Multiple applications for the Malvern MicroCal Auto-iTC200 in a fragment-based drug discovery campaign

Introduction

This work outlines a fragment-based drug discovery (FBDD) approach to the identification and optimization of lead compounds that will inhibit the activity of phosphatidylinositol 3-kinase (Vps34). This kinase has been shown to play an important role in resistance to cancer drugs (1, 2). Vps34 is central to autophagy activation and as such has been identified as a target for therapeutic intervention. Autophagy, or autophagocytosis, is a catabolic process involving degradation of cell components through the lysosomal machinery and is identified as an important mechanism for carcinogenesis.

The  MicroCal Auto-iTC200 played a central role in this campaign primarily because of the minimal steps required to prepare target proteins and the simplicity of the assay design. MicroCal Auto-iTC200 data were used to validate hits from a thermal shift-based primary screen and to accurately rank the affinities of the fragment so that only the strongest binders were advanced to co-crystallization attempts and the structure-based drug discovery program. The ability of this approach to successfully predict hits that would form co-crystal complexes with the target was demonstrated as 12 out of the 14 protein complexes chosen from the Isothermal Titration Calorimetry (ITC) validation were successfully crystallized. The instrument was also used to assess the success of subsequent optimization iterations in the structure-based medicinal chemistry program.

Fig 1. Overview of the FBDD campaign workflow. A fragment library was screened using a thermal shift assay. The resulting fragment hits were validated using the MicroCal Auto-iTC200. The thermal stability of the target protein and the stabilizing effects of the hits were analyzed and validated using the MicroCal VP-Capillary DSC. The hits confirmed by ITC were co-crystallized. An optimization cycle comprising: medicinal chemistry, hit characterization with the MicroCal Auto-iTC200, and co-crystallization, resulted in a set of chemical series. DSC was used to characterize the stability of the target protein and to validate some of the thermal shift data.

Materials and methods

The proteins and fragments were provided by Sprint Bioscience (Stockholm, Sweden). The MicroCal Auto-iTC200 and MicroCal VP-Capillary DSC are available from Malvern Instruments. All experiments were carried out at 25°C.

The MicroCal Auto-iTC200 sample cell was filled with 20 mM Tris pH 7.5, 300 mM NaCl, 10% glycerol, 0.5 mM TCEP, and 2% or 5% DMSO. Those fragments thought to be weak binders (K_D > 10 µM) were injected at a concentration of 4 mM to a protein concentration of 50 µM. The stronger binders were injected at a concentration of 200 µM, to a protein concentration of 20 µM. The injection volumes were 3 µl each for 12 injections, with an injection time of 6 s and a 150 s delay between each injection. Each experiment took approximately 20 minutes in total. The molecular weight range of the fragments was 149 to 333 Da. MicroCal VP-Capillary DSC was used to assess the thermal stability of the target protein in the assay buffer in the absence and in the presence of a selection of fragments. In the DSC scans the protein concentration was 0.2 mg/ml and the fragment concentration 1 mM. The scan rate was 60°C/h.

The primary screen

A library of about 500 fragments was screened against Vps34 using thermal shift approaches, where the protein was heated in the absence and presence of each fragment at a concentration of 1 mM. The protein denaturation process was monitored using differential scanning fluorimetry (DSF). The purpose was to find binders that could successfully be co-crystallized in order to gather structural information for the lead generation and optimization process.

The MicroCal Auto-iTC200 was particularly well suited for this project due to protein instability issues that became apparent during attempts to develop a binding assay requiring protein immobilization. ITC was chosen as a method for hit validation primarily because it requires very few preparative steps, which reduces the risk of protein degradation. Using the MicroCal AutoiTC200 the analyses could be performed easily, rapidly, and in viscous buffers containing glycerol. Basic characterization of the protein stability with MicroCal VP-Capillary DSC uncovered slight thermal stability issues (Tm=50°C in the assay buffer with 4% DMSO). DSC confirmed the stabilizing effects of two fragment hits from the thermal shift assay. One of the fragments identified during the course of this work had been confirmed as a binder by ITC and produced a co-crystal structure, but it proved to give a negative shift in a DSF experiment. During further characterization of this fragment it was shown to stabilize the target protein when tested in DSC. This highlights the importance of orthogonal use of thermal stability assays and the possibility to identify binders also among the fragments that give negative shift in the DSF primary screen.

Follow up of the hits coming from the primary screen was performed in dose response thermal shift studies (Fig 2). A requirement for a successful thermal shift and ITC assay was high solubility of the fragments. A fall-off in the dose-response curve reflected limited solubility of the hit and such fragments could be excluded from further studies. The fragments for which a dose-response curve could be generated were then validated in a secondary assay using the MicroCal Auto-iTC200.

Fig 2. Example of a dose-response plot for two thermal shift experiments. The Tm shift is plotted against fragment concentration. Fragment SB002-41 displayed a fall-off in the curve due to the limited solubility and was therefore excluded from further study.

Hit validation by MicroCal Auto-iTC200

MicroCal Auto-iTC200  was the method of choice for the hit validation primarily because it required very few preparative steps for the protein, reducing the risk of protein inactivation.

