Staple Subunit: rGCN4

rGCN4 is an engineered, reverse, variant of the yeast transcription factor GCN4, featuring a basic region and a leucine zipper dimerization domain. We used rGCN4 to study DNA-binding kinetics in our "Mini staples" that bring two DNA target sites into proximity by binding them simultaneously.

Figure 1: How our part collection can be used to engineer new staples

While synthetic biology has in the past focused on engineering the genomic sequence of organisms, the 3D spatial organization of DNA is well-known to be an important layer of information encoding in particular in eukaryotes, playing a crucial role in gene regulation and hence cell fate, disease development, evolution, and more. However, tools to precisely manipulate and control the genomic spatial architecture are limited, hampering the exploration of 3D genome engineering in synthetic biology. We - the iGEM Team Heidelberg 2024 - have developed PICasSO, a powerful molecular toolbox for rationally engineering genome 3D architectures in living cells, based on various DNA-binding proteins.

The PICasSO part collection offers a comprehensive, modular platform for precise manipulation and re-programming of DNA-DNA interactions using engineered "protein staples" in living cells. This enables researchers to recreate naturally occurring alterations of 3D genomic interactions, such as enhancer hijacking in cancer, or to design entirely new spatial architectures for artificial gene regulation and cell function control. Specifically, the fusion of two DNA binding proteins enables to artificially bring otherwise distant genomic loci into spatial proximity. To unlock the system's full potential, we introduce versatile chimeric CRISPR/Cas complexes, connected either at the protein or - in the case of CRISPR/Cas-based DNA binding moieties - the guide RNA level. These complexes are referred to as protein- or Cas staples, respectively. Beyond its versatility with regard to the staple constructs themselves, PICasSO includes robust assay systems to support the engineering, optimization, and testing of new staples in vitro and in vivo. Notably, the PICasSO toolbox was developed in a design-build-test-learn engineering cycle closely intertwining wet lab experiments and computational modeling and iterated several times, yielding a collection of well-functioning and -characterized parts.

At its heart, the PICasSO part collection consists of three categories.
(i) Our DNA-binding proteins include our finalized Cas staple experimentally validated using an artificial "enhancer hijacking" system as well as "half staples" that can be combined by scientists to compose entirely new Cas staples in the future. We also include our Simple staples comprised of particularly small, simple and robust DNA binding domains well-known to the synthetic biology community, which serve as controls for successful stapling and can be further engineered to create alternative, simpler, and more compact staples.
(ii) As functional elements, we list additional parts that enhance and expand the functionality of our Cas and Basic staples. These consist of staples dependent on cleavable peptide linkers targeted by cancer-specific proteases or inteins that allow condition-specific, dynamic stapling in vivo. We also include several engineered parts that enable the efficient delivery of PICasSO's constructs into target cells, including mammalian cells, with our new interkingdom conjugation system.
(iii) As the final category of our collection, we provide parts that underlie our custom readout systems. These include components of our established FRET-based proximity assay system, enabling users to confirm accurate stapling. Additionally, we offer a complementary, application-oriented testing system based on a luciferase reporter, which allows for straightforward experimental assessment of functional enhancer hijacking events in mammalian cells.

The following table gives a comprehensive overview of all parts in our PICasSO toolbox. The highlighted parts showed exceptional performance as described on our iGEM wiki and can serve as a reference. The other parts in the collection are versatile building blocks designed to provide future iGEMers with the flexibility to engineer their own custom Cas staples, enabling further optimization and innovation in the new field of 3D genome engineering.

Our part collection includes:

1. Sequence overview

2. Usage and Biology

rGCN4 is an engineered variant of the yeast GCN4 transcription factor (BBa_K5237007). In contrast to GCN4 that binds the CRE target sequence with the N-terminal, rGCN4 binds the modified INVii (5' GTCAtaTGAC 3') DNA target sequence with the C-terminal region. rGCN4 is an engineered variant of the yeast GCN4 transcription factor (BBa_K5237007). In our project, we used rGCN4 to study DNA-binding kinetics in our "Mini staples" (BBa_K5237009) that bring two DNA target sites into proximity by binding them simultaneously.

3. Assembly and Part Evolution

The rGCN4 amino acid sequence was taken from literature (Hollenbeck & Oakley 1999) and codon optimized for E. coli. A FLAG-tag (DYKDDDDK) was added to the N-terminus for protein purification. The FLAG-tag can be cleaved off using an Enterokinase, if necessary. Expression of FLAG-GCN4 was done under an IPTG inducible T7 promoter in E. coli BL21 (DE3) cells.

4. Results

4.1 Protein Expression and Purification

The FLAG-rGCN4 protein could be readily expressed in E. coli. The protein was purified using an anti-FLAG resin. Fractions taken during purification were analyzed by SDS-PAGE and the protein concentration of the eluted protein determined with a lowry protein assay. A yield of 3.4 mg/mL was obtained, corresponding to 422 µM of monomeric FLAG-GCN4.

