Ultrafiltration to filter fragments

If you’ve taken our poll on finding fragments you’ll have noticed ultrafiltration as one of the possibilities (and if you haven’t voted yet, please do so on the right-hand side of page). Ultrafiltration is grouped with affinity chromatography and capillary electrophoresis because all of these methods involve affinity-based separation of bound from unbound fragments. The technique is described in a recent paper in Anal. Bioanal. Chem.

The basic idea is simple: mix fragments with your protein of interest and then centrifuge through a membrane which retains large molecules such as proteins (and any bound fragments) but allows small molecules (unbound fragments) to pass through. If the composition of the filtrate differs from the composition of the initial mixture, one can assume that any depleted molecules are bound to the protein.

The researchers, all from the University of Washington, Seattle, have been using the technique on internal targets, and their recent paper gives a thorough account of how to do it and describes the results against two protein targets. In both cases, fragments were grouped into pools of 5 to 10 compounds, with each fragment at 0.098 mM and protein at 0.201 mM concentration; these conditions were chosen such that a 1 mM binder would give a 15% reduction in signal, which was roughly 3-fold over their standard deviation for control experiments. The ultrafiltration was done at 4 ˚C in 96-well plates, and the filtrates were analyzed by HPLC using a UV detector (all the fragments contained a chromophore, though the authors suggest that the experiment could also be done using mass spectrometry as a detector).

For the first protein, riboflavin kinase, the researchers found 4 hits out of 134 fragments tested, of which 3 confirmed as single compounds (ie, not in cocktails). Interestingly, these fragments were competed by the enzymatic product flavin mononucleotide but bound more tightly in the presence of the other product and cofactor, ADP and Mg(II). For the other protein, methionine aminopeptidase 1 from the parasite that causes malaria, 10 hits were found out of 243 fragments tested, of which 9 confirmed. The top 6 of these could be competed by methionine, the enzymatic product.

Overall this seems like an interesting method, though I do have one quibble: the authors do not report the activity of these fragments using other techniques. In theory it should be possible to extract dissociation constants from the % reduction in UV signal, and it would be very interesting to see how these values compare to dissociation constants measured using orthogonal methods. Has anyone out there done this?