First, our team successfully constructed and validated the expression systems for GLP-1 and Bimagrumab, detecting the expression of both proteins extracellularly. Secondly, we designed DPP-4 protein inhibitors at the computational level and verified their affinity to DPP-4 through molecular dynamics simulations, showing superiority over GLP-1. Then, we successfully built an arabinose operon and a low-temperature induced promoter to create a cell suicide system, confirming its functionality. Finally, using pSB1A3 as a vector, we successfully transformed the constructed fat-reduction system, muscle-gaining system, and suicide system into Escherichia coli Nissle 1917. Additionally, we simulated the intestinal environment and conducted characterization analysis on EcN, demonstrating its effective survival in the simulated intestinal conditions.

To obtain engineered strains capable of effectively secreting GLP-1 and Bimagrumab, we designed a genetic circuit containing the sequences for GLP-1 and Bimagrumab (Fig 1). We selected E. coli BL21 as the chassis microorganism for constructing the engineered strains and utilized genetic engineering techniques to clone GLP-1, PelB-GLP-1, Bimagrumab, PelB-Bimagrumab, and PelB-GLP-1-Bimagrumab into E. coli BL21.

Firstly, we obtained the sequences for GLP-1, Bimagrumab, and the secretion signal peptide PelB from NCBI and optimized the codons for E. coli. We then designed PCR primer sequences (Fig 2.A) and had a company synthesize the coding sequences and primer sequences for GLP-1, PelB-GLP-1, Bimagrumab, and PelB-Bimagrumab. We amplified the gene sequences of GLP-1, PelB-GLP-1, Bimagrumab, and PelB-Bimagrumab through PCR. Using overlap PCR, we integrated the PelB-GLP-1 fragment and the Bimagrumab fragment to obtain the PelB-GLP-1-Bimagrumab fragment (Fig 2.B). Subsequently, we cloned the five fragments into the pET23b plasmid via NdeI and XhoI restriction sites. Finally, the recombinant plasmids were transformed into E. coli BL21, and we plated them on LB solid media containing ampicillin, ultimately obtaining single colonies (Fig 2.C).

The results indicate that we successfully amplified the fragments of GLP-1, PelB-GLP-1, Bimagrumab, PelB-Bimagrumab, and PelB-GLP-1-Bimagrumab, and cloned them into the vector pET23b. Ultimately, we obtained single colonies on LB solid media containing ampicillin.

To verify the expression of GLP-1 and Bimagrumab, as well as the functionality of PelB, we measured the concentrations of GLP-1 and Bimagrumab in both intracellular and extracellular environments using ELISA and Western blotting. After centrifuging cultures containing GLP-1, PelB-GLP-1, Bimagrumab, and PelB-Bimagrumab, we obtained supernatants and bacterial pellets, extracting proteins from the pellets using lysis methods. Using an ELISA kit (Shanghai Enzyme Linked, YJ022729), we confirmed GLP-1 expression in the engineered strains. Results showed that without the secretion signal peptide PelB, the intracellular concentration of GLP-1 reached 61.26 pg/mL, while the extracellular concentration was only 9.19 pg/mL; with PelB assistance, extracellular GLP-1 levels increased to 63.81 pg/mL (Fig 3.B). Western blotting of extracted GLP-1 and Bimagrumab confirmed that, without PelB, both proteins were predominantly intracellular. However, after attaching PelB to the N-terminus of GLP-1 and Bimagrumab, their extracellular concentrations significantly increased (Fig 3.A and C). Finally, we verified that GLP-1, PelB-GLP-1, Bimagrumab, and PelB-Bimagrumab fragments did not affect the growth of the strains (Fig 3.D).

The experimental results indicate that the engineered strains successfully expressed GLP-1 and Bimagrumab, with significantly increased extracellular levels of both proteins facilitated by the secretion signal peptide PelB. Additionally, we demonstrated that the genetic engineering did not have a significant impact on the growth of BL21.

