Abstract

The extracellular ligand-binding domain (EPObp) of the human EPO receptor (EPOR) was expressed both in CHO (Chinese Hamster Ovary) cells and in Pichia pastoris. The CHO and yeast expressed receptors showed identical affinity for EPO binding. Expression levels in P.pastoris were significantly higher, favoring its use as an expression and scale-up production system. Incubation of EPO with a fourfold molar excess of receptor at high protein concentrations yielded stable EPO–EPObp complexes. Quantification of EPO and EPObp in the complex yielded a molar ratio of one EPO molecule to two receptor molecules. Residues that are responsible for EPOR glycosylation and isomerization in Pichia were identified and eliminated by site-specific mutagenesis. A thiol modification was identified and a method was developed to remove the modified species from EPObp. EPObp was complexed with erythropoietin (EPO) and purified. The complex crystallized in two crystal forms that diffracted to 2.8 and 1.9 Å respectively. (Form 1 and form 2 crystals were independently obtained at AxyS Pharmaceuticals, Inc. and Amgen, Inc. respectively.) Both contained one complex per asymmetric unit with a stoichiometry of two EPObps to one EPO.

Introduction

Erythropoietin (EPO) is the principal growth factor that induces proliferation and differentiation of erythroid progenitor cells (Koury and Bondurant, 1992). It is a member of the cytokine family that includes interleukins 2–7, G-CSF, GM-CSF, TPO, growth hormone and leptin (Bazan, 1990; Tartaglia et al., 1995). The EPO receptor binds EPO on the surface of erythroid cells with high affinity. The EPO receptor gene encodes a 508 amino acid protein comprised of a 249 amino acid extracellular EPO binding domain, a 22 amino acid transmembrane domain and a 237 amino acid cytoplasmic domain (D'Andrea et al., 1989; Jones et al., 1990; Winkelmann et al., 1990).

Binding of EPO to its receptor triggers signal transduction by ligand mediated receptor dimerization on the cell surface. Point mutations that introduce cysteine residues into the membrane proximal part of the extracellular domain of the EPO receptor, and that result in disulfide linked receptor dimers on the cell surface, are constitutively active; they lead to cell proliferation of EPO dependent cell lines and differentiation of red blood cell precursors in the absence of EPO (Yoshimura et al., 1990; Watowich et al., 1992; Watowich et al., 1994). Expression of these mutants in mice results in erythroleukemia through unregulated activation of the signaling pathway (Longmore and Lodish, 1991; Longmore et al., 1994). Truncated EPO receptors that lack most of their cytoplasmic domains do not lead to signaling, though EPO binding remains the same. When coexpressed with wild-type receptors the truncated form is dominant-negative suggesting a preference for forming inactive heterodimers, presence of which has been demonstrated by immuno precipitation (Barber et al., 1994; Watowich et al., 1994). EPO receptor activation has been shown to follow a sequential dimerization mechanism with binding to a high affinity site 1 on EPO preceding binding of the second receptor to a lower affinity site 2 on EPO (Matthews et al., 1996).

Direct biochemical evidence does not uniformly support the EPO dependent dimerization mechanism. On the one hand, light scattering, sedimentation equilibrium and titration calorimetry demonstrate that a recombinant form of the extracellular EPO receptor domain (EPObp) from mammalian cells forms a 2:1 receptor–EPO complex at high protein concentrations (Philo et al., 1996). On the other hand, other groups have reported that each EPO forms a complex with only one EPOR (Nagao et al., 1992; Yet and Jones, 1993).

Here we report the expression of EPObp in Pichia pastoris and characterize the resulting microheterogeneities that occur within this expression system. Systematic removal of sites of heterogeneity by mutagenesis, coupled with purification resulted in homogeneous EPObp that formed a complex with EPO. This solution complex was shown by analytical HPLC to have the stoichiometry of 2:1 EPObp:EPO. The homogeneity was essential for formation of diffraction quality crystals of the EPO–(EPObp)2 complex (Syed et al., 1998).

Materials and methods

EPOR expression in CHO cells

A human soluble EPO receptor (EPObp), encoding amino acids 1 through 249 of the published sequence (Bazan, 1990), was generated by the polymerase chain reaction (PCR) using the full-length cDNA as template and the oligonucleotide primers, 5′-ATGGACCACCTCGGGGCGT-3′ and 5′-CTAGGGGTCCAGGTC-3′. The amplification product incorporated a TAG termination codon downstream of the extracellular domain. The PCR product was subcloned into expression vector pRc/CMV (Invitrogen, San Diego, CA) and stably transfected into CHO cells. Individual clones secreting EPObp were selected by limiting dilution cloning and seeded into roller bottles (surface area 1700 cm2; Corning, NY). Cells were grown to confluency in RPMI plus 10% FCS (Irvine Scientific). Cells were washed twice and cultured in 200 ml serum-free RPMI-1640. Typical expression levels were 0.6 mg/l of cell supernatant. The product was purified from the supernatant by first diafiltrating with 20 mM Tris–HCl, pH 7.6, followed by a four step protocol which included Q- and Phenyl-Sepharose chromatographies (Pharmacia), an ammonium sulfate precipitation and a final Superdex 200 (Pharmacia) gel filtration. The concentration of EPObp was determined by UV absorbance using an extinction coefficient of 1.6 ml/mg/cm at 280 nm. The concentration was confirmed by amino acid analysis. SDS–PAGE demonstrated that EPObp was purified to apparent homogeneity.

