Genome Scale Metabolic Model (GEM)

To optimize the in-silico production of aerobactin, our first design objective was to identify the environment and genetic conditions that maximize intracellular citrate availability - the initial building block of our siderophore. We selected E. coli as the chassis for heterologous expression of its well characterized metabolism and availability of validated genome-scale models.

We used the iML1515 model, which includes 1516 genes and over 2700 metabolic reactions, as the foundation for our simulations. Our design variables included:

• Carbon sources (glucose, acetate, glycerol, succinate),
• pH and temperature ranges,
• Nitrogen source (NH4+ or inorganic phosphate),
• knockout of key reactions such as ICDHyr (isocitrate dehydrogenase)[1].

We hypothesized that low pH, alternative carbon sources, and specific knockouts could increase citrate accumulation at the expense of biomass growth, a desired trade-off for metabolic overproduction.

Protein Sequences & Structure

At the core of this project lies the set of protein sequences essential to address our research questions. Most notably, these include the four proteins responsible for siderophore biosynthesize, as well as the sequences of the TBDT and TonB. However, limited information was available in the literature regarding siderophore-associated proteins in K. pneumonia. Our initial objective was therefore to gather as much structural information as possible, particularly on the TBDT, to ensure that our prediction closely reflects the naive configuration observed in nature.

To achieve this, one of our first steps was to translate the extensive knowledge available for E. coli to our organism of interest. Leveraging the high structural conservation of TBDTs and other related proteins, we used BLASTp [2] searches restricted to K. pneumonia (taxid:573) to identify homologous sequences. As for aerobactin, our siderophore of interest, its Simplified Molecular Input Line Entry System (SMILE) code was available on PubChem [3]. Additionally, we sought information on potential post-translational modifications (PTMs), disulfide bridges, and functional domains within the TBDT, provind key insights to guide subsequent analyses throughout the iGEM project cycle.

Biomodel Composition

In order to model the TBDT within a bacterial membrane, several additional questions needed to be addressed to ensure the accuracy of the system. Having already identified the TBDT’s sequence and structure, the next considerations concerned the composition of the bacterial membrane and the surrounding solution. Facing the same challenge of limited information available for K. pneumonia, we opted to use a generic Gram-negative membrane composition [4]. For its part, the solution was supplemented with Na+ and Cl- ions, as these are among the most abundant ions found in biological environments. Finally, the CHARMM-GUI [5] platform was selected to construct our systems, given its extensive set of tools and flexibility for various molecular dynamic modelling approaches.

1. pBluescript “backbone” Extraction [6]: We utilized an extraction technique to extract the pBluescript plasmid from a liquid bacterial culture using the Quiagen kit: QUIAprep spin miniprep kit.

2. pBluescript Digestion and Linearization [6]: We used the HindIII-HF enzyme to digest the plasmid and obtain its linear version that will eventually be used for the Gibson Assembly.

3. pBluescript Cleaning [6]: After the digestion with the HindIII-HF enzyme, we cleaned up the DNA with the “BioBasic: EZ-10 Spin Column DNA Cleanup Miniprep Kit” to purify the DNA and remove unwanted components (contaminants).

4. pBluescript Electrophoresis [6]: After the clean-up, we have put the DNA on a 0.8% agarose gel to observe if we had isolated the pBluescript plasmid in its linear form.

5. Oligo Resuspension: We resuspended our gBlocks oligos, along with their forward and reverse primers, to a specific concentration.

6. gBlock PCR Amplification [6.a]: We then used our resuspended oligos and amplified our gBlock to add the homology arms necessary for Gibson assembly.

7. PCR Electrophoresis [6]: Using the same agarose gel recipe from before, we have put our gBlock amplification on the gel to see if it had worked.

8. PCR Touchdown: We needed to upgrade our specificity for the third gBlock, to achieve that, we’ve tried the PCR Touchdown method, meaning that the temperature is higher in the first 9 cycles before dropping at a rate of minus 1 degree Celsius per cycle all the way to the optimal annealing temperature of the primers.

9. Gibson Assembly [6.b]: Our goal was to assemble all of our gBlocks in the pBluescript plasmid, and for that, we’ve used the Gibson assembly kit and method.

10. Chemical Transformation of Competent Cells [6.c]: Using the E. coli DH5α, we have transformed our competent cells with the Gibson assembly product before putting our transformation product on a petri dish culture. After 24h at 37℃, the clones that worked were white, and those that didn’t transform were blue (from the pBluescript plasmid).

11. Plasmid Extraction [6]: We extracted the plasmid from positive mutants using the “QIAgen - QIAprep spin miniprep kit” to be digested afterward.

12. Digestion of Assembled Plasmid [6]: After the plasmid extraction, we digested our plasmid construction to see if the Gibson assembly worked, using the BamI-HF and the EcoRI enzymes.

13. Subsequent Amplification: Before proceeding to the next step of our project, we needed the Gibson assembly to work, and for that, we needed the third gBlock to amplify in PCR. We then tried the inverse touchdown PCR method on the gBlock and the stitching PCR method.

14. Competent Cells [6.c]: After several cycles of troubleshooting, we finally tried to make our own competent cells to then assemble our gBlocks.

GEM

The computational model was built and parameterized using the COBRApy Python package, which enables constraint-based reconstruction and analysis of genome-scale networks. We implemented flux balance analysis (FBA) to calculate steady-state flux distributions maximising either growth or citrate export to assess pathway flexibility [7].

