Wilson Layout - Nature

2 downloads 0 Views 656KB Size Report
Correspondence should be addressed to L.K.J. email: [email protected] and I.A.W. email: wilson@scripps.edu. Receptor oligomerization in the form of ...
1998 Nature America Inc.

• http://structbio.nature.com

articles

An antagonist peptide–EPO receptor complex suggests that receptor dimerization is not sufficient for activation

1998 Nature America Inc.

• http://structbio.nature.com

Oded Livnah1,6, Dana L. Johnson2, Enrico A. Stura1,7, Francis X. Farrell2, Francis P. Barbone2, Yun You4, Kathleen D. Liu4, Mark A. Goldsmith3,4, Wen He5, Christopher D. Krause5, Sidney Pestka5, Linda K. Jolliffe2 and Ian A. Wilson1 Dimerization of the erythropoietin (EPO) receptor (EPOR), in the presence of either natural (EPO) or synthetic (EPO-mimetic peptides, EMPs) ligands is the principal extracellular event that leads to receptor activation. The crystal structure of the extracellular domain of EPOR bound to an inactive (antagonist) peptide at 2.7 Å resolution has unexpectedly revealed that dimerization still occurs, but the orientation between receptor molecules is altered relative to active (agonist) peptide complexes. Comparison of the biological properties of agonist and antagonist EMPs with EPO suggests that the extracellular domain orientation is tightly coupled to the cytoplasmic signaling events and, hence, provides valuable new insights into the design of synthetic ligands for EPOR and other cytokine receptors.

Receptor oligomerization in the form of homodimers — for example, erythropoietin (EPO) receptor (EPOR), growth hormone receptor (GHR) or hetero-oligomers (for example, IL-2R) — is the key event in signaling for the cytokine receptor superfamily. For EPOR, ligand-induced homodimerization of either the extracellular1 or cytoplasmic2 domains is the principal event that leads to activation. However, the orientation of the receptor monomers in the activated state can vary depending on the ligand that mediates dimer formation. Previously, we have described the crystal structure of an EPO-mimetic peptide, EMP1, in a complex with the extracellular binding domain (EPO binding protein, EBP, residues 1–225 of the EPOR). The EBP–EMP1 complex revealed that EMP1 binds two EBP receptor molecules as a peptide dimer generating an almost perfect two-fold symmetric 2:2 assembly3. Each peptide has a close interaction with its peptide partner in the dimer, and with both receptor molecules. The EMP1 dimer formation is apparently the main interaction that holds the EBP receptor dimer together, since very little direct contact is made between the two EBP molecules in this assembly. EPO, like other cytokine hormones, is proposed to have a four-helix-bundle topology4, that would be expected to impose a non-symmetrical dimerization of EPOR. Hence, the symmetrical mode of dimerization observed in the EBP–EMP1 complex and the asymmetrical mode proposed for the EPO–EPOR complex, as was found in the hGH–GHR complex5, represent dimer configurations that allow for signaling3. However, the symmetrical mode of dimerization is not the most efficient one for signaling as judged by the lower potency of EMP1 versus EPO in cell proliferation assays and in peptide dimer studies6–9. Engineering and construction of covalent

Fig. 1 Receptor cross linking and dimerization analysis. Chemical cross linking analysis was used to study the ability of EPO-mimetic peptides to mediate dimerization of the ligand binding domain of the EPO receptor. Peptide-mediated dimerization was stabilized with the homobifunctional sulfhydryl-reactive cross-linking reagent DPDPB and the percent of covalent dimer formed was determined by size-exclusion chromatography. Data are shown for EMP1, EMP6, EMP8 and EMP33 that contain tyrosine, alanine, phenylalanine and 3,5 dibromotyrosine respectively at position 4 of the peptide sequence. The percent dimer population is plotted as the percentage of total protein at each peptide concentration after normalization by dividing the dimer peak area by the total peak area of the monomer and dimer populations.

Department of Molecular Biology and the Skaggs Institute for Chemical Biology, The Scripps Research Institute, 10550 North Torrey Pines Rd., La Jolla, California 92037, USA. 2R.W. Johnson Pharmaceutical Research Institute, Drug Discovery Research, 1000 Route 202, Box 300, Raritan, New Jersey 08869, USA. 3Gladstone Institute of Virology and Immunology, San Francisco, California 94141, USA. 4Department of Medicine, School of Medicine, UCSF, San Francisco, California 94143, USA. 5Robert Wood Johnson Medical School-UMDNJ, 675 Hoes Lane, Piscataway, New Jersey 08854, USA. 6Current address: Department of Biological Chemistry, Institute of Life Sciences, The Wolfson Centre for Applied Structural Biology, The Hebrew University of Jerusalem, Jerusalem 91904, Israel. 7Current address: Departement d’Ingenierie et d’Etude des Proteines, BAT 152 C.E.A./SACLAY, 91191 Gif sur Yvette Cedex, France. 1

Correspondence should be addressed to L.K.J. email: [email protected] and I.A.W. email: [email protected] nature structural biology • volume 5 number 11 • november 1998

993

1998 Nature America Inc.

• http://structbio.nature.com

articles

1998 Nature America Inc.

• http://structbio.nature.com

Table1 EMP1 mutants; binding, activity, and preliminary crystal data Peptide

Sequence1

IC502 [µM] TF-1 cells

IC502 [µM] EBP Beads

EPO ED503 [µM]

Dimer type4

SPG5

EMP16 EMP11 EMP1 EMP14 EMP17 EMP316 EMP20 EMP19 EMP8 EMP327 EMP288 EMP 7 EMP6 EMP339

GGTYSCHFGPLTWVCKPQ GGTYSCHFGPATWVCKPQGG GGTYSCHFGPL TWVCKPQGG GGTYSCHFGPLT F V CKPQGG TYSCHFGPLTWVCKPQGG GGTXSCHFGPLTWVCKPQGG YSCHFGPLTWVCK YSCHFGPL TW VCKP GGTFS CHFGPLTWVCKPQGG GGTXSCHFGPLTWVCKPQGG GGTXSCHFGPLTWVCKPQGG GGTTS CHFGPLTWVCKPQGG GGTASCHFGPLTWVCKPQGG GGTXSCHFGPLTWVCKPQGG

0.07 0.08 0.07 0.3 0.07 0.08 8 9 1.5 ND 13 26 58 15

8 5 5 30 40 40 70 45 60 70 >500 120 120 50

0.0115 0.1 0.1 0.1 0.3 1.0 8 >10 >10 >10 >10 >10 IA IA

A A A A A A A B

19 19 19 19 21 19 18 and 19 18

Conserved amino acids from a phage display selection are highlighted in bold; substitutions are underlined in the different peptides. IC50s were measured for binding to EPO receptor on TF-1 cells or EBP immobilized on beads 9; ND – not determined. EPO ED50 is the concentration of peptide which correspond to the 50% of maximal EPO stimulation; IA- inactive. 4Dimer type A is symmetric, B is asymmetric, - indicates no crystal structure. 5SPG refers to the crystallographic space group numbers. 6In EMP31 X is p-F-Phe. 7In EMP32 X is p-I-Phe. 8In EMP28 X is D-Tyr. 1 2 3

9

In EMP33 X is 3,5-dibromotyrosine.

dimers of EMP17,8 revealed that the covalent peptide dimer does bind much more tightly to cell surface EPO receptors; the 100fold increase in affinity towards the extracellular module of EPOR represents an IC50 of ~2 nM, which is comparable to EPO itself. However, despite the equivalent binding affinity for receptor, a large difference in cell signaling potential still remains between EMP1 dimers and EPO7,8. In order to determine the role of individual residues in the conserved motif of the EMP sequences, the biological activity of a set of peptide deletions and substitutions was tested9. Throughout this study, two hydrophobic amino acids, Tyr 4 and Trp 13, were found to be important for EPO-mimetic activity. The crystal structure of EBP–EMP1 revealed that these hydrophobic residues play a key role in both peptide–peptide and peptide–receptor interactions3. While investigating the role of Tyr 4, a set of peptide analogs at this position was constructed9. One of the Tyr 4 analogs, 3,5-dibromotyrosine, EMP33, was found to be inactive in eliciting cellular growth, yet retained binding activity towards EBP and EPOR. The EBP–EMP33 complex then became a candidate for structural studies in order to explain its lack of biological activity compared to agonist peptides, such as EMP1. Here, we present the crystal structure of the EBP–EMP33 complex along with biochemical and biological data that support formation of a dimeric, but inactive, complex of the peptide and EPOR. The three-dimensional structure has unexpectedly produced new insights into the importance of both proximity and orientation of the extracellular domains of the receptor in the initiation of signal transduction. These data have implications in the design and synthesis of both agonist mimetics or antagonist small molecules for cytokine and other cell surface receptors.

