Solution structure of the transmembrane domain of the mouse erythropoietin receptor in detergent micelles

Erythropoiesis is regulated by the erythropoietin receptor (EpoR) binding to its ligand. The transmembrane domain (TMD) and the juxtamembrane (JM) regions of the EpoR are important for signal transduction across the cell membrane. We report a solution NMR study of the mouse erythropoietin receptor (mEpoR) comprising the TMD and the JM regions reconstituted in dodecylphosphocholine (DPC) micelles. The TMD and the C-terminal JM region of the mEpoR are mainly α-helical, adopting a similar structure to those of the human EpoR. Residues from S216 to T219 in mEpoR form a short helix. Relaxation study demonstrates that the TMD of the mEpoR is rigid whilst the N-terminal region preceding the TMD is flexible. Fluorescence spectroscopy and sequence analysis indicate that the C-terminal JM region is exposed to the solvent. Helix wheel result shows that there is hydrophilic patch in the TMD of the mEpoR formed by residues S231, S238 and T242, and these residues might be important for the receptor dimerization.

The mechanism of ligand induced JAK2 activation was described in a recent study using GHR as a model 10 . In this study, GHR was shown to form a dimer in vivo through its TMD. The TM helices are parallel in the basal state and form a left-hand crossover state when the receptor binds to its ligand. The movement of the TMD helices results in the removal of the pseudokinase inhibitory domain of JAK2 to activate JAK2 10 . The receptor activation mechanism is through the function of TM helix dynamics in a lipid membrane and the EpoR may also be suitable for this model 11 . Based on the accumulated studies, it is obvious that the TMD of the EpoR is important for receptor function. The TMD of the EpoR are also shown to form dimers when it was reconstituted into detergent micelles 12 . We also conducted a structural study on the human EpoR (hEpoR) in dodecylphosphocholine (DPC) micelles. The hEpoR was demonstrated to be able to form dimers in micelles and its JM region formed a helix as predicted 13 .
The TMD of mouse EpoR (mEpoR) and hEpoR share very high sequence homology. It was demonstrated that mEpoR might have a higher binding affinity than hEpoR, which might be one of the reasons that mEpoR is more active than hEpoR 12 . To further understand the structure of the mEpoR, we present a nuclear magnetic resonance (NMR) study of the TMD and JM regions comprising residues 212-259 of the mouse EpoR. Our results show that mEpoR has a similar structure to hEpoR with the exception that its N-terminal region preceding the TMD contains a short helix due to lack of a proline residue. There is a hydrophilic patch formed by residues S231, S238 and T242 in the TMD. Our structural and dynamic information of the mEpoR will be useful to understand the role of the TMD and JM regions in signal transduction.

