Team:USAFA/Model

Template:Usafa

>

Protein Modeling: Dehalogenases

Goal: Investigate crystal structure and PFAS affinity for Dehalogenases from Delftia acidovorans


General Procedure

  1. Estimate Crystal structure of Dehalogenase using the Phyre2 server
  2. Perform protein BLAST of Dehalogenase protein sequence against non-redundant protein database on NCBI.
  3. Perform multiple alignment from top BLAST hits and generate consensus sequence (>95% consensus).
  4. Identify conserved amino acid residues in original dehalogenase protein sequence and visualize with JMOL.
  5. Visually identify potential pockets and perform binding affinity modeling with 1 click docking software.
  6. Visualize PFAS bound to Dehalogenase

Dehalogenase type 1; DeHa1

Modeling DeHa1

The modeling methods used in this procedure are based around a known three-dimensional structure of the enzyme. Given that this protein has not yet been crystalized, the first step was to perform sequence alignment with known structures to estimate the true structure of DeHa1. Phyre2 was used to perform this process which resulted in a wealth of information about the protein of interest, primarily a PDB file output that can be visualized (Figure 1).




Figure 1: PDB model of Dehalogenase 1 from Phyre2 software



Two domains can be identified. A domain containing a central beta-sheet can be observed sitting above a helical domain as seen in Figure 1. In order to generate testable values for future validation methods, a circular dichromism spectra was generated using the PDBMD2CD tool. The generated spectra can be seen below in Figure 2, which generates the % secondary structure content: 46.1% alpha-helix, 11.6% parallel sheets, 9.1% turns, and 28.9% other. In the future the USAFA iGEM team will isolate pure protein and determine secondary structure content by CD spectroscopy to aid in determining if this model accurately depicts reality. Structure elucidation will also be considered should this protein work well in NMR analysis or crystalize cleanly. If this occurs, it would be possible to determine the exact binding location of PFAS, however this can also be modelled for use in this project.




Figure 2: Generated CD spectra for modeled DeHa1.



Determination of Conserved Amino Acids in DeHa1

In order to locate a possible binding center of PFAS to DeHa1, it is critical to identify possible active sites for the enzyme. To perform this, the global environment of the winogradsky column samples were taken into account. Several of the genera present in these columns contain dehalogenases. Therefore, it was of interest to identify regions of high conservation in DeHa1 compared to a consensus sequence of all DeHa1-similar enzymes. Protein BLAST was performed selectively over taxons present in the winogradsky columns compared to DeHa1. Following BLAST, multiple alignment using the COBALT program from NCBI was used to generate a consensus sequence. Once imported to Snapgene, this consensus was modified such that only residues with >95% similarity were considered. After performing a global alignment of the original DeHa1 sequence to the consensus for the winogradsky column genera, several highly conserved residues were obtained. In JMOL these residues were visualized on the original DeHa1 structure (Figure 3). In reference to the rest of the DeHa1 structure, a possible binding pocket can be seen in the interior. Given that this region is also highly conserved amongst Dehalogenase-like peptides, it was hypothesized that PFOA and PFOS may bind in this pocket.


Figure 3: Conserved amino acid residues in winogradsky column dehalogenases visualized in DeHa1. Backbone is shown in cartoon style with side chains colored in standard format using a wireframe style. Non-conserved residues are transparent.



PFAS Binding to Identified Pocket in DeHa1

To investigate the binding affinity of PFOA and PFOS to the conserved pocket in DeHa1, the software 1 click docking by Mcule was utilized to model binding by identification of local energy minima given the protein three dimensional structure and a compound of choice. This software requires an estimated binding location, as it cannot calculate global minima of the ligand and protein. Aspartate 13 was chosen as a binding location in the software, given its generally central location in the conserved region. Below are several figures describing the results of this modeling effort.



The output of MCule is represented in Figure 4A. In this software a stronger affinity is represented by increasingly negative values. The output of PFOA binding had a score of -8.5. The pocket in DeHa1 is clearly observed when a surface mesh is applied to the DeHa-PFOA modeled complex (Figure 4B/C). To visualize the binding of PFOA to DeHa1 in 3-D space, Figure 5 shows PFOA bound to the conserved region identified previously. This shows PFOA localized to an empty region of the conserved region, which overlaps with the identified pocket from MCule. This suggests an active site to be present at this location.



Figure 4: A: Output of 1 Click Docking of PFOA to DeHa1. Binding center was Asp 13 residue. Local energy minima was determined in protein interior. B: Surface mesh applied to docking output. PFOA located in a interior pocket between both domains of DeHa1. C: Interior pocket of DeHa1 with PFOA. Fluorinated tail contained in a hydrophobic region while the carboxylate group is located near a largely polar region of the pocket.