receptor is due to an affinity threshold of peptide binding, the relative affinities of several EMP peptides towards EPO receptor were evaluated in a TF-1 cell binding assay. In this assay, binding of EMPs to the receptor on the cell surface indicates their ability to compete for EPO binding to a receptor that can freely dimerize. EMP33 demonstrates an IC50 of ~15 µM which, although higher than EMP1 (0.07 µM), is within the range found for other peptides that activate the receptor (Table 1). For example, EMP20 (13-mer), with an IC50 of 8.0 µM, and EMP28 (D-tyr), with an IC50 of 13 µM on TF-1 cells, are both capable of stimulating cell proliferation, albeit weakly, through the EPO receptor9. Thus, EMP33 can bind to human EPO receptor on the cell surface and the lack of receptor activation is not simply due to a substantially lower affinity of this peptide for its receptor. We have previously shown that covalent dimerization of EMP1 and other peptides using reactive polyethyleneglycol (PEG) linkers results, as expected, in increased activity; indeed, one apparently inactive EMP1 variant, EMP24, has been converted into a weak agonist through covalent dimerization7. This conversion is still highly sequence-dependent since, for example, EMP6 (Tyr4Ala), which mutates the critical Tyr residue3,8, is not converted into a receptor agonist upon covalent dimerization7. A similar covalent PEG-mediated dimerization of EMP33 does not result in conversion to a receptor agonist, but maintains the inactive properties of the unmodified EMP33 (data not shown). The ability to convert an inactive peptide into an agonist is then not simply a difference in receptor affinity, since EMP33 and EMP24 bind monomeric EBP (Table 1) more weakly than EMP1 by factors of 10 and 18 respectively9. Thus, a weaker receptor binding peptide (EMP24) can be converted to an agonist by covalent dimerization, whereas EMP33 remains inactive in both monomeric and covalent dimer forms.

EPO-mimetic peptide binding assay Dimerization of EBP in solution by EMP peptides To address the question of whether the ability to activate the Previously, we demonstrated that EMP1 mediates the dimeriza994

nature structural biology • volume 5 number 11 • november 1998

1998 Nature America Inc.

• http://structbio.nature.com

articles

1998 Nature America Inc.

• http://structbio.nature.com

a

b

Fig. 2 Bioactivity of EMP1 and EMP33 a, Ligand-induced protein tyrosine phosphorylation. Tyrosine phosphorylation of intracellular receptor and signaling components in the presence of EMP1, EMP33 and EPO. Identification of specific tyrosine-phosphorylated proteins induced by EPO and EMP1 is indicated by an arrow. b, Ligand-induced cell proliferation. Cell proliferation was measured in the presence of EMP1 and EMP33 at the concentrations shown. Cells were pulsed with tritiated thymidine, harvested and counted to assess proliferation.

tion of the extracellular, ligand binding domain of EPO receptor (EBP) and that the dimerization can be captured in solution with the chemical cross-linking reagent DPDPB3,7,9. The crystal structure of the EBP–EMP1 complex revealed that DPDPB is almost exactly the right length (16 Å) to bridge the two equivalent Cys 181 residues (20.7 Å) in the D2 domain of the symmetric EBP dimer3. However, DPDPB has only minimal ability to mediate receptor cross-linking in EBP–EPO complexes (unpublished data), which suggests that this particular technique is sensitive to the relative position, orientation, environment or steric hindrance of the reactive sulfhydryls of the receptor in a particular dimeric assembly. Here, we quantify the dimerization yield with size-exclusion chromatography to assess the ability of any given EMP sequence to promote EBP dimerization (Fig. 1). In the peptide sequences shown, Tyr 4 of EMP1 was replaced by alanine (EMP6), phenylalanine (EMP8) and 3,5-dibromotyrosine (EMP33), while EMP20 is a shorter 13 amino-acid agonist, but still retains the Tyr 4 equivalent (Table 1). Three of these peptides (EMP1, EMP8 and EMP20) activate receptor9 and display DPDPB-detectable EBP dimerization. The percentage of dimer formed, at equivalent peptide concentrations, is similar despite a 12-fold or 14-fold difference in the ability of EMP8 or EMP20 respectively, to compete for [125I]-EPO binding to EBP immobilized on agarose beads9, an assay that reflects binding to monomeric EBP (Table 1). Conversely, in competition for radiolabel ligand binding to cell surface receptors, EMP8 and EMP20 display IC50 values that are 3.3- and 114-fold less than EMP1 respectively, in a system which has uncompromised potential for receptor dimerization9. Collectively, these data suggest that the DPDPB-detectable dimerization ability of a given peptide is relnature structural biology • volume 5 number 11 • november 1998

atively insensitive to peptide affinity in the µM range and that the quantitative ability of a peptide to compete for ligand binding is dependent upon the ability of the assay system to permit receptor dimerization by the ligand (EPO) for which the peptides must compete. Two other peptides, EMP6 and EMP33, differ markedly from each other, as well as from the above peptides, in their ability to promote EBP dimer formation. EMP6 appears incapable of promoting receptor dimerization (Fig. 1) and is not biologically active9. In terms of competitive binding ability, EMP6 is 828-fold less effective than EMP1 on the cell-associated receptor assay but retains EBP monomer binding ability with a 24-fold lower activity than EMP1 (Table 1). EMP33-mediated levels of EBP dimerization is approximately half that obtained with EMP1, EMP8 or EMP20 at 40 µM. Surprisingly, EMP33 was found to be inactive in EPOR activation assays despite demonstrating an IC50 value only 10-fold less than EMP1 on EBP beads9. The EMP33 IC50 value of 15 µM on cell-associated receptors was 214-fold less than EMP1 (0.07 µM), but is only 1.8-fold less than for EMP20 (8 µM), which retains full dimerization ability and receptor agonist activity9. These combined binding and dimerization data suggest that EMP33 can dimerize EBP in solution but in a mode less favorable to formation of a covalent dimer when probed with the DPDPB cross-linker. Small perturbations in the relative orientation, but not distance, of the equivalent Cys 181 residues (Sγ–Sγ distances are 20.7 Å and 21.0 Å for the EMP1 and EMP33 complexes respectively), appear to be correlated with a reduction in detection of a covalent dimer. We have also tested the formation of peptide dimers in solution by analytical ultracentrifugation 995

1998 Nature America Inc.

• http://structbio.nature.com

articles

1998 Nature America Inc.

• http://structbio.nature.com

a

b

Fig. 3 Activation of signaling through native and chimeric EPOR. a, Selective activation of signaling through the EPOR. The 32D/EPOR cell line was used in EMSA with the FcγRI STATresponse element probe to detect activation of the JAK-STAT pathway by interleukin-3 (IL-3), EPO or the indicated peptides. The induced bands (arrow) are specific and represent dimers of Stat5A and Stat5B . b, Selective activation of signaling through heterodimeric and homodimeric receptor chimeras. Transfectants expressing chimeras that represent heterodimeric (IL-2R, IL-7R or IL-9R) or homodimeric (IL-4R) receptor configurations were stimulated with EPO or the indicated peptides, and nuclear extracts were analyzed by EMSA with the FcγRI probe. Supershift analyses demonstrated that the retarded nucleoprotein complexes induced via these receptors represent Stat5A/B (bold arrows for IL-2R, IL7R, IL-9R), Stat6 (bold arrow for IL-4R) and Stat3 and Stat1 (data not shown).

and find that the EMP33 and EMP1 form homodimers with similar dissociation constants (~20 µM) (H. Lashuel, D.L.J., L.K.J., I.A.W. and J. Kelly, in preparation). Thus, differences in the ability of the peptide to dimerize in solution do not contribute to the observed differences in dimerization of the receptor. Thus, these combined data suggest that the inactive EMP33 peptide–receptor dimer assembly differs from that of the active EBP–EMP1 complex. Cell line and phosphotyrosine analysis Tyrosine phosphorylation of intracellular proteins is a rapid, ligand-induced signaling event exhibited by cytokine receptors, including EPOR. EPO binding to its receptor induces tyrosine phosphorylation of several proteins including EPOR, JAK2, Shc, and Stat5 (Fig. 2a). This response is a transient event observed within seconds of ligand stimulation (data not shown). In earlier studies, we showed that EPO-mimetic peptides, specifically EMP1 at low µM concentration, elicit the same temporal intracellular phosphorylation response as EPO6. For this reason, phosphotyrosine analysis can serve as a readout of bioactivity, independent of stability of the peptide, which may complicate analysis of tissue culture experiments. In this assay, EMP33 was unable to elicit a signaling response at concentrations from 10 µM (Fig. 2a) up to 1 mM (data not shown). These results correlate with the inability of EMP33 to stimulate proliferation of 996