Results
Secondary structure of mEpoR. To conduct structural and dynamic studies for the mEpoR (Fig. 1), the construct was purified in the presence of DPC micelles. The backbone assignment of the mEpoR in DPC micelles was achieved using conventional 3D experiments (Fig. 1C) 13 . The assignment has been deposited into the BioMagResBank under accession number 25396. The calculated chemical shift index from the assigned Cα chemical shifts indicated that the mEpoR has similar structure to the hEpoR 13 . Further secondary structural analysis of backbone chemical shifts using TALOS+ indicated that the TMD of the mEpoR forms a helix. Residues compassing S216 to T219 preceding the TMD have a tendency to form a short helix ( Fig. 2A,B). Residue 220 of mEpoR is an alanine instead of a proline in hEpoR, which might favor the formation of this short helix. The sequential NOE connectivity result also confirmed the secondary structural elements in mEpoR (Fig. 2C). The TMD and its C terminal JM region of the mEpoR form a helix, which is similar to the hEpoR. H-D exchange experiment also suggested that residues in the TMD were protected from exchanges (Fig. 2D). Missing the assignment of the backbone amide and amide proton of R251 and lacking inter-proton NOE connectivity between R250 and its preceding residues suggested that residues R250 and R251 might have exchanges with the environment. TALOS+ analysis showed that residues from S216 to T219 have a tendency to be a helix, but the NOEs between Hα of residue i and HN of residue i + 3(4) were not observed. This might arise from the fact that this region is flexible (Fig. 2C).
Backbone dynamics of the mEpoR in DPC micelles. To understand the flexibility of the mEpoR in the picosecond-to-nanosecond timescale, the backbone T 1 , T 2 and 15 N-1 H-heteronuclear NOE (het-NOE) experiments were measured at 600 MHz proton resonance frequencies (Fig. 3). High T 1 , hetNOE and low T 2 values of the TMD suggested it is rigid in solution. The N-terminal region of the construct is dynamic characterized by low T 1 , hetNOE and high T 2 values. For the C-terminal JM region compassing residues L253 to W258, it exhibited a hetNOE value close to 0.6. The generalized order parameter (S 2 ) was obtained using a model-free approach (Fig. 3). Residues from 227 to 248 exhibited S 2 values close to 0.85, suggesting that these residues are rigid in solution. Residues from 252 to 258 exhibited S 2 values close to 0.7, suggesting that this region is not as rigid as the TMD. For the N-terminal region, most of the residues showed low S 2 values and residues 217 and 218 exhibited slightly higher S 2 values than other residues within this region, which might arise from the fact that these two residues are in the short helix (Fig. 3).
To further understand whether the C-terminal JM region is exposed to the solvent, we analyzed our NOE data. It is clear that the amide protons from hydrophilic residues compassing 250 to 257 exhibited NOEs with water protons, demonstrating that these residues were not buried in DPC micelles (Fig. 4A). Interestingly, amide protons of the three conserved hydrophobic residues including L253, I257 and W258 showed no NOE or weak NOEs with water protons, which indicated that these residues may interact with micelles or having exchanges. We then performed bioinformatics analysis of the construct using a half-sphere exposure (HSE) algorithm 14 . In the HSE analysis, a residue is considered as a sphere with two half parts. HES-up corresponds to the direction of the chain side of a residue and HES-down corresponds to the direction of the opposite side. This method also predicts the contact number (CN) that is a measurement of a residue burial in proteins. The HSE-up showed that most residues in the TMD and few residues in both N-and C-termini were protected from exposure. Taking the HES-down and CN results together, the N-and C-termini of the constructs are exposed to the solvent (Fig. 4B). Although this bioinformatics study might not suitable for mEpoR, the information obtained suggested that the C-terminal JM region is exposed to the solvent. There is only one tryptophan residue present (W258) in the protein sequence. Tryptophan fluorescence experiment was carried out. In the fluorescence spectrum, the emission maximum of the mEpoR in DPC micelles was 350 nm (Fig. 4C), indicating that W258 is exposed to the solvent because the emission maximum will be shifted to a lower wavelength if the tryptophan residue is buried in a hydrophobic environment.

Structure of the mEpoR in DPC micelles.
Structures of the mEpoR were solved using NOE restraints derived from NOE experiments, dihedral angle restraints derived from TALOS+ and hydrogen bond restraints based on H-D exchange experiment. The calculated 20 lowest energy structures do not have any distance or dihedral angles violations that exceed 0.5 Å and 5°, respectively. The structural ensemble has been deposited in PDB under accession number 2MXB. Table 1 shows the statistics of the determined mEpoR structures. These structures show an extended transmembrane helix formed by residues from 225-257 with backbone r.m.s.d. of 0.60 ± 0.21 Å. The length of the TMD is 32.8 Å and the C-terminal JM region is helical. Although the N-terminus of the mEpoR is flexible, there is a short helix formed by residues from S216 to S221 (Fig. 5B). Electrostatic surface potential analysis of the mEpoR indicated that residues from 226 to 247 form a hydrophobic surface, which is not surprising for the TMD (Fig. 5C). The N-terminal region of mEpoR contains a negatively charged surface. There is a positive surface formed by R250 and R251 in the JM region and a hydrophobic patch formed by the three hydrophobic residues including L253, I257 and W258 (Fig. 5).
The structure of the mEpoR is similar to that of the human EpoR except that there is a short helix at the N-terminal region (Fig. 6A). Further CD spectra of the mouse and human EpoRs showed that these two proteins had slightly different spectra and mEpoR exhibited slightly lower maximum at 208 nm (Fig. 6B) , suggesting that the helical content in mEpoR is slightly more than hEpoR. Helix wheel presentation showed that the sequence difference at position 238 in the TMD might explain the functional difference observed between mEpoR and hEpoR. S238 in mEpoR (L238 in hEpoR) forms a hydrophilic patch with S231 and T242 (Fig. 6C). This patch may favor TMD-TMD dimerization because it behaves like a small-XXX-small motif that is commonly used in TMD-TMD interactions 15 . The presence of this hydrophilic patch may explain why mEpoR has a higher dimerization affinity than hEpoR. For the JM region, the amino acid difference at position 254 did not alter the position of the three conserved hydrophobic residues including L253, I257 and W258 in both receptors, which still can form a hydrophobic patch that is important for the receptor function (Fig. 6C).

