MALDI-TOF-Based Affinity Selection Mass Spectrometry for Automated Screening of Protein

Abstract

Demonstration of in vitro target engagement for small-molecule ligands by measuring binding to a molecular target is an established approach in early drug discovery and a pivotal step in high-throughput screening (HTS)-based compound triaging. We describe the setup, evaluation, and application of a ligand binding assay platform combining automated affinity selection (AS)-based sample preparation and label-free matrix-assisted laser desorption/ionization time-of-flight (MALDI-TOF) analysis. The platform enables mass spectrometry (MS)-based HTS for small-molecule target interactions from single-compound incubation mixtures and is embedded into a regular assay automation environment. Efficient separation of target–ligand complexes is achieved by in-plate size exclusion chromatography (SEC), and small-molecule ligands are subsequently identified by MALDI-TOF analysis. In contrast to alternative HTS-capable binding assay formats, MALDI-TOF AS-MS is capable of identifying orthosteric and allosteric ligands, as shown for the model system protein tyrosine phosphatase 1B (PTP1B), irrespective of protein function. Furthermore, determining relative binding affinities (RBAs) enabled ligand ranking in accordance with functional inhibition and reference data for PTP1B and a number of diverse protein targets. Finally, we present a validation screen of more than 23,000 compounds within 24 h, demonstrating the general applicability of the platform for the HTS-compatible assessment of protein–ligand interactions.

Keywords

binding assay MALDI-TOF mass spectrometry high-throughput screening PTP1B

Introduction

High-throughput screening (HTS) of large compound libraries against molecular targets has proven itself an efficient approach for early drug discovery campaigns focusing on the identification of novel chemical matter for lead optimization. While sophisticated assay technologies have been developed to enable assessment of test substances for their biological or biochemical effects, the HTS-compatible measurement of protein–ligand interaction has proven to be challenging. Because demonstration of small-molecule target engagement in a primary or secondary assay format is an integral part of HTS-driven drug discovery campaigns, innovative and HTS-capable binding assay formats would be beneficial. This is further emphasized by an increasing demand for binding assays aiming at the de-orphanization of unprecedented targets or the identification of multivalent ligands for targeted protein degradation [e.g., proteolysis-targeting chimeras (PROTACs)]. One major challenge has been the selection of unbiased screening formats and detection principles, because effects attributable to the readout technology, functional labels, and detection reagents can result in assay-specific hit sets with poor overlap among individual assay technologies.¹

Isothermal titration calorimetry (ITC), surface plasmon resonance (SPR), nuclear magnetic resonance (NMR) spectroscopy, and X-ray crystallography (XRC) represent established and label-free biophysical reference methods, which provide valuable information on structural, thermodynamic, and kinetic aspects of the macromolecule–ligand interaction but are not yet compatible with cycle times required for HTS of compound libraries commonly exceeding 100,000 molecules. Historically, displacement assays, in which the competition between a test compound and a labeled tracer molecule for binding to the target of interest is monitored, have been the method of choice for larger screening campaigns.^2–4 Labeling strategies for a tracer ligand of known affinity comprise chemical introduction of functional labels or replacement of specific atoms by their respective radioisotopes for direct or indirect downstream detection. While competitive binding formats usually provide the required throughput, are easy to miniaturize, and can be fully automated, their dependence on a labeled tracer molecule necessitates the availability of a known ligand and constrains the exploration of binding epitopes on the target protein to the respective tracer binding site. An alternative HTS-capable format that has been widely used for detecting in vitro target engagement by small molecules is differential scanning fluorimetry (DSF).⁵ Here, the ligand-dependent stabilization of a protein against thermally induced denaturation is measured with the aid of a hydrophobic dye that emits a fluorescent signal on binding to hydrophobic surfaces on denatured protein. Although this assay does not require a target-specific tracer and is potentially capable of detecting ligands of different binding epitopes, its information value is limited, because the correlation between protein stabilization and ligand binding is not generally valid.⁶ Furthermore, unspecific compound-mediated protein perturbation, due to constraints in buffer composition, and signal interference can exacerbate data interpretation and analysis.

To circumvent such limitations, the use of mass spectrometry (MS) for label-free analyte detection combined with an affinity-based selection procedure, termed affinity selection mass spectrometry (AS-MS), has been suggested for identifying target ligands.^7,8 A significant advantage of this technique is its label-free readout, which is independent of tracers and detection reagents and enables screening of orphan targets, in which no function or endogenous ligand is known. Furthermore, the isolation of protein–ligand complexes from nonbinding buffer components prior to analysis removes constraints in buffer composition and minimizes ion suppression caused by buffer components during MS analysis. Sophisticated applications of AS-MS are the Automated Ligand Identification System (ALIS)^9,10 and the SpeedScreen¹¹ platform, which couple size exclusion chromatography (SEC)-based isolation of protein–ligand complexes to unambiguous detection of bound ligand via liquid chromatography/electrospray ionization MS (LC/ESI-MS). Although technical improvements of ESI instrumentation have shifted analysis cycle times to a few seconds per sample, this throughput is still inadequate for traditional large compound library screens.¹² The above-mentioned technologies therefore use compilations of test substances in compound pools comprising up to 2500 members per experiment.⁹ While this is a powerful approach for screening specially designed mass-encoded libraries, technical implementation of large-scale compound pooling, complex library deconvolution, and ensuring compound integrity and solubility can be cumbersome. In addition, required elevated DMSO concentration (≥2%) might not be compatible with certain targets, and mutual binding competition of compound pool members could decrease assay sensitivity, thereby complicating the identification of low-affinity binders. The integration of AS-MS into regular HTS infrastructure therefore requires the use of MS technology with faster cycle times.