EPObp expression in Pichia pastoris

An EPObp construct, encoding amino acids 1 through 225 of the native EPO receptor (excluding the receptor signal sequence) with three additional N-terminal residues R–3E–2 F–1P+1P+2, was amplified by PCR. The product was cloned into the EcoRI and BamHI sites of the Pichia pastoris expression vector pHIL-S1 (Invitrogen) and transformed into Pichia pastoris spheroplasts. Clones were selected for a homologous recombination event that inactivated the endogenous AOX1 (alcohol oxidase) gene and placed EPObp expression under control of the AOX1 methanol inducible promoter. Twenty clones were assayed for EPObp secretion by western blot analysis and an affinity ELISA using the antibody 2E12 (Schneider et al., 1997). Clones with a high level of EPObp expression were selected for use in scale-up production. Typical expression levels of EPObp were about 40 mg/l with ~75% high molecular weight and ~25% low molecular weight forms, as determined by ELISA. The two different forms can be separated by one chromatography step using a Phenyl Sepharose column (Pharmacia), eluted with loading buffer A [20 mM Tris–HCl, 1.5 M (NH4)2SO4, pH 7.6] to 100% buffer B (20 mM Tris–HCl, pH 7.6) over 20 column volumes. The concentration of EPObp was determined by UV spectroscopy using an extinction coefficient of 1.6 ml/mg/cm at 280 nm. The concentration was confirmed by amino acid analysis. EPObp was purified to apparent homogeneity judging from SDS–PAGE.

All EPObp was identified by western blot analysis using the antibody 2E12 and verified by N-terminal sequencing.

Affinity EPOR ELISA

An ELISA (enzyme-linked immunoadsorbent assay) measuring the binding affinity of EPObp for EPO was developed as follows: EPO was conjugated to horse radish peroxidase (HRP) using standard protocols (Pierce). 96-well plates (Nunc) were coated with EPObp monoclonal antibody 2E12 (1 μg/well) and incubated for 1 h at 37°C. The plate was washed and blocked with PBS containing 20 μg/ml BSA for 1 h at 37°C. Serial dilutions of purified EPObp or cell supernatants containing EPObp were added and incubated for 1 h at 37°C. After washing, 0.1 μg/well EPO–HRP conjugate was added and incubated for 1 h at 37°C. 100 μl TMB plus H2O2 (Pierce) was added to each well and the plate was incubated at room temperature for 5 min. The color reaction was stopped by adding 100 μl 2N H2SO4 to each well and concentrations of EPObp were determined using an absorbance of 450–650 nm on a plate reader (Molecular Devices, Sunnyvale, CA). Pure EPObp, derived from CHO cells and quantified by amino acid composition analysis, was used as a standard for this assay.

Inhibition of 1511 cell proliferation assay—a EPObp binding competition assay

Proliferation assays using the cell line 1511, a BaF3-derived cell line transfected with the human EPO receptor gene, were described by Matthews et al. (1996). Except during the assay, various concentrations of EPObp (0.001–100 nM) were added to compete binding for EPO with the cell surface EPO receptor.

Expression of various forms EPObp in P.pastoris

Two mutant EPObps were constructed by site-specific mutagenesis: a single N52Q construct and the triple mutant EPObp 3D(N52Q, N164Q, A211E). Both proteins were produced under high density conditions in a 10 l fermentor. Typical yield was about 200 mg/l at high density fermentation. These proteins were purified from the P.pastoris supernatant similarly to that described for CHO system except without the use of ammonium sulfate precipitation. In addition, a shorter purification was developed utilizing affinity chromatography. The affinity column was made with the EPObp monoclonal antibody 5C1.8 Mab (a non-neutralizing antibody) (Chaovapong,W. and Schneider,H., personal communication) which was conjugated to azolactone derivatized beads (Pierce) via amine coupling (>90% coupling efficiency). The P.pastoris cells were removed by centrifugation and the supernatants were filtered through a 0.2 μm membrane followed by diafiltration with 200 mM NaCl, 20 mM Tris–HCl, pH 7.6, using a Pellicon system (Millipore). The diafiltrates were run through a Q-Sepharose column prior to loading on the affinity column. A Q-Sepharose step removes significant amounts of contaminating proteins and extends the lifetime of the affinity column significantly. The receptor was eluted with 0.5 M formic acid and the pH was immediately neutralized by the addition of 3 M Tris–HCl, pH 9.0.

Native gel electrophoresis was run in Tris–glycine buffer, pH 8.4. A Multiphore II system (Pharmacia) was employed for isoelectric focusing gels (IEF) with 1 M H3PO4 as anode solution and 1 M NaOH as cathode solution. HPLC reverse phase chromatography was run on a HP1090.

Erythropoietin (EPO), derived both from recombinant CHO cells (size selected for the 35 kDa species) and Escherichia coli, were from Amgen, Inc. (Thousand Oaks, CA). EPOR monoclonal antibody 2E12 was prepared and purified as described previously (Schneider et al., 1997).

