Bicyclic peptide library screening for the identification of Gαi protein modulators

Abstract

Non-canonical G protein activation and inactivation, particularly for the Gαi/s protein subfamilies, has long been a focus of chemical research. Combinatorial libraries were already effectively applied to identify modulators of the guanine nucleotide exchange as can be exemplified with peptides such as KB-752 and GPM-1c/d, so-called guanine nucleotide exchange modulators. In this study, we identified novel bicyclic peptides from a combinatorial library screening which show prominent properties as molecular switch-on/off modulators of Gαi signaling. Among the series of hits, the exceptional paradigm of GPM-3, a protein and state-specific bicyclic peptide, is the first chemically identified GAP (GTPase-activating protein) modulator with high binding affinity for Gαi protein. Computational analyses identified and assessed the structure of the bicyclic peptides, novel ligand-protein interaction sites and their subsequent impact on the nucleotide binding site. This approach can therefore lead the way for the development of efficient chemical biological probes targeting the Gαi protein modulation within a cellular context.

Graphical Abstract

INTRODUCTION

The heterotrimeric Gαβγ proteins undoubtedly have an essential role in numerous signaling pathways. G proteins are classified into the four main subfamilies: the Gαs, Gαi/o, Gαq/11, and Gα12/13^1,2. These proteins are activated by G protein-coupled receptors (GPCRs) and oscillate between the inactive GDP-bound and the active GTP-bound state, with the latter interacting with multiple effector proteins, such as the adenylyl cyclase (AC) and phospholipase Cβ via protein-protein interactions (PPIs)³. Dysregulation or abnormalities of the G protein-mediated signaling pathway can be associated with several diseases, including different cancer types^4–6. As recently described, the development of G protein modulators as chemical tools is fundamental to obtain further insight into the G protein-mediated signal transduction^7–10.

G proteins have been designated in the past as “undruggable” and challenging targets due to their intracellular localization and the close structural homology between the members of the Gα protein families¹¹. In our recent study, we were able to obtain from a one-bead-one compound (OBOC) combinatorial peptide library screening a promising linear peptide as a lead compound (GPM-1, RWLRYLRYP), which was further optimized by cyclization and conjugation to a cell-penetrating peptide (CPP)⁷. Such high-throughput screening (HTS) technologies, including also the one-bead-two-compound (OBTC) libraries and the random nonstandard peptide integrated discovery (RaPID) system^12–15, have been frequently used for the identification of high-affinity cyclic peptide binders to a variety of proteins. This can be exemplified with the bicyclic cyclorasin B4–27¹⁶ acting as an inhibitor for the monomeric G protein K-Ras. Apart from the possibility to insert unnatural amino acids, in particular bicyclic peptides may have the advantage of an enhanced rigidity that may lead to an increased metabolic stability, and even cell permeability^{7, 13,14,17–20}. In this way, a variety of macrocyclic peptides that bind to e.g., the Gαi∙GDP protein have already been identified, such as cycGiBP²⁶, cycPRP-1²⁷, cycPRP-3²⁷, Gα SUPR peptide²⁸, as well as GPM-1b and GPM-1d⁷. Concerning these peptides, no structural investigation has been reported regarding the protein-macrocyclic peptide complex formation for both protein states (active and inactive)^26–28. The aforementioned linear peptide GPM-1 was derived from a peptide library screening against Gαi1·GDP and displayed high similarity to the phage display-derived peptide KB-752^7,29. The latter one was described to act as a guanine nucleotide exchange modulator (GEM), i.e. a guanine nucleotide exchange factor (GEF, Figure 1a) for Gαi and a guanine dissociation inhibitor (GDI) for Gαs⁸, similar to the bifunctional GIV/Girdin, the prototypical member of the GEM family^30–33. Considering the effectiveness of G protein modulators, however, it is required that the molecules are taken up by the cells. There are different mechanisms available by which peptides can penetrate the cell membrane^{14, 19,34–36}; when we tagged GPM-1-derived peptides with a CPP sequence, GPM-1c/d showed a GEM-like activity in cells⁷. This indicates that Gα protein modulators can be obtained from combinatorial peptide library screening and with proper chemical alterations, such as the addition of unnatural amino acids^37,38, can penetrate the cell and modulate intracellular cascades^35,39. The potential of macrocyclic peptides modulating the Gα protein activity is also evident from the selective Gαq inhibitors FR900359 and YM-254890 (both depsipeptides), which have already found widespread utility in various pharmacological studies^40–43. However, since the Gαq subfamily is already well targeted by the natural products, we turned our attention to the Gαi subfamily, which has not yet been adequately addressed by either small molecules or peptides. In this study, we screened a bicyclic peptide library containing both proteinogenic and unnatural amino acids^9,18 against Gαi (in both active and inactive states). This led to the discovery of bicyclic peptides (e.g., GPM-2 and GPM-3) which exhibit promising biological activities, e.g., on the production of second messenger cAMP. Further, computational analyses revealed the binding areas on the protein which could account for this biological activity. To the best of our knowledge, these represent the first peptidyl ligands with GAP activity towards the Gαi protein, in either the active and inactive state.

