A Host-guest Relationship in Bone Morphogenetic Protein Receptor-II Defines Specificity in Ligand-Receptor Recognition

Abstract

One of the most intriguing questions confronting the Bone Morphogenetic Protein family is the mechanism of ligand recognition, since there are more ligands than receptors. Crystal structures of two type II receptors ActR-II and BMPR-II are essentially identical, and a loop structure (A-loop) has been suggested to play a role in determining ligand specificity. Solution biophysical study showed mutations of several A-loop residues in these two receptors exert different ligand binding effects. Thus, the issues of mechanism of ligand recognition and specificity remain unresolved. We examined effects of mutations of residues Y40, G47, and S107 in BMPR-II receptor. These residues are not identified in contact with the ligand in the BMP-7-BMPR-II complex, but are found mutated in genetic diseases. They are likely to be useful in identifying their roles in differentiating the various BMP ligands. Spectroscopic probing revealed little mutation-induced structural change in BMPR-II. Ligand binding studies revealed that Y40 plays a significant role in differentiating three distinct ligands; G47 and S107 affect ligand binding to a lesser extent. The role of the A-loop in ActR-II or BMPR-II is dependent on the host sequence of the receptor extracellular domain (ECD) in which it is embedded, suggesting a host-guest relationship between the A-loop and the rest of the ECD. Computational analysis demonstrated a long-range connectivity between Y40, G47, and S107 and other locations in BMPR-II. An integration of these results on functional energetics and protein structures clearly demonstrate, for the first time, an intra-domain communication network within BMPR-II.

Activins (Acts) and Bone Morphogenetic Proteins (BMPs) are members of the transforming growth factor-β (TGF-β) superfamily^1–3 and are multifunctional proteins, playing key roles in numerous biological processes, such as embryonic development, cell differentiation, proliferation, morphogenesis, tissue repair, homeostasis, and apoptosis.^4–6 Based on the extent of their amino acid sequence homology, numerous BMPs have been identified and classified into six subfamilies. BMPs exert their effects on cells by forming a complex with the extracellular domains (ECDs) of two different types of serine/threonine kinase receptors, known as type I and type II.^7–9 Four mammalian type I and three type II receptors have been shown to bind BMPs.¹⁰ Since there are multiple ligands and receptors, one of the most intriguing questions confronting this system is the mechanism of ligand recognition and the structural insights on the mechanism. Studies on the ligand-receptor complexes have identified protein regions and amino acid residues in the receptor essential for ligand binding.^11–16 Structures of more than a dozen receptor-ligand complexes in the TGF-β superfamily have been determined at atomic resolution.^11,13–21 In brief, these studies have shown that the ECD of the type I receptors contains two β-sheets and an α-helix organized around 5 disulfide bonds, and the ECD of the type II receptors contains 3 two-stranded β-sheets (three-finger toxin-fold) also organized around 5 disulfide bonds.^22,23 These studies have revealed that the free receptors have a similar overall backbone fold (see Fig. 1) and further suggested that the different BMPs bind to the same overall ligand binding domain without inducing significant structural changes. Ligands are always bound to the concave surface of the receptor ECD. The hydrophobic residues in Finger 2 are always intimately involved as the binding interface (Fig. 1). However, there are still incongruous conclusions derived from structural and solution studies. For example, there is a disagreement on the role of specific residues or structural elements such as the loops in the receptor in determining ligand specificity.^12,24

Figure 1. Alignment of the structures of ActR-II receptor (PDB:1LX5, red) and BMPR-II receptor (PDB:2HLQ, green).

The locations of natural single site mutants are marked as green dots while grey spheres are contact sites in the receptor within 4Å with BMP-7 in a BMP-ActR-II complex (PDB: 1LX5). Hydrophobic patch of the ligand interface is marked as cyan spheres at the lower part of Finger 2. The A-loop is indicated and is located at the bottom of the structure.

The goal of the present study is to compare and contrast the BMPR-II and ActR-II receptors whose ECDs have essentially the same fold, although their sequences are only 24% identical. Although the apo structures of BMPR-II and ActR-II are known, the structure of BMP-7-ActR-II complex is the only holo structure known for these two receptors. We utilized four strategies to identify the structural elements and molecular mechanism on ligand recognition: First, we included mutations identified in genetic diseases in order to perturb the structure of the BMPR-II receptor, assuming that these sites are important, since mutations lead to disease states. Recent discoveries of mutations in the BMPR-II receptor in patients with different diseases, including pulmonary veno-occlusive disease, congenital heart disease, and pulmonary arterial hypertension have provided a clue about amino acid residues that might play a role in the structure and function of the receptor.^25–29 Thus, these naturally occurring mutations identified in genetic diseases in the BMPR-II receptor are expected to serve as excellent models to elucidate the role of these structural elements in this three-finger toxin fold. Second, we focused on the role of the A-loop in BMPR-II receptor and in the closely related ActR-II receptor.^{11,13–16,24} Although the primary structures of BMPR-II and ActR-II receptors are different (Fig. 2A), the 3-dimentional folds are very similar as shown in the superimposed model of the two structures (Fig. 1). The present strategy is to study the A-loop deletion type II receptors and chimeric type II receptor molecules with substitution of the A-loop region (Fig. 2B). Third, we employed a structure-based computation algorithm, COREX/BEST, to probe the structural connectivity among various regions of the BMPR-II ECD in order to provide a rationale for the structure-function relationship of these mutants. Fourth, we employed the Autodock algorithm to identify potential contact residues in the ligand-BMPR-II ECD interface.

Figure 2. (A) Sequence alignment of human ActR-IIB receptor and BMPR-II receptor.

The secondary structure (β1 to β7) indicated on top of the sequence is adopted from Greenwald et al.²² Amino acid residues in ActR-IIB that have been shown by mutagenesis and analysis of crystal structures of ligand-receptor complexes to be important for ligand binding are shaded in light green.^11,16,18 Amino acid residues that have been shown by mutagenesis to be critical for ligand binding and are ligand specificity determinants are shaded in red and magenta, respectively.²⁴ (B) Schematics of chimera constructs of receptors ActR-IIB and BMPR-II. Numbers represent amino acid residue positions. BΔB is the BMPR-II with its A-loop deleted. BAB is the chimera of BMPR-II containing the ActR-IIB A-loop. ABA is the chimera of ActR-IIB containing the BMPR-II A-loop. AΔA is the ActR-IIB with its A-loop deleted.

Results of these studies revealed communications among the loops and other structural domains in these receptors. The structure-based computation algorithm revealed a network of connectivity among various structural elements. Such a network could serve as an important rationale for different receptors to differentiate the various ligands. The network may also serve as the molecular vehicle to implement the message imparted by the different ligands for further execution of downstream biological activities.