FDC-P1/HER cells at concentrations up to 500 µM (Fig. 2b). Furthermore, EMP33 did not stimulate proliferation of the EPO hypersensitive cell line, FDC-P1/trER, that expresses a truncated form of human EPOR with a 40-residue deletion at the C-terminus9. Activation of JAK-STAT pathway by chimeric receptors Activation of the JAK-STAT signal transduction pathway by EPOR is a rapid consequence of productive receptor–ligand engagement, and, therefore, serves as a reliable indicator for comparison of agonist function among EPO-mimetic peptides. The electrophoretic mobility shift assay (EMSA), using oligonucleotides containing canonical STAT-binding motifs, represents a sensitive method for detecting induction of DNA-binding activity by Stat5, which is the predominant STAT factor coupled to EPOR10. Both recombinant EPO (5 U ml–1) and EMP1 (20 µM) promote robust and rapid induction of nuclear Stat5 activity in transfected 32D cells expressing the murine EPOR (Fig. 3a). Both EMP1 and EMP33 bind to the cell-associated murine EPOR with IC50 values of 0.1 µM and 61 µM, respectively (data not shown). Interestingly, EMP33 (20 µM) elicited no detectable STAT activity in these cells despite its ability to promote dimer formation. Although inhibitory activity was not apparent when cells were exposed to both EPO and EMP33 simultaneously (Fig. 3a), the absence of agonist function reveals that EPOR subnature structural biology • volume 5 number 11 • november 1998

1998 Nature America Inc.

• http://structbio.nature.com

articles b

c

1998 Nature America Inc.

• http://structbio.nature.com

Stat1α

a

Fig. 4 Identification of EMP33 as an antagonist of the EPOR. a, Kinetics of Stat1α activation in EAR-13 cells. EAR-13 cells were treated for 15 min at 37 oC with 1 µM of EMP1 or EMP33. NP-40 whole cell extracts were made and the electrophoretic mobility shift assay performed. The intensity of the Stat1α signal was quantified with a phosphoimager. Background values with untreated lysates were subtracted from all the measurements. Solid circles represent Stat1α activation as a function of time with EMP1; open circles represent Stat1α activation as a function of time with EMP33. The ordinate represents Stat1α activation in arbitrary phosphoimager density units. b, Competition of EMP33 with EMP1. EAR-13 cells were treated with various levels of EMP33 mixed with EMP1 at 200 nM for 15 mins at 37 °C. The intensity of the Stat1α signal after EMSA was determined with a phosphorimager as described in the text. Signals from cells treated with only EMP1 represented the average of three independent samples determined in two separate EMSA assays; so the value in the absence of competitor is the mean of six individual assays. Solid circles represent Stat1α activation as a function of EMP33 concentration (µM). The ordinate represents Stat1α activation in arbitrary phosphorimager density units. c, Competition of EMP33 with EPO. EAR-13 cells were treated with various levels of EMP33 mixed with EPO at 490 pM for 15 min at 37 °C. The intensity of the Stat1α signal after EMSA was determined with a phosphorimager as described in the text. Signals from cells treated with only EMP1 represented the average of three independent samples determined in two separate EMSA assays so the value in the absence of competitor is the mean of six individual assays. Solid circles represent Stat1α activation as a function of EMP33 concentration (µM). The ordinate represents Stat1α activation in arbitrary phosphorimager density units.

unit dimerization per se is not sufficient to trigger the transmembrane signaling process. To investigate further the relationship between receptor dimerization and signal transduction, these ligands were tested against a panel of chimeric receptors in which the ligand-binding extracellular domain of the murine EPOR had been fused to the transmembrane and cytoplasmic portions of various interleukin (IL-) receptors. These EPO-responsive chimeric receptors recapitulate the signaling programs of receptors that require heterodimeric or hetero-oligomeric configurations (the IL-2, IL-7 or IL-9 receptors) or that can function in a homodimeric arrangement (the IL-4 receptor), and as such represent novel probes of dimer-dependent signal transduction11–13 (Goldsmith, M.A. unpublished data). For the mandatory heterodimers, stable cell lines were prepared that express a fusion protein containing the intracellular domain of the shared γc chain (EPORγ) in conjunction with fusion proteins representing the IL-2Rβ (EPORβ), IL-7Rα (EPOR7) or IL-9Rα (EPOR9) subunits respectively; the IL-4R was represented by transient expression of a fusion protein of the IL-4Rα (EPOR4) chain. In every case, EPO and EMP1 readily induced the nuclear DNA-binding activities of specific STAT factors normally linked to the cognate receptors (Stat5 by IL-2R, IL-7R; Stat1, Stat3 and Stat5 by IL-9R; Stat6 by IL-4R) (Fig. 3b). As observed with the native EPOR, EMP33 failed to elicit a detectable STAT response with any of the chimeric receptors. One explanation for the lack of activation of the chimeric receptors by EMP33 is that peptide is somehow unable to bind the EPOR extracellular domains of the chimera or that the peptide forms a dimer complex that is inactive for cell signaling. To investigate this further, the ability of EMP33 to compete for nature structural biology • volume 5 number 11 • november 1998

EMP1- and EPO-induced STAT activation was evaluated with an EPOR-γ interferon receptor chimera, EPOR/γR2(p91) described previously14. Stat1α activation as a function of time was determined for EMP1, EMP33 and EPO. Stat1α activation appeared linear up to 15 minutes with 1 µM EMP1, but no activation of Stat1α was seen with EMP33 (Fig. 4a); the increase in Stat1α activation for EPO (490 pM) continued for 30 min (data not shown). Therefore, to evaluate competition of EMP33 with EMP1 and EPO, we chose the 15 min time point, as the curve was approximately linear with respect to EMP1. Furthermore, to determine which concentration of EMP1 and EPO to use for the competition experiments, we determined Stat1α activation as a function of EMP1 and EPO concentration. From these doseresponse curves (data not shown), we chose 200 nM of EMP1 and 490 pM of EPO for the competition experiments (Fig. 4b,c). EMP33 inhibited Stat1α activation by EMP1 with 50% inhibition at ~150 µM EMP33, which is ~750 times the concentration of EMP1. EMP33 also inhibited Stat1α activation by EPO, with ~150 µM EMP33 representing 50% inhibition, at ~300,000 times the concentration of EPO. In conjunction with the earlier findings, these observations with a range of chimeric receptors demonstrate that ligand-dependent dimerization of receptor subunits is not sufficient to initiate transmembrane signaling. Taken together, these results demonstrate that EMP33 binds to the EPOR chimeric receptors and can act as a modest antagonist of both EPO and EMP1-dependent STAT activation. EMP33 antagonist activity is not detectable in cell proliferation studies that are performed over the course of several days (data not shown). The STAT activation studies examine the pre-equilibrium events of ligand-induced receptor activation and are not kinetically limited by the apparent dilution of EMP33 antagonist 997

1998 Nature America Inc.

• http://structbio.nature.com

articles

1998 Nature America Inc.

• http://structbio.nature.com

a

b

F > 1σF in the resolution range of 50–2.7 Å with relatively good stereochemistry (Table 2). No non-glycine residues lie in the disallowed region of the Ramachandran plot, as analyzed by PROCHECK15, and 79% are in the most favorable region. Residues 1–9, 164–166, 225 for EBP1 and 1–9, 133–134, 166–167, 121–125 for EBP2 and the N- and C-terminal glycine spacers, residues 1–2, 19–20, from the two EMP33 peptides did not have any clear electron density and are not included in the structure. These residues are similarly disordered in the EBP–EMP1 complex3. The individual EBP domains of the two molecules in the dimer are very similar in tertiary structure but have a small difference in their D1–D2 intra-domain arrangement. After superimposing equivalent D2 domains, a slight rotation along the long axis of the β-structure of the D1 domain is required to overlap the equivalent D1 domains. An overlap of the Cα atoms of the individual β-structure cores for D1 and D2 of the dimer gives r.m.s.d.’s of 0.54 Å (94 pairs, D1/D1) and 0.47 Å (92 pairs D2/D2 ) respectively, for the Cα atoms.