Compared to ESI-based MS, matrix-assisted laser desorption/ionization time-of-flight (MALDI-TOF) MS provides the sampling speed required for HTS and has been established as a valuable readout strategy for large-scale screening campaigns against the catalytic activity of enzymatic targets.^13–17 The adoption of the technology to interrogate protein–ligand interaction has been the subject of a number of proof-of-principle studies using ultrafiltration or affinity capture-based workflows to isolate protein–ligand complexes from incubation mixtures.^18–20 Although these studies have demonstrated the general feasibility of using MALDI-TOF MS for label-free small-molecule detection, a report of its combination with an HTS-capable automation platform is unprecedented.

In this work, we present how the speed of MALDI-TOF MS combined with fully automated sample preparation and software-aided compound tracking can be leveraged to perform label-free protein–ligand binding assays in standard single-compound incubation HTS format on a regular automation platform. We describe the setup and validation of an automated SEC-based workflow in 384-well format for efficient and reproducible sample preparation, as well as the optimization of sample/MALDI matrix co-crystallization for MALDI-TOF small-molecule detection. Using the protein phosphatase PTP1B as a model system, we cross-compare the obtained AS-MS results to two orthogonal binding assay formats and a functional activity-based readout, and further demonstrate the capability of the platform in a validation study screening >23,000 small molecules on a single day.

Materials and Methods

Materials

Disposable HTS MALDI target plates (no. 1847006) and HTS MALDI Adapter (no. 8283496) were purchased from Bruker Daltonics (Billerica, MA). 384-well polypropylene (PP) microwell plates (no. 784201) were obtained from Greiner Bio-One (Frickenhausen, Germany). Recombinant human PTP1B_CA (PTP1B_C215A_h2-321_NHis8TEV) was purchased from Trenzyme GmbH (Konstanz, Germany). The catalytic domain of PTP1B (PTP-1B_h2-321_NHis8TEV) was produced in-house at Boehringer Ingelheim. α-Cyano-4-hydroxycinnamic acid (4-HCCA; no. 70990), acetonitrile (no. 34851), DMSO (no. D5879), ammonium acetate (no. 73594), bovine gamma globulin (BGG; no. G5009), HEPES (no. H3375), isopropanol (no. 34863), and SYPRO Orange (no. S5692) were obtained from Sigma Aldrich (St. Louis, MO). Trifluoroacetic acid (TFA; no. 6957) was purchased from Carl Roth (Karlsruhe, Germany). Bovine serum albumin (BSA; no. 11945.03) and dithiothreitol (DTT; no. 20710.04) were obtained from Serva (Heidelberg, Germany). Bio-Gel P10 Media (Fine: 45–90 µm; no. 150-4144) was purchased from BioRad (Feldkirchen, Germany). Multiscreen HTS 384-well filter plates [0.45 µm polyvinylidene fluoride (PVDF) membrane; no. MZHVN0W10], Tris (no. 108382), and MgCl₂*6 H₂O (no. 105833) were obtained from Merck Millipore (Burlington, MA). Peptides (custom synthesis), KCl 2M solution (no. AM9640G), and TCEP [tris(2-carboxyethyl)phosphine; no. 20491] were purchased from Thermo Fisher Scientific (Waltham, MA). AlphaScreen reagents (Streptavidin Donor beads, no. 676002; and anti-6xHis AlphaLISA Acceptor beads, no. AL128M) were obtained from PerkinElmer (Waltham, MA).

MALDI-TOF AS-MS Screening Workflow

The setup of incubation mixtures, the in-plate SEC-based separation process, and the transfer of 384-well plates into 1536-well format were performed on a fully automated system featuring a modular design and robotic plate transportation. At the beginning of each iterative cycle, a 384-well compound plate was transported to a CyBio Well vario liquid-handling system (Analytik Jena, Jena, Germany) equipped with a 384-well format capillary head to transfer 50 nL of 5 µg/µL or 10 mM compound solutions or DMSO to the wells of a 384-well PP assay plate. The compound plate was then put back into storage, and the assay plate was transported to a Thermo Multidrop Combi (Waltham, MA) dispenser, where 5 µL of 10 µM (0.40 mg/mL) PTP1B_CA in 50 mM HEPES (pH 7.4), 100 mM NaCl, 2 mM DTT, 0.01% BGG, 0.001% Tween20, and 25 µM quality control (QC) peptide (assay buffer) were added to every well. The assay plate was then vortexed for 30 s at 1000 rpm on a Bioshake 3000 (Quantifoil Instruments, Jena, Germany) device, covered with a plastic lid, and incubated at 24 °C for 40 min in a humidified chamber. Just before the completion of the incubation step, a loaded 384-well filter plate was stacked onto an empty 384-well PP plate, and both were loaded into a centrifuge. After centrifugation for 2 min at 800 rcf, the PP plate was discarded, and the filter plate was placed in a CyBio Well vario liquid-handling system (Analytik Jena) equipped with a 384-well format and 40 µL head. After transportation of the assay plate to the multichannel pipettor, incubation solutions were quantitatively transferred to the filter plate. On completion of this step, the filter plate was immediately stacked onto an empty 384-well PP plate (elution plate), and the combined plates were loaded into the centrifuge and spun at 800 rcf for 2 min. Subsequent to the separation step, the filter plate was restored, and the elution plate containing the eluted samples was transported to a CyBio FeliX liquid-handling system (Analytik Jena). There, 3 µL of every well was transferred from the 384-well elution plate to one quadrant (Q1/Q2/Q3/Q4) of a 1536-well plate. After four consecutive transfers from four individual elution plates (elution plate 1 → Q1, elution plate 2 → Q2, etc.), 3 µL of a mixture containing internal standard (500 nM final concentration) and the standard for the SEC quality control (QCref; 250 nM final concentration) in acetonitrile were added to every well of the 1536-well plate using a Certus Flex Micro Dispenser (Gyger, Gwatt, Switzerland). Plates were then sealed and stored until preparation of the MALDI target plates. A flowchart of the automated binding assay and a Gantt chart of the process sequence are provided in the online supporting information ( Suppl. Fig. S2A and Suppl. Fig. S2B ).

