View all hidden authors and organizations

Edited by Robert J. Lefkowitz, HHMI, Durham, NC, and approved on October 22, 2021 (reviewed on December 31, 2020)

We look for β2AR agonists to treat obstructive lung diseases (such as asthma), in which the receptor relaxes airway smooth muscle (ASM) cells and opens the airway. Compared with activating β-arrestin (which limits its effectiveness due to receptor desensitization), agonists that facilitate Gs coupling (causing airway relaxation) were studied in a 40 million compound screening library. Among the several agonists identified, one clearly favors β-arrestin. Compared with other agonists, the agonist-receptor-G protein model reveals different receptor interactions. The apparently biased beneficial effects of this agonist are demonstrated in the physiological system (ASM relaxation). These studies point out that a different structural type of β-agonist can be used to treat obstructive pulmonary disease without the adverse reactions associated with rapid immunity.

G protein-coupled receptors display multifunctional signals and provide the potential of promoting conformational selectivity toward output for agonist structures. For β2-adrenergic receptor (β2AR), the unbiased agonist stabilizes the conformation, which leads to coupling with Gαs (cyclic adenosine monophosphate [cAMP] production / human airway smooth muscle [HASM] cell relaxation) and β-arrestin Participate, the latter acts to quench the Gαs signal and help receptor desensitization/rapid defense. We screened a sorted library of stents containing 40 million compounds and revealed unexpected agonists with dihydroimidazolyl-butyl-cyclic urea stents. Through the four methods, the S-stereoisomer of compound C1 showed no detectable β-arrestin involvement/signaling. However, C1-S retains the Gαs signal-a difference in output that is beneficial for the treatment of asthma. Functional studies of the two models confirmed this bias: β2AR-mediated cAMP signaling desensitized the unbiased agonist salbutamol instead of C1-S, and the desensitization of HASM cell relaxation was observed with salbutamol, but No observations for C1-S. These HASM results indicate the biologically relevant deviation of C1-S in the human cell type of interest in the context of relevant physiological responses. Therefore, compared with salbutamol and C5-S, C1-S is clearly biased towards β-arrestin. Modeling and simulation of the C1-S structure revealed binding differences compared to the unbiased adrenaline of transmembrane (TM) segments 3, 5, 6, 7 and ECL2. C1-S (R2 = cyclohexane) is repositioned in the pocket. Compared to the similar unbiased C5-S (R2 = phenyl), it loses the TM6 interaction and gains the TM7 interaction, which seems to help The C1-S bias comes from β-arrestin. Therefore, an unknowable large chemical space library identifies agonists with receptor interactions, leading to the splitting of signals related to the action of β2AR, which is beneficial to the treatment of obstructive lung disease.

Most G protein coupled receptors (GPCRs) are now considered multi-signal sensors (1, 2). The early concept of agonist-receptor interaction was based on the idea that any combination of agonists would induce a single "active" receptor conformation, resulting in an interaction with the heterotrimeric G protein and a universal single signal . Generally, the alpha subunit of G protein is considered to be the main activator (or inhibitor) of the effector after its dissociation, thereby generating intracellular signals. Subsequently, due to the specific molecular determinants (3⇓ –5) of the receptor triggering independent mechanisms, a single agonist may produce multiple signal transduction results, that is, activating a given GPCR will produce multiple signal transduction results. With the determination of these multiple functions, it is obvious that agonists with different structures can act on a given receptor to preferentially activate one signal, while other signals are the least involved. This feature is later called signal bias (6⇓ – 8). Because of this signal selectivity, biased agonists may have important advantages over non-biased agonists, which can activate specific therapeutic pathways while minimizing unnecessary or harmful signaling. The pathway selectivity of biased agonists is thought to be established by a set of different interactions with unbiased (also called balanced) agonists (9⇓ ⇓ –12) to stabilize the specific conformation of the agonist-receptor complex. Although it is conceivable that small modifications to established homologous agonists may produce this particular signal, it may require a significant deviation from the common agonist structure to achieve this goal (13).

The signal/function of a given GPCR that may be sought to be selectively activated is defined by the cell type, the disease, and the ultimate physiological function required. In asthma and chronic obstructive pulmonary disease (COPD), the contraction of active human airway smooth muscle (HASM) cells restricts airflow, which is the main cause of morbidity and mortality. The β2-adrenergic receptors (β2ARs) expressed on HASM cells are targets of therapeutically administered β-agonists, which relax cells through a mechanism mediated by cyclic adenosine monophosphate (14). Beta-agonists are used for the treatment of acute bronchospasm and long-term prevention. However, the response of HASM bronchodilators to acute β-agonists is weakened by receptor desensitization (15). Typical treatment of human or isolated HASM cells results in loss of receptor function over time (16) ⇓ –18), clinically called rapid allergic reaction.

The desensitization of β2AR (and other GPCRs) promoted by agonists is due to the partial uncoupling of the receptor and G protein, which is caused by G protein-coupled receptor kinase (GRK) to the receptor's intracellular Ser/Thr residues Initiated by phosphorylation (19, 20). GRK phosphorylated β2AR recruits β-arrestin1 or β-arrestin2 to these receptors, and the subsequent interaction appears to compete with the receptor for binding to the Gα subunit, thereby weakening the intracellular response (11, 21). β2AR (22, 23) strongly infers this competition and is attractive for rhodopsin-arrestin interactions (24). In addition, the binding of β-arrestin to GPCRs can initiate receptor internalization and other events through its polyprotein linker function, such as receptor activation of ERK1/2 (25). Therefore, the participation of β-arrestin can be regarded as the early “second signal” of β2AR, and also the desensitizing initiator that weakens Gs signal. Agonists that prefer Gas coupling (cAMP production and airway smooth muscle [ASM] relaxation) and away from β-arrestin binding (desensitization) are desirable in the treatment of obstructive pulmonary disease, because the efficacy will not be rapidly diminished and rapid hypersensitivity The response will not gain experience from long-term treatment. Although some GPCRs (such as μ-opioids and type 1 angiotensin II receptors) have been described as biased agonists for G protein or β-arrestin (6) signaling, most of them use catecholamines In the study, the Gαs bias was not obvious to β2AR. Therefore, we have little information on whether the two β2AR pathways can be differentially activated by effective agonists in a selective manner. From a structural point of view, it is also unclear what strategies can be used to design biases in this way. Body agonist.

In order to find this type of bias in β2AR, we screened a library of 40 million compound scaffold ranking (SR), which is not related to the known β2AR agonist structure. We found a scaffold in which the substitution of certain R groups resulted in a single compound that was clearly a Gas-biased agonist of β2AR, while there was no obvious β-arrestin involvement in the model system. Other studies on HASM cells have shown that the lead compound lacks a rapid defense against relaxation compared to the most widely used β2AR agonist salbutamol. The structure of this biased agonist is very different from that of catecholamine-like agonists. In order to determine the mechanism that may lead to this biased activity, we used structural modeling and molecular simulation, and studied homologous compounds with different R groups and receptor mutagenesis to predict the interaction site with activated β2AR. This type of research reveals different structural features that may lead to bias effects.