EBP–EMP33 structure The structure of the EBP monomer in the EBP–EMP33 inactive complex (Fig. 5a,b) is very similar to that in the EBP–EMP1 active complex3. The EBP consists of two fibronectin 16 Fig. 5 Superposition of an active and inactive EPOR assembly. The EBP–EMP1 (cyan), and (FBN)-III folds , with the N-terminal (D1) and EBP–EMP33 (red) crystal structures are superimposed only on their D2 domains of one C-terminal (D2) domains being connected at receptor monomer. The side chains of Trp 4 (EMP1) and DBY 4 (EMP33) from both pepright angles by a short helical linker. The EMP33 tides are shown. a, Side view of the two complexes show a difference in the dimer configuration between the two complexes. The second receptor monomer (EBP2) and two peptide is a hairpin structure consisting of two peptides in the EMP33 complex are rotated by 15º relative to the EMP1 complex after short antiparallel β-strands connected by a overlapping the first monomers (EBP1) of both complexes. b, The bottom view is rotated disulfide bridge at one end and a slightly distortby 90º along the horizontal axis. The α-helix at the N-terminus of the D1 domain, and the ed type-1 β-turn for residues Gly P9–Pro entire D2 domain are removed for clarity. P10–Leu P11–Thr P12 at the other. The β-turn segment contains one of the conserved motifs of this EMP peptide family (GPXTW), with activity over time. The off-rate of EMP33 can be assumed to be approximate type 1 main-chain torsional angles for Pro P10 very rapid compared to that of EPO which effectively eliminates (i+1) of φ = -62º; -84º; ψ = -52º; -51º and for Leu P11 (i+2) of the antagonist behavior of the peptide under equilibrium condi- φ = -37º; -82º; ψ = -71º; -59º, for the two EMP33 peptides of tions achieved in cell culture. Alternatively, this inability to the dimer respectively. The conformation of the two EMP33 detect inhibitory activity could also be due to peptide instability monomers is similar between the disulfide bridge but deviates under the conditions of cell culture. In either case, the antagonist slightly at the ends (Fig. 6). In particular, the 3,5 dibromotyrobehavior of EMP33 can be clearly demonstrated at the early sine P4 (DBY P4) conformation differs between the two pepphases of receptor signaling. tides so that the equivalent Cα–Cα distance is 1.2 Å with an altered side chain rotamer (Oζ–Oζ distance of 3.5 Å). Crystal structure determination Comparison of the EMP33 and EMP1 structures also shows In order to determine the structural correlates of agonist ver- high similarity for both main chain and side chain conformasus antagonist peptides, co-crystals of the complex between tions within the disulfide bridge but with some deviation at EBP and EMP33 were grown and X-ray data collected to 2.7 Å their ends. The r.m.s.d. between the two EMP1 and the two resolution (Table 2). The structure was determined by molecu- EMP33 peptides of the dimer in their corresponding complexlar replacement (MR) using only one EBP monomer (EBP1) es is 0.51 Å and 0.57 Å for their Cα atoms respectively. from the refined EBP–EMP1 complex3 as a starting model The quaternary structure of the EBP–EMP33 structure con(PDB code 1ebp). An orientation and translation solution for sists of a 2:2 complex, where two EMP33 peptides (peptide 1, two EBP molecules (EBP1, EBP2) in the asymmetric unit was peptide 2) bind to two receptor molecules (EBP1, EBP2) in determined from the MR search. Two EMP33 peptide order to form the dimeric assembly. The most striking obsermonomers were independently interpreted from Fo - Fc differ- vation is that the two receptor molecules of the non-active ence maps after rigid body refinement of the two EBP EBP-EMP33 dimer are related by an approximate 165º rotamonomers. The current R-value is 20.0% and Rfree is 29.9% for tion, in contrast to the almost perfect two-fold axis (180º) of 998

nature structural biology • volume 5 number 11 • november 1998

1998 Nature America Inc.

• http://structbio.nature.com

articles

1998 Nature America Inc.

• http://structbio.nature.com

Fig. 6 Stereo representation of the superposition between the two EMP33 peptides in EBP–EMP33 complex structure. Several of the side chains are labeled for reference.

the EBP–EMP1 dimer (Fig. 5a,b). We have now crystallized several other active EBP–peptide complexes (data not shown) and all have a two-fold symmetric dimer configuration, irrespective of the crystallization conditions, crystallographic space group or cell constants (Table 1). As in the EBP–EMP1 complex, the two EMP33 peptides have a close association, with average buried surface areas of 350 Å2 per EMP33 peptide, which corresponds to ~25% of each peptide’s molecular surface (1,260 Å2 and 1,235 Å2 for peptides 1 and 2 respectively). Specific interactions between the peptides of the dimer consist of four interchain hydrogen bonds between the main chains of the peptide β-strands and a large hydrophobic interface which contains the disulfide bridge and the side chains of DBY P4, Phe P8, and Trp P13. This hydrophobic core also provides the framework for hydrophobic interactions between the EMP33 peptide dimer and the EBP (Fig. 7). The peptide dimer serves as the mediator of receptor dimerization with the combined contact surfaces of 863 Å2 for the two EBP’s and 843 Å2 for the two EMP33 molecules. The two EBP molecules have only a slight increase in self-association (100 Å2) compared to the EBP–EMP1 complex (75 Å2 from Leu 175 and Arg 178 of each receptor); the EBP–EMP33 interface also contains additional hydrophobic residues that reflect the non-symmetrical mode of dimerization. The buried interface residues are Ser 152, His 153, Leu 175, Glu 176, Arg 178 for EBP1 and His 153, Leu 175, Glu 176, Arg 178, Pro 203 for EBP2. Implications for signal transduction We have shown that the antagonist EMP33 peptide dimerizes in solution with a dissociation constant (~20 µM) that is similar to that for the agonist peptide EMP1 (Lashuel, H. et al., in preparation). EMP33 can dimerize the EPO receptor in solution at a peptide concentration of 40 µM (Fig. 1), which is in the range of the binding IC50 of 15 µM. No receptor-specific tyrosine phosphorylation activity is observed for 10 µM (Fig. 2a), 50 µM and 1 mM EMP33 peptide (data not shown), and no cell proliferation is observed for EMP33 over the range from 10–10 to 10–3 M (Fig. 2b). Covalent dimerization of EMP33 with PEG also does not lead to agonist activity, unlike other weaker binding pepnature structural biology • volume 5 number 11 • november 1998

tides, such as EMP24. Hence, we conclude that the mode of dimerization must be different between EPO, EMP1 and EMP33 complex, as clearly indicated from the substantial differences in biological potency of the natural versus synthetic ligands, even when their receptor-binding affinities are equivalent7,8. EMP1 provides a qualitative mimicry of EPO through dimerization and activation of EPOR (EPO ED50 = 0.1 µM), but the signal is not nearly of the same magnitude as the natural hormone (EC50 = 20 pM for EPO)6–8. When two EMP1 peptides are covalently dimerized to provide an entropic advantage of reduced molecularity, the IC50 of 10 nM for the covalent EMP1dimer approaches the IC50 of 2 nM for EPO8. However, the covalent peptide dimer does not attain the same bio-activity in terms of cell proliferation (EPO ED50 = 10 µM) compared to EPO (EC50 = 20–40 pM)6–8. Hence, it appears that the mode of dimerization, not absolute affinity, governs the efficiency of activation. The major structural difference in the EBP–EMP33 inactive (antagonist) complex versus the EBP–EMP1 active (agonist) complex is that the mode of receptor dimerization differs by a relative rotation of ~15° from the almost perfect two-fold dimerization seen in the EBP–EMP1 complex3. Indeed, formation of a receptor dimer with the EMP33 peptide was unexpected, as EMP33 did not trigger signaling by the EPO receptor, unlike EMP1. The overall structures of the EBP monomers in the EBP–EMP1 and EBP–EMP33 complexes are very similar to each other. Therefore, large conformational differences in their tertiary structures cannot account for the vast difference in biological activity of these related peptides. It would then appear that the difference in dimer configuration accounts for the non-proliferative properties of the EBP–EMP33 complex. We previously proposed3 that more than one mode of productive dimerization can lead to receptor activation. However, this new inactive dimer complex allows us to go further and suggests that, even if the two receptors are brought into equivalent close proximity with one another, differences in their relative orientations can lead to marked differences in their signaling outcomes. This result is consistent with the substantial differences in biological potency for EPO and EMP13,6,8 that presumably are correlated with even larger differences in the relative orientation and translation of the extracellular domains of the receptor molecules that consti999

