Structure-guided engineering of snake toxins for selective modulation of adrenergic and muscarinic receptors

Zhong, Yixuan; Tao, Huihui; Zhang, Yu; He, Binbin; Jiao, Haizhan; Wang, Dandan; Teng, Maikun; Hu, Hongli; Guo, Qiong; Tao, Yuyong

doi:10.1038/s41467-025-61695-0

Download PDF

Article
Open access
Published: 14 July 2025

Structure-guided engineering of snake toxins for selective modulation of adrenergic and muscarinic receptors

Yixuan Zhong ORCID: orcid.org/0000-0001-7474-0536¹^na1,
Huihui Tao¹^na1,
Yu Zhang¹,
Binbin He¹,
Haizhan Jiao²,
Dandan Wang¹,
Maikun Teng¹,
Hongli Hu²,
Qiong Guo¹ &
…
Yuyong Tao ORCID: orcid.org/0000-0002-9257-1438¹

Nature Communications volume 16, Article number: 6478 (2025) Cite this article

7953 Accesses
2 Citations
4 Altmetric
Metrics details

Subjects

Abstract

Adrenergic receptors (ARs) and muscarinic acetylcholine receptors (mAChRs) are essential G protein-coupled receptors (GPCRs) that regulate a wide range of physiological processes. Despite their significance, developing subtype-selective modulators for these receptors has been a formidable challenge due to the high structural and sequence similarities within their subfamilies. In this study, we elucidated the recognition and regulatory mechanisms of ARs and mAChRs by muscarinic toxin 3 (MT3), a cross-reactive ligand derived from snake venom. By leveraging the distinct toxin-receptor interfaces, we engineer a panel of toxin variants capable of selectively modulating α2A and M₄AChR using computational design and directed evolution. These subtype-selective toxins not only provide valuable tools for basic research but also hold therapeutic potential for diseases associated with these GPCRs. This study further underscores the effectiveness of structure-guided approaches in transforming venom-derived scaffolds into receptor-specific modulators.

Molecular mechanism of human α_1A-adrenoceptor inhibition by Mamba snake toxin AdTx1

Article Open access 16 July 2025

The unconventional activation of the muscarinic acetylcholine receptor M4R by diverse ligands

Article Open access 23 May 2022

Structural and dynamic insights into supra-physiological activation and allosteric modulation of a muscarinic acetylcholine receptor

Article Open access 23 January 2023

Introduction

Adrenergic receptors (ARs) and muscarinic acetylcholine receptors (mAChRs) are two types of G protein-coupled receptors (GPCRs) that regulate a variety of physiological processes, including smooth muscle contraction, heart rate, and blood pressure^1,2,3. Of them, ARs are divided into three subfamilies, each containing three members (α1A/B/D, α2A/B/C, and β1/2/3)^4,5,6. Although all these receptors are activated by catecholamines such as norepinephrine and epinephrine, they exhibit different distributions and functional behaviors in the human body^7,8,9. For example, α2A receptors are predominantly located in presynaptic nerve terminals and human platelets, and blocking of α2A can reduce the inflammatory responses in sepsis and reverse pain and morphine analgesia^10,11. In contrast, α2B receptors are abundantly found in blood vessels, and their activation induces an increase in blood pressure, which is the opposite effect of α2A¹². Therefore, the development of subtype-selective drugs to modulate the cognate receptors involved in the physiological processes is necessary to achieve the desired clinical outcomes¹³.

Although α and β adrenergic receptors (αARs and βARs) have significant differences in their orthosteric sites, facilitating the design of selective drugs for αARs or βARs, obtaining subtype-selective drugs within each subfamily remains highly challenging^13,14. This is particularly true for the α2AR subfamily, where ligand-binding pockets exhibit great sequence and structural conservation^15,16. This phenomenon of receptors being highly similar and making it difficult to obtain receptor-specific ligands also occurs in the mAChR family, the receptor for the neurotransmitter acetylcholine. MAChRs consist of five members (M₁AChR-M₅AChR), each with distinct tissue distribution and functional roles^17,18. For instance, while M₂AChR is primarily found in the heart, regulating heart rate and contractility^19,20,21, its close relative M₄AChR is mainly located in the central nervous system (CNS) and regulates cognitive processes, making it a promising therapeutic target for movement disorders including Parkinson’s disease (PD) and dystonia^22,23,24. Despite the discovery of numerous orthosteric ligands for mAChRs, few exhibit true subtype selectivity^25,26,27. To address this issue, researchers have focused on exploiting the less conserved allosteric sites in mAChRs to develop allosteric ligands^28,29,30. Additionally, combining chemical groups that target both orthosteric and allosteric sites to create bitopic ligands has been explored to achieve direct and specific modulation of mAChRs^31,32,33.

Protein ligands typically have a larger buried surface area on target receptors than chemical compounds, enabling them to reach into less conserved regions on target receptors and achieve specificity³⁴. Muscarinic toxins (MTs) are a group of small proteins identified from mamba snake venoms that have been shown to modulate ARs and mAChRs³⁵. Interestingly, despite being highly similar in sequence and structure, MTs exhibit distinct specificity profiles and regulatory mechanisms with ARs and mAChRs^36,37. For example, MT7 specifically recognizes M₁AChR and modulates it in an allosteric manner^38,39. In contrast, MT3 exhibits a broader specificity profile, with nanomolar affinity for M₄AChR, α1A, α2A, and α2C, and a 10-fold lower affinity for α1B and α1D, acting as an antagonist^37,40,41. The mechanisms by which MTs achieve such different specificity profiles and exert distinct regulatory effects remain unclear. Elucidating these mechanisms not only enhances our understanding of GPCR regulation but also guide the development of ARs- and mAChRs-selective modulators, which is of great urgency in both basic and translational research.

In this work, we determine the structures of MT3 in complex with α1A, α2A and M₄AChR. Guided by the information of toxin-receptor interface, we successfully engineer a group of variants that are selective for α2A and M₄AChR using computational design and directed evolution. These subtype-selective modulators can serve as valuable tools and drug candidates in basic research and drug development.

Results

Structural basis of MT3 antagonism against α1A, α2A and M₄AChR

To elucidate the antagonistic regulatory mechanism of MT3 on target receptors and provide clues for engineering of receptor subtype-selective MT3 variants, we first set out to determine the structures of MT3 binding to each target receptor. Therefore, we purified the recombinant MT3 toxin and verified its binding to the target receptors (Supplementary Fig. 1a, b). Following this validation, we selected representative members, namely α1A, α2A, and M₄AChR, from the three target receptor subfamilies for structural characterization. The fusion and gluing strategy recently developed by our group was applied to these three receptors⁴². Briefly, receptors were engineered by inserting mBril between TM5 and TM6 to create continuous helices within the receptor. Additionally, a K3-ALFA helical tag was introduced at the C-terminus of H8 in the desired configuration (Supplementary Fig. 1c–e). To facilitate the preparation of the toxin-receptor complex, MT3 was fused directly to the N-terminus of the receptors (Supplementary Fig. 1c–e). The 1B3 Fab and glue molecules were then added during purification. The α1A, α2A, and M₄AChR were assembled with the “4-9,” “6-10,” and “6-12” glue molecules, respectively, where the hyphenated numbers indicate the length of the two linkers in the glue molecules (as defined in Supplementary Fig. 1f). These complexes exhibited high-quality particles in the 2D classification analysis (Supplementary Fig. 1g). Furthermore, a density on the extracellular side, which can be assigned as MT3, was observed in all these samples (Supplementary Fig. 1g), suggesting that MT3 should form a stable complex with the target receptors, allowing further cryo-EM studies. Finally, a cryo-EM map with a global resolution of 3.2 Å, 3.4 Å and 3.6 Å was obtained for the MT3-α1A, MT3-α2A and MT3-M₄AChR complex (Supplementary Figs. 2-4), respectively, and guided by these maps, the models of MT3-α1A, MT3-α2A and MT3-M₄AChR were ultimately built (Fig. 1a).

