Structural basis of α1A-adrenergic receptor activation and recognition by an extracellular nanobody

The α1A-adrenergic receptor (α1AAR) belongs to the family of G protein-coupled receptors that respond to adrenaline and noradrenaline. α1AAR is involved in smooth muscle contraction and cognitive function. Here, we present three cryo-electron microscopy structures of human α1AAR bound to the endogenous agonist noradrenaline, its selective agonist oxymetazoline, and the antagonist tamsulosin, with resolutions range from 2.9 Å to 3.5 Å. Our active and inactive α1AAR structures reveal the activation mechanism and distinct ligand binding modes for noradrenaline compared with other adrenergic receptor subtypes. In addition, we identified a nanobody that preferentially binds to the extracellular vestibule of α1AAR when bound to the selective agonist oxymetazoline. These results should facilitate the design of more selective therapeutic drugs targeting both orthosteric and allosteric sites in this receptor family.

Recent progress in the structural characterization of GPCRs including the adrenergic receptors has clarified the mechanism of ligand recognition and G protein activation. The βARs are extensively well-characterized GPCRs and a number of structures have been determined in the active and inactive states [8][9][10][11][12][13] . Moreover, recent structures of active and inactive α 2A AR 14,15 , active α 2B AR-G i/o 16 and inactive α 2C AR 17 demonstrated the subtype selectivity of ligand recognition between α 2 ARs and βARs. However, little is known about the structure and mechanism of activation of the α 1 AR subtypes. The only available structure is the inactive α 1B AR structure 18 . Here we present three cryo-electron microscopy (cryo-EM) structures of active α 1A AR bound to oxymetazoline and the endogenous agonist noradrenaline, along with inactive α 1A AR bound to tamsulosin. We also discovered a nanobody (a single domain antibody) Nb29 that binds to the extracellular vestibule of the agonist-binding pocket. Antibodies against GPCRs have attracted particular interest for pharmaceutical applications 19 , and are useful research tools for stabilizing GPCR conformations for structural analysis 20,21 . Nevertheless, only a few class A GPCR structures in complex with the extracellular antibody are available [22][23][24][25][26] . These findings may guide the development of more effective drugs for the α 1A AR.

Structure determination of active and inactive α 1A AR
We expressed human α 1A AR in baculovirus-infected Spodoptera frugiperda (Sf9) insect cells. We constructed a variant of human α 1A AR lacking residues 371-466 in the C-terminus, and three N-linked glycosylation sites (N7Q, N13Q and N22Q) in the N-terminus were mutated. To further stabilize the receptor, we discovered a conformationally selective nanobody from a library of synthetic nanobodies displayed on the surface of Saccharomyces cerevisiae 27 . After two rounds of magnetic-activated cell sorting (MACS) and four rounds of fluorescence-activated cell sorting (FACS) with oxymetazoline-and tamsulosin-bound α 1A AR, we identified Nb29 as the most enriched clone ( Fig. 1a and Supplementary Fig. 2; See Method). An on-yeast titration assay indicated that Nb29 has selectivity for oxymetazolinebound α 1A AR compared with apo, noradrenaline-, and antagonist (tamsulosin and phentolamine)-bound states (Fig. 1b and Supplementary Table 1a). In a ligand binding assay, Nb29 induced the left shift of the agonist competition curves for α 1A AR over α 1B -and α 1D AR ( Fig. 1c and Supplementary Table 1b). Although Nb29 on its own competed for the antagonist [ 3 H]prazosin binding for α 1A AR and might affect the competition binding results, we observed a larger left shift of oxymetazoline competition curves than those for noradrenaline, which is in agreement with the on-yeast titration result (Fig. 1d, e and Supplementary Table 1c).
To solve the structure, we formed the α 1A AR complex with Nb29 and obtained the initial cryo-EM structure at 4 Å resolution, and found that Nb29 binds to the extracellular region of the α 1A AR. To further stabilize the receptor, we used α 1A AR construct fused with a minimal-T4 lysozyme (mT4L) in the intracellular loop 3 (ICL3) 28 ( Supplementary Fig. 3a), and formed a complex with engineered minimal Gsq protein (miniGsq) in which mini-Gs was substituted with the 15 residues of carboxyl-terminal α5 helix of Gq protein 29 (Methods and Supplementary Fig. 3b, c). The α 1A AR in complex with the heterotrimeric G q/11 proteins was not stable enough for structure determination. Finally, we obtained the cryo-EM structures of active α 1A AR bound with oxymetazoline (Nb29-α 1A AR-miniGsq) at a global resolution of 2.9 Å (Fig. 2a, Table 2). The cryo-EM map allowed the model building of most of the regions with a clear electron density for the ligand. The map density of the extracellular region of α 1A AR is relatively clear because of the bound Nb29, whereas there is poor density for the fused mT4L due to map refinement by masking out the mT4L. Subsequently, we also solved the Nb29-α 1A AR-miniGsq complex bound with noradrenaline at a global resolution of 3.5 Å (Fig. 2c, Table 2). Although the map quality allowed the model building of the receptor with clear electron density for the ligand, the map density for Nb29 is weak. The weaker map density is in agreement with the observation that Nb29 preferentially binds to oxymetazoline-bound receptors over noradrenaline-bound receptors (Fig. 1b). The resolution of the Nb29dissociated α 1A AR-miniGsq complex is much lower (~6 Å) than that of the Nb29-bound complex ( Supplementary Fig. 3h).