1998 Nature America Inc.

• http://structbio.nature.com

1998 Nature America Inc.

• http://structbio.nature.com

articles Fig. 7 The hydrophobic interaction surface between EBP and EMP peptides. a, Comparison between the receptor–peptide interactions in the EBP–EMP1 (left) and EBP–EMP33 (right) complexes. In the EBP–EMP1 complex, the receptor and peptides are colored in blue and cyan and for EBP–EMP33 in magenta and gold. The Tyr P'4 and DBY P'4 are contributed from the other peptide partner and, in addition, to their essential role in the hydrophobic interaction, form the only side chain specific hydrogen-bond (dotted line) between the peptide and receptor. The overall architecture of the main hydrophobic interaction patch, consisting of residues Phe 93, Met 150, and Phe 205 from the EBP and Phe P8, Trp P13, and Tyr/DBP P'4 from the peptides remain similar in the two EMP1 and EMP33 structures with the largest variation being in the actual conformation of the DBY compared to Tyr 4. b, Stereoview of a 3Fo - 2Fc electron density map calculated at 50–2.7 Å resolution, contoured at 1.0σ (blue) and 4.5σ (red) with superimposed coordinates from the refined structure. The orientation for the DBY residue is easily defined by the positions of the two bromine atoms on the tyrosine ring, as observed from the high 4.5σ level contour in the electron density map (red). At that cutoff level, the sulfur from Met 150 is also observed. Phe P8 has been omitted for clarity.

a

b

tute their respective dimers. It is still unclear how flexible the connecting regions are between the extracellular globular D1, D2 domains and the intracellular module. However, the EMP1 and EMP33 data suggest that there is insufficient flexibility to permit optimal signaling as deduced from the increased potency of the EPO dimer complex. In eight crystal structures of EBP–peptide complexes that we have now analyzed, the orientation of the two EBP receptors appear to be induced by and restricted by interaction with a peptide dimer, since very little direct interaction occurs between the receptor molecules. The two-fold dimerization for the active EBP–EMP1 complex must then be only one of several possible EPOR dimer configurations that can result in cell proliferation3. EPO, the natural ligand for EPOR, is predicted to have the same non-symmetrical four-helix bundle topology as other related cytokine hormones, such as hGH and prolactin, and most likely induces a substantially non-symmetric dimerization of the two EPOR molecules. This proposed mode of assembly is based on the structure of the hGH–hGHBP binding protein complex5, where growth hormone induces a non-symmetrical dimerization of the soluble receptor molecules that includes substantial deviations in both the rotational and transitional components from a symmetric two-fold relationship. 1000

In EPOR, as in other single-pass transmembrane cytokine receptors, the ligand appears to induce specific receptor dimer/oligomerization in order to allow the cytoplasmic domains to position and orient themselves so as to become substrates for phosphorylation by two molecules of JAK-2. These events initiate the cascade that leads to intracellular signaling and cell proliferation. For EPOR, ligand-induced homodimerization is the key event that governs activation1,17,18. Constitutive activation of EPOR by covalent crosslinking or dimerization of the extracellular domain in the absence of EPO, or any other ligand, can also lead to signaling1,19. However, EPOR can be dimerized by a limited number of point mutations, R129C, E132C, E133C (murine) and R130C (human)1 that allow formation of an intermolecular disulfide bridge. These mutational data support the proposal that homodimerization of EPOR can lead to cell activation but suggest that only certain receptor residues, when cross-linked, permit formation of a complex that is competent for signal transduction. These Cys point mutations occur in a loop region between strands A and B in the bottom part of the C-terminal domain. In the crystal structures of the human EBP complexes with EMP1, EMP33 and other EMPs (data not shown), this loop region is extremely flexible and, in several cases, part of this segment cannot be modeled due to lack of nature structural biology • volume 5 number 11 • november 1998

1998 Nature America Inc.

• http://structbio.nature.com

articles

1998 Nature America Inc.

• http://structbio.nature.com

Table 2 Data collection and refinement statistics for EBP–EMP33 complex Resolution range Unique reflections Redundancy Rsym (I) Completeness I/σ R-factor (F > 1σF) Rfree2 R.m.s.d. from ideality Bond length Bond angle Dihedrals Improper dihedrals Ramachandran Plot (PROCHECK) Favored Allowed Generously allowed Disallowed 1 2

50–2.7 Å 16,100 3.3 8.9% [30%]1 99.0% [99.0%]1 19.2 [4.1]1 20.0 29.9 0.008 Å 1.55º 26.2º 1.3º 78.7% 20.7% 0.6% 0.0%

Statistics for outer shell of 2.8–2.7 Å data. For 10% of data not used in the refinement (1,498 reflections).

interpretable electron density. This loop region in the murine EPOR is the same size as in the human receptor and we assume has similar flexibility. A disulfide bridge connecting these equivalent loop regions can then keep two receptor molecules in close proximity, but the inherent flexibility of the loop could also permit the receptors to accommodate themselves in an orientation that would result in a productive dimer configuration of the extracellular domains. Another example of receptor dimerization in the absence of EPO is the activation of EPOR with bivalent monoclonal antibodies that were raised against the extracellular domain of EPOR19,20. In one study, 96 antibodies were found to bind EPOR, but only four were capable of activating the receptor19. Thus, only ~4% of antibodies were capable of activating the EPOR; the other 96% bind EPOR and probably several, if not all, can induce receptor dimerization, but not activation. Bivalent attachment was essential, as Fabs from these murine IgG’s bound but did not activate the receptor. The IgG molecule has substantial flexibility about the hinge region21,22 which can allow for such adjustments in order to bind and dimerize EPOR. Thus, these data provide additional support for a limited range of EPOR dimeric arrangements that result in constitutive signaling, but many that do not permit activation. Studies with chimeric receptors have shown that dimerization of the cytoplasmic domain of EPOR is also an essential condition for signaling. After substitution of the extracellular domains of EPOR by epidermal growth factor receptor (EGFR), stem cell factor receptor (c-kit)23,24 and IFN-γR214, phosphorylation of intracellular substrates can be retained when the respective cytokine hormones are present. In one of these studies, the orientation of the extracellular domain was proposed to have a direct effect on the arrangement of the receptor cytoplasmic domains14. In order to determine whether inactive (antagonist) and active (agonist) peptides induce different receptor assemblies, we have determined crystal structures of seven agonist–EPOR complexes and one antagonist complex. The symmetric receptor dimerization is common for all active peptides, irrespective of the crystalnature structural biology • volume 5 number 11 • november 1998

lization conditions, space group and cell dimensions (Table 1). The EBP–EMP33 antagonist complex is the only one that forms a significantly different, non-symmetric assembly that correlates with the observed differences in biological properties. The lack of EMP33 bioactivity cannot be attributed solely to a substantially lower affinity towards EPOR compared to other active peptides. EMP32, EMP8 and EMP20 have lower competitive binding (IC50 = 70, 60 and 70 µM respectively, for monomeric EBP) towards EBP than EMP33, yet still activate the receptor (Table 1). The crystal structures of EBP complexed with EMP32 and EMP20 peptides (data not shown) show the same symmetrical two-fold dimerization as in the EBP–EMP1 complex3. EBP dimerization is also observed in solution (Fig. 1). The formation of different dimer assemblies for active versus non-active peptides is also inferred from the use of DPDPB as a cross-linking reagent. EMP8 has a similar IC50 to EMP33 (Table 1) but shows a much higher dimerization efficiency than the EMP33 complex for this particular cross-linker (Fig. 1). The EBP–EMP1 and EBP–EMP8 complexes are symmetrical (Table 1), providing the correct distance and angle for the crosslinker to bridge the free Cys 181 residues of each receptor. In the case of the EBP–EMP33 complex, we do not observe a substantial change in the Sγ–Sγ distance (~21 Å), compared to the EBP–EMP1, but the relative orientation of the receptor subunits has changed from the symmetrical dimer in the EBP–EMP1 complex. Hence, it appears that this particular cross-linker can be a sensitive indicator of subtle differences in the orientation of the receptor molecules in active and non-active peptide-EBP assemblies. The cross-linking data cannot be explained by differences in the ability of the peptide to form dimers, as EMP1 and EMP33 dimerize equivalently in solution (Lashuel, H. et al., in preparation). Therefore, the crystal and solution data concur that different dimerization modes exist for active (agonist) versus inactive (antagonist) complexes. As this particular cross-linker (DPDPB) does not capture significant amounts of covalent receptor dimer in the presence of the natural ligand EPO, the EBP–EPO dimerization mode must also be substantially different from the symmetrical EBP–EMP1 complex, substantiating our previous hypothesis that there is more than one mode of productive receptor dimerization3. An additional illustration of the role of receptor dimer configuration on signaling is emphasized by studies on insulin receptor (IR)25,26. IR is a member of the receptor tyrosine kinase family and functions as a heterotetrameric association of two α and two β subunits (α2β2) which are linked into a covalent assembly through disulfide bridges. Mild reduction loosens the assembly into a conformation that results in non-proliferative dimerization, although insulin is still bound to the α chains25. Finally, the crystal structure and solution studies, the bifunctional cross-linking data, and the receptor chimera data demonstrate that the antagonist EMP33 can still dimerize the EPOR, albeit in an altered mode. Thus, homodimerization, although a requirement for EPOR activation is, in and of itself, not a sufficient condition for activation of the receptor. It appears then that the mode of extracellular dimerization can have a direct influence on the ability of the cytoplasmic domains to assemble into a functional complex. Perhaps some types of extracellular dimerizations of EPOR, as demonstrated by the EBP–EMP33 structure, can impose an assembly that does not allow sufficient flexibility for forming a productive dimer configuration. The resultant non-productive complex of EPOR with EMP33 appears not to provide a correct docking environment for two molecules of JAK2, hence inhibiting the first event of the signal1001