MATERIALS & METHODS

Materials

All reagents were of molecular biology grade. Activin was purchased from R&D Systems (Minneapolis, MN). BMP-7 and GDF-5 were provided by Stryker Biotech (Hopkinton, MA) and were dissolved in filtered deionized, distilled water. Oligonucleotide primers were synthesized by the Advanced Nucleic Acids Core Facility of The University of Texas Health Science Center at San Antonio. Thrombin and competent E. coli strain BL21 (DE3) cells were purchased from Novagen (Madison, WI). Ni-NTA resin was purchased from Qiagen (Valencia, CA). Superdex^™ 200 column and C18 column were purchased from Amersham (Piscataway, NJ). bis-ANS was purchased from Molecular Probe (Eugene, OR) and was dissolved in Dimethylformamide to make a 10 mM stock. Prior to use, the bis-ANS stock solution was diluted with RDB buffer (50 mM Tris-HCl, pH 7.5, 150 mM NaCl).

Cloning and mutagenesis

The human recombinant BMPR-II ECD sequence, corresponding to nucleotides +76 to +450 downstream from the translation start codon ATG (A is +1), was cloned into a modified pET32 (ΔKpn) vector as previously described.^24,30 The human recombinant ActR-IIB ECD sequence, corresponding to nucleotides +55 to 402 downstream from the translation start codon ATG (A is +1), was also cloned into the same vector. Desired mutants were constructed using the QuikChange site-directed mutagenesis kit (Stratagene, La Jolla, CA). Locations of these mutation sites are shown in Fig. 1. Schematics of chimera constructs of the ECD of the ActR-IIB and BMPR-II receptors are shown in Fig. 2B. Each mutant plasmid was characterized by restriction enzyme mapping and double-stranded DNA sequencing to be certain that no inadvertent changes had occurred.

Protein expression and purification

Both BMPR-II and ActR-II proteins were expressed in E. coli and purified to homogeneity as previously reported.^24,30 Briefly, E. coli BL21 (DE3) cells contained these DNA clones were grown and protein expression was induced by 0.2 mM IPTG for 3 hours. All ECD proteins were expressed as a fusion protein with a N-terminal TRX sequence and a C-terminal hexahistidine tag. The fusion protein TRX-ActR-IIB and TRX-BMPR-II ECD contained 262 and 271 amino acid residues with a theoretical molecular mass of 29 and 30 kDa, respectively. The fusion protein was first purified by Ni-NTA affinity chromatography and digested with thrombin. The resulting free receptor was separated from the TRX protein by Ni-NTA affinity chromatography followed by Superdex^™ 200 gel filtration column chromatography. The monomeric receptor ECD was further purified to homogeneity with a C18 reverse phase column. The purified ActR-IIB and BMPR-II ECD contained 133 and 142 amino acid residues with a theoretical molecular mass of 15 and 16 kDa, respectively. Protein concentrations were determined by the absorbance at 280 nm using Nanodrop 1000 (Thermo Scientific, Waltham, MA). Purity of the protein was analyzed on Tricine SDS gel. Protein samples were stored at −80°C until used.

Osteocalcin promoter-Luciferase (OC promoter-Luc) reporter activity assay

The in vivo biological activity of BMPR-II mutant receptor ECDs was tested. Fetal rat calvaria (FRC) cells were cultured in 24-w plate with αMEM+10%FBS and transfected with FuGENE 6 (Roche, Indianapolis, IN) according to the instruction provided by the manufacturer. Briefly, the OC promoter-Luc reporter DNA complex was prepared by incubating with FuGENE 6 for 30 min at room temperature. FRC cells were rinsed with HBSS, and mixed with 0.2 ml OPTI medium plus 8 Units of Hyaluronidase. The DNA complex was added drop-wise to the cells. After 5 h, fresh αMEM+10%FBS medium was added and incubated overnight. Transfected cells were treated with 6 nM BMP-7 in the absence or presence of 20 molar excess of wild type or mutant soluble BMPR-II receptor for 24 h. The OC promoter activity in the cell lysate was measured by the Luciferase activity using the Dual-Light System (Applied Biosystems, Foster City, CA).

Ligand binding activity by Surface Plasmon Resonance (SPR)

Ability of the different receptors to bind Activin, BMP-7, and GDF-5 ligands that were immobilized onto a CM5 sensor chip was measured by SPR as previously described.²⁴ Briefly, the CM5 sensor chip was equilibrated with HBS-EP buffer (10 mM HEPES, 150 mM NaCl, 3 mM EDTA, 0.005% P-20, pH 7.4) at a flow rate of 5 μl/min. A final surface of ~300 RU was used to perform flow-through assay of receptor ECD on the Biacore 3000 with control software (Piscataway, NJ). The receptor ECD of various concentrations was diluted with HBS-EP buffer, injected for 6 min at a flow rate of 5 μl/min, followed by a 6 min dissociation period. The surface was regenerated with 4 M Guanidine-HCl for 30 sec at a flow rate of 10 μl/min and buffer 30 sec at 100 μl/min. Flow cell #1 on the CM5 chip without immobilized ligand was used as a reference. The data from flow cell #1 (activated and blocked) without immobilized ligand were referenced and subtracted using Scrubber version 2.0a BioLogic software from The University of Utah. The results were plotted using Origin 7.0 (Northampton, MA). All data in triplicates were analyzed with reference subtracted.

Near and far UV CD spectroscopy and fluorescence determinations

CD measurements were performed on an AVIV model 60DS Spectropolarimeter (Lakewood, NJ) using published procedures.³¹ Fluorescence measurements were performed on a Horiba Jobin Yvon Fluoromax-3 Fluorometer (Edison, NJ) following published procedures.³² Briefly, fluorescence spectra from 310 to 600 nm were obtained with 295 nm excitation at 25°C. Binding constants were calculated following the Hill model using the equation F_obs = F_max [bis-ANS]ⁿ/(K′ + [bis-ANS]ⁿ), where F_obs and F_max represent the peak fluorescence intensity observed at a defined bis-ANS concentration and at a defined amount of protein, respectively. [bis-ANS] is the concentration of the dye. The Hill coefficient n in the above equation represents the number of classes of bis-ANS binding site on the protein, and is unity in the case of a single non-interactive site. A value that is greater than unity suggests positive cooperativity. K′ is the apparent binding constant.

Computation analysis

Stability calculation

The connectivity of different structural elements in BMPR-II was studied by computation in order to provide a rationale for the functional behavior of the BMPR-II mutants. The computation was accomplished by using the COREX/BEST algorithm in a server located at UTMB (http://best.utmb.edu/BEST/), where conformational ensembles were generated for the wild-type and mutant BMPR-II ECDs. The basic approach followed that employed for the study on DHFR in our laboratory.³³ Briefly, the residue stability constant, K_f,j, is the ratio of the summed probability of all states in the ensemble in which a particular residue j is in a folded conformation (ΣP_,f,j) to the summed probability of all states in which that residue is in an unfolded conformation (ΣP_nf,j,):

$${\kappa }_{f,j}=\frac{\sum P_{f,j}}{\sum P_{nf,j}}$$

The important feature of the residue stability constants is that they provide a measure of the local stability around each residue which can be experimentally verified by comparison to hydrogen exchange protection factors. Using the high resolution X-ray structure of BMPR-II receptor¹² as a template, an ensemble of BMPR-II EDC conformations was generated and the residue stability constants was calculated with a modified version of COREX that employs a Monte Carlo sampling strategy. High stability constants signify residues which are folded in the majority of highly probable states under native condition, whereas lower stability constants signify residues that are unfolded in many of those states.