Use untransfected CHW-1102 cells (CHW) to screen for unique agonists acting on β2AR, CHW-1102 cells do not endogenously express the receptor, and the same cell line stably transfected with human β2AR complementary DNA (cDNA), Express ~2,000 fmol/mg receptor protein (called CHW-β2). The prior criteria for a compound (or a mixture of compounds) to be eligible for further research are as follows: 1) The stimulation of cAMP in CHW-β2 cells exceeds the statistical significance of baseline, 2) The untransfected CHW cells do not stimulate cAMP, 3) The cAMP reaction from the compound (or compound mixture) blocked by the βAR antagonist propranolol, 4) The stereoisomer of a single compound, compared with the parent stereoisomer, shows a reduced cAMP reaction, 5) The concentration-response data of cAMP can be fitted to a four-parameter least squares regression equation (sigmoid curve with R2> 0.9), and 6) The average Hill coefficient of the compound fitting curve is between 0.8 and 1.3.

Our method of screening 40 million compounds is to use the SR library, where each well in the sample plate contains thousands of compounds arranged in a scaffold type system, so each well contains only the compounds of a specific scaffold (26). In this case, we studied 87 different SR library sample wells composed of small molecules, cyclic peptides, and linear peptides. From the screening of the SR library (Figure 1), the 1319-well dihydroimidazolyl-butyl-cyclic urea scaffold (Figure 2A) increased cAMP by more than 6 times in CHW-β2 cells compared to the base, but not The significant increase in cAMP in CHW cells prompted further investigation of these compounds. Although compared with CHW cells, several other sample wells caused an increase in cAMP in CHW-β2 (see the magenta to blue transition in Figure 1), but the increase was <1.5 times, so we focused on the compound in well 1319 . It is recognized that potential agonists in these other wells may be missed in this screening. When there are many compounds in each well, the concentration of these agonists may be very low. Nevertheless, this method has successfully identified many unique agonists of GPCRs (27, 28), and in any case, the agonists in well 1319 seem to be suitable for further study based on efficacy.

The cAMP screening of the SR library containing 40 million compounds identified the dihydroimidazolyl-butyl-cyclic urea scaffold containing β2AR agonist. A total of 116 wells, each containing a compound mixture (250 μg/mL) arranged by scaffold type system for the treatment of cells transfected with β2AR (CHW-β2 cells, bottom row) and untransfected CHW cells (Top row). The cAMP response of all wells is shown as a heat map, indexed to the cAMP response of 10 μM forskolin using the color scheme shown. One hole (SR library hole 1319) increased cAMP in CHW-β2 cells by 6 times the basal level (0.25% dimethylformamide [DMF] carrier). As shown, this compound mixture does not stimulate cAMP in untransfected CHW cells. Shown are the average results from the two sets of measurements.

Scaffold structure of compound and compound C1-S in hole 1319 of SR library. (A) Brackets with indicated positions (see also SI appendix, Figure S2), denoted as R1, R2, and R3. (B) The structure of a single compound represented as C1-S (1 of 12), which was determined from the deconvolution of the cAMP response of the synthetic PS library from well 1319 of the SR library.

To predict the relative efficacy of each of these compounds from the 1319 wells in the SR library, we used a calculation method and a customized position scan (PS) library, as described previously (29, 30). The PS library is used to predict the relative activity of all compounds in the library by using exponentially fewer samples. For the SR library sample well 1319, we used 116 sample wells to evaluate the relative efficacy of all 56,610 different dihydroimidazolyl-butyl-cyclic ureas. 56,610 compounds are derived from 34 different functional groups at the R1 position, 37 functional groups at the R2 position, and 45 functional groups at the R3 position (34 × 37 × 45 = 56,610). The samples from 1319 holes, named 1319-1 to 1319-34, contain all the dihydroimidazolyl-butyl-cyclic ureas arranged in their R1 functional group system. For example, a given sample well contains all 1,665 dihydroimidazolyl-butyl-cyclic ureas with S-methyl fixed at position R1, while another contains all 1,665 compounds with S-benzyl fixed at R1 . In a similar way, create sample holes with fixed R2 or fixed R3 positions.

The cAMP screening resulted in several positive sample wells in the PS library. An example is shown in Figure S1 of the SI Appendix, where the response to the 1319 scaffold compound is arranged in groups located at R1 (Figure 2A). We deconvolve the PS library screening results by selecting multiple functions from each of the R1, R2, and R3 positions and combine them. In order to do this, we first sorted the samples of each of the three positions by cAMP reactivity, and then analyzed the data on the structure-activity relationship trend, as described (29, 30). Using this method, 12 S-compounds (represented as C1-S to C12-S) and R-stereoisomers at the R1 position (represented as C1-R to C12-R, respectively) appeared for synthesis and Further study (see the structure in SI) appendix, Figure S2). According to the PS data, the S compound is predicted to be active, while the R compound is predicted to be significantly less active (if any). The response of cAMP to individual compounds C1-S to C12-S at various concentrations was determined (Figure 3). The average parameters of a single curve fit are shown in Table 1. Each S-stereoisomer stimulated cAMP to varying degrees, as shown in the figure. Additional cAMP studies were performed using C1 to C12 S- and R-stereoisomers to determine stereoselectivity and antagonism The reaction in the presence of propranolol (Figure 4). We noticed that the R analogs caused little or no stimulation of cAMP at baseline (Figure 4). In addition, with the exception of C7-S and C8-S, most agonists showed complete blockade of propranolol (the cAMP level was not different from the baseline level). The C7-S response (Figure 3) cannot fit the sigmoid curve, and the Hill coefficients of some other S simulations exceed the pre-specified range (Table 1). Based on the lack of complete antagonism of Propranolol and/or Hill coefficient, the seven S-isomer compounds were not further studied.

Dose-response curves of cAMP in CHW-β2 and untransfected CHW cells derived from individual compounds denoted as C1-S to C12-S, derived from the deconvolution of the PS library. Refer to the SI appendix, the structure of Figure S2 and the summary of the curve fitting derived parameters in Table 1. The result is the mean ± SE of four to six experiments.

The functional characterization of agonists C1-C12 is determined by the production of cAMP in whole cells

Pharmacological properties of single S and R stereoisomer compounds C1 to C12. CHW-β2 cells were pretreated with vehicle or 10 μM βAR antagonist propranolol for 5 minutes. Then determine the response of cAMP to a 50 μM single S-active compound or its R-stereoisomer at the R1 position (Figure 2A and SI appendix, Figure S2). *, the response to the compound (vehicle pretreatment, black bar) is greater than (P <0.01) the baseline (0.25% dimethylformamide [DMF]); for all gray bars except C7-S and C8-S, Propranolol pretreatment did not cause significant stimulation of the compound to baseline (P> 0.05); †, P <0.05 compared to baseline. The result is the mean ± SE from three experiments.

