Module 2: Protein Engineering of GLP-1
The short plasma half-life (~2 minutes) of native GLP-1, caused by rapid degradation from enzymes like DPP-IV and neprilysin (also known as NEP-24.11, NEP) limits its therapeutic potential. While the mechanism and structural basis of DPP-IV has been well-studied, strategies against NEP-24.11 remain less explored. To address this, we rationally designed a GLP-1 variant, GLP-1Z (A8H, E27K, K34R) based on literature review. Our integrated strategy uses computational modeling methods such as molecular docking and dynamics simulations to identify optimal mutations, particularly the incorporation of unnatural amino acids, followed by subsequent experimental validation of the mutants’ stability and activity.
Part 1: In Silico Screening for Optimal Mutants and Unnatural Amino Acid Combinations
The computational modeling part of our protein engineering module focuses on two objectives. First, to rationally design mutations by analyzing the GLP-1(Z)-NEP complex interface to reduce enzymatic cleavage and extend GLP-1’s in vivo circulation half-life. Second, to employ computational scanning methods to evaluate the binding free energy changes (ddG) associated with incorporating unnatural amino acids across the entire GLP-1 sequence, thereby identifying mutation sites that optimize protein stability.
1.Molecular Docking
Molecular docking simulations were performed to predict the binding conformations of core complexes of GLP-1(Z) with NEP and the GLP-1 receptor (GLP-1R), providing structural insights for sequence optimization.
(1)Docking of GLP-1 with NEP
After multi-platform docking screening, we obtained one ideal docking conformation from each of the ClusPro 2.0 and HADDOCK platforms (denoted as Model I and Model II respectively). We then performed molecular dynamics simulations of 300 ns on these two conformations to screen for stable conformations. The simulation results showed that Model I reached a stable state within 300 ns, whereas Model II exhibited persistent instability and was therefore discarded.
Figure 1. Two docking conformations of GLP-1 with NEP. A. Docking conformation from the HADDOCK platform: blue represents NEP, pink represents GLP-1, and the red region indicates the active site of NEP. B. Docking conformation from the ClusPro 2.0 platform: yellow-green represents NEP, blue represents GLP-1, and the green region indicates the active site of NEP.
(2)Docking of GLP-1Z with NEP
Docking analysis was performed on the variant GLP-1Z. Its structure was first predicted using AlphaFold3. Following a similar workflow, its complex with NEP was analyzed to identify key interaction interfaces, providing a theoretical basis for screening sites for unnatural amino acid incorporation. A stable docking model was ultimately obtained using the ClusPro 2.0 platform.
Figure 2. Optimal molecular docking conformation model of GLP-1-Z with NEP from the ClusPro 2.0 platform, where NEP-24.11 (pink) has its active site marked in purple, and the GLP-1Z peptide chain is in green.
(3)Docking of GLP-1(Z) with GLP-1R
To ensure that the engineered sequence retained significant receptor activation potential, we conducted molecular docking analysis of GLP-1(Z) with GLP-1R. This helped determine the conservation of key binding residues, allowing for the selection of modification sites that balance enhanced stability with maintained biological activity.
Figure 3. Comparison of docking conformations of GLP-1, GLP-1-Z with GLP-1R. A. Docking conformation of GLP-1 with GLP-1R (PDB ID: 9A3F): light purple surface structure is GLP-1R, and dark purple is GLP-1 peptide chain. B. Docking conformation of GLP-1-Z (structure predicted by AlphaFold3) with GLP-1R: light purple surface structure is GLP-1R, and gold is the engineered GLP-1-Z peptide chain.
