Demonstration of a Direct Interaction between β2-Adrenergic Receptor and Insulin Receptor by BRET and Bioinformatics

Glucose metabolism is under the cooperative regulation of both insulin receptor (IR) and β2-adrenergic receptor (β2AR), which represent the receptor tyrosine kinases (RTKs) and seven transmembrane receptors (7TMRs), respectively. Studies demonstrating cross-talk between these two receptors and their endogenous coexpression have suggested their possible interactions. To evaluate the effect of IR and prospective heteromerization on β2AR properties, we showed that IR coexpression had no effect on the ligand binding properties of β2AR; however, IR reduced β2AR surface expression and accelerated its internalization. Additionally, both receptors displayed a similar distribution pattern with a high degree of colocalization. To test the possible direct interaction between β2AR and IR, we employed quantitative BRET2 saturation and competition assays. Saturation assay data suggested constitutive β2AR and IR homo- and heteromerization. Calculated acceptor/donor (AD50) values as a measure of the relative affinity for homo- and heteromer formation differed among the heteromers that could not be explained by a simple dimer model. In heterologous competition assays, a transient increase in the BRET2 signal with a subsequent hyperbolical decrease was observed, suggesting higher-order heteromer formation. To complement the BRET2 data, we employed the informational spectrum method (ISM), a virtual spectroscopy method to investigate protein-protein interactions. Computational peptide scanning of β2AR and IR identified intracellular domains encompassing residues at the end of the 7th TM domain and C-terminal tail of β2AR and a cytoplasmic part of the IR β chain as prospective interaction domains. ISM further suggested a high probability of heteromer formation and homodimers as basic units engaged in heteromerization. In summary, our data suggest direct interaction and higher-order β2AR:IR oligomer formation, likely comprising heteromers of homodimers.


Introduction
The functional interplay between different classes of receptors represents a means of fine-tuning the control of cellular functions that could be fundamental for understanding pathologic conditions and responses to therapeutic agents that interact with cellsurface receptors. Originally, 7 transmembrane receptors (7TMRs) and receptor tyrosine kinases (RTKs), together with their respective downstream effectors, were thought to represent distinct and linear signaling units; however, recent studies have provided evidence for functional crosstalk between these different types of receptors [1].
Glucose metabolism is under the cooperative regulation of the RTK insulin receptor (IR) and b 2 -adrenergic receptor (b 2 AR), a representative 7TMR. The anabolic action of insulin promotes glycogen synthesis, glucose uptake in skeletal muscle and lipid storage. By contrast, catecholamines act in opposition to insulin, stimulating glycogen breakdown, gluconeogenesis, and lipolysis [2]. They play an important role in counter-regulation of insulininduced hypoglycemia. The ability of insulin to counter regulate catecholamine action at the tissue/cellular level is crucial for the ''tight'' regulation of serum glucose levels, and interaction between IR and b 2 AR could be a required action in this process.
At the molecular level, the ability of insulin to counter regulate b 2 ARs can be exerted through insulin-stimulated phosphorylation of b 2 AR and its subsequent internalization. An early in vitro study demonstrated that, in the presence of insulin, IR catalyzes the phosphorylation of b 2 AR predominantly at residues located in the cytoplasmic tail of b 2 AR-i.e., Tyr350/Tyr354, and Tyr364 [3]-a finding that was consistent with the hypothesis that IR directly interacts with and phosphorylates b 2 AR. Phosphorylated b 2 AR Tyr350 creates a phosphotyrosine Src homology 2 (SH2) binding domain, which effectively precludes b 2 AR from coupling to its cognate G-protein (Ga S ) and enables binding of the adaptor Grb2, the p85 regulatory subunit of phosphoinositide 3-kinase (PI3K) or the GTPase dynamin; all of the latter molecules are involved in 7TMR trafficking [4,5,6]. In addition, insulin-induced activation of MAPKs, ERK 1 and ERK 2, is potentiated by b 2 AR via a mechanism that requires the integrity of Tyr350 [4]. Thus, the functional cross-talk between b 2 AR and IR may be to fine tune the signals from multiple receptor signaling pathways. Recently, Fu et al. [7] described an IR:b 2 AR complex in the mouse heart, indicating that cross-talk between these two receptors could provide a molecular basis for the pathophysiology of metabolic and cardiovascular dysfunction under insulin-resistant states. b 2 AR and IR are also endogenously coexpressed in other cell types-i.e., pancreatic beta cells, adipocytes, liver cells, skeletal muscle cells and astrocytes [8,9,10]. Therefore, the formation of heteromeric complexes consisting of these two receptors may be plausible.
In contrast to 7TMR-based BRET studies, BRET has been less exhaustively employed for RTK-based studies, and an overview of BRET studies associated with RTKs was recently reviewed by Issad et al. [26] and Siddiqui et al. [27]. RTKs often employ dimerization as a key factor in their activation; in fact, members of the IR subfamily are pre-dimerized by disulfide bonds (reviewed by De Meyts [28]). In IR studies, BRET was used to monitor the activation state of IR [29,30,31] and its interactions with different intracellular binding partners, including protein tyrosine phosphatase-1B [32], Grb14 [33] and IRS1, IRS4 and Shc [34]. IR can also form hybrid heterotetramers with the insulin-like growth factor-1 (IGF-1) receptor, exemplifying that IR can comprise heteromers that are products of independent genes [35].
In addition to interfamily receptor heteromerization-i.e., 7TMR:7TMR and RTK:RTK pairs-a physiologically relevant, direct physical interaction between the 7TMR adenosine A 2A receptor (A 2A -AR) and RTK fibroblast growth factor receptor (FGFR) was demonstrated [36]. Evidence for other 7TMR:RTK complexes was also provided [37,38,39]. Computer-assisted analysis suggested that the N-and C-termini, as well as the third intracellular loop (ICL3) of 7TMRs show a propensity toward unstructured conformation, and could potentially interact with domains of other receptors/receptor-interacting proteins (reviewed in [40]).
To test whether b 2 AR and IR are prospective interaction partners, we employed the following methods: i) radioligand binding and enzyme-linked immunosorbent assay (ELISA) to determine the receptor-binding properties and surface expression, as well as to quantify receptor internalization processes; ii) confocal microscopy to detect possible colocalization between b 2 AR and IR; iii) BRET 2 proximity assays consistent with the direct interaction between b 2 AR and IR; and iv) the informational spectrum method (ISM), a virtual spectroscopy method to investigate protein-protein interactions [41,42], to identify important informational characteristics of the interaction between b 2 AR and IR and to predict domains potentially involved in their interaction. Prior studies have proved this bioinformatics approach to be effective in identifying protein-protein interaction partners and some of them were also confirmed experimentally [43,44].