Figure 1:

Development of a consensus sequence for Gαi binding peptides based on the peptide library screening. (a) Schematic representation of Gα protein signal transduction. (b) One-bead-two-compound (OBTC) library used in this study on TentaGel microbeads. The outer layer displays a unique bicyclic peptide, and the inner layer contains the corresponding linear peptide as an encoding tag. B: β-alanine; M: methionine; R: arginine; X: random amino acid residues; F: phenylalanine; Hmb, hydroxymethylbenzoyl; Pra, propargylglycine^21–24. (c) Example of a representative PED-MALDI-MS spectrum of a hit sequence, i.e. Ser-D-Pro-D-Thr-D-Phe-Lys*-D-Ala-D-Pro-Fpa-D-Phe. (d) Amino acid preferences towards Gαi1·GDP (top) and Gαi1·GMPPNP (bottom) at the different positions are given relative to the central Lys/Orn. Relative frequency of amino acid properties (≥ 20 %) are given in symbols derived from Aasland et al.²⁵ as follows: Φ: hydrophobic (V, I, L, F, Fpa, Nal, Phg, W, Y, M), Ω: aromatic (F, Fpa, Nal, Phg, W, Y), Ψ: large aliphatic (V, I, L, M), ζ: uncharged hydrophilic (N, Q, S, T), [+]: basic (H, K, Orn, R) and [−]: acidic amino acids (D, E). The derived trend is presented above. (e) Selected hit sequences from each screening experiment.

RESULTS & DISCUSSION

Identification of state-selective Gαi1 binding peptides by screening of an OBTC library.

We previously demonstrated that linear peptidyl modulators of the Gα protein can be obtained by screening an OBOC combinatorial peptide library⁷. However, linear peptides are proteolytically labile and have poor cell permeability⁴⁴. To overcome these limitations, we opted to screen a bicyclic peptide library, as the latter approach has recently led to the discovery of potent, cell-permeable, and metabolically stable bicyclic peptidyl inhibitors against the monomeric G protein K-Ras (Figure 1b)⁹. Given the structural similarity between monomeric G proteins and the GTPase domain of the Gα-subunit of heterotrimeric G proteins, we hypothesized that Gα-binding peptides and modulators of Gαi might also be identified by screening a bicyclic peptide library. We chose a bicyclic OBTC peptide library previously reported by Lian et al.¹⁸ and Upadhyaya et al.⁹ In this library, each peptide ring contained three to five random residues and each library bead displayed a unique bicyclic peptide (~50 pmol) on its surface and the corresponding linear peptide (~50 pmol) in its inner sphere. The peptide library was synthesized on 90-μm TentaGel beads (2.86 × 10⁶ beads per gram) and has a theoretical diversity of 6.6 × 10¹³. The peptide library was subjected to two rounds of screening against biotinylated (Btn-) Gαi in the inactive (GDP-bound) or active (GMPPNP-bound) state. In the first round, the bicyclic library (~5.4·10⁵ beads) was incubated with Btn-Gαi protein and streptavidin-coated magnetic particles followed by isolation of the positive beads by magnetic sorting¹⁸. Next, the positive beads from above were incubated again with Btn-Gαi and a streptavidin-alkaline phosphatase (SA-AP) conjugate^18,24. Subsequent incubation with 5-bromo-4-chloro-3-indolyl phosphate (BCIP)²⁴ resulted in a turquoise color on positive beads, which were manually isolated under a dissecting microscope. Thirty-five beads for Gαi·GDP and thirty-six for Gαi·GMPPNP were isolated and their sequences were determined by the partial Edman degradation (PED)-mass spectrometry method (Figure 1c)^45,46 to give twelve and fifteen complete, unambiguous sequences, respectively (Table S2–3). Four representative peptides against Gαi-inactive (peptides 1–4) and four against Gαi-active (peptides 5–8) were selected for re-synthesis and testing (Figure 1e). Details on the establishment of the consensus sequence and peptide bioanalytical characterization are described in the Supporting Information (Figures S1–S4, Table S4).