Autodock Suite

Autodock, Autogrid, and Autodock Tools were used to reproduce and predict binding sites in complex of BMP-BMPR receptors and involved the following 4 steps:

• Step 1Preparing coordinate files for ligand and receptors. As control experiments to verify the ability of Autodock to capture the correct structures of BMP-BMPR receptors in the literature, the structures were downloaded from Protein Data Bank (PDB). In case of a PDB structure that contains the ligand-receptor complex, the ligand was translated and rotated with arbitrary vectors using PyMol. Thus the calculations were not biased to find the binding site. Then both the ligand and the receptor were saved in different ‘pdb’ files. This was done for control structures. For other structures, where ligand and receptor were in different files, it is necessary to verify that there were no clashes between ligand and receptor structures. In case they showed clashes, translation and rotation were carried out on ligand. In order to launch the calculations, the PDB files were modified to delete water molecules, hetero atoms and added hydrogen atoms. This was carried out by Autodock Tools [see references 34 and 35 for details]. Control calculations were conducted with the following five sets of complexes: (1). Activin Receptor II – BMP-7 (PDB: 1LX5); (2). BMP Receptor IA + Activin Receptor IIB – BMP-2 dimer (PDB: 2H62); (3). TGF-β Receptor Type II - TGF-β3 (PDB: 1KTZ); (4). TGF-β Receptor Type I + TGF-β Receptor Type II - TGF-β3 (PDB: 2PJY); (5). Activin Receptor IIB - Activin dimer (PDB: 1S4Y).
• Step 2Calculating atomic affinity potentials. Using the program Autogrid, pre-calculations of atomic affinity potential for each atom in ligand was conducted, i.e. the receptor was embedded in a virtual box to calculate the affinity between receptor and ligand surfaces using a probe atom.³⁵
• Step 3Docking ligand on receptor. With the program Autodock the final calculations and positions of ligand on receptor were conducted to generate a set of data based on interacting energy. This program provided a log file with coordinates and energy values for each configuration of the ligand-receptor. By default the log file contained the top ten possible conformations.
• Step 4Analyzing results. The log file obtained by previous steps was loaded into Autodock Tools and analyzed for geometric consistency of the ligand-receptor complex. If one of these structures were found in nature as dimer or trimer, the calculated conformation must keep space to these other subunits. Then, according to the ranking of energy values the best conformation was chosen as the final result.

RESULTS

I Biological activities of mutants

The bioactivity of these receptor proteins was measured using two different assays: (A) the cell-based osteocalcin promoter activity assay, and (B) the in vitro SPR studies.

A. Cell-based bioactivity assay

Naturally occurring single-amino-acid-substitution mutants

Three representative naturally occurring mutants Y40X, G47N, and S107P were selected for the present study. A key rationale for selecting these mutants for the present study was that the mutated amino acid residues are located in two different loop regions of the BMPR-II receptor structure with respect to the hydrophobic patch in Finger 2 which has persistently been identified in other receptor-ligand complexes as part of the ligand-receptor interface (Fig. 1). Furthermore, these mutations resulted in pulmonary arterial hypertension; thus, they most likely play important roles in the normal functions of BMPR-II.

The osteocalcin promoter activity assay determines the ability of the soluble BMPR-II ECD mutants to compete with the wild-type cell surface receptor for a fixed concentration of the BMP-7 ligand, resulting in a change in the available BMP-7 ligand to bind cell surface receptors. BMP-7, a BMPR-II ligand known to stimulate osteocalcin (OC) gene expression in a concentration-dependent manner in primary FRC cultures was used as a test ligand.^24,36 Accordingly, FRC cultures were transfected with an OC promoter -Luc reporter DNA construct followed by treatment with a fixed concentration of BMP-7 (6 nM) in the absence or presence of 20 molar excess of soluble wild-type or mutant BMPR-II ECD proteins. In agreement with published results, BMP-7 alone stimulated OC promoter activity by 14-fold beyond the empty plasmid control. For ease of comparison, the rest of the data obtained in the presence of the different soluble BMPR-II ECDs are expressed as relative activity with respect to that of the BMP-7 alone in the absence of any competing soluble receptor (as 1). As shown in Fig. 3A, in the presence of the wild-type (WT) BMPR-II ECD, the BMP-7-stimulated OC promoter activity was reduced to ~60% of the empty plasmid control¹. The observation indicates that the soluble wild-type BMPR-II ECD was capable of binding BMP-7, thus reducing the “effective” BMP-7 ligand to act on the cell-surface receptor and stimulating the OC-Luc reporter. In the presence of Y40Q and Y40R, the BMP-7-stimulated OC promoter activity was reduced to ~60 and ~70 %, indicating that the ECDs of these two mutant receptors were as effective in binding BMP-7 as the wild-type BMPR-II. On the other hand, in the presence of G47N, the BMP-7-stimulated OC promoter activity was not changed, indicating that the mutant was not effective in binding BMP-7 under the present experimental condition. In the presence of S107P, the BMP-7-stimulated OC promoter activity was reduced to ~80%. The present data, which are obtained by a cell-based assay, allow assessment of the relative activity of mutant receptors. Hence, the Y40Q, Y40R, and S107P mutants are effective competitors of the WT BMPR-II receptor in binding BMP-7 (6 nM) with an apparent affinity similar to that of the WT BMPR-II ECD, while G47N mutant is not. Fig. 3C shows the difference in stimulation of the OC promoter activity in the presence of the Y40, G47, and S107 receptor mutants.

Figure 3. Osteocalcin (OC) promoter-Luc reporter activity assay.

The bioactivity of the mutant ECDs of (A) BMPR-II natural mutants and (B) A-loop mutants to bind ligand was measured by their ability to bind BMP-7 ligand and stimulate OC promoter activity in FRC cells. Confluent FRC cultures were transfected with an OC-Luc reporter plasmid. After 5 h, transfected cells were recovered in αMEM complete medium with 10% FBS overnight, and then treated with 6 nM of BMP-7 in the absence or presence of 20-fold molar excess of wild-type (WT) or mutant ECDs. Luciferase reporter activity was measured as described in Materials and Methods. (C) Differences in stimulation of the OC-Luc reporter activity (Δ Stimulation = Stimulation_mutant − Stimulation_wt) in the presence of the different mutant receptors are calculated using the data shown in (A) and (B). Values represented the mean±SEM of three independent measurements.