We simulated multiple conditions by systematically varying media composition and environmental parameters, applying gene knockouts, and recording resulting fluxes for:

• Specific growth rate (h⁻¹),
• Citrate export flux (mmol/gDW/h),
• Siderophore secretion rate (mmol/gDW/h),

Structure Predictions

Prediction of the aerobactin siderophore was performed using Chai-1 [8], which accepts SMILEs strings as valid ligand inputs. The TBDT structure was predicted using both the Chai-1 and Alphafold3 [9] servers . Since both accept the one letter code of amino acid, it permitted us to assert a sort of consensus on the resulting prediction. Residue-level predictions of membrane embedding were obtained using the OPM web server [10].

Molecular Dynamics

Using the information gathered during the Design phase, along with the predicted 3D structure of TBDT, we constructed a model representing the in vivo composition of a Gram-negative bacteria membrane with the TBDT embedded at its core. The system was further surrounded by a sufficiently large water box to minimize potential interaction artifacts and supplemented with the appropriate ions. CHARMM-GUI provided the system, its parameters, and the optimized force field, enabling a high-quality and physically realistic model for subsequent simulations.

*From this point forward, the test and learn section for the wet lab are hypothesis about what the next steps should be about and what we could learn from them.*

With our assembled plasmid carrying the three gBlocks, and within them all the necessary genes to produce and export the aerobactin siderophore, the next step is to put the bacterial strain in a rich citrate environment to induce the production of the aerobactin. Within the culture, we’ll need to detect the presence of siderophore using Mass-Spectrometry (MS) and other chemical tests for the hydroxamate siderophore-type, like the Csaky Test that depends on the formation of NO_2- via the oxidation of hydroxylamine (NH₂OH) by an iodine solution and the formation of colored dye, the Ferric Perchlorate Assay that identify, by the appearance of an orange-red or purple color, the hydroxamate-type siderophores when putting Fe³⁺ ion and low pH perchloric acid in the test tube, or the Tetrazolium Test that only uses a few drops of NaOH and the sample to the tetrazolium salt creating a deep-red color [11]. Eventually, we’ll need to conjugate our siderophore to a gold nanoparticle and try to find a way to activate the complex aerobactin-Fe³⁺-AuNP inside K. pneumoniae.

GEM

Simulations showed that maximum citrate secretion (18.17 mmol/gDW/h) occurred under acidic conditions (pH ≈ 5) with glycerol as the carbon source and Pi as the nitrogen source. In contrast, the wild-type (WT) model consistently achieved the highest growth rate (~1.85 h⁻¹) under neutral pH with ammonium, confirming a clear metabolic trade-off between citrate production and biomass formation.

Molecular Dynamics

We used NAMD [11] to simulate three independent replicates of the TBDT system with the domain, as well as a second set of three replicates without the plug domain, each for a minimum duration of 350 ns for each trajectories. Upon completion of the simulations, the membrane systems trajectories were visualized using VMD (Visual Molecular Dynamics) [13]. Finally, most computational analyses were performed using Wordom [14].

Our Wet lab experiments were concluded short because of the unattainable assembled plasmid three gBlocks construction. However, the entire sequence should be revised to know why the third gBlock is so difficult to amplify before trying any PCR methods. We suspect having some more difficulties with the expression of the siderophore and the conjugation with the AuNPs, but nothing unachievable. The attending results should be the successful binding of the siderophore-AuNPs and the formation of the aerobactin-Fe³⁺-AuNP complex that will use the TBDT-TonB system to enter K. pneumoniae. After this step, we should be able to activate the complex with a NIR laser (808nm) that will be absorbed by both AuNPs and K. pneumoniae. The NIR laser will activate the conjugate AuNPs that will, then, vibrate, produce heat and create a cellular lysis.

By then, we suppose that we will have learned if the use of a NIR laser is effective or another method should be used and if the AuNPs are optimal for the binding and if the lysis effect is the one that the Totally [Fe]rocious team search.

GEM

From this first modelling iteration, we learned that citrate accumulation can be enhanced in acidic, glycerol-rich environments but at the cost of growth. This insight indicates that two-stage fermentation strategy, first growing cells in neutral conditions, then switching to acidic media for production. It would likely optimize total aerobactin yield. We also learned that maintaining siderophore flux despite ICDHyr deletion suggests possible redundancy in the citrate synthesis routes, warranting further exploration via flux variability analysis and double knockouts.

Molecular Dynamics

The dry lab investigation of the plug domain’s contribution to the TBDT structure provided valuable insight at the residue level, clarifying how its presence influences the β-barrel and overall protein stability. However, a more rigorous experimental design could have yielded a stronger dataset to analyse. Initiating the calculation earlier on, for instance, would have enabled the generation of longer trajectories, allowing for the observation of more stable protein conformations, even within the limited timeframe of an iGEM cycle. Similarly, increasing the number of replicates to five would provide us a clearer picture of the interactions governing the TBDT dynamics and its plug domain.

The scope of our research was also constrained by the absence of parametrization for the siderophore-Fe³⁺ complex. Future work could focus on developing optimized parameters for various siderophores in complexes with Fe³⁺, enabling the study of a broader range of bacterial systems. Incorporating these newly parametrized complexes into simulations would provide a deeper insight into the molecular interactions governing siderophore translocation through the TBDT. Ultimately, this would improve our understanding of the siderophore transport across the outer membrane and could extend to characterizing their passage through the inner membrane, across an ABC transporter.