1998 Nature America Inc.

• http://structbio.nature.com

articles

1998 Nature America Inc.

• http://structbio.nature.com

ing process. Another possibility would be formation of an extracellular complex that does not permit optimal docking with a yet unidentified accessory molecule27, that could play a role in the signaling complex. Thus, the mode of extracellular receptor dimerization appears to play a key role in receptor activation, a factor that has not been clear until now, and has significant implications in our understanding of how dimerization or oligomerization leads to productive signaling. Clearly, both proximity and orientation of the receptor molecules (that is, effective molarity) appear to be a requirement for forming productive complexes. Indeed, the biological potency of active ligands appears to be highly correlated with the configuration of the ligand–receptor complex3,6–8. Such findings may be important in many other systems where the design of agonists or antagonists can be used to control receptor activation and biological processes. Note added in proof: The crystal structure of the EPO–EPObp (~EBP) complex44 and the NMR structure of erythropoietin45 have appeared this past month. The EPO–EPObp structure reveals a substantially asymmetric assembly (~120°) of the receptors (~120°) in the 2:1 complex that adds further support to the notion that receptor orientation is correlated with the level of activation. The free Cys 181 residues that are cross-linked by the DPDPB (Fig. 1) are ~26 Å apart in the EPO–EPObp structure (R.M. Stroud, pers. comm.) compared to ~22 Å in EMP1 and EMP33, and confirm that this cross-linker is exquisitely sensitive to both distance and orientation of the receptors in order to capture a covalent dimer. As the DPDPB only cross links a very small percentage of the high affinity EPO–EPObp complex dimers (data not shown), there must be relatively little flexibility in the homodimeric assembly that would allow for the adjustment required to form a covalent dimer. The human EPO structure is confirmed to be a left-handed fourhelix bundle from these studies44,45, similar to other members of the hematopoietic growth factor family4. The complex structure44 shows interaction of five to six binding loops (L1–6) of the EPObp with EPO (depending on site 1 or site 2) with buried surfaces of 920 Å2(site 1) and 660 Å2 (site 2) that are substantially larger than the peptide for the EBP–EMP1 complex3 (~420 Å2 for each site), as expected. Interestingly, the only contacts between receptors in the EPO–EPObp homodimer are between equivalent loops in the D2 domains that contain Arg 130, which when mutated to Cys allows formation of a constitutive dimer. The Arg 130 residues in this structure are 27 Å apart and would require a substantial rearrangement for a covalent dimer to be formed44. Thus, the crystal structure of EPO–EPObp complex structure44 provides further evidence that receptor orientation relates to the effectiveness of the intracellular signal that is generated. Methods EPO-mimetic peptide binding assay. TF-1 cells were maintained in RPMI 1640, 10% fetal calf serum, 1% glutamine, 1% penicillin, 0.1% streptomycin, and 1 ng ml–1 of GM-CSF. [125I]-EPO was obtained from NEN Research Products. Cells were centrifuged and washed with binding buffer (RPMI 1640, 5% BSA, 10 mM Hepes, pH 7.0, 0.02% sodium azide) and counted using trypan blue as an indicator of viability. Binding assays were essentially as described7,9. Briefly, binding reactions contained 5 × 105 cells, [125I]-EPO (0.5nM) in the absence or presence of peptide as competitor, in increasing concentrations in binding buffer in a final volume of 200 µl. The binding reactions, in duplicate, were incubated at 4 ºC overnight then centrifuged for 2 min at 12,000 r.p.m. and the supernatant removed. The cell pellets were resuspended in binding buffer and layered onto 0.7 ml of bovine calf serum. The tubes were centrifuged at 12,000 r.p.m. for 5 min. The supernatant was removed, the bottom

1002

of the tube snipped off and the pellets counted in a Micromedic ME plus gamma counter. Dimerization of EBP by EPO-mimetic peptides. The non-water soluble homobifunctional sulfhydryl-reactive cross-linking reagent DPDPB [(1,4-di-[3'-(2'-pyridyldithio) propionamido]butane, Pierce Chemical Co.] was used to detect peptide-mediated dimerization of EBP3,7,9. Briefly, EBP (11 µM) was incubated in the presence of DPDPB (0.5 µM, stock prepared in DMSO) and variable concentrations of test peptide (stock prepared in 0.1% trifluroacetic acid) in 75 µl of PBS, pH 7.5, with all reactions and controls run in a final concentration of 4.4% DMSO and 0.007% TFA. Samples were incubated for 4 h at room temperature and stored at 4 ºC for at least 12 h before analysis. The samples were analyzed by high performance size exclusion chromatography (HP-SEC) on a Waters 625 HPLC system equipped with a Waters 996 detector. Separations were performed at ambient temperature on a G-3000 SWXL (7.8 × 300 mm) column (Supelco, Bellfonte). The column was equilibrated in 10 µM sodium phosphate, pH 7.2, 150 µM NaCl, at a flow rate of 1 ml min–1 and was monitored at 220 nm. Under these conditions, EBP monomer was eluted at ~9.6 min, while the dimer product eluted earlier at ~8.9 min. These elution times correspond to an molecular mass of 25,000 Mr for EBP monomer and 52,000 Mr for the dimer product based on calibration of the column using a commercial HP-SEC standard mixture28. The percent of detectable dimer product was calculated by adding the integrator-determined peak area for the monomer and dimer protein peaks to determine total protein peak area. The dimer area value was divided by the total protein peak area and multiplied by 100 to yield the % dimer observed in each reaction mixture. The reported value for each concentration is the average of three different chromatographic separations and the error bars are the standard deviation value for the three experiments. Cell line and phosphotyrosine analysis. FDC-P1/HER cells were maintained in erythropoietin (EPO) at 1 U ml–1 as described29. For the phosphotyrosine analysis study, cells are grown to stationary phase, collected by centrifugation, washed and incubated in fresh medium in the absence of EPO for 18 h as described9. Cells were collected, washed, and suspended in RPMI 1640 in the absence of serum for an additional 4 h. Stock solutions of EMP1 and EMP33 were prepared as 10 mM solutions in water. EPO is prepared as a 10 U µl–1 solution in serum free medium. Briefly, cells were diluted to a concentration of 0.5 × 106 cells ml–1. One ml of cells were stimulated with EPO, EMP1 or EMP33 at 10 U ml–1 (EPO) or 10 µM (peptide) for 10 min at 37 °C. Cells were collected by centrifugation and resuspended in 50 µl of 2× electrophoresis sample buffer (250 µM Tris, pH 6.8, 2% SDS, 10% glycerol, 0.006% bromophenol blue and 2% β-mercaptoethanol). Samples were loaded onto an 8% SDSPAGE gel (Fig. 2a), transferred to nitrocellulose and blotted with antiphosphotyrosine antibody (4G10, Upstate Biotechnology) at a dilution of 1:1000. Reactive proteins were visualized with ECL (Amersham Life Sciences). Cell proliferation analysis. FDC-P1/HER cells were grown, washed and incubated in the absence of erythropoietin as described above for phosphotyrosine analysis. Cells were plated in a 96 well microtiter dish at 40,000 cells per well. Stock solutions of EMP1 and EMP33 were prepared as 10 mM solutions in water. Peptide was added to cell wells in triplicate at the concentrations as shown in Fig. 2b. Cells were grown in presence of peptide for 42 h at 37 °C at which time 1 µCi per well of [3H]-thymidine was added and the incubation continued for an additional 6 h. Cells were harvested and counted to assess [3H]-thymidine incorporation as a measure of cell proliferation. Chimeric receptor constructs, cell lines and transfection. Constructs encoding chimeric cytokine receptors which activate Stat5A were prepared by conventional molecular techniques, and expressed using the pCMV4Neo expression vector30. In each case the coding sequence representing the extracellular domain of the murine EPOR was fused at a unique NheI restriction site to