Chimeric and deletion A-loop mutants

Our earlier published study with various single mutations in the A-loop did not show any perturbations of binding, leading to our conclusion that the A-loop of the BMPR-II receptor did not appear to play a significant role in ligand binding.²⁴ However, based on sequence alignment with other type II receptors of known structures but without the benefit of a holo BMPR-II structure, Mace et al.¹² suggested that 5 residues in the A-loop should be involved in ligand binding. Moreover, published data indicated that the A-loop in the ActR-II receptor and that of BMPR-II receptor may play different roles in ligand binding.^{11,13–16,24} The A-loop of BMPR-II receptor extends from S86 to E98 while that of ActR-II receptor is three residues shorter and extends from D62 to Q70 (Fig. 2A). We tested the validity of the results in the literature by constructing chimeric receptor molecules in which the A-loop was swapped between ActR-IIB receptor and BMPR-II receptor. The chimeric receptor ABA consists of the ActR-II sequence except substituting its A loop sequence with the BMPR-II A-loop sequence. The chimeric receptor BAB consists of the BMPR-II sequence except substituting its A loop with the ActR-IIB A-loop (Fig. 2B).

The ability of the soluble chimeric A-loop mutant receptors to bind to 6 nM of BMP-7 ligand thus reducing the ligand concentration to bind cell surface receptors was tested using FRC cells.^24,36 As shown in Fig. 3B, in the presence of the wild-type BMPR-II, the BMP-7-stimulated OC promoter activity was reduced to 60%, indicating that the soluble wild-type BMPR-II was capable of binding BMP-7, thus reducing the “effective” concentration of BMP-7 to act on the cell-surface receptor and stimulating the OC promoter. In the presence of BAB and BΔB BMPR-II ECD, the BMP-7-stimulated OC promoter activity also reduced to about 55%, indicating that these two mutants were as effective in binding to 6 nM BMP-7 as the wild-type BMPR-II ECD. On the other hand, in the presence of the wild-type ActR-II, the BMP-7-stimulated OC promoter activity was reduced only to about 70%, indicating that the soluble ActR-II was not as effective as WT BMPR-II ECD in binding to 6 nM BMP-7. In the presence of the soluble ABA ActR-II ECD, the BMP-7-stimulated OC promoter activity was reduced to 63%. Most interestingly, in the presence of the soluble AΔA ActR-II ECD, the BMP-7-stimulated OC promoter activity was not inhibited, suggesting that the affinity of soluble AΔA ActR-II ECD was low and was not capable of binding to 6 nM BMP-7. Taken together, these results indicate that the BMPR-II ECD is capable of binding BMP-7 ligand equally well with or without the A-loop; whereas in the ActR-II ECD, the presence of the A-loop from BMPR-II is essential. Without the A-loop from BMPR-II, the ActR-II ECD is not capable of binding or binds only with significantly weakened affinity for BMP-7. For ease of comparison, Fig. 3C shows the difference in stimulation of the OC promoter activity in the presence of the different A-loop receptor mutants.

B. Ligand binding energetics of mutants

Naturally occurring single-amino-acid-substitution mutants

SPR assays were also used to evaluate effects of the single amino acid substitution of BMPR-II receptor on ligand binding activity and specificity. Three ligands (Activin, BMP-7, and GDF-5) were selected for the study mainly because they are members from different BMPs subfamilies. Representative sensorgrams are included in the Fig. S1. It is important to note that the sensorgrams all returned completely to the baseline after buffer injection. These results indicated the complete reversibility of receptor-ligand interactions in this study.

Figs. 4A to 4C show the interaction between the immobilized Activin, BMP-7, and GDF-5, respectively, with varying concentrations of the wild-type and mutant receptors. The relative apparent equilibrium binding constants (K_{d app})² derived from the data on each mutant receptor for Activin, BMP-7, and GDF-5 are shown in Table 1.

Figure 4. Ligand binding activity of wild-type, Y40G, Y40P, Y40Q, Y40R, G47N, and S107P BMPR-II mutant receptors.

The bioactivity of the mutant BMPR-IIs was determined by SPR with (A) Activin, (B) BMP-7, and (C) GDF-5 immobilized on CM-5 chips. The response at equilibrium (R_eq) of each sensorgram was plotted against the concentration of each receptor protein. All data in triplicates were analyzed with reference subtracted. The points are actual experimental data and the curves are the results of best-fits of nonlinear regression analysis of the data.

TABLE 1.

Apparent K_d values for the interaction of wild type and BMPR-II variants with immobilized Activin, BMP-7, and GDF-5 as determined by SPR

Receptor	Activin (μM)	BMP-7 (μM)	GDF-5 (μM)
BMPR-II
Wild type	26 ± 1 (1.0)a	135 ± 8 (1.0)	221 ± 13 (1.0)
Y40G	54 ± 2 (2.1)	138 ± 8 (1.0)	123 ± 6 (0.6)
Y40P	70 ± 4 (2.7)	157 ± 7 (1.2)	200 ± 9 (0.9)
Y40Q	35 ± 2 (1.3)	129 ± 5 (1.0)	175 ± 9 (0.8)
Y40R	46 ± 2 (1.8)	117 ± 5 (0.9)	117 ± 5 (0.5)
G47N	30 ± 2 (1.1)	92 ± 5 (0.7)	136 ± 7 (0.6)
S107P	23 ± 1 (0.9)	43 ± 2 (0.3)	90 ± 5 (0.4)
Chimera BABb	5 ± 1 (0.2)	59 ± 2 (0.4)	197 ± 8 (0.9)
BΔBb	94 ± 5 (3.6)	27 ± 1 (0.2)	BDL
ActR-IIB
Wild type	6 ± 1 (1.0)	16 ± 1 (1.0)	BDL
Chimera ABAb	BDL	117 ± 5 (7.3)	BDL
AΔAb	BDL	30 ± 2 (1.9)	BDL

^aNormalized K_d values shown in parenthesis were obtained by normalization to the wild type values. Values represent the mean ± SEM of three independent determinations with two different ligand chips and three different receptor preparations. BDL, below detection limit (~ apparent K_d = 1300 μM).

^bBAB = chimeric receptor consisting of BMPR-II with the ActR-IIB A-loop. BΔB = BMPR-II receptor without its A-loop. ABA = chimeric receptor consisting of ActR-IIB with the BMPR-II A-loop. AΔA = ActR-IIB receptor without its A-loop.

Effects on ligand binding affinities of single amino acid substitution in BMPR-II were further evaluated by calculating changes in ΔG (ΔΔG) values, using the following equations: ΔG = −RT ln(K_d) and ΔΔG = ΔG _mutant − ΔG _{wild type} (Fig. 5A). A positive or negative value for ΔΔG means that the mutation weakens or enhances the affinity over that of the wild type ligand, respectively. The data show that all of the substitutions of amino acid residue Y40 differentially disrupted ligand binding affinities, namely, suppression of affinity for Activin but enhancement of affinity for GDF-5 with little or no effect for BMP-7. However, substitutions at G47 and S107 enhanced the affinity for only BMP-7 and GDF-5. Thus, residues Y40, G47, and S107 all seem to play a role in differentiating various ligands. The SPR results are in general agreement with the trend of the cell-based bioactivity assay data with respect to BMP-7, i.e. suppression of affinity by Y40 mutations as compared to the other mutations.

Figure 5. Changes in the binding free energy (ΔΔG) of different BMP ligands for (A) the different natural mutants and (B) the A-loop chimera mutants.