**Fig. 1: Structural basis of MT3 regulation on target receptors.**

As anticipated, MT3 bound to the extracellular regions of the three target receptors (Fig. 1a). Similar to MT7, MT3 utilized its three finger loops to interact with the target receptor (Fig. 1a). However, unlike the “lying” conformation with which MT7 engages M₁AChR, MT3 encounters the target receptors with a more “standing” conformation (Fig. 1b). Additionally, MT3 displays a counterclockwise rotation and deeper insertion into the receptor core compared to MT7 in the MT7-M₁AChR complex (Fig. 1b). Notably, the finger loop 2 of MT3 extends through the extracellular vestibule in all three receptors, with the tip residue R34, the residue that inserts deepest into the receptor core, reaching the orthostatic site (Fig. 1c). However, unlike chemical compounds that typically fully occupy the ligand-binding pocket, R34 of MT3 occupies only a fraction of the ligand-binding pocket (Fig. 1c). Nevertheless, MT3 engagement will still prevent the binding of the receptor to other ligands. This distinctive binding mode of MT3 elucidates its role as an antagonist rather than an allosteric modulator like MT7 and may also be responsible for the insurmountable antagonism observed in MTs⁴¹. In line with the antagonistic activity of MT3, all three target receptors adopt an inactive conformation, characterized by the close packing of TM6 with the intracellular receptor core (Fig. 1d). Superposition of the toxin receptor structures with those stabilized by chemical compounds reveals subtle conformational rearrangements in the extracellular region across all three receptors (Fig. 1d). Taking the M₄AChR as an example, its intracellular side exhibits no apparent conformational differences, but its ECL2 and TM7 are displaced outward by 2-3 Å to accommodate the toxin (Fig. 1d). Collectively, through the partial occupancy of the orthostatic pocket with finger loop 2, MT3 antagonistically modulates target receptors.

Distinct toxin-receptor interface

In contrast to the positioning of finger loop 1 in MT7, which extends along the exterior of TM4/5 of the M₁AChR, the counterpart in MT3 finds its location nestled between TM5 and TM6 in α1A, α2A, and M₄AChR (Fig. 1b and Supplementary Fig. 5a–c). This, coupled with the unified insertion of “finger loop 2 into the receptor core”, results in MT3 adopting similar interaction modes with the three target receptors (Supplementary Fig. 5a–c). Except for TM1 and TM4, all extracellular structural elements in these three receptors actively contribute to shaping the toxin-receptor interface (Fig. 2a–d). Notably, among them, finger loop 2 is tightly embraced by ECL2 and TM6/7 on both sides, playing a pivotal role in establishing toxin-receptor interactions (Fig. 2c). Taking the MT3-M₄AChR structure as an example, at the bottom layer, the finger loop 2 residue P33 is surrounded by a series of hydrophobic residues from M₄AChR: Y89^2.61, Y92^2.64, W108^3.28, L109^3.29, and Y439^7.39 (Fig. 2c). At the same time, the adjacent residue R34 also mediates hydrophobic contact with the receptor with the aliphatic portion of its side chain (Fig. 2c). In addition, R34 also forms a hydrogen bond with Y416^6.51 of M₄AChR (Fig. 2c). Directly above P33 and R34, MT3 residues V31, I32, and Y35 form an additional hydrophobic patch in conjunction with M₄AChR residues W435^7.35 and Y439^7.39 (Fig. 2c). On the opposing side, MT3 residues Y30 and T36 stack directly with ECL2 by forming an interaction network with ECL2 residues P181, Q184, and F186 (Fig. 2c). Clearly, these interactions are crucial for maintaining the optimal conformation of MT3, thereby ensuring it sitting above the receptor core. An on-yeast binding assay indicated both Y30 and Y35 mutations significantly impaired the receptor binding ability of MT3 (Fig. 2e).

**Fig. 2: Interaction details between MT3 and target receptors.**

Although a nearly identical set of residues of finger loop 2 was used for binding to the receptor, the corresponding amino acids on the receptor side show variability across the three receptors. A prominent discrepancy is observed in the residues that constitute the binding environment for MT3’s P33: these are S83^2.61, F86^2.64, W102^3.28, A103^3.29, and F312^7.39 in α1A, and S105^2.61, N108^2.64, Y124^3.28, L125^3.29, and F427^7.39 in α2A, contrasted with Y89^2.61, Y92^2.64, W108^3.28, L109^3.29, and Y439^7.39 in M₄AChR (Fig. 2c and Supplementary Fig. 5d). Another notable difference lies in the ECL2 residues that directly interact with MT3’s Y30 and T36: they are Q184^ECL2 and F186^ECL2 in M₄AChR, but I175^ECL2 and Q177^ECL2 in α1A, and R202^ECL2 and E204^ECL2 in α2A (Fig. 2c and Supplementary Fig. 5e). The ECL3 region in all three receptors represents the last region where both significant sequence and conformational divergence were observed (Supplementary Fig. 5f). Collectively, the distinct configurations of binding pockets and interacting residues in the three receptors give rise to varied conformational states of MT3 in the respective complexes (Supplementary Fig. 5a-c). For example, the trajectory of finger loop 2 in MT3 binding to α1A deviates significantly from the corresponding motifs in the other two complexes (Supplementary Fig. 5a-c).

While finger loop 2 plays a dominant role, the other two finger loops contribute differently to the toxin-receptor complexes. Firstly, finger loop 1, leveraging the tip residue I9 to anchor itself to the outer edges of TM5 and TM6, mediates a key hydrophobic contact with the receptors in all three complexes (Fig. 2b). This interaction is further supported by the loss of binding ability observed in the mutant during the on-yeast binding assay (Fig. 2e). In contrast, finger loop 3 exhibits the most significant conformational disparities among the three complexes (Supplementary Fig. 5a–c) and plays distinct roles in the interactions between the toxin and receptors (Fig. 2d). In particular, in the MT3-α2A complex, finger loop 3 establishes minimal contacts with α2A; however, in MT3-α1A and MT3-M₄AChR complex, its Y52 and S54 appear to pack against the ECL1 residue Y91^α1A or Y97^M4AChR and ECL2 residues P181^M4AChR, N183^M4AChR, and Q184^M4AChR with different strength (Fig. 2d). In summary, MT3 toxin engages its target receptors through a relatively conserved set of interaction residues, while the amino acids on the receptor side exhibit variability. This variability leads to the emergence of different conformational states and interactions across receptor complexes.

Computational redesigning MT3 for α2A selectivity

Driven by the binding differences between MT3 and target receptors, we next decided to engineer MT3 into α2A selective modulators. Inspired by the recent progress of protein design with deep learning network, we exploited the use of denoised diffusion method RFdiffusion, which has been shown to be effective in improving the interface between target and binding proteins^43,44. By applying partial diffusion mode to regions of both the toxin and receptor, we reasoned that it might be possible to generate structural states and sequence compositions that could not be sampled by experimentally displayed libraries, thereby generating novel MT3-based modulators. To implement the redesign, the four pairs of cystines were retained to maintain the folding of the toxin, and all residues between disulfide bonds in the three finger loops were selected for optimization (Supplementary Fig. 6a). After filtering the design results using AlphaFold2 prediction, the remaining sequences did show great diversity, with little homology to wild-type MT3 (Supplementary Fig. 6b). We synthesized 30 variants and screened them using yeast surface display. However, none of them showed significant α2A selective binding, and most of them even showed nonspecific binding to M₂AChR and M₄AChR, as reflected by the inability of orthosteric ligands to compete off yeast binding (Supplementary Fig. 6c). We reasoned that these redesigned toxin variants might be unable to fold on the yeast surface and that refolding may be required for the formation of a functional protein.

To minimize the impact of sequence changes on toxin folding, making it easier for subsequent experimental validation, we next focused solely on residues from finger loop 2 and loop 3 in close vicinity to α2A. We conducted two parallel redesigns: the first involved finger loop 2 residues H27-E37 and loop 3 residues Y52-S54, while the second involved the same set of residues but kept the finger loop 2 tip residues I32/P33/R34 constant to ensure that the resulting variants still function as antagonists (Fig. 3a), given the role of these three residues in occupying the orthostatic site. Following the same design and filtering procedures, these focused redesigns yielded significantly less diverse outputs compared to the three-loop redesign (Supplementary Fig. 6d). We then synthesized 20 representative candidates from each redesign and tested their binding to α2A through on-yeast binding assay. Ultimately, we obtained a binder named A91 from the first batch that showed α2A-specific binding; however, its binding affinity is relatively low (Fig. 3b). In contrast, three α2A selective binders were obtained from the second batch, with one named RD7. Purified recombinant RD7 exhibiting an affinity close to that of MT3 (Fig. 3b, c). These results suggest that while computational redesign can identify receptor subtype-selective binders, the sequence composition in these binders may still not be optimal. Most importantly, instead of converting all the residues in the interface, a refinement of a limited number of MT3 residues can likely lead to receptor subtype selectivity.