Hit Confirmation Workflow

Process steps and plate handling were executed as described above. For hit confirmation experiments, 50 nL of 5 µg/µL or 10 mM compound solutions were transferred to the wells of two individual 384-well PP assay plates (sample plate and low control plate). A third plate received 50 nL DMSO per cavity (high control plate). Sample plates and high control plates were further processed by dispensing 5 µL of 0.4 mg/mL PTP1B_CA in assay buffer into every well of the plates, while low control plates received 5 µL BGG (0.4 mg/mL) instead. After incubation, mixtures were transferred to assay-ready filter plates using automated liquid handling. Filter plates containing samples or low controls were then stacked onto 384-well PP plates containing 50 nL DMSO per cavity, while high control–containing filter plates were stacked onto plates containing 50 nL of 10× diluted (0.5 µg/µL or 1 mM) compound solutions in DMSO. Plates were then centrifuged, and eluates were subsequently transferred to quadrants of a 1536-well microtiter plate (samples → Q1, low controls → Q2, and high controls → Q3). Q4 received 3 µL of 10 mM NH₄Ac buffer (pH 7.3) as negative control. Addition of standards and further plate processing were conducted as described above.

MALDI Target Preparation

4-HCCA was dissolved at 3 mg/mL in a 1:1 mixture of acetonitrile and 5 mM NH₄H₂PO₄/0.1% TFA (v/v) with the aid of vigorous vortexing. The CyBio Well vario liquid-handling system (Analytik Jena) equipped with ceramic tips and operated in 1536-well format was used to conduct dried droplet spotting, providing highly homogeneous spot shapes. Here, assay plates were centrifuged at 1000 rpm for 60 s, and the seals were removed before 100 nL matrix solution and 100 nL sample were aspirated successively from the matrix reservoir and the assay plate, respectively, and dispensed together onto the plain steel MALDI target plates. The target plates were dried in a vacuum chamber before the spotting process was repeated. Between the first and second spotting steps and after the second spotting step, the ceramic tips were washed three times with isopropanol/0.1% aqueous TFA (70/30, v/v). Following successful transfer and drying of matrix–analyte mixtures, an on-target washing procedure was applied using 10 mM NH₄H₂PO₄/0.3% TFA. Processed MALDI target plates were stored until analyzed. After every preparation cycle, three repetitive washing cycles using 0.1 M NH₄OH followed by three cycles of 70% isopropanol and 30% water/0.1% TFA (v/v) were carried out on the liquid-handling system to prevent carryover and clogging of the tips.

MALDI-TOF-Based HTS Readout

Mass spectra were acquired with a rapifleX MALDI-TOF/TOF instrument from Bruker Daltonics, including a Smartbeam 3D laser. FlexControl (v. 4.0), flexAnalysis (v. 4.0), and rapifleX MALDI PharmaPulse (MPP; v. 2.2), all from Bruker Daltonics, were used for MS acquisition and data analysis. Target plates were loaded onto an Orbitor RS (Thermo Fisher Scientific) robotic system controlled by the laboratory automation software Momentum (v. 4.2.3, Thermo Fisher Scientific) and automatically inserted into the MALDI-TOF device. Mass spectra were acquired in the mass range of m/z 200–1100 to cover target analyte, internal standard, and controls of the binding assay. Therefore, 2500 laser shots per sample spot were accumulated in 20-shot increments across a small spiral spot raster in positive or negative ionization mode using a 160 ns pulsed ion extraction with a 10 kHz laser frequency, a digitizer setting of 5.00 GS/s, and an M5 defocus Smartbeam parameter at a 50×50 µm scan range, which resulted in a 124×124 µm field size. The laser power was adjusted manually before every start of a batch process to reach a sufficient signal intensity for the internal standard (~2×10⁵ cts in positive ionization mode; ~5×10⁴ cts in negative ionization mode). The acquired spectra were processed in MPP2.2 Synthesis Screening mode or using flexAnalysis with a centroid peak detection set to a signal-to-noise ratio (S/N) ≥ 15 and a peak width of m/z 0.05.