Isothermal titration calorimeters measure the heat change that occurs when interacting molecules go from a free to a bound state. The isotherm resulting from the measurements is fitted to a binding model to generate the affinity (K_D), stoichiometry (N), and the enthalpy of interaction (ΔH). In this study, the affinities were used to rank the fragments to select the stronger binders for crystallization.

Of the 47 fragments selected from the thermal shift screen, 33 provided good binding data that could be used to quantitate the affinity and 14 showed either no heat of binding or unusual isotherms (Fig 3). Inspection of MicroCal Auto-iTC200 raw data allowed early identification and elimination of false positives among the hits produced by thermal shift assay. The 33 promising fragments were ranked in order of affinity and the 20 strongest binders were rescreened to assess the robustness of the affinity determination. The data from the first and second screens showed reasonable agreement (Fig 4) despite the weakness of the binders and the fact that two different batches of the protein were tested.

Fig 3. Examples of fragments analyzed with MicroCal Auto-iTC200. The left and the middle graphs show raw data (upper panels) and binding isotherms (lower panels) of a fragment with an affinity of approximately (A) 0.9 mM and (B) 4 µM. Of these compounds the stronger binder was selected for further study. The data for the low affinity fragment (A) represents an example of a low-C titration setup, where a protein at a concentration < K_D is titrated with large excess of a compound (3). (C) shows raw data from a fragment that was rejected because of its unusual isotherm. The peaks are broad and do not return to baseline before the next injection. This indicates a slow event, either a slow (ligand induced) aggregation process or a slow dissociation of the ligand. The latter behavior could be caused by the small molecule aggregating at higher concentrations in the syringe and then dissociating upon dilution. There is no clear indication of binding and some evidence of unusual behavior, which makes it likely that it was a false positive in the primary screen. (D) shows raw data from a fragment that exhibited exothermic behavior in the first 2 injections and endothermic behavior for subsequent injections. This complex behavior might be caused by a false positive, and the compound was therefore removed from further study.

Fig 4. Robustness of measured fragment affinities. Two measurements of the same protein are plotted against each other to assess reproducibility of the affinity determinations.

Based on this data, 14 fragments were chosen for co-crystal structure determination. Of these, 12 were successful (Fig 5), demonstrating the usefulness of the MicroCal Auto-iTC200  as a good predictor of successful determination of protein-fragment co-crystal structures and for advancing fragments into further chemical development.

Fig 5. Crystal structure of a fragment binding to Vps34.

Lead generation

The 47 hits put through the secondary screen represented five different chemical series. The tightest binders identified in the secondary screen originated from two of these series, which were further optimized in a medicinal chemistry program. The next iteration of the compounds contained structural elements of both series and was validated for binding to the target protein with the MicroCal Auto-iTC200 (Fig 6). The second generation of fragments had markedly improved binding affinities to the target protein, demonstrating the applicability of ITC as the structure-activity relationship (SAR) driving assay.

Fig 6. Example of recombination of the two chemical series with the tighter binders. ITC measurements showed improved affinity for the next iteration.

Conclusions

The MicroCal Auto-iTC200  was successfully incorporated into the drug discovery workflow for the fragment-based campaign reported here. The instrument offers the convenience of requiring minimal assay development and little sample preparation, supporting a streamlined workflow that delivers consistent, reliable results. It was used successfully as a secondary screening tool in hit validation for proteins that demonstrated limited stability in other orthogonal assays. Data from the system allowed the early identification and elimination of false positives produced by the primary screen and the technique also proved to be a good predictor of the successful determination of protein-fragment co-crystal structures.

In summary, ITC is a SAR (structure-activity relationship) assay that delivers K_D values which can be used to prioritize fragment series and to guide the chemical development of the next iterations of fragments as part of the drug discovery process. Designed for ease-of-use and exceptional sensitivity, the fully automated MicroCal Auto-iTC200  is a valuable tool for its implementation.

Acknowledgements

The protein material and fragments along with the data from the thermal shift assays and X-ray studies were kindly provided by Sprint Bioscience.

Sprint Bioscience is a drug discovery company building a portfolio of innovative oncology projects, focusing on cancer metabolism. With its fragment-based drug discovery platform, Sprint Bioscience can rapidly identify molecules with properties that are suitable for drug development. These molecules provide a basis for the drug discovery programs. In an iterative process that includes protein science, fragment screening, medicinal chemistry, X-ray crystallography, and relevant biology studies Sprint Bioscience modifies molecules to create candidate drugs ready to enter the development process. In addition, Sprint Bioscience offers contract research services with access to its in-house knowledge and technology platform.

References

1. Funderburk, S.F. et al. The Beclin 1-VPS34 complex – at the crossroads of autophagy and beyond. Trends Cell Biol. 20, 355-362 (2010)
2. Yang, S. et al. Pancreatic cancers require autophagy for tumor growth. Gene. Dev. 25, 717-729 (2011)
3. Tumbull, B.W and Daranas, A.H. On the value of c: Can low affinity systems be studied by isothermal titration calorimetry? J. Am. Chem. Soc. 125, 14859-14866 (2003)