Validation of library hits by membrane-based assays and SPR.

To identify bona fide Gα modulators, peptides 1 – 8 were initially tested in a cell-free ELISA for their effect on the function of AC^7,47. As depicted in Figure 2, all assays were performed using a membrane preparation from NG108–15 hybrid cells stably expressing the human β₂-AR, where the final readout is taken into consideration. Three different systems were tested: first a receptor-independent and then, a β₂-AR-bound (specific for Gαs) system, where NG108–15 membrane preparations were stimulated with forskolin (Fsk, Figure 2a) and isoproterenol (Iso, Figure 2b), respectively, stimulating the subsequent cAMP production. Thirdly, a δ-opioid-bound (specific for Gαi) approach was used, where DADLE binds to the receptor and prevents the activation of AC (stimulated by Fsk, Figure 2c). The cell-free assay showed that, when compared to the net Fsk-stimulation, peptide 3 (referred to as GPM-2 hereafter, 10 μM) decreased the cAMP production (or activated Gαi signaling) by ~40%, whereas peptide 8 (GPM-3) increased the cAMP level (Fsk and Fsk + DADLE stimulation) by ~30% (Figure 2d). The peptide specificity against Gαi signaling was tested upon Iso stimulation, where no notable change was distinguished. Peptide 1 exhibited no significant effect on Gαi signaling and was used as a negative control in subsequent studies.

Figure 2:

Preliminary studies for peptide selection (10 μM). (a-c) Illustration of membrane-based assays on the NG108–15 membrane preparations^7,47. (a) Forskolin (Fsk) stimulates adenylyl cyclase (AC) and increases the cAMP level in a receptor-independent manner. (b) Upon Isoproterenol (Iso) binding to the β₂-adrenergic receptor, Gαs signaling is getting activated, which in turn stimulates AC leading to an increased cAMP level. (c) Fsk + DADLE: Fsk stimulates cAMP production by direct activation of AC, whereas DADLE binds to the δ-opioid receptor, inhibits the direct AC activation by Fsk and activates the Gαi signaling with subsequent intracellular cAMP decrease. (d) Induced cAMP levels with or without ligand incubation, normalized to Fsk (no peptide state) stimulation. Shown are percentage values of membranes incubated with the respective peptide in the presence of Iso, Fsk and Fsk + DADLE. Error bars represent SD for n = 3. Statistical analysis was performed using the two-way ANOVA test, with **p<0.001, ***p<0.0003 and ****p<0.0001 for comparisons with the control (w/o). (e, f) SPR data of immobilized Gαi1∙GDP with GPM-2 (e) and Gαi1∙GMPPNP with GPM-3 (f), n = 1. (↑): upregulation, (↓): downregulation.

We next tested four of the eight peptides (1 as negative control, 2, GPM-2 and GPM-3) for binding to Gαi by surface plasmon resonance (SPR) analysis. Gαi∙GDP/GMPPNP was immobilized onto the surface of a sensorchip (via amine coupling, Figure S6a–f, Table S5) and varying concentrations of the peptides were flown over the surface. GPM-2 and peptide 2 bound to Gαi∙GDP with K_D values of 5.0 ± 0.34 (Figure 2e) and 24.7 ± 0.95 μM (Figure S6b, Table S5), respectively. In the case of GPM-2, the observed binding affinity lies within the range of previously reported dissociation factors of the used screening library (1–10 μM)^{9, 18,38}. GPM-3 bound to Gαi∙GMPPNP with a K_D value of 0.71 ± 0.37 μM (Figure 2f); as expected, it bound to Gαi∙GDP with much lower affinity (K_D = 25.5 μM, Figure S6c, Table S5). Also as expected, the negative control peptide 1 did not bind Gαi∙GDP (Figure S6a). Binding assays were also performed in the reverse direction, by immobilizing the peptides on the SPR surface and flowing protein solutions over the surface. No binding was observed, presumably because the immobilization procedure modified the peptide structure thereby interfering with protein binding (Figure S6g,h,i). To our knowledge, GPM-3 represents the first potent, selective peptidyl ligand with a sub-μM binding affinity for the GMPPNP-bound Gαi protein^{9, 23,26–28}.