EPO–EPObp complex formation and characterization

For all the analytical studies, protein samples were analyzed by gel filtration HPLC at room temperature using a Bio-Sil sec-250, 300×7.8 mm column (Bio-Rad). A mixture of thyroglobulin, IgG, ovalbumin, myoglobin and vitamin B12 (Bio-Rad) was used for the molecular weight calibration. Column void volume was determined by using Blue dextran 2000 (Pharmacia). Plots of ratio of elution volume and void volume (Ve/Vo) versus log MW were generated and used as a standard for determining the apparent molecular weight of various forms of proteins. Different molar ratios of EPO and EPObp were mixed and allowed to reach equilibrium over a 2 h period at room temperature before injecting onto the column for separation.

To obtain relatively large quantities of 2:1 EPObp/EPO complex, EPO was typically incubated with a fourfold molar excess of EPObp. Total protein concentration was approximately 94 μM for EPObp and 22 μM for EPO. The complex was purified by gel filtration with a Superdex 75 (Pharmacia) column equilibrated in 20 mM Tris–HCl, pH 7.5, 150 mM NaCl and 10 mM EDTA. Thus, the EPO/(EPObp)2 complex was eluted in the excluded volume. Column peaks were analyzed by SDS–PAGE and compared with EPO and EPObp standards. A typical dilution factor was three as calculated by the ratio of width (ml) at half-height of the peak to the injection volume.

The ratio of EPO to EPObp was determined by injecting the purified complex onto a C-4 (YMC-Pack C4-AP, 5 mm, 300 Å, 250×4.6 mm ID, YMC, Inc.) reverse phase analytical column in buffer A (0.05% TFA in water). EPO and EPObp were eluted with a gradient of 5–90% buffer B (0.05% TFA in acetonitrile). The EPO and EPObp protein peaks were quantified spectrometrically by UV absorption at 214 nm. The EPO and EPObp ratio in the complex was calculated using the equation 

formula
where R is the ratio of EPO to BEPObp. AEPO and AEPObp are the integrated peak areas at 214 nm of EPO and EPObp respectively. BEPO and BEPObp are the number of peptide bonds in EPO (166) and EPObp (228), respectively.

This equation was validated by generating a standard detector response curve of each protein at 214 nm. The peak area for each protein in the complex after dissociation was compared with the above standard curve to determine the protein ratio. This equation should have general applications in quantifying the stoichiometric relationship of proteins in the complex.

Sugar determination

Neutral hexose was determined as described previously (Zhan et al., 1990), using mannose as a standard.

Free thiol analysis

Conditions for Cys-specific fluorescence labeling and oxidative disulfide formation reactions were carried out according to the published methods (Zhan et al., 1994).

Mass spectrometry

All mass spectrometry was performed on a Finnigan-MAT (San Jose, CA) TSQ-7000 triple quadrapole mass spectrometer equipped with a Finnigan-MAT electrospray ionization source, coupled to a Hewlett-Packard 1050 liquid chromatography system. The samples were scanned over a range of m/z 500–2000 in 1.5 s.

EPO–EPObp (3D) complex crystallization, diffraction measurements and data processing

The EPO/(EPObp)2 complex was crystallized in two separate crystal forms.

Form 1

Crystals of the EPO–(EPObp)2 complex were grown by hanging drop vapor diffusion at 21°C. EPO was from the E.coli expression system. Extensive screening of conditions, incorporating crystal screens I and II (Hampton Research) were used to establish initial crystallization conditions. Typically, 3 μl of a 10 mg/ml complex solution was mixed with an equal volume of the crystallization well-buffer containing 13–15% PEG 4000, ±0.2 M CaCl2 and 0.1 M Na MES pH 6.5–7.0. Once an initial stock of crystals had been established, microseeding was used to initiate crystal growth. For data collection, crystals were harvested at room temperature in reservoir buffer.

Crystals were screened for their ability to diffract using a Siemens IPC X1000 multiwire area detector mounted on a Rigaku RU200 generator using a rotating copper anode target tube operating at 50 kV, 60 mA. The data sets were collected with an R-AXIS II image plate system mounted on the same generator. Data were collected at room temperature and required three crystals for a complete data set. All diffraction data were indexed, integrated, scaled and merged with the R-AXIS associated data reduction software Biotex (Molecular Structure Corporation, The Woodlands, TX 77381).

Form 2

Crystals were grown in hanging drops from 2:1 EPObp/EPO complex (4.7 mg/ml) using cryoprotectant conditions, 32% PEG 1500, 280 mM ammonium sulfate, 100 mM MES buffer (pH 6.5) at 20°C.

Data were collected with a single crystal (containing one complex per asymmetric unit with a solvent content of 46%) at 100 K on an R-Axis IV imaging system installed on an 18 kW Rigaku RU300 generator. Data were indexed, integrated and scaled with the programs DENZO and SCALEPACK.

Results and discussion

Expression of EPObp

Two systems were assessed for their ability to yield high levels of EPObp expression. An immunoblot analysis of the supernatant from CHO cell expression revealed a single 30 kDa EPObp band (predicted non-glycosylated weight of 25.1 kDa) and yields were typically 0.6 mg/l as measured by ELISA. To explore expression systems that would allow for increased EPO receptor secretion, EPObp was also expressed in the yeast P.pastoris. Shaker flasks typically yielded an expression level of 40 mg/l, showing a 70-fold improvement over CHO cells. In addition, P.pastoris culture supernatants contained very little contaminating protein which greatly facilitated purification. Therefore, P.pastoris became the organism of choice for large scale production. SDS–PAGE and immunoblotting of the Pichia cell supernatant revealed two EPObp bands, a 30 kDa band similar in size to the CHO-produced EPObp and a diffuse 60 kDa band. The appearance of this diffuse 60 kDa band suggests a hyperglycosylated form of EPObp (Figure 1).