Materials
Molecular biology reagents, as well as the tissue culture media and reagents, were purchased from Sigma-Aldrich (St. Louis, MO, USA) and Gibco Invitrogen Corporation (Breda, the Netherlands), unless otherwise specified. [ 125 I]-Iodopindolol was obtained from PerkinElmer (Boston, MA, USA). The ligands pindolol and isoproterenol were purchased from Sigma-Aldrich, and recombinant human insulin S100 was from Novo Nordisk A/S (Bagsvaerd, Denmark). Coelenterazine 400a was purchased from Biotrend Chemikalien GmbH (Köln, Germany). Anti-hemagglutinin (HA) high-affinity rat monoclonal antibodies were from Roche (Basel, Switzerland). Anti-rat horseradish peroxidase (HRP)-conjugated antibodies and anti-rat TRITC-conjugated antibodies were from Sigma-Aldrich.

Receptor fusion constructs
Human HA-tagged b 2 AR (HA-b 2 AR) cDNA in the vector pcDNA3.1 was purchased from Missouri S&T cDNA Resource Center (University of Missouri-Rolla, USA). HA-b 2 AR C-terminally tagged with Renilla luciferase 8 (RLuc8) (b 2 AR-RLuc8) was made using standard molecular biology techniques and was verified by sequencing. WT human IR isoform A that lacks exon 11 and IR C-terminally tagged with the green fluorescent protein variant 2 (IR-GFP 2 ) were generated and verified at Novo Nordisk A/S. C-terminally RLuc8-tagged IR (IR-RLuc8), C-terminally GFP 2 -tagged b 2 AR (b 2 AR-GFP 2 ) and the membrane-inserted GFP 2 -tagged construct (GFP 2 -17aa) were the same as described previously [24,34,45]. All of the generated cDNA clones were inserted into the expression vector pcDNA3.1(+).

Cell culture and transfection
HEK-293 cells (European Collection of Animal Cell Cultures, Salisbury, UK) were routinely maintained and passaged in Dulbecco's modified Eagle's Medium (DMEM) supplemented with 10% (v/v) heat-inactivated fetal calf serum, 2 mM Glutamax-I, penicillin (100 U/mL) and streptomycin (100 mg/mL) at 37uC in a humidified atmosphere of 5% (v/v) CO 2 . For transient transfection, HEK-293 cells were seeded at a density of ,1610 6 cells per 60-mm tissue culture dish or at a density of 4610 6 cells per 75-cm 2 flask and transfections were performed the following day using Lipofectamine 2000 according to the manufacturer's instructions. Cells were harvested 48 h after transfection, the cell number was determined using a cover-slipped hemocytometer, and the cells were resuspended in Dulbecco's PBS supplemented with Ca 2+ /Mg 2+ , 1 g/L glucose and 36 mg/L sodium pyruvate to a density of 1610 6 cells/mL, unless otherwise stated.

Luminescence and fluorescence measurements
The expression levels of RLuc8-and GFP 2 -tagged receptor constructs were monitored by total luminescence and fluorescence measurements as previously described [46]. For luminescence measurements, ,2610 5 of resuspended cells were distributed in 96-well microplates (white Optiplate; Packard BioScience, Meriden, CT, USA). After the addition of coelenterazine 400a to a final concentration of 5 mM, total luminescence was measured using a TriStar LB 942 microplate reader (Berthold Technologies, Bad Wildbad, Germany). For fluorescence measurements, ,2610 5 of resuspended cells from the same transfections were plated in black 96-well FIA plates (Greiner Bio-One, Frickenhausen, Germany). Total fluorescence was measured using the TriStar LB 942 microplate reader with an excitation filter at 380 nm and an emission filter at 515 nm. Background values obtained with mock-transfected HEK-293 cells were subtracted in both measurements, and the mean values of triplicate wells/sample were then calculated.

Receptor Binding Assay
To establish a relationship between the luminescence/fluorescence signals generated by RLuc8 and GFP 2 -tagged b 2 ARs and cell-surface receptor number, radioligand binding assays were carried out on whole cells as previously described [47]. After transfections with increasing amounts of cDNA (0.01 to 2 mg) for either b 2 AR-RLuc8 or b 2 AR-GFP 2 , cells were plated onto 24-well plates at a density of ,1610 5 cells per well. An aliquot of cells (,5610 5 ) from each transfection was also transferred into a 60mm dish for total luminescence/fluorescence measurements as described above. For the whole-cell radioligand binding assay, cells were washed once with assay buffer (HEPES-modified DMEM with 0.1% bovine serum albumin (BSA)) before being incubated with the b 2 AR antagonist [ 125 I]-iodopindolol (30,000 cpm/well) and increasing concentrations of unlabeled pindolol (10 -12 to 10 25 M final concentration) in assay buffer for 2 h at 4uC. Cells were then washed 3 times with ice-cold PBS and solubilized with 0.2 M NaOH and 1% sodium dodecyl sulfate (SDS) solution, and then the radioactivity levels were determined using a c counter (LKB Wallac, Turku, Finland). Determination of the radioactivity levels was performed in triplicate. Binding parameters were determined from displacement curves generated by a sigmoidal dose-response curve fit (GraphPad Prism 5.0). Receptor density (B max ), expressed as the receptor number per cell, was calculated as previously described by Ramsay et al. [48]. Whole-cell radioligand binding assays were also performed with cells expressing b 2 AR alone or in combination with IR. Displacement curves were generated using [ 125 I]-iodopindolol (30,000 cpm/well) and increasing concentrations of isoproterenol or isoproterenol and insulin, and IC 50 values were then determined using GraphPad Prism 5.0.

