<p>In bioremediation and directed evolution, we were mainly looking at six different MTs, from the species <i>Mytilus edulis, Mytilus galloprovincialis, Callinectes sapidus, Danio rerio, Pseudomonas fluorescens</i> and <i>Saccharomyces cerevisiae</i>. Structures for the <i>C. sapidus</i> MT in complex with Cd<sup>2+</sup> [12], and the <i>S. cerevisiae</i> MT in and out of complex with Cd<sup>2+</sup> exist on the PDB [13], however not for any of the other species. Given the strong conformational changes which occur with ligand binding, we decided to use AlphaFold to predict the structures of all our MTs. We generated 10 models for each MT, and selected the model which scored the highest in the Local Distance Difference Test (lDDT) to use for our docking simulations.</p>
<p>For biosensing, we were chiefly interested in four different TFs from <i>Cupravidus metallidurans</i>; arsR, pbrR, merR and a mutated form of merR (<ahref="http://parts.igem.org/Part:BBa_K1724002">BBa_K1724002</a>, henceforth referred to as mut_merR), which mainly bind Arsenic, Lead, Mercury and Cadmium ions respectively. arsR [14] and merR [3] have been structurally characterized, but from different species. pbrR has an existing AlphaFold predicted PDB structure, and mut_merR has never been structurally characterized. We generated de novo structures for each of these using AlphaFold as well and chose to perform our docking simulations with both the existing TF structures not from <i>C. metallidurans</i>, and the <i>de novo</i> predicted AlphaFold structures.</p>
<p>For biosensing, we were chiefly interested in four different TFs from <i>Cupravidus metallidurans</i>; ArsR, PbrR, MerR and a mutated form of MerR (<ahref="http://parts.igem.org/Part:BBa_K1724002">BBa_K1724002</a>, henceforth referred to as mut_merR), which mainly bind Arsenic, Lead, Mercury and Cadmium ions respectively. ArsR [14] and MerR [3] have been structurally characterized, but from different species. PbrR has an existing AlphaFold predicted PDB structure, and mut_merR has never been structurally characterized. We generated de novo structures for each of these using AlphaFold as well and chose to perform our docking simulations with both the existing TF structures not from <i>C. metallidurans</i>, and the <i>de novo</i> predicted AlphaFold structures.</p>
<p>PDB structures were already dimerized, as our TFs are found as dimers in vivo [5, 6, 20]. However, normal AlphaFold produces monomeric structures [7], so we needed to dimerize and energy minimize these structures. We chose to do this with the <ahref="http://huanglab.phys.hust.edu.cn/hsymdock/">HSYMDOCK Server</a>, choosing C2 symmetry for each TF [1,9,10,15,17,19,21]. Energy minimization was performed with the <ahref="http://www.yasara.org/minimizationserver.htm">YASARA energy minimization server</a> [8]. The resulting structures were extracted using YASARA view and used for the rest of the docking simulations.</p>
<h3id="2.3"class="anchor">Heavy metal docking simulations</h3>
<p>We used the same approach to dock heavy metals to both our MTs and TFs. Heavy metal ions are coordinated by cysteines, so we decided to dock on every cysteine in each protein and used the free energy of binding to see which cysteines could be successfully docked by each heavy metal ion. Unlike normal docking simulations which use mainly van der Waals interactions to determine docking, we needed to slightly modify the approach because heavy metal ions form a dative covalent bond. We prepared ligand PDBs with AutoDock’s <i>prepareCovalent.py</i> [2], removed hydrogens and added gasteiger charges for the heavy metals, and then used MGLTools <i>prepare_receptor4.py</i> to prepare the ligand PDBQT [11], in accordance with the Covalent Docking protocol. We then prepared the protein PDBQT with MGLTools <i>prepare_receptor4.py</i> with the -A Hydrogens flag. We then generated flexible docking files with MGLTools <i>prepare_flexreceptor4.py</i>, and generated Grid Parameter Files and Docking Parameter Files (GPFs and DPFs) with MGLTools <i>prepare_gpf4.py</i> and <i>prepare_dpf4.py</i>, specifying the flexible receptor files [11]. We then used AutoGrid to calculate atomic affinities, and finally AutoDock 4.2 to preform docking [11] See Figure 1 for a diagramatic workflow. The docked coordinates of the heavy metal and corresponding gamma sulfur for the best docked model were extracted from the resulting docking logs. Heavy metals which could spontaneously bind (negative free energy of binding) were added back into the original structures to generate the final docked structures. Full annotated shell scripts are available on our <ahref="https://gitlab.igem.org/2022/software-tools/edinburgh-uhas-ghana/-/tree/main">GitLab software page.</a></p>
...