Purification and characterization of EPObp

CHO cell expressed EPObp was purified using three separate chromatographic steps: first Q-Sepharose (anion exchange), then phenyl Sepharose (hydrophobic interaction) followed by ammonium sulfate precipitation, and finally Superdex-200 size exclusion chromatography. The resultant EPObp was purified to apparent homogeneity as demonstrated by SDS–PAGE, and N-terminal protein sequencing. Purification of P.pastoris-expressed EPObp and separation of the 30 and 60 kDa EPObp species required only a single Phenyl Sepharose chromatographic step to achieve apparent homogeneity (Figure 1). Thus the high expression levels in P.pastoris translate into high levels of purified EPObp. Carbohydrate analysis indicates that polysaccharide leads to the hyperglycosylated high molecular weight form (data not shown).

EPO-binding affinities of the 30 kDa CHO EPObp species and those of the 30 and 60 kDa P.pastoris EPObp species were analyzed by a binding competition cell proliferation assay. IC50 values for the 30 kDa CHO and P.pastoris EPObp are identical (~1.5 nM) in agreement with published data (Harris et al., 1992; Yet and Jones, 1993; Johnson et al., 1996). IC50 for the 60 kDa EPObp is twofold lower (~3 nM) suggesting that hyperglycosylation slightly reduces its affinity for EPO.

EPO–EPObp complex formation and determination of stoichiometry

EPO/(EPObp)2 was formed at room temperature within 2 h by incubation of EPO (22 μM) with a fourfold molar excess of EPObp (94 μM), which should favor complex formation. Gel filtration separated EPO–(EPObp)2 from free EPObp (Figure 2). To precisely quantify the ratio of EPO and EPObp in the complex, the two proteins from peak A were separated by C4 reverse phase chromatography and the protein peaks were quantified by UV absorption at 214 nm wavelength. The molar ratios of EPObp to CHO EPO in these complexes were 2.2:1.0 for CHO EPObp (Figure 2B), 2.0:1.0 for P.pastoris 30 kDa EPObp and 2.1:1.0 for 60 kDa P.pastoris EPObp.

Analytical HPLC gel filtration carried out on a Bio-Sil sec-250 column was used to further confirm the stoichiometry by measuring the hydrodynamic size of the proteins. The apparent molecular weights were 65 and 35 kDa for P.pastoris-produced high molecular weight and low molecular weight EPObp, respectively, 35 kDa for CHO EPObp and 65 kDa for CHO EPO. The molecular weights for all forms of EPObp, based on the hydrodynamic radii, are in good agreement with those determined by SDS–PAGE. However, the molecular weights determined for CHO EPO differed significantly (34 kDa by SDS–PAGE versus 65 kDa from the hydrodynamic radius). It was also observed by Davis et al. (1987) that the Stokes' radius of CHO EPO was 32 Å by gel filtration, much larger than the 20 Å radius calculated for a sphere based on the SDS–PAGE-derived molecular weight. The difference may highlight the contribution of carbohydrate (~40% of total mass) to the EPO hydrodynamic volume. The apparent molecular weight of the EPO–(EPObp)2 complex was 150 kDa, consistent with two EPObp (2× 35 kDa) and one EPO (65 kDa) (Figure 3).

Identifying and eliminating the glycosylated residue on EPObp by site-directed mutagenesis (N52Q)

Although P.pastoris expressed protein has the advantage of preventing O-linked glycosylation, N-linked glycosylation is frequent. There is a consensus -NYS- N-glycosylation sequence in EPObp. Enzymatic removal of the N-linked polysaccharide with Endo-H reveals a sharp band on SDS–PAGE at the expected deglycosylated molecular weight. For structural studies, the mutation N52Q was produced by site-directed mutagenesis to abrogate glycosylation. N-terminal sequencing showed that the protein had been correctly processed. Quantitative amino acid analysis confirmed that the molar ratios of stable amino acids are consistent with the predicted composition. Electrospray mass spectrometry gave a molecular mass of 25 195.3 Da (predicted Mr 25 172), suggesting that there is no major post translational modification to EPObp. As with wild-type EPObp, the N52Q mutant EPObp binds to EPO with similar affinity and forms a 2:1 complex, consistent with the observation that N-glycosylation-defective EPO receptor can induce the ligand-dependent cell proliferation signal (Nagao et al., 1995). Crystallization of this complex was successfully achieved, although the resolution of X-ray diffraction was poor, indicating that some heterogeneity remained.

Elimination of isoformylization by changing Asn-Gly to Gln-Gly

The EPObp N52Q protein, which gave a single sharp band on SDS–PAGE, was separated on a native gel (Figure 4A and B) and a IEF gel into two distinct species. Since the two species have similar molecular weights, finding a difference between the two focused on their relative charges. Charge differences could be due to two isomers created from the isoformylization reaction, in which the amide of the Asn side chain in the sequence Asn164–Gly165 can substitute for the peptide amide rearrangement. It is known that, due to the absence of a glycine side chain favoring succinimide formation on the neighboring residue, the deamination reaction of the asparagine side chain can also occur 30–50-fold more rapidly. We confirmed this to be the case by treating EPObp with hydroxylamine, known to hydrolyze the Asn–Gly peptide bonds by nucleophilic attack. After incubation, the deaminated form, that has one more negative charge, remained intact while the other form was preferentially cleaved (Figure 4C). The two isoforms were partially separated based on their differences in pI by chromatography on monoP (pH from 4.0 to 7.0). No conversion was noticeable in 20 mM Tris–HCl, pH 7.2–8.0 and both forms were stable under these storage conditions. These two forms were detected in the fermentation mixture by western blot analysis of a native gel. Therefore both isoforms may be secreted into the media simultaneously by P.pastoris or the isomerization may take place during the fermentation process. This further source of heterogeneity was eliminated by site-directed mutagenesis (N164Q).