To solve the inactive α 1A AR structure, our crystallographic and cryo-EM experiments using fusion protein and/or the other nanobodies selected from the synthetic nanobody library were unsuccessful. Thus, we utilized a recently described engineering strategy to enable the binding of nanobody 6 (Nb6) that engages the ICL3 of the inactive-state κ-opioid receptor (kOR) 21 . Based on the cryo-EM structure of engineered neurotensin receptor 1 (NTSR1)-Nb6 complex, we swapped the same site from C205 5.59 of transmembrane (TM) 5 (T247 5.59 of kOR) to G275 6.38 of TM6 (L277 6.38 of kOR) including intracellular loop (ICL) 3 (superscripts indicate Ballesteros-Weinstein numbering for GPCRs 30 ). In addition, we introduced two thermostabilizing point mutations, S113R 3.39 to mimic allosteric sodium ion binding 31 , and M115W 3.41 to increase expression 32 . These mutations were used for solving other inactive GPCR structures such as prostaglandin E receptor EP4 for the R 3.39 mutation 22 , and dopamine D2 receptor for double mutation of positions 3.39 and 3.41 25 . In radioligand binding studies, the α 1A AR-kOR mutant showed enhancement of [ 3 H]prazosin binding by Nb6, but did not significantly alter the binding affinities for the agonist oxymetazoline or the antagonist tamsulosin ( Supplementary Fig. 5a, b). We formed the α 1A AR-Nb6 complex and solved the cryo-EM structure of the inactive α 1A AR bound to tamsulosin at a global resolution of 3.3 Å resolution.  Table 2). The cryo-EM map allowed the model building of most of the regions with a clear electron density for tamsulosin; the map density of the inactive α 1A AR-Nb6 complex is relatively weak for ECL2 and well-defined in the cytoplasmic region including Nb6 binding domain where ICL3 and the cytoplasmic sides of TM5 and 6 were exchanged to those of the kOR. Although Nb6 binds to a similar site as in the kOR-Nb6 complex 33  Nb29-α 1A AR-miniGsq and tamsulosin-bound α 1A AR-Nb6 complexes. The cryo-EM density maps and structure models of the Nb29-α 1A AR-miniGsq complexes bound to the agonists oxymetazoline (a, b) and noradrenaline (c, d), and α 1A AR-Nb6 complex bound to the antagonist tamsulosin (e, f). The detergent micelle (a, c, e) and unmodelled mT4L (b) are shown in gray. The densities of the ligands (shown as sticks) are depicted as surfaces. Color code for the proteins is as follows: oxymetazoline-bound active α 1A AR (blue), noradrenaline-bound active α 1A AR (green), inactive α 1A AR (orange), miniGsq (pink), Nb29 (yellow), and Nb6 (purple). Small molecules are colored as follows: oxymetazoline in magenta, noradrenaline in gray, and tamsulosin in cyan.
(Supplementary Fig. 6a-e). The active α 1A AR structures enable us to model a putative cholesteryl hemisuccinate (CHS) molecule bordering TM3-5 in the active structures, in contrast, we observed only a weak density in the inactive α 1A AR structure, since the side chain of M115W 3.41 overlaps the regions corresponding to the lipid tail of CHS ( Supplementary Fig. 6g, h). The side chain of S113R 3.39 is located in the putative sodium ion binding site as designed ( Supplementary  Fig. 6i, j) 22,31 . Structural comparison of active and inactive states of α 1A AR exhibits 14.5 Å outward displacement of an intracellular segment of TM6 that is characteristic of receptor activation (Fig. 3a-c). The TM6 movement is accompanied by a small rotation of the helix, as well as inward movements of TMs 3, 5, and 7 toward TM6. α 1A AR also exhibits other characteristics of the activation of class A GPCRs 9,14,16 . We observed the displacement of the side chain of W285 6.48 (Fig. 3d), a highly conserved residue that contributes to conformational changes associated with activation for some GPCRs. We also observe conformational changes in the conserved PIF (P196 5.50 I114 3.40 and F260 6.44 ) interaction, as well as the NPxxY (N322 7.49 , P323 7.50 , and Y326 7.53 ) and DRY (D123 3.49 , R124 3.50 and Y125 3.51 ) motifs (Fig. 3d, e). In the active α 1A AR, R124 3.50 forms hydrogen bond networks with Y204 5.58 , Y326 7.53 and C329 7.56 (Fig. 3e). These structural changes allow the C-terminal helix (α5 helix) of Gα to engage the receptor core, as described below. In the extracellular view, due to the Nb29 binding, the conformation of ECL2, TMs 4 and 7 in the active state is closer to the receptor core, as discussed later in detail.
