Development of “plug and play” fiducial marks for structural studies of GPCR signaling complexes by single particle cryo-EM

Summary

“Universal” synthetic antibody (sAB)-based fiducial marks have been generated by customized phage display selections to facilitate the rapid structure determination of G-protein coupled receptor (GPCR) signaling complexes by single particle cryo-electron microscopy (SP cryo-EM). sABs were generated to the two major G-protein subclasses: trimeric G_i and G_s, as well as mini-G_s and were tested to ensure binding in the context of their cognate GPCRs. Epitope binning revealed that multiple distinct epitopes exist for each G(αβγ)-protein. Several Gβγ-specific sABs, cross-reactive between trimeric G_i and G_s, were identified suggesting they could be used across all subclasses in a “plug and play” fashion. sABs were also generated to a representative of another class of GPCR signaling partner, G-protein receptor kinase 1 (GRK1) and evaluated further, supporting the generalizability of the approach. EM data indicated that the subclass specific sABs provide effective single and dual fiducials for multiple GPCR signaling complexes.

Graphical Abstract

eTOC Blurb

High affinity sABs generated to different classes of GPCR signaling partners are exploited as “universal” fiducials to enable structure determination of multiple GPCR-G-protein and GRK signaling complexes by cryo-EM. The power of the approach is that these complexes with sABs can be assembled in a “plug and play” fashion.

Introduction

Structural studies of GPCRs and their signaling complexes provide insights into the fundamental processes determining their functions. Although their mechanisms are very similar, GPCRs differ in detail in how they are stimulated, modulated and downregulated (Erlandson et al., 2018). These receptors are characterized by a similar seven transmembrane architecture and their ability, upon stimulation, to recruit a set of cytoplasmic binding partners that modulate different activities (Ritter and Hall, 2009). These activities derive from three principal classes of signaling partners: (i) heterotrimeric G-protein complexes, (ii) G-protein coupled receptor kinases (GRKs) and (iii) β-arrestins. The coupling of a GPCR to each type of binding partner results in a different biological response (Hilger et al., 2018). Notably, the resulting complexes are generally transient and their lifetimes are dependent on other regulatory factors. This paints a picture where the functions are derived from a series of highly dynamic events involving multiple components. Interestingly, structural studies seem to suggest that subtle changes in structure can often confer very different signaling outcomes (Katritch et al., 2013; Latorraca et al., 2017; Thal et al., 2018). However, these are very dynamic and complex systems and to establish the definitive linkages between structures and their relevant functions will require significantly more examples, both for the different states of the receptors themselves and the changes that occur in the context of their engagement with their signaling partners.

While the Protein Data Bank (PDB) is relatively sparse with structures of GPCR, it is rapidly becoming more populated with the development of approaches that exploit the use of engineered fusion constructs that help crystallization (Liu et al., 2012; Chun et al., 2012; Ghosh et al., 2015) and antibody based fiducial marks for SP cryo-EM (Kang et al., 2018; Koehl et al., 2018; Maeda et al., 2018). Although this represents a significant step forward, the undertaking of each individual system is challenging. Further, the structures of the GPCRs with their signaling partners offer more detailed information about the structural linkages between activation and signaling than can’t be inferred from the structure of the isolated components. Not surprisingly, structure determination of these GPCR signaling complexes is significantly more technically challenging than the respective GPCRs, which are formidable by themselves. Thus, while substantial resources will be expended on attempts to obtain structures of these complexes at near-atomic resolution, successes will be limited and protracted without developing more powerful approaches.

Thus, we have endeavored to develop a multi-faceted platform to expand the throughput for determining the structures of GPCR complexes. A key element of the platform is a phage display pipeline that generates tailored Fabs (Fragment antigen binding) referred to as “synthetic antibodies” or sABs. The phage display selection protocols can be tuned to bind to the antigen target in a defined conformational state or regio-specific fashion (Rizk et al., 2011; Paduch et al., 2013; Rizk et al., 2015; Rizk et al., 2017; Mukherjee et al., 2018). Also, it is possible to generate sABs that stabilize transient complexes to facilitate their structure determination (Mateja et al., 2015). A second element of the platform is the use of SP cryo-EM to determine the structures. The recent advances in cryo-EM have been transformative for structural biology studies of large proteins and complexes (Kühlbrandt, 2014; Thompson et al., 2016). However, GPCRs and their complexes remain challenging systems because of their relatively small size, low stability and high flexibility. In this regard, we have demonstrated that sABs can be utilized as extremely potent fiducial marks to increase the size of the “particle”, facilitate its orientation, increase its stability and in some cases, enhance the binding of the receptor to its signaling partners (Shukla et al., 2014; Kang et al., 2018).

Even coupling the attributes of sAB fiducial marks with the powerful capabilities of cryo-EM, the structure determination of GPCR signaling complexes is likely never to be high throughput. A major impediment is that given the number of important GPCRs of interest, it would not be feasible to generate sABs to each system individually. Thus, a key step in eliminating many of the arduous tasks is to adopt a strategy to exploit the attributes of sABs without having to generate them for each of the individual complexes. So, we introduce the concept of “universal” sABs, which takes advantage of the fact that while there are a number of different GPCRs, they associate with relatively few families of signaling partners. For instance, the majority of GPCRs bind to one of the four subclasses of heterotrimeric G-proteins: G_s, G_i/o, G_q/11 and G_12/13 (Syrovatkina et al., 2016). So, instead of generating unique sABs to each GPCR, it is only required to have sABs to each of trimeric G-protein subclasses to get a full coverage of these complexes. Likewise, for the GRKs, there are seven subclasses (Komolov and Benovic, 2018) and sABs to each one of them would virtually cover across all GPCR/GRK complexes. For the trimeric G-proteins, there is a further potential simplification. The trimeric G-proteins consist of α,β and γ subunits. Whereas GPCR/G-protein interaction is highly α subunit dependent (Rasmussen et al., 2011; Flock et al., 2017), there is no clear evidence that composition of the Gβγ dimer impacts receptor coupling. Consequently, if a sAB can be generated that targets only the Gβγ dimer in the most commonly used combination of Gβ₁γ₂, then this sAB can be used as a fiducial mark in all complexes that use this particular combination of the subunits.

We had previously shown the value of a sAB in delineating the hallmarks of β-adrenergic receptor/β-arrestin complex formation and dynamics (Shukla et al., 2013; Shukla et al., 2014). This work focuses on the two other families of GPCR signaling partners: heterotrimeric G-proteins and GRKs. We describe the methods for generating high affinity sABs to the inhibitory (G_i) and stimulatory (G_s) G-protein families. While many of these sABs are specific to the particular complex they were generated against, some of them are cross-reactive as they recognize only the Gβγ dimer. Epitope binning shows that both G_i and G_s have at least two orthogonal epitopes so that both can accommodate two sABs that could be used as dual fiducials. One of the sABs (G50) was used in the cryo-EM structure determination of the rhodopsin (Rho)/G_i complex (Kang et al., 2018). Because of its general utility for structure determination of receptor/G_i complexes, we further improved its binding characteristics through affinity maturation. Similarly, we generated sABs to GRK1 that can be used as universal GPCR/GRK1 fiducials.

Our work has produced “off the shelf” universal sABs to a subset of signaling partners and lays the groundwork for developing and utilizing these sABs broadly across all signaling partner families in a “plug and play” fashion. The ultimate aim is to develop the platform to a level where the difficult step in the structure determination pathway is producing the GPCR sample, everything else is turnkey. This is an ambitious goal and will be subject to system specific caveats, but the ability to accomplish this will be transformative to both the structural and biological efforts aimed at better elucidation of the functional mechanisms of this important class of receptors and their effectors.

Results

1. Choice of target set for generating sABs to GPCR signaling components

The target set consisted of: (i) dominant negative mutant of heterotrimeric inhibitory G-protein (G_i1-DN), (ii) α subunit of the stimulatory G-protein (Gα_s), (iii) mini-G_s and (iv) GRK1. The trimeric G_i belonging to the adenyl cyclase-inhibitory G-protein family (G_i/o) was one of the targets of choice as G_i-coupling receptors consists of the largest group among GPCRs (Flock et al., 2017). The DN mutant was used because of its ability to enhance GPCR/G-protein complex formation by reducing nucleotide affinity, and consequently limiting dissociation of the components (Liang et al., 2018b). Additionally, sABs against trimeric G_s will be useful for receptors coupling to the stimulatory class of G proteins. However, instead of the trimeric G_s, only the Gα_s subunit of the protein was subjected to selection, as the Gβ₁γ₂ subunits of the trimeric G_s and G_i under study are identical. Finally, we wanted to obtain sABs against mini-G_s that is an engineered version of the G_s (Carpenter et al., 2016; Carpenter and Tate, 2016).

Among the GRKs, we focused on GRK1, a member of the GPCR kinase family that couples to Rho and is well characterized (He et al., 2017; Komolov and Benovic, 2018). The K46L mutant was used for the initial selections because this mutant has been shown to increase the affinity of the kinase towards the receptor (He et al., 2017). However, the kinase activity of this mutant is relatively low, probably due to electrostatic interactions between kinase domains that form the “ionic lock”. Thus, in the later selections, a more catalytically active triple mutant (K46L/R461A/Q462A) that eliminates the ionic lock interaction and increases complex stability was used. Biotinylated G_i, Gα_s, mini-G_s and the GRK1 were expressed and purified as described in Methods. The labeling efficiency was determined by a pull-down assay on streptavidin (SA) magnetic beads (Figure S1a, S7a).