GPM-3 induces Gαi-GTP hydrolysis.

The cell-free studies suggest that GPM-3 may possess GAP activity. To test this notion, we subsequently assayed the selected bicyclic peptides in HeLa cells. To assess the AC activation, we performed a series of receptor-independent bioluminescence resonance energy transfer (BRET) studies by utilizing a well-established system (Figure 3a), which monitors the intracellular cAMP levels in real time and in live cells with a CAMYEL (cAMP sensor using YFP-Epac-RLuc) construct^48,49. The cells were first incubated with varying concentrations of the peptides (0 – 10 μM, Figure S7a) and cAMP production was measured upon Fsk- stimulation. These experiments revealed that for all three peptides, 2.5 μM ligand is optimal for exerting an effect on AC (Figure 3b,c). GPM-3 was most active in this assay and increased the cAMP levels by ~10%, in agreement with the results from the membrane-based assays. Peptides GPM-2 and 2 caused a decrease of the cAMP level, although with lower potencies compared to the data from the membrane assays. Again, peptide 1 had no significant effect. This observation, together with the fact that GPM-3 binds preferentially to the active state of Gαi, suggests that GPM-3 is a potential GAP modulator.

Figure 3:

(a) Schematic representation of the BRET-based assays using a CAMYEL construct to detect the intracellular cAMP levels in live cells upon Fsk stimulation^48,49. (b, c) The bar graph (b) represents the % cAMP production from CAMYEL-based BRET assays, where HeLa cells were incubated for 30 min with 2.5 μM (or without) ligand. The resulting BRET signal was measured after cell stimulation with Fsk. A significant cAMP increase was observed for GPM-3. A concentration-dependent effect on cAMP level is illustrated on panel c) for GPM-2/-3. (d, e) The bar graph (d) represents the % GRK3 (free Gβγ biosensor) production from BRET assays (30 min incubation with 2.5 μM (or w/o) ligand. No significant change was observed for GPM-3, assuming the Gαi heterotrimeric reassociation. A concentration-dependent effect on GRK3 level is illustrated on panel e) for GPM-3 and peptide 1 (neg. control). (f) End-point cellular studies (TR-FRET) for intracellular-cAMP level determination upon Fsk stimulation utilizing the LANCE cAMP kit. The quantification was facilitated by comparison to a cAMP calibration curve. Error bars represent ± S.E.M. for n = 3. Statistical analysis was performed using the one-way ANOVA test (Tukey’s multiple comparison test), with *p < 0.05 for comparisons with the control (w/o).

To assess the selectivity of GPM-3 for Gαi, we performed the same BRET assay with cells expressing β₂-AR (which is Gαs specific) and stimulated the cells with Iso, in a manner similar to the membrane assays. The results indicated the subfamily specificity of GPM-2/-3 (Figure S7b, c). The cellular activity of the bicyclic peptides indicates that they are permeable to the cell membrane. Inspection of their structures revealed the presence of both aromatic hydrophobic and positively charged amino acids, which are the key features of CPPs.¹⁴ It should be noted that the modest activity of peptides 2 and GPM-2 could result from their lower cell permeability compared to GPM-3 and that further optimization of the bicyclic peptides, e.g., conjugation to a highly active CPP, may further enhance the cellular activity of these peptides. None of the bicyclic peptides showed significant cytotoxicity up to 80 μM concentration (by the MTT assay⁷). GPM-3 showed excellent stability in human plasma, with no significant degradation after 180 min (Figure S8).⁵⁰ These attributes led us to select GPM-3 as a lead compound for further investigation.

GPM-3 likely prevents heterotrimer reassociation.