ΔΔG values were calculated from ΔΔG = ΔG_mutant − ΔG_wild-type where ΔG_wild-type and ΔG_mutant are the free energies for dissociation of ligand-receptor complexes with wild-type and mutant receptor, respectively. Values are calculated using the formula ΔG = −RTln(K_d). Values for Activin, BMP-7, and GDF-5 are shown in red, green, and blue columns, respectively for each mutant. An estimate for the upper detection limit is about 3 Kcal/mol. Some of the values are close to zero and thus are not visible in the graph.

Chimeric and deletion A-loop mutants

Interactions between immobilized Activin, BMP-7, and GDF-5 ligands with varying concentrations of the ActR-IIB, BMPR-II receptors and their respective chimeric and deletion mutants are shown in Figs. 6A to 6C, respectively. Representative sensorgrams are included in Fig. S2. The relative apparent equilibrium binding constants (K_{d app}) derived from the data on each receptor for these ligands are shown in Table 1. The binding results are expressed as ΔΔG, i.e. FG_{mu t} − FG_wt., as shown in Fig. 5B. The data show that substitution of the A-loop of ActR-IIB with that from BMPR-II (ABA) led to a positive ΔΔG value for BMP-7, i.e. replacement with the BMPR-II A-loop suppressed its affinity for BMP-7. Surprisingly, substitution of the A-loop of BMPR-II with that from ActR-IIB (BAB) led to a negative ΔΔG value, i.e. increased its affinity for BMP-7. These results imply that there is a guest-host relationship between the A-loop and the rest of the receptor molecule in defining the affinity for the ligand. In another word, there is communication between the A-loop and the rest of the receptor molecule. Similar conclusion can be derived from the results for the ligand Activin, which showed essentially no affinity for the chimeric receptor ABA but an enhancement for BAB. Activin and BMP-7 binding was strongly affected by deletion of the A-loop of BMPR-II (BΔB) with a ΔΔG value that was essentially equal in magnitude but opposite in direction, i.e. suppression in affinity for Activin but enhancement for BMP-7. Another interesting observation is that the deletion of the A-loop in the ActR-IIB leads to enhancement of affinity for BMP-7 ligand. Since the A-loop is more dynamic, as reflected by the B-factor in X-ray crystallographic data and the computation results shown in Fig. 7, the results of BMP-7 ligand binding to the A-loop deletion mutants imply that BMP-7 binding is favored by a receptor that is less dynamic, but the opposite conclusion might be derived for Activin. That GDF-5 binding was below the level of detection in these A-loop mutants suggests that GDF-5 binding to these two receptors is very sensitive to the A-loop.

Figure 6. Bioactivity of wild-type, chimera and A-loop deleted ActR-IIB and BMPR-II as determined by SPR with immobilized ligands.

The bioactivity of the different receptors was determined by SPR with (A) Activin, (B) BMP-7, and (C) GDF-5 immobilized on CM-5 chips. Experimental conditions were similar to those described in Fig. 4.

Figure 7. Computational analysis of the stability of human recombinant BMPR-II receptor (A) wild-type and (B) Y40G, Y40P, Y40Q, G47N, and S107P mutants.

Calculated residue stability constants (ln K_f) for the different BMPR-II molecules are on the Y-axis as a function of residue number (X-axis). The locations of secondary structural elements and the locations of the disulfide bonds are also indicated.

II Structure probing of mutants

A. Structural integrity of mutants

Naturally occurring single-amino-acid-substitution mutants

The ECDs of these mutant proteins were produced in E. coli and purified to homogeneity using a multiple chromatographic step procedure.^24,30 The wild-type BMPR-II receptor proteins produced using this protocol are properly folded as analyzed by NMR³⁰ and are biologically active.²⁴ Similarly, all mutants in this study are properly folded. In the present study we used several biochemical means to examine the folding of the mutant proteins, viz. circular dichroism (CD) and fluorescence spectroscopy.

Far UV analysis, which reveals secondary structure of the protein, showed that all ECDs of these mutant receptors exhibited similar negative ellipticity at 205 nm corresponding to β-sheets (Fig. S3). Near UV analysis which mostly monitor the environments surrounding aromatic amino acid residues, showed that all ECDs exhibited similar positive ellipticity at 250 nm. Very minor changes were detected in Y40G and Y40Q, and no significant structure changes were detected in the other mutants. Thus, we can conclude that the mutants are folded mainly like the wild-type protein with minor perturbations in the environment of the aromatic residues in the G and Q mutations in residue Y40.

Authors of the published X-ray crystallographic studies proposed that interactions between ActR-IIs receptor and BMP ligands are driven mainly by hydrophobic forces.^11,14 The proposed thermodynamic driving force is based on that many of the residues that lie within the hydrophobic patch have been shown to be critical for ligand binding (Fig. 1).^11,13 Based on these published results, we thus used bis-ANS fluorescence to examine whether any of the mutations under investigation affected the integrity of the hydrophobic patch of BMPR-II. This approach was shown previously to be able to monitor the structural integrity of the hydrophobic regions of the BMPR-II receptor.³² Accordingly, we determined the fluorescence intensity of bis-ANS upon interaction with the wild-type and the mutant receptors. The fluorescence intensity increased as a function of bis-ANS concentration with negligible inner filter effect (data not shown). The data were further analyzed using the Hill equation and yielded K′ values for bis-ANS binding (Table 2). Based on the K′ values, the property of the bis-ANS-binding hydrophobic region appeared to be affected marginally by the different substitutions of Y40 with a maximum of 2 to 3 fold changes in affinity. Taken together, these results suggest that the microenvironment of the hydrophobic patch is marginally modulated by mutations. If the hydrophobic patch in BMPR-II is indeed involved as the interface for ligand interaction, then the binding of ligands are expected to be modulated by mutations which seem to perturb the microenvironment of the hydrophobic patch.

TABLE 2.

K_d values of bis-ANS for BMPR-II

Receptor	K′a (μM)	K′b (μM)	IMax (arbitrary units)
BMPR-II
Wild type	19 ± 3 (1.0) c	19 ± 3 (1.0)	6 × 104 (1.0)
Y40G	6 ± 1 (0.3)	10 ± 1 (0.6)	6 × 104 (1.1)
Y40P	34 ± 4 (1.8)	30 ± 5 (1.6)	6 × 104 (1.1)
Y40Q	45 ± 10 (2.4)	36 ± 7 (1.9)	8 × 104 (1.4)
Y40R	21 ± 2 (1.1)	17 ± 2 (0.9)	5 × 104 (0.8)
G47N	55 ± 13 (2.9)	44 ± 8 (2.3)	9 × 104 (1.7)
S107P	19 ± 3 (1.0)	17 ± 5 (0.9)	4 × 104 (0.8)
BMPR-II
Wild type	10 ± 2 (1.0)	12 ± 2 (1.0)	9 × 104 (1.0)
Chimera (BAB)d	6 ± 1 (0.5)	6 ± 1 (0.5)	8 × 104 (0.9)
BΔBd	1 ± 0 (0.1)	2 ± 0 (0.1)	10 × 104 (1.1)
ActR-IIB
Wild type	3 ± 1 (1.0)	3 ± 1 (1.0)	2 × 104 (1.0)
Chimera (ABA)d	7 ± 1 (2.0)	13 ± 1 (3.8)	6 × 104 (3.0)
AΔAd	4 ± 1 (1.2)	5 ± 1 (1.4)	5 × 104 (2.5)

^aCalculated according to Hill Eq. F_obs = F_max [bis-ANS]ⁿ/(K′ + [bis-ANS]ⁿ)

^bCalculated from energy transfer from tryptophan in BMPR-II to bis-ANS data.