Since there are no naturally occurring protein inhibitors that can competitively bind to DPP-4, our team chose to design a DPP-4 protein inhibitor from scratch using deep learning models. We established a comprehensive workflow based on advanced deep learning models: RFdiffusion for generating the protein backbone, ProteinMPNN for side-chain generation, ESMfold for screening high-quality sequences, Amber for molecular dynamics simulations, and MM/PBSA to compare the binding energies of GLP-1 and the inhibitor with DPP-4 (Fig 4.A). During the backbone generation with RFdiffusion, we utilized the Practical Considerations for Binder Design module, selecting critical residues E167, E168, and Y624 in the DPP-4 pocket as hotspots. We generated 200 sequences of random lengths (10-100 amino acids) and filtered for promising scaffolds. The selected scaffolds were subjected to side-chain generation using ProteinMPNN's FastRelax protocol, resulting in 18 sequences suitable for molecular dynamics simulations based on RMSD < 15 and pLDDT > 80 (Fig 4.B). We processed these sequences with AmberTools and conducted long-term molecular dynamics simulations using Amber, with a 100 ns equilibration and 100 ns of production (totaling 200 ns). MM/PBSA calculations yielded one sequence with a binding energy superior to that of GLP-1 with DPP-4: DPP-4PI had a binding energy of -6522.0487 kcal/mol, compared to GLP-1's -6474.9481 kcal/mol (Fig 4.C). Finally, we analyzed the molecular dynamics trajectories, confirming that RMSD and RMSF values remained within acceptable ranges (Fig 4.D and E). We selected a stable frame from the simulation for structural observation, revealing the binding mode between DPP-4 and its protein inhibitor (Fig 5).

The results indicate that, through the established workflow and utilizing existing deep learning models, we successfully designed a DPP-4 protein inhibitor and validated its feasibility at the computational level.

To prevent the leakage of genetic material from the engineered strains, we connected the arabinose operon and low-temperature inducible promoter in an "OR" gate configuration, allowing the engineered strains to express lysozyme under specific conditions for self-destruction (Fig 6A). First, we surveyed the iGEM parts library and identified the arabinose operon pBAD (BBa_I13453) and the low-temperature inducible promoter pCspA (BBa_K4987003), which we then amplified (Fig 6B).

We measured the expression of red fluorescent protein induced by the arabinose operon (Fig 7.A). As the concentration of arabinose increased, the Fluorescence/OD600 values consistently rose, demonstrating that the arabinose operon can effectively induce the expression of the corresponding protein in response to arabinose (Fig 7.B) [3].

First, we validated the functionality of the low-temperature inducible promoter. Our experiments demonstrated that the optimal growth temperature for E. coli DH5α is 37°C, with growth slowing as the temperature decreases (Fig 8.B). We introduced the low-temperature inducible promoter pCspA and linked it to the red fluorescent protein to assess its function (Fig 8.A). The results showed that under low-temperature conditions, the OD600 value of DH5α reached only 0.5 within 12 hours, while at 37°C, it reached 2.3. However, the Fluorescence/OD600 value at low temperature reached 278, significantly higher than the value at 37°C (Fig 8.C). This indicates that the low-temperature inducible promoter CspA can effectively induce the expression of the red fluorescent protein under low-temperature conditions.

We obtained the sequences for two types of lysozymes, T4 Holin and T4 Lysozyme, from the T4 bacteriophage via the NCBI website and amplified them through PCR (Fig 6.B). Next, we genetically connected the arabinose operon pBAD and the low-temperature inducible promoter pCspA to the T4 Holin and T4 Lysozyme sequences. We then validated the survival status of the bacteria under the induction of both promoters. The results showed that in the presence of 0.5 mM/L arabinose, the OD600 of the culture significantly declined after 5 hours, approaching zero by the 20th hour (Fig 9.A). In contrast, under low-temperature conditions at 16°C, the engineered strains containing the lysozyme sequence maintained a low density, with the OD600 remaining around 0.3 (Fig 9.B).

In summary, our team successfully constructed a suicide system using pBAD and pCspA as promoters, with T4 Holin and T4 Lysozyme sequences as suicide genes, and effectively validated its functionality in E. coli DH5α.

Since our ultimate goal is to create a probiotic, we needed to select components suitable for E. coli Nissle 1917 (EcN). By searching the iGEM parts library, we identified a suitable promoter (BBa_J23100), ribosome binding site (BBa_B0034), and terminator (BBa_B0015) for EcN, and created a gene circuit diagram incorporating both the fat-reduction and muscle-building systems (Fig 10.A). Additionally, we chose the EcN-specific pSB1A3 as the vector and illustrated the plasmid map (Fig 10.B).

(B) Plasmid map suitable for EcN

We assessed the expression of GLP-1 in the engineered EcN. The ELISA results indicated that the engineered EcN successfully expressed GLP-1 and, with the help of PelB, secreted it extracellularly (Fig 11.A). Additionally, we evaluated whether the engineered strain affected the normal growth of EcN. The results showed that during the first 20 hours, the OD600 of the engineered EcN grew at a slower rate compared to the wild type, but after 20 hours, the OD600 of the engineered strain approached that of the wild type (Fig 11.B).