To assess the role of GPM-3 as a GAP for Gαi, we performed additional BRET assays and monitored the levels of free GRK3 (C-terminus of G protein-coupled receptor kinase-3, also GRK3ct) and the release of the Gβγ dimer (Figure S9a), or the levels of free monomeric Gαi subunit (Figure S9b). GRK3ct associates with free Gβγ and therefore acts as a G protein dissociation indicator⁵¹. This is a Gβγ activation assay, where the cells bearing the Rluc-fused GRK3ct or Rluc-Gαi, upon incubation with the respective ligand, should lead to heterotrimer dissociation and subsequently release of the Gβγ dimer (and interaction with the GRK3) or the Gαi subunit inducing eventually, the BRET signal. The data indicate that GPM-3 consistently (albeit only slightly) reduced the levels of both free Gβγ dimer (Figure 3d,e) and monomeric Gαi (Figure S9c,d), suggesting that GPM-3 may function by preventing the reassociation of the heterotrimer. Also, release of Gβγ dimer could also support the activation of the Akt/ERK pathway⁵². Immunoblots of HeLa cell lysates incubated with peptides 1, 2, GPM-2 and GPM-3 for phosphorylated (p-) and total (t-) Akt and ERK showed a slight effect of GPM-2 and no effect of GPM-3 on the Akt/ERK pathway (Figure S10), enhancing our suggestion of heterotrimer reformation in the latter case.

Quantitation of the GAP activity of GPM-3.

The ability of GPM-3 to induce GTP hydrolysis was quantitatively assessed with different assays. First, we determined the endpoint intracellular cAMP levels with a TR-FRET (time-resolved fluorescence resonance energy transfer) assay⁵³. HeLa cells were incubated with different concentrations of GPM-3 and cAMP production was stimulated with Fsk (or without). The intracellular cAMP levels were measured with the LANCE Ultra cAMP kit (Figure 3f, compared to a cAMP standard curve). At lower concentrations (<2.5 μM), GPM-3 stimulated the cAMP production, reaching a maximum of ~120% at 2.5 μM (Fsk-stimulated); still higher concentrations of GPM-3, however, resulted in dose-dependent reduction of cAMP level, with an IC₅₀ value of 6.92 μM.

Identification of novel binding sites by in silico analyses.

No structural information is currently available for the binding of macrocyclic peptidyl ligands to the Gαi protein^26–28. We therefore turned to computational modeling to gain some insight into the binding of peptides 1, 2, and GPM-2 to Gαi·GDP by using a previously established homology model^2,7 and docking GPM-3 to the existing structure of Gαi1·GTPγS (PDB: 1GIA⁵⁴, Figures S11–17, Tables S6–7). Based on the results of a 100-ns molecular dynamics (MD) simulation, we observed that GPM-3 and GPM-2 form stable complexes with Gαi via distinct stabilizing interactions (Figure 4a–c, d–f, respectively). Conversely, peptide 1 dissociates from its binding site over the course of the simulation with a conformational change of 1.64 Å (Figure 4g–i), while peptide 2 formed an unstable complex with the inactive Gαi protein (Supporting Information). GPM-3 binds between the αA (α-helical domain, α-HD) and switch III region (SWIII, GTPase domain) and thus interacts with both domains of Gαi1·GTPγS (Figure S14). This binding site has not previously been occupied by any of the other peptidic GEM-modulators^7,32. GPM-3 persistently interacts with the α-HD through hydrophobic interactions between the aromatic side chain of Nal1 and residue Arg86 in αA (Figure 4a–c). Additional interactions are mediated by hydrogen bonds between the following pairs of residues: Arg7-Glu238, Phe8-Glu238, Dap10-Ala235 and Dap10-Glu236, and Lys11-Asp237, which occur at frequencies >80% (Figure 4b–c, Video S2). The critical role of Lys11 explains why immobilization of GPM-3 to the sensorchip prevented Gαi binding in our SPR experiments. Residues Ala235, Glu236, and Asp237 are closely associated with the GTP hydrolysis mechanism^54–57 and interaction with the GAP protein RGS4^7,55. The observation that GPM-3 interacts with these three residues is consistent with GPM-3 acting as a GAP. Three critical residues, Arg_cat178, Thr181, and Gln_cat204, are known to be key players during GTP hydrolysis^56,58,59. Herein, Arg_cat178 forms a salt bridge with Glu43 (P-loop) and acts as a “seatbelt”, stabilizing the nucleotide at the binding site²⁹. This bridge persists throughout the simulation trajectory indicating that the transition state is indeed conserved (Figure S16, Video S3).