The β-arrestin binding promoted by the agonists of the remaining S-isomer compounds was further evaluated. We first used the previously described proximity ligation assay (PLA) (31) to transfect the carboxy-terminal GFP-tagged β2AR and carboxy-terminal myc-tagged β-arrestin2 cDNA expression constructs into human embryonic kidney (HEK)-293T cells. When the distance between the two proteins is at least 30 nm, the confocal microscope will show red emission. Here, we use the β2AR-specific partial agonist salbutamol as the benchmark agonist because it is widely used as an asthma bronchodilator, and the selected compound is also a partial agonist. The representative results are shown in Figure 5A, and the quantitative results are shown in Figure 5B. Studies using PLA revealed that the binding of the receptor to β-arrestin2 is caused by the agonists C5-S and C6-S, which are comparable to salbutamol, while C3-S is less recruited. In contrast, C1-S (see the structure in Figure 2B) showed no β-arrestin2 binding. The results of the dose response experiment using PLA with salbutamol, the full agonist isoproterenol and C1-S are shown in Figure S3 in the SI Appendix. A dose-dependent increase in β-arrestin binding was found in salbutamol and isoproterenol, but no C1-S signal was detected at any concentration tested.

The agonist C1-S favors β2AR Gαs rather than β-arrestin interaction. (AH) C1-S cannot promote the interaction between β2AR and β-arrestin. (A) Representative PLA comes from HEK-293 cells transfected with the unbiased agonist albuterol (ALB, 10 μM) or the specified compound (300 μM). The red signal represents the proximity of β2AR-GFP and β-arrestin2-myc promoted by the agonist, which is only found in cells expressing β2AR-GFP (green signal). (B) Red spectrum imaging results from five independent PLA experiments, the agonist concentration is the same as in A; *P <0.05, **P <0.001; PLA signal is different from salbutamol. Refer to the SI appendix, Figure S3, for the results of agonist dose response. (CF) The binding of β-arrestin2 to β2AR, as determined by ECA (n = 3 to 5). The response curves of isoproterenol (ISO), ALB, and C5-S showed a concentration-dependent increase in β-arrestin signal (CE), while C1-S did not (F). (G) Image of full βAR agonist ISO (1μM) and partial agonist ALB (10μM) instead of partial agonist C1-S (300μM) agonists promote β-arrestin-GFP recruitment to cell surface spots). The image represents four experiments. (H) Agonist-dependent, β-arrestin-mediated ERK1/2 activation was observed in ISO and ALB, but not in C1-S. The bar graphs are from five experiments, and the Western blot shows the activation of Erk1/2 observed in a representative experiment. * P <0.05 compared with vehicle; NS, meaningless. (I) The deviation graph of the combination of cAMP and β-arrestin2 for ISO, C1-S and C5-S, the index is the maximum ISO response. The data shown are averages and 95% CI from three to four experiments. C1-S does not show detectable β-arrestin2 binding and has a flat curve with non-overlapping values ​​at most concentrations (see illustration). In some cases, the 95% CI bars are smaller than the drawing symbols and are not displayed. (J) The deviation factor (β) of the designated agonist is calculated using the data from the BRET2 detection of Gs coupling and ECA for β-arrestin binding, using the formula shown in the SI appendix, SI extension method. See also the SI appendix, Figure S5. Since C1-S β-arrestin binding was not detected, the calculated value of β is uncertain but greater than zero. The results are shown as the mean ± 95% CI of four to five experiments. ALB and C5-S are not statistically different from 0 (ie, balanced agonists). (The scale bar in A and G, 10 μm.)

To further explore this C1-S phenotype, a dose-response experiment was performed using the enzyme complementation test (ECA, "PathHunter"), which has been shown to be highly sensitive to the detection of β-arrestin binding to GPCRs (32). Here, we use transfected Chinese hamster ovary (CHO) cells to express β-arrestin2, which is fused with fragment-deficient β-galactosidase and β2AR labeled with a complementary fragment at its carboxy terminus. After the agonist promotes the binding of β-arrestin to the receptor, β-galactosidase is reconstituted to produce an active enzyme, which is detected by luminescence. Using this method, we detected the binding of dose-dependent β-arrestin to β2AR from isoproterenol and salbutamol, with half maximum effective concentrations (EC50) of ~100 nM and ~1,800 nM, respectively (Figure 5C and D ). These results indicate that there is a 1.22 log difference in the effectiveness of these two agonists to promote the binding of β-arrestin to β2AR, which is similar to the 1.11 log difference previously reported by others using similar assays (32). In addition, note that C5-S compounds promote β-arrestin binding (Figure 5E), consistent with PLA. However, β-arrestin ECA lacks a response to C1-S, and no concentration results in a signal greater than the base (vehicle alone, Figure 5F). Given that both PLA and ECA show a lack of β-arrestin binding to C1-S, we speculate that within the limitations of these two assays, compared to isoproterenol or salbutamol, this agonist actually does not cause Significant β-arrestin response. It has been reported (33) that when GRK2 is overexpressed, some agonists of a given GPCR, but not all [e.g., D2 agonist 75A (33)], show increased β-arrestin binding reactivity, which may be due to the assay Sensitivity increase. Therefore, we performed additional ECA experiments using C1-S agonists, which resulted in GRK2 overexpression> 8 times (SI appendix, Figure S4A). We found that the β-arrestin signal from C1-S did not consistently increase under these conditions (R2 value of global best fit = 0.217). Compared with the endogenous expression at each C1-S concentration, the examination of the changes in the luminescence signal of GRK2 overexpression showed that the signal amplitude has very small fluctuations (SI appendix, Figure S4B), which is not statistically significant. Significantly, even under the condition of GRK overexpression, the obvious deviation of C1-S at β2AR is confirmed. We also studied C1-S using a qualitative method, in which the redistribution of GFP-β-arrestin2 cells in response to agonists was monitored with fluorescent confocal microscopy over time in living cells, expecting that there would be no such morphological event in C1- S. As we and others have described (15, 34), the recruitment of β-arrestin2 to β2AR under these conditions is related to the change in the cytoplasm from uniform distribution of GFP-β-arrestin2 to punctate accumulation on the cell surface. This reaction was observed with isoproterenol and to a lesser extent with salbutamol, but not with C1-S agonists (Figure 5G). Finally, it has been determined (25) that β2ARs activate ERK1/2 through dual independent mechanisms: cAMP/PKA-Gas-dependent pathways and β-arrestin-dependent, Gas-independent pathways. Therefore, cAMP-independent activation by β2AR of ERK1/2 serves as an alternative measure for agonist-promoted β-arrestin binding to the receptor. Therefore, we measured ERK1/2 phosphorylation in HASM cells treated with the PKA inhibitor H89 to isolate the β-arrestin component. Both isoproterenol and salbutamol activate ERK1/2 under these conditions (Figure 5H), consistent with the previously described agonist-dependent, β-arrestin-mediated events (25). The signal of salbutamol is smaller than that of isoproterenol, which supports the less effective β-arrestin signal of salbutamol measured by other methods. In contrast, consistent with the other three different assays mentioned above, C1-S at a concentration of 300 μM failed to promote ERK1/2 phosphorylation at baseline (Figure 5H). In summary, these experiments used different methods to determine the binding of β2AR to agonist-dependent β-arrestin, and all indicated that C1-S is clearly biased towards β-arrestin signaling.

