05 August 2015

The Value of DSF

Science is based upon incremental advances of previous work.  A year ago, Dan blogged about worked on BioA.  The key take home from that work was that a hydrazine fragment ended up destabilizing the target by 18C.  It ended up being, as expected, a reversible, SAM-competitive inhibitor with modest potency.  As Dan concluded:
This is a very nice paper, and it will be fascinating to try to understand how the fragments so effectively destabilize the protein despite binding tightly, and how this translates into inhibition. The researchers suggest that finding ligands that destabilize proteins could be generally useful for turning off proteins.
In this paper, the same group is back (This work was also presented at DDC in San Diego in April). Interestingly, they seemed to have abandoned the hydrazine.  Taking the same approach (DSF-Xray-ITC) they identify different fragments (2% hit rate from a 1000 screened).  9 were stabilizers (average of +3.8C) and 12 were destabilizers (average of -13.8C(!)).  5 fragments were able to be crystallized by soaking, co-crystallization was able to add one more structure (Figure 1).  Interestingly, the calorimetry showed that only F5's binding is strongly, enthalpically driven.
Figure 1.  Crystallographically Confirmed Fragment Hits
The authors make several interesting observations:
  • Little correlation between magnitude of Tm shift and confirmation by crystallization
  • Stabilizing and destabilizing compounds were confirmed by Xray
  • No correlation between magnitude of the Tm shift and calorimetry determined Kd.
  • Conformational flexibility in the target active site need to be taken into account.
This is not surprising to me; I have seen/heard this many times.  What does this mean for DSF in general? 


Matthieu D said...

Maybe it just means that DSF (like others methods) must be confirmed by (at least) one orthogonal / independent method(s).

Yikai Wang said...

what if a DSF screening only yielded destabilizers without a single stabilizer? can you treat the destabilizers as hits and how to prioritized them?

Dan Erlanson said...

It is interesting that 5 of the 9 stabilizing fragments yielded crystal structures, while only 1 of the 12 destabilzing fragments did. This is consistent with anecdotal reports that stabilizers are more likely to yield crystal structures than destabilizers.

That said, fairly subtle chemical changes can transform a stabilizer to a destabilizer (F5 vs F5.1, for example), while the affinity, binding mode, and thermodynamic parameters remain roughly the same.

One possibility the authors suggest is that the melting point depression might be due to effects other than protein binding, such as reversible bond formation between the primary amine in fragment F5.1 and the PLP cofactor at high temperatures. Alternatively (or additionally), perhaps F5.1 binds preferentially to the unfolded form of the protein, which would also lead to a decrease in Tm.

sgcox said...

In our hands DSF works really well when compound concentration is below 100 uM. Once you go above it, as in the paper with 5 mM (!), you get all sort of artifacts. DSF with environment sensitive dyes can not be run with detergent and compounds aggregation and direct interaction with Sypro orange can simply dominate effect of protein unfolding and data go all other the place. This of course limits the utility of DSF to relatively potent compounds. I am not surprise by all negative experiences when people use it for high concentration screening

AL said...

I Second sgcox above.
Lots of strange things can happen at high concentrations.
When following up hits with DSF dose-response curves, its quite common with bad behavior above 1mM (which i think is the absolute max screening conc for a very soluble Frag library for DSF) and even worse beyond 2mM.