Motivation

The core of our project hinges on how the expressed foreign Thioesterase interacts with the native Fatty Acid Synthase (FAS) complex of yeast. The thioesterases of J.curcas are part of Fatty Acid Synthesis Type I (FAS I) pathway while the FAS complex from Y.lipolytica is part of a Fatty Acid Synthesis Type II (FAS II) pathway. There are two types of FAS's:

Fig: Table highlighting the features of the 2 FAS pathways

The essential difference between these two systems implies that the interactions brought in by our genetic interventions are "unusual" from an evolutionary perspective. Hence its reasonable to assume that there is scope for further improvement in the efficiency of this interaction via the principles of protein engineering. But any attempt in doing this is hindered by the lack of a structural model for these interactions.

No crystal structure has been determined for the integrated complex and hence we decided to insilico model this integration with the objective of using the predicted model for further engineering interventions. As our first attempt we structurally characterized the interaction between J. curcas thioesrerase B (JcFatB) and the FAS complex from Y. lipolytica.

Protein – Protein Docking

The first step in the modelling of this integration was to generate structural predictions for the combined complex comprising both JcFatB from J.curcas and the Fatty Acid Synthase(FAS) complex from Y.lipolytica.

Predicting this structure from the FASTA sequence using AlphaFold 2.0 turned out to be not feasible, because of the huge size of the complex (24,186 residues). Hence, we decided to use protein - protein docking as the strategy to predict the integrated complex structure. Owing to its capability to handle large input files, we decided to use PatchDock server for this purpose.^[1]

PatchDock

The PatchDock algorithm draws inspiration from object recognition and image segmentation methods commonly used in the field of Computer Vision. When provided with two molecules, it divides their surfaces into patches based on their shape, mirroring the idea of visually distinctive puzzle piece patterns. After identifying these patches, the algorithm can align them using shape complimentarity techniques.^[2][3]

The Docking Algorithm

The algorithm can be divided into three major steps:

• Molecular Shape Representation - In this step the molecular surface of the molecule is computed. Next, a segmentation algorithm is applied for detection of geometric patches (concave, convex and flat surface pieces). The patches are filtered, so that only patches with 'hot spot' residues are retained.

• Surface Patch Matching - A hybrid of the Geometric Hashing and Pose-Clustering matching techniques are applied to match the patches detected in the previous step. Concave patches are matched with convex and flat patches with any type of patches.
• Filtering and Scoring - The candidate complexes from the previous step are examined. All complexes with unacceptable penetrations of the atoms of the receptor to the atoms of the ligand are discarded. Finally, the remaining candidates are ranked according to a geometric shape complementarity score.

For our purpose of modelling the integration, a PatchDock protein – protein docking was performed with the following inputs:

• JcFatB structure
• Fatty acid Synthase (FAS) complex structure

Now these proteins (JcFatB from J.curcas and FAS complex from Y.lipolytica) were not structurally characterized in literature. For JcFatB we ran an AlphaFold 2.0 prediction and used the prediction with highest confidence score as the input structure.As the second input, the structure of FAS complex from S.cerevisiae (available in Protein Data Bank, PDB ID: 6QL6) was used since the actual structure was too large to be predicted using AlphaFold 2.0 and a 99% query cover was found between the sequences from these yeast species.

Fig: JcFatB(left) and FAS complex(right) structures

The server generated 3000 docked structures as output, which were filtered in the next step to obtain the most probable structure.

Validation of the Docking Structures

Fig:Progression of the validation process

Once we obtained 3000 output structures as the possible configurations in which JcFatB could interact with the FAS complex, the next task at hand was to validate one as the most probable model of interaction. A combination of docking score cutoffs and biologically valid rationale was used for this purpose. The series of filters used in this process are as follows:

Docking Score Cutoff:

• PatchDock assigns each of its predicted structure with a docking score in terms of a geometric shape complementarity score, which is a numerical measure that assesses how well two molecular structures or complexes fit together geometrically. It quantifies the compatibility of the shapes of the interacting molecules in a docking simulation. From the 3000 structures predicted, top 50 structures (in terms of the docking score) were filtered out and used in the later stages of validation.

Hypothesis from Literature:

• Literature hypothesizes the mechanism of the integration as by the localisation of the heterologous Thioesterase into the reaction chambers of yeast’s native Fatty acid Synthase complex, facilitating access to acyl-ACP substrates from their catalytic sites.^[5] Of the top 50 structures obtained from the last step, 4 structures were identified which satisfied this constraint.
  

  Fig:4 structures satisfying the hypothesis

Active Site Analysis:

• The accessibility of the active site of JcFatB to the substrates present in the reaction chamber of the FAS complex was chosen as the next criteria to validate the structures. Jatropha’s thioesterase belongs to the TE14 family which are characterised by substrate specificity towards short to long acyl chains bound to ACP rather than CoA. TE14 thioesterases consist of two tandem hot dog domains, termed the N-terminal hotdog and the C-terminal hotdog domains. Literature identifies a network of 4 catalytic residues (Asp535, Asn537, His539, and Glu573) located at the periphery of the C-terminal Hotdog domain as the active site.^[6]


Fig: Spatial arrangement of the catalytic residues



Fig: Structure of JcFatb with the catalytic residues in yellow

• In our analysis we found that the structure ranked 2nd among the 4 structures obtained from the last step had its active site buried into the walls of the FAS complex, rendering it inaccessible to the ACP type substrates present inside the reaction chamber. Hence this structure was eliminated at this stage of validation.

Redundancy Analysis:

• The FAS complex is an α6β6 type protein complex and has a barrel-shaped hollow body separated at the equator by a wheel-like structure with three spokes. The six α subunits form the central wheel in the assembly, and the β subunits form domes on the top and bottom of the wheel, creating six reaction chambers, 3 above and 3 below. Because of this overall symmetry of the system, the possibility for redundancies in the obtained output structures cannot be neglected and we analysed the interaction sites for the same. We found that the top two structures obtained from the last step were equivalent, with the binding residues conserved between the two. The third structure represented a different binding configuration. Since we observed a convergence of results in the case of the first two structures and because they were ranked higher in terms of docking, we validated structure 1(and 2) as the most probable model of integration.


Fig: The most probable structure of integration, with the JcFatB in yellow

• As described above, following a combination of docking score cutoffs and biological rationale, we were able to validate one structure from the 3000 predicted structures as the most probable model for this integration. This structure was further used to identify the residues on JcFatB which interact with FAS complex, giving us insights into engineering strategies that could be undertaken to improve the efficiency of this interaction.

Identification of Interacting Residues

The most probable model of integration was analysed using PyMol and interacting residues were identified.

Fig: The most probable structure of integration, with JcFATB colored yellow and the interacting residues highlighted in red

With the information of these interacting residues (and the active residues) we propose to engineer the JcFatB via random mutations with objective of improving its binding affinity towards the FAS complex without affecting its catalytic function.

In summary, this research sheds light on the intriguing interaction between the thioesterase (JcFatB) from J.curcas and the Fatty Acid Synthase (FAS) complex from Y.lipolyitca . By using computational modeling techniques, we tackled the challenge of understanding their interaction, despite lacking a physical structure. Our validated model provides a practical framework for exploring and optimizing this unique interaction, with potential applications in further improving the efficiency of our biomanufacturing process.