Use data from cAMP and β-arrestin ECA measurements for isoproterenol, C1-S, and C5-S (Figure 5I) to create a deviation graph. The isoproterenol curve represents a complete, balanced agonist. By examining the C1-S response, it is clear that cAMP production occurs at a dose where β-arrestin is least or undetected (see also inset), which distinguishes C1 from isoproterenol because it favors β-inhibition Protein is involved. Qualitatively, C5 appears to be balanced for both responses (Figure 5I). To quantitatively assess agonist bias, we used the ECA method to measure β-arrestin participation and the bioluminescence resonance energy transfer (BRET2)-based method (35) ("TRUPATH", see method). Here, the cells are transfected to express β2AR, RLuc8-Gαs (energy donor), Gβ and Gγ-GFP2 (energy acceptor). As the Gs heterotrimer dissociates, the agonist exhibits a dose-dependent decrease in BRET2 (35). Although both the cAMP test and the BRET2 test were performed with transfected (overexpressed) cells, the latter showed less systematic bias (SI appendix, Figure S5A). The transduction coefficient of β2AR signal to the agonists of the two pathways is used to calculate β with isoproterenol as the balanced reference agonist, by the logical equivalence method (32, 36) (see methods and SI appendix, SI extension method ). The results are summarized in Figure 5J and SI Appendix, Figure S5 B and C. Using this measurement, β=0 for balanced (unbiased) agonists. When β> 0, the Gs signal is more favored than the β-arrestin signal, and β <0 indicates that the β-arrestin signal is more inclined than the Gs signal. Therefore, it was found that salbutamol is balanced (Figure 5J), which is consistent with previous reports, such as the report of Rajagopal et al. (32). Since PLA and ECA (and confocal recruitment and ERK1/2 studies) do not have β-arrestin binding to C1-S, β cannot be defined for the agonist itself (see equation in the SI appendix, SI extension method). However, since C1-S stimulates cAMP (Figure 3) and effectively relaxes HASM (see next section), there is no detectable β-arrestin signaling. The compound appears to be biased towards β-arrestin while retaining the resistance to Gαs. Signal conduction, and therefore, β>0. It was found that the related compound C5-S showed balanced signal transduction (Figure 5J).

C1-S shows the required signal bias (Gs coupling with minimal or no β-arrestin binding), which may translate into the therapeutic effect we seek, namely HASM relaxation (activation of cAMP) without desensitization. In order to verify the lack of C1-S-mediated β-arrestin binding in the desensitization function assay, β2AR transfected cells attached to the culture dish were pretreated with carrier (control), salbutamol and C1-S for 10 minutes, washed, The cells were then stimulated with isoproterenol and cAMP levels measured 15 minutes later. As shown in Figure S6 of the SI Appendix, salbutamol pretreatment resulted in a decrease in cAMP response to isoproterenol, which was consistent with the desensitization response promoted by short-term agonists, reaching 68%. In contrast, as determined by the accumulation of cAMP promoted by isoproterenol, C1-S pretreatment was not associated with a significant loss of β2AR function. To address the physiological consequences of this apparently favorable bias, magnetic torsion cytometry (MTC) was used to quantify the changes in the cytoskeletal stiffness of HASM cells. Here, the cells are labeled with ferrimagnetic microbeads, and the decrease in cell stiffness ("relaxation") in response to the β-agonist is measured (37, 38) (see Figure 6A and method). First, we confirmed that C1-S can relax HASM cells as expected by β2AR agonists. It was found that C1-S relaxes HASM by approximately 35% from baseline, confirming the efficacy in the physiological model (SI appendix, Figure S7). In order to solve the desensitization promoted by agonists, we used our previously described protocol for these HASM cells (39), summarized in Figure 6A. Cells were treated with vehicle, salbutamol (positive β-agonist control) and C1-S for 30 minutes and 4 hours. The cells were then washed and subsequently exposed to the full βAR agonist isoproterenol to assess receptor function by real-time measurement of cell relaxation. Pretreatment with salbutamol for 30 minutes resulted in a loss of 35±6.0% of the relaxation response to isoproterenol (ie, ~35% desensitization, Figure 6B and D). In contrast, C1-S pretreatment resulted in an insignificant (12±7.9%) reduction in the subsequent β2AR response to isoproterenol (Figure 6B and D). With 4 hours of agonist pretreatment, the desensitization of salbutamol was >70%, while the cells exposed to C1-S responded to isoproterenol without a significant degree of desensitization (Figure 6 C and D). These results confirm the biased agonist phenotype of C1-S, which uses physiologically relevant functions to have endogenous expression of receptors, GRK and β-arrestin in target cells of interest.

The agonist C1-S cannot cause HASM β2AR-mediated relaxation desensitization. (A) The single-cell mechanics of HASM cells was studied using MTC. RGD-coated ferrimagnetic beads are attached to the integrin receptor. The cells are magnetized horizontally and then twisted in a vertical magnetic field. The reduction in torsional force is quantified by lateral bead displacement in response to the application of various βAR agonists added to the medium. The decrease in stiffness is related to ASM cell relaxation. The desensitization protocol is shown by the big arrow (see also method). (B and C) ALB, but not C1-S, will cause β2AR-mediated desensitization of HASM cell relaxation before exposure. HASM was pretreated with vehicle (control) or 1.0 μM ALB or 100 μM C​​1-S for 30 minutes (B) or 4 hours (C) and washed, and then the β2AR relaxation response to 10 μM ISO was measured. (D) Maximum desensitization to ALB and C1-S. ALB caused desensitization of the relaxation response under 30 minutes and 4 hours of pretreatment, while C1-S did not cause statistically significant desensitization in any pretreatment time. * P <0.01 Compare the response of vehicle pretreatment to ISO. The results are from 103 to 387 cells measured under each condition.

Although the potency of C1-S is a bit low, we still expect C1-S and C1-R to compete with 125I-iodocyanindolol (125ICYP) (40) in the binding pocket to some extent. As shown in Figure S8 of the SI Appendix, both C1-S and C1-R compete for 125ICYP on the membrane-expressed β2AR. The C1-R compound does not stimulate cAMP more than the baseline, and the Ki competing with 125ICYP is about 10 times higher than that of the C1-S (SI appendix, Figure S8). These radioligand binding competition results and the functional results shown in Figure 4 indicate that C1-R actually did bind to the receptor but failed to activate it, which means that it can act as an antagonist. To further solve this possibility, we studied the cAMP stimulation of the agonist salbutamol at its EC50 on CHW-β2 cells in the absence and presence of C1-R to determine whether the binding of C1-R is sufficient to prevent salbutamol Close to the binding site required for receptor receptor-Gas coupling. These studies show a dose-dependent blockade of the cAMP response to salbutamol, with the highest concentration of C1-R leading to complete blockade comparable to the antagonist propranolol (SI appendix, Figure S9).

