of the alanine mutations, D18A and H20A, were found TABLE II Kinetic and equilibrium binding constants for binding of mAbs to TPO as determined from BIAcore Full-length TPO was randomly immobilized to the sensor chip via lysine residues at densities ranging from 1500 and 500 resonance units. Dilutions of each mAb were passed over the immobilized TPO and kinetic constants were obtained by nonlinear regression of association/ dissociation phases. Association rates were determined by linear regression of rates for various mAb concentrations. Errors represent the standard error of the linear regression analysis. Dissociation rates were determined using a two-component fit analysis and average k off was calculated using k off(ave) ϭ f (1) k off(1) ϩ f (2) k off (2) , where f is the fractional response attributed each rate. Errors are given as the standard deviation using at least three determinations. Apparent affinities were determined by K d ϭ k off(ave) /k on . to improve binding to the TPO receptor by 4-and 2-fold, respectively. Similarly, it was found that alanine mutations at several contact residues in helix-1 and helix-4 in hGH gave affinity improvements for hGH receptor binding (10,28). Interestingly, these steric hindrance residues in hGH (His 18 , Phe 25 , Gln 29 , Glu 65 , and Glu 174 ) are known to be critical for hGH binding to the prolactin receptor (8) or for forming a dimeric complex that chelates Zn 2ϩ (37). Thus, it is likely that Asp 18 and His 2O are buried at the TPO-receptor interface, and it is possible they play additional roles as yet to be determined.
Although a number of alanine mutations reduce the affinity of TPO for its receptor, most were found to have no effect (Fig.  3B). These "nonfunctional" residues are found in loops and throughout helices-1, -2, and -3. A number of these residues are involved in binding one or more of the neutralizing mAbs. Two of the antibodies (mAbs 1653 and 1831) that map to the Nterminal end of helix-2 and helix-4 are neutralizing for both receptor binding and TPO receptor activation (Fig. 5). These results suggest that this face of TPO is required for TPO receptor binding and activation, analogous to site 1 in hGH. mAbs 1643 and 1645 map to regions on helix-3 and do not block receptor binding but do inhibit receptor activation. Correlating this result to the receptor alanine-scanning data strongly suggests that residues near Ser 87 , Ser 88 , Gln 92 , and Ser 94 are involved in binding of possibly a second TPO receptor. In addition, mAb 1646 blocked TPO activity but was only moderately effective at competing for receptor binding. This antibody maps to helix-1 and helix-3 further indicating that residues on this side of the molecule (opposite that of site 1) are required for receptor activation.
A report by Alexander et al. (38) suggests that TPO activates its receptor by homodimerization. By engineering cysteine residues into a conserved receptor dimer interface region proximal to the membrane-spanning helix in the TPO receptor, receptor mutants were found that were constitutively active in a cell-based assay. Similar findings were first reported for the EPO receptor (39). Furthermore, chimeric receptors consisting of the extracellular domains of the granulocyte colonystimulating factor receptor (6,40) and the hGH receptor (41) fused to the intracellular region of the TPO receptor have demonstrated that TPO receptor homodimerization precedes signal transduction. CONCLUSIONS Using a structural alignment and alanine-scanning mutagenesis we have begun to dissect the interactions between TPO and its receptor. As for a number of other ligand-receptor systems within the hematopoietic family (hGH, EPO, and granulocyte colony-stimulating factor receptor), our data suggest that activation of the TPO receptor occurs by a sequential homodimerization mechanism. These results provide a foundation for further analysis of TPO-receptor interactions, and suggest yet another example of how a single hormone molecule can have multiple receptor binding sites to facilitate receptor oligomerization and consequent activation.
FIG. 5. Competition of TPO receptor binding and activity. A, TPO receptor binding inhibition assay. The TPO receptor (MPL-IgG) was captured on ELISA plates using an anti-hFc antibody. Each anti-TPO 1-153 mAb was serially diluted and then incubated with biotinylated full-length TPO (35 pM). Mixtures were then transferred to TPO receptor coated wells and amount of TPO bound was measured spectrophotometrically. Data was normalized to maximum signal (A 450 ) and is expressed as percent inhibition. Each data point represents the average between two experiments and the solid line is the best fit using a four parameter analysis. IC 50 values determined for mAb 1643 (q), 1645 (å), 1646 (□), 1653 (ࡗ), and 1831 (E) are 86.7, 51.2, 13.5, 1.7, and 1.3 nM, respectively. B, MPL/Rse KIRA assay for measurement of receptor intracellular domain phosphorylation. CHO cells were transfected with a gD-MPL/Rse chimeric receptor. Treatment of these cells with TPO results in activation of the fusion receptor via phosphorylation of the tyrosines in the Rse intracellular domain. Activity is measured spectrophotometrically after treatment and cell lysis using a gDcapture ELISA format with an anti-phosphotyrosine mAb. Each mAb was tested for the ability to block TPO activity (150 pM). Shown as controls are three mAbs 1664 (å]), 1682 (छ), and 1908 (ç) that recognize TPO within the COOH domain. Data represents the average of three trials. Symbols for the neutralizing mAbs are as denoted above.