nature structural biology • volume 5 number 11 • november 1998

1998 Nature America Inc.

• http://structbio.nature.com

1998 Nature America Inc.

• http://structbio.nature.com

articles sequences encoding six extracellular linker amino acids and the transmembrane and cytoplasmic domains of the γc (EPORγ), IL-2Rβ (EPORβ), IL-4Rα (EPOR4), IL-7Rα (EPOR7) or IL-9Rα chains (EPOR9), as described11–13 (Goldsmith, M.A., unpublished data). pEPORNeo was prepared by subcloning the cDNA encoding the murine EPOR (kindly provided by G. Longmore, Washington University) into pCMV4Neo. The transfected cell lines HT-2EPOβ/EPORγ and HT-2EPOR7/EPORγ were prepared and maintained as described11–13. HT-2EPOR9/EPOγ was prepared with the EPO9 fusion protein by the same methods. 32D/EPOR cells were prepared by electroporation of the cytokinedependent pro-myeloid cell line 32D31 as described11,12 with pEPORNeo (see below) and growth selection in EPO without other cytokines. Transient transfection of COS7 cells (ATCC) employed Lipofectamine (GIBCO/BRL) per the manufacturer’s instructions. The 16-9 hamster x human somatic cell hybrid line is the Chinese hamster ovary cell (CHO-K1) hybrid containing a translocation of the long arm of human chromosome 6 including the gene encoding the ligand binding chain of the human interferon gamma receptor (IFN-γR1) and a transfected human HLA-B7 gene32. The EAR-13 cell line represents 16-9 cells expressing the chimeric receptor EPOR/γR2(p91) that consists of the extracellular domain of the erythropoietin receptor (EPOR) and the intracellular domain of the second chain of the interferon gamma receptor (IFN-γR2) and the Stat1α recruitment site of the ligand binding chain of the IFN-γ receptor (p91)14. EAR-13 cells were maintained in Ham’s F-12 medium (Life Technologies, Inc.) containing 10% heat-inactivated fetal bovine serum (Sigma). Electrophoretic mobility shift assay. Electrophoretic mobility shift assays (EMSA) for the EPOR chimeric receptors which activate Stat5A or Stat6 were performed using an oligonucleotide probe representing the FcγRI STAT-response sequence and nuclear extracts prepared from resting or stimulated cells, as described33. Stimulations were performed for 10–15 min at 37 oC with either EPO (5 U ml–1) or the indicated peptide (20 µM). For the chimeric receptors which activate Stat1α, EMSAs were performed with a 22 bp sequence containing a Stat1α-binding site corresponding to the gamma activation sequence (GAS element) in the promoter region of the human IRF-1 gene (5'-GATCGATTTCCCCGAAATCATG-3') as described14. Briefly, cells were grown to confluence in 12-well or 24-well plates. Medium was then removed by aspiration. For determination of Stat1α activation, prewarmed media containing the ligands (EPO; or EMP1) and/or competitor (EMP33) were then added. For the competition assay, incubation was performed for 15 min at 37 °C. After incubations, the wells were washed with 1.0 ml of phosphate-buffered saline per well before 100 µl of lysis buffer (10% glycerol, 50 µM Tris·HCl, pH 8.0, 0.5% NP-40, 150 µM NaCl, 0.1 µM EDTA, 1 µM DTT, 0.2 µM PMSF, and 1 µM Na3VO4, 3 mg ml–1 aprotinin, 1 mg ml–1 pepstatin, 1 mg ml–1 leupeptin) was added to each well. Cells were then harvested by scraping, left on ice for 30 min, then stored at -70 °C until used for the EMSA assay. For each assay, 2.5 µl of the whole cell lysate was used in the reaction prior to the EMSA14,34. After incubation at 24 °C for 20 min, 11 µl of reaction mixture was electrophoresed at 450 volts for 3 h at 4 °C on a 5% polyacrylamide (19:1, acrylamide:bisacrylamide) gel. The Stat1α band on the dried gel was quantified with a BIO-RAD GS-525 phosphorimager and the molecular analyst software. For the competition curves (Fig. 4), two separate gels were run, the phosphorimager densities converted to relative values with the density in the absence of competitor arbitrarily set at 100, then the means of the values for the two independent EMSA assays plotted (Fig. 4).

nature structural biology • volume 5 number 11 • november 1998

Crystallization and data collection. Human recombinant EBP was expressed in E. coli as described28. The EPO-mimetic peptides were synthesized and activity determined as described9. Crystals of the EBP–EMP33 complex were grown by the sitting drop vapor diffusion method in inverted glass cups35,36 in a constant temperature incubator at 22 °C. Small crystals were nucleated with a reservoir solution consisting of 12% monomethylether polyethylene glycol (MPEG) 5000, 100 µM sodium cacodylate, pH 6.5. The seeds were then transferred to other pre-equilibrated drops set-up at 12% MPEG 5,000, 0.2 M imidazole malate, pH 7.6. Crystals do not nucleate spontaneously under such conditions. The complex was prepared by mixing 10 mg ml–1 peptide with 7 mg ml–1 EBP in a volumetric ratio of 10:1. Drops of 2.5 µl of reservoir and complex solution were mixed and allowed to equilibrate for 24 h before seeding. The crystals belong to the orthorhombic space group P21212. A comparison of the cell constants with the previously published EBP–EMP1 complex, in space group P212121, shows the cell dimensions changed to a = 58.9 Å, b = 75.0 Å, c = 130.5 Å from a = 59.2 Å, b = 75.5 Å, c = 132.2 Å (permuted to P22121 for comparison). As for the EBP-EMP1, the complex consists of two EBP molecules and two EMP33 peptides in the asymmetric unit, with a VM = 2.8 Å3 Mr–1 (56% solvent)37. Data were collected from a single crystal on a 30 cm MAR image plate mounted on a Siemens generator operating at 50 kV and 80 mA, with a crystal-to-plate distance of 150 mm. A total of 83 frames were collected with an oscillation range of 1.5° per frame. Data were integrated, scaled and reduced (Table 2) with DENZO and SCALEPACK38,39. Structure determination and analysis. The EBP–EMP33 complex structure was determined by molecular replacement (NMR) methods using AMoRe40, as implemented in the CCP4 package, using the previously determined structure3 of unliganded EBP monomer 1 (Brookhaven PDB code 1ebp) as the search model. The MR solution revealed an EBP dimer in the asymmetric unit. The structure was initially refined without peptide in X-PLOR41 using rigid body refinement in the resolution range of 8.0–4.0 Å. An EMP33 peptide dimer could be easily fitted to 3Fo - 2Fc electron density maps (Fig. 7b) and the entire complex of two EBP and two EMP33 molecules refined using rigid body refinement. The structure was further refined using simulated annealing with the slow cooling protocol in X-PLOR v3.1 and v3.8 with data from 50 to 2.7 Å. The Fobs were scaled anisotropically (B11 = -2.2 A2, B12 = -3.6 A2, B22 = -4.4 A2, B13 = 2.5 A2, B23 = 0.1 A2, B33 = 4.7 A2) and a bulk solvent correction was applied42. The structure was built into the electron density maps (3Fo - 2Fc and Fo - Fc) using the graphics program O43. After several cycles of refinement and model building, individual B values were refined. The refinement statistics are summarized in Table 2. Coordinates. The coordinates have been deposited in the Brookhaven PDB Databank (accession number 1eba).