Orthosteric ligand-binding pocket of α 1A AR Extensive site-directed mutagenesis studies have identified amino acids that form the binding pocket of the α 1A AR, including residues responsible for subtype selectivity [34][35][36][37][38][39][40][41][42] . Our structures largely confirm these observations. α 1A AR structures bound to the endogenous agonist noradrenaline, selective partial agonist oxymetazoline, and selective antagonist tamsulosin are shown in Fig. 4. All three ligands form polar interactions with D106 3.32 which is a highly conserved residue involved in ligand binding in all aminergic receptors ( Fig. 4 and Supplementary Fig. 7). The binding pocket of noradrenaline is formed by residues in TMs 3, 5, 6, and 7 (Fig. 4a, e). The noradrenaline has two catechol hydroxyl groups. The para-hydroxyl forms a hydrogen bond with S188 5.42 , whereas meta-hydroxyl does not form polar interaction but is close to M292 6.55 , a unique residue among ARs ( Supplementary Fig. 7). Previous mutagenesis [34][35][36] and [ 13 C ε H]methionine labeling NMR studies 37 support the role of S188 5.42 and M292 6.55 in ligand binding. The chiral β-hydroxyl forms a hydrogen bond with D106 3.32 and the amino group of the noradrenaline forms cation-π stacking with the phenyl ring of F312 7.39 and a hydrogen bond with the backbone carbonyl of F312 7.39 . Noradrenaline forms extensive nonpolar interactions with highly conserved aromatic residues among ARs, including Y184 5.38 , F288 6.51 , and F289 6.52 .
The oxymetazoline binds in a similar site (Fig. 4b, f), but the paracatechol hydroxyl is replaced by tertiary butyl, leading to the lack of polar interaction with S188 5.42 , which may account for its partial agonism in receptor activation 12  extensive van der Waals interaction with Y184 5.38 , V185 5.39 , S188 5.42 , M292 6.55 and the backbone carbonyl of N179 ECL2 . Position 5.39 is a valine in α 1A AR, α 2A AR, and β 3 AR, but is an alanine/isoleucine in other subtypes of ARs ( Supplementary Fig. 7). A previous mutagenesis study showed that V185A 5.39 and M292L 6.55 mutations resulted in decreased oxymetazoline binding but not noradrenaline binding, and the equivalent A204V 5.39 and L312M 6.55 mutants of α 1B AR increased agonist binding 35,36 . In place of the chiral β-hydroxyl and the amino group of noradrenaline, oxymetazoline has an imidazoline ring which forms polar interactions with D106 3.32 and C110 3.36 , π-π stacking with F312 7.39 , and aromatic interactions with W285 6.48 , F288 6.51 and Y316 7.42 . In α 1 -and α 2 ARs, C110 3.36 and F312 7.39 are conserved residues and involved in ligand recognition for imidazoline-type agonists [14][15][16] . Consistent with this result, a mutagenesis study indicated that F312A 7.39 and F312N 7.39 (the equivalent residues for βARs) mutations decreased oxymetazoline binding but not adrenaline binding 38 . This report also demonstrated that F308 7.35 , a residue above the F312 7.39 , influences oxymetazoline binding, even though it is too far away for a direct interaction 38 .
The antagonist tamsulosin has two aromatic groups on each side of an ethyl-aminopropyl backbone (Fig. 4c, g). Similar to agonist binding, the ethylamine group of tamsulosin forms a hydrogen bond with D106 3.32 . The ethoxyphenoxy group forms van der Waals interaction with V107 3.33 , S188 5.42 , M292 6.55 , F288 6.51 and F289 6.52 (Supplementary Fig. 7). Unlike agonist binding, reorientation of F312 7.39 enlarges the binding pocket and enables the antagonist to bind towards the extracellular vestibule, which is also called an exosite or secondary binding pocket with less conserved residues compared to the orthosteric pockets 12,43 (Fig. 4d). The sulfonamide group forms a polar interaction with the backbone carbonyl of C176 45.50 and a nonpolar interaction with F308 7.35 . The methoxybenzene group forms nonpolar interactions with F86 2.64 , E87 2.65 , W102 3.28 , F312 7.39 and the backbone carbonyl of S83 2.61 ( Fig. 4c and Supplementary Fig. 7). F86 2.64 is a unique residue to α 1A AR relative to other ARs and was previously identified as a determinant for the interaction of the α 1A AR with various antagonists including HEAT and prazosin 39-41 ( Supplementary  Fig. 1). In addition, another mutation study indicated that three nonconserved residues (Q177 45.51 , I178 45.52 , N179 45.53 ) in ECL2 are responsible for the α 1A AR selectivity of phentolamine and WB4101 over α 1B AR 42 , but are not involved in binding tamsulosin.