**Fig. 3: Binding of redesigned MT3 variants to α2A.**

Directed evolution of MT3 for improved affinity for α2A

Based on the sequences of the redesigned MT3 variants, finger loop 2 residues at positions 29, 30, 36, and 37, along with finger loop 3 residues at positions 52 and 54, frequently appeared as amino acids with different properties compared to those in wild-type MT3 (Supplementary Fig. 6b, d). This suggests that these positions play dominant roles in determining selectivity. Indeed, inspecting the environment of the four finger loop 2 residues reveals that they make direct contact with regions that are least conserved in ARs and mAChRs (Supplementary Fig. 7a-f). Therefore, a rigorous refinement of these residues can probably yield subtype-selective binders while maintaining high affinity. We thus constructed a directed yeast display library by randomizing these residues (Supplementary Fig. 7g). The library was precleared with α1A/M₄AChR protein and screened with α2A protein. After one round of magnetic-activated cell sorting (MACS) and three rounds of fluorescence-activated cell sorting (FACS) (Supplementary Figs. 7h, i and 13a), we identified several variants that showed strong α2A binding and selectivity. Among them, one variant named A2S5 contains five mutations: three in finger loop 2 (H29S, Y30H, T36V) and two in finger loop 3 (Y52G and S54R) (Fig. 4a-c). To further verify the specificity and affinity of A2S5, we prepared the recombinant A2S5 protein and tested its binding with the entire adrenergic and muscarinic receptor families. Compared to the computational redesigned variants, A2S5 indeed displays an improved affinity for α2A (Figs. 3c and 4c) while demonstrating exclusive selectivity for α2A, contrasting with MT3’s broad specificity profile (Fig. 4b). Notably, A2S5 retained an affinity for α2A comparable to that of MT3 (Fig. 4c). Accordingly, mutation of the α2A residues that interact with A2S5 crippled the α2A binding activity of A2S5 (Supplementary Fig. 7j, k). Additionally, consistent with the fact that the residues at the finger loop 2 tip region of A2S5 are identical to those in MT3, A2S5 also modulates α2A in an antagonistic manner, with an IC₅₀ of ~5 nM, which is twice as potent as MT3 (Fig. 4d).

**Fig. 4: Characterization of the α2A selective modulators.**

In addition to this finger loop 2/3 variant, we surprisingly identified several variants with mutations solely in finger loop 3 that exhibited exclusive α2A binding in yeast binding (Fig. 4a, e). These variants were probably incorporated during the construction of the library, in which wild-type MT3 was used as a template to amplify the finger loop 2 and loop 3 fragments (Supplementary Fig. 7g). Notably, all these finger loop 3 variants presented the consensus Y52G mutation (Fig. 4a), identical to that observed in computational- and library-yield variants A91, A64, A66 and A2S5 (Figs. 3b and 4a). We prepared the variant A2-3 that shares the same finger loop 3 as A2S5 and validated again its affinity and selectivity for α2A. Interestingly, A2-3 retained selectivity for α2A (Fig. 4f), although it exhibited moderate residual binding to α1A at saturating concentrations on yeast binding assay (Fig. 4f). However, the affinity and regulatory activity of purified A2-3 were slightly weaker compared to MT3 (Fig. 4g, h). Based on the structural model, this G52 appears to mediate minimal contact with α2A (Supplementary Fig. 7l), so the α2A selectivity of the Y52G substitution may be conferred by other factors, such as reshaping the configuration of the finger loops, particularly finger loop 2, in a way that only fits the receptor core of α2A. Nevertheless, this unexpected finding prompted us to explore whether finger loop 2-only mutations could also yield α2A selectivity. We therefore converted the finger loop 3 of A2S5 back to that of MT3 and checked the binding ability of this A2S5-MT3 chimera with ARs and mAChRs. Again, this A2S5-MT3 demonstrates α2A selectivity in yeast binding (Fig. 4f), while the purified toxin exhibits affinity and regulatory activity comparable to that of MT3 (Fig. 4g, h). These results demonstrate that fine-tuning the sequence of individual finger loop 2 and 3, probably through different mechanisms, is sufficient to generate subtype-selective binders. Furthermore, in competitive ligand binding assays, A2S5, A2-3 and A2S5-MT3 effectively displaced ligand binding, demonstrating direct competition for the same orthosteric site (Supplementary Fig. 12a). In sum, through the directed evolution of MT3, we obtained a series of variants that can selectively modulate α2A in an antagonistic manner.

Engineering of M₄AChR selective modulators

After successfully obtaining the α2A-specific antagonists, we next aimed to develop M₄AChR-selective modulators. Given the independent and determining role of finger loop 2 in mediating the selective recognition of α2A by MT3-based variants, we focused on the four finger loop 2 residues (29/30/36/37) and constructed a library using the hard randomization mode to identify M₄AChR-specific binders (Supplementary Fig. 8a). Using a selection procedure similar to that for α2A (Supplementary Figs. 8b, c and 13b), we eventually identified three variants that demonstrated strong and selective binding to M₄AChR (Fig. 5a–d and Supplementary Figs. 8d-f, 12b). All these variants retained the original histidine at position 29 and had an arginine substitution at position 37 (Fig. 5a). Among them, M4S3 exhibited the highest affinity for M₄AChR, with an EC₅₀ comparable to that of MT3 (Fig. 5c). Additionally, M4S3 also demonstrated antagonist activity against M₄AChR, with an IC₅₀ of 14.4 nM (Fig. 5d). Based on these finger loop 2-only variants, we next attempted to extend the randomization region to loop 1 and loop 3 to probe how they impact the receptor specificity and whether they can further improve the affinity for the receptor. To this end, we constructed two additional libraries, targeting residues 7–9 in finger loop 1 and 52-54 in finger loop 3, respectively (Supplementary Fig. 8g, h). After selections, we identified two M₄AChR-selective variants, M129 and M120 through on-yeast binding assay (Supplementary Fig. 8i), which exhibited different residue compositions in finger loop 1 (Fig. 5e). However, compared to MT3, neither purified M120 nor M129 showed a significant improvement in affinity (Fig. 5e). Surprisingly, for the finger loop 3 library, after two rounds of selections, all sequenced variants reverted to the wild-type sequence—a phenomenon completely different from that observed in the α2A selections. This suggests that the naturally evolved finger loop 3 may already be optimally configured for M₄AChR binding.

Fig. 5: Characterization of the M4AChR selective modulators. — **Fig. 5: Characterization of the M₄AChR selective modulators.**

Mapping the substituted residues onto the structure revealed that while it is straightforward to understand how M4S3 and the related variants can discriminate M₄AChR from ARs (Supplementary Fig. 9a, b), the mechanism by which they differentiate M₄AChR from other mAChRs remains elusive. Particularly, M₂AChR shares a 79% sequence identity with M₄AChR in the toxin-engaging region but cannot bind MT3 and M4S3 (Supplementary Fig. 9c). To investigate this, we created a panel of M₄AChR-M₂AChR swapped chimeras and tested their binding with the purified toxins (Fig. 5f). Replacement of the ECL1 or ECL3 of M₄AChR with that of M₂AChR did not significantly impair their association with MT3 or M4S3 (Fig. 5f). Consistently, grafting these M₄AChR regions into M₂AChR also did not confer toxin-binding ability (Fig. 5f). In contrast, swapping ECL2, specifically the four residues PDNQ (EDGE in M₂AChR), led to a different result: M₄AChR-M₂AChR^EDGE lost MT3 binding and M₂AChR-M₄AChR^PDNQ gained MT3 binding (Fig. 5f). Interestingly, different from the observation with MT3, M4S3 can still bind M₄AChR^ECL2 and M₄AChR-M₂AChR^EDGE (Fig. 5f), suggesting the three converted residues in M4S3 probably render M4S3 a different binding entropy or enthalpy for M₄AChR. Given these four residues (PDNQ in M₄AChR) control receptor specificity, we screened a directed library containing focused randomizations of the MT3 residues facing these four residues (Supplementary Fig. 9d, e); however, we did not obtain any high-affinity binders (Supplementary Figs. 9f and 13c). As these four residues (PDNQ in M₄AChR) make minimal contact with the toxin (Supplementary Fig. 9d), they probably contribute to establishing different conformational states in the two receptors, and that in M₂AChR probably prevents the binding of MT3. Together, through directed evolution, we obtained a series of M₄AChR-selective modulators, which are of great value for M₄AChR-related basic research and drug development.

Discussion

During the long-term evolution driven by natural selection, venomous animals have accumulated thousands of toxins for predation or defense. Due to their regulatory function on ion channels, GPCR receptors, and other targets, venom toxins represent a vast source of agents that can be developed into therapeutics^45,46. For example, exenatide, derived from the Gila monster, and ziconotide, derived from Conus magus, are used for the treatment of type 2 diabetes and chronic pain, respectively^47,48. MTs, a family of small protein toxins, have been shown to modulate adrenergic and muscarinic acetylcholine family receptors with different specificities and mechanisms. Most importantly, adopting a scaffold belonging to the family of three-finger-fold (3FT), MTs can potentially be engineered into variants possessing subtype selectivity, as reflected by the naturally existing MT3 and MT7 being able to target different receptors. While redirecting the allosteric regulation of MT7 from M₁AChR to M₂AChR has been recently realized through structure-guided engineering, the engineering of MTs to fulfill orthosteric and selective modulation of target receptors has not been explored⁴⁹. Here, we took MT3, which has the broadest receptor specificity, as the object and determined its complex structure with α1A, α2A, and M₄AChR. Structural characterization revealed that despite adopting analogous β-strand conformations in MT3’s finger loop 2 and the CDR3 domain of nanobody-derived VHHs, their receptor engagement mechanisms diverge fundamentally in spatial orientation within the orthosteric binding pocket. Structural superposition (PDB ID: 8Z74, 8XQJ, 9EAH for VHH-GPCR, this study for MT3-GPCR) demonstrate that MT3 stabilizes receptor interactions through a rigid β-strand architecture, in contrast to the conformationally flexible CDR3 loops of VHHs, which exhibit hypervariable sequence plasticity (Supplementary Fig. 10). Driven by different toxin-receptor interfaces, we conducted a computational redesign of MT3 to obtain subtype-selective modulators. However, despite the redesign of MT3 for α2A outputting sequences possessing subtype selectivity, it did not generate binders with high affinity. The reason may be the limited number of residues for redesign and the all-beta-strand structure of the target, which is still challenging using the available computational design methods. Nevertheless, the output sequences provide other information, the hotspot residues worthy of optimization for desired selectivity. Based on this information, a directed library-based screening eventually led to both α2A and M₄AChR selective modulators.