2.Molecular Dynamics Simulations
Molecular dynamics (MD) simulations were employed to assess intermolecular interactions and dynamic stability, thereby obtaining stable conformations and inferring the susceptibility of enzymatic sites within GLP-1. The simulation results showed that the NEP-GLP-1 complex remained stable throughout the 300 ns simulation, with the root mean square deviation (RMSD) of its backbone atoms fluctuating within a range of 0.2-0.3 nm and averaging 0.24 ± 0.019 nm. Notably, after 100 ns, the RMSD fluctuation amplitude was less than 0.5 Å, indicating a well-equilibrated complex that provides a stable structural basis for NEP-mediated cleavage. In contrast, the NEP-GLP-1Z complex exhibited a significantly higher average RMSD (0.28 ± 0.016 nm), despite a similarly small fluctuation amplitude. This suggests reduced binding stability between GLP-1Z and NEP, potentially indicating lower enzymatic cleavage susceptibility.
The hydrogen bond network is also crucial for maintaining the conformation of the zinc-binding motif, the hydrophobic pocket, and the specific catalytic mechanism of the NEP active site. To investigate changes in the binding mode between GLP-1(Z) and NEP, we analyzed a 300 ns MD simulation trajectory. A distance of ≤ 3.5 Å between donor and acceptor atoms was used as the criterion for hydrogen bond formation. The results indicated that H7, K26, and K34 of GLP-1 are the primary residues involved in hydrogen bonding with NEP. Given that position 7 is a well-established key residue for GLP-1 receptor interaction, our analysis suggests K26 and K34 as promising modifiable sites. In the engineered GLP-1Z, residues H8, E21, and K26 still exhibit a high propensity to form hydrogen bonds with NEP, facilitating binding and subsequent proteolysis. Therefore, further modifications at these sites could be explored to reduce affinity for NEP.
Figure 4. MD simulation results on interactions between NEP and GLP-1(Z). A. The RMSD curve of NEP-GLP-1 complex over 0 to 300 ns MD simulation. B. 3D structure of molecular docking between NEP and GLP-1. C. The RMSD curve of NEP-GLP-1Z complex over 0 to 300 ns MD simulation. D. 3D structure of molecular docking between NEP and GLP-1Z. E. Critical hydrogen bonds at the NEP-GLP-1 binding interface. F. Critical hydrogen bonds at the NEP-GLP-1Z binding interface.
3.Library screening with unnatural amino acids and ΔΔG validation via Rosetta
The Rosetta energy function is an optimized scoring system used to evaluate the stability and reliability of protein structural models. It calculates interatomic interaction energies to provide a quantitative assessment of model quality. We identified optimal sites for unnatural amino acid substitution through a computational analysis using two Rosetta protocols. Cartesian ddG assessed mutational effects on the stability of the isolated GLP-1(Z) molecule, while Flex ddG evaluated the resulting changes in binding free energy within the GLP-1(Z)-GLP-1R complex. This allowed us to calculate ΔΔG for mutations at every position across the GLP-1(Z) sequence.
Based on preliminary wet-lab results, we focused the Rosetta analysis primarily on 5-HTP substitutions. Mutation sites yielding negative ΔΔG values were selected, as this indicates a potential beneficial effect on protein stability. From these, two top candidate sites for 5-HTP substitution in GLP-1Z , L20 and R34, exhibited negative ΔΔG values in both methods and are predicted to enhance both intrinsic protein stability and GLP-1R binding affinity, whereas no sites in GLP-1 produced similarly favorable results.
Figure 5. Rosetta ΔΔG analysis of 5-HTP incorporation in GLP-1(Z) with Cartesian ddG and Flex ddG. A. Favorable ΔΔG heat map for 5-HTP incorporation in GLP-1. B. Favorable ΔΔG heat map for 5-HTP incorporation in GLP-1Z.
4.Summary
The aforementioned computational simulations provided us with deeper insights into the binding modes and spatial conformations of GLP-1 and its variant GLP-1Z with both NEP and GLP-1R. These findings identified key residues involved in intermolecular interactions and several candidate sites suitable for unnatural amino acid substitution. Subsequent wet-lab experiments could introduce amber stop codons at these positions to experimentally validate the predictions.