In order to understand the structural basis of C1-S agonism and C1-R, we conducted a series of molecular dynamics (MD) studies based on our prediction of μ-opioid receptor (μ-OR) (11), δ-opioid Receptor (δ-OR) (41) and adenosine receptor subtypes (42, 43). For the current work, we start with the activation structure of the unbiased agonist BI167107, which is compounded with β2AR-Gs and defined by its crystal structure (Protein Database [PDB] 3SN6) (44), and then remove the agonist, leaving only the receptor Initial docking of body C1-S and C1-R structures. Then we use the anchored G protein to rebuild the receptor complex. Assuming that the imidazole functional group is protonated, the predicted structures of C1-S and C1-R docked with β2AR-G are optimized. We use partial charges from Quantum Mechanics (QM) (B3LYP style of density functional theory) and use the 6-31G** basis set of Jaguar (from Schrödinger).

We used the DarwinDock complete sampling method to predict the binding site of the active (Gαs binding) β2AR, as described in other GPCRs (42, 45). After replacing the hydrophobic residues with Ala, DarwinDock samples a large number of poses (~50,000) in the possible binding regions, and then selects the side chain direction of these hydrophobic residues for each of the best 100 poses (45). Due to the size and geometry of the atypical C1 ligand, we first docked the imidazole and urea subunits separately, as shown in Figure S10 in the SI Appendix. Then we combine the two parts of the ligand with the cyclohexane-alkyl chain to form a C1-S compound, and we use the charge from QM to minimize it (see method and SI appendix, modeling method). The binding site of the C1-R form is predicted in the same way as the C1-S. After adding C1 or C5, we reconnect the Gs protein and reinsert the ligand-β2AR-Gs complex into the lipid membrane and water tank at physiological pH and salt. We include the S-palmitoyl-cysteine ​​lipid anchor at the carboxyl end of helix 8 (SI appendix, Figure S11) and the N-terminus of Gαs. Then we performed a series of calculations (minimization and MD simulation) to predict the conformational changes promoted by C1 and C5, and provide energy-favorable receptor conformations. It is expected that there is a deviation from understanding why C1-S and C5-S are balanced. And C1-R is an antagonist. The final predicted three-dimensional (3D) structure of the activated agonist-receptor-G protein complex (called Σact) in the C1-S membrane is shown in Figure 7. The global view of the intra-membrane complex is shown in Figure 7A, while the position of C1-S and the interaction seen from the various views are shown in Figure 7 BE. These include the interaction of SB in Asp1133.32 and Phe193ECL2 and HBs with Ser2035.42 and Asn3127.39. The intra-receptor HB between Asp1133.32 and Tyr3167.43 (the same as the adrenaline-binding receptor) was also achieved (Figure 7C and D). Each system uses GROMACS to balance 800 ns with MD (see SI appendix, modeling method). The last 100 ns of Σact and C1-S and its stereoisomer C1-R are shown in the SI appendix. Figure S12 depicts the relationship between the 11 most important residues and each of the seven transmembrane (TM) regions. Time-dependent interaction energy. We found that the strong SB of C1-S to Asp1133.32 is more stable than C1-R. In addition, compared with C1-R, the HB of Ser2035.42 is more stable, and the average amplitude of C1-S is larger, as is the interaction with Asn3127.39. The Σact status of C1-S and C1-R and the pharmacophore of the unbiased agonist epinephrine are shown in Figure 8A-C, respectively. SB to Asp113 are shared by C1-S, C1-R and epinephrine. The C1-S aromatic group shown forms a π-π stacking network with Phe193ECL2 (Figure 8A), but no similar interaction of C1-R is observed (Figure 8B). Finally, C1-R did not observe HB (Figure 8B) from C1-S to Asn3127.39 (Figure 8A).

The predicted binding site of C1-S binding to β2AR is coupled to the dominant membrane and Gs in the water. (A) Minimized activated β2AR (blue)-C1-S (yellow)-Gs protein (Gαs [red], Gβ [grey] and Gγ [orange] in the phosphatidylcholine (POPC) membrane (light blue) ]). (BD) The selective interaction between C1-S and activated β2AR: (B) imidazole forms SB with Asp1133.32, forms HB with Asn3127.39, (C) urea forms HB with Ser2035.42, and π agonist The -π stacking of the two aromatic rings with each other and Phe193ECL2 is obvious, (D) Asp1133.32 interacts with the internal receptor of Tyr3167.43, and (E) the top view of the compound in the TM bag.

Comparison of the predicted binding site pharmacophores of the three ligands bound to the active state β2AR. (A) The C1-S binding site includes HB (pink arrow) to Ser2035.42 and Asn3127.39; SB (purple line) to Asp1133.32; and Pi-Pi is stacked at Phe193ECL2 with internal aromatic bonds ( green Line). (B) The main feature of the C1-R binding site is SB to Asp1133.32. (C) Adrenaline binding sites include HB to Ser2035.42, Ser2075.46, Asn2936.55, Asn3127.39 and Asp1133.32; SB to Asp1133.32; the interaction of cation-π with Phe193ECL2; and Phe2906. The π stacking interaction of 52.

Our experiment found evidence of C1-R as an antagonist. To explore this possibility computationally, we never started with apo-β2AR in the high-resolution crystal structure of engineered inactive human β2AR (PDB 2RH1) complexed with G protein. We added the S-palmitoyl-cysteine ​​lipid anchor and inserted apo-β2AR into the lipid membrane and solution, and then minimized it. Then, we inserted the C1-R ligand into apo-β2AR and minimized the ligand-Σ0 complex, resulting in the structure shown in Figure 9A (the complex depicted in the membrane) and the combination of C1-R and Apo-β2AR. The residues indicated by each view (Figure 9B and C). We performed 300 ns MD to balance the two structures. Then we performed a 0.5-μs MD simulation for each complex and averaged the interaction energy between the residues and TM in the past 300 ns. The interaction energy diagram in the SI appendix, Figure S13 shows the significantly different energy distribution between C1-S and C1-R, consistent with the 3D structure shown in Figure 3 and Figure 3. 7 and 9. For C1-R-Σ0 (SI appendix, Figure S13), we observed interactions with His2966.58, Asn2936.55, Asn3127.39, and strong aromatic interactions with Phe193ECL2 and Tyr3087.35. Except for Asn3127.39, these interactions have greater energy, or have not been observed, with C1-S-Σ0 (SI appendix, Figure S13). In addition, C1-S and C1-R both retain the combination with Asp1133.32 in the Σ0 state. In summary, these results indicate that the binding of C1-R to β2AR in the Σ0 state is more energy-efficient than C1-S, which is consistent with the experimental data showing that C1-R fails to activate but binds to the receptor, and therefore acts as A villain. The pharmacophore of C1-S and C1-R at Σ0 is shown in Figure S14 in the SI Appendix. Asp1133.32 SB was observed for both ligands. Both C1-R and C1-S also show stable aromatic bonds with Tyr3087.35. However, as shown in the SI appendix, Figure S13, the energy is favorable for C1-R binding. No interaction between C1-R and Phe193ECL2 was observed with C1-S (SI appendix, Figure S13). The interaction of the inverse agonist karaazole alcohol (2RH1) with the binding site of β2AR showed several commonalities with those we identified as inactive with C1-R, including Asp1133.32, Asn3127.39, Tyr3167. 43 HBs, and aromatic interaction with Phe2906. 52 and Tyr3087.35 (46).