In this study, we only included α2A and M₄AChR to engineer their selective ligands. Moving forward, the same strategy could be applied to other MT3-binding receptors, such as α1A, α1B, α1D, and α2C, to develop subtype-selective ligands. However, caution is needed when applying this approach to other related receptors, as demonstrated by the unsuccessful redirection of MT3 for M₂AChR in this study. This failure may not be due to the strategy itself but rather to the nature of the target receptors. GPCRs often have great flexibility, and if the extracellular core resides in a closed conformation unsuitable for protein ligand engagement, the ligand will not bind, regardless of how well it matches the pocket. Nevertheless, leveraging protein design and engineering techniques to develop subtype-selective modulators, as exemplified by α2A and M₄AChR in this study, represents a promising direction for drug discovery.

Methods

Expression, purification and biotin labeling of MT3 and engineering variants

The coding sequence of muscarinic toxin 3 (MT3) was cloned into the pME vector (derived from AbVec2.0-IGHG1) with an N-terminal human trypsinogen-2 (TRYP2) signal peptide, maltose-binding protein (MBP), and a C-terminal 6xHis tag. Recombinant plasmids were transfected into HEK293F cells using PEI (Yeason, CAS 49553-93-7). After 96 h, the media were harvested, supplemented with 1/10 supernatant volume of 10×PBS and 2 mM CaCl₂, and incubated at 4 °C for 1 h. The precipitate was removed by centrifugation, and the supernatant was incubated with Ni-NTA resin for 2 h at 4 °C. After incubation, the resin was washed with 10 column volumes of PBS, followed by 10 column volumes of PBS with 10 mM imidazole, then MT3 was eluted with PBS containing 250 mM imidazole. The eluted protein was further purified by size-exclusion chromatography (SEC) using a Superdex 200 Increase 10/300 GL column (Cytiva) pre-equilibrated with PBS. Fractions corresponding to monomeric MBP-MT3-His were pooled and concentrated using a 10-kDa MWCO centrifugal filter (Milipore). Purified protein purity (>95%) was confirmed by SDS-PAGE, and protein concentration was determined by UV absorbance at 280 nM (NanoDrop OneC, Thermo Fisher).

For biotin labeled MBP-MT3-His, purified MBP-MT3-His was diluted to 5 mg mL^-1 and incubated with EZ-link™ NHS-Biotin (Thermo Fisher, CAS: 20217) at 4 °C overnight. The labeled protein was then subjected to size-exclusion chromatography (SEC) to remove excess biotin. Biotinylation efficiency was assessed by incubating labeled proteins with a 5-fold molar excess of streptavidin (25 °C, 30 min), followed by SDS-PAGE analysis under reducing conditions.

All engineered variants were subjected to identical purification and biotinylation protocols to ensure methodological consistency across functional assays.

Expression and purification of the 1B3 Fab and glue molecules

The glue molecule used in this study is a fusion protein composed of the E3 helix (PDB 1UOI), a Fab nanobody (PDB ID: 6AND), and an ALFA helix nanobody (PDB 6I2G). Three variants of the glue molecule, “4-9,” “6-10,” and “6-12,” were utilized, where the numbers indicate the amino acid count in the flexible linkers connecting the components: the first number represents the linker between the E3 helix and Fab nanobody, and the second denotes the linker between the Fab nanobody and ALFA helix nanobody. The 1B3 Fab (PDB ID: 8J7E) was used as the binding Fab. Both the glue molecules and the 1B3 Fab were developed in our recent research⁴².

For protein purification, the coding sequences of Fabs or glue molecules were cloned into the pET-22b (+) vector with an N-terminal pelB signal peptide and a C-terminal 6xHis tag. Plasmids were transformed into E. coli BL21 (DE3) cells. Cells were grown to OD₆₀₀ = 0.8 at 37 °C in LB media supplemented with 1‰ ampicillin. Cells were induced by the addition of 1 mM IPTG and incubated for 24 h at 16 °C. Cells were harvested and disrupted by sonication in a buffer containing 20 mM HEPES pH 7.5 and 150 mM NaCl. Both 1B3 Fab and glue molecules were purified by Ni-NTA chromatography. First, the unwanted proteins were removed with wash buffer (20 mM HEPES pH 7.5, 150 mM NaCl, 20 mM imidazole), and then the target protein was eluted with wash buffer supplemented with 300 mM imidazole. The elution was concentrated to 20 mg mL^-1 using a 10-kDa molecular weight cutoff concentrator (Millipore) for the assembly of complexes.

Receptor construct design

For structural studies, the coding sequence of human α1A, α2A and M₄AChR were cloned into pME vector. To enhance the stability of the complex, the MT3 sequence was added to the N-terminus of each receptor, allowing co-expression of MT3 with the receptors. MT3 and each receptor were connected by a 15 glycine-serine (GS) linker. For further modifications, the intracellular loop 3 (ICL3) residues 215-261 of α1A, residues 242-378 of α2A, and residues 220-393 of M₄AChR were deleted and replaced by mBRIL, respectively. Additionally, the C-terminal residues 342-466 of α1A, residues 457-465 of α2A, and residues 470-479 of M₄AChR were deleted and replaced by a K3-ALFA tag, respectively.

For library screening, human α1A, α2A and M₄AChR constructs retained mBRIL insertions and C-terminal truncations but excluded MT3 fusion.

All constructs featured an N-terminal HA signal peptide followed by a FLAG tag and a C-terminal 6×His tag to standardize expression, purification, and detection across experiments.

For functional assays, wild-type human α1-, α2- and mAChRs construct were subcloned into pME vector with N-terminal HA signal peptide and cMyc tag and utilized for functional characterization.

Structural complex formation and purification

Recombinant plasmids were transfected into HEK293F cells using PEI (Yeason, CAS 49553-93-7). After 60 h, the cells were spun down, and the cell pellets were stored at −80 °C until use. For purification of the fusion proteins, thawed cell pellets were resuspended in lysis buffer (10 mM HEPES pH 7.5, 1 mM PMSF, 0.5 mM EDTA) and centrifuged at 4 °C. Cell membranes were collected and homogenized in solubilization buffer containing 20 mM HEPES pH 7.5, 500 mM NaCl, 10% glycerol, 1% (wt/vol) n-dodecyl-B-D-maltoside (DDM, Anatrace, CAS 69227-93-6), and 0.1% (wt/vol) cholesteryl hemisuccinate tris salt (CHS, Anatrace CAS 102601-49-0) and then spun at 4 °C for 2 h. The supernatant was isolated by centrifugation at 8000 g for 60 min and incubated with Ni-NTA chromatography at 4 °C for 2 h. The resin was collected in a column, washed with wash buffer containing 20 mM HEPES pH 7.5, 150 mM NaCl, 10% glycerol, 0.03% (wt/vol) DDM, and 0.02% (wt/vol) CHS, and eluted in wash buffer supplemented with 250 mM imidazole. Then, 2 mM CaCl₂, excess purified 1B3 Fab and glue molecules were added to the elution and incubated with the Anti-FLAG M1 affinity resin (M1 resin, Sigma, A4596) at 4 °C for 60 min. M1 resin was then collected and washed with wash buffer containing 20 mM HEPES pH 7.5, 150 mM NaCl, 0.03% DDM, 0.01% CHS, and 2 mM CaCl₂. Buffer exchange from DDM buffer to lauryl maltose neopentyl glycol (LMNG, Anatrace, CAS 1257852-96-2) buffer was performed in a stepwise manner. After buffer exchange, the M1 resin was further washed with LMNG buffer containing 20 mM HEPES pH 7.5, 150 mM NaCl, 0.01% (wt/vol) LMNG, 0.001% (wt/vol) CHS, and 2 mM CaCl₂. The complex was eluted with elution buffer containing 20 mM HEPES pH 7.5, 150 mM NaCl, 0.00075% (wt/vol) LMNG, 0.00025% (wt/vol) glycol-diosgenin (GDN, Anatrace CAS 1402423-29-3), 0.0001% CHS, 5 mM EDTA and 200 µg mL^-1 synthesized Flag peptide. The complex was further purified by size-exclusion chromatography using a Superdex 200 Increase 10/300 column (Cytiva, 28-9909-44) preequilibrated with buffer containing 20 mM HEPES pH 7.5, 150 mM NaCl, 0.00075% (wt/vol) LMNG, 0.00025% (wt/vol) GDN, and 0.0001% (wt/vol) CHS. The complex fractions were collected and concentrated to 10 mg mL^-1 using a 100-kDa molecular weight cutoff concentrator (Millipore) for electron microscopy experiments.