Data Analysis

For data evaluation, signal area values [area under the curve (AUC)] extracted from flexAnalysis result files were referenced against corresponding standard signals for small molecules [internal standard (IS)] and quality controls (QCref), and reported as ratio values. When spectra were processed by MPP2.2 Synthesis Screening mode, ratio values for detected compound signals and QC signals were extracted straight from result files. In all cases, signals measured with an m/z value deviating more than 0.032 from the calculated theoretical value were rejected and not used for further calculations. The ratio of QC peptide signal and QC standard signal had to stay lower than the threshold of 1.0 (representative for 1% SEC leakage) for measured samples to be accountable for further analysis. For signals within the mass deviation threshold, the proportional increase of the signal ratio in sample wells compared to the mean of negative control–containing wells was calculated and expressed as Fold Change (FC) values, according to the following equation:

Fold Change (FC) = \frac{{Ratio}_{Cpd}}{mean {Ratio}_{Negative control}}

(1)

Small-molecule analyte signals reaching FC values of ≥2.5 were considered as hits in the screening workflow or in compound detectability analysis. In cases in which signals at compound m/z were observable in sample measurements but not in negative controls, the FC value calculation was omitted, and the compound was counted as an assay hit. Measurements failing the QC threshold or not containing any measurable IS were reported together with compound information and aggregated in a dropout list to enable retesting in a separate assay run.

Data from hit confirmation workflow experiments were analyzed in a similar fashion as described for screening workflow data. Ratio values were calculated for compound and QC signals from Exp. A (sample), Exp. B (low control), and Exp. C (high control) measurements. In addition, a FC value was calculated from high control compound signals and the mean of observable negative control ratio values. Relative binding affinities (RBAs) for specific compound–protein interactions were calculated via the following equation:

RBA [%] = \frac{{Ratio}_{Cpd} [E x p . A] - {Ratio}_{Cpd} [E x p . B]}{{Ratio}_{Cpd} [E x p . C]}

(2)

To qualify as a hit in the hit confirmation workflow, test compound data had to fulfill the following criteria: m/z deviation ≤ 0.032, internal standard detected in all three experiments (A–C), QC passed (QC ratio ≤ 1.0) in all three experiments, FC ≥ 2.5, and RBA ≥ 5%. Data were further processed using GraphPad Prism (v. 8.30; GraphPad Software, La Jolla, CA). A more detailed description on data handling and processing is provided online in the supporting information. The assignment of compounds to the corresponding measurements was achieved by software-aided deconvolution of every 1536-well assay plate to the corresponding 384-well substance plates in the course of data analysis, as we have described in Ref. ²¹.

Further Methods

A summary table compiling information on components and controls of the presented MALDI-TOF AS-MS concept targeting PTP1B ( Suppl. Table S1 ) and method information on the MALDI-TOF MS calibration procedure, the loading of SEC resin to 384-well filter plates, AlphaScreen experiments, differential scanning fluorimetry experiments, the MALDI-TOF-based PTP1B activity assay, isothermal titration calorimetry experiments, NMR spectroscopy, and further data handling and processing can be found in the Supplemental Methods section of the Supplemantal Material.

Results and Discussion

MALDI-TOF AS-MS Workflows

SEC-based separation performed directly in microtiter plates is a straightforward and effective approach that is amenable to automation on current screening platforms. To enable screening of small-molecule libraries for noncovalent binding to a protein target, two workflows were designed based on an in-plate SEC-enabled separation procedure in 384-well format. A schematic depiction of both workflows is provided in Figure 1 . In the first workflow (Screening workflow), individual incubations of protein and test compounds are transferred to SEC resin-filled 384-well filter plates, and the components are separated according to the size exclusion cutoff on centrifugation. Compounds interacting with the target pass the SEC resin and can be detected in subsequent analysis. To control for efficient cutoff separation, a quality control (cutoff QC) is added to the assay buffer at a fixed concentration, which allows for quantification of SEC leakage. To account for intrinsic MALDI-TOF signal variabilities, two reference standards are added to every sample after elution from the filter plate. While the IS is referenced against compound signals (Ratio_sample), the quality control standard (QCref) defines the threshold for acceptable SEC leakage and is used to precisely quantify the amount of QC in eluted sample fractions. Compounds detected at levels higher than certain thresholds in sample wells compared to DMSO control wells are considered a screening workflow hit. Because this first workflow provides only a qualitative measure of protein binding and does not account for unspecific interactions, a second assay (Hit confirmation workflow) including controls for every test compound was implemented. Here, in addition to the regular sample incubation of test compound and protein target, a low control that detects unspecific binding and a high control that resembles the maximum achievable compound concentration after elution are included. While for low controls the protein of interest is substituted for an unrelated protein or an unfavored protein isotype, the high control is performed with the target but in the absence of test compound. After elution from the filter plate, dilutions of test compounds are spiked into high controls at concentrations that mimic full 1:1 target saturation. After analysis, controls are used to calculate the specific RBA, which is a quantitative measure of specific protein binding. Combining the two presented workflows conceptually enables screening of small-molecule compound libraries for qualitative target binding followed by stringent hit verification and relative quantification of ligand affinity of primary screening hits using a label-free MALDI-TOF MS readout. Following workflow design, we evaluated critical method parameters for the SEC-based in-plate separation process and for MALDI-TOF MS-enabled small-molecule detection.

Figure 1.