ELISA
ELISA assays for the measurement of surface-expressed HAb 2 AR and quantification of receptor internalization were performed as described previously [49]. Briefly, HEK-293 cells were transiently transfected with either b 2 AR-RLuc8 alone or b 2 AR-RLuc8 and IR-GFP 2 . One microgram of each receptor was used, and the total amount of cDNA used for transfection was kept uniform by adding empty pcDNA3.1 vector. After transfection, cells were seeded at a density of ,1610 5 cells per well in a 24-well plate. An aliquot of cells (,5610 5 ) from each transfection was also transferred into a 60-mm dish for total luminescence/fluorescence

Receptor internalization assay
The b 2 AR internalization assay was performed as described previously [47]. Briefly, HEK-293 cells were transiently transfected with either b 2 AR alone or b 2 AR and IR. One microgram of each receptor was used, and the total amount of cDNA used for transfection was kept uniform by adding empty pcDNA3.1 vector. After 2 days, cells were first subjected to a 2-h starvation period in assay medium (HEPES-modified DMEM with 0.01% BSA) before being incubated with 10 mM isoproterenol or a combination of 10 mM isoproterenol and 0.1 mM insulin S100 in assay medium for time intervals ranging from 5 min to 60 min at 37uC. Cells were then placed on ice, washed 3 times with ice-cold PBS and incubated for 2 h with [ 125 I]-iodopindolol in the presence or absence of 10 mM pindolol at 4uC. Specific binding in each fraction was determined as the difference between radiolabeled ligand detected in the presence and absence of 10 mM pindolol.
Receptor sequestration was then defined as the decrease in specific [ 125 I]-iodopindolol binding compared with the total binding obtained in untreated cells. The amount of internalized receptors as a function of time was fitted using a one-site binding (hyperbola) curve fit (GraphPad Prism 5.0) to estimate the half-time of internalization (t 1/2 ). All of the time points were performed in triplicate.

Confocal microscopy
HEK-293 cells were transiently transfected with constructs encoding HA-tagged and/or GFP 2 -tagged receptor constructs, trypsinized, and plated on poly-L-lysine-coated glass coverslips in complete DMEM. After 48 h, cells were treated as required with either 10 mM isoproterenol, 0.1 mM insulin S100 or the combination of both in HEPES-modified DMEM for 10 min at 37uC. Upon treatment, cells were washed with ice-cold PBS and fixed with 4% paraformaldehyde for 20 min at 4uC. Following washing (3 times in PBS), cells were permeabilized with PBS containing 0.01% Triton X-100 for 20 min. To reduce the nonspecific binding, cells transfected with the HA-tagged receptors were incubated in blocking solution (PBS containing 1% BSA) for 30 min. Subsequently, cells were incubated with a 1:100 dilution of primary rat anti-HA antibodies in PBS overnight at 4uC. Following washing (3 times in PBS), cells were incubated with a 1:50 dilution of secondary rabbit anti-rat TRITC-conjugated antibodies for 60 min at room temperature in the dark. Cells were then mounted using an anti-fading ProLong Gold reagent (Molecular Probes, the Netherlands), sealed and examined under an oil immersion objective (Planapo 406, N.A. = 1.25) using a Leica multispectral confocal laser microscope (Leica TCS NT, Heidelberg, Germany). The sequential detection of GFP 2 -and TRITC-stained receptors was achieved using excitation laser lines at 488 nm (argon) and 543 nm (helium-neon), respectively. The fluorescence from the channels was collected sequentially, and images were produced using an 8-fold frame averaging a resolution of 102461024 pixels. Optical sections (1.0 mm) were acquired, and representative sections corresponding to the middle of the cells were presented using Adobe Photoshop 7.0 computer software.

BRET 2 saturation and competition assays
To derive BRET 2 saturation curves, HEK-293 cells were transiently cotransfected with constant amounts of the constructs encoding RLuc8-tagged receptors together with increasing amounts of the constructs encoding GFP 2 -tagged receptors. For BRET 2 competition assays, HEK-293 cells were cotransfected with constant amounts of RLuc8-and GFP 2 -tagged receptor constructs and with increasing amounts of untagged receptors. BRET 2 assays were performed as described previously [24,46,47]. Briefly, 180 ml of resuspended cells containing ,2610 5 cells was distributed in 96-well microplates (white Optiplate; Packard BioScience, Meriden, CT, USA). After the addition of coelenterazine 400a to a final concentration of 5 mM using an injector, readings were collected (TriStar LB 942 microplate reader, Berthold Technologies, Bad Wildbad, Germany). Signals at 410 nm (RLuc8 luminescence signal) and 515 nm (emission of light from excited GFP 2 ) were measured sequentially, and 515/ 410 ratios (BRET 2 signal) were calculated. The results were expressed in milliBRET units (mBU); BRET ratio 6 1000. The expression levels of RLuc8-and GFP 2 -tagged constructs for each experiment were assessed by total luminescence and fluorescence measurements as described above. Determinations were made in triplicates. Figure 2. Visualization of b 2 AR and IR cellular localization by confocal microscopy. IR-GFP 2 and HA-b 2 AR-RLuc8 cellular localization is shown in untreated (control) cells (upper panels) and cells concomitantly treated with isoproterenol (10 mM) and insulin (0.1 mM) for 10 min at 37uC (lower panels). The green color indicates IR-GFP 2 ; red indicates HA-b 2 AR-RLuc8; yellow/orange is the overlapping region indicating colocalization of IR-GFP 2 and HA-b 2 AR-RLuc8. Note that both receptors have comparable localization in untreated and agonist-stimulated cells and that they exhibit a high degree of colocalization. However, in agonist-stimulated cells a proportion of intracellular receptors did not colocalize. Objective 406and zoom factor 4 apply for all images. doi:10.1371/journal.pone.0112664.g002 BRET 2 assay data evaluation The BRET 2 values were fitted using the following equation for dimers: BRET = BRET max (1/(1+AD 50 /X)), where X is the ratio of acceptor (A; Receptor-GFP 2 ) to donor ((D; Receptor-RLuc8) molecules. BRET max is the maximum BRET 2 signal when all of the donor molecules are interacting with acceptor molecules. AD 50 value corresponds to the acceptor/donor ratio providing 50% of the BRET max and reflects the relative affinity of the acceptor (GFP 2 -tagged receptor) for the donor molecules (RLuc8-tagged receptor). Fitting parameters were compared using Welch's t-test.