Sample preparation and data collection for cryo-EM

To prepare the cryo-EM sample, 3.5 μL purified protein was applied onto the amorphous alloy film grid (CryoMatrix nickel titanium alloy film, R1.2/1.3, Zhenjiang Lehua Electronic Technology Co., Ltd.), which was glow-discharged for 60 s in a Tergeo-EM plasma cleaner. At 8 °C and 95% humidity, the sample was left to stand for 4 s and blotted for 4 s. Grids were plunge-frozen in liquid ethane cooled by liquid nitrogen using Vitrobot Mark IV (Thermo Fisher Scientific) and then stored in liquid nitrogen until checked. Movies were collected on a 300 kV Titan Krios Gi3 microscope (Thermo Fisher Scientific FEI, the Center for Integrative Imaging, Hefei National Laboratory for Physical Sciences at the Microscale, University of Science and Technology of China). operated at 300 kV with a GIF energy filter (Gatan) at a magnification of 81000× (corresponding to a calibrated sampling of 0.535 Å per pixel). The target defocus range was set to −1.3 to −2.2 μm.

Cryo-EM data processing

For the MT3-α1A, MT3-α2A, and MT3-M₄AChR complexes, 5826; 4746; and 6120 movies were collected, respectively. Motion correction and contrast transfer function (CTF) estimation were performed using the patch-based algorithms in cryoSPARC⁵⁰. Initial particle picking and extraction at bin2 resolution yielded 6,296,155; 5,193,378; and 8,362,657 particles for the respective complexes. These particles underwent 2D classification, heterogeneous refinement, and non-uniform (NU) refinement. The resulting maps were used for an additional round of 2D classification, from which high-quality particles were selected for Topaz training⁵¹ to enhance particle picking. After combining all particles and removing duplicates, heterogeneous and non-uniform (NU) refinements were repeated to select the best-resolved class. Selected particles were re-extracted at full pixel size and subjected to global CTF refinement, local CTF refinement and another round of non-uniform (NU) refinement. The final datasets comprised 866,595 (MT3-α1A), 458,495 (MT3-α2A), and 490,478 (MT3-M₄AChR) particles. These were used for final NU and local refinements in cryoSPARC, yielding density maps at nominal resolutions of 3.19 Å, 3.41 Å, and 3.63 Å, respectively, based on the FSC using the 0.143 criterion.

Model building and refinement

The predicted structures of MT3-α1A, MT3-α2A and MT3-M₄AChR were used as initial models for model rebuilding and refinement against the electron microscopy map. All the models were first docked into the EM density map using Chimera⁵², followed by iterative manual adjustment and rebuilding in COOT⁵³ and phenix.real_space_refine in Phenix⁵⁴. The final model statistics were validated using MolProbity. Model refinement statistics are summarized in Supplementary Table 1. The molecular graphic figures were prepared with PyMOL.

Partial diffusion to optimize MT3

To optimize MT3 for selective binding to α2A, we utilized RFdiffusion⁴³ to diversify the sequence of MT3. The complex structure of MT3-α2A served as the input for partial diffusion, with the entire toxin and α2A residues within 5 Å of MT3 set as the noising region. All other sequences in the toxin and α2A remained unchanged. We applied 10 and 20 noising timesteps out of a total of 50 timesteps in the noising schedule, followed by denoising using “diffuser.partial_T” input values of 10 and 20. Approximately one thousand partially diffused designs were generated. This library of backbones was subsequently sequence-designed using ProteinMPNN⁵⁵ and then modeled with AlphaFold2⁵⁶. The resulting designs were filtered based on AlphaFold2 criteria, including an ipTM > 0.7, pLDDT > 80, and Cα RMSD < 2 Å.

Construction and selection of α2A affinity maturation library

Library construction: The coding sequence of MT3 was used as the template for constructing the α2A affinity maturation library. Forward and reverse primers (GenScript) containing degenerate codons and flanking regions were designed according to the desired library design and were annealed to the MT3 sequence to initiate PCR (Primers are provided in Source Data file). Three rounds of PCR amplification were performed to scale up the PCR products, with extensions containing homology to the pYD1 vector (Addgene #73447). The PCR products were purified using agarose gel electrophoresis, combined with linearized pYD1 vector DNA, and the DNA mixture was electroporated into EBY100 yeast, yielding 2 × 10⁷ transformants. The electroporated yeast cells were amplified in SDCAA selection media and induced in SGCAA induction media. Expression levels of toxins estimated by staining with anti-Myc antibody conjugated with FITC (Miltenyi) and analyzing by flow cytometry (Beckman CytoFLEX).

Protein preparation for selection: Purified and biotinylated receptor proteins (α1A, α2A, M₄AChR, M₂AChR) served as antigens. For α2A, recombinant plasmids encoding α2A sequence for library screening were transfected into HEK293F cells using PEI. Proteins were isolated via NI-NTA affinity chromatography, anti-Flag M1 resin and SEC under identical conditions to complex purification (omitting Fab/glue molecule addition), then concentrated to 5 mg mL^-1 using a 100-kDa molecular weight cutoff concentrator (Millipore). Half the purified α2A was taken to biotin label, α2A was incubated with EZ-link™ NHS-Biotin at 4 °C overnight. Excess biotin was removed by SEC. Biotinylated proteins were incubated with 5× molar excess streptavidin (25 °C, 30 min) and analyzed by SDS-PAGE. Identical protocols were applied to α1A and M₄AChR.

Selection process: Induced yeast cells were washed and resuspend in a selection buffer (PBS, 0.5 mM EDTA, 0.02% LMNG, 0.02% CHS and 0.5% bovine serum albumin). Clones that specific bound to α2A were enriched by one round of Magnetic Activated Cell Sorting (MACS) and three rounds of Fluorescence Activated Cell Sorting (FACS). For the first round MACS, 2×10⁸ yeast cells were precleared by incubating with 1 μM biotin labeled α1A and M₄AChR and passed LS column (Miltenyi) to remove unexpected binders. Flowed-through yeast cells were washed with the selection buffer, then incubated with 1 μM biotin labeled α2A and Streptavidin microbeads (Miltenyi). After incubation at 4°C for 2 h, yeast cells were loaded on the LS column, washed with the selection buffer and the elute cells by plunger were expanded and induced for further selection. For the selection rounds 2 and 4, FACS was performed to enrich high-affinity binders by CytoFLEX SRT (Beckman). Induced yeast cells were stained with 0.5 μM biotin labeled α2A, Streptavidin conjugated with Alexa Fluor^TM 647 (Thermo Fisher, S21374) as well as anti-Myc antibody conjugated with FITC (Miltenyi). Additionally, 1 μM unlabeled α1A and M₄AChR were added for competitive selection. Over 200,000 cells were collected in the selection round 2 and amplified in SDCAA. The selection round 3 and 4 were performed in the same way as in round 2 but with 0.1 μM and 0.05 μM biotin labeled α2A as antigen. The post selection cells were amplified and 30 clones were randomly sequenced to find the mutations contribution to the maturation. The consensus mutations were re-cloned into the pME vector with an N-terminal TYRP2-signal peptide, MBP, and 6xHis tags, and expressed and purified using the same protocol as for MT3.

Construction and selection of M₄AChR specific binder library

The same protocol was employed for constructing the M₄AChR specific library and selecting the specific binders. Clones specifically bound to M₄AChR were enriched by one round of MACS and four rounds of FACS. In the first round of selection, 2 × 10⁸ yeast cells were precleared by incubating with 1 μM biotin labeled α1A and α2A, while 1 μM biotin labeled M₄AChR was used for positive selection. For the subsequent FACS selection rounds, the concentration of M₄AChR was reduced to 0.5 μM, 0.2 μM, 0.1 μM and 0.05 μM in rounds 2 to 5. Additionally, unlabeled α1A and α2A were included for competitive selection. Following five rounds of selection, the cells were amplified and 30 clones were randomly sequenced to find the mutations contribution to the maturation and selectivity. The consensus mutations were further validated using the same protocol as previously described.