Ligand recognition of adrenergic receptor subtypes
Although all adrenergic receptors are activated by endogenous adrenaline and noradrenaline, their binding pockets are not identical. Comparisons of the key residues of the noradrenaline binding pockets in α 1A AR, α 2A AR 14 and β 1 AR 11 reveal similar but different mechanisms of noradrenaline recognition (Fig. 5a, b). Compared to M 6.55 in α 1A AR, Y 6.55 in α 2A AR (conserved in all α 2 ARs) forms a hydrogen bond with the meta-hydroxyl of the catechol ring, while the para-hydroxyl is involved in hydrogen bonds with S 5 TM3 TM3 TM3   TM4 TM4 TM4   TM5 TM5 TM5   TM6 TM6 TM6  TM7 TM7 TM7   TM3 TM3 TM3  TM4 TM4 TM4   TM6 TM6 TM6   TM7 TM7 TM7   TM2 TM2 TM2   TM1 TM1 TM1   TM3 TM3 TM3  TM4 TM4 TM4   TM5 TM5 TM5   TM6 TM6 TM6  TM7 TM7 TM7   TM2 TM2 TM2   TM1 TM1  : Highly conserved residues among all ARs : Cation-π or π-π interaction : Polar interaction : Highly conserved residues among α 1, 2 ARs the pose of the catechols. In α 1A AR, the β-hydroxyl group interacts with D 3.32 , and the amino group forms a hydrogen bond and a cation-π interaction with F 7.39 (Figs. 4a and 5a). In contrast, only the amino group forms hydrogen bonds with both D 3.32 and Y 7.43 in α 2A AR. The noradrenaline binding pose of β 1 AR is different from that of α 1A AR (Fig. 5b). The meta-hydroxyl forms polar interaction networks with S 5.42 , S 5.43 and N 6.55 , and the para-hydroxyl forms a hydrogen bond with S 5.46 in β 1 AR. Previous α 1A AR mutagenesis studies indicated that double mutation of S188A 5.42 and S192A 5.46 decreased agonist binding rather than S188A 5.42 or S192A 5.46 alone 34 . Non-aromatic N 7.39 interacts with the amine group of noradrenaline through the polar interaction networks with D 3.32 and Y 7.43 in β 1 AR. Moreover, the bulkier F 45.52 in ECL2 of β 1 AR (conserved in all βARs) forms a non-polar interaction with noradrenaline 10,11 .
We next compare the binding pose of imidazoline-type partial agonist oxymetazoline in α 1A AR and α 2A AR 14 (Fig. 5c). The oxymetazoline presents high selectivity for α 1A AR and α 2A AR over other αARs 4 . As mentioned before, oxymetazoline replaces the para-hydroxyl with the hydrophobic tertial-butyl group which no longer forms the polar interaction but engages in hydrophobic interactions with partially conserved V 5.39 in α 1A AR. In α 2A AR, oxymetazoline also does not form polar interaction with TM5, but C 5.43 is involved in the hydrophobic interaction. This position is A189 5.43 in α 1A AR, while S 5.43 or C 5.43 in other ARs (Supplementary Fig. 7). In both α 1A AR and α 2A AR, the imidazoline ring is stabilized by π-π stacking with F 7.39 along with polar interaction with D 3.32 , in contrast, C 3.36 (conserved in all αARs and V in βARs) forms weak polar interaction with the imidazoline ring only in α 1A AR. Although most of the residues have the same orientation between noradrenaline and oxymetazoline binding in α 1A AR, the orientation of Y 6.55 is shifted in α 2A AR. This Y 6.55 is involved in G i/o -biased signaling over βarrestin recruitment for oxymetazoline among the other agonists such as noradrenaline, brimonidine and dexmedetomidine in α 2A AR 14 .
Compared to α 1 AR agonists, the α 1 AR antagonists have higher subtype-selectivity because they extend to the extracellular vestibule (Fig. 5d). As mentioned above, inactive α 1A AR bound to tamsulosin reveals that unique (F86 2.64 , and M292 6.55 ) and partially conserved (S83 2.61 , E87 2.65 , W102 3.28 , I178 45.52 , and F312 7.39 ) residues are involved in subtype selectivity (Fig. 3c, f and Supplementary Fig. 7). Recent crystal structure of inactive α 1B AR bound to its selective inverse agonist (+)-cyclazosin, along with the chimeric α 1B AR-α 2C AR mutagenesis studies indicated that non-conserved residues L 2.64 , W 3.28 , A 3.29 , V 45.52, and L 6.55 are important for the selectivity in α 1B AR 18 (Supplementary Fig. 7). The (+)-cyclazosin is a derivative of prazosin in which piperazinyl quinazoline scaffold is introduced in a bulky cycloaliphatic group ( Supplementary Fig. 1), leading to 100-1000 fold selectivity for α 1 ARs over α 2 ARs, and a slight preference for α 1B AR over α 1A AR. When comparing the α 1A AR and α 1B AR (Fig. 5e), both the tamsulosin and (+)-cyclazosin extend into the extracellular vestibule. Tamsulosin interacts with F86 2.64 more closely than the furan group of (+)-cyclazosin, which is consistent with the binding selectivity 44   stabilizing mutation of F 7.39 to L 7.39 in the α 1B AR structure might affect the (+)cyclazosin binding mode 18 . In contrast to α 1 ARs, the antagonists of α 2A AR and β 2 AR do not interact with TM2 (Fig. 5e, f) 8,15 . The positions 2.64, 3.28, 3.29, and 45.52 are different from α 1 ARs and likely involved in the ligand selectivity. In addition, residues M 6.55 in α 1A AR and N 6.55 in β 2 AR allow antagonist interactions with the extracellular side of TM6, in contrast to the bulkier Y 6.55 in α 2A AR.
It is known that the other α 1A AR ligands such as A61603 (agonist) and silodosin (antagonist) have high selectivity for α 1A AR (Supplementary Fig. 1) 37,44 . In these compounds, the phenyl rings corresponding to catechol have much bulkier substituents, suggesting that they may exhibit selectivity through interaction with α 1A AR unique residues such as M292 6.55 , A189 5.43 and the non-conserved residue V185 5.39 .