A third mutation, A211E, was introduced into the WSXWS box, highly conserved among the cytokine receptor family, since it had been demonstrated to increase cell surface expression of the full length receptor by enhancing folding in the endoplasmic reticulum in both transfected and normal hematopoietic cells (Hilton et al., 1995). The three mutations from the wild-type sequence (N52Q, N164Q, A211E; pI 5.01) lead to a protein that migrated at the identical position on a native gel as the deaminated EPObp N52Q, consistent with the predicted charge (Figure 4D).

Identifying glutathione modifications and free cysteine

Further analysis of the native, or of the IEF gel, reveals a relatively small amount of a third more negatively charged form, which varies in quantity between different fermentation runs. The two forms are separated by high resolution anion exchange chromatography, based on a single charge difference, using a source 15Q resin (Pharmacia) (Figure 5). Mass spectrometry of this species showed it to have a molecular weight ~300 Da greater than the major form (Figure 6A and B). The quantity of this minor species is diminished on treatment with the reducing agent DTT, consistent with the idea that an unpaired cysteine can be modified by glutathione, the principal redox buffer in the endoplasmic reticulum (Meister and Anderson, 1983; Hwang et al., 1992).

In EPObp, there is one unpaired cysteine (C181) which may serve as the glutathione modification site. A solvent exposed cysteine, like the EPObp C-terminal engineered cysteine, can be oxidized during the P.pastoris fermentation (data not shown). However, an embedded or partially buried cysteine is more likely to be found in the reduced state. If so, this site would be of potential use as a heavy metal derivative site for structure determination. If it is oxidized from SH to SO2 or SO3, the charge difference should be detected on a native or IEF gel. To prove this experimentally, different thiol specific reactive compounds were reacted with the protein. It was found that this protein can be readily reacted with organomercury (Figure 7A) or the hydrophobic thiol-reactive compound fluorescein-5-maleimide (Figure 7B), suggesting that the sulfhydril was reduced, and at high concentration (>10 μM), the EPObp protein was subjected to conditions that would favor oxidative disulfide formation (Zhan et al., 1994). However, no covalent dimers were found by non-reducing SDS–PAGE, indicating that the protein has a naturally protected thiol group that may be useful for heavy metal derivation to assist X-ray structure determination.

Assessment of homogeneous complex formation by crystallization form 1

Crystals of the EPO–EPObp complex, using E.coli-produced EPO and P.pastoris-produced EPObp (N52Q, N164Q, A211E), were obtained after an extensive and broad screen of conditions, followed by optimization. Using a set of optimized PEG-based conditions, and nucleation by seeding, crystals grew to a maximum size (approximately 0.5×0.2×0.1 mm) within 2 weeks. The complex crystallizes in space group P212121 with average unit cell dimensions 73.3×80.3×134.9 Å and diffraction to 2.8 Å. However, crystals grown even within the same drop diffract to different resolutions with each crystal, often to only 7–8 Å resolution and the unit cell dimensions varied co-ordinately by up to 3%, as if different degrees of hydration expand the three cell dimensions co-ordinately. Occasionally different crystals within the same solution had similar cell dimensions and diffracted to ~2.8 Å resolution. These crystals (Figure 8A) were washed, dissolved and analyzed by quantitative reverse phase HPLC and found to contain EPO/EPObp in the ratio 1:2 (Figure 8B). These crystals gave rise to the first structure of the complex (Syed et al., 1998).

Crystallization form 2

Higher resolution data (1.9 Å) was eventually obtained from a new crystal form, which used an EPO analog with two additional mutations, P121N and P122S, introduced into the CD loop. This analog of EPO, when complexed with P.pastoris EPObp in 1:2 molar ratio (determined by absorbance at 280 nm), crystallized in a new unit cell (a, 58.3 Å; b, 79.4 Å; c, 136.5 Å; P212121; form 2) with a diffraction limit beyond 1.9 Å. These crystals have extremely similar b and c cell dimensions, but are shorter in the a cell dimension by ~15 Å due to a different packing in this direction. Crystals were grown in hanging drops from 2:1 EPObp–EPO complex (4.7 mg/ml) using cryoprotectant conditions, 32% PEG 1500, 280 mM ammonium sulfate, 100 mM MES buffer (pH 6.5) at 20°C. These crystals grow in flower like cluster formation to an approximate size of 0.7×0.4×0.2 mm. The crystal structures of forms 1 and 2 were subsequently determined as we report (Syed et al., 1998).