On-yeast binding assay

An on-yeast binding assay was used as a preliminary test to evaluate the binding ability and selectivity of MT3 and its engineered variants. Specifically, the coding sequences of MT3-WT, MT3-I9R, MT3-Y30A, MT3-Y35A and computationally redesigned variants A91, RD7, A64, and A66 (65 residues) were cloned into the pYD1 vector and transformed into EBY100 cells. The transformed yeast cells were cultured in SDCAA medium, and surface expression was induced in SGCAA medium for 48 h. For variants derived from selection, including A2S5, A2S5-MT3, A2-3, A2-6, A2-10, M4S3, M4-3, M4-9, M129 and M120 (65 residues), clones were similarly amplified in SDCAA medium and induced in SGCAA medium for 48 h.

Purified biotin-labeled target GPCRs, identical to those used in yeast selection, were utilized for binding and selectivity testing. The induced yeast cells were washed twice with selection buffer and incubated with biotin-labeled target GPCRs at their respective concentrations for 30 min at 4 °C. After incubation, the cells were washed twice with the cold buffer and further incubated with anti-Myc antibody conjugated with FITC (Miltenyi) and Streptavidin conjugated with Alexa Fluor^TM 647 (Thermo Fisher, S21374) for 1 h. The cells were then washed twice with cold buffer and the fluorescence signal was analyzed by CytoFLEX (Beckman).

On-cell binding assay

Biotin labeled recombinant toxins (e.g., MBP-MT3-His) were used to measure their binding affinity and specificity using flow cytometry. Specifically, the coding sequence of α1A, α1B, α1D, α2A, α2C, M₁AChR, M₄AChR and all mutant receptors were cloned into the pME vector with N-terminal HA signal peptide and cMyc tag. HEK293 cells (ATCC CRL-1573) were seeded at a density of 0.8 × 10⁶ cells per well in 6-well culture plates and incubated at 37 °C with 5% CO₂ 1 day prior to transfection. Cells were transiently transfected with plasmids encoding receptors using Lipo8000 (Beyotime, C0533). After 48 h of growth, the cells were harvested and divided into equal parts. Labeled toxins was added in serial dilution and incubated with cells at 4 °C for 1 h. After incubation, the cells were washed with ice-cold PBS three times. Streptavidin and Alexa Fluor^TM 647 conjugate (Thermo Fisher, S21374) were added to the cells and incubated at 4 °C for 1 h. The cells were washed and resuspended in ice-cold PBS, and the fluorescence intensity of single cells was quantified by a CytoFLEX (Beckman) with a 647 nm laser. The flow cytometry data were recorded and analyzed with CytExpert 2.3.0.84. The percentage of toxin-binding positive cells from 10000 cells per sample were used for analysis. Nonspecific binding was removed along with the corresponding fluorescence binding values in the final analysis. The binding affinity data were normalized with the highest value to 100 within the clones and fit by a one-site binding curve using Prism 8.0 (GraphPad Software Inc., San Diego).

Competitive ligand binding assay

Biotin labeled recombinant toxins (e.g., MBP-MT3-His) were used to measure their competitive ligand binding affinity via flow cytometry. Specifically, the coding sequence of wild-type α2A or M₄AChR was cloned into the pME vector with N-terminal HA signal peptide and cMyc tag. HEK293 cells (ATCC CRL-1573) were seeded at a density of 0.8 × 10⁶ cells per well in 6-well culture plates and incubated at 37 °C with 5% CO₂ 1 day prior to transfection. Cells were transiently transfected with plasmids encoding receptors using Lipo8000 (Beyotime, C0533). After 48 h of growth, the cells were harvested and divided into equal parts. The antagonists cabergoline (Topscience, T14853) for α2A and atropine (Topscience, T0375) for M₄AChR were added in serial dilution and incubate with cells at 4 °C for 1 h. Then, 250 nM biotin labeled toxins were added and incubated for 1 h at 4 °C. After incubation, the cells were washed with ice-cold PBS three times. Streptavidin and Alexa Fluor^TM 647 conjugate (Thermo Fisher, S21374) were added to the cells and incubated at 4 °C for 1 h. The cells were washed and resuspended in ice-cold PBS, and the fluorescence intensity of single cells was quantified by a CytoFLEX (Beckman) with a 647 nm laser. The flow cytometry data were recorded and analyzed with CytExpert 2.3.0.84. The percentage of toxin-binding positive cells from 10000 cells per sample were used for analysis. Nonspecific binding was removed along with the corresponding fluorescence binding values in the final analysis. The data were normalized with the highest value to 100 within the clones and fit a four-parameter sigmoidal concentration-response curve using Prism 8.0 (GraphPad Software Inc., San Diego).

NanoBiT-G-protein dissociation assay

GPCR-induced G-protein activation was measured by a NanoBiT-G-protein dissociation assay⁵⁷, in which interaction between a Gα subunit and a Gβγ subunit was monitored by a NanoBiT system (Promega). Specifically, a NanoBiT-G protein consisting of a large fragment (LgBiT)-containing Gα subunit (Gα_i1 for α2A, M₄AChR) and a small fragment (SmBiT)-fused Gγ₂ subunit (C68S mutant) was expressed together with untagged Gβ₁ subunit and with a test GPCR.

Agonist activity measurement: HEK293T cells were seeded in a 10 cm culture dish at a concentration of 2 × 10⁵ cells mL^-1 (10 mL per well) in DMEM (Gibco) supplemented with 10% fetal bovine serum (Gibco), glutamine, penicillin and streptomycin 1 day before transfection. Transfection solution was prepared by combining 20 µL of PEI and a plasmid mixture consisting of 500 ng LgBiT-containing Gα subunit, 2.5 µg Gβ₁, 2.5 µg SmBiT-fused Gγ₂ (C68S), and 1 µg a test GPCR (wild-type α2A or wild-type M₄AChR) in 1 mL of Opti-MEM (ThermoFisher Scientific). After a 1 day incubation, transfected cells were harvested with 0.5 mM EDTA containing Dulbecco’s PBS, centrifuged and suspended in 8 mL assay buffer (HBSS containing 0.01% BSA and 5 mM HEPES (pH 7.4)). The cell suspension was dispensed in a white 96-well plate at a volume of 60 µL per well and loaded with 20 µL of 50 µM furimazine (MCE) diluted in the assay buffer and the baseline luminescence was measured using Spectramax L luminometer (Molecular Devices). Then, titrated agonists (brimonidine (Topscience, T20157) for α2A, bethanechol chloride (Topscience, T3126) for M₄AChR) were diluted in the assay buffer at a 5× the final concentration and added at a volume of 20 μL per well and the second luminescent measurement was taken. The second luminescence counts were normalized to the initial count, and fold-change signals over vehicle treatment were used to plot the G-protein dissociation response (Supplementary Fig. 11a).

Antagonist activity measurement: Purified recombinant toxins (e.g., MBP-MT3-His) were used in this assay, chemical antagonists (cabergoline for α2A and atropine for M₄AChR) served as positive control (Supplementary Fig. 11b). Specifically, HEK293T cells were seeded in a 10 cm culture dish at a concentration of 2 × 10⁵ cells mL^-1 (10 mL per well) in DMEM (Gibco) supplemented with 10% fetal bovine serum (Gibco), glutamine, penicillin and streptomycin 1 day before transfection. Transfection solution was prepared by combining 20 µL of PEI and a plasmid mixture consisting of 500 ng LgBiT-containing Gα subunit, 2.5 µg Gβ₁, 2.5 µg SmBiT-fused Gγ₂ (C68S), and 1 µg a test GPCR in 1 mL of Opti-MEM (Thermo Fisher Scientific). After a 1 day incubation, transfected cells were harvested with 0.5 mM EDTA containing Dulbecco’s PBS, centrifuged and suspended in 8 mL assay buffer. The cell suspension was dispensed in a white 96-well plate at a volume of 60 µL per well. To measure antagonist activity of chemical antagonists, MT3, A2S5, A2-3, A2S5-MT3, M4S3, M4-3 and M4-9, titrated toxins or chemical antagonists were diluted in the assay buffer at a 10× the final concentration and added at a volume of 10 μL per well. After 1 h incubation, 20 µL of 50 µM furimazine were loaded to each well and the baseline luminescence were measured using Spectramax L luminometer (Molecular Devices). After that, a test agonist (10 μL, final concentration of 100 nM brimonidine for α2A; final concentration of 10 μM bethanechol chloride for M₄AChR) was manually added. Following an incubation of 3–5 min at room temperature, the second luminescent measurement was taken. The second luminescence counts were normalized to the initial count, and fold-change signals over vehicle treatment were used to plot the G-protein dissociation response. The G-protein dissociation signals were fitted to a four-parameter sigmoidal concentration-response curve using Prism 8 software (GraphPad Software Inc., San Diego), from which EC₅₀ and IC₅₀ values were calculated to determine mean and SEM.