Structural insight into Nb29 binding
Nb29 binds to the extracellular side of α 1A AR which is topologically distinct from the orthosteric agonist pocket (Fig. 2). This site has been shown to bind to allosteric modulators for muscarinic receptors [45][46][47][48] . We do observe a left shift of the agonist competition binding curves in the presence of Nb29 (Fig. 1c-e); however, these experiments are complicated by the fact that Nb29 is a competitive inhibitor of the radioligand [ 3 H] prazosin. In cell signaling assays, Nb29 exhibits no agonist activity on its own, has no effect on EC 50 for oxymetazoline or noradrenaline, and slightly reduces the maximum efficacy of α 1A AR activation ( Supplementary Fig. 8a-d), suggesting that Nb29 appears to antagonize receptor activation or possibly block the ligand entry into the orthosteric pocket. It should be noted that the radioligand competition assay was performed in equilibrium and the agonists had a longer incubation time to access the orthosteric pocket than in the signaling assay. In both assays, the effects of Nb29 are larger for oxymetazoline compared with noradrenaline, which is consistent with Nb29's binding selectivity towards the oxymetazoline-bound state of the α 1A AR in the titration assay (Fig. 1b-e and Supplementary Fig. 8). Thus, Nb29 might be considered a weak positive allosteric modulator (PAM) or a neutral allosteric modulator, given that it binds to a known allosteric binding pocket in other GPCRs.
The Nb29 binding interactions are almost identical in oxymetazoline-and noradrenaline-bound Nb29-α 1A AR complexes, but the map resolution is relatively poor in the noradrenaline-bound state ( Fig. 1 and Supplementary Figs. 3 and 4). Thus, we used the oxymetazolinebound state for structural analysis. Nanobodies consist of three complementarity-determining regions (CDRs). The relatively long CDR3 interacts with a broad range of residues from ECL2 and with E305 7.32 at the top of TM7 (Supplementary Figs. 9a-c and 10a, b). Among them, seven amino acids (R166, Q167, E171, T174, Q177, N179, and E305 7.32 ) are non-conserved residues in α 1 AR subtypes, suggesting that these residues are involved in nanobody specificity (Supplementary Figs. 7, 9a-c). Nb29 binding also stabilized the polar interaction network between ECL2 and R96 3.22 , which is not observed in the inactive α 1A AR structure without Nb29 ( Supplementary Fig. 9d-f). Four residues of CDR3 (Y100, R101, D102 and H103) bind to the extracellular vestibule from the agonist-binding pocket (Fig. 6a)  interaction with F308 7.35 of α 1A AR, stabilizing the inward conformation of TM7 (Fig. 6a, b). As mentioned above, F308 7.35 and the close-lid conformation of F 7.39 is important for the agonist binding in αARs (Fig. 4) 38 , the π-π stacking of oxymetazoline with F312 7.39 might contribute to the Nb29 binding selectivity compared with the cation-π stacking of noradrenaline with F312 7.39 .
The residues at position 7.35 have also been identified as critical residues for both PAM and negative allosteric modulator (NAM) binding of muscarinic acetylcholine receptors (MRs) by stabilizing the extracellular side of TM7 either in inward or outward conformations, respectively 45,46 . The LY2119620, a PAM for M 2 R, binds to the extracellular vestibule and changes the conformation of W422 7.35 by an aromatic stacking, whereas those of the other residues are almost identical between the LY2119620-M 2 R-iperoxo and M 2 R-iperoxo complexes ( Fig. 6c and Supplementary Fig. 10c, d) 45 . In contrast to this PAM, a peptide toxin MT7, the NAM for M 1 R activation, stabilizes the outward displacement of TMs 6 and 7 through interactions with W400 7.3546 (Fig. 6d and Supplementary Fig. 10e, f). Moreover, a mutagenesis study indicated that F330 7.35 in α 1B AR is involved in the allosteric binding for conotoxin ρ-TIA, a selective NAM for α 1B AR among α 1 ARs 49 . The Nb29 binding site also overlaps with the aryloxyalkyl tail of the selective agonist salmeterol binding in β 2 AR (Fig. 6e, f) 12 . The smaller N 7.39 in β 2 AR enables the salmeterol to extend into the extracellular vestibule to make aromatic interactions with F194 ECL2 , H296 6.58 and Y308 7.35 , which are unique to β 2 AR (Supplementary Fig. 7). Among aminergic receptors, the aromatic amino acid at position 7.39 is observed in αARs (F 7.39 ), muscarinic receptors (Y 7.39 ) and histamine H 3 and H 4 receptors (F 7.39 ) 12,43,45,50 . The selective agonists targeting the extracellular vestibule of αARs are limited compared to other GPCRs such as βARs and muscarinic receptors 12,13,43,50 (Fig. 6g-j and Supplementary Fig. 1).