Simplistic BRET model for trimers with different association affinities
The BRET model for trimers is based on the Veacht and Steyer (1977) concept [50]. The theoretical saturation curve for trimers with the same association affinity is: where A, D and E are acceptor, donor and energy transfer efficiency, respectively. If we have different affinities for the formation of AAD trimer we introduce enhancement factor X: BRET competition assay simulation for trimers with different energy transfer ratio The BRET model for trimer competition is based on the Veacht and Steyer (1977) concept [50]. For receptors with equal affinities to form trimers we determine the frequency of each trimer type (A = acceptor, D = donor and W = wild type receptor): BRET is defined as a ratio of light emission from acceptor divided by that of the donor. E is resonance energy transfer ratio: If the energy transfer ratio E is different for ADW trimers the following equation is obtained: Figure 3. BRET 2 saturation curves of b 2 AR and IR homo-and heteromers. HEK-293 cells were transiently cotransfected with a constant amount of RLuc8-tagged and increasing amounts of GFP 2 -tagged receptor encoding constructs. BRET 2 values were plotted as a function of the ratio between the acceptor and donor fusion proteins (A/D ratio). A dataset is composed of 3-5 independent saturation experiments that were fitted using a nonlinear regression equation assuming a single binding site using GraphPad Prism 5.0 (Fitting results are found in Table 3 Informational spectrum method (ISM) The ISM is based on a model that assigns to each amino acid a defined parameter describing a physico-chemical property involved in the biological activity of the protein and corresponding to electron-ion interaction potential (EIIP). These values determine the electronic properties of the amino acids responsible for their intermolecular interactions [51].
The obtained numerical sequence, representing the primary structure of a protein, is then subjected to a discrete Fourier transformation defined as follows: where x(m) is the m-th member of a given numerical series, N is the total number of points in this series, and X(n) are discrete Fourier transformation coefficients. These coefficients describe the amplitude, phase and frequency of sinusoids, which comprised the original signal. The absolute value of the complex discrete Fourier transformation defines the amplitude spectrum and phase spectrum. The complete information concerning the original sequence is contained in both spectral functions. However, in the case of protein analysis, relevant information is presented in an energy density spectrum (for a review, see [51]) defined as follows: S n ð Þ~X n ð ÞX Ã n ð Þ~1=2X n ð Þ 1 =2 2 , n~1,2,:::,N=2: Thus, the initial information defined by the sequence of amino acids is now presented in the form of the informational spectrum (IS), representing the series of frequencies and their amplitudes.
The IS frequencies correspond to the distribution of structural motifs with defined physico-chemical characteristics responsible for the biological function of a protein. When comparing proteins that share the same biological or biochemical function, ISM allows the detection of code/frequency pairs that are specific for their common biological properties or correlate with their specific interaction. This common informational characteristic of sequences is determined by a cross-spectrum (CS) for two proteins or consensus informational spectrum (CIS) for two or more proteins-i.e., the Fourier transformation of the correlation function for the spectrum. In this way, any spectral component (frequency) not present in all of the compared ISs is eliminated. Peak frequencies in CIS are common frequency components for the analyzed sequences. A measure of similarity for each peak is the signal-tonoise ratio (S/N), representing the ratio between the signal intensity at one particular IS frequency and the main value of the whole spectrum. If a CIS is calculated for a group of proteins with different primary structures, and strictly defined peak frequencies are found, the analyzed proteins likely participate in a mutual interaction or have a common biological function. The ISM was, thus far, successfully applied in the structure-function analysis of different protein sequences [51], prediction of new protein interactors [43] and identification of protein domains responsible for long-range interactions [52,53].

Computational peptide scanning
Computational peptide scanning was used to define linear protein regions that contribute the most to the amplitude and signal to noise ratio at the characteristic frequency and, therefore, are responsible for the interaction(s) described by the particular spectral characteristic. To identify the regions with the highest amplitudes at predefined Fourier frequencies, the entire sequences of b 2 AR and IR were scanned by the ISM algorithm with overlapping windows of different lengths, leading to the identification of regions with the highest amplitudes at predefined Fourier frequencies.

Datasets
The sequences used for bioinformatics analysis were retrieved from the Uniprot database with the following accession numbers: P07550 (human b 2 AR) and P06213 (human IR isoform A that lacks exon 11).

Statistical analysis
Statistical significance was determined using Student's t-test and Welch's t-test. Differences were considered statistically significant at a p value less than 0.05.