2. Phage Display Selection and primary validation of the sABs

The biotinylated targets were subjected to selection campaigns using a synthetic phage library (Miller et al., 2012). Four rounds of biopanning were performed. To ensure specificity and high affinity, the antigen concentrations were gradually dropped from 200 nM in the first round to 5 nM in the last round. After four rounds of selection, the enrichment, which is a measure of the target specific binders over the background, was above 100-fold in all cases (Figure S1b). Ninety-six clones from each selection were analyzed by single-point phage ELISA. As judged from the signal intensity from the wells coated with targets compared to the control wells lacking any target, most of the clones were specific with minimal off-target binding. Phage ELISA results of representative clones are shown in Figure 1. DNA sequencing identified 59, 14, 24 and 23 clones for G_i, Gα_s, mini-G_s and GRK1 (K46L) respectively. Selected clones having the highest signal to background ratio were subsequently cloned into bacterial expression vector pRH2.2, expressed in E. coli BL21 (Gold) cells, purified by protein A or protein G followed by cation exchange chromatography for further characterization. CDR sequences of the clones are in Table S1.

Figure 1. Phage display antibody selection.

Representative results of single point phage ELISA: (a) Gα_s, (b) G_i1, (c) mini-G_s and (d) GRK1.

3. Characterization of the sABs against G proteins

The binding kinetics of the sABs to their targets were determined by surface plasmon resonance (SPR) (Table 1). A characteristic of the most promising sABs was their slow dissociation rates (k_off < 10⁻³ s⁻¹). Differential scanning fluorimetry (DSF) studies show the sABs are thermally stable having melting temperatures (T_m) above 70°C (Figure 2a). They also form high affinity complexes with the respective targets as evident from the monodispersed elution peaks in analytical size-exclusion chromatography (aSEC) (Figure 2b).

Table 1.

Kinetic parameters of the characterized sABs for G-proteins measured by SPR.

Target	sAB	kon (M−1s−1)	koff (s−1)	KD (nM)
Gi	G2	1.6 x 105	3.7 x 10−4	2.3
	G27	7.2 x 104	5.3 x 10−4	7.3
	G33	2.0 x 106	2.0 x 10−3	1.0
	G39	1.8 x 105	2.4 x 10−3	13.7
	G46	7.9 x 105	4.0 x 10−3	5.0
	G50	1.2 x 106	1.4 x 10−3	1.2
	G51	1.2 x 105	1.7 x 10−3	14.3
	G56	8.3 x 105	8.8 x 10−4	1.1
Gαi	G2	9.4 x 104	4.5 x 10−3	48.5
	G6	4.1 x 105	2.7 x 10−3	6.6
	G46	9.0 x 105	6.0 x 10−3	6.6
	G50	6.3 x 105	1.5 x 10−3	2.5
Gs	G39	3.7 x 105	2.3 x 10−3	6.2
	G56	8.0 x 105	1.2 x 10−3	1.4
	Gs3	1.0 x 106	8.9 x 10−3	8.7
	Gs5	9.9 x 105	8.2 x 10−3	8.3
	Gs6	9.3 x 105	4.3 x 10−5	<1 (0.05)
Mini-Gs	M1	1.2 x 105	4.0 x 10−4	3.4
	M3	1.1 x 105	1.2 x 10−3	11.8
	M4	2.6 x 105	6.6 x 10−4	2.5
	M7	3.7 x 105	8.1 x 10−4	2.2
	M15	8.1 x 105	1.6 x 10−3	2.0
	M19	2.4 x 106	3.9 x 10−3	1.6
	M20	1.2 x 105	2.2 x 10−3	18.3

Figure 2. Validation of the G-protein variants specific binders.

(a) Thermal melting curves of selected sABs against G-proteins monitored by DSF. (b) aSEC profiles show that multiple sABs form monodispersed complexes with G_i. (c) Cross-reactivity of the sABs obtained for Gαs and G_i with G_s analyzed by single point protein ELISA. Data are represented as mean ± SD. (d, e) Epitope binning by SPR. (d) G_i and (e) G_s is immobilized on the NTA sensor chip. Net increase in response units is observed for sABs that are binding to epitopes different from the first sAB. (f) Representative 2D class averages from negative-stain EM shows that two sABs G51 and G50 recognizing different epitopes, can bind simultaneously to G_i. Scale bar, 50 Å. (g) Epitope binning of mini-G_s sABs by SPR identified three epitopes. Sensograms for representative sABs from each epitope are shown.

To test whether the sABs generated against the Gα_s also bind to trimeric G_s, protein ELISA was performed (Figure 2c). Four out of six sABs tested against the Gα_s, viz. Gs1, Gs3, Gs5, Gs6 also bind well to the trimeric variant suggesting that their epitope on Gα_s is not occluded when combined with the Gβγ subunits. The binding of sAB Gs10 and Gs13 to the trimeric G_s is compromised suggesting that their epitope is different from the other four binders and presumably interferes with the interaction between Gβ and Gα_s.

Similarly, the reduced binding of most of the trimeric G_i sABs to the trimeric G_s indicate that these sABs bind primarily to the Gα_i subunit as the Gβγ subunits in both heterotrimeric G_s and G_i under study are identical (Figure 2c). So, if sABs selected for the heterotrimeric G_i bind to G_s or vice versa, it is very likely that the epitope resides on Gβγ. Thus, this particular class of cross-reactive sABs can be used as “versatile” fiducial marks for receptors bound to all classes of G-proteins containing identical Gβγ subunits. These results were confirmed by epitope binning experiments using SPR (Figure 2d, e). Sixteen sABs were binned to identify the number of potential epitopes. From these experiments, we were able to categorize the sABs into three distinctive groups, represented by: (i) heterotrimeric G_i specific sABs: G2 and G6; (ii) Gβγ specific sABs: G33, G39, G56; and (iii) Gα subunit binding sABs: G50, G46 that bind to Gα_i, and Gs6, Gs3 that bind to Gα_s. sABs recognizing multiple, non-overlapping epitopes to the G-proteins provides an opportunity to pair them together as “dual-fiducials” for SP cryo-EM, thereby adding a mass of 100 kDa and providing a higher precision of particle orientation (Figure 2f).

To determine the diversity of the epitopes on mini-G_s, we binned the twenty-three sABs against mini-G_s using SPR (Figure 2g). From the results, we were able to classify mini-G_s sABs into three groups. Thirteen sABs (M1, M3, M4, M5, M6, M9, M10, M11, M12, M15, M16, M17, M21, M22, M23) were clustered on the same epitope, another two binders (M7 and M19) bind to a different epitope. A potential third epitope, which partially overlaps with the other two can be identified for the remaining four sABs (M13, M14, M18 and M20).

As the selections were done on the G-proteins and not on the GPCR/G-protein complexes, there is a chance that some of the sABs might interfere with the binding between the receptor and the G-proteins. To identify sABs that disrupt complex formation, we used aSEC followed by negative-stain EM. Rho and Dopamine D1 (D1R) receptors were chosen as model systems since they couple with trimeric G_i and G_s respectively. sABs tested for complex formation with the receptor/G-protein complexes were chosen based on their expression level, thermal stability, affinity and having different epitopes. For example, sAB G46 has K_D = 5 nM to trimeric G_i and stabilized it by 7.5°C (Figure S2a).

Successful sAB binding to a receptor/G-protein complex was confirmed by the shift of the aSEC retention volume of the GPCR/G-protein/sAB complexes compared to GPCR/G-protein and co-elution of all components, as analyzed by SDS-PAGE. Three of the sABs against Rho/G_i - G39, G46 and G50, form monodispersed complexes that can be isolated by aSEC (Figure 3a, b). 2D class averages from negative-stain EM showed a characteristic bimolecular feature of the variable and constant domains of the sAB bound to the Rho/G_i complex (Figure 3d). Since sAB G39 binds to the trimeric G_s as well (Table 1), this proves that it binds to Gβγ and can be used as a “versatile” fiducial. Similarly, sAB Gs6, a G_s binder of high affinity (K_D < 1 nM), forms a monodispersed complex with D1R/G_s (Figure 3c), which was subjected to negative-stain EM after aSEC. Gs6 also displays characteristic features of a sAB based on negative-stain EM 2D class averages (Figure 3d). Moreover, Gs6 binds to D1R/G_s in presence of a nanobody (Nb35) (Figure 3c), which is used for the stabilization of GPCR/G_s complexes (Rasmussen et al., 2011; Glukhova et al., 2018). This provides an opportunity to take advantage of both nanobody mediated complex stabilization and effective fiducial mark provided by the sAB.

Figure 3. sABs binning to GPCR/G-protein complexes.

(a,b,c) aSEC profiles and SDS-PAGE analyses of the GPCR/G-protein/sAB complexes. (a) Rho/G_i binds to sABs G46 and G50. (b) Rho/G_i binds to sAB G39 and can also bind to sAB G39 and G50 simultaneously. (c) D1R/G_s binds to sAB Gs6 alone and also in presence of Nb35. (d) Representative negative-stain EM 2D classification of the GPCR/G-protein/sAB complexes. Scale bar, 50 Å. The corresponding micrographs are in Figure S3.