Figure 4:

Overview-illustration of protein-ligand complexes formed between bicyclic peptidic modulators and Gαi1·GTPγS/GDP. In (a, d, g), the binding sites of GPM-3 (cyan), GPM-2 (green) and peptide 1 (negative control, red) are shown on the Van-der-Walls surface (grey) of the respective Gαi1 structure. Structures (b, e, h), illustrate the interactions between the side chains of the interacting residues on the receptor with the respective sides of the bicyclic ligands. The respective bound-nucleotide (magenta) and the interacting residues (color-coding as shown above) are labeled. Hydrogen bonding and hydrophobic interactions are represented by black and green dotted lines, respectively. At the same time, switch regions are highlighted in yellow and α3-helix is displayed in orange. In (c, f, i), the chemical structure of each bicyclic peptide as well as their atom-interactions (oxygen in red, nitrogen in blue) with surrounding protein residues at the binding sites are depicted in details. PDB IDs: Gαi1·GTPγS (1GIA⁵⁴), homology model of Gαi1·GDP⁷ (PDB IDs: 1Y3A²⁹, 5JS8⁶⁰ and 3UMS⁶¹).

From another perspective, the heterotrimer formation was examined via simulations based on the Gβγ behavior. Commonly, the “attraction” between the SWII and SWIII regions is enhanced by the interactions of Glu236 and Asp237 (SWIII) with Arg205 (SWII)⁵⁷. Surprisingly, GPM-3 potentiates the interactions between Arg205-Glu236 and Arg205-Asp237 by stabilizing the orientations of the side chains of Glu236 and Asp237 converging them in SWII (Figure S17, Video S2). Another interesting property of GPM-3 is the enfeebling of the Gly203-Arg208 interaction within the G-R-E motif (Gly203-Arg208-Glu245)⁵⁷. The formation of the G-R-E triad usually favors the dissociation of Gβγ from the heterotrimer by reducing the strength of the polar network formed between Lys209/210 (SWII) and Gβγ dimer. The weakening of the interaction of Gly203-Arg208 leads to triad perturbation and subsequently suppression of the Gβγ release (Figure S17). This triad state may be an indication of the protein transition from the active to the inactive state. The persistence of the salt bridge (stable transition state) in the simulation trajectory and the prevention of triad formation (heterotrimer reassociation) strengthen our assertion of GPM-3 as a GAP for the Gαi protein. The above results provide unique information on the computationally assisted structural depiction of the bicyclic peptide interaction with the active state of the Gαi protein. Unlike GPM-3, GPM-2 binds between both domains but is “localized” in front of the nucleotide binding pocket, another undescribed binding site (Figure 4d, S14). The peptide participates in multiple hydrogen bonds (50%) through Lys6, as well as in stable hydrophobic interactions (>80%) via both Lys6 and Nal1 at this binding site. Since Lys6 contributes to both types of interactions, it appears to be the essential residue for complex stability. In particular, this residue interacts through its Cβ atom with the Cα of Gly42, which is entangled in the catalytic motif in the P-loop (Figure 4e–f), whereas Nal1 forms a hydrophobic bond with Arg205 and stabilizes the complex. Additionally, it was observed that different residues within the peptide interact with Glu43 leading to interruption of the salt bridge Arg178-Glu43. Thus, the peptide may enable GDP exchange by partially interfering with the salt bridge formation²⁹. To evaluate the topological behavior of GPM-2 for the nucleotide exchange, we also conducted a 100 ns MD simulation of the Gαi-”empty pocket” state. The results denoted that the peptide slowly lost its interactions with the surrounding protein residues and dissociated from the binding site already after 10 ns and continued over the entire simulation trajectory (Video S4). Here, the peptide did not block the nucleotide binding site. These data corroborate the binding and cellular studies, where GPM-2 binds to the protein and moderately activates the Gαi signaling through nucleotide exchange. Moreover, the instability and dynamic behavior of peptides 1 and 2 in their respective complexes during simulation proved the findings of the experimental studies (detailed description provided in the Supporting Information).

