MODEL

Overview

Preliminary testing of our DhdR-DhdO system showed unexpected behavior with minimal repression, which motivated us to carry out modeling studies of the protein structure and interactions. To validate the protein’s activity and to verify that protein dimerization and DNA binding could occur as expected, we completed three computational modeling analyses, which will allow us to move forward with targeted refinements in our wet lab work:

  1. Structural comparison of DhdR in monomeric and homodimeric forms, with and without the FLAG-NLS sequence attached to the protein.Using relaxed AlphaFold2 predictions, we showed that there were differences in the predicted structures across the two tools. Furthermore, by comparing predicted structures both with and without the attached FLAG/NLS sequences, we showed that there are possible conformational changes resulting from this sequence modification that may impact the ability of a DhdR homodimer to effectively bind to the DNA dhdO binding site as desired.
  2. Sequence homology modeling to compare structures of related transcription factors bound to DNA.Using an available PDB structure of a GntR family transcription factor bound to DNA and SWISS-Model-generated predicted structures, we were able to better understand the specific residues involved in DhdR dimer binding to the dhdO binding site sequence.
  3. Docking study of D-2HG into the predicted structure of DhdR homodimer.Using the SWISS-Dock tool, we generated a list of potential docking sites for the D-2HG ligand, obtaining potential insight into the mechanism by which D-2HG binding causes DhdR dissociation from its binding site.

Analysis of Monomeric and Homodimeric Predicted Structures

Because we were working with a newly described transcription factor, DhdR, we wanted to perform structural predictions in order to visualize the shape of the protein. To do this, we used the AlphaFold2 tool, which allowed us to generate predicted structures based on the protein sequence [1]. These predictions were compared to the known protein structure for a related GntR transcription factor [2], as well as to homology-modeled structures generated using SWISS-Model (Figure 1) [3].

Figure 1. SWISS-Model alignment results for the DhdR dimer sequence with PDB structure 6AZ6, which was later validated to be a related GntR family transcription factor protein

The results of this analysis validated the AlphaFold2 generated structures, suggesting that they were reasonably close to the expected structures (Figure 2).

Figure 2. Comparison of the DNA binding domain conformations between relaxed AlphaFold2 predictions (green), the SWISS-Model homology-based prediction (blue), and the known GntR structure from PDB (green, PDB code 6AZ6)

This served as initial confirmation for our ability to use AlphaFold2 as a tool to probe the structure of DhdR, both with and without sequence modifications. Next, we determined whether the inclusion of the NLS and FLAG tags to the N- and C-termini of the protein would impact the dimerized structure of DhdR. Using AlphaFold2 and PyRosetta, predicted dimer structures were generated and relaxed (Figure 3).

Figure 3. Relaxed lphaFold2 predicted structures of the modified DhdR homodimer, with the NLS and FLAG tags attached to the C-terminus (left) and the N-terminus (right)

The visualized structures show that addition of the NLS-FLAG sequence to the N-terminal domain (NTD) significantly impacts the conformation of this portion of the protein, which is problematic given that the NTD is the DNA-interacting domain of the homodimer. Indeed, the conformation appears to completely block the region where DNA would normally form its required binding interactions, suggesting that addition of the FLAG-NLS sequence to the N-terminal end of the protein may prevent DNA binding and the DhdR transcription factor from actually carrying out its functions. This sheds light on one potential factor that may have caused our proof of concept experiment to not function as expected.

AlphaFold2 was run using the publicly available ColabFold notebook, available here. [4] PyRosetta for structure relaxation was run using this Colab notebook, which was modified from the original PyRosetta Jupyter notebooks available here. [5]

Comparison to Known DNA-bound Structures and Validation of Binding Interactions

Based on the identified structure modifications that were generated in the previous portion of our modeling study, we proceeded to determine possible DNA binding sites within the protein structure. With this information, we would be able to better determine how the changes caused by the FLAG tag may have impacted the function of the transcription factor. To do this, we consulted existing literature on GntR transcription factors, which characterized the presence of canonical helix-turn-helix (HTH) binding domains within the N-terminal portion of the protein. [6] We then validated the presence of this motif within our structure, shedding light on the potential DNA-binding interactions occurring in our DhdR system (Figure 4).

Figure 4. An annotated view of the HTH motif in a related GntR transcription factor, showing relevant DNA-interacting residues in a previously characterized protein structure (PDB: 1HW2) (left, [6]) and in the AlphaFold2 predicted protein structure of DhdR (right)

Given the presence of conserved positively-charged arginine residues at positions 35, 49, and 64, along with consensus modeling to previously elucidated structures, we can conclude that this N-terminal domain is likely the DNA binding domain and is less engineerable. This also provides a scaffold to engineer the DhdR transcription factor to bind novel sequences through mutational or directed evolution studies, both of which are beyond the current scope of the project.

In order to visualize the possible binding conformation of the transcription factor, we found a GntR family transcription factor that had a DNA-bound conformation solved and deposited (PDB code: 1HW2) [7]. Then, we aligned and superimposed the two structures in ChimeraX, allowing us to create a simple predictive structure of how the DhdR transcription factor would bind to a DNA sequence (Figure 5).