Part 2: Validation of the Unnatural Amino Acid Incorporation System
To verify the feasibility of the unnatural amino acid incorporation system, we first validated the insertion of sulfotyrosine (sTyr) and 5-hydroxytryptophan (5-HTP) into superfolder GFP (sfGFP) in E. coli BL21(DE3) using an amber codon suppression strategy. This system was then implemented in the engineered chassis E. coli Nissle 1917 (EcN) to evaluate its applicability and efficiency. Following this validation, we would proceed to incorporate sTyr and 5-HTP into target sites within GLP-1(Z).
1.Construction of expression plasmids for the unnatural amino acid incorporation system
We acquired the unnatural amino acid incorporation plasmid pET-GFP2TAG (BBa_25R2L2YD), pUltra-sTyr and pUltra-CNF from our Primary PI Dr. Haoran Yu’s lab. To generate the pUltra-5-HTP plasmid, the aaRS gene in pUltra-CNF was replaced with a synthesized orthogonal 5-HTP aminoacyl-tRNA synthetase gene ScTrpRS (BBa_25QIIBEZ), and its tRNA sequence was subsequently engineered to the orthogonal SctRNA5-HTP (BBa_255G11B4) via circular PCR. After verifying the target plasmids by sequencing, we co-transformed pUltra-sTyr or pUltra-5-HTP with pET-GFP2TAG into E. coli BL21(DE3). Transformants were selected on LB plates with ampicillin and spectinomycin at 37 °C, and positive colonies were cultured for further analysis.
Figure 6. Construction of expression plasmids for ncAA incorporation. A. Plasmid map of pET-GFP2TAG. B. Plasmid map of pUltra-sTyrRS-tRNATyr. C. Plasmid map of pUltra-Sc (5-HTP)RS-tRNATyr. D. Sequencing results of the engineered tRNA in the pUltra-Sc (5-HTP)RS-tRNATyr plasmid.
Since native EcN lacks T7 RNA polymerase, the original pET-GFP2TAG expression system with T7 promoter is non-functional in this chassis. To address this, we replaced the T7 operator in the pET-GFP2TAG plasmid with the arabinose operon via circular PCR and Gibson assembly, thereby obtaining the functional plasmid pET-araC-PBAD-GFP2TAG suitable for EcN.
Due to EcN’s thick cell wall, which leads to very low chemical transformation efficiency, we adopted electroporation. Following optimization of electroporation parameters including cell concentration and incubation time, and a sequential plasmid transformation protocol, we successfully introduced both the aaRS-tRNA expression plasmid and the GFP reporter plasmid into EcN, enabling subsequent characterization of unnatural amino acid incorporation.
Figure 7. Optimization and transformation of the expression plasmids in EcN. A. Results of electrophoretic bands of the arabinose operon fragment (~2.1 kb) and the vector with modified sfGFP (~4.5 kb) after PCR. B. Plasmid map of pET-araC-PBAD-GFP2TAG. C. Sequencing results of the pET-araC-PBAD-GFP2TAG plasmid. D. Successfully transformed EcN grew on the LB plates with AmpR and SpeR.
2.Characterization and optimization of the unnatural amino acid incorporation system
We picked single colonies of transformed E. coli BL21(DE3) and EcN and grew them in 3 mL of LB medium with ampicillin and spectinomycin at 37 °C for 8-10 hours. Cells were then harvested, resuspended in antibiotic-supplemented LB or GMML medium, and induced overnight with ncAAs (1 mM sTyr or 5-HTP), along with arabinose (6 mM, 30 °C) or IPTG (0.5 mM, 37 °C). After 12 hours, we measured OD600 and fluorescence intensity in a microplate reader and normalized fluorescence to OD600 to evaluate ncAA incorporation efficiency.
