Erythropoietin Receptor Activation by a Ligand-Induced Conformation Change

See allHide authors and affiliations

Science  12 Feb 1999:
Vol. 283, Issue 5404, pp. 990-993
DOI: 10.1126/science.283.5404.990


Erythropoietin and other cytokine receptors are thought to be activated through hormone-induced dimerization and autophosphorylation of JAK kinases associated with the receptor intracellular domains. An in vivo protein fragment complementation assay was used to obtain evidence for an alternative mechanism in which unliganded erythropoietin receptor dimers exist in a conformation that prevents activation of JAK2 but then undergo a ligand-induced conformation change that allows JAK2 to be activated. These results are consistent with crystallographic evidence of distinct dimeric configurations for unliganded and ligand-bound forms of the erythropoietin receptor.

The erythropoietin receptor (EpoR) shares both structural and functional features with the cytokine receptor superfamily that includes the interleukins, human growth hormone (hGH), and colony-stimulating factor (1, 2). Crystal structures and biochemical analysis have led to the generally accepted dimerization model of growth factor–mediated receptor activation, where monomeric receptors remain inactive until ligand binds to and oligomerizes the receptors, allowing autophosphorylation of receptor-associated intracellular kinases (3). However, dimerization or oligomerization of receptors is a required but not necessarily sufficient condition for receptor activation.

Studies of hGH bound to human growth hormone receptor (hGHR) and of Epo, agonist, or antagonist peptides bound to EpoR have shown that all of these ligands bind to receptor dimers with different dimer interface orientations, which correlate with differences in the efficiency of receptor activation (4). Livnah et al. have also determined the crystal structure of unliganded EpoR and show that it is also a dimer, but with a dramatically different arrangement of the two subunits (5). A significant feature of the unliganded extracellular dimer is that the COOH-termini of the monomers, the points of insertion into the membrane, are separated by 73 Å, compared with 39 Å when the agonist peptide EMP1 is bound to EpoR. On the basis of this finding and the ligand-bound structures, they propose an alternative allosteric mechanism of receptor activation in which ligand-induced reorganization of the dimer brings the intracellular domains into closer proximity, allowing associated JAK2s to come into contact and autophosphorylate (Fig. 1A).

Figure 1

Schematic representation of the crystallographic structure–based Epo receptor activation hypothesis and of the experimental strategy. (A) Receptors are dimers in their unliganded state (5). The extracellular domain exists in a conformation that holds the intracellular domains and associated JAK2s separated from each other by ∼73 Å. On binding ligand (Epo or peptide agonist EMP1), the extracellular dimer is reorganized, bringing the intracellular domains to within 39 Å of each other, allowing autophosphorylation and activation of the JAK2s. (B and C) The extracellular and transmembrane domains of murine EpoR [EpoR(1-270)] are fused to one of two complementary fragments of murine DHFR (F[1,2] or F[3]) through flexible linkers (gray lines) consisting of (Gly-Gly-Gly-Gly-Ser)N repeats whereN = 1, 2, or 6 to generate the following: EpoR(1-270)-5aa-F[1,2] or -F[3], EpoR(1-270)-10aa-F[1,2] or -F[3], and EpoR(1-270)-30aa-F[1,2] or -F[3]. Cells transfected with these fusions express receptors at the membrane surface. Fluorescein-methotrexate (fMTX) is taken up by cells and binds to reconstituted DHFR (F[1,2]+F[3]) and is retained in the cell. Unbound fMTX is rapidly released from the cells by active transport (8). Panel (B) shows how fusions in which DHFR fragments are connected to receptors by a 5-aa linker cannot complement in the unliganded receptor. When receptors bind to Epo or EMP1, DHFR fragment complementation can take place. In panel (C), fusions with the 30-aa linker allow complementation of DHFR fragments whether receptors are ligand-bound or not. (D) Results of complementation experiments with EpoR extracellular and transmembrane domains should be reproducible with complete EpoR receptor complex, including associated JAK2. Here is shown one such experiment in which DHFR fragments are fused to the COOH-terminal of JAK2 and coexpressed in cells along with full-length EpoR.

To test this model, we used a fluorescent assay based on dimerization-induced complementation of designed fragments of the murine enzyme dihydrofolate reductase (DHFR) (6, 7). The basis for the assay is that complementary fragments of DHFR, when they are fused to interacting proteins and expressed in cells, will reassemble and bind to the high-affinity (dissociation constantK d = 540 pM) fluorescein-conjugated inhibitor methotrexate (fMTX) in a 1:1 complex. The fMTX is retained in cells by this complex, whereas unbound fMTX is actively transported out of the cells (8). fMTX-DHFR complexes can be monitored by fluorescence microscopy, fluorescence-activated cell sorting (FACS), or spectroscopy. To test the allosteric model we reasoned as follows: if the receptor dimer transmembrane domains are separated by the distance observed in the crystal structure of unliganded EpoR (73 Å), then DHFR fragments (F[1,2] and F[3]) fused to the COOH-termini of the transmembrane domains will complement only if ligand induces the conformation change that allows the fragments to fold into the precise three-dimensional structure of DHFR (Fig. 1B) (7, 9). This requires that the NH2-termini of the DHFR fragments be 8 Å apart. Insertion of flexible linker peptides between the transmembrane domain and DHFR fragments allowed us to probe the distance between the insertion points of the extracellular domain dimer. Linkers of 5, 10, and 30 amino acids allowed us to probe the range of possibilities as they correspond to extended lengths of 20, 40, and 120 Å, respectively (∼4 Å per peptide bond). It would be predicted that with 5–amino acid (5-aa) linkers, complementation of the DHFR fragments would only occur in the presence of ligands (Fig. 1B), with 10–amino acid (10-aa) linkers some complementation could occur in the absence of ligand, but with 30–amino acid (30-aa) linkers it would be independent of ligand (Fig. 1C).