Figure 5: PFOA cound to conserved pocket in winogradsky column dehalogenases visualized in DeHa1. Backbone is shown in cartoon style with side chains colored in standard format using a wireframe style. Non-conserved residues are transparent. Ligand is visualized according to 1 click docking.



Binding affinity of PFOA to DeHa1

In order to scale the effectiveness of PFOA binding to DeHa1, two different approaches were taken. From the protein data bank, stabilized A2A adenosine receptor A2AR-StaR2-bRIL in complex with caffeine (PDB, 5ZMP) was chosen to act as a standard for binding given that caffeine has been shown to bind as an antagonist to the adenosine receptor. The same protein structure was modeled to bind caffeine near the location identified in the known crystal structure. The resultant enzyme-ligand complex had a binding score of -5.7 (Figure 6). Using this receptor/ligand pair as a standard to compare against, it is expected that PFOA binds tightly to the interior pocket of DeHa1.




Figure 6: A: Estimated binding location of caffeine antagonist to adenosine receptor (PDB, 5ZMP) calculated by 1 click docking. B: Crystal structure of adenosine receptor bound to caffeine (PDB, 5ZMP). Caffeine is bound in identical locations with a score of -5.7.



Additionally, an effort to better quantify the binding energies of PFOA to DeHa1 was made in a follow-up analysis. The SwissDock program was used to identify a variety of local energy minima by means of changing the location and orientation of the PFOA ligand around the entire hypothetical DeHa1 structure. The program also calculates an estimated &Delta G value for each binding location/ligand conformation. A lower ΔG corresponds to a higher binding energy, and thus a better location overall for PFOA to bind to the enzyme. For PFOA complexed to DeHa1, 250 unique binding locations which were determined and visualized with the software UCSF Chimera. Upon visualization, approximately 56 general binding zones (clusters) were elucidated (Figure 7A). A majority of these clusters accomodate a large variety of PFOA conformations and exist mostly on the surface of the enzyme. Among these, a single cluster can be seen that is interior to the dehalogenase containing binding modes with the greatest ΔG values. This suggests that the interior binding site is the most favorable for this model, which agrees with the conclusions drawn from the previous, less computationally intensive models. The top three results for PFOA-DeHa1 complexes are shown below in Figure 8.



Figure 7: Binding Energy of PFOA to DeHalogenase 1 as calculated by SwissDock and visualized with a combination of UCSF Chimera and JMOL. A: All identified clusters for PFOA binding. B-D: Models 9, 240, and 241 of PFOA - DeHa1 Complex. Clusters are 1, 45, and 46, respectively. DeHa1 is shown in cartoon style and PFOA is shown in ball and stick style using standard coloration.



Figure 8: Top three binding locations for PFOA in enzyme complex. Model numbers are represented by:'Model.Cluster'. Binding energies are the inverse of ΔG, with Model 46.241 having the strongest binding.




Binding affinity of PFOS to DeHa1

In a similar manner to PFOA, Affinity of PFOS was modeled using JMOL, MCule, SwissDock, and UCSF Chimera to determine the capability of PFOS binding to DeHa1. Initial modeling of PFOS to the internal pocket of DeHa1 was determined by the MCule software using 1 click docking, with similar results that can be seen in Figure 9. PFOS binds to the identical pocket in DeHa1 as PFOA with a slightly increased binding score (-8.7 vs. -8.5 for PFOA), and has a more curved conformation, possibly due to the larger sulfonate group compared to a carboxylate group on PFOA; however it is expected that both compounds have high capacity to bind in-between the two domains of DeHa1.


Figure 9: A: Output of 1 Click Docking of PFOS to DeHa1. Binding center was Asp 13 residue. Local energy minima was determined in protein interior. B: Surface mesh applied to docking output. PFOS located in a interior pocket between both domains of DeHa1. C: Interior pocket of DeHa1 with PFOA. Fluorinated tail contained in a hydrophobic region while the sulfonate group is located near a largely polar region of the pocket.



Follow up analysis using SwissDock and UCSF Chimera was performed in a similar fashion described earlier. A similar result was seen for the most likely binding site, with PFOS in the interior of the structure. However the second and third highest binding energies were found in the same cluster on the exterior of the structure in the helical domain (Figure 10). The relative energies are shown in Figure 11; all three models are very close in ΔG, suggesting that the true binding site for PFOS is less clear using this approach. One clear downfall in these models is that the dehalogenases conformation remains constant, and the lidang's location and conformation are the only variables adjusted to find a global minima in energy. This limitation may be due to computational resources available for these analysis techniques, or that given proteins often adopt active confirmations that are not representative of a minimum energy conformation it is not helpful to model the modeled protein as having performed such actions. One clear method to validate or disprove this model is to elucidate the tertiary structure of the free dehalogenase and the dehalogenase-PFAS complex by NMR or X-ray crystallography, which will be pursued in future research endeavors.