Characteristics of the b 2 AR fusion constructs
Pharmacological characterization of the human HA-tagged b 2 AR (HA-b 2 AR) fused at the C-terminus with either the energy donor RLuc8 or energy acceptor GFP 2 was performed using BRET 2 data from saturation assays were fitted using the following equation for dimers: BRET = BRET max (1/(1+AD 50 /X)) where X is the ratio of acceptor (A; Receptor-GFP 2 ) to donor (D; Receptor-RLuc8) molecules. The BRET max is the maximal BRET obtained for a given pair and AD 50 value corresponds to the A/D ratio providing 50% of the BRET max . The best-fit parameters and standard errors were derived from the data presented in Fig. 3. Fitting parameters were compared using Welch's t-test. Statistical analysis shows that AD 50  ð5Þ radioligand binding assays ( Table 1). The IC 50 values of pindolol for HA-b 2 AR-RLuc8 and b 2 AR-GFP 2 were in agreement with those obtained for HA-b 2 AR and were also in the range previously reported for b 2 AR-RLuc fusion construct [47]. IR fusion constructs were previously characterized [34]. Radioligand binding assays were also performed with HEK-293 cells cotransfected   (Table 2).

b 2 AR surface expression and the internalization effect of IR coexpression
The effect of IR coexpression on b 2 AR surface expression and internalization properties was evaluated by ELISA in intact cells and by radioligand binding assays. As shown in Fig. 1A, a significant approximately 35% decrease in b 2 AR surface expression was observed in HEK-293 cells cotransfected with IR compared with b 2 AR-expressing cells (p,0.05). However, total b 2 AR expression was not affected by IR coexpression as shown by luminescence measurements, whereas IR expression was validated by fluorescence measurements. Analogous observations in b 2 AR surface expression were detected using radioligand binding assays, where b 2 AR surface expression decreased by around 30% (27.8962.11) in HEK-293 cells cotransfected with IR compared with b 2 AR-expressing cells (p,0.05).
Additionally, ELISA measurements suggested that IR coexpression moderately increased isoproterenol-induced b 2 AR internalization, whereas concomitant treatment with isoproterenol and insulin evoked a comparable internalization rate. Treatment with insulin did not induce marked b 2 AR internalization ( Figure S1). To further examine the internalization kinetics of b 2 AR, HEK-293 cells expressing either b 2 AR or b 2 AR and IR were treated with either isoproterenol (10 mM) or a combination of both isoproterenol (10 mM) and insulin (0.1 mM) for varying periods of  Interaction between b 2 AR and IR PLOS ONE | www.plosone.org time (5 min to 1 h), and the number of cell surface binding sites were determined ( Fig. 1B and C). We found that isoproterenolinduced steady-state proportion of internalized receptors was approximately 38% and 44% in b 2 AR-and in b 2 AR-and IRexpressing cells, respectively. The rate of isoproterenol-induced b 2 AR internalization was significantly increased by coexpression of IR (p,0.05), as the estimated t 1/2 was 8.661.3 and 5.460.6 min in b 2 AR-and in b 2 AR-and IR-expressing cells. Comparable results were obtained when either b 2 AR-or b 2 ARand IR-expressing cells were concomitantly treated with isoproterenol and insulin (see Fig. 1B and C). In addition, confocal fluorescence microscopy demonstrated predominant plasma membrane localization of the HA-b 2 AR-RLuc8 and IR-GFP 2 and their colocalization in untreated cells (Fig. 2, upper panels; untreated cells). Concomitant treatment of b 2 AR-and IRexpressing cells with isoproterenol and insulin (Fig. 2, lower panels) promoted receptor redistribution to the cytoplasm, indicating receptor internalization. Again, agonist-redistributed b 2 AR and IR showed a similar distribution pattern; however, a proportion of intracellular receptors did not colocalize.