Subsequently, we compared the nanobody binding site with available class A GPCR structures in complex with extracellular nanobodies or antibody Fab fragments (Supplementary Fig. 10). Consistent with the review for GPCR antibodies 19 , peptide-binding GPCRs are more frequently targeted by antibodies because they have relatively large binding pockets compared with the small-molecule binding GPCRs such as aminergic GPCRs. Only two structures in complex with extracellular nanobodies have been reported in class A GPCRs, which are the apelin receptor (APJ) 24 and the orexin receptor 2 (OX 2 R) 51 , although more structures of the intracellular binding nanobodies have been published for stabilizing the GPCR active conformations as G protein mimetics, such as Nb9-8 for M 2 R 10,20,45 (Supplementary Fig. 10c-j). Anti-APJ nanobody JN241 antagonizes APJ through extensive interactions with extracellular loops of APJ and the insertion of CDR3 into the peptide-binding site 24 , whereas anti-OX 2 R nanobody Sb51 is positioned above the small-molecule agonist and partially overlaps with natural-peptide orexin B binding site 51 . In contrast to nanobodies, conventional antibodies and Fab fragments consist of heavy (CDRs H1-3) and light (CDRs L1-3) chains ( Supplementary  Fig. 10k-r). The antibody for protease-activated receptor 2 (PAR2) behaved as an antagonist by blocking ligand access from the extracellular region by both heavy and light chains (H1, H3, L2, and L3) 26 . In the angiotensin II type 2 receptor (AT2) 23 , EP4 22 , and D 2 R structures 25 , their antibodies allosterically enhance the ligand binding. The antibody for AT2 bound to the ECL1 and β-hairpin motif of ECL2 to stabilize the peptide agonist binding pocket, while the antibody for EP4 stabilizes the occluded β-hairpin of ECL2, leading to enhanced antagonist binding. In the D 2 R-Fab3089 structure, the CDR-H2 stabilizes the antagonist spiperone which binds toward the extracellular vestibule.
Taken together, the Nb29 structure provides insights into allosteric binding and antibody recognition for α 1A AR. Nb29 covers the extracellular surface of α 1A AR like anti-APJ nanobody JN241 and anti-PAR2 Fab, whereas the CDR3 loop of Nb29 binds to a similar site of PAM for M 2 R (Fig. 6 and Supplementary Fig. 10). It should be noted that nanobodies are amenable to optimization due to the single variable domain. For example, the G protein-mimicking nanobody for β 2 AR was optimized by directed evolution to increase its affinity to the receptor 10 ; and the APJ nanobody antagonist JN241was rationally engineered into an APJ agonist by structure-guided site-directed mutation of CDR3 24 .

Selectivity of G protein interactions with adrenergic receptors
Noradrenaline-and oxymetazoline-bound α 1A AR-miniGsq complexes are almost identical at the interfaces with miniGsq, as observed in α 2A AR-Go complexes with different agonists 14 . Thus, we used the oxymetazoline-bound active state for structural analysis. Of the 15 residues at the carboxyl-terminal α5-helix of miniGsq (Fig. 7a) 29 , seven residues are specific to Gq, including K H5.12 , L H5.16 , Q H5.17 , N H5.19 , E H5.22 , N H5.24 , and V H5.26 (superscript, CGN G protein numbering system 52 ); five residues [D H5.13 , I H5.15 , L H5.20 , Y H5.23 , and L H5.25 ] are conserved between Gs and Gq, and the other three residues [I H5.14 , M H5.18 and R H5.21 ] are located on the opposite side of the interface. When comparing the interactions of α 1A AR-miniGsq complexes with α 2A AR-Go 14 and β 2 AR-Gs complexes 9 , the α5-helix of miniGsq is slightly shifted towards helix 8 of α 1A AR (Fig. 7b-h). In the α 1A AR-miniGsq complex, α 1A AR forms six hydrogen bonds with the residues corresponding to Gq: between R213 5.67 and Q H5.17 , between the backbone carbonyl of G127 3.53 and N H5. 19 , between T273 6.36 and the backbone carbonyl of N H5. 24 ; the side chain of N H5.24 forms hydrogen bond networks with the backbone carbonyl of C328 7.55 , the backbone carbonyl of S330 8.47 , and the side chain of Q331 8.48 (Fig. 7c, d). These residues of α 1A AR are not conserved in α 2 ARs and βARs (Supplementary Fig. 11). The R 3.50 forms cation-π stacking interaction with Y H5.23 , which is also observed in the β 2 AR-Gs complex, but not in α 2A AR-Go complex as this position is C H5.23 in Go protein (Fig. 7c-h). In the α 2A AR-Go complex, polar interactions are observed between S 3.53 and N H5. 19 , and hydrophobic interactions are predominant (Fig. 7e, f). In the β 2 AR-Gs complex, one side of the α5-helix of the Gs protein forms a cluster of hydrogen-bond interactions with TM3 (I 3.54 and T 3.55 ) and TM5 (E 5.64 , Q 5.68 , and K 5.71 ), leading the α5-helix to shift towards TM5 (Fig. 7b, g, h). In addition to α5-helix, the N-terminal helix of G α subunits are also involved in GPCR-G protein interactions 9,14 . The previous study indicates that polybasic cluster at the C terminus of M 1 R, which is conserved among most G q/11 -coupling GPCRs, interacts with the G-protein Gα 11 /β interface 53 . However, our structure lacks these regions and there are currently no other active structures of α 1 ARs. While α 1A AR is predominantly coupled to G q/11 proteins, a few studies have identified G 12/13 54 and β-arrestin signaling pathways 2,55 . Further studies will be required to better understand the mechanism of activation and selectivity of α 1 AR signaling.