The interaction between C1-R and inactive β2AR (apo-β2AR). (A) Minimized inactive β2AR (blue)-C1-S (yellow) complex in phosphatidylcholine (POPC) membrane (light blue). (B and C) The binding pocket of C1-R in inactive β2AR includes Asp1133.32, His2966.58 and the aromatic interaction with Phe193ECL2, located in the TM3-4-5-6-7 region. See also the energy diagram in Figure S13 in the SI Appendix and the pharmacophore in Figure S14 in the SI Appendix.

Compared with the biased agonist C1-S, the difference in the interaction between the endogenous and unbiased agonist adrenaline and the activated receptor is shown in Figure 8. However, adrenaline also interacts with the HB produced by this residue and the β-carbon hydroxyl group of catecholamines (Figure 8C). Both agonists interact with Asn3127.39, Ser2035.42 and Phe193ECL2. However, C1-S acts as a proton acceptor at Ser2035.42, while epinephrine acts as a proton donor at the same residue. In addition, for C1-S, two aromatic π-π stacking interactions with Phe193ECL2 were observed, while adrenaline has a cationic-π interaction with this residue from the terminal amino group of the agonist. Adrenaline also interacts with residues Ser2075.46 and TM6 residues Asn2936.55 and Phe2906.52, which are not present in C1-S.

We noticed that the difference between C1-S and the agonist C5-S is only in the R2 group (SI appendix, Figure S2), where C1-S has a cyclohexane and C5-S has a benzene. Since C1-S favors β-arrestin and C5-S does not, we simulated the interaction of C5-S with activated receptors to explore the differences between the two compounds that might explain the C1-S phenotype (SI Appendix ,figure 2). S15 A–C). We found that SB interacts with C5-S at Asp1133.32 and the monoaromatic π stacking with Phe193ECL2 is similar to C1-S. However, although C1-S shows HB with Ser2035.42, C5-S interacts with Ser2075.46. In addition, C5-S has a HB with Asn2936.55 instead of the Asn3127.39 residue of C1-S. In fact, the interaction energy between Asn-2936.55 HB and C5-S TM6 is greater than that found in C1-S (compare SI appendix, Figures S12 and S15D). In contrast, the TM7 interaction energy of C1-S is greater than that of C5-S, which is mainly the result of Asn312 interaction.

The predicted interaction of C1-S with Asn312 instead of Asn293, and the interaction of C5-S with Asn293 instead of Asn312 suggests that mutations of these two residues will have different effects depending on the agonist, which will further confirm the modeling. Asn293-to-Ala substitution results in the loss of high-affinity binding to isoproterenol, as determined by 125ICYP binding to the membrane in the absence of guanosine triphosphate (GTP) (SI appendix, Figure S16). This result is consistent with the binding of catecholamines to Asn293 (and Asn312; Figure 8). Compared with the wild-type (WT) of C5-S, the mutant Ala293 receptor also loses high-affinity binding, which is consistent with the model showing that C5-S binds to Asn293 in an activated state (SI appendix, Figure S16). It has been previously reported (47, 48) that cAMP has an impaired response to isoproterenol with the Asn293 mutation compared to the WT receptor. Our results with this agonist confirmed these observations, and we also found that the cAMP signal of the C5-S agonist was impaired (SI appendix, Figure S17). These results are consistent with modeling and competitive studies showing the important interaction between C5-S and Asn293. The C1-S compound does not show a high-affinity binding site for the WT receptor (SI appendix, Figure S16), so this specific phenotype cannot be compared with the mutant. However, compared with WT, Ala293 receptor has the same degree of competition and low affinity Ki value as C1-S. Importantly, there is no difference in cAMP response between C1-S WT and Ala293, which is in contrast to C5-S (SI appendix, Figure S17). Overall, the data indicate that C5-S is dependent on Asn293, not C1-S, and β2AR activation. Consider also conducting experiments to explore the TM7 interaction on Asn312. However, the Ala312 mutant receptor failed to bind 125ICYP or another antagonist radioligand 3H-dihydroaprolol, and Western blotting verified the cell surface expression of the mutant receptor (SI appendix, Figure S16D). In addition, compared with WT, the cAMP response of the Ala312 receptor was reduced by >80%, and the curve shifted to the right by >2-logs (SI appendix, Figure S17A). In short, the mutation at Asn312 seems to cause the antagonist and agonist binding pocket to be distorted, and the function is significantly impaired, so no further studies on this receptor have been carried out.

Agonists that act on β2AR expressed on ASM cells are currently the only direct bronchodilators that can be used to treat obstructive airway diseases (49). Given that there are more than 300 million people with asthma and COPD, β2AR may be the most common GPCR prescribed for agonists worldwide. However, over time, beta-agonist treatment has been associated with some results related to loss of effectiveness. Both short-acting and long-acting agonists are associated with rapid defense (16⇓-18), increased airway responsiveness to contraction stimuli (50, 51), loss of bronchial protection (52, 53), and increased morbidity and mortality (54⇓ ⇓ –57). Extensive cellular and biophysical studies of β2AR have determined that β-arrestin binding is the initial event that leads to rapid attenuation of Gα signaling and long-term downregulation through receptor internalization (58). Therefore, therapeutic agents that are Gas promoters but prefer β-arrestin should improve airway response and reduce rapid defense. Several biased agonists of GPCRs have been found, including those that are biased toward or away from G protein signaling or β-arrestin action. For example, the μ-OR agonist TRV130 appears to be biased towards G protein (Gαi) coupling and away from β-arrestin. It is equivalent to morphine in providing analgesia, but shows fewer targeted side effects, such as respiratory depression, which is attributed to bias (59). In fact, we found that the binding site of TRV130 on μ-OR differs from that of morphine in several key areas (11). We also noticed that on α2CAR, the agonists UK14304 and B-HT920 have almost no structural similarity, but both prefer Gi coupling rather than Gs coupling (7). Such results indicate that a similar quantitative bias of agonists for a given receptor can be achieved from compounds with different structures. It is also possible to achieve bias towards β-arrestin and away from G protein coupling, which is considered to be therapeutically beneficial for certain receptors and diseases (60). For example, TRV027 (also known as TRV120027) has been shown to be a type 1 angiotensin II receptor agonist that favors β-arrestin (60), and lacks significant signaling for Gαq. The biased activation of this receptor increases the contractility of cardiomyocytes in vitro and reduces blood pressure in vivo, but shows a preserved cardiac output (unlike unbiased agonists) (60⇓ –62). Other β-arrestin-biased agonists of this receptor, such as SII, have been shown to improve cardiac contractility (63). These effects are related to β-arrestin-mediated activation of ERK1/2, Src, Akt, PI3 kinase and endothelial nitric oxide synthase (64, 65). Given that chronic Gαq activation in the heart can exacerbate heart failure (66), this selective agonist may be helpful in the treatment of hypertension or cardiomyopathy in some cases.