^cRelative values shown in parenthesis were calculated by normalization to the wild-type values.

^dSame abbreviations as listed in Table 1.

Additionally, we used fluorescence energy transfer (FRET) to examine structural integrity of the BMPR-II mutant ECD. Previously we showed presence of one bis-ANS binding site per BMPR-II ECD protein molecule.³² In the current study, we observed FRET as evident from the overlap of the tryptophan fluorescence emission spectrum of all the BMPR-II ECD (wild-type and the mutants) and the excitation spectrum of bis-ANS. The structural integrity of W85 and W87 in the BMPR-II receptor might be important, since they have been postulated by Mace et. al¹² to interact with ligand. The present observation is consistent with published results indicating the proximity of the tryptophan residues and the bis-ANS binding site.³² The observed energy transfer between bis-ANS and tryptophan is consistent with assigning the hydrophobic residues in Finger 2 as the primary site for bis-ANS binding. The two tryptophan residues reside at the edge of the hydrophobic patch located in the β-sheets which constitute Finger 2. The calculated K′ values by energy transfer are shown in Table 2. The K′ values determined by the titration method and FRET were in agreement. The findings also suggested that all mutant ECDs exhibited similar, if not identical, folding regarding to the bis-ANS reactive hydrophobic region and the tryptophan residues.

Chimeric and deletion A-loop mutants

The folding of these chimeric receptors was studied using the same fluorescence approaches, namely, the binding of bis-ANS and FRET. Analysis of the bis-ANS binding data using the Hill equation yielded K′ values as summarized in Table 2. The results suggest that the A-loop of BMPR-II can be replaced by that of ActR-II but with significant perturbation of the accessibility of the bis-ANS binding hydrophobic region. Deletion of the A-loop from BMPR-II also altered the microenvironment of the hydrophobic region. The A-loop of ActR-IIB also can be replaced by that of BMPR-II but with significant perturbation of the microenvironment of ActR-IIB molecule. In contrast to BMPR-II, deletion of the A-loop of ActR-IIB did not appear to perturb this region significantly.

In the FRET experiments, we observed fluorescence energy transfer from the overlap of the tryptophan fluorescence emission spectrum of the BMPR-II ECD wild-type as well as the chimera receptors and the excitation spectrum of bis-ANS. The calculated K′ values by energy transfer are shown in Table 2. The apparent dissociation constants determined by the fluorescence titration method and FRET for all chimeric receptors, except the ABA chimera, were consistent with each other. For the ABA chimera, the k′ value determined by FRET is almost twice as that determined by fluorescence titration, suggesting that the folding of the ABA molecule differed from that of the wild-type, and the spatial relationship between the tryptophan residues and the bis-ANS binding site was altered. However, folding of the AΔA mutant did not appear to be perturbed. The BAB chimera reacted with bis-ANS with a smaller K′ value, suggesting that the hydrophobic regions of the chimeric receptor are more exposed. Deletion of the A-loop from the BMPR-II resulted in a dramatic exposure of hydrophobic regions, suggesting that the A-loop in BMPR-II is an important structural determinant.

B. Structural stability of mutants

The ensembles-based COREX/BEST algorithm³³ was used to evaluate effects of mutation on protein stability of BMPR-II receptor (Figs. 7A and 7B). The analysis expresses the stability (Y-axis) of each residue in BMPR-II (X-axis); amino acid residues with lower ln k_f values indicate low stability. Fig. 7A shows the results of wild type BMPR-II. Regions around residues 40–46, 70–80, and 88–96 are less stable than regions around residues 66–70. The three amino acid residues of interest (Y40, G47, and S107) are located in two different regions with low stability. It is interesting to note that residues L69, I77, L79, V80, I88, V100, V101, and I108 in the hydrophobic patch in Finger 2 reside mostly in regions of low stability. Actually, residues 40 to 46, 71 to 76, and 102 to 112 constitute the tips of Fingers 1, 2, and 3, respectively. In addition, the A-loop which consists of residues 89 to 98 is also shown to be less stable. The computation analysis suggests that these regions are dynamic; a conclusion consistent with the distribution of B-factors in the X-ray structure.¹² Thus, the results derived from the COREX/BEST algorithm can capture the essence of dynamics of the structure defined by high resolution techniques such as X-ray crystallography.

Fig. 7B shows the effects of mutations on the stability profiles of BMPR-II ECD. The general stability profile of these mutants has not changed significantly, although the specific value of ln k_f has changed, particularly that of Y40G. The most interesting and significant observation is that the perturbation of stability not only occurs at the mutation site but also is propagated to other parts of the molecule, i.e. there is long range communication among these sites. For example, the stability of regions consisting of residues 65–70 and 95–125 is affected by Y40G mutation. However, residue Y40 is located in Finger 1 whereas residues 65–70 and 95–125 are in Finger 2 and 3, respectively.

III. Computation docking of ligands to BMPR-II ECD

To further elucidate the understanding of the results of our functional and solution structural studies at the molecular level, we conducted a computational docking exercise. In the control set of calculations we used the following five complexes with known structures at the atomic level: (1). Activin Receptor II – BMP-7 (PDB: 1LX5); (2). BMP-Receptor IA + Activin Receptor IIB – BMP-2 dimer (PDB: 2H62); (3). TGF–β Receptor Type II - TGF–β3 (PDB: 1KTZ); (4). TGF-β Receptor Type I + TGF-β Receptor Type II - TGF–β3 (PDB: 2PJY); and (5). Activin Receptor IIB - Activin dimer (PDB: 1S4Y). In all cases the computation results yielded a structure of the complex with a RMSD <0.1Å in reference to their crystal structures. Thus, we have been successful in capturing the atomic structures of the complexes with 1:1 stoichiometry, but it was difficult to obtain unique solutions with the dimeric ligands because the computation resulted in more than one structure for the ligand-receptor complex with very similar energetics. Since the systems under investigation are 1:1 complexes, we proceeded to explore the molecular structures of complexes of the BMPR-II ECD with Activin, BMP-7, and GDF-5 as ligands. The emphasis was on the general interfacial contacts within a radius of 4Å between the residues in the receptor and ligands. Results of this exercise are summarized in Table 3. In all cases there are two structures of equal probability as gauged by the energetic of complex formation. In Fig. 1, the structure of BMPR-II ECD is shown to highlight the 3 fingers and A-loop. The contacts of the two most probable complexes of Activin-BMPR-II are shown in Table 3. The interfacial interactions in the first probable complex involved BMPR-II receptor residues D43, L44, I46 in Finger 1, residues L69 of the hydrophobic core of Finger 2, and residues P106 and Y113 at the tip of Finger 3. The alternative complex of Activin-BMPR-II involved contacts with I46 in Finger1, K81 in Finger 2, P106 and Q109 in Finger 3, and the A-loop of the receptor. The interfacial contacts in the BMP-7-BMPR-II probable complexes involved Finger 3 and A-loop of the receptor, but the alternative probable BMP-7-BMPR-II complex inferred that contacts involve Fingers 1 and 3 and residues in the A-loop of the receptor (Table 3). The interesting exception was the lack of probable contacts with Finger 2 of the receptor. The contacts of the two probable complexes of GDF-5-BMPR-II are shown in Table 3. In one probable complex, no contacts were probable with Finger 1 although contacts were probable with residues in Fingers 2 and 3 and A-loop of the receptor. In the alternative complex of GDF-5-BMPR-II the pattern was different with probable contacts with all fingers and A-loop of the receptor.