In summary, we have established a general protocol for the production of soluble extracellular EPO receptor domains (EPObp) and demonstrate that this molecule dimerizes upon EPO binding to form stable complexes, both in solution and in the crystalline state. Detailed analysis of EPObp has resulted in the construction of a triple mutant EPObp molecule which eliminates the micro-heterogeneity encountered in the Pichia expression system. The enhancement of EPO–(EPObp)2 purity led to diffraction quality crystals that led to structural determination by X-ray crystallography, and then to improvement in resolution of the structure by subsequent mutational tuning. The iterative pathway we followed is important, as so often the pathway to successful characterization and crystallography is detailed, processive and iterative, often taking years of diligent molecular biology and protein chemistry, as was the case here.

Fig. 1.

SDS–PAGE analysis of P.pastoris produced EPObp. Lane 1 shows the supernatant containing the two distinct EPObp species. After a single step Phenyl Sepharose chromatography purification, 60 kDa apparently hyperglycosylated form (lane 2) can easily be separated from the 30 kDa form (lane 3). The gel was stained with Coomassie brilliant blue.

Fig. 1.

SDS–PAGE analysis of P.pastoris produced EPObp. Lane 1 shows the supernatant containing the two distinct EPObp species. After a single step Phenyl Sepharose chromatography purification, 60 kDa apparently hyperglycosylated form (lane 2) can easily be separated from the 30 kDa form (lane 3). The gel was stained with Coomassie brilliant blue.

Fig. 2.

Analysis of EPO–EPObp complex formation. A fourfold molar excess of CHO EPObp was incubated with CHO EPO. The mixture was injected onto a Superdex 75 (Pharmacia) FPLC gel filtration column and two peaks were observed (A). Protein peaks were monitored at 280 nm. The elution time (x-axis) was expressed in minutes. Peaks A and B were collected and analyzed by SDS–PAGE and show two bands for peak A (lane A), a higher molecular weight band corresponding to EPO and a lower molecular weight band corresponding to EPObp, while peak B contained only excess EPObp (lane B). Identical results were obtained using the 30 and 60 kDa P.pastoris EPObp proteins (data not shown). Peak A which was in the excluded volume and peak B were collected and analyzed by reducing SDS–PAGE stained with Coomassie brilliant blue. Lane A represents peak A containing the EPO–EPObp complex and lane B represents peak B containing the excess EPObp. (B) Quantitative analysis of EPO–EPObp complex. After gel filtration, peak A was collected and its components were separated by reverse phase chromatography. The integrated peak areas were used to calculate the molar ratio of EPObp to EPO. Signals were monitored at 214 nm UV. The ratio of EPO/EPObp was 1 to 2.

Fig. 2.

Analysis of EPO–EPObp complex formation. A fourfold molar excess of CHO EPObp was incubated with CHO EPO. The mixture was injected onto a Superdex 75 (Pharmacia) FPLC gel filtration column and two peaks were observed (A). Protein peaks were monitored at 280 nm. The elution time (x-axis) was expressed in minutes. Peaks A and B were collected and analyzed by SDS–PAGE and show two bands for peak A (lane A), a higher molecular weight band corresponding to EPO and a lower molecular weight band corresponding to EPObp, while peak B contained only excess EPObp (lane B). Identical results were obtained using the 30 and 60 kDa P.pastoris EPObp proteins (data not shown). Peak A which was in the excluded volume and peak B were collected and analyzed by reducing SDS–PAGE stained with Coomassie brilliant blue. Lane A represents peak A containing the EPO–EPObp complex and lane B represents peak B containing the excess EPObp. (B) Quantitative analysis of EPO–EPObp complex. After gel filtration, peak A was collected and its components were separated by reverse phase chromatography. The integrated peak areas were used to calculate the molar ratio of EPObp to EPO. Signals were monitored at 214 nm UV. The ratio of EPO/EPObp was 1 to 2.

Fig. 3.

Analytical HPLC analysis of the EPO–EPObp complex. A mixture of CHO EPObp (1.9 nmol) and CHO EPO (0.34 nmol) in 20 μl after incubation was separated by gel filtration. CHO EPO was incubated with an ~4-fold molar excess of CHO EPObp and the mixture was separated by analytical gel filtration chromatography. The elution time (x-axis) was expressed in minutes. The first peak at 17.58 min has an apparent molecular weight of 150 kDa and represents the 2:1 receptor–ligand complex. The second peak at 20.5 min corresponds to the excess free EPObp. The peak around 25 min represents the Tris buffer. Peaks were monitored at 214 nm UV. Analysis of peak areas showed that 1.9 nmol CHO EPObp produced an area of 22 821 mAUs (milli area units×s) while the 0.34 nmol EPO resulted in a peak area of 3070 mAUs. Assuming that all the EPO was bound in the 2:1 complex, the calculated area of the complex peak would be 11 238 mAUs, which is in good agreement with the observed complex peak area of 11 145 mAUs. Free EPObp has a peak area of 14 653 mAUs, consistent with the observed area of 14 263 mAUs for the second peak.

Fig. 3.