Scheme of matrix-assisted laser desorption/ionization time-of-flight (MALDI-TOF) affinity selection mass spectrometry (AS-MS) assay workflows. (A) Screening workflow: Individual incubation solutions are transferred from 384-well assay plates onto size exclusion chromatography (SEC) resin-loaded 384-well filter plates. Protein and protein–ligand complexes are separated from unbound components via centrifugation, and eluates are captured. After addition of standards, samples are spotted onto MALDI target plates and analyzed by MALDI-TOF MS. Compounds binding to the protein target (Case 1) produces mass specific signals in MS analysis compared to samples in which the compound is retained by the SEC resin (Case 2). Intactness of the SEC-based separation procedure is monitored by quantifying a quality control (QC) peptide to detect false positives caused by inefficient separation (Case 3). (B) Hit confirmation workflow: In addition to regular sample incubation (Exp. A), controls containing either an unrelated target (Exp. B) or pure DMSO instead of test compound (Exp. C) are prepared. In the case of high controls, compound is spiked to eluted high controls to simulate full target binding. The amount of protein binding for samples (Exp. A) is then quantified in relation to low and high controls, respectively.

SEC-Based In-Plate Separation Process

Efficient retention of unbound ligand and buffer components is a prerequisite for robust data generation by the two presented workflows. Optimization of SEC separation efficiency was carried out with buffer containing fluorescein as a tracer. A bed volume of 65 µL SEC resin per well caused >99.5% retention of the fluorescent tracer and was used for further experiments ( Suppl. Fig. S1A ). To consistently track the separation capacity of the assay, a control peptide (QC; ETDYYRKG-amide; M = 1029.458 Da) and its reference standard (QCref; ETDYYR[K]G-amide; M = 1039.488 Da) with mass values well outside the typical mass window for small-molecule analytes were selected. Linearity of QC concentration-dependent signal response was confirmed by MALDI-TOF MS, and a lower detection limit (LoD) of 78 nM was assessed ( Suppl. Fig. S1B ). The threshold for maximum acceptable SEC leakage was set to 1%. Bradykinin2-9 (PPGFSPFR; M = 903.455 Da) served as a general internal standard. Stable retention of >99% was revealed for the selected setup by challenging a fully SEC resin-loaded filter plate with either 1 µM fluorescein or 100 µM QC peptide ( Fig. 2A ). Following demonstration of efficient component retention, protein recovery after SEC passage was investigated for proteins of different molecular weight. Using SEC resin with a 10 kDa molecular cutoff, average protein recovery reached around 70% for proteins exceeding 14 kDa in size ( Suppl. Fig. S1C ). Together, the data indicated robust and trackable retention of assay components and stable size-dependent protein recovery in eluates after centrifugation.

Figure 2.

Evaluation of in-plate size exclusion chromatography (SEC) and matrix-assisted laser desorption/ionization time-of-flight (MALDI-TOF) mass spectrometry (MS) small-molecule detection. (A) Retention profile of eluates from a 384-well filter plate filled with 65 µL SEC bed volume per cavity and challenged with 10 µL of 1 µM fluorescein solution or 100 µM quality control (QC) peptide, respectively. (B) Example spectra of a seven-component mixture of five small molecules (C01–C05) and the standards IS (internal standard) and QCref (the SEC quality control) (500 nM each). Blank spectra of pure co-crystallization matrix (black) are overlaid with sample spectra (red). Signals used for internal calibration are marked with asterisks. (C) Comparison of detection rates for small-molecule analytes in sets of 973 and 4821 compounds, respectively, using optimized MALDI matrix composition and spotting parameters.

Automation of the in-plate SEC procedure was a prerequisite for handling larger compound numbers and enabling HTS-capable sample processing. We therefore transferred the screening workflow to our in-house HTS system comprising components of regular automation platforms for process steps that include liquid handling, sample mixing, incubation, centrifugation, robotic plate handling, plate storage, and plate sealing. A flow chart of the automated process is provided in the supporting information ( Suppl. Fig. S2A ). Because protein–ligand complex dissociation during SEC due to fast k_off kinetics can compromise the detection of respective ligands, a rapid separation of the protein–ligand complex from nonbinding components after sample transfer to filter plates is crucial. The automation protocol was therefore established, with special focus on the time needed for stacking the 384-well filter plates onto 384-well collection plates after sample transfer and transporting the plate stacks to the centrifuge. A transfer time of <25 s was achieved, thus ensuring fast component separation and therefore minimizing the impact of fast dissociation kinetics on ligand identification. To enable a more comprehensive MALDI-TOF MS analysis, eluates from four individual filter plates were consecutively transferred into quadrants (Q1–Q4) of a 1536-well plate, which was further used for MALDI target plate preparation. Simulating the assay on the automated system yielded iteration cycle times of 74 min per 1536-well plate, which translated into a theoretical throughput capacity of 29,890 samples per day per MALDI instrument (19 1536-well plates, 76 assay plates/filter plates, and 26,700 test compounds; 2.9 s/sample; Suppl. Fig. S2B ). We would like to note that additional process optimization, parallelization of process steps, and the introduction of filter plates in higher-density format together hold the potential for further increasing the daily throughput of the platform. Here, we demonstrate that HTS-capable sample preparation by automated in-plate SEC is achievable on a standard automation platform without major modifications.