Discussion
Here, we present three cryo-EM structures of α 1A AR in both active and inactive states. These structures reveal several structural aspects of α 1A AR. The ligand-binding modes of the endogenous agonist noradrenaline and the imidazoline-type agonist oxymetazoline demonstrated a key aromatic interaction involving F312 7.39 , which is conserved in αARs, and distinct ligand recognition by the unique residue M292 6.55 . The inactive α 1A AR bound to tamsulosin reveals the subtype selectivity of antagonist binding pockets involving F86 2.64 . Our results also provide structural insights into nanobody recognition for α 1A AR. Nb29 binds to the extracellular vestibule of α 1A AR and the cationic residue of CDR3 stabilizes F308 7.35 which is the equivalent binding site of the positive allosteric modulator for M 2 R. Finally, our active α 1A AR structures provide insight into G protein binding selectivity by comparisons with α 1A AR-miniGsq, α 2A AR-Go, and β 2 AR-Gs structures. Together with our α 1A AR structures and previously published structures of α 2 ARs and βARs, the active and inactive structures of the major subtypes of the adrenergic receptor family have been determined. These results should facilitate the design of more selective and effective therapeutic drugs targeting both orthosteric and allosteric sites in this receptor family.

Methods
Construction C-terminus truncated human α 1A AR (residues 1-370, full length: 466) was modified by mutation of the N-linked glycosylation sites to glutamine (N7Q, N13Q and N22Q), N-terminal addition of the hemagglutinin signal peptide, FLAG-tag epitope, and C-terminal addition of the 8 × His-tag. For the active α 1A AR structure, residues 223-261 of intracellular loop 3 were replaced with minimal T4L 28 . For the inactive α 1A AR structure, we swapped residues from C205 5.59 to G275 6.38 with residues from T247 5.59 to L277 6.38 of kOR 21 . In addition, we introduced two thermostabilizing point mutations S113R 3.39 and M115W 3.41 which were previously used for structure determination of several inactivestate GPCRs 22,25,31 . The primers used in this study were obtained from Rui Biotech (Beijing, China), Xianghong Biotech (Beijing, China) or Genewiz (Beijing, China). The DNA sequencing analysis was performed at the Rui Biotech (Beijing, China).

Expression and purification of α 1A AR
Recombinant baculovirus was generated using the Bac-to-Bac Baculovirus Expression System (Thermo Fisher Scientific). Sf9 insect cells at a cell density of 4 × 10 6 cells/ml in ESF-921 insect media (Expressions Systems) with 20 μg/ml gentamycin and 1 μM ligand were infected with baculovirus and shaken at 27°C for 2 days. Cells were harvested by centrifugation and stored at −80°C until use.

Discovery of the conformationally selective α 1A AR nanobody
The synthetic nanobody library displayed on the surface of BJ5465 yeast strain was obtained from Drs. A. C. Kruse (Harvard University) and A. Manglik (University of California San Francisco) 27 . The yeast cells were recovered in tryptophane dropout (-Trp) medium [prepared by Yeast Synthetic Drop-out Medium Supplements without tryptophane (sigma) and Yeast Nitrogen Base without amino acids (BD Difco) at pH 6.0] with 2% (w/w) glucose at 30°C, and the nanobody was induced by -Trp medium with 2% (w/w) galactose at 25°C. Expression levels of nanobody were estimated by staining with anti-HA antibody  TM6 TM6 TM6   TM6 TM6 TM6  TM6 TM6 TM6  TM5 TM5 TM5  TM5 TM5 TM5   TM6 TM6 TM6   TM3 TM3 TM3  TM3 TM3 TM3   S 3 TM6 TM6  TM5 TM5 TM5  TM3 TM3 TM3   ICL2 ICL2 ICL2  ICL2 ICL2 ICL2   TM2 TM2 TM2  TM2 TM2 TM2  TM2 TM2 TM2   TM2 TM2 TM2   TM5 TM5 TM5  TM3 TM3 TM3   TM5 TM5 TM5   TM6 TM6 TM5 TM5 TM5   TM7 TM7 TM7   H8 H8 H8   TM7 TM7 TM7   H8 H8 H8   TM7 TM7 TM7   H8 H8 H8   ICL2 ICL2 ICL2   ICL2 ICL2 K  K  K  R  T  T  K  K   D  D  D  D  D  D  D  D   I  T  T  I  V  I  T  T   I  I  I  I  I  I  I  I   L  L  L  Q  I  I   (Cell Signaling Tech) and analyzing by flow cytometry with an Accuri C6 (BD Biosciences) Induced yeast cells were washed and resuspended in a selection buffer (20 mM HEPES pH 7.5, 150 mM NaCl, 0.05% MNG, 0.005% CHS, 2.8 mM CaCl 2 , 0.1% (w/v) bovine serum albumin and 5 mM maltose). Nanobody clones against the purified FLAG-tagged α 1A AR (C-terminus truncated after residue 370) bound to oxymetazoline were enriched by two rounds of magnetic-activated cell sorting (MACS) and four rounds of fluorescence-activated cell sorting (FACS) (See Fig. 1a). For the first round of the MACS, 5 × 10 9 yeast cells were precleared by incubating with Alexa Fluor-647 conjugated anti-FLAG M1 antibody (M1-647, prepared by anti-FLAG M1 antibody and Alexa Fluor 647-NHS ester) and anti-Alexa Fluor-647 microbeads (Miltenyi) and passed LD column (Miltenyi) to remove nonspecific nanobody. Flowed-through yeast cells were washed with the selection buffer, then incubated with 0.2 µM α 1A AR bound to oxymetazoline, the antibody and the microbeads. After incubation at 4°C for 30 min, yeast cells were loaded on the LD column, washed with the selection buffer and the eluted yeast cells (3.4 × 10 6 cells) by plunger were expanded and used in a subsequent round of MACS. The second round of MACS was performed similarly to the first, but beginning with 4 × 10 8 yeast cells and using biotin-conjugated anti-FLAG M1 antibody Fab fragment (anti-FLAG M1 antibody was digested by papain then labeled with biotin-NHS ester), streptavidin microbeads (Miltenyi) and LS column (Miltenyi) were used, and 5 × 10 6 yeast cells were eluted.