Analytical HPLC analysis of the EPO–EPObp complex. A mixture of CHO EPObp (1.9 nmol) and CHO EPO (0.34 nmol) in 20 μl after incubation was separated by gel filtration. CHO EPO was incubated with an ~4-fold molar excess of CHO EPObp and the mixture was separated by analytical gel filtration chromatography. The elution time (x-axis) was expressed in minutes. The first peak at 17.58 min has an apparent molecular weight of 150 kDa and represents the 2:1 receptor–ligand complex. The second peak at 20.5 min corresponds to the excess free EPObp. The peak around 25 min represents the Tris buffer. Peaks were monitored at 214 nm UV. Analysis of peak areas showed that 1.9 nmol CHO EPObp produced an area of 22 821 mAUs (milli area units×s) while the 0.34 nmol EPO resulted in a peak area of 3070 mAUs. Assuming that all the EPO was bound in the 2:1 complex, the calculated area of the complex peak would be 11 238 mAUs, which is in good agreement with the observed complex peak area of 11 145 mAUs. Free EPObp has a peak area of 14 653 mAUs, consistent with the observed area of 14 263 mAUs for the second peak.

Fig. 4.

Identifying EPObp heterogeneities produced from P.pastoris. (A) Purified EPObp N52Q migrated at the position of 25.5 kDa as a single band on a 12% reducing SDS–PAGE. (B) However, two distinct bands were separated by native gel electrophoresis (lane 1). The protein failed to bind to ConA. It flowed through the ConA Sepharose column in the binding buffer: pH 7.5, 50 mM Tris, 100 mM NaCl, 1 mM CaCl2, 1 mM MnCl2, consistent with the lack of polysaccharide modification. (C) EPObp N52Q was incubated at pH 9.0 at 42°C for 16 h in the absence of 2 M hydroxylamine (lane 1) or presence of 2 M hydroxylamine. The protein was separated by native gel electrophoresis, showing that the slower migrating band was preferentially cleaved. (D) Comparison of electrophoretic mobility profiles on a native gel. Lane 1, N52Q and lane 2, triple mutant (N52Q, N164Q, A211E).

Fig. 4.

Identifying EPObp heterogeneities produced from P.pastoris. (A) Purified EPObp N52Q migrated at the position of 25.5 kDa as a single band on a 12% reducing SDS–PAGE. (B) However, two distinct bands were separated by native gel electrophoresis (lane 1). The protein failed to bind to ConA. It flowed through the ConA Sepharose column in the binding buffer: pH 7.5, 50 mM Tris, 100 mM NaCl, 1 mM CaCl2, 1 mM MnCl2, consistent with the lack of polysaccharide modification. (C) EPObp N52Q was incubated at pH 9.0 at 42°C for 16 h in the absence of 2 M hydroxylamine (lane 1) or presence of 2 M hydroxylamine. The protein was separated by native gel electrophoresis, showing that the slower migrating band was preferentially cleaved. (D) Comparison of electrophoretic mobility profiles on a native gel. Lane 1, N52Q and lane 2, triple mutant (N52Q, N164Q, A211E).

Fig. 5.

A native PAGE showing that the glutathionylated EPObp 3D can be readily separated from the unmodified form by anion exchange chromatography using a 15Q source column (Pharmacia). The modified version moved faster on native PAGE. The unmodified species was step eluted out at the conductance of 15.5 ms/cm, while the one modified was eluted out at 17 ms/cm. The buffer was 20 mM Tris–HCl, pH 7.6 and salt was NaCl.

Fig. 5.

A native PAGE showing that the glutathionylated EPObp 3D can be readily separated from the unmodified form by anion exchange chromatography using a 15Q source column (Pharmacia). The modified version moved faster on native PAGE. The unmodified species was step eluted out at the conductance of 15.5 ms/cm, while the one modified was eluted out at 17 ms/cm. The buffer was 20 mM Tris–HCl, pH 7.6 and salt was NaCl.

Fig. 6.

ESI mass spectrum. Spectrum (A) shows mass of the unmodified EPObp 3D is 25 266.7 (calculated Mr 25 244) and (B) shows the modified form with mass of 25 577.1. The difference of the mass was 310.4.

Fig. 6.

ESI mass spectrum. Spectrum (A) shows mass of the unmodified EPObp 3D is 25 266.7 (calculated Mr 25 244) and (B) shows the modified form with mass of 25 577.1. The difference of the mass was 310.4.

Fig. 7.

EPObp 3D free cysteine detection. (A) SDS–PAGE. EPObp in 20 mM Tris–HCl, 150 mM NaCl (lane 1) was bound to Hg-derivative gel (Affi-gel 501) (Bio-Rad). The protein was then eluted with the same buffer, in the presence of 20 mM DTT (lane 2) or absence of DTT (lane 3). (B) Fluorescence labeling. A thiol-specific compound, fluorescein-5-maleimide, was reacted with EPObp 3D in the absence of DTT (lane 1) or in the presence of DTT (lane 2). The fluorescence of protein bands in the SDS–PAGE gel was visualized with an UV transilluminator, and the gel was subsequently stained with Coomassie brilliant blue to confirm the protein bands.

Fig. 7.

EPObp 3D free cysteine detection. (A) SDS–PAGE. EPObp in 20 mM Tris–HCl, 150 mM NaCl (lane 1) was bound to Hg-derivative gel (Affi-gel 501) (Bio-Rad). The protein was then eluted with the same buffer, in the presence of 20 mM DTT (lane 2) or absence of DTT (lane 3). (B) Fluorescence labeling. A thiol-specific compound, fluorescein-5-maleimide, was reacted with EPObp 3D in the absence of DTT (lane 1) or in the presence of DTT (lane 2). The fluorescence of protein bands in the SDS–PAGE gel was visualized with an UV transilluminator, and the gel was subsequently stained with Coomassie brilliant blue to confirm the protein bands.

Fig. 8.

