Looking for protein stabilizing drugs with thermal shift assay

Thermal shift assay can be used for the high‐throughput screening of pharmacological chaperones. These drugs are small molecules that bind a mutant protein and stabilize it. We demonstrated the robustness, reproducibility and versatility of the method using two molecules that are in clinical trial for Fabry or Pompe disease, Deoxygalactonojirimycin and N‐Butyldeoxynojirimycin, and their target enzymes, lysosomal alpha‐galactosidaseA and alpha‐glucosidase, as test cases. We assessed the influence of solvents and of scanning rate on the measures. We showed that a value that is equivalent to the melting temperature can be obtained by the first derivatives of raw data. We discuss the advantages of the method and the precaution to be taken in running the experiments. © 2015 The Authors Drug Testing and Analysis Published by John Wiley & Sons Ltd.

it can also stabilize the wild type protein, which can conveniently replace mutants for screening purposes. Human wild type recombinant lysosomal alpha-galactosidaseA and alpha-glucosidase are commercially available with the names of Fabrazyme® and Myozyme® (Genzyme Corporation, Cambridge, MA, USA), respectively. In Figure 1, panels A, B, and C, we show the melting profiles of Fabrazyme® recorded at 0.5°C/min with or without ligands, (0.04 mM DGJ (SIGMA, Milan, Italy) or 100 mM galactose). Raw data are unpractical for comparison. The unfolded fraction can be calculated as fu(T) = f(T)-fn(T)/fd(T)-fn(T) where f(T) is the fluorescence at temperature T, fn(T), and fd(T) are the values of fluorescence extrapolated at temperature T from the native and unfolded regions of the melting profile (data not shown) or, more simplistically, normalizing the melting profiles using the equation fu(T) = f(T)-fn/fd-fn, where fn represents the minimum value of the fluorescence before the transition and fd represents the maximum value after the transition ( Figure 1, panels A,B,C,F; and Figure 2, panels A,B,C,F). From these curves we calculated the midpoint of the protein melting transition, T 0.5 and observed that the effect of the simplification on T 0.5 is negligible if compared with variability, on average ±1°C, among our replicas. In order to assess the reproducibility of the measures and the influence of the technique employed, we compared our results (T 0.5 =48°C neutral pH no ligand; T 0.5 =56°C acidic lysosomal pH no ligand; T 0.5 =61°C neutral pH plus DGJ; T 0.5 =70°C acidic lysosomal pH plus DGJ) with those obtained by Valenzano et al. [14] (T 0.5 =47°C neutral pH no ligand; T 0.5 =58°C acidic lysosomal pH no ligand; T 0.5 =58°C neutral pH plus DGJ; T 0.5 =72°C acidic lysosomal pH plus DGJ) exploiting TSA or by Petsko et al. [15] (T 0.5 =48.0°C neutral pH no ligand; T 0.5 =60.2°C acidic lysosomal pH  no ligand; T 0.5 =60.6°C neutral pH plus DGJ; T 0.5 =73.5°C acidic lysosomal pH plus DGJ) exploiting differential scanning calorimetry. Taken together the three sets of experiments cover a broad range of protein concentrations (10 μM our data, 2 μM Valenzano et al. [14] 47.5 μM or 8.6 μM Petsko et al. [15] ) and scanning rates (0.5°C/min our data, 1.0°C/min Valenzano et al. [14] 1.0°C/min or 1.5°C/min Petsko et al. [15] ). Denaturation induced during TSA is not a reversible process and the parameters which are measured cannot be defined as thermodynamic properties of the system, nonetheless the coincidence of T 0.5 measured by three groups with different experimental procedures allows us to consider TSA as a sufficiently robust method to assay lysosomal alpha-galactosidaseA stability under different conditions. The parallelism of the results obtainable with TSA and with calorimetry can be pushed further. Examining the melting curve profiles, we observed that the fraction of unfolded apo-enzyme increases from 10% to 90% over 5-6°C both at pH 7.4 ( Figure 1, panel C) and 5.2 (Figure 1, panel B) in a highly cooperative fashion whereas a more complex unfolding process occurs in the presence of DGJ. This confirms the results obtained by Petsko et al. [15] who found a coincidence of calorimetric and van't Hoff curves for the free enzyme, but not for the complexed enzyme and suggested that ligand binding leads to the preferential stabilization of the TIM barrel domain where binding takes place. The effect is more evident at pH7.4 than at pH 5.2 both with scanning calorimetry [15] and with TSA. TSA is useful to compare the stabilizing effect of different ligands and infer the binding modes. Galactose has a lesser stabilizing effect than DGJ; this reflects the fact that galactose has lower affinity for alpha-galactosidaseA than DGJ, K i 16 mM and 39 nM, respectively. [16] Melting temperature variations, ΔT 0.5 =T 0.5DGJ -T 0.5galactose , are not dependent on pH over the range explored (Figure 1, panel D). DGJ and galactose have a different heteroatom in the ring, N or O, respectively, with a different protonation state. If a titrable residue was responsible for the higher affinity of DGJ, [16] it would be expected that ΔT 0.5 depended on pH. Experiments at different pH are important to show that both DGJ and galactose are not the optimal chaperones for Fabry disease, because they bind the enzyme under neutral and acidic conditions. These molecules in fact are reversible inhibitors of the enzyme. Ideally, this type of chaperone should bind alpha-galactosidaseA in the endoplasmic reticulum, stabilize it during its transport to lysosomes and dissociate in the acidic compartment where the enzyme encounters its substrate and exerts its biological activity.
In Figure 2, panels A, B, or C, we show the melting profiles of Myozyme® recorded at 0.5°C/min or 1.0°C/min in the presence or in the absence of a ligand (0.04 mM NB-DNJ (SIGMA, Milan, Italy)). Contrary to the case represented by Fabrazyme®, results depend on the scanning velocity and the apparent higher stability is inferred from the curves recorded at 1.0°C/min. Nonetheless results are reproducible provided that data are collected at the same scanning velocity (1.0°C/min), as we observed comparing T 0.5 measure by Flanagan et al. (T 0.5 =53°C neutral pH no ligand; T 0.5 =68°C acidic lysosomal pH no ligand) [17] and by us (T 0.5 =53°C neutral pH no ligand; T 0.5 =68°C acidic lysosomal pH no ligand). In relative terms results are robust since the ΔT 0.5 observed with or without the ligand at the three pHs is the same and do not depend on the scanning velocity. The experiments carried out at various pHs consents to demonstrate that NB-DNJ stabilizes Myozyme® better at neutral pH than at acidic pHs ( Figure 2, panel D). A possible explanation could be that a hydrogen-bond acceptor, which is protonated at low pH, plays a role in the binding at neutral pH. As a consequence of this, NB-DNJ might preferentially stabilize alpha-glucosidase mutants in the endoplasmic reticulum than in the lysosomes, leaving the enzyme more available to act on the substrate.
So far we have shown normalized melting profiles (Figure 1, panels A,B,C and Figure 2, panels A,B,C) obtained with the previously described equation fu(T) = f(T)-fn/fd-fn. But in order to evaluate the stabilizing effect of ligands it is sufficient to process raw data with software that are provided with any Real Time PCR equipment. In Figure 1, panel E and Figure 2, panel E we show the curves that can be obtained calculating the first derivatives of raw data; the temperature at which the minimum is observed corresponds with T 0.5 and differences between the two values are negligible if compared with variability, on average ±1°C, among our replicas.
High throughput screenings are often carried out in the presence of dimethyl sulfoxide (DMSO). The effect of the solvent on Fabrazyme® (Figure 1, panel F) and Myozyme® (Figure 2, panel F) is different for the two enzymes suggesting that it must be evaluated case by case when planning a screening.
In conclusion, TSA offers advantages over other methods, differential scanning calorimetry, [15] isothermal titration calorimetry, [18] circular dichroism, [16] chemical induced denaturation followed by intrinsic fluorescence detection [16] or limited proteolysis, [19] because it can process many samples containing a small amount of protein at the same time, is fast and requires equipment which is largely available in the majority of biomedical laboratories and processing of data is very simple. The method is robust, reproducible and versatile. It should be considered that the effect of scanning rates and the effect of solvents can vary case by case. In relative terms, binding of ligands and stabilization of the protein target can be detected with fast scanning rates, which are commonly employed and might be preferable for high throughput screening plans when thousand of molecules must be tested.