Acknowledgments This work was supported in part by the NIH (I.A.W., M.A.G. and S.P.), and a New Jersey State Commission on Cancer Research grant (C.D.K.). O. L. was supported by a Rueff-Wormser postdoctoral fellowship and K.D.L. was supported by the N.I.H. Medical Scientist Training Program and the Program in Biological Sciences at U.C.S.F. We thank J. Tullai and F. McMahon for technical assistance, K. Hoey for peptide synthesis, H. Lashuel and J. Kelly for peptide ultracentrifugation and R.M. Stroud, J.A. Wells, S.L. Schreiber and D.C. Wiley for helpful discussions. This is publication 11174-MB from the Scripps Research Institute. Received 30 July, 1998; accepted 6 October, 1998

1003

1998 Nature America Inc.

• http://structbio.nature.com

1998 Nature America Inc.

• http://structbio.nature.com

articles 1. Watowich, S.S., Hilton, D.J. & Lodish, H.F. Activation and inhibition of erythropoietin receptor function: role of receptor dimerization. Mol. Cell. Biol. 14, 3535–3549 (1994). 2. Blau, C.A., Peterson, K.R., Drachman, J.G. & Spencer, D.M. A proliferation switch for genetically modified cells. Proc. Natl. Acad. Sci. USA 94, 3076–3081 (1997). 3. Livnah, O. et al. Functional mimicry of a protein hormone by a peptide agonist: the EPO receptor complex at 2.8 Å. Science 273, 464–471 (1996). 4. Boissel, J.-P., Lee, W.-R., Presnell, S.R., Cohen, F.E. & Bunn, H.F. Erythropoietin structure-function relationships-mutant proteins that test a model of tertiary structure. J. Biol. Chem. 268, 15983–15993 (1993). 5. de Vos, A.M., Ultsch, M. & Kossiakoff, A.A. Human growth hormone and extracellular domain of its receptor: crystal structure of the complex. Science 255, 306–312 (1992). 6. Wrighton, N.C. et al. Small peptides as potent mimetics of the protein hormone erythropoietin. Science 273, 458–463 (1996). 7. Johnson, D.L. et al. Amino-terminal dimerization of an erythropoietin mimetic peptide results in increased erythropoietic activity. Chem. Biol. 4, 939–950 (1997). 8. Wrighton, N.C. et al. Increased potency of an erythropoietin peptide mimetic through covalent dimerization. Nature Biotech. 15, 1261–1265 (1997). 9. Johnson, D.L. et al. Identification of a 13 amino acid peptide mimetic of erythropoietin and description of amino acids critical for the mimetic activity of EMP1. Biochemistry 37, 3699–3710 (1998). 10. Ihle, J.N., Witthuhn, B.A., Quelle, F.W., Yamamoto, K. & Silvennoinen, O. Signaling through the receptors. Ann. Rev. Immun. 13, 369–398 (1995). 11. Lai, S.Y. et al. Interleukin-4-specific signal transduction events are driven by homotypic interactions of the interleukin-4 receptor alpha subunit. EMBO J. 15, 4506–4514 (1996). 12. Lai, S.Y. et al. The molecular role of the common gamma c subunit in signal transduction reveals functional asymmetry within multimeric cytokine receptor complexes. Proc. Natl. Acad. Sci. USA 93, 231–235 (1996). 13. Lai, S.Y., Molden, J. & Goldsmith, M.A. Shared gamma(c) subunit within the human interleukin-7 receptor complex. A molecular basis for the pathogenesis of X-linked severe combined immunodeficiency. J. Clin. Invest. 99, 169–177 (1997). 14. Muthukumaran, G., Kotenko, S., Donnelly, R., Ihle, J.N. & Pestka, S. Chimeric erythropoietin-interferon gamma receptors reveal differences in functional architecture of intracellular domains for signal transduction. J. Biol. Chem. 272, 4993–4999 (1997). 15. Laskowski, R.A., MacArthur, M.W., Moss, D.S. & Thornton, J.M. PROCHECK: a program to check the stereochemical quality of protein structures. J. Appl. Crystallogr. 26, 283–291 (1993). 16. Bork, P., Holm, L. & Sander, C. The immunoglobulin fold. Structural classification, sequence patterns and common core. J. Mol. Biol. 242, 309–320 (1994). 17. Wojchowski, D.M. & He, T.C. Signal transduction in the erythropoietin receptor system. Stem Cells 11, 381–392 (1993). 18. Heldin, C.-H. Dimerization of cell surface receptors in signal transduction. Cell 80, 213–223 (1995). 19. Elliott, S., Lorenzini, T., Yanagihara, D., Chang, D. & Elliott, G. Activation of the erythropoietin (EPO) receptor by bivalent anti-EPO receptor antibodies. J. Biol. Chem. 271, 24691–24697 (1996). 20. Schneider, H. et al. Homodimerization of erythropoietin receptor by a bivalent monoclonal antibody triggers cell proliferation and differentiation of erythroid precursors. Blood 89, 473–482 (1997). 21. Huber, R., Deisenhofer, J., Colman, P.M., Matsushima, M. & Palm, W. Crystallographic structure studies of an IgG molecule and an Fc fragment. Nature 264, 415–420 (1976). 22. Harris, L.J. et al. The three-dimensional structure of an intact monoclonal antibody for canine lymphoma. Nature 360, 369–372 (1992).

1004

23. Ohashi, H., Maruyama, K.K., Liu, Y.C. & Yoshimura, A. Ligand-induced activation of chimeric receptors between the erythropoietin receptor and receptor tyrosine kinases. Proc. Natl. Acad. Sci. USA 91, 158–162 (1994). 24. Maruyama, K., Miyata, K. & Yoshimura, A. Proliferation and erythroid differentiation through the cytoplasmic domain of the erythropoietin receptor. J. Biol. Chem. 269, 5976–5980 (1994). 25. Boni-Schnetzler, M., Scott, W., Waugh, S.M., DiBella, E. & Pilch, P.F. The insulin receptor. Structural basis for high affinity ligand binding. J. Biol. Chem. 262, 8395–8401 (1987). 26. Lee, J.T., O’Hare, T., Pilch, P.F. & Shoelson, S.E. Insulin receptor autophosphorylation occurs asymmetrically. J. Biol. Chem. 268, 4092–4098 (1993). 27. Dong, Y.J. & Goldwasser, E. Evidence for an accessory component that increases the affinity of the erythropoietin receptor. Exp. Hematol. 21, 483–486 (1993). 28. Johnson, D.L. et al. Refolding, purification and characterization of human erythropoietin binding protein produced in E. coli. Prot. Express. Purif. 7, 104–113 (1996). 29. Carroll, M.P. et al. Erythropoietin induces Raf-1 activation and Raf-1 is required for erythropoietin induced proliferation. J. Biol. Chem. 266, 14964–14969 (1991). 30. Goldsmith, M.A., Xu, W., Amaral, M.C., Kuczek, E.S. & Greene, W.C. The cytoplasmic domain of the interleukin-2 receptor beta chain contains both unique and functionally redundant signal transduction elements. J. Biol. Chem. 269, 14698–14704 (1994). 31. Dexter, T.M., Garland, J., Scott, D., Scolnick, E. & Metcalf, D. Growth of factordependent hemopoietic precursor cell lines. J. Exp. Med. 152, 1036–1047 (1980). 32. Soh, J. et al. Identification and sequence of an accessory factor required for activation of the human interferon gamma receptor. Cell 76, 793–802 (1994). 33. Gaffen, S.L. et al. Signaling through the interleukin 2 receptor beta chain activates a Stat-5-like DNA-binding activity. Proc. Natl. Acad. Sci. USA 92, 7192–7196 (1995). 34. Kotenko, S.V. et al. Interaction between the components of the interferon gamma receptor complex. J. Biol. Chem. 270, 20915–20921 (1995). 35. Stura, E.A. et al. Preliminary crystallographic investigations of glycinamide ribonucleotide transformylase. J. Biol. Chem. 264, 9703–9706 (1989). 36. Stura, E.A. & Wilson, I.A. The streak seeding technique in protein crystallization. J. Cryst. Growth 110, 270–282 (1991). 37. Matthews, B.W. Solvent content of protein crystals. J. Mol. Biol. 33, 491–497 (1968). 38. Otwinowski, Z. Oscillation Data Reduction Program. In Proceedings of the CCP4 Study Weekend (eds. Sawyer, L., Isaacs, N. & Bailey, S.) 56-62 (SERC Daresbury Laboratory, Daresbury, England, 1993). 39. Otwinowski, Z. & Minor, W. Processing of x-ray diffraction data collected in oscillation mode. Meth. Enz. 276, 307–326 (1997). 40. Navaza, J. AMoRe: an automated package for molecular replacement. Acta Crystallogr. A50, 157–163 (1994). 41. Brünger, A.T. X-PLOR, Version 3.1: A system for X-ray and NMR. 3.1.1 edn (Yale University Press, New Haven, CT, 1992). 42. Jiang, J.S. & Brunger, A.T. Protein hydration observed by X-ray diffraction. Solvation properties of penicillopepsin and neuraminidase crystal structures. J. Mol. Biol. 243, 100–115 (1994). 43. Jones, T.A., Zou, J.Y., Cowan, S.W. & Kjeldgaard, M. Improved methods for building protein models in electron density maps and the location of errors in these models. Acta Crystallogr. A47, 110–119 (1991). 44. Syed, R.S. et al. Efficiency of signaling through cytokine receptors depends critically on receptor orientation. Nature 395, 511–516 (1998). 45. Cheetham, J.C. et al. NMR structure of human erythropoietin and a comparison with its receptor bound conformation. Nature Struct. Biol. 5, 861–866 (1998).

nature structural biology • volume 5 number 11 • november 1998