Figure 10: Binding Energy of PFOS to DeHalogenase 1 as calculated by SwissDock and visualized with a combination of UCSF Chimera and JMOL. A: All identified clusters for PFOS binding. B-D: Models 17, 168, and 170 of PFOS - DeHa1 Complex. Model 17 is in cluster 2 while models 168 & 170 are both in cluster 20. DeHa1 is shown in cartoon style and PFOS is shown in ball and stick style using standard coloration.




Figure 11: Top three binding locations for PFOS in enzyme complex. Model numbers are represented by:'Model.Cluster'. Binding energies are the inverse of ΔG, with Model 17.2 having the strongest binding.


Conclusions

These results suggest that Dehalogenase 1 is capable of binding PFOA and PFOS in a highly conserved sie amongst dehalogenase-like peptides. These results also support the hypothesis that Dehalogenase 1 from Delftia acidovorans is capable of defluorination. Additionally, according to previous research into haloalkane dehalogenases, they are classified as α/Β hydrolases, and require only H2O as a co-substrate. The catalytic mechanism is known for several halogenated substrates, and relies on 2 critical aspartate (occasionally glutamate) residues and a histidine residue to perform catalysis. These three critical residues are referred to as the catalytic triad (nucleophilic aspartate, acidic aspartate, and basic histidine). A possible mechanism for PFOA Β Carbon-fluoride hydrolysis is shown in Figure 12 below. This mechanism may occur in DeHa1 given the presence of the catalytic triad being present in the DeHa1 binding pocket. An important note is that in theory any C-F bond is open to hydrolysis given that a nucleophilic aspartate is responsible for catalysis without mechanistic reliance on the carboxyl group of PFOA, although positioning of the substrate in the binding pocket will dictate which bonds are more subject to hydrolysis. Further research is needed to show experimentally that hydrolysis occurs. Currently, the team is focussed on assaying catalytic capacity of dehalogenase 1 by fluoride concentration monitoring, with plans to follow up via HPLC-MS analysis of degradation products.



Figure 12: Possible mechanism of C-F bond hydrolysis by DeHa1 residues. Catalysis is shown on β carbon of PFOA. Relative positioning of residues are not shown in this model.



Dehalogenase type 2; DeHa2

Modeling DeHa2

In a similar method to DeHa1, Phyre2 was utilized to model the 3-dimensional structure of DeHa2. The result is shown below in Figure 13A. Figures 14B/C show the result of modeling PFOA/PFOS respectively in an identical location used for DeHa1, which did not result in an interior pocket binding location. Scores were -6.4 and -6.8, respectively. This suggests that PFOA/PFOS bind less tightly to DeHa2, and bind to the exterior surface of the enzyme, which does not contain a catalytic triad. This result suggests that DeHa2 may not be capable of binding these substrates and catalytically hydrolyzing C-F bonds. A follow up analysis using SwissDock (Figure 13) resulted in a similar trend with no cluster modeled in which PFOA or PFOS exist in the internal conserved site of DeHa2. From this modeling we expect that DeHa2 to not have catalytic activity on PFOA or PFOS.


Figure 13: structures of DeHa2 from Phyre2 modeling. A: No ligand docked. B: PFOA modeled to bind to DeHa2 using 1 click docking; no interior pocket detected. C: PFOS modeled to bind using 1 click docking; no interior pocket detected.



Figure 14: SwissDock results of A: PFOA and B: PFOS binding to DeHa2 modeled crystal structure. Several clusters were can be seen on the enzyme surface.



Final Conclusions

From our modeling analysis, we would expect catalytic activity of Dehalogenase type 1 (DeHa1) from Delftia acidovorans against both PFOA and PFOS in some capacity given that an internal binding pocket was identified in a highly conserved region of similar proteins from bacterial genera that showed growth in PFOA/PFOS-containing media. Additionally, the conserved region contains the expected catalytic triad from the literature available for this class of enzymes (α/β hydrolases). For both PFOA and PFOS, the internal binding site of DeHa1 provides the greatest ΔG value, indicating a strong binding interaction. This also provides evidence for our expectation for catalytic activity. In sharp comparison, no internal binding site can be identified for PFOA/PFOS in dehalogenase type 2 (DeHa2) from Delftia acidovorans. Given that no binding can be seen in the internally conserved site, we expect no catalytic activity. It should be mentioned again that these modeling attempts are not guaranteed to be a true reflection of reality, as throughout the analysis the dehalogenases structure is not changed. In reality proteins adopt a variety of conformations that is often sensitive to the presence of ligands. As such, no concrete conclusions can be drawn; however the modeling shown here provides a first step in understanding these enzymes and provides testable hypotheses such as the catalytic activity of DeHa1 vs. DeHa2, and the expected secondary structure content via circular dichromism spectroscopy.