The fluorescence results revealed a significant strain-dependent specificity for UAA incorporation. In BL21(DE3), the addition of sTyr resulted in a stable and significant increase in fluorescence. In contrast, the 5-HTP group showed a lower fold-increase in signal intensity with substantial background fluorescence, indicating lower incorporation fidelity and specificity. Conversely, in EcN, 5-HTP incorporation was more efficient, achieving a signal-to-noise ratio of approximately 4-fold in GMML medium, whereas sTyr incorporation was largely ineffective. These results clearly highlight the critical influence of the host chassis on unnatural amino acid incorporation efficiency. Further optimization of the plasmid expression system and aaRS-tRNA expression levels is required to for specific host-ncAA pairs.
Figure 8. Characterization of ncAA incorporation efficiency over different strains and medium. A. E. coli BL21(DE3) cultured in LB medium. B. E. coli BL21(DE3) cultured in GMML medium. C. EcN cultured in LB medium. D. EcN cultured in GMML medium.
3.Summary
Our experiments have successfully established a functional unnatural amino acid incorporation platform in EcN and identified 5-HTP as our primary candidate for subsequent dry and wet lab investigations. Due to our timeframe and the availability of high-cost ncAA agents, we did not include validating the de novo biosynthesis pathway for 5-HTP. However, its feasibility can be reasonably supported by existing literature. To build on these promising findings, subsequent work could focus on strengthening the robustness of our system and optimizing incorporation efficiency through further parameter refinement.
Part 3: Purification and Characterization of GLP-1(Z)
The expression and purification of the small-molecular peptide GLP-1 in E. coli are challenging due to its poor stability and low recovery. To address this, we fused an N-terminal His-SUMO tag (BBa_25DRMK5X and BBa_2504KELH) to both GLP-1 and GLP-1Z sequences. The incorporation of a His-tag enables protein enrichment via nickel-affinity chromatography, while the small ubiquitin-like modifier (SUMO) tag increases the molecular weight and improves the intracellular solubility of the recombinant protein. The fusion tag was employed to enhance overall protein expression and solubility, thereby facilitating subsequent concentration and purification steps, such as nickel-affinity purification and ultrafiltration.
1.Construction of expression plasmids with an N-terminal His-SUMO tag
The DNA fragment encoding the His-SUMO tag was obtained from our Primary PI Dr. Yu's lab. After PCR amplification and verification by gel electrophoresis, the fragment was inserted into specific sites of the pET-GLP-1 and pET-GLP-1Z plasmids via Gibson assembly. The assembled products were then transformed into E. coli BL21(DE3). Positive clones were selected on kanamycin-resistant LB plates and verified by DNA sequencing to select the desired recombinant plasmids: pET-His-SUMO-GLP-1 (BBa_25DRMK5X) and pET-His-SUMO-GLP-1Z-CPP (BBa_2504KELH).
Figure 9. Construction of expression plasmids for GLP-1(Z) with an N-terminal His-SUMO tag. A. Plasmid map of pET-His-SUMO-GLP-1. B. Plasmid map of pET-His-SUMO-GLP-1Z with a cell-penetrating peptide (CPP).
2.Concentration and purification of recombinant GLP-1(Z)
Following successful sequencing, the plasmid was transformed into E. coli BL21(DE3). Single colonies were inoculated and cultured in LB liquid medium containing kanamycin. When the culture density reached an OD600 of 0.6-0.8, we induced the protein expression with 0.5 mM IPTG overnight.
After induction, the cells were lysed by ultrasonication, and the supernatant containing soluble proteins was collected by centrifugation. Leveraging the N-terminal His-tag, the supernatant was subjected to initial purification via nickel-affinity chromatography, involving column equilibration, binding, washing, and imidazole gradient elution. To obtain the mature, tag-free GLP-1(Z) protein, we cleaved the SUMO fusion tag from the nickel-column-purified product using SUMO protease. The cleavage mixture was then concentrated and further purified through ultrafiltration.