4. Characterization of interactions between sAB G50 and heterotrimeric G_i

The structure of Rho/G_i/G50 was the first GPCR/G-protein complex cryo-EM structure with a fully resolved α-helical domain (AHD) (Kang et al., 2018). Since G50 interacts with both the Gα_i-AHD and the Gβ, this effectively traps the AHD in a particular defined position (Figure 4a, b). The AHD domain is predominantly stabilized by the CDRs – H1, H2 and H3 and L3, which are rich in aromatic, hydrophobic and hydrophilic residues. Notably, additional interactions between the CDR-H3 and L2 of G50 and the Gβ, lock the AHD domain in this particular position (Figure 4c, d).

Figure 4. sAB G50 stabilize AHD of G_i.

(a) Structure of the Rho-G_i complex stabilized by sAB G50 [PDB ID: 6CMO]. (b) Close-up view of the stabilizing interaction between the sAB and G_i. (c, d) Important interactions between CDRs of the sAB and G_i subunits: (c) Gα_i-AHD and (d) Gβ. (e) Paratope scanning of sABG50. (f) Surface representation of the AHD with side chains color coded according to their energetic contributions from epitope scanning. Gβ is colored green. ΔΔG_mut-wt were calculated from SPR experiments for (e) and (f). (g) Traces from molecular dynamics simulation of the AHD in presence and absence of sAB G50. Results show AHD is more flexible in absence of the sAB. (yellow trace in bottom panel: −sABG50).

The interface energetics of the G-protein/sAB G50 was probed by Ala scanning mutagenesis. Based on the interface analysis of the cryo-EM structure of Rho/G_i/G50 in PISA (Krissinel and Henrick, 2007), 36 residues distributed across all the six CDRs of G50 and 13 residues from the Gα_i were selected for Ala scanning. We expected CDR-H3 to be the major contributor to G50 binding although only a relatively small fraction of the residues (~25%) had side chains that are in direct contact with G_i. However, all the 17 residues of CDR-H3 were included in the Ala scanning to assess the degree to which the non-contacting side chains packed together to confer a highly defined conformation on this CDR. We used only Gα_i in mapping the energetic footprint of the paratope, because the binding affinity of G50 to the Gα_i subunit and trimeric protein were comparable (Table 1) suggesting that the principal contributions to the binding affinity were through this subunit. Energetic contributions of each residue were measured from changes in free energy of the Ala mutants relative to that of wt sidechain. The ΔΔG_mut-wt values were calculated based on the kinetic data (Table S2, S3) from SPR and summarized in the Figure 4e, f.

As expected, Ala scanning of the paratope showed that CDR-H3 has the largest impact on binding energetics (Table S2, Figure 4e). Ala substitutions at 13 positions (Y102, P103, W104, Y105, W106, W107, M108, P111, Y112, L113, Y116, G117, M118) caused more than 10-fold decrease in binding affinity with mutations at W106 and W107 completely abolishing the binding. The residues that are not interacting directly with the trimeric G_i in the structure are still energetically important probably in maintaining the proper structure and orientation of the CDRs that in turn facilitate interactions with the target. Three of the contacting residues: W104, W107, P111, interact with Gα_i-AHD (Figure 4c), while L113 contacts the Gβ subunit (Figure 4d). Two additional energetic hot spot residues were identified - Y34 in CDR-H1 and Y57 in CDR-H2, which directly interact with the AHD. Typically, serine residues are energetically neutral compared to aromatic and hydrophobic residues. However, in CDR-L3, which consists of four serine residues, Ala scanning indicates that two of them (S92, S94), result in ~10-fold reduced affinity to Gα_i. Notably, the phage library was designed to introduce diversity into only four of the six CDR loops: L3, H1, H2 and H3, but the structure also shows direct interactions between G_i and CDRs L1 and L2. Interestingly, Ala substitutions in these non-diversified CDRs revealed four residues with ΔΔG_mut-wt >1 kcal/mol. Two of them residing in CDR-L2 (Y50 and S51) are involved in the interaction with the AHD and the two in CDR-L1 (S29 and V30) interact with the Gβ, effectively “stitching” the AHD in the position observed in the structure.

Ala scanning of the epitope on Gα_i-AHD identified several side chain residues that contribute to sAB G50 binding (Figure 4f, Table S3). In general, single point Ala substitutions in the epitope had a far less energetic consequence on the sAB-antigen interaction than the paratope mutations. Interestingly, D106A had a stabilizing effect, as evidenced by the negative ΔΔG_mut-wt, while the side chains of the other residues have small favorable contributions. The dominant contributor here is F108, where the Ala mutation completely eliminates the binding, highlighting the role of this aromatic residue in antigen binding. Overall, this study revealed that all the energetic contributors are clustered on the most exposed, bottom part of the AHD (Figure 4f).

Finally, to have insights into the stabilization effect by sAB G50, we performed allatom simulations of the G_i protein complex both with and without the sAB. Simulation without the sAB resulted in a dissociation event of the AHD from the Gβ subunit and a significant increase in heterogeneity of its position (Figure 4g). Indeed, these results highlight the importance of G50 in directly stabilizing of the AHD in one defined position.

5. Affinity maturation of sAB G50

Although sAB G50 was successfully used as a fiducial mark to determine the Rho/G_i complex structure, its effectiveness could be further improved by optimizing its binding kinetics through affinity maturation. G50 binds to G_i with high affinity (~1.2 nM); however, the off-rate (1.4 x 10⁻³ s⁻¹) is suboptimal compared to other sABs we have used for fiducials with other systems. So we undertook a structure guided affinity maturation strategy to obtain variants with improved off-rate kinetics. Based on the structure of the Rho/G_i/G50 complex and the Ala scanning results (Figure 4e) discussed above, affinity maturation phage display libraries were designed using sAB G50 as the template. Five different sub-libraries were generated by diversifying CDRs -L1, L2, L3, H1 and H2. Since CDR-H3 is extensively involved in antigen interaction, it was not included in the affinity maturation process. The new libraries were designed to incorporate residues via “hard” and “tailored” randomizations. The choice of the codons at specific positions was based on the results from the Ala scanning. Residues that were less affected by Ala substitution were subjected to “hard” randomization, thereby introducing more diversity into the generated phage pool whereas positions that showed marked preference to any specific pattern of residues (hydrophobic/polar/acidic/basic) were “tailored” to present similar characteristics in the randomized variants (Table S4).

Clones with improved off-rates were enriched by using two selection strategies. In the first strategy, the stringency was increased by reducing the protein concentration from 10 nM to 10 pM over the course of three rounds. The improved affinity of the generated variants by this strategy can be attributed to either increased on-rates or decreased off-rates or a combination of both. However, since the aim of this campaign was to generate variants with slower off-rates, we additionally used an “off-rate selection” strategy (Lee et al., 2004). Here, additional selection pressure was introduced by incubating the phage pool with a molar excess of the non-biotinylated target for an extended time during each round. This effectively acted as a soluble competitor by “soaking up” the binders with faster off-rates from the phage pool.

The two strategies resulted in different outputs as monitored by the enrichment levels (Figure S4). The first strategy yielded enrichment for all five libraries. However, the total amount of eluted phage particles varied between libraries with the best results for Lib-L3 and Lib-H1 and Lib-H2. For the off-rate protocol, which introduced much stronger selection pressure, positive enrichment was obtained for only two libraries; Lib-L3 and Lib-H2. Considering the extra selection pressure imposed in off-rate selection, this observation was not surprising. 96 clones from each successful selection were picked and analyzed by single-point competition phage ELISA (Figure S5). Subsequently, clones displaying the strongest ELISA signal and competition ratio were sequenced and further characterized by SPR. Analysis of the unique binders revealed high level of diversity in almost all libraries except Lib-H1, where mutations were introduced in only two positions – 33 and 34 (Figure 5a). All the new clones obtained from Lib-L1, independent of selection strategy, contained Val in position 30, which the Ala scan data previously identified as a critical residue (Figure 4e). A similar phenomenon was observed for position 92 in CDR-L3 and 57 in CDR-H2 (Figure 5a), where the energetically important residues Ser and Tyr returned almost exclusively in all clones.

Figure 5. Affinity maturation of sAB G50.

(a) Weblogo plots showing the distribution of the residues of four successfully matured CDRs. (b) Representative SPR sensograms of the affinity matured variants of the sAB G50 from Lib-L1, Lib-L3, Lib-H1, Lib-H2. (c) aSEC profiles and SDS-PAGE analyses of the Rho/G_i/affinity matured sAB complexes.

SPR data showed that the affinities (K_D) of all the affinity matured variants were substantially improved over wt G50. Many had K_D values of <50 pM (Table S5). Further, the increase in affinity is due to a 10- to 1000-fold improvement in the k_off (Figure 5b). Additional quality controls on the variants were performed based on T_m and expression level. Variants with lower T_m and poor expression compared to the wt sAB G50 were triaged and the following clones were chosen for further studies: G50.23 from Lib-L1 (K_D = 220 pM), G50.09 from Lib-L3 (K_D = 50 pM), G50.27 form Lib-H1 (K_D = 270 pM) and G50.12 from Lib-H2 (K_D = 12 pM).