Overall, based on our experimental and computational observations, we hypothesize that the new binding sites may be suitable regions for the development of peptidic modulators with novel biological activity.

CONCLUSION

The present study provides a novel series of moderately potent, selective, cell-permeable, and metabolically stable bicyclic peptidyl Gαi modulators. One of the peptides, GPM-3, is biologically active in cellular assays and has an AC stimulating effect on the active state of the Gαi protein but no significant effect on the Gαs subunit. On the other hand, peptides 2 and GPM-2 target the Gαi∙GDP state and exhibit a partial AC inhibitory effect, with the latter also exhibiting the ability to promote GDP exchange. Computational studies provided key insights into the binding of the bicyclic peptides to the Gαi protein and structural basis for GPM-3 acting as a GAP for Gαi. We anticipate that GPM-3 may be further improved through structural optimization, e.g., by conjugation with a CPP to improve its cell penetration properties.

EXPERIMENTAL SECTION

One-bead-two-compounds (OBTC) library screening.

Screening experiments with the Gαi protein were performed using a bicyclic peptide library^9,18 according to Lian et al.¹⁸ and Qian et al.³⁶. Therefore, the bicyclic peptide library was subjected to two rounds of screening against the biotinylated Gαi protein. During the first round 180 mg (approach with Btn-Gαi·GDP: 272 mg) of the bicyclic library was incubated with biotinylated 0.5 μM Gαi [inactive GDP bound or the active state of GMPPNP-bound (guanosine-5’-[(β,γ)-imido]triphosphate)] and streptavidin-coated magnetic particles. The resulting magnetic beads were isolated from the library by magnetic sorting, during which the positive beads were attracted to the wall while the negative beads settled at the bottom of the container. The positive beads were washed, incubated again with the biotinylated G protein (0.2–0.5 μM), and subjected to a second round of screening. This screening approach is based on an on-bead enzyme-linked assay and a streptavidin-alkaline phosphatase (SA-AP) conjugate in which binding of the G protein to a bead recruited SA-AP to the bead surface. Upon the addition of BCIP (5-bromo-4-chloro-3-indolyl phosphate), SA-AP produced a blue-colored precipitate on the protein-bound bead. The intensely colored beads were manually isolated with a micropipette and subjected to PED-MALDI-MS analysis for sequencing the resin-bound peptides according to Sweeney et al.⁴⁶ and Thakkar et al.⁴⁵. The identified hit sequences are listed in Table S2–3 along with the establishment of the consensus sequence.

Peptide synthesis and purification.

Solid-phase peptide synthesis according to Lian et al.²³ was performed on a Rink Amide MBHA resin (0.53 mmol·g⁻¹) with HATU (4 eq.) as coupling reagent and NMM (8 eq.) as base. For the coupling of phenylglycine, COMU/TMP was used as described previously⁶². In order to reduce the resin loading to 0.25 – 0.33 mmol·g⁻¹, Fmoc-Lys(Boc)-OH (0.6 eq.) was coupled first followed by acetylation (1 eq. acetic anhydride, 2 eq. N-methylimidazole in DMF, 30 min) to block the remaining free amino groups. Subsequently, Fmoc-Dap(Aloc)-OH and the respective amino acid sequence (Fmoc-protected amino acids) as well as trimesic acid were coupled as reported earlier²³. The Aloc group was cleaved by Pd(PPh₃)₄ (0.5 eq.) and phenylsilane (10 eq.) in DCM and then, cyclization with PyBOP (10 eq.) and HOBt (10 eq.) was performed on the resin. The cleavage of the side-chain protecting groups together with the respective peptides from the resin was performed with reagent K as described before²¹. The crude peptides were purified by semi-preparative reversed-phase HPLC using a Shimadzu LC-8A instrument and the subsequent purity of the peptides (>95%) was confirmed by analytical RP-HPLC from a Shimadzu LC-20AD system. The collected fractions were combined, freeze-dried, and stored at −20 °C. Detailed information on the purification and analysis of individual peptides can be found in Table S4.

Live cell biological activity assay.