Optimization of MALDI Matrix Composition and Deposition for Small-Molecule Detection

Interferences in the low mass region arising from background signals caused by the matrix and chemotype-dependent low ionization efficiency have been major issues in MALDI-TOF MS analysis of small molecules. A multitude of approaches to minimize background interference while maximizing compound ionization have been reported, suggesting a variety of matrix constituents, additives, solid supports, and preparation procedures.^22–26 Many of these improvements, however, are valid only for specific analyte chemotypes, require special treatment of MALDI target plates, or interfere with automation processes. Therefore, a set of 10 different MALDI matrices, which have proven to be compatible with the applied automated spotting procedure, were evaluated for suitability. Co-crystallization with α-cyano-4-hydroxycinnamic acid (HCCA) produced acceptable S/N values for each of eight chemically diverse small molecules and the QC peptide (1 µM in demineralized H₂O) in positive and/or negative ionization mode ( Suppl. Fig. S3A ). Further optimization showed that best results were obtained using a solution of 3 mg/mL HCCA in acetonitrile/H₂O 1:1 containing 10 mM NH₄H₂PO₄ + 0.3% trifluoroacetic acid and applying a double dried droplet spotting procedure followed by an on-target washing step ( Suppl. Figs. S3B and S4 ). Because HCCA produces intense mass signals around m/z 190 (corresponding to [M+H]⁺ and [M-H]⁻ of the HCCA molecule), measurements were focused on a mass range of m/z 200–1100. Figure 2B depicts representative mass spectra recorded in positive and negative ionization mode for a five-component compound mixture plus IS and QC standard (QCref) peptides (500 nM each, in demineralized H₂O) overlaid with blank spectra recorded from the HCCA matrix. Prominent signals that could be assigned to individual components based on calculated mass and isotope pattern can be identified in the compound mix spectra compared to matrix background. To take advantage of remaining background signals and to increase mass accuracy, a five-point internal recalibration procedure was applied, using the signals of the standard peptides and three reproducible background signals originating from matrix ion species for positive and negative ionization mode, respectively. This allowed for clear differentiation of compound signals from matrix background with a mass accuracy of <30 ppm, which translated to a maximum mass deviation of 0.032 Da within the observed mass range.

ESI-based liquid chromatography (LC)/MS methods operating in positive ionization mode have been reported to be capable of detecting >85% of library compounds under screening conditions.¹¹ To achieve a better estimate of the analytical coverage accessible by MALDI-TOF MS, a chemically diverse selection of 973 compounds was tested for detectability at three different concentrations ranging from 1 to 100 µM ( Fig. 2C ). Detection rates increased with compound concentration from 72% (1 µM) to 90% (100 µM). At 10 µM, 85% of compounds were identified, and the detection rate was stable through repetitive testing during a period of up to 3 days ( Suppl. Fig. S5 ). To confirm these observations and to eventually draw conclusions from statistical analysis of physicochemical compound properties, a larger set of 4821 small-molecule substances was tested. Measurements revealed a comparable detection rate of 82% (3949 compounds) at 10 µM compound concentration. The vast majority of compounds was detectable in positive ionization mode, and only 2% (107 compounds) were exclusively detected under negative ionization conditions. A total of 3.3% of test set members had a molecular weight lower than 200 Da and were therefore not amenable to analysis. The results from both test sets are summarized in the table included in Figure 2C . Comparing the molecular weight, clogP value, fraction of sp³-hybridized atoms (FSP3), and number of hydrogen bond donors and acceptors for detected and nondetected small-molecule analytes did not reveal any major differences, but showed a mild trend toward higher FS3P values for detectable molecules ( Suppl. Fig. S6 ). This is consistent with the assumption that compounds carrying fewer aromatic moieties but more sp³-hybridized functional groups (amines, hydroxyl groups, and thiols) are more efficiently ionized during MALDI. We conclude that occasional incompatibility of individual small-molecule chemotypes with MALDI is unavoidable, and a consequential chance for false negatives has to be tolerated. Computational preselection of chemotype representatives with proven amenability to MALDI could likely minimize the detection bias and increase overall detection rates. The high proportion of detectable compounds achieved at reasonable compound concentrations in the low micromolar range implies, however, that chances to generate false negatives are manageable and are outweighed by the potential advantages of performing label-free HTS for ligands of orphan targets. The combined results for the two compound sets demonstrate MALDI-enabled detection rates for small-molecule screening collection representatives that, at the concentrations used, approach those reported for ESI-based methods. This emphasizes the suitability of MALDI-TOF MS as a readout strategy for small-molecule detection.

Comparison of MALDI-TOF AS-MS to Alternative Assay Formats

We have previously demonstrated that MALDI-TOF MS is a versatile drug discovery readout for targeting the enzymatic activity of human protein tyrosine phosphatase PTP1B.²⁷ To investigate whether the technology can be used as a readout strategy for in vitro target engagement analysis, we selected PTP1B as a model system to set up a MALDI-TOF MS-enabled binding assay, according to the described workflows. Here, an active site mutant of PTP1B (PTP1B_CA) was used as the target of interest, and BGG was chosen as an unrelated protein for low control experiments. Two previously published PTP1B inhibitors were used as reference compounds (Ref_Cpd.), and their binding toward PTP1B_CA was confirmed by ITC and NMR experiments ( Suppl. Figs. S7 and S8 ).^28–30 Ref_Cpd. 1 is a phosphotyrosine mimic and orthosteric ligand of the PTP1B active site, whereas Ref_Cpd. 2 has been described as an allosteric PTP1B modulator preventing the enzyme from adopting an active conformation. Together, these reference compounds served as tools to evaluate the capability of the assay system to detect ligands with different binding modes. To ensure uncompromised control of SEC integrity, the absence of intrinsic binding affinity of the QC peptide to PTP1B_CA was confirmed by DSF ( Suppl. Fig. S9A ). To ensure high detection rates of MALDI-TOF MS hits and sufficient assay sensitivity, analyte concentrations in the low micromolar range were required, as assessed by compound detection analysis. Therefore, a protein concentration of 10 µM (0.4 mg/mL for 40 kDa PTP1B_CA) and a test ligand concentration of 100 µM were selected. The high protein demand of the assay format was partially compensated by a small assay volume (5 µL/sample). With these settings, the applicability of the MALDI-TOF AS-MS concept for PTP1B was tested using Ref_Cpd. 1 as a reference ligand.