To evaluate the side-chain contributions of the four best affinity matured variants, we performed Ala scanning on the randomized CDRs from all four variants (G50.23, G50.09, G50.27, G50.12) (Table S6). For G50.23, G50.09 and G50.27, all the residues contribute measurably to the antigen binding. Conversely, for sAB G50.12, surprisingly Ala substitutions in five (G53A, G56A, G59A, T61A and S62A) out of the ten positions resulted in improved binding affinity. Although five residues in this variant appear to have a negative contribution to binding, this sAB binds to the trimeric G_i with 12 pM affinity (Table S6).

Finally, we combined all four successfully randomized CDRs in order to generate a “supermutant” of the sAB G50, called G50X. Analysis of the binding kinetics of the sAB G50X with trimeric G_i by SPR (Table S5) showed a K_D value of ~ 50 pM. The combination of the mutations enhanced the expression of the supermutant in comparison to that of each individual affinity matured variants. sAB G50X stabilizes the target antigen by 2 °C as monitored by DSF (Figure S2b). The binding of the affinity matured variants and the supermutant to the Rho/G_i complex were verified by aSEC (Figure 5c).

6. Generation of conformation-specific sABs for GRK1

The phage display selection on GRK1 (K46L) yielded twenty-three high affinity sABs. Ten sABs, ranked by their affinities, were further tested on aSEC to confirm binding to the Rho/GRK1 complex. The best expressing sAB in the cohort was 1F1. It has an affinity of 2.7 nM (Table 2) to the K46L mutant and forms a stable complex withthe Rho/GRK1 with well-defined characteristic features, as analyzed by negative-stain EM (Figure 6d, e).

Table 2.

Kinetic parameters of the characterized sABs for GRK1 (K46L) and GRK1 (K46L/R461A/Q462A) measured by SPR.

	GRK1 (K46L)			GRK1 (K46L/R461A/Q462A)
sAB	kon (M−1s−1)	koff (s−1)	KD (nM)	kon (M−1s−1)	koff (s−1)	KD (nM)
1F1	4.0 x 105	1.0 x 10−3	2.7	2.9 x 105	1.0 x 10−3	3.6
GR2	NB	NB	NB	9.2 x 105	2.4 x 10−4	0.3
GR6	NB	NB	NB	1.8 x 105	1.0 x 10−4	0.6

Figure 6. Characterization of sABs against GRK1.

(a) Conformational specificity of sABs determined by phage ELISA. sABs show stronger binding to the triple mutant of GRK1 over K46L mutant. (b) Epitope binning by SPR - The triple mutant was immobilized. sABs GR2 and GR6 share same epitope. sAB 1F1 has a different footprint and can bind in presence of either of them. (c) DSF melting curves of the GRK1 triple mutant shows stabilization by selected sABs. (d) aSEC profiles of the Rho/GRK1(triple mutant)/sAB complexes overlaid with that of Rho/GRK1(triple mutant). Arrows indicate main peaks. (e, f) Representative negative-stain EM 2D classes of Rho/GRK1(triple mutant)/sAB complexes: (e) with sAB 1F1 and (f) with sABs 1F1 and GR6. (g) Representative cryo-EM 2D classes of Rho/GRK1(triple mutant)/1F1/GR6 from 673 images. Scale bar, 50 Å. The corresponding micrographs of (e), (f) and (g) are in Figure S3.

The effect of the GRK1 (K46L) mutant is to increase its binding to the receptor, but it has low catalytic activity. An additional set of mutations encompassed in GRK1 (K46L/R461A/Q462A) allows the conversion to a more catalytically competent state. However, this comes with some additional conformational flexibility in the kinase with compromised thermal stability (Komolov et al., 2017). Thus, we endeavored to generate a cohort of conformation specific sABs to the triple mutant that can effectively trap and stabilize GRK1 in a particular, low energy (receptor bound) conformation.

Disrupting the ionic lock introduced conformational change in GRK1 triple mutant. sAB 1F1 generated from the previous selection on GRK1 (K46L) is conformationally agnostic, as it binds with equal affinity to K46L and the triple mutant (Table 2) indicating that its footprint is conformationally insensitive. Since our goal was to obtain conformationally specific sABs, we designed three parallel selection strategies to maximize the probability of success (Figure S6). Strategy 1: Competitive selection - Since K46L and the triple mutant are in different conformation states, we did the phage display selection on the triple mutant and used 100-fold molar excess of K46L mutant as a soluble competitor to enrich for binders that preferentially bind to the triple mutant over K46L. Strategy 2: Epitope masking selection - As sAB 1F1 binds to a conformationally insensitive epitope, 1 μM sAB 1F1 was used to occlude the epitope, thereby directing the phage pool binders to other epitopes that are potentially sensitive to conformational change. Strategy 3: Regular selection – Harsh selection pressure for conformation bias can sometimes lead to generating few, if any binders. So, we also did a selection where no additional selection pressure was introduced.

All the three selection strategies had >50-fold enrichment (Figure S7b) and in total produced 36 unique binders as determined from single point competition phage ELISA and DNA sequencing. Binding of the 15 clones that showed strong preference for the triple mutant over K46L were further tested in presence of saturating amount of sAB 1F1 (Figure 6a, 6b). All of the 15 conformational specific binders recognized epitopes different from that of sAB 1F1. SPR showed that some of the sABs displayed high affinity and selectivity. For example, sAB GR2 and GR6, bound only to the triple mutant with sub-nanomolar affinity and did not bind to the K46L at all (Table 2). Most of these sABs have slow off-rate kinetics. Since the ionic lock is disrupted upon complex formation with the GPCR (Komolov et al., 2017), we hypothesize that these high affinity sABs specific for the triple mutant with disrupted ionic lock, mimic the bound state conformation and will assist in the stabilization of the receptor/kinase complex for further studies.

The triple mutant of GRK1 is characterized by very low T_m (29°C) (Figure 6c) that is presumably due to the disruption of the ionic lock leading to instability and high conformational flexibility. To investigate the sAB-mediated stabilization on the triple mutant, we used DSF to determine the T_m of the triple mutant in absence and presence of the sABs (Figure 6c). While sAB 1F1 increased the T_m by 10°C (T_m ~ 39° C), the conformational selective sABs GR2 and GR6 increased the T_m of the target by 24°C. Thus, these sABs were able to overcome the adverse effects of the mutations on the conformational flexibility by stabilizing/trapping the ionic lock mutant in a stable conformation state.

Finally, we tested binding of these new set of sABs to the Rho/GRK1 triple mutant complex in presence of sAB 1F1 by aSEC and negative-stain EM (Figure 6d, e, f). Nine of the tested sABs formed complexes with the Rho/GRK1 triple mutant in presence of 1F1. The representative 2D class averages of the complexes show characteristic features of all the components: receptor in detergent micelle, kinase and both sABs (Figure 6e, f). Class averages of a very small cryo-EM data set from the screening microscope highlights the utility of these sABs as potent fiducials (Figure 6g). Considering the high-affinity, conformational-specificity, imparting thermal stability to the target and forming stable complexes with Rho/GRK1 triple mutant, we conclude that this cohort of sABs embody the pertinent attributes required of stabilizing agents and fiducial marks for structural studies of GPCR/GRK1 complexes.

Discussion

Besides being the largest family of membrane proteins in the genome, GPCRs represent the richest set of therapeutic targets (Sriram and Insel, 2018; Insel et al., 2019). To approach the underlying questions of GPCR biology from a structural perspective requires a two-pronged strategy. First, to further understanding of the general structure-function elements that define GPCRs as a family, a significant number of structures are required, chosen to cover multiple examples of each subclass. On the other hand, the type of high-level structural information that will be necessary for drug development will require details that can’t be inferred from the structure of even a highly related family member. Thus, it is imperative to determine the structures of each of these potential therapeutic targets. At a level, these are conflicting objectives requiring different emphases. Focusing on one objective to the detriment of the other will lead to a situation where new insights will be incremental and overall progress protracted.

With this in mind, we developed an experimental platform that effectively satisfies each of these objectives without compromising the other; that is, greatly increasing throughput, while simultaneously improving accuracy and detail. Employing phage display, we generated “universal” sABs to two sets of different signaling molecules: trimeric G-proteins and GRKs. For the G-proteins, high affinity sABs recognizing multiple epitopes were produced for variants of G_i, G_s and mini-G_s. For the GRK family, multiple sABs against GRK1 were produced. Generally, the sABs for both sets of targets had affinities < 10 nM, with many having K_D values below 1 nM. sABs to trimeric G_i bind to at least four unique epitopes: two on the Gα_i subunit and two that appear to be located on the Gβγ. This latter group has the potential to be versatile fiducials across all forms of trimeric G-proteins. For Gα_s, four sABs were identified to cross-react with the trimeric G_s construct.

The format for the phage display selections put no restrictions on the location of the footprints of the sABs so that some of these might bind an epitope that would interfere with the association of the G-protein or GRK to its cognate receptor. To identify this group, the fidelity of the receptor/G-protein (GRK) complexes in the presence of each sAB was evaluated and the ones that failed this test were triaged. For instance, four of the eight tested sAB binders selected against G_i formed stable complexes with the receptor/G-protein. Likewise, all four tested sABs to Gα_s bound to the receptor/G_s complex.