Interaction between the b 2 AR and IR-BRET evidence
Considering the observed effects of IR coexpression on b 2 AR surface expression and internalization, as well as colocalization of both receptors in HEK-293 cells, we next investigated whether proximity indicative of direct interaction occurs between b 2 AR and IR by performing BRET dilution, saturation and competition assays. To obtain an absolute value of donor [D] and acceptor [A] molecules-i.e., Receptor-RLuc8 and Receptor-GFP 2 -we first generated correlation curves between total luminescence and fluorescence versus the number of receptor-binding sites determined by radioligand binding. The linear correlation was obtained between the number of HA-b 2 AR-RLuc8 and b 2 AR-GFP 2 binding sites and the total luminescence and total fluorescence, respectively ( Figure S2).
The derived correlation factors to convert total luminescence and total fluorescence into receptor number were 0.00031 and 0.0089, respectively. Thereafter, BRET 2 dilution assays with a constant [A]/[D] ratio were performed to set the concentration range for the saturation assays and distinguish monomers from dimers as previously described [54]. Due to the increasing noise in calculated BRET at low luminescence intensities, the lowest amount of HA-b 2 AR-RLuc8 and IR-RLuc8 cDNA used for transfection in saturation assays was determined at 0.05 and 0.015 mg, respectively. BRET 2 saturation assays were performed to investigate homoand hetero-merization of b 2 AR and IR. The BRET 2 signal was measured in live HEK-293 cells that were transiently transfected with constant amounts of RLuc8-tagged (HA-b 2 AR-RLuc8, IR-RLuc8) and with increasing amounts of GFP 2 -tagged (b 2 AR-GFP 2 , IR-GFP 2 ) receptor encoding constructs. In dimers and higher oligomers, the probability of BRET interaction increases with increasing acceptor/donor ratio until all of the acceptors have pairs, and the maximum BRET 2 level (BRET max ) or plateau is reached. The AD 50 value (also designated as BRET 50 ; see methods) was also included to compare the affinity for homo2/ heteromer formation. BRET 2 saturation experiments data are shown in Fig. 3 and the expression levels of RLuc8-and GFP 2tagged receptor constructs are shown in Figure S3. For all of the receptor combinations, the BRET 2 signal was plotted as a function of the A/D ratio increased as a hyperbolic function reaching a saturation level (Fig. 3). Control experiments for the specificity of the interaction were performed by cotransfecting RLuc8-tagged receptors with increasing amounts of membrane-inserted GFP 2tagged construct (GFP 2 -17aa). Plasma membrane localization of the GFP 2 -17aa construct was previously demonstrated [45]. The BRET 2 signal increased linearly with the increase in fluorescence/ luminescence (GFP 2 /RLuc8) ratio, most likely reflecting random collisions between the RLuc8-tagged receptors and control unrelated GFP 2 -tagged construct (GFP 2 -17aa) ( Figure S4). Saturation assay data were fitted using a dimer model in the approximation of small energy transfer [54] (Table 3). We observed the highest BRET max -i.e., 437 mBU in the b 2 AR homologous saturation assay and considerably lower BRET max values for other receptor combinations. It should be stressed that the BRET max value depends on the distance between BRET pairs and is not a measure of the strength of interaction between the acceptor-and donor-tagged receptors. In homologous saturation Table 4. The affinity of interaction between the b 2 AR and IR homo-and heteromers characterized by the signal-to-noise ratio (S/ N) at the characteristic frequency (F) in the consensus informational spectrum (CIS). assays, the value of AD 50 for the IR homomer was significantly lower (p#0.05) than that for the b 2 AR homomer (Table 3). For the receptor complex consisting of b 2 AR-RLuc8:IR-GFP 2 , the AD 50 value was significantly lower (p#0.05) than the AD 50 value for the IR-RLuc8:b 2 AR-GFP 2 . For heterodimers, there should be no difference in the relative affinity between AD and DA dimers. To interpret our data, we developed a simplistic trimer model for interpretation of AD 50 values in a BRET saturation assay. According to this model, different AD 50 values are derived from theoretical BRET saturation curves with different affinities for various types of trimer formation ( Figure S5). This model should be used with caution because it does not take into account the simultaneous formation of dimers, trimers and higher-order oligomers. Stimulation with the agonists insulin or isoproterenol did not promote any detectable change in the BRET 2 signal (data not shown), indicating that the receptor dimers/oligomers form constitutively and that addition of agonists does not induce a detectable change in the conformational or oligomerization state of the receptor complexes. However, it needs to be stressed that BRET only provides information concerning a steady-state population of dimer/higher-order oligomers and is not suited to monitor the rapid, ''real-time'' dynamic type of di2/oligomerization [55].
To further support the findings from the saturation assays, we performed homologous and heterologous BRET 2 competition assays where HEK-293 cells were co-transfected with a constant amount of RLuc8-and GFP 2 -tagged receptor while increasing the amount of untagged receptor (Fig. 4A and B). Competition experiments were carried out at the constant Receptor-GFP 2 / Receptor-RLuc8 expression ratio to avoid possible variations in the BRET 2 signal due to fluctuation in the relative expression levels of the energy donor and acceptor ( Fig. 4C and D). It is expected that the BRET 2 signal would decrease if untagged receptors compete with the tagged receptors for the binding in complexes. In homologous BRET 2 competition assays BRET 2 signal decreased with increasing amount of competitor (un-tagged receptor). In the IR homologous competition assay, we observed a typical competition curve for dimers, where introduction of the same amount of untagged receptor produced an approximately 50% reduction in the observed BRET signal. In the b 2 AR homologous competition assay, the reduction of the BRET signal was smaller. This observation cannot be explained by a simple dimer or trimer model; it could be attributed to clustering of the b 2 AR where several acceptors can interact with each donor. In heterologous competition assays, we observed a transient increase in the BRET signal with a later hyperbolical decrease (see Fig. 4A and B). Untagged b 2 AR caused approximately 1.4-fold increase in the IR BRET signal, while the effect of untagged IR on the b 2 AR BRET signal was less obvious; maximal observed increase was less than 1.2-fold. The transient increase in the BRET signal with a subsequent hyperbolical decrease is theoretically predicted for trimers or higher-order oligomers, where the donor, acceptor and competitor are all present in the same complex ( Figure S6).

Interaction between b 2 AR and IR characterized by ISM
To support our experimental evidence with the bioinformatics data, we next applied the informational spectrum method (ISM), a virtual spectroscopy method to investigate protein-protein interactions and to analyze the structure/function relationship of proteins. ISM was utilized to identify important informational characteristic of the interaction between b 2 AR and IR and identify the structural determinants potentially involved in receptor heteromerization. The primary structure of proteins encodes the information represented by the informational spectrum (IS) frequencies that correspond to the protein biological function. Mutually interacting proteins share common information that is represented by peaks in their cross-spectrum [51]. The informational spectrum of b 2 AR is presented in Fig. 5A. It contains two characteristic peaks at the frequency F(0.216) and F(0.355). Fig. 5B represents the IS of the IR. By performing cross-spectral analysis of b 2 AR and IR, we have identified that these two molecules share common information corresponding to the IS frequency F(0.216) (Fig. 5C). To further evaluate the importance of the peak at F(0.216) we performed CS analysis of b 2 AR and IR with scrambled IR and b 2 AR proteins as negative controls. The scrambled proteins with the identical amino acid composition to that of b 2 AR and IR were created by random permutation of original proteins. CS analysis between wild type receptors and randomly selected scrambled b 2 AR and IR are presented in Figure S7. The intensity of whole spectrum and the value of amplitudes at the characteristic peak F(0.216) is higher in CS of two wild type proteins compared to CS of original and scrambled proteins confirming the importance of the characteristic peak at the F(0.216) for interaction between b 2 AR and IR.