CHO DUKX-B11 (DHFR) cells were cotransfected with EpoR extracellular and transmembrane domains [EpoR(1-270)] fused to peptide linkers and one of the two DHFR fragments F[1,2] or F[3]. Cotransfectants were selected for survival in the presence of Epo (2 nM), to induce DHFR complementation, in nucleotide-free medium (selection for DHFR activity). Fluorescence flow cytometric analysis of the cotransfectants incubated with fMTX showed significant increases (by a factor of ∼8) in fluorescence over background levels when cells were treated with Epo or EMP1. In the absence of ligands, cells transfected with EpoR-DHFR fragment fused through 5-aa linkers showed no fluorescence, compared with nontransfected cells; cells with a 10-aa linker showed a weak fluorescence, whereas those expressing fusions with a 30-aa linker showed the same level of fluorescence in the presence or absence of ligands (Fig. 2). These results are consistent with the allosteric model of Livnah et al. (5) and cannot be due to steric hindrance caused by proteins at the membrane intracellular surface that might prevent DHFR fragment complementation for the shorter 5-aa and 10-aa linkers. If this was the case, the signal seen in the presence of ligand for the 30-aa linker, and to some extent for the 10-aa linker, would be higher than for the 5-aa linker. In fact, the fluorescence intensity and number of activated receptors as determined spectroscopically (about 8500 receptors per cell) is the same in all three cases (10). Furthermore, dose-response analyses based on FACS data showed saturable single-site binding withK d's of 164 pM and 168 nM for Epo and EMP1, respectively (10). These results are consistent with previous binding studies (1, 11).

Figure 2

Fluorescence flow cytometric analysis of CHO DUKX-B11 cells labeled with fMTX. Cells were transfected with EpoR(1-270)-5aa-F[1,2] and -F[3] (top), with EpoR(1-270)-10aa-F[1,2] and -F[3] (middle), and with EpoR(1-270)-30aa-F[1,2] and -F[3] (bottom). Histograms are based on analysis of fluorescence intensity for 10,000 cells at flow rates of ∼1000 cells per second. Data were collected on a Coulter XL 4 color FACS analyzer (Coulter-Beckman) with stimulation by an argon laser tuned to 488 nm with emission recorded through a 525-nm band width filter. Histograms represent the response in absence of ligands (white histogram), with 10 nM Epo (light gray histogram), or with 10 μM EMP1 (dark gray histogram). Constructs and methods are described in (16).

To demonstrate that the allosteric model applies to the complete receptor complex, we coexpressed full-length EpoR and JAK2 in COS-7 cells fused to complementary F[1,2] and F[3] fragments through the 5-aa linker (Fig. 3). Coexpression of EpoR-F[1,2] with EpoR-F[3] gave a response in the absence of ligands. These results are consistent with circular dichroism and nuclear magnetic resonance studies showing that the intracellular domain of EpoR is partially unfolded in the absence of JAK2 (12). Taken with our results presented here, we suggest that the 236–amino acid intracellular domain acts as a very long linker itself, resulting in reconstitution of DHFR. Fluorescence was also seen when JAK2-F[1,2] was coexpressed with EpoR-F[3] in the absence of ligands, consistent with previous studies indicating that this interaction is constitutive (13). JAK2-F[1,2] and JAK2-F[3] coexpressed with full-length EpoR (not fused to either F[1,2] or F[3]; see Fig. 1D) showed an induced response to Epo or EMP1. This key result demonstrates that coupling of a ligand-induced conformation change in the extracellular domain allows two JAK2s associated with the receptor intracellular domain to come into proximity. Finally, coexpressed alone, JAK2-F[1,2] and JAK2-F[3] showed no detectable fluorescence. Cotransfected EpoR and JAK2 was also shown to function normally with the attached F[1,2] or F[3] fragments because ligand-induced phosphorylation of JAK2 and EpoR were detected (12).

Figure 3

Fluorescence microscopy of COS-7 cells expressing EpoR-DHFR and JAK2-DHFR fragment fusions. Cells were grown on 18-mm glass coverslips to ∼3 × 105 in Dulbecco's modified Eagle's medium (Life Technologies) enriched with 10% cosmic calf serum (Hyclone) in 12-well plates. Plasmids pMT3 harboring full-length EpoR or EpoR and JAK2 fused through a 5-aa linker to F[1,2] or F[3] were created and cells were transiently transfected with the different constructs. Cells were prepared as described in (16) for FACS except that after the last wash with PBS, cells were mounted on glass slides. Fluorescence microscopy was performed on live cells with a Zeiss Axiophot microscope (objective lens Zeiss Plan Neofluar 40×/0.75).

The results presented here are consistent with the structural allosteric model of Livnah et al. (5) (Fig. 1A). The model is not contradictory to dimerization models; dimerization is a required, but not necessarily sufficient, condition for receptor activation. Fluorescence resonance energy transfer studies have demonstrated that other receptors may exist as preformed dimers, including those for interleukin-1, interleukin-2, and epidermal growth factor (14). Investigation with strategies such as those described here will help to determine if these and other receptors may be activated by a similar mechanism as that of EpoR.


View Abstract

Navigate This Article