To test the sABs as fiducial marks for GPCR/G-protein complexes, we selected the Rho/G_i signaling complex as a model system. Rho is a well-characterized GPCR and there had been no previous structures involving G_i associated complexes with any GPCR. The combined mass of Rho (~39 kDa) and G_i (~70 kDa) is only slightly over 110 kDa, making it a difficult SP cryo-EM problem based on size alone. Additionally, the receptor is embedded in a solubilization medium, in this case a detergent, which obscured its structural features (Herzik et al., 2019). Further, G_i has few discernable features and adopts high structural flexibility upon receptor binding. These factors taken together made orientation of the particle a formidable task in the absence of a fiducial mark. In this regard, the impact of adding sAB G50 to Rho/G_i was game changing. The benefit of adding 50 kDa to the mass is obvious since it is well established that signal-to-noise ratio in SP images scales roughly to molecular weight (Henderson et al., 2011). However, the contribution of the sAB’s shape to the orientation of the particle is perhaps an under-appreciated attribute (Wu et al., 2012). A sAB’s structural framework consists of two 25 kDa units (constant and variable domains) that has approximate dimensions of 35 Å x 70 Å. The sAB appears as an appendage extending from the target particle that is easily distinguished in micrographs. Additionally, even at moderate resolutions the region between the variable and constant domains appears as a distinct “hole” thereby adding a higher order feature to help to more accurately align the particle.

Using sAB G50 as a fiducial mark, the cryo-EM structure of the Rho/G_i complex at 4.5 Å shows a well-defined density for the receptor, the entire trimeric G_i assembly and the sAB. The binding site of the sAB is distal to Rho/Gα_i interface confirming that it does not affect the conformation of Rho. Interestingly, the position of the functionally important AHD (Dror et al., 2015; Du et al., 2019) of Gα_i is well defined, presumably since it is stabilized through the sAB G50 as observed in molecular dynamics simulation. It is known that receptor mediated activation of the G-protein leads to GDP release and separation of the AHD from the Ras-like Gα domain inducing flexibility (Westfield et al., 2011; Gao et al., 2017). In other cryo-EM structures of GPCR/G-protein complex, this domain exhibits significant flexibility (Liang et al., 2017; Zhang et al., 2017; Liang et al., 2018a; Koehl et al., 2018; Draper-Joyce et al., 2018; Kumar et al., 2019; Tsai et al., 2019; Zhao et al., 2019). The question, therefore, is whether the AHD conformation observed in this structure is an artifact or represents a true low energy conformation, albeit one of several. We would argue the latter is more probable. We note that AHD is discernable in lower resolution structures in a position similar to the one we observe, but it appears averaged out in the higher resolution reconstructions (Liang et al., 2017; Liang et al., 2018a; Tsai et al., 2019). Moreover, such placement of the AHD is in good agreement with DEER studies of the Rho/G_i complex (Van Eps et al., 2011). The mapping of the interacting interface revealed that the majority of interactions of sAB G50 with G_i are through the AHD of the Gα_i. The contacts between Gβ and G50 were few, energetically neutral and the sAB binds to Gα_i alone with virtually the same affinity as it does to the trimeric G_i (2.5 nM vs. 1.2 nM). These observations suggest that instead of dictating the conformation of the AHD directly, sAB G50 rather provides a low energy packing surface to Gβ that restricts its variability. The recently published structure of Smoothened GPCR/trimeric Gi complex bound to sAB G50 at 3.8 Å reiterates the utility of the chaperone in stabilizing the AHD (Qi et al., 2019).

In contrast to the selection format used for G-proteins, the one used for generating sABs for GRK1 was designed to capture a particular conformation state. The ionic lock between two major domains of GRK stabilizes the inactive state. Receptor induced disruption of the ionic lock and domain separation stabilizes its interaction with the receptor and converts the enzyme into its more catalytically competent state which is the desired conformation. However, as this state is highly flexible, generating conformationally specific sABs presented a formidable challenge. The essential first step was to produce a GRK1 mutant that biased the enzyme into its more catalytically competent conformation. Few selection strategies were used that effectively eliminated binders to other conformational states. This resulted in a cohort of sABs that not only trapped the desired conformational state, but stabilized it significantly, in some cases by over 20°C. Virtually, all the complex specific sABs share a single epitope; however, several non-conformationally selective sABs were also identified. These bound to a different epitope on GRK1 and can bind simultaneously with the conformational specific sABs. This provides for the possibility of producing a dual fiducial that adds 100 kDa to the particle while still confining the GRK to its more active state.

Taken together, it is evident that sABs generated to different classes of signaling partners can be effectively utilized as universal fiducial marks that can be employed in a “plug and play” fashion for most GPCR systems. This greatly simplifies the process of structure determination of signaling complexes by SP cryo-EM for this large family of important receptors. Once high quality samples are produced, assembly with optimized and engaged sAB becomes a straightforward task. This will greatly reduce the time and resources necessary for determining high resolution structures of GPCR signaling complexes and will open up opportunities to study different subfamilies of GPCRs more broadly in a systematic manner.

STAR METHODS

LEAD CONTACT AND MATERIALS AVAILABILITY

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Anthony A. Kossiakoff (koss@bsd.uchicago.edu).

EXPERIMENTAL MODEL AND SUBJECT DETAILS

E. coli BL21(DE3)-RIL cells were used for expression of human Gα_s, human Gα_i1 and mini-G_s. sABs were expressed in E. coli BL21 (Gold) cells. Bacterial cells were cultured using standard protocols in 2xYT or TB-glycerol media. Human D1R, human Rho, chimeric G_s (human-Gα_s, rat-Gβ₁, bovine-Gγ₂) and chimeric G_i1 (human-Gα_i1, rat-Gβ₁, bovine-Gγ₂) were overexpressed in Sf9 insect cells. Heterotrimeric Gs (human-Gα_s, human-Gβ₁γ₂) was expressed in High Five cells. All insect cells were cultured in ESF921 media (Expression systems) at 27°C. See “Method Details” section for more details on construct choice and protein expression.

MATHOD DETAILS

Protein expression and purification

The genes of interest, the short isoform of human Gα_s, human Gα_i1 and mini-G_s with N-terminal 6x His-tag followed by Tobacco Etch Virus (TEV) protease cleavage site were cloned into pET28a bacterial expression vector. Additionally, if biotinylated protein was required, Avi-tag was introduced between TEV protease cleavage site and the gene of interest. All proteins were expressed in E. coli strain BL21(DE3)-RIL as described by Carpenter et al. (2016). The Avi-tagged proteins were co-expressed with E. coli biotin ligase BirA in presence of 50 μM D-biotin for in-vivo biotinylation. Proteins after purification over Ni-NTA resin (Qiagen) were loaded onto a HiLoad 16/600 Superdex 200 column pre-equilibrated in SEC buffer (20 mM HEPES; pH 7.4, 100 mM NaCl, 5% glycerol, 1 mM MgCl₂, 10 μM GDP and 100 μM TCEP). Protein containing fractions were pooled, flash-frozen and stored at −80°C until use.

Heterotrimeric Gi-DN and Gs proteins were expressed in Sf9 insect cells and purified as previously described (Liu et al., 2016). Briefly, two separate baculoviruses were generated using Bac-to-Bac system. The human Gα_i1 containing DN mutations (G203A and A326S) or the short isoform of human Gα_s were fused with the C-terminus of bovine Gγ₂ via short flexible linker, cloned in pFastBac-Duet vector, which contained a cassette for expression of 8x His-tag, Green Fluorescence Protein (GFP) and TEV protease cleavage site followed by Avi-tag at the N-terminus of Gγ subunit. Second virus was generated using the rat Gβ₁ subunit without any tag cloned into pFastBac1 vector. For the co-expression of the Gγ-Gα and Gβ subunits Sf9 cells at a density ~3 X 10⁶/ml were infected with viral stock in 1:1.2 ratio and incubated at 27°C for 48h. Proteins were purified over Ni-NTA resin followed by TEV protease cleavage and SEC in 20 mM HEPES; pH 8.0, 200 mM NaCl, 10 μM GDP and 100 μM TCEP. Peak protein fractions were collected and stored at −80°C.

For the complex formation with D1R, heterotimeric G_s was expressed in High Five insect cells by using two viruses - one expressing human Gα_s subunit and another expressing human Gβ₁γ₂ subunits with a 6x His-tag inserted at the N-terminus of Gβ₁. Cells were harvested after 48h from infection. The cells were lysed in hypotonic buffer and heterotrimeric G_s was extracted in buffer containing 20 mM HEPES; pH 7.5, 100 mM NaCl, 1% sodium cholate, 0.05% DDM, 5 mM MgCl₂, 5 mM β-Mercaptoethanol, 10 μM GDP and 1 unit CIP. The G_s heterotrimer was purified over Ni-NTA resin followed by ion exchange chromatography on MonoQ HR (GE healthcare). The target fractions were concentrated to 10 mg/ml, flash frozen with glycerol and stored at −80°C.

The human MBP-GRK1 fusion construct containing K46L mutation or K46L/R461A/Q462A mutations with N-terminal Avi-tag was sub-cloned into pFastBac-Duet vector and expressed in Sf9 insect cells for 48h. Subsequently, the protein was affinity purified on amylose resin following standard procedures of purifying MBP-fused proteins (Kang et al., 2015).