Discussion
The signal transduction through EpoR includes the movement of its TMD. To investigate the mechanism, we solved the solution structure of the mEpoR in DPC micelles. DPC has been used in many structural studies of membrane proteins due to its close structure to the cell membrane 16 . The DPC micelles might not be an ideal system for the mEpoR to form dimers because only a minor population of dimer species  Residues in additionally allowed regions 9.6 Residues in generously allowed regions 0.8

Residues in disallowed regions 0
Average RMSD to mean (Å) was observed in the sample 17 (Fig. S1). It was also shown that hEpoR could form dimer or trimers in different micelles, suggesting that the membrane mimicking system affects receptor dimerization 13 . As the length of the mEpoR used in this study is short, the structural information of the mEpoR obtained in micelles will be close to the physiological conditions 18 . Previous studies on GHR and full-length cytochrome P450 have suggested that the structures of the TMDs were not altered during signal transduction 10,19 . A study on the TMD of the TpoR-a homology of EpoR showed that it could form three stable conformations. Rotation of the TMD might also be one of the mechanisms during receptor activation 20 . The rigidity of the mEpoR TMD also suggested that there might not be any conformational changes in the TMD during signal transduction. There is only seven amino acids difference between the mEpoR and hEpoR that were used in structural studies 13 . Both receptors have similar structures and dynamics (Fig. 6). The residues preceding the TMD are demonstrated to be dynamic for both mEpoR and hEpoR (Fig. 4). As predicted from a previous study, TMD and the JM region formed an extended helix 8 . Dynamic study show the C-terminal JM regions both receptors containing residues L253 to W258 are stable, but not as rigid as the TMDs in DPC micelles (Fig. 4). Although the JM region contains several conserved hydrophobic residues that are important for signal transduction, this region was shown to be exposed to the solvent (Fig. 4). These three hydrophobic residues might bind JAK2 under certain conditions. Structural difference was also observed between the human and mouse receptors (Fig. 6). For the mEpoR, there is a short helix preceding the TMD (Fig. 2), which affected the length between the extracellular domain and the TMD. Having a short helix will shorten the length of the linker region, which might be one of the reasons that mouse EpoR has a stronger dimerization affinity than human EpoR. Although TMDs of both mEpoR and hEpoR adopt similar structures, there are three different residues present in this domain (Fig. 1). In mEpoR residue 231 is a serine (S231) in place of a leucine, which forms a hydrophilic patch with other residues including S238 and T242. This patch can favor TMD-TMD interactions (Fig. 6C), which might make mEpoR have a stronger dimerization affinity than hEpoR.
The TMD of the EpoR was shown to be essential for the dimerization of the entire dimer 7 . By replacing the extracellular domain of the EpoR with a dimeric coiled coil, Seubert et al. have identified three different TM conformations representing fully active, partially active and inactive receptors 21 . Residues including S231, S238, T242 and S248 were predicted to be important for dimerization 21 . In a double mutation study, Kubatzky et al. showed that residues L240 and L241 are important for EpoR TMD dimerization 22 . In their study, it was shown that the dimerization mechanisms might be different for GHR and EpoR 22 . We also investigated the EpoR dimerization in DPC micelles. The EpoR dimeric species was observed when the mEpoR was reconstituted in DPC micelles in a cross-linking study (Fig. S1). Several residues exhibited line broadening in the 1 H-15 N-HSQC spectra at different protein to DPC ratios suggesting that mEpoR can form dimers in DPC micelles (Fig. S2). A left-hand dimer structure of mEpoR can be built based on few inter-molecular NOEs observed from residues including S231, S238 and T241 (Fig. S3). It has been noted that the number of the inter-molecular NOEs observed in this system could not generate high resolution structures. This might arise from the fact that DPC is not an ideal system for mEpoR to form dimers because the mEpoR dimer species were low even at a high protein to DPC ratio and in the presence of a cross linker (Fig. S1). Although detergent micelles is suitable for solution NMR structural study of membrane proteins due to the size of protein-detergent complex, detergents tend to destabilize protein-protein or protein-lipid interactions. Bicelle systems such as isotropic bicelles might be a suitable system for studying EpoR dimer structures because this system is able to sustain structure and function of membrane proteins containing both TMD and large water soluble regions [23][24][25] . Further structural study of this construct in other membrane systems or a new construct containing the extracellular domain and TMD using both solution and solid states NMR 26 will provide more information to understand signal transduction through the EpoR.
In summary, we solved the solution structure of the TMD the JM regions of mEpoR in DPC micelles. The mEpoR showed similar structure to hEpoR with the exception that the N-terminal region preceding the TMD contains a short helix. The N-terminal region of the EpoR localized between the extracellular domain and the TMD is flexible. The TMD forms a rigid helix in DPC micelles. The C-terminal JM region is helical and stable in solution. The TMD of mEpoR contains a hydrophobic patch formed by residues S231, S238 and T242 and this patch may be important for dimerization.