Subsequently, we performed four rounds of FACS by FACSAria II (BD Biosciences) (See Supplementary Fig. 2). For the selection rounds 3 and 6, yeast cells were stained with Alexa Fluor-488 or −647 conjugated anti-HA antibody (Cell Signaling Tech) and 0.1 µM FLAGtagged α 1A AR with anti-FLAG M1-647 or −488. For the selection rounds 4 and 5, in order to enrich for conformational selective nanobodies, yeast cells were stained with two different populations of α 1A AR labeled with anti-FLAG M1-488 and −647 fluorophores, one bound with oxymetazoline and another bound to tamsulosin. Staining yeast cells for each round of FACS experiments were the following; 5 × 10 7 cells for round 3 and 1 × 10 7 cells for rounds 4-6. After round 6, the sorted yeast cells were diluted and plated on -Trp agar plates. Single clones were sequenced and cloned into the periplasmic expression vector pET26b, containing an N-terminal pelB signal sequence and a C-terminal histidine tag, and transformed into BL21(DE3) Escherichia coli. Cells were induced in Terrific Broth medium with 2 mM MgCl 2 , 0.1% glucose and 50 µg/ml kanamycin at an OD600 of 0.7 with 1 mM IPTG and incubated with shaking at 25°C for 20 h. Periplasmic protein was obtained by osmotic shock in a buffer containing 0.2 M Tris pH 8.0, 0.5 mM EDTA and 0.5 M sucrose at 4°C for 1 h, then diluted 4 times and incubated for another one hour. The lysate was centrifuged and the supernatant was purified by Ni-NTA resin and size-exclusion chromatography.
For the on-yeast titration assay, Nb29-displayed yeast cells were stained with the Alexa Fluor-647 conjugated anti-HA antibody and several concentrations of purified α 1A AR fused at the C-terminus to an enhanced green fluorescent protein in the presence or absence of 500 μM ligands in the selection buffer. Yeast cells were analyzed by Accuri C6 and the ratio of double-positive yeast cells among anti-HA positive cells was calculated.
value of each image, which was set from −1.3 to −1.8 μm during data collection, was determined by Gctf 59 .

Model building and refinement
The atomic coordinate of the Nb29-α 1A AR-miniGsq and the α 1A AR-Nb6 complexes was generated by combining homology modeling and de novo model building. An initial structure model for the active α 1A AR was predicted by the homology model from GPCRdb (gpcrdb.org) 60 , α 1A AR-kOR was generated by AlphaFold2 61 and the structure model of Nb29 was predicted by the homology model from swiss-model 62 . The initial models of miniGsq and Nb6 were imported from miniGs (PDB code: 5G53) 29 and Nb6 (PDB code: 6VI4) 33 , respectively. The cryo-EM model was docked into the electron microscopy density map using UCSF Chimera 63 , followed by iterative manual adjustment and rebuilding in PHENIX 64 and COOT 65 . The structures were refined against the corresponding map using PHENIX and COOT in real space with secondary structure and geometry restraints. Figures were created using the PyMOL Molecular Graphics System v.2.4.0 (Schrӧdinger, LLC), UCSF Chimera and the UCSF Chimera X1.3 package.

Glo-sensor signaling assay
To determine the signaling profile of Nb29 against α 1A AR, we used a cAMP Glo-sensor kit (Promega) with an engineered Gsq protein in which 15 residues at the C-terminus of Gs protein were replaced with those of Gq protein. The Gsq protein could be activated by the α 1A AR and stimulates intracellular cAMP production. In brief, pGlo-Sensor™−22F plasmid, receptor plasmid, and Gsq plasmid were transfected into HEK293T cells. At 24 h after transfection, the cells were switched into a CO 2 -Independent Medium (Gibco) and incubated with GloSensor TM cAMP reagent. The mixture was then transferred to a 96-well white plate. The 96-well plate is placed at 37°C in the dark for 1 h, then placed at room temperature in the dark for 1 h before use. The luminescence signal was measured by Ensight TM plate reader (Perki-nElmer) around 10-15 min after the addition of the agonist and/or Nb29. The result curves were calculated and fitted by GraphPad Prism 9.

Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.