Constitutively active mutant of Rho (N2C, E113Q, M257Y, N282C) fused with thermally stabilized apocytochrome b562RIL (BRIL) at the N-terminus was cloned into pFastBac1 vector, which contained a cassette for expression of a HA signal sequence followed by a FLAG-tag, a 10x His-tag and a TEV protease recognition site at the N-terminus of the BRIL sequence. Protein was expressed in Sf9 cells as described by Kang et al. (2018). Isolated membranes were thawed and diluted in 20 mM HEPES; pH 7.4, 400 mM NaCl, 10% glycerol, 10 μM all trans-Retinal (ATR) supplemented with Complete protease inhibitors. Detergent stock was added to final 0.5% DDM and 0.1% CHS, and incubated for 3h with gentle mixing. The sample was clarified by ultracentrifugation and the supernatant incubated with TALON resin over night at 4°C for binding. The resin was washed with 20 CV of 20 mM HEPES; pH 7.4, 400 mM NaCl, 10% glycerol, 0.03% DDM, 0.01% CHS, 25 mM imidazole, 5 μM ATR. Protein was eluted in minimal volume with 20 mM HEPES; pH 7.4, 400 mM NaCl, 10% glycerol 0.03% DDM, 0.01% CHS, 250 mM imidazole, 5 μM ATR. Eluted sample was concentrated using 100 kDa cut-off Amicon concentrator and loaded onto a Superdex 200 column pre-equilibrated in 20 mM HEPES, pH 7.4, 100 mM NaCl, 0.03% DDM, 0.01% CHS, 5 μM ATR. Peak fractions were pooled and kept at 4°C for further experiments.

Human full-length D1R with an N-terminal FLAG-tag and a C-terminal 8x His-tag was expressed in Sf9 cells using the Bac-to-Bac system. Cells were infected with virus and cultured for 48h with a reversible antagonist SCH23390 at the concentration of 100 nM in the medium. The induced cells were collected by centrifugation and lysed in buffer containing 20 mM Tris-HCl, pH 7.5, 0.2 μg/ml leupeptin, 100 μg/ml benzamidine and 1 μM SCH23390. After centrifugation, the pellet was re-suspended and solubilized in buffer containing 20mM HEPES; pH 7.5, 750 mM NaCl, 1% DDM, 0.2% sodium cholate, 0.2% CHS, 20% glycerol, 0.2 μg/ml leupeptin, 100 μg/ml benzamidine, 1 μM SCH23390 and 100 μM TCEP at 4°C for 2h. The supernatant was isolated by centrifugation and purified over Ni-NTA resin. The eluate after His-tag purification was directly applied to an anti-FLAG M1 antibody resin (coupled in-house) after adding 2 mM CaCl₂. Ligand was slowly exchanged to 1 μM agonist SKF83959 on the M1 resin. The receptor was eluted with buffer containing 20 mM HEPES; pH 7.5, 100 mM NaCl, 0.1% DDM, 0.01% CHS, 1 μM SKF83959, 200 μg/ml FLAG peptide and 5 mM EDTA. The purified receptor was collected after SEC using Superdex 200 Increase column in 20 mM HEPES; pH 7.5, 100 mM NaCl, 0.05% DDM, 0.005% CHS and 1 μM SKF83959 and was concentrated to ~20 μM for complex formation.

The fusion construct of BRIL-Rho-GRK1 was generated by fusing GRK1 to the C-term of the BRIL-Rho receptor construct (described above) via a flexible 4x GSA linker and the protein was expressed in Sf9 cells for 48h at 27°C. After expression, the Sf9 cell pellets were lysed in 20 mM HEPES; pH 7.5. The supernatant was centrifuged at 160,000g for 30 min to collect the membranes. The membranes were washed by homogenization in 20 mM HEPES, pH 7.5, 1 M NaCl followed by centrifugation at 160,000g for 30 min. The washed membranes were then solubilized in 20 mM HEPES; pH 7.5, 200 mM NaCl, 0.5% DDM, 0.1% CHS for 2h at 4°C. The supernatant was isolated by a centrifugation at 160,000g for 1h, and then was incubated with Ni-NTA resin for 2h at 4°C. After binding, the resin was washed with 20 mM HEPES; pH 7.5, 200 mM NaCl, 0.02% DDM, 0.004% CHS, 50 mM imidazole. The complex sample was eluted with 20 mM HEPES; pH 7.5, 200 mM NaCl, 0.1% digitonin, 250 mM imidazole. The elution was concentrated and loaded onto Superdex S200 10/300 GL column equilibrated with 20 mM HEPES; pH 7.5, 200 mM NaCl, 0.1% digitonin. Purified fractions of the complex were collected for further experiments.

SA Pull – down assay

The pull-down assay was performed using SA magnetic beads at RT and all incubation steps were for 15 min. 3 μg of target was incubated with 100 μl of SA magnetic beads with gentle mixing. Unbound protein fraction (flow-through) was collected and beads were washed three times. The extent of biotinylation was analyzed by comparing the amount of protein in the different fractions (input, flow-through, wash and beads) by SDS-PAGE.

Phage display selection

Selection for all targets was performed according to published protocols (Fellouse et al., 2007, Paduch et al., 2013). In the first round of selection, 200 nM of target was immobilized on 250 μl SA magnetic beads. This was followed by rigorous washing steps to remove unbound protein followed by a 5 min incubation with 5 μM D-biotin to block unoccupied SA sites on the beads to prevent nonspecific binding of the phage pool. The beads were then incubated with the phage library (Miller et al., 2012), containing 10¹²10¹³ virions/ml for 30 min at RT with gentle shaking. The resuspended beads containing bound phages after extensive washing were used to infect freshly grown log phase E. coli XL1-Blue cells. Phages were amplified overnight in 2YT media with 50 μg/mL ampicillin and 10⁹ p.f.u./mL of M13 KO7 helper phage. To increase the stringency of selection, three additional rounds of sorting were performed with decreasing target concentration in each round (2^nd round: 100 nM, 3^rd round: 50 nM and 4^th round: 10 nM and 5 nM) using the amplified pool of virions of the preceding round as the input. Unlike solid capture in the first round, the 2^nd – 4^th rounds involve solution capture. These rounds are carried out in a semi-automated platform using the Kingfisher instrument. From 2^nd round onwards, the bound phages were eluted using 0.1 M glycine; pH 2.7. This technique often risks the enrichment of nonspecific and SA binders. In order to eliminate them, the precipitated phage pool from 2^nd round onwards were negatively selected against 100 μL of SA beads. The “precleared” phage pool was then used as an input for the selection.

Additionally, for the mini-G_s variants and GRK1, which were fused with MBP, competition strategy was applied in order to exclude all potential MBP binders. All buffers used during the selection process from 2^nd round were supplemented with 1 μM MBP. Similarly, in the selection of the conformationally specific sABs against ionic lock mutant of GRK1, 100-fold molar excess of either a soluble competitor (GRK1 (K46L)) or a masking reagent (sAB 1F1) was used from the 2^nd round of selection campaign.

Enzyme-linked immunoabsorbent assays (ELISA)

All ELISA experiments were performed in a 96-well plate coated with 50 μL of 2 μg/mL neutravidin in Na₂CO₃; pH 9.6 and subsequently blocked by 0.5% in PBS. A single-point phage ELISA was used to rapidly screen the binding of the obtained sABs in phage format. Colonies of E. coli XL1-Blue harboring phagemids were inoculated directly into 400 μL of 2YT broth supplemented with 100 μg/ml ampicillin and M13 KO7 helper phage. The cultures were grown at 37 °C for 16-20h in a 96-deep-well block plate. Culture supernatants containing sAB phage were diluted 20-fold in ELISA buffer and transferred to ELISA plates that were incubated with 50 nM of biotinylated target proteins in experimental wells and buffer in control wells for 15 min at RT. The ELISA plates were incubated with the phage for another 15 min and then washed with PBST. The washed ELISA plates were incubated with HRP-conjugated anti-M13 mouse monoclonal antibody (1:5000 dilution in PBST) for 30 min. The plates were again washed, developed with TMB substrate and quenched with 1.0 M HCl, and absorbance (A₄₅₀) was determined. The background binding of the phage was monitored by the absorbance from the control wells.

For the single-point competitive phage ELISA, sAB phage were diluted in PBST with or without with 50 nM soluble competitor (non-biotinylated target) and incubated for 15 min at RT. Subsequently, the mixtures were transferred to ELISA plates with immobilized 25 nM of target and incubated for additional 15 min at RT. Further steps were performed as described above.

Protein based multipoint ELISA was performed to estimate the affinity of the generated unique sABs to their cognate antigens. A fixed concentration of the immobilized target (50 nM) on ELISA plate was incubated with 3-fold serial dilutions of the purified sABs starting from 1μM for 15 min. The plates were washed and the bound antigen-sAB complexes were incubated with a secondary HRP-conjugated anti-human F(ab’)₂ monoclonal antibody (1:5000 dilution in PBST). As with phage ELISA, the plates were again washed, developed with TMB substrate and quenched with 1.0 M HCl, and absorbance (A₄₅₀) was determined. To determine the affinity, the data were fitted in a non-linear sigmoidal function with variable slope in GraphPad PRISM and EC₅₀ value was calculated.

Cloning, overexpression and purification of sABs

Positive clones selected based on a phage ELISA were sequenced at DNA Sequencing Facility at The University of Chicago. Unique clones were reformatted using Infusion cloning (Raman and Martin, 2014) in pRH2.2, an IPTG inducible vector for bacterial expression of sABs.