...
@@ -139,16 +139,16 @@
<h3id="3.2"class="anchor">Transcription Factors - Heavy Metals</h3>
<p>Like with MTs, Heavy Metal regulated TFs also coordinate metal ions using cysteine. However, MTs are reasonably nonspecific in their Heavy Metal coordination, as that is what they have evolved to do, and hence can bind multiple. Dimeric Heavy Metal regulated TFs on the other hand, only ever coordinate two heavy metal ions, one for each symmetrical face. To determine which cysteine residues are normally involved in coordination, we used the existing structures of merR and arsR in complex with their heavy metal.</p>
<p>Like with MTs, Heavy Metal regulated TFs also coordinate metal ions using cysteine. However, MTs are reasonably nonspecific in their Heavy Metal coordination, as that is what they have evolved to do, and hence can bind multiple. Dimeric Heavy Metal regulated TFs on the other hand, only ever coordinate two heavy metal ions, one for each symmetrical face. To determine which cysteine residues are normally involved in coordination, we used the existing structures of MerR and ArsR in complex with their heavy metal.</p>
<figcaption><b>Figure 3: </b>Structures of a) <i>Bacillus megaterium</i>merR in complex with two Hg<sup>2+</sup> ions (grey spheres, PDB: <ahref="https://www.rcsb.org/structure/4UA1">4ua1</a>). b) <i>Cornyebacterium glutamicum</i>arsR in complex with two As<sup>3+</sup> ions (purple spheres, PDB: <ahref="https://www.rcsb.org/structure/6J0E">6j0e</a>). Pictured are an image of the entire TF bound to two ions (left) and a closeup of the the ion coordinated by cysteines, shown in yellow (right).</figcaption>
<figcaption><b>Figure 3: </b>Structures of a) <i>Bacillus megaterium</i>MerR in complex with two Hg<sup>2+</sup> ions (grey spheres, PDB: <ahref="https://www.rcsb.org/structure/4UA1">4ua1</a>). b) <i>Cornyebacterium glutamicum</i>ArsR in complex with two As<sup>3+</sup> ions (purple spheres, PDB: <ahref="https://www.rcsb.org/structure/6J0E">6j0e</a>). Pictured are an image of the entire TF bound to two ions (left) and a closeup of the the ion coordinated by cysteines, shown in yellow (right).</figcaption>
</figure>
<br>
<p>The crystal structures reveal that merR normally coordinates heavy metal ions using CYS79 of one chain, and CYS114 and CYS123 of the other chain for each heavy metal ion (Figure 3a), and arsR coordinates with CYS34 and CYS37 of one chain, and CYS91 of the other chain (Figure 3b). We then found the free energies of heavy metal docking to these cysteines, for existing and AlphaFold predicted structures, and generated structures of docked TFs (Figure 4).</p>
<p>The crystal structures reveal that MerR normally coordinates heavy metal ions using CYS79 of one chain, and CYS114 and CYS123 of the other chain for each heavy metal ion (Figure 3a), and ArsR coordinates with CYS34 and CYS37 of one chain, and CYS91 of the other chain (Figure 3b). We then found the free energies of heavy metal docking to these cysteines, for existing and AlphaFold predicted structures, and generated structures of docked TFs (Figure 4).</p>
<figcaption><b>Figure 4: </b>Structures of a) PDB structure of merR in complex with Hg<sup>2+</sup>, b) PDB structure of arsR in complex with As<sup>3+</sup>, c) PDB structure of arsR in complex with As<sup>5+</sup>, d) AlphaFold structure of merR in complex with Hg<sup>2+</sup>, e) AlphaFold structure of mut_merR in complex with Cd<sup>2+</sup>, f) AlphaFold structure of arsR in complex with As<sup>3+</sup>, g) AlphaFold structure of arsR in complex with As<sup>5+</sup>. Structures of pbrR with docked Pb<sup>2+</sup> could not be generated.</figcaption>