First, the elution profile of a 384-well filter plate was recorded for 5 µL fractions after applying 100 µM Ref_Cpd. 1 alone or in mixture with 10 µM PTP1B_CA ( Suppl. Fig. S9B ). An average concentration of close to 7 µM was detected in the first fraction of protein containing samples but not in the absence of protein, again demonstrating efficient SEC-based separation and protein-dependent compound detection. Next, plate uniformity was assessed across eluates of a full 384-well assay plate containing incubation mixtures of 100 µM Ref_Cpd. 1 and 10 µM PTP1B_CA, which were were used to evaluate the influence of intrinsic MALDI-TOF signal variability on the assay readout ( Suppl. Fig. S9C ). Signals of Ref_Cpd. 1 showed a coefficient of variation (CV) of 34.3% around an average AUC of 12,191 (S/N = 593). Normalization of AUC values to the internal standard peptide decreased signal variability to CV = 16.2%, thereby providing an appropriate approach to correct for signal variations attributable to the MALDI process. Because the suitability of the selected IS likely varies for different small-molecule chemotypes, a higher signal CV compared to values typically observed in enzyme activity-targeted HTS was anticipated.

To assess data quality and to compare assay hits identified by MALDI-TOF AS-MS to orthogonal HTS-capable binding assay formats, the diversity set of 973 small molecules (971 compounds + 2 reference compounds) was subjected to the testing scheme outlined in Figure 3A . All assays were performed in biological duplicates on two individual days to ensure concise and representative data.

Figure 3.

Matrix-assisted laser desorption/ionization time-of-flight (MALDI-TOF) affinity selection mass spectrometry (AS-MS) comparison to alternative assay formats using protein tyrosine phosphatase 1B (PTP1B) as a model system. (A) Testing scheme for the set of 973 compounds. Results from MALDI-TOF AS-MS are compared to an AlphaScreen displacement assay and DSF analysis for the detection of compound target engagement. An activity assay is used to additionally evaluate the transfer of target binding into functional inhibition. (B) Venn diagram of hit sets identified in HTS-capable binding assay formats. Reference compounds are indicated in orange. (C) Comparison of results for the confirmed MALDI-TOF AS-MS hits (0.31% final hit rate) across different assay readouts. Values that comply with predefined hit criteria are highlighted in green. (D) Correlation of activity assay results for identified inhibitors. Hits from MALDI-TOF AS-MS (reference compounds = orange; JTT-551 = green) are also functional inhibitors of PTP1B activity.

For MALDI-TOF AS-MS, a total of 19 primary hits (1.95% hit rate) were identified, and ratio values showed acceptable correlation for individual experiments ( Suppl. Fig. S10 ), indicating reproducible hit identification. From these initial hits, 16 exhibited similar signals in samples and low controls, and only three compounds were confirmed in the hit confirmation workflow, which resulted in a reasonable final hit rate of 0.31% and demonstrated the efficiency of the combined workflows for hit triaging. The hit set of confirmed protein binders comprised both reference compounds (Ref_Cpd. 1, RBA = 96.9%; and Ref_Cpd. 2, RBA = 70.0%), demonstrating the capability of the assay to unambiguously identify ligands with orthosteric or allosteric binding mode ( Suppl. Fig. S11 ). Furthermore, the structurally unrelated small molecule JTT-551 (RBA = 84.2%), which has been reported as an inhibitor of PTP1B catalytic activity (K_i = 0.22 ± 0.04 µM) before, was confirmed as an assay hit.³¹ Although no proof of target engagement had been provided so far, comparing MALDI-TOF AS-MS spectra for JTT-551 samples, high controls, and low controls (n = 3, each) at m/z 606.282 ([M+H]⁺, 1. isotope) clearly indicated PTP1B_CA-dependent compound detection and therefore JTT-551 binding to its target protein ( Suppl. Fig. S11 ).

To compare results to a displacement assay format, an AlphaScreen technology-based assay was used, using a biotinylated tracer based on the structure of active-site ligand Ref_Cpd. 1. Compounds were also tested in a TruHits format that allowed for quantification of compound-induced signal quenching by using a biotinylated His-tag as a tracer–protein complex surrogate. AlphaScreen competition data were corrected for average signal quenching, and a hit threshold of 80% of controls [percentage of controls (PoC)] was defined. With this approach, 15 compounds were identified as hits (1.54% hit rate), among which JTT-551 and the orthosteric binder Ref_Cpd. 1 but not the allosteric binder Ref_Cpd. 2 were found ( Suppl. Fig. S12A ). This is consistent with the conceptual bias of a displacement assay in which only ligands preventing the binding of the specific tracer can be identified. In the case of the here-applied model system PTP1B, this led to the rejection of allosteric binders and created false-negative assay results.