Figure 2: SDS-PAGE Analysis of FLAG-GCN4 Purification Fractions analysed are the raw lysate, flow through and eluate. Depicted is GCN4 (this part), rGCN4 (BBa_K5237008), and bGCN4 (BBa_K5237009). Protein size is indicated next to construct name and purified band with protein of interest highlighted by a red box.

4.2 Electrophoretic Mobility Shift Assay

Figure 3: Overview Image of Electrophoretic Mobility Shift Assay (EMSA)

The Electrophoretic mobility shift assay (EMSA) is a widely adopted method used to study DNA-protein interactions. EMSA functions on the basis that nucleic acids bound to proteins have reduced electrophoretic mobility, compared to their counterpart. (Hellman & Fried, 2007). Mobility-shift assays can both be used to qualitatively assess DNA binding capabilities or quantitatively to determine binding stoichiometry and kinetics such as the apparent dissociation constant (K_d) (Fried, 1989).

To analyze the binding DNA affinity an EMSA was performed, in which rGCN4 was incubated in binding buffer with a 20 bp DNA probe containing the INVii rGCN4 binding sequence (5' GTCAtaTGAC 3') until equilibration. Subsequently the formed protein-DNA complexes were loaded on a native PAGE. Afterwards the DNA bands were stained with SYBR-safe.
To further analyze DNA binding, quantitative shift assays were performed for GCN4 and rGCN4 (BBa_K5237008). 0.5 µM DNA were incubated with varying concentrations of protein until equilibration. After electrophoresis, bands were stained with SYBR-Safe and quantified based on pixel intensity. The obtained values were fitted to equation 1, describing formation of a 2:1 protein-DNA complex:

Θ_app = Θ_min + (Θ_max - Θ_min) × (K_a² [L]_tot²) / (1 + K_a² [L]_tot²) Equation 1

Here [L]_tot describes the total protein monomer concentration, K_a corresponds to the apparent monomeric equilibration constant. The Θ_min/max values are the experimentally determined site saturation values (For this experiment 0 and 1 were chosen for min and max respectively).

Figure 4: Quantitative EMSA. Quantitative assessment of binding affinity for GCN4 and rGCN4. Proteins of different concentrations were incubated with 0.5 µM DNA until equilibrium and fraction bound analyzed, after gel electrophoresis, by dividing pixel intensity of bound fraction with pixel intensity of bound and unbound fraction. At least three separate measurements were conducted for each data point. Values are presented as mean +/- SD.

rGCN4 binds to INVii with an apparent dissociation constant K_D of K_d of (0.298 ± 0.030) × 10^-6 M, which is almost identical to the GCN4 dissociation constant of (0.293 ± 0.033) × 10^-6 M Comparing them to literature values, our dissociation constants are approximately a factor 10 higher then those described in literature ((9 ± 6) × 10^-8 M for GCN4 and (2.9 ± 0.8) × 10^-8 M for rGCN4) (Hollenbeck et al., 2001). The differences could be due to the lower sensitivity of SYBR-Safe staining compared to radio-labeled oligos. Most likely, the protein concentration was miscalculated due to the presence of additional (lower intensity) bands in the SDS-PAGE analysis, indicating co-purification of small amounts of unspecific proteins.

The FLAG-tag fusion to the N-terminus of proteins could potentially decrease binding affinity, likely due to steric hindrance affecting the interaction with DNA. Interestingly, the differences in binding affinity between GCN4 and rGCN4 appear negligible. Since GCN4 binds to DNA via its N-terminus and rGCN4 binds C-terminally, the FLAG-tag likely does not directly influence DNA binding. However, it may influence the dimerization of the proteins, which is necessary for DNA binding. To further investigate this, the FLAG-tag can be cleaved using an enterokinase and potential changes in binding affinity analyzed. Furthermore, coiled coil formation, and the amount of dimeric and monomeric proteins could be further analyzed with circular dichroism spectroscopy (Greenfield, 2006).