Identification of the key protein domains responsible for the interaction between b 2 AR and IR
Computational peptide scanning of b 2 AR and IR was performed to identify the regions of proteins essential for information corresponding to the frequency F(0.216). The computer-assisted peptide scanning survey of the primary structure of b 2 AR with overlapping windows of different lengths revealed that the region encompassing residues 325-364 is essential for the information represented by the frequency F(0.216) (Fig. 6A). Further peptide scanning of IR identified three principal regions as important for the information represented by the frequency F(0.216), however region encompassing residues 1269-1314 represents the most probable domain involved in this interaction (Fig. 6B).

Affinity of interaction between protomers
Peak frequencies in CIS represent common information encoded by the primary structures of analyzed proteins. Significance of information is determined by the signal-to-noise ratio (S/ N), representing the ratio between the signal intensity at one particular IS frequency and main value of the whole spectrum. A higher S/N value at the characteristic frequency (F) in CS/CIS of two or more proteins suggests a higher propensity for their interaction.
The current analysis showed that the interaction affinities between the homomers of b 2 AR and IR are similar and that at the level of dimer formation, both b 2 AR and IR displayed a higher propensity toward homodimerization than heterodimerization ( Table 4). Considering that in the CS of the b 2 AR and IR, the two receptors share common information corresponding to the IS frequency F(0.216), it can be assumed that this frequency is equally important for heterodimerization and for higher-order heterooligomer formation. The obtained S/N values at F(0.216) in the CS of b 2 AR:b 2 AR:b 2 AR:IR and b 2 AR:b 2 AR:IR:IR tetramers were comparable, whereas that for the IR trimer displayed a considerably lower affinity for interaction with the b 2 AR monomer ( Fig. 7 and Table 4).