In the current study, we seek a β2AR agonist that favors Gαs coupling and is far away from β-arrestin. The goal is to improve the treatment of obstructive pulmonary disease and at the same time establish a structural basis for the selective pathway of the receptor. We screened a library of 40 million compounds, sorted by scaffold structure, the library was not related to the known ligand structure of β2AR agonists, and then assessed the agonist propensity of a single compound to induce β-arrestin participation. One compound (C1-S) is particularly noteworthy due to the obvious absence of β-arrestin, and also shows the coupling of Gas and cAMP and the relaxation of ASM. Since C1-S has no detectable β-arrestin signal, the bias factor β cannot be calculated, but it is clear that C1-S has Gs coupling, and there is no β-arrestin signal within the multiple limits. It was determined that this agonist was placed in a strong preference for β-arrestin events. The functional evidence confirming this biased phenotype is minimal short-term, agonist-promoted desensitization. This was tested in two different ways in our research. First, we measured any loss of β2AR cAMP response after pretreatment with C1-S or the positive control agonist salbutamol. Although the balanced agonist salbutamol caused desensitization, C1-S did not (SI appendix, Figure S6). In the second confirmation of the C1-S biased phenotype, we studied the biologically relevant physiological response (relaxation) of HASM cells, whose β2AR is a therapeutic target of β-agonists. We verified that C1-S is an agonist in this system, and then found that pretreatment with C1-S for 30 minutes (or 4 hours) does not cause desensitization of the relaxation response, while salbutamol promotes the loss of β2AR function (Figure 6 ). In the third confirmation, we evaluated another functional result of β2AR-mediated β-arrestin binding and cAMP-independent activation of ERK1/2. Traditional full and partial agonists promote ERK1/2 activation, while C1-S does not (Figure 5H). In summary, the results of three different types of experiments measuring the interaction between β-arrestin and β2AR (Figure 5A-H) and the results of three different types of experiments measuring the functional consequences of β-arrestin binding (Figures 5H and 6 and SI Appendix, Figure S6) is consistent. Therefore, we are very confident to conclude that C1-S favors β-arrestin, but still maintains sufficient Gs coupling to relax ASM cells. As far as we know, this bias of β2AR agonists has not been reported before. Salmeterol and formoterol are long-acting β-agonists with extended R group substitutions. Studies have been found to favor β-arrestins (8, 32), which is not considered clinically in the current situation Is advantageous. The detected differences may explain some of the differences related to salmeterol (8, 32, 47). Carvedilol is classified as a β1/β2-antagonist and has inverse agonist activity in β2AR. It weakly recruits β-arrestin but does not stimulate cAMP (34). By definition, this can be called "bias", but in any case, it is against β-arrestin.

Our MD simulations revealed the favorable energy conformation of C1-S, and strong interactions with Asp1133.32 and Ser2035.42, both of which were observed with agonists that activate Gs coupled with β2AR (44, 67, 68) . C1-S shows HB with Ser2035.42, but without Ser2075.46, which also interacts with epinephrine (44, 67, 68). This lack of Ser2075.46 HB may change the inward bulge of TM5 that occurs when combined with adrenaline, which is related to the outward movement of TM6 (44, 67). Adrenaline also binds to Phe2906.52 in TM6, but C1-S does not. In addition, active adrenaline-binding receptors show HB between Asn2936.55, Ser2035.42, Ser2075.46 and the hydroxyl group of catecholamines, and within the receptor between Ser2045.43 and Asn2936.55. These HBs have been proposed as part of the polar network of this balanced agonist (47). In contrast, the C1-S binding receptor only shows Ser2035.42 HB. Interestingly, mutation of Ser2045.43 to Thr or Ala can disrupt the network by eliminating HB in the receptor, significantly reducing the binding of adrenaline-promoted β-arrestin to β2AR, and has little effect on Gs coupling (47 ). Therefore, the C1-S signal transduction phenotype may also be affected by the atypical polarity network between TM5 and TM6 because it lacks HB with Asn2936.55 and HB in the receptor. In our research on the substitution of R groups on selected scaffolds, we found a compound (C5-S) similar to C1-S, but showed no bias. The only difference between these two compounds is the R2 group (benzene or cyclohexane respectively). Comparing the interaction of the C5-S agonist with activated β2AR with the C1-S interaction, we found that the C5-S agonist may be redirected due to the greater exposure of benzene to the solvent. This seems to be because the imidazole of C5-S has a strong simultaneous interaction with Asp1133.32 in TM3 and Asn2936.55 in TM6, which draws TM3 and TM6 on the upper part of the receptor closer. C1-S works in a similar way to imidazole, but it combines with Asp1133.32 and Asn3127.39 to connect TM3 to TM7 instead of TM6. The end result is that the C1-S position is far away from the TM3-5-6 pocket formed with C5-S, which may lead to the difference in the activation of β-arrestin promoted by the two compounds.

In short, by screening a library of non-targeted scaffolds containing 40 million compounds, we identified the C1-S β2AR agonist, which favors β-arrestin participation, but still actively couples with Gs, stimulates cAMP and relaxes ASM. We expect that if an agonist can be found that can selectively separate β-arrestin and Gs signals to avoid the effect of β-arrestin, then it will be in the absence of the rapid response observed with typical β-agonists Work under. In fact, studies of C1-S and cAMP accumulation or HASM cell relaxation have shown the lack of rapid defense, which represents a beneficial pharmacological effect in the treatment of obstructive lung disease. The basis for this deviation may be multiple factors, including the loss and increase of binding sites that may affect the hydrophobic network, and the positioning of the agonist in the binding pocket away from TM6.

CHW cells are transfected to stably express human β2AR under G418 selection and are maintained in a monolayer as previously described (69). Use Lipofectamine 2000 (Invitrogen) as previously described (31, 70, 71) to transiently transform HEK-293T cells with constructs expressing human β-arrestin2-myc, human β-arrestin2-GFP, human β2AR and β2AR-GFP Stain and study after 48 hours. CHO cells stably expressing modified β2AR and modified β-arrestin2 for ECA were purchased from DiscoverX. In some experiments, these cells were transiently transfected with 5 μg GRK2 cDNA, as described (72). Primary HASM cells were derived from donor lungs obtained from the National Disease Exchange Registry, maintained as a single layer as previously reported (37), and used for physiological studies between the 3rd and 6th passages. Other experiments used the D9 telomerase reverse transcriptase immortalized HASM cell line, as described (70).

The SR and PS libraries consist of a mixture of compounds in the wells and are designed according to the previously detailed report (26, 29, 30). The SR library contains about 40 million compounds, and the PS library contains fewer compounds (the compound of each sample is more Less) Okay), based on the selected candidate SR library samples. The composition of the SR library has been previously published (26). The PS library is custom-designed by selecting multiple parts at each R1, R2, and R3 position and then synthesizing all of these combinations for the next round of cAMP stimulation test. We sorted the samples according to each of the three positions of the cAMP reaction, and evaluated the results of the structure-activity relationship trend through calculation methods (73).