Reporting summary

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

Data availability

The structural data generated in this study have been deposited in the Electron Microscopy Database (EMDB) and the protein data bank (PDB) with the following accession codes: 9IQV, EMD-60796 for the MT3-α1A; 9IQR, EMD-60793 for the MT3-α2A; 9IQS, EMD-60794 for the MT3-M₄AChR. The PDB datasets used for analysis in this study include 6WJC, 7YMJ, 6KUX, 5DSG, 8Z74, 8XQJ and 9EAH. All the other data generated in this study are provided in the Supplementary information and Source data are provided with this paper.

References

Strosberg, A. D. Structure, function, and regulation of adrenergic receptors. Protein Sci. 2, 1198–1209 (1993).
Article CAS PubMed PubMed Central Google Scholar
Wess, J., Eglen, R. M. & Gautam, D. Muscarinic acetylcholine receptors: mutant mice provide new insights for drug development. Nat. Rev. Drug Discov. 6, 721–733 (2007).
Article CAS PubMed Google Scholar
Wess, J. Novel muscarinic receptor mutant mouse models. Handb. Exp. Pharmacol. 2012, 95−117 (2012).
Ruffolo, R. R. Jr., Nichols, A. J., Stadel, J. M. & Hieble, J. P. Pharmacologic and therapeutic applications of alpha 2-adrenoceptor subtypes. Annu Rev. Pharm. Toxicol. 33, 243–279 (1993).
Article CAS Google Scholar
MacDonald, E., Kobilka, B. K. & Scheinin, M. Gene targeting–homing in on alpha 2-adrenoceptor-subtype function. Trends Pharm. Sci. 18, 211–219 (1997).
Article CAS PubMed Google Scholar
Furchgott, R. F. in Catecholamines (eds H. Blaschko & E. Muscholl) 283−335 (Springer Berlin Heidelberg, 1972).
Docherty, J. R. Subtypes of functional alpha1- and alpha2-adrenoceptors. Eur. J. Pharm. 361, 1–15 (1998).
Article CAS Google Scholar
Philipp, M., Brede, M. & Hein, L. Physiological significance of alpha(2)-adrenergic receptor subtype diversity: one receptor is not enough. Am. J. Physiol. Regul. Integr. Comp. Physiol. 283, R287–R295 (2002).
Article CAS PubMed Google Scholar
Maletic, V., Eramo, A., Gwin, K., Offord, S. J. & Duffy, R. A. The role of norepinephrine and its alpha-adrenergic receptors in the pathophysiology and treatment of major depressive disorder and schizophrenia: a systematic review. Front Psychiatry 8, 42 (2017).
Article PubMed PubMed Central Google Scholar
Neyama, H., Dozono, N., Uchida, H. & Ueda, H. Mirtazapine, an alpha2 antagonist-type antidepressant, reverses pain and lack of morphine analgesia in fibromyalgia-like mouse models. J. Pharm. Exp. Ther. 375, 1–9 (2020).
Article Google Scholar
Zhang, F., Wu, R., Qiang, X., Zhou, M. & Wang, P. Antagonism of alpha2A-adrenoceptor: a novel approach to inhibit inflammatory responses in sepsis. J. Mol. Med. (Berl.) 88, 289–296 (2010).
Article CAS PubMed Google Scholar
Link, R. E. et al. Cardiovascular regulation in mice lacking alpha2-adrenergic receptor subtypes b and c. Science 273, 803–805 (1996).
Article ADS CAS PubMed Google Scholar
Proudman, R. G. W., Akinaga, J. & Baker, J. G. The signaling and selectivity of alpha-adrenoceptor agonists for the human alpha2A, alpha2B and alpha2C-adrenoceptors and comparison with human alpha1 and beta-adrenoceptors. Pharm. Res Perspect. 10, e01003 (2022).
Article CAS Google Scholar
Proudman, R. G. W., Akinaga, J. & Baker, J. G. The affinity and selectivity of alpha-adrenoceptor antagonists, antidepressants and antipsychotics for the human alpha2A, alpha2B, and alpha2C-adrenoceptors and comparison with human alpha1 and beta-adrenoceptors. Pharm. Res. Perspect. 10, e00936 (2022).
Article CAS Google Scholar
Gentili, F., Pigini, M., Piergentili, A. & Giannella, M. Agonists and antagonists targeting the different alpha2-adrenoceptor subtypes. Curr. Top. Med. Chem. 7, 163–186 (2007).
Article CAS PubMed Google Scholar
Wu, Y., Zeng, L. & Zhao, S. Ligands of adrenergic receptors: a structural point of view. Biomolecules 11, 936 (2021).
Caulfield, M. P. Muscarinic receptors–characterization, coupling and function. Pharm. Ther. 58, 319–379 (1993).
Article CAS Google Scholar
Caulfield, M. P. & Birdsall, N. J. International union of pharmacology. XVII. Classification of muscarinic acetylcholine receptors. Pharm. Rev. 50, 279–290 (1998).
Article CAS PubMed Google Scholar
Dhein, S., van Koppen, C. J. & Brodde, O. E. Muscarinic receptors in the mammalian heart. Pharm. Res. 44, 161–182 (2001).
Article CAS Google Scholar
Saternos, H. C. et al. Distribution and function of the muscarinic receptor subtypes in the cardiovascular system. Physiol. Genom. 50, 1–9 (2018).
Article CAS Google Scholar
Volpicelli, L. A. & Levey, A. I. Muscarinic acetylcholine receptor subtypes in cerebral cortex and hippocampus. Prog. Brain Res. 145, 59–66 (2004).
Article CAS PubMed Google Scholar
Kruse, A. C. et al. Muscarinic acetylcholine receptors: novel opportunities for drug development. Nat. Rev. Drug Discov. 13, 549–560 (2014).
Article CAS PubMed PubMed Central Google Scholar
Foster, D. J., Jones, C. K. & Conn, P. J. Emerging approaches for treatment of schizophrenia: modulation of cholinergic signaling. Discov. Med. 14, 413–420 (2012).
PubMed PubMed Central Google Scholar
Dencker, D. et al. Muscarinic acetylcholine receptor subtypes as potential drug targets for the treatment of schizophrenia, drug abuse and parkinson’s disease. ACS Chem. Neurosci. 3, 80–89 (2012).
Article CAS PubMed Google Scholar
Myslivecek, J. Multitargeting nature of muscarinic orthosteric agonists and antagonists. Front. Physiol. 13, 974160 (2022).
Shekhar, A. et al. Selective muscarinic receptor agonist xanomeline as a novel treatment approach for schizophrenia. Am. J. Psychiatry 165, 1033–1039 (2008).
Article PubMed Google Scholar
Heinrich, J. N. et al. Pharmacological comparison of muscarinic ligands: historical versus more recent muscarinic M1-preferring receptor agonists. Eur. J. Pharm. 605, 53–56 (2009).
Article CAS Google Scholar
De Amici, M., Dallanoce, C., Holzgrabe, U., Trankle, C. & Mohr, K. Allosteric ligands for G protein-coupled receptors: a novel strategy with attractive therapeutic opportunities. Med. Res. Rev. 30, 463–549 (2010).
Article PubMed Google Scholar
Birdsall, N. J. & Lazareno, S. Allosterism at muscarinic receptors: ligands and mechanisms. Mini Rev. Med. Chem. 5, 523–543 (2005).
Article CAS PubMed Google Scholar
Conn, P. J., Christopoulos, A. & Lindsley, C. W. Allosteric modulators of GPCRs: a novel approach for the treatment of CNS disorders. Nat. Rev. Drug Discov. 8, 41–54 (2009).
Article CAS PubMed PubMed Central Google Scholar
Lane, J. R., Sexton, P. M. & Christopoulos, A. Bridging the gap: bitopic ligands of G-protein-coupled receptors. Trends Pharm. Sci. 34, 59–66 (2013).
Article CAS PubMed Google Scholar
Valant, C., Sexton, P. M. & Christopoulos, A. Orthosteric/allosteric bitopic ligands: going hybrid at GPCRs. Mol. Inter. 9, 125–135 (2009).
Article CAS Google Scholar
Steinfeld, T., Mammen, M., Smith, J. A., Wilson, R. D. & Jasper, J. R. A novel multivalent ligand that bridges the allosteric and orthosteric binding sites of the M2 muscarinic receptor. Mol. Pharm. 72, 291–302 (2007).
Article CAS Google Scholar
Davenport, A. P., Scully, C. C. G., de Graaf, C., Brown, A. J. H. & Maguire, J. J. Advances in therapeutic peptides targeting G protein-coupled receptors. Nat. Rev. Drug Discov. 19, 389–413 (2020).
Article CAS PubMed Google Scholar
Bradley, K. N. Muscarinic toxins from the green mamba. Pharm. Ther. 85, 87–109 (2000).
Article CAS Google Scholar
Servent, D. & Fruchart-Gaillard, C. Muscarinic toxins: tools for the study of the pharmacological and functional properties of muscarinic receptors. J. Neurochem. 109, 1193–1202 (2009).
Article CAS PubMed Google Scholar
Blanchet, G. et al. Polypharmacology profiles and phylogenetic analysis of three-finger toxins from mamba venom: case of aminergic toxins. Biochimie 103, 109–117 (2014).
Article CAS PubMed Google Scholar
Max, S. I., Liang, J. S. & Potter, L. T. Purification and properties of m1-toxin, a specific antagonist of m1 muscarinic receptors. J. Neurosci. 13, 4293–4300 (1993).
Article CAS PubMed PubMed Central Google Scholar
Fruchart-Gaillard, C., Mourier, G., Marquer, C., Menez, A. & Servent, D. Identification of various allosteric interaction sites on M1 muscarinic receptor using 125I-Met35-oxidized muscarinic toxin 7. Mol. Pharm. 69, 1641–1651 (2006).
Article CAS Google Scholar
Olianas, M. C., Adem, A., Karlsson, E. & Onali, P. Rat striatal muscarinic receptors coupled to the inhibition of adenylyl cyclase activity: potent block by the selective m4 ligand muscarinic toxin 3 (MT3). Br. J. Pharm. 118, 283–288 (1996).
Article CAS Google Scholar
Nareoja, K. et al. Adrenoceptor activity of muscarinic toxins identified from mamba venoms. Br. J. Pharm. 164, 538–550 (2011).
Article CAS Google Scholar
Guo, Q. et al. A method for structure determination of GPCRs in various states. Nat. Chem. Biol. 20, 74–82 (2024).
Article CAS PubMed Google Scholar
Watson, J. L. et al. De novo design of protein structure and function with RFdiffusion. Nature 620, 1089–1100 (2023).
Article ADS CAS PubMed PubMed Central Google Scholar
Vázquez Torres, S. et al. De novo design of high-affinity binders of bioactive helical peptides. Nature 626, 435–442 (2024).
Article ADS PubMed Google Scholar
Newman, D. J. & Cragg, G. M. Natural products as sources of new drugs from 1981 to 2014. J. Nat. Prod. 79, 629–661 (2016).
Article CAS PubMed Google Scholar
Gilles, N. & Servent, D. The European FP7 venomics project. Future Med. Chem. 6, 1611–1612 (2014).
Article CAS PubMed Google Scholar
Gallwitz, B. Exenatide in type 2 diabetes: treatment effects in clinical studies and animal study data. Int J. Clin. Pr. 60, 1654–1661 (2006).
Article CAS Google Scholar
Rubiu, E. et al. A benefit/risk assessment of intrathecal ziconotide in chronic pain: a narrative review. J. Clin. Med. 13, 1644 (2024).
Maeda, S. et al. Structure and selectivity engineering of the M(1) muscarinic receptor toxin complex. Science 369, 161–167 (2020).
Article ADS CAS PubMed PubMed Central Google Scholar
Punjani, A., Rubinstein, J. L., Fleet, D. J. & Brubaker, M. A. cryoSPARC: algorithms for rapid unsupervised cryo-EM structure determination. Nat. Methods 14, 290–296 (2017).
Article CAS PubMed Google Scholar
Bepler, T., Kelley, K., Noble, A. J. & Berger, B. Topaz-Denoise: general deep denoising models for cryoEM and cryoET. Nat. Commun. 11, 5208 (2020).
Article ADS CAS PubMed PubMed Central Google Scholar
Pettersen, E. F. et al. UCSF Chimera–a visualization system for exploratory research and analysis. J. Comput. Chem. 25, 1605–1612 (2004).
Article CAS PubMed Google Scholar
Emsley, P. & Cowtan, K. Coot: model-building tools for molecular graphics. Acta Crystallogr. D. Biol. Crystallogr. 60, 2126–2132 (2004).
Article ADS PubMed Google Scholar
Adams, P. D. et al. PHENIX: a comprehensive Python-based system for macromolecular structure solution. Acta Crystallogr. D. Biol. Crystallogr. 66, 213–221 (2010).
Article ADS CAS PubMed PubMed Central Google Scholar
Dauparas, J. et al. Robust deep learning-based protein sequence design using ProteinMPNN. Science 378, 49–56 (2022).
Article ADS CAS PubMed PubMed Central Google Scholar
Jumper, J. et al. Highly accurate protein structure prediction with AlphaFold. Nature 596, 583–589 (2021).
Article ADS CAS PubMed PubMed Central Google Scholar
Inoue, A. et al. Illuminating G-protein-coupling selectivity of GPCRs. Cell 177, 1933–1947.e1925 (2019).
Article CAS PubMed PubMed Central Google Scholar