E. coli BL21 (Gold) cells were transformed with sequence-verified clones of sABs in pRH2.2. sABs were grown in 2YT media with 100 μg/mL ampicillin at 37°C for 2-2.5 h during which OD₆₀₀ reached 0.6–0.8, induced with 1 mM IPTG and further grown for 4.5h at 37°C. The sABs were purified using affinity (protein A / protein G) followed by ion-exchange chromatography using an automated program on ÄKTA explorer system (Mukherjee et al., 2015). Briefly, pellets were resuspended in PBS, supplemented with 1 mM PMSF, 1 μg/mL DNase I. The suspension was lysed by ultrasonication. The cell lysate was incubated at 63°C for 30 min. Heat-treated lysate was then cleared by centrifugation, filtered through 0.22 μm filter and loaded onto a HiTrap MabSelect SuRe (GE Healtchcare) column or HiTrap NHS-Activated HP affinity column (GE Healtchcare) coupled with protein G-A1 (Bailey et al., 2014) pre-equilibrated with 20 mM Tris; pH 7.5, 500 mM NaCl. The column was washed extensively with 20 mM Tris; pH 7.5, 500 mM NaCl followed by elution of sABs with 0.1 M acetic acid for protein A purification or 0.1 M glycine; pH 2.7 when protein G-A1 resin was used. Fractions containing protein were directly loaded onto an ion-exchange Resource S 1-ml column pre-equilibrated with 50 mM NaOAc; pH 5.0. Column was washed with the equilibration buffer and sABs were eluted with a linear gradient 0–50% of 50 mM NaOAc; pH 5.0, 2 M NaCl. Purified sABs were dialyzed overnight against 20 mM HEPES; pH 7.5, 150 mM NaCl. The quality of purified sABs was analyzed by SDS-PAGE.

Epitope binning and binding kinetics by SPR

All SPR experiments were performed at 20°C using ei ther BIACORE 3000 (GE Healthcare) or MASS-1 (Bruker) instrument. All targets, expect GRK1 were immobilized onto a nitrilotriacetic acid (NTA) sensor chip via His-tag. Because the construct of GRK1 lacks a His-tag but has an Avitag, immobilization was done using Biotin CAPture Kit (GE Healthcare). In the binning experiments, after injection of the saturating concentration of the first sAB, the equal molar mixture of the first sAB and a tested sAB was injected. For the sABs, which were binding to epitopes different from the first sAB, net increase in response units was observed.

For kinetic experiments, 2-fold serial dilutions of the sAB were injected following ligand immobilization on the sensor chip. For each kinetic assay at least five dilution of the sAB were tested.

Thermal shift assay

Thermal stability of samples was analyzed by Differential Scanning Fluorimetry (DSF) using a real-time PCR instrument (Bio-Rad CFX384) (Niesen et al., 2007). Protein samples were mixed with 4x Sypro Orange at a final 4 μM concentration. For target/sAB complexes, 1.5-fold molar excess of sAB was used. All reactions were performed in triplicates in a 384-well plate. Thermal melting curves were monitored from 25°C to 95°C in steps of 0.5°C/30 s. Wavelengths of 490 and 575 nm were used for excitation and emission, respectively.

GPCR/G-protein complexes formation

Purified and concentrated to ~ 20 μM, constitutively active mutant of Rho was mixed with 1.5-fold molar excess of G_i-DN and incubated for 30 min at RT. Subsequently, to allow the formation of the nucleotide-free complex, 25 mU/ml Apyrase (NEB) was added and the mixture was incubated for additional 1h at RT. Formed Rho/G_i complex was isolated by SEC in 20 mM HEPES; pH 7.4, 100 mM NaCl, 0.03% DDM, 0.01% CHS, 5 μM ATR. Peak fractions were concentrated and used for sAB binding experiments.

Purified SKF83959-bound D1R was mixed with G_s at a molar ratio of 1:2 for 1h at RT. Then 25 mU/ml Apyrase (NEB) was added and the mixture was incubated for additional 2h at 4°C to allow the formation of the nucleotide-free complex. The complex was purified by M1 resin to remove excessive G-protein. The eluate was applied to size exclusion chromatography on Superdex 200 column in 20 mM HEPES; pH 7.5, 100 mM NaCl, 0.05% DDM, 0.005% CHS, 100 μM TCEP and 1 μM SKF83959. Peak fractions were collected and concentrated to at least 0.2 mg/ml to further test the binding of various sABs.

Target/sAB complexes formation

Target/sAB complexes were formed by mixing 2-fold molar excess of the sAB with respective target followed by 30 min to 1h incubation on ice. If binding of more than one sAB (or Nb) was tested each sAB was added in 2-fold molar excess over the target protein/ protein complex. Subsequently, complex was injected onto a Superdex 200 increase column pre-equilibrated in target buffer. Formation of target/sAB complex was determined by analyses of retention volume of the complex with respect to that of target alone and co-elution of the individual components on SDS-PAGE.

Negative-staining EM analysis

The target/sAB complex was applied to a freshly glow-discharged carbon coated copper grid for 1 min before being reduced to a thin film by blotting. Immediately after blotting, 3 μl of a 1% solution of uranyl formate was applied to the grid for 1 min followed by blotting off and air dry. The negatively-stained sample was imaged at RT with a Tecnai Spirit electron microscope at Van Andel Institute operated at 120 kV. Images were recorded at a magnification of 30 000x and processed using Relion 2.1 software.

Cryo-EM analysis

To ascertain how well the target/sAB complexes perform as cryo-EM samples after vitrification, 3 μl of purified Rho/GRK1/1F1/GR6 complex was applied to a glow-discharged 400-mesh grids (Quantifoil R1.2/1.3) and subsequently vitrified using a Vitrobot. Cryo-EM imaging was performed on a Talos Arctica microscope operated at 200 kV at magnification of 120,000x using a Falcon camera, corresponding to a pixel size of 1.21 Å. A small dataset of only 673 image stacks were obtained with a defocus range of −1.0 to −3.0 μm. Image stacks were subjected to beam-induced motion correction using MotionCor2 and CTF parameters for each micrograph were determined byCTFFIND-4.1. Particle selection and two-dimensional classification were performed using Relion 3.0 software.

Affinity maturation of sAB G50

Sub-libraries for affinity maturation were prepared using “stop template” version of the phagemid encoding sAB G50 using described methods (Sidhu et al., 2004). For preparation of a stop template, stop codons (TAA) were introduced separately in all five CDRs chosen for construction of libraries (CDR-L1, CDR-L2, CDR-L3, CDR-H1 and CDR-H2). Kunkel mutagenesis (Kunkel et al., 1987) was used to repair stop codons and introduce desired mutations, detailed in Sup. Table 3. Purified reaction products were electroporated in E. coli SS320 cells and after 1h of recovery at 37°C, phag e production was initiated by addition of M13 KO7 helper phage. Cells were grown for 20h at 37°C in 2YT media supplemented with 100 μg/ml ampicillin with vigorous shaking. Phage were harvested using PEG/NaCl precipitation method and re-suspended in 6 mL of buffer containing 50% glycerol. The actual diversity of the libraries was determined by infecting log phase XL-1 Blue cells with the phage library. All data are summarized in Table S4.

Selection for affinity maturation was done using two independent protocols – the regular selection and “off-rate selection”. For each sub-library, regular selection campaigns were performed for three rounds with different target (biotinylated Gi-DN) concentrations – (a) 10 nM in first round, 1 nM in 2^nd round and 100 pM in 3^rd round, (b) 1 nM in 1^st round, 0.1 nM in 2^nd round and 10 pM in the 3^rd round. The solution capture technique where the biotinylated target was first incubated with the phage pool was employed in all the rounds.

Off-rate selection strategy was also tested in parallel to the regular selection campaign as described by Lee et al. (2004). Here also, two different concentrations of target (biotinylated Gi-DN) were tested for each round for each sub-library – (a) 10 nM and 1 nM in 1^st round (b) 1 nM and 0.1 nM both in 2^nd and 3^rd rounds. In the 1^st round, the biotinylated target was mixed with 10¹¹ virions from each sub-library. Subsequently, a 1000-fold molar excess of the non-biotinylated target was added to and incubated with the mixture of the biotinylated target and the phage pool at RT for 15 min with gentle shaking. This step ensures elimination of the binders with faster off-rates from the phage pool by the excess of non-biotinylated target acting as a competitor in solution. Although the target concentration was not dropped from the 2^nd to the 3^rd round, the incubation time with the non-biotinylated target (competitor in solution) was extended to 30 min and 2 h in 2^nd and 3^rd round, respectively. The phage clones obtained from the regular selection protocol and the off-rate maturation strategy were analyzed by single point competitive ELISA.