For radioligand competition studies, cell membranes expressing β2AR were incubated with 125ICYP (40 pM) in 75 mM Tris (pH 7.4), 12 mM MgCl2, and 2 mM ethylenediaminetetraacetic acid (EDTA) buffer, as shown , With or without 100 μM GTP and as described in (69), different concentrations of compounds were kept at 25 °C for 1.5 hours. Using a cell harvester (Brandell), at 4 °C, through a glass fiber filter and 5 mM Tris (pH 7.4) and 2 mM EDTA buffer, the bound 125ICYP and free radioactive ligand were separated by vacuum filtration. For cAMP experiments, β2AR transfected and untransfected cells were inoculated into serum-free Dulbecco's modified Eagle's medium at 20,000 cells per well, and the phosphodiesterase inhibitor 3-isobutyl was used the next day -1-Methylxanthine (100 μM, 30 minutes). By exposing to various reagents at 37°C for 10 minutes, cAMP is produced from the cells, and the reaction is stopped by cell lysis. As described in (15), cAMP was measured by a competitive immunoassay (Molecular Devices). By first treating the cells with the PKA inhibitor H89 (10 μM), then passing the cells at 37 °C with the carrier, 10 μM isoproterenol, 300 μM salbutamol or 100 μM C​​1-S and the protein to pass 12% ten. Sodium dialkylsulfate gel electrophoresis was separated and transferred as described (74).

PLA (Duolink, Sigma) was performed according to the most recent description (31). In short, HEK-293T cells were transfected on coverslips with β2AR-GFP and β-arrestin2-myc. Each labeled primary antibody is incubated with the transfected cells, and then a pair of oligoconjugated secondary antibodies are added to the culture dish. When the oligonucleotides of two antibody couplings are very close, the hybrid oligonucleotides connect them. The ligase generates a signal after forming a circular DNA, which is then amplified by rolling circle PCR. The fluorescently labeled oligonucleotide hybridizes to the product. The cells were exposed to the vehicle or the specified concentration of agonist at 37°C for 10 minutes and fixed with 4% paraformaldehyde. The cells are imaged by a fluorescent confocal microscope. ECA (PathHunter, DiscoverX) is performed as another method for detecting receptor β-arrestin interactions (32). In short, attached CHO cells were studied in 96 wells, stably transfected to express β-arrestin 2 fused to β-galactosidase lacking peptide fragments, and a complementary β- Β2AR labeled with a galactosidase fragment. plate. The agonist and the cells were incubated for 30 minutes at 37°C. After the agonist promotes the binding of β-arrestin to the receptor, β-galactosidase is recombined to produce an active enzyme, which can be detected by the luminescence on the FlexStation3 microplate reader. The recruitment assay based on β-arrestin-GFP was performed as described above (15).

The activation of Gαs by β2AR was quantified by using BRET2 to determine the dissociation of Gs heterotrimers. Use Lipofectamine 2000 to transfect HEK-293T cells, which contains a cDNA construct encoding β2AR, which is a short form of Gαs fused with RLuc8, Gβ, and GFP2 (Addgene) at a ratio of 1:1:1:1, as described by Olsen述等。 Said and so on. (35). The next day, they were taken out and seeded on a 96-well plate, 48 hours after transfection, RLuc8 substrate was added, and the cells were treated with multiple doses of agonist at room temperature in quadruplicate for 5 minutes. The Rluc8 signal was collected on FlexStaion3 at 395 nm, and the GFP2 signal was collected at 510 nm. BRET2 is calculated as the ratio of GFP2 to RLuc8 signal.

We use MTC to measure the dynamic changes in cytoskeletal stiffness as a substitute for agonist-induced relaxation of single cells, as we have previously verified (32, 33, 37, 38). Ferrous magnetic microbeads (4.5 μm in diameter) coated with synthetic peptides containing Arg-Gly-Asp (RGD) are connected to cell surface integrin receptors to form adhesion spots and are closely connected to the underlying cytoskeletal network (Figure 6A). The beads are magnetized horizontally to the cell plating, and then twisted in a vertically arranged magnetic field, which changes sinusoidally with time. Forced bead movement is optically detected with a spatial resolution of ~5 nm, and their changes are monitored in real time in response to β-agonists (HASM cell relaxation). The maximum relaxation change from baseline observed at any point in the time course was used to quantify the maximum HASM response to isoproterenol.

We first merged the data from the β2AR crystal structure (PDB 3SN6). The DarwinDock complete sampling method is used to predict the binding sites and energetics of C1-S, C1-R and C5-S binding to β2AR, as we have previously described for other GPCRs (41⇓ –43, 45) (SI appendix, SI extension method). In order to define the site and energetics of the activated receptor (Σact), the ligand-receptor-Gs complex is modeled, and for the inactive state (Σ0), the ligand-receptor complex without the G protein is modeled The object is modeled. As described in (41), use GROMACS (Uppsala University) for MD simulation, and the optimization is shown in the SI appendix, SI extension method.

The cAMP values ​​from the screen were compared by ANOVA, and then a post hoc t test was performed using Tukey's correction for multiple comparisons. Iterative four-parameter least squares logistic regression is used to obtain the base (Rmin), the maximum value (Rmax), the concentration that leads to the half-maximum response (EC50), the Hill coefficient and the fitting of each concentration-response curve. R2 uses Prism (GraphPad) Fit the sigmoid curve. The Emax (Rmax-Rmin) and EC50 values ​​were compared by t-test using the same software. For PLA, ECA, and Gs activation studies, the data fits a three-parameter logistic function, and the Hill coefficient is set to 1.0. Calculate the deviation factor from the Gαs activation and ECA data using the logical equivalent method (Method 3 in Reference 32, see the equation in the SI appendix, SI extension method). The other data shown were compared with Tukey's correction for multiple comparisons by t-test. A P value of <0.05 is considered statistically significant. Unless otherwise stated, data from multiple experiments are shown as mean ± SE.

All research data is included in the article and/or SI appendix.

We thank Michel Bouvier for helping with the construction of Ala293, Himeshkumar Patel and Christine Tam for helping with manuscript preparation, and NIH for supporting grants HL045967, HL155532 and HL114471 (to SBL).

↵1D.K. and AT have made the same contribution to this work.

↵2 Current address: Biophysics Project, University of California, San Francisco, California 94102.

Author contributions: DK, AT, LKL, NK, MNB, MAG, J.-AAW, SSA, WAG and SBL design research; DK, AT, LKL, HRS, NK, MNB, MAG, J.-AAW, SSA, WAG and SBL conducted research; DK, AT, MAG, AM, J.-AAW, SSA, WAG and SBL contributed new reagents/analysis tools; DK, AT, LKL, HRS, NK, MNB, MAG, AM, J.-AAW, SSA, WAG and SBL analysis data; DK, AT, MAG, AM, J.-AAW, SSA, WAG and SBL wrote this paper.

The author declares no competing interests.

This article is directly contributed by PNAS.

This article contains online support information https://www.pnas.org/lookup/suppl/doi:10.1073/pnas.2026668118/-/DCSupplemental.

This open access article is distributed under Creative Commons Attribution 4.0 (CC BY).

Thank you for your interest in advertising on PNAS.

Note: We only ask you to provide your email address so that the people you recommend the page to know that you want them to see it and that it is not spam. We do not capture any email addresses.

Feedback privacy/legal

Copyright © 2021 National Academy of Sciences. Online ISSN 1091-6490. PNAS is a partner of CHORUS, COPE, CrossRef, ORCID and Research4Life.