Samples from each purification stage were analyzed by SDS-PAGE to assess protein expression levels and purification efficiency. SDS-PAGE results indicated the following: The native GLP-1 sequence (3.37 kDa) appeared as a diffuse band. The SUMO tag (13.7 kDa) was visible in Box I. The uncut SUMO-GLP-1 fusion protein (16.96 kDa) was present at high concentration, suggesting the cleavage reaction required optimization. The target GLP-1Z-CPP protein (5.70 kDa), after SUMO tag removal, also showed a diffuse band with lower intensity than GLP-1 as shown in Box II, potentially due to structural constraints hindering efficient protease cleavage. Furthermore, low sample concentration contributed to faint band visibility.
Figure 10. SDS-PAGE results of GLP-1(Z) samples during purification. A. Enzymatic cleavage results of His-SUMO-GLP-1. (1: Unincubated SUMO-GLP-1 protein sample; 2: His-SUMO-GLP-1 sample incubated with SUMO protease; 3: Effluent sample after nickel-affinity purification; 4: Impurity-washing sample with 50 mM imidazole; 5: Eluted sample with 250 mM imidazole; 6: Supernatant from ultrafiltration of Sample 3; 7: Precipitate from ultrafiltration of Sample 3; 8: Supernatant from ultrafiltration of Sample 4; 9: Precipitate from ultrafiltration of Sample 4; 10: Supernatant from ultrafiltration of Sample 5; Bands corresponding to the cleaved SUMO tag are present in Samples 8 and 10.) B. Enzymatic cleavage results of His-SUMO-GLP-1Z-CPP. (1: Unincubated His-SUMO-GLP-1Z-CPP sample; 2: His-SUMO-GLP-1Z-CPP sample incubated with SUMO protease; 3: Effluent sample after nickel-affinity purification; 4: Impurity-washing sample with 50mM imidazole; 5: Eluted sample with 250mM imidazole; 6: Supernatant of Sample 4 after ultrafiltration; 7: Precipitate of Sample 5 after ultrafiltration.)
3.Summary
These results demonstrate that the SUMO protease removes both the SUMO tag and the adjacent His-tag during cleavage, making subsequent characterization such as His-tag pull-down assays or repurification of the target GLP-1 protein challenging. Based on this, we propose an optimized design by inserting the His-tag between the SUMO tag and the GLP-1(Z) sequence, creating a SUMO-His-GLP-1(Z) fusion construct. Theoretically, after cleavage, the GLP-1 protein would retain the His-tag, allowing it to be separated from the SUMO protease by nickel-affinity chromatography, with final isolation achievable via ultrafiltration based on molecular weight differences.
Conclusions
Our project aimed to integrate dry and wet lab experiments for engineering the therapeutic peptide GLP-1. Our strategy combined computational design with experimental validation to develop a robust platform for its optimization. We initiated the process by selecting a patented GLP-1 variant (GLP-1Z) as a starting scaffold and employed molecular docking, MD simulations and ΔΔG calculations to identify promising sites for unnatural amino acid incorporation, such as 5-HTP.
In parallel, we carried out systematic wet experiment step by step for verification. This involved developing a His-SUMO fusion tag system, which was successfully validated for enhancing the solubility and facilitating the purification of recombinant GLP-1 via nickel-affinity chromatography and ultrafiltration. Furthermore, we established a functional ncAA incorporation system in E. coli, using an sfGFP reporter harboring two amber codons to confirm the feasibility of site-specific incorporation and optimizing aaRS-tRNA expression plasmids for improved efficiency.
Our project outlines a clear path for future work. Possible next steps include optimizing GLP-1 expression and SUMO protease cleavage conditions, followed by experimental validation of top-predicted ncAA sites in GLP-1Z within the EcN chassis. Crucially, the stability-enhanced variants must be rigorously characterized for retained GLP-1R binding affinity and extended plasma half-life to assess therapeutic potential. Ultimate validation will require in vivo efficacy and safety studies in animal models. For an integrated system, our long-term goals involve engineering EcN with genetic circuits for optimized GLP-1 expression, quantified using strategies such as mass spectrometry and ELISA. Further refinement is essential to enable the intelligent production of long-acting GLP-1 therapies via engineered probiotics.