HeLa cells from American Type Culture Collection (ATCC) were cultured and maintained accordingly⁴⁸ in DMEM supplemented with 10% FBS. For the intracellular cAMP and Gβγ measurement assay, a BRET (bioluminescence resonance energy transfer) based experimental set up was used⁴⁹. The cAMP-based sensor was a generous gift from Dr. Tracy Handel and Dr. Irina Kufareva (University of California San Diego). On day one cells were plated (400K cells per well). On the second day, these cells were transfected with a total of 2000 ng of total DNA with Gαi (C351I)/β₂-AR (1000 ng) or pcDNA and modified-CAMYEL (1000 ng). For the Gβγ release measurements, the cells were transfected with Venus fused Gβ (500 ng), Gγ (500 ng) and Rluc8 fused Gαi (500 ng) plasmids⁵³. Transfection was performed according to the manufacturer’s protocol for TransIT^®-X2 transfection reagent (Mirius, USA). On day three, cells were washed and replated on a 96 well-plate (white, clear bottom) supplemented with DMEM with 10% FBS at a density of 40K cells per well. On day four, culture medium was removed and 70 μL assay buffer (0.027 mM glucose, 250 μL BSA and 50 mL PBS pH 8.0) was added. Peptide solutions (final concentrations of 10, 5 and 2.5 μM, diluted in assay buffer) were added (10 μL) to the cells and incubated for 30 min at RT. The luminescence scan was performed after the peptide treatment on the plate reader (Tecan, Spark M20, Switzerland). The initial cellular luminescence was measured for 7 min after the addition of 10 μL mixture of 100 μM IBMX (for cAMP detection) and 10 μM CTZ (in assay buffer). With regard to the CAMYEL release, the BRET was measured again for 15 min after the addition of 10 μL of 5 mM Fsk or 1 mM isoproterenol. The final BRET ratio was calculated from its excitation and emission scan. The bar graphs were generated from normalizing the area under the curve (AUC) values of the no peptide and peptide treated conditions. The assay was validated in triplicates and the final outcome was plotted. Detailed description of further BRET-based assays can be found in the Supporting Information.

In silico studies.

Each bicyclic peptide was generated manually on YASARA (Structure, Version 21.8.27⁶³ based on the spatial organization of the amino acid residues (Figure S6). The dynamic behavior and stability of the elucidated bicyclic peptides was monitored after 50 ns MD simulations. The derived peptide structures were then used for docking studies and binding site investigations. For these purposes, a homology model (HM) of Gαi1∙GDP, which was generated as previously⁷ from PDBs: 1Y3A²⁹, 3UMS⁶¹, 5JS8⁶⁰, was used as the target protein for the bicyclic peptides 1, 2 and GPM-2 (binders of the inactive protein state), whereas the Gαi1·GTPγS structure (PDB: 1GIA⁵⁴) was tested for GPM-3. The peptides were first docked to the hydrophobic cleft between the SWII/α3-helix (focused docking approach in YASARA). Due to no stable binding, further studies targeting potential alternative binding surfaces were conducted and additionally supported by SeeSAR (BioSolveIT GmbH, Version 11). Blind and ensemble docking studies were performed to determine the optimal peptide binding poses. The most suitable ones were determined, and the protein-ligand complexes were further subjected to 500 ps refinement simulations with the YAMBER⁶⁴ force field to increase the accuracy and quality of the selected complex. The resulting protein-ligand complex with the lowest energy and the highest structural quality was then subjected to 100 ns MD simulations to monitor the stability of the complexes formed and to investigate the effects of the peptides on the protein. A detailed description of the computational methodology is provided in the Supporting Information.

ABBREVIATIONS

AC

adenylyl cyclase

BRET

Bioluminescence Resonance Energy Transfer

cAMP

cyclic adenosine monophosphate

CPP

cell-penetrating peptide

GAP

GTPase-activating proteins

GDI

guanine-nucleotide dissociation inhibitor

GEF

guanine nucleotide exchange factor

GEM

guanine-nucleotide exchange modulator

GPCR

G protein-coupled receptor

Fsk

Forskolin

Iso

Isoproterenol

MD

Molecular Dynamics

SPR

surface plasmon resonance

SW

switch regions

TR-FRET

Time-Resolved Fluorescence Resonance Energy Transfer