Since our engineered strain needs to function in the gut, we simulated the intestinal microenvironment to test its stability. First, we evaluated the antioxidant capacity of the strain using the DPPH scavenging assay (DPPH radical scavenging capacity kit, A153-1-1, Nanjing Jianchen). Next, we conducted acid resistance tests by resuspending the strains at different pH levels and measuring OD600. Finally, we assessed the bile salt tolerance by adding equal amounts of bacteria to LB media containing different concentrations of bile salts (0.0%–0.5%) and incubating at 37°C for 4 hours, followed by OD600 measurement. The results showed that the antioxidant capacity (Fig 11.C), pH tolerance (Fig 11.D), and bile salt tolerance (Fig 11.E) of the engineered strain did not significantly differ from those of the wild-type EcN.

In summary, our team successfully integrated the fat-reduction system, muscle-building system, and suicide system into the pSB1A3 vector and transformed it into EcN. Furthermore, we validated the extracellular expression of GLP-1 and the growth characteristics of the engineered EcN. Finally, we simulated the intestinal microenvironment and demonstrated that the engineered EcN maintained comparable antioxidant capacity, acid tolerance, and bile salt tolerance to the wild type, indicating its potential to survive in the gut.

Our team successfully constructed and validated expression systems for GLP-1 and Bimagrumab, designed a high-affinity DPP-4 inhibitor, and established an effective cell suicide system. By transforming the fat-reduction, muscle-building, and suicide systems into E. coli Nissle 1917, we confirmed the engineered strain's survival in a simulated intestinal environment, ensuring its efficacy and safety. Additionally, the DPP-4 inhibitor designed using deep learning models demonstrated superior binding energy compared to GLP-1, highlighting the rationality and promise of this design. Experimental results indicate that the engineered strain can efficiently express and secrete target proteins while maintaining growth, showcasing its potential applications in medicine and biotechnology, thus providing a solid foundation for future research.

[1] D.A. Case, H.M. Aktulga, K. Belfon, D.S. Cerutti, G.A. Cisneros, V.W.D. Cruz eiro, N. Forouzesh, T.J. Giese, A.W. Götz, H. Gohlke, S. Izadi, K. Kasavajhala, M.C. Kaymak, E. King, T. Kurtzman, T.-S. Lee, P. Li, J. Liu, T. Luchko, R. Luo, M. Manathunga, M.R. Machado, H.M. Nguyen, K.A. O’Hearn, A.V. Onufriev, F. Pan, S. Pantano, R. Qi, A. Rahnamoun, A. Risheh, S. Schott-Verdugo, A. Shajan, J. Swails, J. Wang, H. Wei, X. Wu, Y. Wu, S. Zhang, S. Zhao, Q. Zhu, T.E. Cheatham III, D.R. Roe, A. Roitberg, C. Simmerling, D.M. York, M.C. Nagan*, and K.M. Merz Jr.* AmberTools. J. Chem. Inf. Model. 63, 6183-6191 (2023).

[2] D.A. Case, H.M. Aktulga, K. Belfon, I.Y. Ben-Shalom, J.T. Berryman, S.R. Brozell, D.S. Cerutti, T.E. Cheatham, III, G.A. Cisneros, V.W.D. Cruzeiro, T.A. Darden, N. Forouzesh, M. Ghazimirsaeed, G. Giambaşu, T. Giese, M.K. Gilson, H. Gohlke, A.W. Goetz, J. Harris, Z. Huang, S. Izadi, S.A. Izmailov, K. Kasavajhala, M.C. Kaymak, A. Kovalenko, T. Kurtzman, T.S. Lee, P. Li, Z. Li, C. Lin, J. Liu, T. Luchko, R. Luo, M. Machado, M. Manathunga, K.M. Merz, Y. Miao, O. Mikhailovskii, G. Monard, H. Nguyen, K.A. O'Hearn, A. Onufriev, F. Pan, S. Pantano, A. Rahnamoun, D.R. Roe, A. Roitberg, C. Sagui, S. Schott-Verdugo, A. Shajan, J. Shen, C.L. Simmerling, N.R. Skrynnikov, J. Smith, J. Swails, R.C. Walker, J. Wang, J. Wang, X. Wu, Y. Wu, Y. Xiong, Y. Xue, D.M. York, C. Zhao, Q. Zhu, and P.A. Kollman (2024), Amber 2024, University of California, San Francisco.

[3] Hecht A, Endy D, Salit M, et al. When wavelengths collide: bias in cell abundance measurements due to expressed fluorescent proteins[J]. ACS synthetic biology, 2016, 5(9): 1024-1027.