(A) A photograph of a typical crystal of the EPO–(EPObp)2 complex. EPO was produced from E.coli, with a molecular weight of ~18 kDa. The bar represents 0.15 mm. (B) Crystals were isolated, washed and dissociated either on a reverse phase HPLC (2 EPObp/EPO was observed) or on SDS–PAGE.

Fig. 8.

(A) A photograph of a typical crystal of the EPO–(EPObp)2 complex. EPO was produced from E.coli, with a molecular weight of ~18 kDa. The bar represents 0.15 mm. (B) Crystals were isolated, washed and dissociated either on a reverse phase HPLC (2 EPObp/EPO was observed) or on SDS–PAGE.

4
To whom correspondence should be addressed

We thank Dr John Winklemann for the EPO receptor cDNA clone, Dr Mary McGrath for helpful discussions, Dr Bill Moore for critical reading and Dr Mike Venuti for strong support.

References

Bazan,J.F. (
1990
)
Proc. Natl Acad. Sci. USA
 ,
87
,
6934
–6938.
Barber,D.L., DeMartino,J.C., Showers,M.O. and D'Andrea,A.D. (
1994
)
Mol. Cell. Biol.
 ,
14
,
2257
–2265.
D'Andrea,A.D., Lodish,H.F. and Wong,G.G. (
1989
)
Cell
 ,
57
,
277
–285.
Davis,J.M., Arakawa,T., Strickland,T.W. and Yphantis,D.A. (
1987
)
Biochemistry
 ,
26
,
2633
–2638.
Harris,K.W., Mitchell,R.A. and Winkelman,J.C. (
1992
)
J. Biol. Chem.
 ,
267
,
15205
–15209
Hilton,D.J., Watowich,S.S., Murray,P.J. and Lodish,H.F. (
1995
)
Proc. Natl Acad. Sci. USA
 ,
92
,
190
–194.
Hwang,C., Sinsker,A.J. and Lodish,H.F. (
1992
)
Science
 ,
257
,
517
–523.
Johnson,D.L., Middleton,S.A., McMahon,F., Barbone,F.P., Kroon,D., Tsao,E., Lee,W.H., Mulcahy,L.S. and Jolliffe,L.K. (
1996
)
Protein Exp. Purif.
 ,
7
,
104
–113.
Jones,S.S., D'Andrea,A.D., Haines,L.L. and Wong,G.G. (
1990
)
Blood
 ,
76
,
31
–35.
Koury,M.J. and Bondurant,M.C. (
1992
)
Eur. J. Biochem.
 ,
210
,
649
–663.
Longmore,G.D. and Lodish,H.F. (
1991
)
Cell
 ,
67
,
1089
–1102.
Longmore,G.D., Pharr,P.N. and Lodish,H.F. (
1994
)
Mol. Cell. Biol.
 ,
14
,
2266
–2277.
Matthews,D.J., Topping,R.S., Cass,R.T. and Giebel,L.B.. (
1996
)
Proc. Natl Acad. Sci. USA
 ,
93
,
9471
–9476.
Meister,A. and Anderson,M.E. (
1983
)
Annu. Rev. Biochem.
 ,
52
,
711
–760.
Nagao,M., Masuda,S., Abe,S., Masatsugo,U. and Sasaki,R. (
1992
)
Biochem. Biophys. Res. Commun.
 ,
188
,
888
–897.
Nagao,M., Morishita,E., Hanai,Y., Kobayashi,K. and Sasaki,R. (
1995
)
FEBS Lett.
 ,
373
,
225
–228.
Philo,J.S., Aoki,K.H., Arakawa,T., Narhi,L.O. and Wen,J. (
1996
)
Biochemistry
 ,
35
,
1681
–1691.
Schneider,H., Chaovapong,W., Matthews,D.J., Karkaria,C., Cass,R.T., Zhan,H., Boyle,M., Lorenzini,T., Elliott,S.G. and Giebel,L.B. (
1997
)
Blood
 ,
89
,
473
–482.
Syed,R.S. et al. (
1998
)
Nature
 ,
395
,
511
–516.
Tartaglia,L.A. et al. (
1995
)
Cell
 ,
83
,
1263
–1271.
Watowich,S.S., Yoshimura,A., Longmore,G.D., Hilton,D.J., Yoshimura,Y. and Lodish,H.F. (
1992
)
Proc. Natl Acad. Sci. USA
 ,
89
,
2140
–2144.
Watowich,S.S., Hilton,D.J. and Lodish,H.F. (
1994
)
Mol. Cell. Biol.
 ,
14
,
3535
–3549.
Winkelmann,J.C., Penny,L.A., Deaven,L.L., Forget,B.G. and Jenkins,B.B. (
1990
)
Blood
 ,
76
,
24
–30.
Yet,M.G. and Jones,S.S. (
1993
)
Blood
 ,
82
,
1713
–1719.
Yoshimura,A., Longmore,G. and Lodish,H.F. (
1990
)
Nature
 ,
348
,
647
–649.
Zhan,H., Gray,J.A., Levery,S.B., Rolfe,B.G. and Leigh,J.A. (
1990
)
J. Bacteriol.
 ,
172
,
5245
–5253.
Zhan,H., Choe,S., Huynh,P.D., Finkelstein,A., Eisenberg,D. and Collier,R.J. (
1994
)
Biochemistry
 ,
33
,
11254
–11263.