possu Lv 1
We have previously described the design problem of improving fibronectin and we chose to produce a design by BootsMcGraw. The results are in, but unfortunately the product did not show an improvement in stability. The bacteria made the protein just fine, but it was largely insoluble. This happens usually when the protein produced by the bacteria is not properly folded and started to aggregate as it was being produced so they never had the chance of folding to the target structure. What usually happens next is that a biochemist will carefully solubilize the protein and refold it by keeping it at a low concentration while reducing the amount of denaturants. This way the protein will hopefully fold to its shape first before clumping together and crash out of the solution. To be thorough, our collaborator at USC did exactly that, but we never obtained any soluble protein even after this process.
As described previously, the design by BootsMcGraw was chosen because it has the most balanced qualities according to our force field, namely it was very well packed, got good scores, and was evaluated to have few unsatisfied hydrogen bonding groups. We knew from the start that the sequence was more hydrophobic than what one would expect from an all-beta protein, but we decided to try it since the model buried the hydrophobic groups quite nicely and it was folded correctly by the automated fold-prediction algorithm. The experimental evidence suggests that the scores can be deceptive and if we allow backbones to move relatively freely, it was easy for users to over-pack the core and made them insoluble. This is of course only part of the problem, and the same happens when we run automated design. There were also published evidence that the loops we allow users to mutate may be important for stability (we allow users to change them since FoldIt is setup for users to explore the folding landscape, we are hoping to get better answers than the wildtype). All these should be taken into account for our next round.
Subsequently we've posted a puzzle on the same protein. we locked the backbone structure while imposing constraints to allow mutations only within the amino acids used by proteins with high homology. The results we recovered were quite encouraging, since we got back designed sequences very much like the wildtype even when a poly-alanine was given to the users to start. Interestingly, after collecting the top designed sequences, we realized that most of the mutations suggested by FoldIt players were already tested in a research paper. Some mutations maintained the stability, but none of which improved it.
We are tweaking the puzzle and will try again soon.
happy holidays.
Possu
