It indicates that there is an interaction between the ligand and the protein that is destabilizing to the protein's tertiary structure. This could be due to a specific, 1:1 ligand-protein interaction, or it could be due to a non-specific, non-stoichiometric, chaotropic effect of the ligand. Try to determine the stoichiometry of the ligand binding by ITC or some other method, if possible. There are also a variety of ways to investigate whether or not the interaction is specific.

Dear Megha, as Adam told you it probably indicates that there is an interaction between the ligand and the protein. 

Just one question. 

I which buffer the ligand is it dissolved? in the same buffer of the protein or in some other solvent and could be the change in the buffer composition and not the interaction with the ligand that cause the Tm  negative shift?

The ligand is in DMSO and then added to the protein sample, such that the final DMSO conc. is of 2% in sample with compound and sample without compound (control with protein but no compound).

OK. 

But in your control is present the DMSO at 2%? o just the protein? 

When you says: a negative Tm transiton. The Tm shift is calculated respect the protein in his own buffer or the protein with 2% of DMSO but with out ligand? 

Because when you want check the effect of a ligand binding is essential that all the component in all well are the same (therefore also the DMSO concentration) and the only difference is represented from the ligand amout, otherwise you have a change that is due to the sum of 2 different factors and you cannot find out the contribute of each single factor. 

control has protein with DMSO. I have DMSO 2% maintained in all my samples whatsoever.

This is a very common phenomenon

Wenchao lu, have you some scientific paper references about this common phenomenon please?

In high-throughput screening(HTS),i mean screening against compound library with enough scaffold diversity, certain compounds may bind to the allosteric site and some other places( usually not the substarte pocket which is related to the positive Tm shift) .|Tm|>2 degree makes sense and compounds should present dose-dependent behavior.

however in literature, i only read 2 two paper publish the negative shift result 

doi: 10.1038/nchembio.2070

http://dx.doi.org/10.1016/j.cub.2012.08.007

Although these results i mean the transition curve is not very standard

specific binding shoud not influence curve shape i.e. baseline, 

Rigorously speaking, a ligand binding to a protein always stabilizes the protein. The question is which conformational state does the ligand stabilize? How does the ligand binding shifts the conformational equilibrium and alters the conformational landscape of the protein? Depending on that the Tm will increase or decrease.

If the ligand binds preferentially to the native state (in principle, the conformational state with the lower energy and, therefore, the more populated state), then the ligand binding will increase the Tm.

If the ligand binds preferentially to a less populated conformational state (e.g. a partially unfolded state not too far in energy from the native state), then the ligand binding will decrease the Tm. The protoptypical example is urea or guanidinium chloride binding preferentially to the unfolded state of a protein and lowering the Tm. The same occurs with protons: reducing the pH will lower the Tm, although in this case, as it may occur with guanidinium, there will be an electrostactic effect.

For a detailed formulation of the conformational equilibrium modulated by ligand binding and resulting in stabilization or destabilization, depending on the preferential interaction of the ligand, see:

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2547457/

but take into consideration that this is just the result of the coupling between conformational and binding equilibria, together with conformational preferential interaction of the ligand.

When we say "the ligand stabilizes the protein and increases the Tm", we want to say "the ligand binds to the protein preferentially to the native state, and it stabilizes (increases the molar fraction or the population of) the native state". Alternatively, when we say "the ligand destabilizes the protein and decreases the Tm", we want to say "the ligand binds to the protein preferentially to the unfolded state or non-native states, and it stabilizes (increases the molar fraction or the population of) the unfolded or non-native states". But, as you see, the ligand always stabilizes the conformational state to which it binds preferentially. However, the Tm may increase or decrease.

Additionally, if the ligand forms colloidal aggregates or micelles, then, the protein may adsorb onto those structures and show a reduced stability. See publications by Brian Shoichet warning about potential artifacts in stabilization screening due to non-specific interactions.

Thank you very much for the answers!!!

I agree with Adam, you should verify the interaction with an orthogonal technique, such as ITC, MST, etc. DSF is typically used to obtain yes/no answers. In order to get information on binding affinity, stoichiometry, etc., you need to apply other methods. They might also answer your question on the observed negative Tm shift.

Article An Automated Microscale Thermophoresis Screening Approach fo...

What about the case where the protein, in its native state, partially unfolded - creating binding surfaces for the dye. Then when ligand is added, it causes the protein to fully fold - eliminating some binding sites for the dye. This would result in the compound lowering the Tm of the protein, even though it is actually stabilizing the fully folded state. Is there is flaw in this logic? Does anyone have an example of this happening?

This is an idea you could easily test isothermally. Does the protein bind less dye when the ligand is present than when it is absent? Just measure the effect of the compound concentration on dye fluorescence intensity.

Thanks Adam. We've done this and do see a concentration-dependent effect of the ligand on fluorescence at low (~25 C) temperatures. It's that this is so unusual that I can't find literature to support my interpretation.

@Pamela,

If the ligand induces refolding in the protein, you might see a reduction in dye fluorescence when adding the ligand. But if the ligand binds to the fully folded protein you should observe an increase in Tm. For observing a decrease in Tm the ligand should preferentially bind to the unfolded protein or to a partially unfolded protein. Another possibility is for the ligand to form micelles, if it is very hydrophobic, which promote unfolding/aggregation of the protein.

By the way, the interaction of hydrophobic fluorescent dyes with proteins is not always straightforward. For example, ANS can be used in DSF as an extrinsic probe. Usually ANS increases its fluorescence emission upon protein unfolding. But there are some proteins showing a decrease in fluorescence intensity upon protein unfolding: prothymosin alpha, nucleobindin, calcium- and integrin binding protein, Tim15...

If anyone is still interested in this question, they should read this article: Article Selective protein unfolding: A universal mechanism of action...