Molecular dynamics simulation

All-atom atmospheric simulations of the G_i protein complex both with and without sAB G50 were performed using GROMACS 5.0.6 (Abraham et al., 2015) in the NPT ensemble with periodic boundary conditions and the CHARMM36m force field (Huang et al., 2016). The system was prepared by removing Rho and all heteroatoms from the Rho/Gi/G50 complex cryo-EM structure (PDB ID: 6CMO) leaving only the G_i protein subunits and sAB G50. Titratable residues were left in their dominant state at pH 7.0 and all His were represented with hydrogen atoms on the epsilon nitrogen. The resulting systems were solvated in a box of TIP3P3 waters allowing for at least 14 Å of padding on all sides with 150 mM NaCl, and neutralized by removing appropriate ions or counter ions using the Desmond system builder within Maestro (Schrödinger Release 2018-1: Maestro, Schrödinger). Final system dimensions for the G_i and G_i/G50 complex systems were 117 Å x 108 Å x 114 Å and 112 Å x 100 Å x 117 Å and comprised of 139,181 and 128,336 atoms, respectively.

As the receptor was removed from the G_i complex, positional restraints (1,000 kJ mol⁻¹ nm⁻²) were placed on all Gα_i backbone atoms within 5 Å of the Rho interface. Before production simulations, 50,000 steps of energy minimization were performed followed by equilibration in the NVT and NPT ensembles for 10 and 50 ns, respectively, with positional restraints (1,000 kJ mol⁻¹nm⁻²) placed on heavy atoms. A second round of NTP equilibration for 50 ns was run with positional restraints (1,000 kJ mol⁻¹nm⁻²) on backbone atoms to allow for sidechain relaxation. System temperature was maintained at 310 K using the v-rescale method with a coupling time of 0.1 ps and pressure was maintained at 1 bar using the Berendsen barostat with a coupling time of 1.0 ps and compressibility of 4.5 × 10⁻⁵ bar⁻¹ with isotropic coupling. Simulations were performed with a 4-fs time step with hydrogen mass repartitioning (Hopkins et al., 2015) to increase sampling and all bond lengths were constrained using LINCS. Electrostatic interactions were computed using the particle mesh Ewald (PME) method with non-bonded interactions cut at 10.0 Å.

For each system, 2 independent production simulations were performed for approximately 1.5- μs using the Parrinello–Rahmam barostat with a coupling time of 5.0 ps. During production, trajectory snapshots were saved every 10 ps. Simulation analysis was performed using MDTraj 1.7.2 (McGibbon et al., 2015) and VMD 1.9.2 (Humphrey et al., 1996). Plots were generated using the R statistical package.

QUANTIFICATION AND STATISTICAL ANALYSIS

Error bars in Figure 2c represents mean ± SD. Data obtained from independent experiments were analyzed in GraphPad Prism. Normalization of SEC profiles (Figure 3 and 6) was performed using GraphPad Prism. The melting temperatures (Figure 6 and S2) were inferred from first derivative of the melting curves using CFX software provided by the manufacturer. SPR data after double-reference subtraction data were fitted to a 1:1 model and analyzed with BiaEvaluation or Sierra Analyser software.

For details on cryo-EM and molecular dynamics data analysis see “Methods Details”.

DATA AND CODE AVAILABILITY

Molecular dynamics simulation system parameters and trajectories are available upon request.

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
HRP-conjugated anti-M13 mouse monoclonal antibody	GE Healthcare	Cat# 27942101
HRP-conjugated anti-human F(ab’)2 monoclonal antibody	Jackson ImmunoResearch	Cat# 109-035-006
Various Fab-based antibodies (sABs)	This paper	N/A
Bacterial and Virus Strains
E. coli BL21(DE3)-RIL	Agilent Technologies	Cat# 230245
E. coli XL1-Blue	Agilent Technologies	Cat# 200236
E. coli BL21 (Gold)	Agilent Technologies	Cat# 230130
E. coli SS320	Lucigen	Cat# 60512-1
M13 KO7 helper phage	NEB	Cat# N0315S
Chemicals, Peptides, and Recombinant Proteins
D-biotin	Sigma-Aldrich	Cat# 47868
Ni-NTA resin	Qiagen	Cat# 30230
TALON resin	Clontech	Cat# 635507
anti-FLAG M1 antibody resin	This paper	N/A
FLAG-peptide	GenScript	Cat# RP10586
GDP	Sigma-Aldrich	Cat# G7127
TCEP	GoldBio	Cat# TCEP50
Sodium cholate	Sigma-Aldrich	Cat# C1254
n-Dodecyl-β-D-Maltopyranoside (DDM)	Anatrace	Cat# D310
Cholesteryl hemisuccinate (CHS)	Anatrace	Cat# CH210
Digitonin	Sigma-Aldrich	Cat# D141
CIP	NEB	Cat# M0290S
all trans-Retinal	Sigma-Aldrich	Cat# R2500
SCH23390	Cayman Chemical	Cat# 15631
cOmplete™, EDTA-free Protease Inhibitor Cocktail	Roche	Cat# 11836170001
Leupeptin	Sigma-Aldrich	Cat# L2884
Benzamidine	Sigma-Aldrich	Cat# B6506
Streptavidin MagneSphere Paramagnetic Particle	Promega	Cat# Z5482
NeutrAvidin	Thermo Fisher	Cat# 31000
Sypro Orange	Thermo Fisher	Cat# S6651
Apyrase	NEB	Cat# M0398S
IPTG	GoldBio	Cat# I2481C50
Uranyl formate	Electron Microscopy Sciences	Cat# 16984-59-1
Critical Commercial Assays
In-Fusion® HD Cloning Plus	Takara Bio	Cat# 639650
Deposited data
6CMO	Kang et al., 2018	RCSB PDB ID: 6CMO
Experimental Models: Cell Lines
Insect cell line Sf9	Expression Systems	Cat# 94-001S
Insect cell line High Five	Thermo Fisher	Cat# B85502
Oligonucleotides
L1_randomization mutagenesis primer	IDT	CTGCCGTGCCAGTCAGNNCNTTDVTNVTGCTGTAGCCTGGTATC
L2_randomization mutagenesis primer	IDT	GCTCCGAAGCTTCTGATTTDKVVCRSTNNCNNCNTTNNCTCTGGAGTCCCTTCTCGC
L3_randomization mutagenesis primer	IDT	CTTATTACTGTCAGCAANNSNNSNNSNNSNTTNTTACGTTCGGACAGGGTAC
H1_randomization mutagenesis primer	IDT	CAGCTTCTGGCTTCAACNTTNNKNNKNNKNNKATACACTGGGTGCGTC
H2_randomization mutagenesis primer	IDT	CTGGAATGGGTTGCARSTNTTNNKNNKNVTNNTRSTVVCVVCTCTTATGCCGATAGCGTC
Recombinant DNA
pET28a-miniGs	This study	N/A
pET28a-Gαs	This study	N/A
pET28a-Gαi1	This study	N/A
pFastBacI-Gβ1	Liu et al., 2016	N/A
pFastBac Duet- Gγ2Gαi-DN/BirA	Liu et al., 2016	N/A
pFastBac Duet- Gγ2Gαs/BirA	This study	N/A
pFastBacI-Gαs	This study	N/A
pFastBac Duet-Gβ1γ2	This study	N/A
pFastBac-Duet-GRK1/BirA	This study	N/A
pFastBacI-Rho	Kang et al., 2018	N/A
pFastBacI-D1R	This study	N/A
pFastBac1-Rho_GRK1 fusion construct	He et al., 2017	N/A
RH2.2-various sABs	This study	N/A
Software and Algorithms
Prism 8.2	GraphPad	https://www.graphpad.com
BiaEvaluation	GE Healthcare	https://www.biacore.com/lifesciences/service/downloads/software_licenses/biaevaluation/
Sierra Analyser	Bruker	https://www.bruker.com/products/surface-plasmon-resonance.html
CFX software
Relion 2.1	Fernandez-Leiro and Scheres, 2017	https://www3.mrc-lmb.cam.ac.uk/relion/index.php/Main_Page
MotionCor2	Zheng et al., 2017	https://emcore.ucsf.edu/ucsf-motioncor2
CTFFIND-4.1	Rohou and Grigorieff, 2015	http://grigoriefflab.janelia.org/ctffind4
GROMACS 5.0.6	Abraham et al., 2015	http://www.gromacs.org
CHARMM36m	Huang et al., 2016	https://www.charmm.org/charmm/
Maestro	Schrödinger	Release 2018-1: Maestro
MDTraj 1.7.2	McGibbon et al., 2015	http://mdtraj.org/1.7.2/index.html
VMD 1.9.2	Humphrey et al., 1996	http://www.ks.uiuc.edu/Research/vmd/
R statistical package	The R Foundation	http://www.R-project.org
Other
DNA Sequencing	Sequencing Facility at The University of Chicago	http://cancer-seqbase.uchicago.edu
Electron Microscopy	David Van Andel Advanced Cryo-Electron Microscopy Suite	https://cryoemcore.vai.org
Ultra-Thin copper grids, 300 mesh	Electron Microscopy Sciences	Cat #CF300-Cu-UL
R1.2/1.3 holey copper film grids, 400 mesh	Electron Microscopy Sciences	Cat #Q4100CR1.3
FEI Vitrobot	Thermo Fisher	N/A
FEI Talos Arctica with Falcon detector	Thermo Fisher	N/A
FEI Tecnai Spirit	Thermo Fisher	N/A

Highlights.

• sABs to trimeric G-proteins, miniG and GRK have been generated and characterized

• sABs can be used as “plug and play” fiducial marks for GPCR/G-protein complexes

• sABs to GRK bind to multiple epitopes in a conformation sensitive manner

• EM data demonstrate that sABs are potent fiducials for GPCR-signaling complexes