Download references

Acknowledgements

Cryo-EM data were collected at the Center of Cryo-Electron Microscopy, University of Science and Technology of China. We thank staff members at the Center for Integrative Imaging, Hefei National Laboratory for Physical Sciences at the Microscale and University of Science and Technology of China for cryo-EM sample examination. This work was supported by the National Key Research and Development Program of China Grants (2024YFC3405500, 2022YFF1203100 to Y.T.), the National Natural Science Foundation of China (32401024 to Q.G.), Anhui Provincial Natural Science Foundation (2308085QC77 to Q.G.), the Center for Advanced Interdisciplinary Science and Biomedicine of IHM (QYZD20220006 to Y.T.), and the Fundamental Research Funds for the Central Universities (WK9100000070 to Q.G., WK9990000165 to D.W.).

Author information

These authors contributed equally: Yixuan Zhong, Huihui Tao.

Authors and Affiliations

Department of Laboratory Medicine, The First Affiliated Hospital of USTC, MOE Key Laboratory for Membraneless Organelles and Cellular Dynamics, Hefei National Center for Cross-disciplinary Sciences, Biomedical Sciences and Health Laboratory of Anhui Province, Center for Advanced Interdisciplinary Science and Biomedicine of IHM, Division of Life Sciences and Medicine, University of Science and Technology of China, 230027, Hefei, P.R. China
Yixuan Zhong, Huihui Tao, Yu Zhang, Binbin He, Dandan Wang, Maikun Teng, Qiong Guo & Yuyong Tao
Kobilka Institute of Innovative Drug Discovery, School of Medicine, The Chinese University of Hong Kong, Shenzhen, China
Haizhan Jiao & Hongli Hu

Authors

Yixuan Zhong
View author publications
Search author on:PubMed Google Scholar
Huihui Tao
View author publications
Search author on:PubMed Google Scholar
Yu Zhang
View author publications
Search author on:PubMed Google Scholar
Binbin He
View author publications
Search author on:PubMed Google Scholar
Haizhan Jiao
View author publications
Search author on:PubMed Google Scholar
Dandan Wang
View author publications
Search author on:PubMed Google Scholar
Maikun Teng
View author publications
Search author on:PubMed Google Scholar
Hongli Hu
View author publications
Search author on:PubMed Google Scholar
Qiong Guo
View author publications
Search author on:PubMed Google Scholar
Yuyong Tao
View author publications
Search author on:PubMed Google Scholar

Contributions

Y.Zhong. and H.T. designed, expressed, purified, and prepared the cryo-EM samples. Y.Zhong. collected the cryo-EM data and helped with data processing. H.H., Y.Zhong. and H.J. processed the cryo-EM data. Y.Zhang. conducted the computational work. H.T. and Y.Zhong. performed yeast selection for specific binders. Y.Zhong. and H.T. conducted the biochemical experiment. B.H., D.W., and M.T. help with data analysis. Y.T. and Q.G. conceived and designed the overall project, supervised the research, and wrote the manuscript with input from all co-authors.

Corresponding authors

Correspondence to Hongli Hu, Qiong Guo or Yuyong Tao.

Ethics declarations

Competing interests

Y.T., Q.G., Y.Zhong, and H.T. have filed an invention patent for the toxin variants described in this work (application number: CN202510869589.7; status: pending). The other authors declare no competing interests.

Peer review

Peer review information

Nature Communications thanks Nicolas Gilles, Shoji Maeda, Joseph Watson and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

Supplementary Information (download PDF )

Peer Review file (download PDF )

Reporting Summary (download PDF )

Source data

Source Data (download XLSX )

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.

Reprints and permissions

About this article

Cite this article

Zhong, Y., Tao, H., Zhang, Y. et al. Structure-guided engineering of snake toxins for selective modulation of adrenergic and muscarinic receptors. Nat Commun 16, 6478 (2025). https://doi.org/10.1038/s41467-025-61695-0

Download citation

Received: 25 January 2025
Accepted: 30 June 2025
Published: 14 July 2025
Version of record: 14 July 2025
DOI: https://doi.org/10.1038/s41467-025-61695-0