Figure 5. A predictive structure of how the DhdR transcription factor would bind to DNA based on structural comparison and alignment.

Further experiments are necessary to validate the specific DNA binding domain and residues responsible for DhdR binding to DhdO sequences. These may include evolution studies or binding assays using surface plasmon resonance (SPR) to determine dissociation constants. However, we have high confidence that N-terminal modifications to DhdR severely impact our system’s function.

D-2HG Docking Study

Finally, after characterizing the DNA binding domain of the DhdR homodimer, we wanted to determine how the introduction of D-2HG to the system could potentially lead to disruptions of DhdR binding to DNA as a part of its allosteric activity. To do this, we used the SWISSDock tool, which generated the following likely binding configurations (Figure 6).

Figure 6. Predicted structures generated from the docking of D-2HG (red-green/brown spheres) into the DhdR heterodimer, showing the most energetically favorable (left) and second most energetically favorable (right) docking configurations.

Overall, the predicted location of the D-2HG ligand in the DNA binding domain suggests that the presence of the ligand may inhibit DhdR binding through disruption of binding pocket interactions. However, further experimentation would likely be required to determine the exact mechanism by which this occurs, whether by preventing DhdR dimerization or by blocking favorable interactions between the transcription factor and the binding site. This provided more evidence that N-terminal fusions would significantly impact DhdR function, both in its binding activity and allosteric binding of D-2HG, the oncometabolite of interest to our study.

Conclusion

The results of the different structural prediction assays provide insight into the DhdR transcription factor that forms the backbone of our reporter system. We were able to better understand the mechanisms by which DhdR both associates and dissociates from its binding site. Furthermore, we gained potential insight into the impacts that our sequence modifications may have produced on the structure of the DhdR homodimer, revealing how its binding ability could have possibly been eliminated through the addition of the NLS and FLAG tag to the N-terminal end. Overall, the results reveal the power that modeling studies can offer, validating them as a meaningful complement to traditional wet lab experimentation and showing their utility in revealing future avenues of exploration.

References

  1. Jumper, J., Evans, R., Pritzel, A., Green, T., Figurnov, M., Ronneberger, O., Tunyasuvunakool, K., Bates, R., Žídek, A., Potapenko, A., Bridgland, A., Meyer, C., Kohl, S., Ballard, A. J., Cowie, A., Romera-Paredes, B., Nikolov, S., Jain, R., Adler, J., Back, T., … Hassabis, D. (2021). Highly accurate protein structure prediction with AlphaFold. Nature, 596(7873), 583–589. https://doi.org/10.1038/s41586-021-03819-2
  2. Little, M. S., Pellock, S. J., Walton, W. G., Tripathy, A., & Redinbo, M. R. (2018). Structural basis for the regulation of β-glucuronidase expression by human gut Enterobacteriaceae. Proceedings of the National Academy of Sciences of the United States of America, 115(2), E152–E161. https://doi.org/10.1073/pnas.1716241115
  3. Waterhouse, A., Bertoni, M., Bienert, S., Studer, G., Tauriello, G., Gumienny, R., Heer, F. T., de Beer, T., Rempfer, C., Bordoli, L., Lepore, R., & Schwede, T. (2018). SWISS-MODEL: homology modelling of protein structures and complexes. Nucleic acids research, 46(W1), W296–W303. https://doi.org/10.1093/nar/gky427
  4. Mirdita, M., Schütze, K., Moriwaki, Y., Heo, L., Ovchinnikov, S., & Steinegger, M. (2022). ColabFold: making protein folding accessible to all. Nature methods, 19(6), 679–682. https://doi.org/10.1038/s41592-022-01488-1
  5. Le, K. H., Adolf-Bryfogle, J., Klima, J. C., Lyskov, S., Labonte, J., Bertolani, S., Burman, S., Leaver-Fay, A., Weitzner, B., Maguire, J., Rangan, R., Adrianowycz, M. A., Alford, R. F., Adal, A., Nance, M. L., Wu, Y., Willis, J., Kulp, D. W., Das, R., Dunbrack, R. L., Jr, … Gray, J. J. (2021). PyRosetta Jupyter Notebooks Teach Biomolecular Structure Prediction and Design. Biophysicist (Rockville, Md.), 2(1), 108–122. https://doi.org/10.35459/tbp.2019.000147
  6. Zheng, M., Cooper, D. R., Grossoehme, N. E., Yu, M., Hung, L. W., Cieslik, M., Derewenda, U., Lesley, S. A., Wilson, I. A., Giedroc, D. P., & Derewenda, Z. S. (2009). Structure of Thermotoga maritima TM0439: implications for the mechanism of bacterial GntR transcription regulators with Zn2+-binding FCD domains. Acta crystallographica. Section D, Biological crystallography, 65(Pt 4), 356–365. https://doi.org/10.1107/S0907444909004727
  7. Xu, Y., Heath, R. J., Li, Z., Rock, C. O., & White, S. W. (2001). The FadR.DNA complex. Transcriptional control of fatty acid metabolism in Escherichia coli. The Journal of biological chemistry, 276(20), 17373–17379. https://doi.org/10.1074/jbc.M100195200