TABLE 3.

Contact residues in BMPR-II ECD between Ligand Activin, BMP-7, or GDF-5 and BMPR-II receptor predicted by Autodock

Ligand-Receptor	Complex	Location of residues in structural domain
		Finger 1	Finger 2	Finger 3	A-loop
Activin-BMPR-II	I	Y40, D43 L44, I46	L69	P106, Y113	W85, S86, D90 P91, Q92
	II	I46	K81	P106, Q109	W85, Q92
BMP-7-BMPR-II	I	Y40	Y67, K81 C84	P105, P106, Q109 N110, Y113, F115	W85, D90, P91
	II	L44, I46		P105, P106, Q109	W85, S86, H87 D90, P91, Q92
GDF-5-BMPR-II	I		C84	P105, P106, I108, Q109	W85, H87 D90, P91, Q92
	II	D38, Y40 I46	Y67, L69, E71 V80, K81, C84	Y113	W85, H87, I88 G89, D90, P91

Residues in bold denote the residues common to all three predicted receptor-ligand complexes.

Another interesting result derived from this computation exercise is that residues Y40 and I46 in Finger 1 and W85 in the A-loop of the receptor were identified to be the common residues in contact with all three ligands. Consistent with the computation results are others and our experimental data showing that in this current study the effects of mutation of Y40 on the ability of BMPR-II receptor to distinguish ligands Activin, BMP-7, and GDF-5. We and others previously also demonstrated the effect of mutating W85 on the affinity of binding to BMP-2, BMP-7, and GDF-5.^8,24

Fig. 8 summarizes the results of this docking exercise. The three residues in Fingers 1 and 2 of the receptor that make contact with all three ligands are shown as spheres while the other residues in contact with at least one ligand is shown as dots. It is interesting to note that all three fingers have residues that make contacts with ligands and they all reside on one face, the concave side, of BMPR-II receptor.

Figure 8. Summary of important residues involved in interfacial interaction between the BMPR-II receptor and ligands.

The three residues in Fingers 1 and 2 that make contact with all three ligands are shown as spheres while the other residues in contact with at least one ligand is shown as dots. The color codes for the residues are: blue spheres for residues Y40 and I46, red sphere for W85.

DISCUSSION

In the present study, we aimed to elucidate the molecular mechanism used by the two type II BMP receptors in defining specificity and/or affinity in ligand binding. Although the ECD of BMPR-II receptor has only 24% sequence identical to that of its closely related ActR-IIB receptor (Fig. 2A), they have essentially identical folding (Fig. 1). One may expect that these ECDs to have similar functional and structural behaviors. In the present study, we tested the validity of this assumption.

Effects of natural mutations

This study focused on two different structural elements, namely, Fingers 1 and 3. Specifically, mutants involving residues Y40, G47, and S107 were studied since previously they were detected in patients suffering from pulmonary arterial hypertension. These amino acids are not among the residues predicted to be part of the binding interface as a consequence of analogy analysis with other known structures of ligand-receptor complexes or sequence alignment. Residues in the ActR-IIB receptor that are in contact with BMP-7 are shown in Fig. 1¹¹ and do not include Y40, G47, and P107.

If indeed these residues in BMPR-II were not in contact with the ligands but mutations of these residues affect ligand binding as shown in the present study, then one interpretation is that the effect of perturbations resulted from the mutation must be the consequence of a long-range communication from these residues to the binding interface, which involves at least the hydrophobic patch identified in most ligand-receptor complexes. The current consensus in the literature is that intra-molecular long range communication is likely because proteins exist as an ensemble of microstates in dynamic equilibrium. We hypothesize that BMPR-II most likely can be represented by such an ensemble of dynamic states. The validity of this hypothesis is supported by the ability of the COREX/BEST algorithm to capture the dynamic nature of the BMPR-II structure that is defined by the atomic structure of the receptor determined by high resolution X-ray crystallography, as shown in Fig. 7. Namely, the high dynamic nature of the finger tips and the A-loop is highly coincident of the structural elements showing low lnk values (low stability) in the computation results. The computation results are consistent with experimental data such as high B-factor in the X-ray data.²⁴ Thus, COREX/BEST is capable of capturing the essence of BMPR-II receptor as a dynamic entity.

A direct corollary to a dynamic entity is that each molecule exists simultaneously in multiple microstates, the distribution of which is governed by equilibrium processes which can be modulated by environmental factors such as pH, temperature, ligand binding or mutation. Results from the present study provide the expected data. For example, as observed in Fig. 7, the effect of mutation of Y40 is not localized to the tip of Finger 1 but also tips of Fingers 2 and 3. The regions, whose stability is predicted to be the most perturbed, include residues 40–50, 60–70, 95–125. Furthermore, the more stable regions around residues 65, 100, and 120 are further stabilized by the Y40 mutation. The present analysis also shows that the expected binding interface in BMPR-II, assuming BMPR-II and ActR-IIB use the same surface area for ligand binding, includes approximately residues 60–80, 85, 91–96 and several residues in the C-terminus. Thus, the stability of the predicted residues is affected by the mutation, and these residues are within the predicted regions of the ligand binding interface. In addition, different amino acid side-chain substitution at Y40 leads to different magnitudes of perturbations of different parts of BMPR-II receptor. The magnitude of perturbation decreases in the order of G>P>Q. It is gratifying to note that the binding affinity of bis-ANS for the hydrophobic patch is affected by mutations at Y40 and the magnitude of change seems to be dependent on the nature of the substituted side chain. Thus, it is not surprising that different mutations of Y40 alone can lead to a diverse set of functional perturbations, e.g. enhancing the affinity for GDF-5 but suppressing that of Activin. Hence, in this study the computation results of the COREX/BEST algorithm apparently can capture the essence of the functional perturbation as a consequence of mutation. This observation is consistent with the fact that the COREX/BEST algorithm has been employed successfully to capture many phenomena which require long range intra-molecular communications.^33,37–40

The current consensus opinion in the literature is that a hydrophobic patch in the two type II receptors ActR-IIB and BMPR-II is involved in ligands binding. The location of this patch resides in Finger 2 (Fig. 1). Based on crystallographic data of the apo BMPR-II receptor, Mace et al. proposed that Y67, W85, and F115 as the residues in the hydrophobic patch that are in contact with ligands.¹² They further proposed that K81, S86, E93, and Y113 are also involved. It is interesting to note that the stability of residue 67 and 115 is significantly perturbed by a Y40G mutation. This significant change in stability is correlated to the most significant changes in the binding affinity of ligands BMP-7 and GDF-5 in opposite directions as shown in Fig. 5A. Furthermore, the same Y40G mutation leads to the largest change in the binding affinity for bis-ANS (Table 2). Thus, there is an apparent correlation among perturbation of residue stability, environmental integrity of the hydrophobic patch and ligand binding affinity.