4.3 In Silico Characterization Using DaVinci

Figure 5: Molecular Dynamics Simulation of GCN4

We developed DaVinci, an in silico model, for rapid engineering and optimization of our PICasSO system. DaVinci serves as a digital twin to PICasSO, analyzing the forces acting on the system, refining experimental parameters, and identifying optimal interactions between protein staples and target DNA. The model was calibrated using literature data and experimental affinity results from rGCN4 EMSA assays with purified proteins.
DaVinci operates in three phases: static structure prediction, all-atom dynamics simulation, and long-range DNA dynamics simulation. We applied the first two phases to our components, allowing us to characterize the structure and dynamics of the DNA-binding interactions.
For our bivalent DNA-binding Mini staple (BBa_K5237009), consisting of GCN4 fused via a GSG-linker to rGCN4 (BBa_K5237008), we predicted the structure and binding affinity and tested various linker options. We evaluated the flexibility and rigidity of the constructs using pLDDT values from the predictions. Flexible linkers, like ('GGGGS')n, and rigid linkers, like ('EAAAK')n, were assessed (Arai et al., 2001). Predictions were colored by pLDDT scores, providing insights into chain rigidity (Akdel et al., 2022; Guo et al., 2022). Construct C (Fig. 5) was tested in the wet lab as part of BBa_K5237009, but it failed to bind DNA due to excessive rigidity, which inhibited subunit dimerization.

Figure 6: Variation of Linkers Connecting Our Mini Staples. Panels A (BBa_K5237007) and B (BBa_K5237008) show orientations of the leucine zipper, each bound to DNA. Panels C to I display linker variations colored by their pLDDT confidence score, which serves as a surrogate for chain flexibility (Akdel et al., 2022). Note that panels H and I are not bound to the second DNA strand. All structures were predicted using the AlphaFold server (Google DeepMind, 2024).

4.4 Comparison of ΔG Calculated by Wet- and Dry Lab

The apparent dissociation constant K_D, calculated from the EMSA experiments, was used to calculate the Gibbs free energy (ΔG) of the rGCN4-DNA interaction.

G = RT ln(K_D) Equation 2

Our DaVinci model calculated ΔG by using an all-atom molecular dynamics simulation to compute the dissociation of DNA from its bound state, capturing the transition between states by constructing a custom weight function to sample the configuration space in a Monte Carlo scheme (Frenkel & Smit, 2023).This approach enables precise estimation of the energy landscape around the bound state and, furthermore, the calculation of the thermodynamic quantities.

Comparing the wet lab with the dry lab results showed some discrepancies. The calculated ΔG based on the measured K_D of the wet lab experiments was 9.256 ± 10.676 kCal mol^-1 which compares well to literature with an estimate of G = 10.7 ± 11.48 kCal mol-1 (Jessica J. Hollenbeck et al., 2001). Even though the dry lab calculation is within the 1.6 deviation of the experimental data, this is due to a large error in the calculation. Moreover, when visualizing the trajectory, DNA strand separation was observed, suggesting an unnatural event. Thus, this calculation will require further investigation to improve accuracy

Figure 7: Visualized Pulling Simulation All-Atom Pulling Simulation

5. References

Akdel, M., Pires, D. E. V., Pardo, E. P., Janes, J., Zalevsky, A. O., Meszaros, B., Bryant, P., Good, L. L., Laskowski, R. A., Pozzati, G., Shenoy, A., Zhu, W., Kundrotas, P., Serra, V. R., Rodrigues, C. H. M., Dunham, A. S., Burke, D., Borkakoti, N., Velankar, S., … Beltrao, P. (2022). A structural biology community assessment of AlphaFold2 applications. Nat Struct Mol Biol, 29(11), 1056–1067. https://doi.org/10.1038/s41594-022-00849-w

Arai, R., Ueda, H., Kitayama, A., Kamiya, N., & Nagamune, T. (2001). Design of the linkers which effectively separate domains of a bifunctional fusion protein. Protein Engineering, Design and Selection, 14(8), 529–532. https://doi.org/10.1093/protein/14.8.529

Chen, X., Zaro, J. L., & Shen, W.-C. (2013). Fusion protein linkers: Property, design and functionality. Advanced Drug Delivery Reviews, 65(10), 1357–1369. https://doi.org/10.1016/j.addr.2012.09.039

Google DeepMind. (2024). AlphaFold Server. https://alphafoldserver.com/terms

Guo, H.-B., Perminov, A., Bekele, S., Kedziora, G., Farajollahi, S., Varaljay, V., Hinkle, K., Molinero, V., Meister, K., Hung, C., Dennis, P., Kelley-Loughnane, N., & Berry, R. (2022). AlphaFold2 models indicate that protein sequence determines both structure and dynamics. Scientific Reports, 12(1), 10696. https://doi.org/10.1038/s41598-022-14382-9

Hollenbeck, J. J., Gurnon, D. G., Fazio, G. C., Carlson, J. J., & Oakley, M. G. (2001). A GCN4 Variant with a C-Terminal Basic Region Binds to DNA with Wild-Type Affinity. Biochemistry, 40(46), 13833 - 13839.

Hollenbeck, J. J., & Oakley, M. G. (2000). GCN4 Binds with High Affinity to DNA Sequences Containing a Single Consensus Half-Site. Biochemistry, 39(21), 6380 - 6389. https://doi.org/10.1021/bi992705n