<figcaption><b>Figure 4: </b>Structures of a) PDB structure of MerR in complex with Hg<sup>2+</sup>, b) PDB structure of ArsR in complex with As<sup>3+</sup>, c) PDB structure of ArsR in complex with As<sup>5+</sup>, d) AlphaFold structure of MerR in complex with Hg<sup>2+</sup>, e) AlphaFold structure of mut_MerR in complex with Cd<sup>2+</sup>, f) AlphaFold structure of ArsR in complex with As<sup>3+</sup>, g) AlphaFold structure of ArsR in complex with As<sup>5+</sup>. Structures of PbrR with docked Pb<sup>2+</sup> could not be generated.</figcaption>
</figure>
<br>
<p>The results indicate that using this method, heavy metals can be successfully docked into merR, arsR and mutated merR. pbrR could not be successfully docked, as the model generated a positive value for the free energy of binding, indicating that the ligand cannot bind the structure, which is not true. This could arise from the assumption that pbrR and merR use the same binding site because pbrR comes from the merR family of transcriptional repressors, whereas in reality the binding site of pbrR is not known. The docking results also show that As<sup>3+</sup> binds more strongly to arsR than As<sup>5+</sup>, which is consistent with known results [18]. There exists variation between the AlphaFold predicted and PDB extracted structures, but this could also be due to the known TFs being from a different species than our AlphaFold predicted TFs.</p>
<p>Overall, these docking simulations have helped us gain deeper insight into the nature of both our biosensor and bioremediation device, showing which metallothioneins could be best to include in a hydrogel, and validating some assumptions about our biosensors. The failure of pbrR docking also could indicate it is not the best TF to be using for inducible constructs such as our biosensor. Importantly, this model has also further characterized mut_merR, both in terms of structure and binding affinity to cadmium and has showed that this part could be useful for Cd<sup>2+</sup> inducible constructs. </p>
<p>The results indicate that using this method, heavy metals can be successfully docked into MerR, ArsR and mutated MerR. PbrR could not be successfully docked, as the model generated a positive value for the free energy of binding, indicating that the ligand cannot bind the structure, which is not true. This could arise from the assumption that PbrR and MerR use the same binding site because PbrR comes from the MerR family of transcriptional repressors, whereas in reality the binding site of PbrR is not known. The docking results also show that As<sup>3+</sup> binds more strongly to ArsR than As<sup>5+</sup>, which is consistent with known results [18]. There exists variation between the AlphaFold predicted and PDB extracted structures, but this could also be due to the known TFs being from a different species than our AlphaFold predicted TFs.</p>
<p>Overall, these docking simulations have helped us gain deeper insight into the nature of both our biosensor and bioremediation device, showing which metallothioneins could be best to include in a hydrogel, and validating some assumptions about our biosensors. The failure of PbrR docking also could indicate it is not the best TF to be using for inducible constructs such as our biosensor. Importantly, this model has also further characterized mut_MerR, both in terms of structure and binding affinity to cadmium and has showed that this part could be useful for Cd<sup>2+</sup> inducible constructs. </p>