On the other hand, our experimental data showed that a G47N mutation impacts on ligand binding by lowering the affinity for all ligands, though to different extents depending on the identity of the ligand. In Fig. 7B the COREX/BEST computation results showed minor perturbation on the stability of a smaller number of residues in BMPR-II receptor. The same correlation between functional and structural stability can be concluded with the S107P mutation. Thus, our results seem to indicate that Y40, G47, and S107 play different roles to differentiate ligands, and alterations of these residues are detrimental to ligand binding.

Another very interesting observation is that the affected regions seem to be anchored by the 5 disulfide bonds in the ECD of the BMPR-II receptor; most of them seem to be strategically located in junctions between a stable and a highly dynamic structural element. This finding implies that one function of the disulfide bonds might be to transmit the signal to specific regions of the BMPR-II molecule in addition to serve a structural role.

Structural relationship between the A-loop and the hydrophobic patch

Although both BMPR-II and ActR-IIB receptors contain an A-loop structure, results from previous published studies from different laboratories are incongruent on the role of the A-loop in binding of BMPs.^{11,13–16,24} The current results of functional and computational studies showing a guest-host relationship between the ECD and A-loop suggest the presence of intra-molecular long-range communication. Such communication can provide a rationale for the discrepancies reported in the literature.

The A loop in the BMPR-II receptor is two amino acid residues longer than that in the ActR-IIB. The current study showed that deletion of the A-loop from the ActR-IIB receptor (AΔA) did not appear to affect its global fold, but substitution of the ActR-IIB A-loop with the BMPR-II A-loop (ABA) resulted in a significant alteration in the accessibility of the hydrophobic site by bis-ANS. According the structure model, the distance between the A-loop and the hydrophobic patch is fairly substantial, thus, the change in the bis-ANS binding site must be attributed to a long range communication between these two structural elements within the receptor. Furthermore, the observed perturbations in the hydrophobic patch are important because this region has been shown as part of the binding interface in other ligand-receptor complexes. The finding of the present study showing that the binding of ligands is affected is thus consistent. Thus, results obtained from mutation of a single amino acid residue and from A-loop substitution/deletion lead to the same correlation between the structural integrity of the hydrophobic patch and ligand binding.

Function of A-loop

Our results on the ActR-IIB and BMPR-II chimera indicate that the A-loop in the two ECD proteins play complex and distinct functional roles. Specifically, our observations suggest that the BMPR-II receptor is more tolerant and is not that sensitive to the nature of the A-loop or whether it is present. As shown in Table 1, the wild-type BMPR-II receptor, the BFB deletion mutant receptor, and the chimeric BAB receptor can all bind ligands. The data further suggest that both the BMPR-II receptor and the BMP-7 ligand are quite promiscuous in their binding behavior. Specifically, deletion of the A-loop essentially eliminated the ability to bind ligands Activin and GDF-5 but enhanced the affinity for ligand BMP-7. Replacement of the A-loop with the shorter loop from the ActR-IIB receptor renders the chimeric receptor a better Activin binder. The observation suggests that the A-loop is dynamic and a less dynamic A-loop is better for Activin binding and the change in the dynamic nature of the A-loop is not important for BMP-7 binding. However, computational docking analysis of the BMP-7-BMPR-II complex shows a great deal of contacts between the receptor and the ligand; thus, one might expect a decrease in ligand binding affinity due to the decrease in contact as a result of the shorter A-loop from ActR-IIB receptor. Yet the binding data for the chimeric receptor showed that it binds stronger. One plausible explanation of these findings is that the A-loop in BMPR-II is very dynamic and thus does not affect BMP-7 binding affinity. Consistent with this supposition are the results of published results indicating that mutations of several residues (viz. S86, H87, D90, and Q92) within the A-loop did not affect binding of BMP-2, BMP-7, and GDF-5 significantly.²⁴ On the contrary, mutations of L61 and D63 (which corresponds to S86 and H87 in BMPR-II, respectively) in the A-loop of ActR-IIB significantly disrupted BMP-2 and BMP-7 binding.¹⁴ Studies of the crystal structure of the BMP-7-ActR-IIB complex also revealed that several residues (L61, D63, I64, and N65) in the A-loop region of ActR-II receptor interact with BMP-7 ligand.²² These findings suggest that the A-loop in the two related receptors might play distinct role.

By comparison, the ActR-IIB receptor is not tolerant at all. Only the wild type ActR-IIB can bind both Activin and BMP-7. A replacement of the A-loop from the ActR-IIB receptor with that of the BMPR-II receptor (ABA) eliminated its ability to bind ligands Activin but still showed a demonstrable affinity for ligand BMP-7. A deletion of the A-loop renders the ActR-IIB receptor (AΔA) totally inactive towards ligands Activin and GDF-5 but shows a strong affinity towards ligand BMP-7.

Results on the chimeric receptors reveal two novel observations, namely, the role of the A-loop of the receptor is dependent on the three-finger fold of the host receptor and both the BMPR-II receptor and BMP-7 ligand have less stringent requirement for binding. In the context of ActR-IIB, the A-loop is critical for Activin and GDF-5 binding, but contributes to BMP-7 binding.

Summary statement

In summary, the present study has identified regions of the BMPR-II molecule in defining ligand specificity. The different regions play different roles in ligand binding. An integration of these results on functional energetics and protein structures identify, for the first time, an intra-domain communication network within the BMPR-II molecule. Identification of such a communication network is not only important for understanding the working of the receptor but also for protein engineering and drug design. Furthermore, we are currently investigating one of the more intriguing and fundamental issues on the mechanism and structural entity that define promiscuity and specificity of both the ligand and receptor. Are the information imbedded in the folds? Can the 3-finger fold be responsible for promiscuity? What is the mechanism that enables amino acids to exert dominant effect on the A-loop? What is the role of the disulfide bonds in the natural mutants? These are the issues that we are actively pursuing.

ABBREVIATIONS

BMP

Bone Morphorgenetic Protein

BMPR-II

BMP receptor-II

bis-ANS

4, 4′-dianilino-1, 1′-bisnaphthyl-5, 5′ disulfonic acid

SPR

Surface Plasmon Resonance