Discussion
7TMRs form the largest and most important pharmacotherapeutic target in drug discovery. Continual discovery of receptor heteromers expands the repertoire of functional 7TMR units and offers promising new targets for drug development (reviewed in [56]). Increasing evidence for the existence of functional 7TMR:RTK heteromers [7,36,37,38,39] adds another level of complexity to this system. In the present study, we combined in vitro experimental and bioinformatics approaches for the investigation of the interaction between b 2 AR and IR. Initially, it was observed that IR coexpression reduces b 2 AR surface expression and accelerates its internalization. It could be assumed that IR increased constitutive internalization of the b 2 AR [57], thereby reducing its surface expression. The extent of isoproterenol-induced b 2 AR internalization was comparable to the previously reported internalization rate for b 2 AR [58,59]. The estimated t 1/2 of ,9 min for isoproterenol-induced b 2 AR internalization is also within the range reported for the same receptor using a fluorimetric assay [60]. Augmented agonistinduced b 2 AR internalization promoted by IR coexpression is in good agreement with that described in a previous report [61]. The amount of surface-expressed b 2 AR lost to internalization in response to insulin in HEK-293, ,15%, correlated well with results obtained in DT 1 MF-2 smooth muscle cells and CHO cells [61,62] but was smaller than the reported value of ,30% in epidermoid carcinoma A431 cells [6]. Our colocalization results in untreated cells correlate well with recent coimmunoprecipitation findings demonstrating that b 2 AR and IR form complexes in the mouse heart, in which the stimulation of IR resulted in the reduction of the association of the b 2 AR:IR complex [7].
BRET results provided further evidence for the direct interaction between b 2 AR and IR. Hyperbolic saturation assay curves are indicative of a specific constitutive interaction between IR and b 2 AR. The BRET max obtained for b 2 AR homomerization is higher than that described in previous reports [19,23,24], whereas the relative affinity of b 2 AR for homomerization is within the previously reported range (AD 50 of 1.760.5 vs. 1.2 6 0.33 [23]). The RLuc8/GFP 2 BRET 2 donor/acceptor pair used in our study has the largest Förster distance R 0, leading to 50% of energy transfer from the donor to acceptor, thereby extending the maximum distance that the BRET 2 system can accurately measure [63]. Lower reported BRET max values were obtained with original BRET 1 [19] or BRET 2 [23,24] combinations. The BRET max values generated in the homologous IR saturation assay were also slightly higher than that previously reported [29]. The same study revealed that the substantial insulin-induced effect on the BRET signal could only be detected with partially purified fusion IRs [29]. Single-molecule analysis of fluorescently labeled 7TMRs recently demonstrated that b 2 AR had a high tendency of forming dimers, even at low densities comparable to receptor expression in native tissue; however, at higher densities, b 2 AR formed a mixture of di-, tri-, and tetramers [55]. Membranedriven spatial organization of b 2 AR also favors its stable/extensive oligomerization [64]. The inability of agonists to modify the oligomerization status of b 2 AR was also reported in studies that utilized methods other than BRET (reviewed by Ferre et al. [56]). Both receptors displayed a similar propensity to form homomers as assessed by AD 50 and S/N values. Affinities for the formation of b 2 AR:IR and IR:b 2 AR heteromers differed substantially, a finding that could not be explained by a dimer model that assumed a heteromer stoichiometry of 1:1. The lower AD 50 value observed for the b 2 AR-RLuc8:IR-GFP 2 pair could suggest either a higher affinity for heteromer formation or, alternatively, complex oligomer formation. The melatonin MT 2 receptor also displayed a higher propensity for heteromerization with the MT 1 receptor [65], whereas certain receptor pairs (i.e., b 1 AR:b 2 AR [23,24], b 2 AR:b 3 AR [20]), oxytocin and vasopressin receptors [66] showed a similar propensity, and muscarinic acetylcholine receptors [67] showed a slightly lower propensity for heteromerization. However, simplistic model for interpretation of AD 50 values and heterologous competition assays data provided evidence for higher-order oligomeric complex formation. A transient increase in the BRET signal suggests that the energy transfer efficiency E is increased in hetero-oligomeric complexes due to the smaller distance between the BRET donor/acceptor pair in the complex (see Figure S6). A similar effect-i.e., an increased homo-dimer BRET signal induced by an unrelated, untagged receptor-was previously observed with the gastric inhibitory polypeptide receptor [24]. Therefore, both the change in affinities and transient increase in the BRET signal could be a reflection of higher-order oligomeric complexes with affinities for distinct associations such as with trimers and tetramers with 2:1 and 2:2 stoichiometry, respectively, as proposed by Breitwieser et al. [68]. At the level of dimer formation, both b 2 AR and IR displayed a higher propensity toward homodimerization than heterodimerization, suggesting homodimers as the basic units engaged in heteromerization. Affinity calculation for trimers and tetramers highlighted differences between b 2 AR and IR. Apparently, neither IR dimers nor trimers could form high affinity interactions with the b 2 AR monomer. However, IR monomers/homodimers can form high affinity heteromers with b 2 AR di2/trimers. Considering that both receptors displayed a higher propensity toward homodimerization and that IR is present in the plasma membrane as a disulfide-linked dimer [28], it is plausible to suggest that high-order 7TMR:RTK oligomers most likely comprise heteromers of homodimers (2:2 stoichiometry). The computer-assisted peptide scanning survey of the primary structure of b 2 AR and IR revealed the domain encompassing residues 325-364 and 1269-1314 as prospective interaction domains. The identified region is located at the end of the 7 th TM domain and C-terminal tail of b 2 AR and almost completely overlaps with helix 8 (helix adjacent to TM7 running along the internal membrane surface) of b 2 AR (residues L324-N357) [69]. A recent study also identified b 2 AR helix 8 as an important dimerization surface region [70]. In addition, this region contains major sites (Tyr350/Tyr354 and Tyr364) for IR-mediated phosphorylation of b 2 AR [3,71] and consensus sites of Aktcatalyzed phosphorylation (Ser345/Ser 346) [72]. These results are in accordance with previous findings showing that b 2 AR is a substrate for IR and proposing direct interaction between these proteins [3,71]. The C-terminal cytoplasmic tail-more specifically, the 15-amino acid motif (residues 342-356)-was sufficient to confer the b 1 AR-b 2 AR tail chimera the ability to be regulated by insulin [62].
The prospective interaction domain identified in IR (residues 1269-1314) is positioned in the cytoplasmic part of the IR b chain. This region encompasses the terminal end of the tyrosine kinase catalytic domain and is a part of the C-terminal tail. The Cterminal domain of the IR b subunit has been found to play a key role in the regulation of tyrosine kinase activity [73,74]. The construct based on proteolytic cleavage used to solve the crystal structure of the IR tyrosine kinase domain of the human IR Bisoform [75] ends at residue 1283, which corresponds to the amino acid at position 1271 in the human IR isoform A used in our study. Therefore, it could be hypothesized that the terminal end of the IR tyrosine kinase domain and a part of the C-terminal tail are involved in the interaction with b 2 AR.
The involvement of intracellular domains (C-terminal tail and ICL3) was found to be fundamental for heterodimerization between the cannabinoid CB 1 receptor, adenosine A 2A and dopamine D 2 receptors, thus favoring the idea that electrostatic interactions between intracellular domains are more predominant in receptor heteromers and constitute a general mechanism for receptor heteromerization [76].
In summary, BRET data and ISM bioinformatics provided evidence for direct interaction and higher-order b 2 AR:IR oligomer formation that we hypothesize comprise heteromers of homodimers and identified prospective intracellular interaction domains engaged in heteromerization. In this regard, 7TMR:RTK heteromers could potentially generate a basis for the design of new therapeutics that can compete with today's epidemics, such as type-2 diabetes, obesity and cardiovascular diseases.  Figure S3 Relationship between receptor-RLuc8 and receptor-GFP 2 constructs expression. Expression levels of RLuc8-and GFP 2 -tagged constructs used in BRET 2 saturation assays were also monitored by luminescence and fluorescence measurements. Total luminescence was measured after the addition of the RLuc8 substrate coelenterazine 400a. Total fluorescence was measured with an excitation filter at 380 nm and an emission filter at 515 nm. Data are expressed as the means6S.E. of 3-5 independent saturation experiments. (TIF) Figure S4 Random collisions between the RLuc8-tagged receptors and membrane-inserted GFP 2 -tagged construct (GFP 2 -17aa). HEK-293 cells were transiently cotrans-fected with a constant amount of RLuc8-tagged receptors and increasing amounts of GFP 2 -17aa encoding construct. BRET 2 values were plotted as a function of the ratio between the total fluorescence/total luminescence (GFP 2 /RLuc8 ratio). Total luminescence was measured after the addition of the RLuc8 substrate coelenterazine 400a. Total fluorescence was measured with an excitation filter at 380 nm and an emission filter at 515 nm. Increasing the concentration of GFP 2 -17aa in cells expressing either the IR-RLuc8 or b 2 AR-RLuc8 resulted in high, but nonspecific linear increase of the BRET 2 signal. Data are expressed as the means 6 S.E. from three independent experiments performed in triplicate. Representative BRET 2 saturation curves of b 2 AR and IR homomers are shown for comparison. (TIF) Figure S5 Comparison of theoretical BRET saturation curves with different affinities for trimer formation. Shown are simulated BRET saturation curves for case with the same affinity for AAD and ADD formation (solid line) and two special cases with different affinities for formation of AAD compared to ADD (hatched and dotted lines). Note that in all three cases the AD 50 Figure S7 Cross-spectrum (CS) of (A) wild type b 2 AR and IR, (B) scrambled b 2 AR and wild type IR and (C) scrambled IR and wild type b 2 AR. Note that the value of amplitudes at the characteristic peak F(0.216) is higher in CS of two wild type proteins (panel A) compared to the CS of wild type and scrambled proteins (panels B and C). (TIF)