Experimental Procedures
Materials. The cDNA encoding residues S212-P259 of the mEpoR were synthesized by Genscript.
Expression vector such as pET-29b plasmid was purchased from Merck. The SDS-PAGE system was purchased from Invitrogen. Protein sample loading dye, molecular weight standards were purchased from Bio-rad. Bl21 (DE3) competent cells were purchased from StrataGene. β -D-1-thiogalactopyranoside (IPTG), Dithiothreitol (DTT) and detergents including dodecylphosphocholine (DPC), deuterated DPC (D-DPC) were purchased from Anatrace or Avanti. The 15 NH 4 Cl, 13 C-glucose and D 2 O were purchased from Cambridge Isotope Laboratories. All other chemicals were purchased from Sigma.
Expression and purification of the EpoR. The mouse EpoRs were cloned into the NdeI and XhoI sites of pET29b and the resulting plasmid pET29-mEpoR encodes a protein sequence containing S212-P259 of mouse EpoR (mEpoR) with six histidine residues (HHHHHH) at its C-terminus. The EpoR was expressed and purified as described previously 13,27 . To prepare for a sample with low DPC to protein ratio, mEpoR was eluted with an elution buffer containing 20 mM sodium phosphate, 300 mM imidazole and 5 mM DPC. Purified fractions with low DPC to protein ratio were collected and was further purified through a gel filtration chromatography and concentrated for further studies.
NMR experiments and structure determination. Uniformly 13 C/ 15 N-labeled mEpoR was prepared using the aforementioned method. Sample was concentrated to 0.5 mM in a buffer containing 20 mM sodium phosphate, pH 6.5, 1 mM DTT, 200 mM D-DPC and 10% D 2 O. NMR spectra were recorded at 313 K on a Bruker Avance 600 or Avance 700 spectrometer with a cryogenic triple resonance probe. Data acquired were processed with NMRPipe 28 and analyzed using NMRView 29 . Secondary structure was identified by analysis of 13 C secondary chemical shifts 30 and TALOS+ 31 . Distance constraints were collected from a 3D 15 N-edited NOESY (mixing time = 100 ms) experiment under the aforementioned conditions. The NOE connection was plotted using CYANA 32 . The hEpoR structure determination was carried out using XPLOR-NIH with python interface [33][34][35] . The backbone dihedral angles restraints were generated using TALOS+ based on the backbone chemical shifts 31 . The nuclear Overhauser effect (NOE) peaks were picked manually from a 3D 15 N-edited NOESY and calibrated using NMRView 29 . Structure determination was carried out using a randomized template. Simulated annealing was performed and energy minimization were carried out as previously described 36 .H-D exchange experiment was carried out using the method described 37 . T 1 , T 2 and hetNOE experiments 38 were measured at 313K on a Bruker Avance II 600 MHz spectrometer. For T 1 measurement, the relaxation delays of 10, 50, 100, 200, 400, 800, 1200, 1400, 1600 and 1800 ms were recorded. For T 2 measurement, the data was acquired with delays of 16.9, 34, 51, 68, 85, 102, 119, 136 and 153 ms. The hetNOE was obtained using two datasets that were collected with and without initial proton saturation for a period of 3 s 39 . Protein structures were analyzed by using PROCHECK-NMR 40 . The relaxation was further analyzed using model-free 41,42 . Circular dichroism (CD) Spectroscopy. Far-UV CD spectrum of the peptide at a concentration of 20 μ M was analyzed in a buffer containing 20 mM sodium phosphate, pH6.5 and 0.5% dodecylphosphocholine (DPC). The instrument was referenced against a cuvette containing buffer without EpoR. The CD spectra were recorded at 25 °C and 40 °C.
Bioinformatics analysis of the mEpoR sequence. Predicting amino acid exposure based on the amino acid sequence was carried out using half-sphere exposure (HES) method 14 . The protein sequence was analyzed using the web server (http://sunflower.kuicr.kyoto-u.ac.jp/~sjn/hse/). The obtained HES-up, HSP-down, and contact number (CN) were plotted against residue number.
Fluorescence spectroscopy. The fluorescence measurement was conducted as previously described 43,44 . The fluorescence emission spectra were measured in a 96-well plate. Purified hEpoR was prepared in 50 μ M in a buffer that contained 20 mM sodium phosphate, pH6.5 and 40 mM DPC. Samples with 100 μ l volume were subjected to analysis. Excitation wavelength was 280 nm and the emission was scanned from 305 to 400 nm. Experiment was carried out at 37 °C.