As a second orthogonal binding assay, DSF was performed in detergent-free buffer using the hydrophobic SYPRO Orange dye. Test substances were tested for their effect on thermally induced protein denaturation, indicated by a shift of the protein melting temperature (ΔT_m). Screening the 973-compound library with a hit threshold of ΔT_m > 0.5 °C produced 21 hits (2.15% hit rate; Suppl. Fig. S12B ). Ref_Cpd. 1 showed the most pronounced temperature shift (ΔT_m = 10.6 °C), while JTT-551 was neutral and Ref_Cpd. 2 was found among a set of 13 test substances promoting strong protein destabilization (ΔT_m < −1.5 °C). Although protein destabilization usually points toward physicochemical protein perturbation, the specific interaction of Ref_Cpd. 2 with PTP1B has been proven via co-crystallization experiments.²⁹ The ligand binds to an allosteric pocket on the protein, trapping the enzyme in an inactive conformation. This conformation is characterized by reduced intramolecular interaction of the enzyme’s WPD loop with the active site, which could explain the observed decrease in ΔT_m on Ref_Cpd. 2 binding.³² As a result, Ref_Cpd. 2 produced a false-negative response in the DSF assay, underlining the shortcoming of this technology when screening for ligands with different binding modes.

To test whether protein binding also translates into functional inhibition, we performed a MALDI-TOF-based PTP1B activity assay for the compound test set. Substances were tested for inhibition of PTP1B-mediated tyrosine dephosphorylation. From the set of 973 compounds, 30 hits impeded catalytic activity by more than 20% (3.08% hit rate), including the two reference inhibitors ( Suppl. Fig. S12C ). A table listing the results of alternative assay formats (AlphaScreen, DSF, and MALDI activity) is provided in Supplemental Figure S12 .

Figure 3B shows a comparison of the hit sets generated by MALDI-TOF AS-MS, AlphaScreen, and DSF. Little agreement was found for hits identified by the three different technologies, with Ref_Cpd. 1 as the only mutual hit. Notably, allosteric protein binding by the known ligand Ref_Cpd. 2 was detectable solely by MALDI-TOF AS-MS, but not by alternative binding assay formats. The published inhibitor JTT-551 showed sufficient signal inhibition in the AlphaScreen assay and the functional readout, but did not promote any ΔT_m shift in DSF analysis. A possible reason could be insufficient compound solubility, because DSF analysis prohibits the use of detergents in buffer solutions. Together with the results from the functional assay, summarized in Figure 3C , the compiled data illustrate the capability of MALDI-TOF AS-MS to identify functional PTP1B ligands originating from different modes of action, which further underscores the competitive advantages of the presented platform. Comparing the distribution of MALDI-TOF AS-MS hits to the hit set of the functional enzyme activity assay showed reproducible inhibition in line with reported values (Ref_Cpd. 1 IC₅₀ = 0.12 µM;²⁸ Ref_Cpd. 2 IC₅₀ = 8 µM;²⁹ JTT-551 K_i = 0.22 µM;³¹ Fig. 3D ). The hit rate of the functional assay was, however, 10 times higher than for the binding assay (3.08% vs. 0.31%, respectively). This suggests that MALDI-TOF AS-MS not only is useful for primary screening but also could provide a powerful approach in combination with enzymatic screens. For example, binding assays for primary screening against enzymatic targets are inherently biased toward a binding hypothesis that defines the components included in an assay setup. In contrast to activity-based methods, such approaches tend to fail in identifying ligands with certain binding modes that require the presence of specific assay components for efficient protein binding (e.g., uncompetitive binding). Initial functional screening against the enzymatic activity followed by MALDI-TOF AS-MS hit confirmation could aid the identification of inhibitors with specific binding affinities toward the protein target and promote the deconvolution of confirmed hits according to their mode of action by testing their assay response in the presence and absence of respective components.

Because the MALDI-TOF AS-MS platform can conceptually be adjusted to any target with a size greater than the SEC cutoff, we assessed the versatility of the assay by simply exchanging the protein of interest for either cyclic guanosine monophosphate–adenosine monophosphate synthase (cGAS), carbonic anhydrase II, human serum albumin, or T-cell PTP1B isotype (TcPTP1B). Determining RBA values for corresponding reference compounds demonstrated apparent protein binding and allowed comparison with literature values ( Suppl. Fig. S13 ).^17,33–36 Together, the results propose MALDI-TOF AS-MS as an attractive supplementary technology for affinity-based screening.

HTS Validation Study

To evaluate the performance of the assay under HTS conditions and in a fully automated environment, we performed a 24-h screening run on our in-house automation platform with subsequent MALDI target spotting and MALDI-TOF MS analysis.

A total of 23,444 compounds were tested in the automated screening workflow at 100 µM for binding to 10 µM PTP1B_CA, including reference compounds. Figure 4A shows the distribution of ratio values for positive controls (Ref_Cpd. 1) on every 1536-well plate. Signals scattered around an average ratio value of 0.96 with a CV of 23.0%. This signal variation was slightly increased compared to our findings from plate uniformity analysis, which could be attributable to the limited number of controls per plate and the combination of four individual assay plates into 1536-well format. Two positive controls (<0.19%) were not detected during MALDI-TOF MS analysis due to flawed sample spotting. The predefined QC threshold of <1% SEC leakage failed in one instance (<0.01%), and the corresponding spectrum was excluded from analysis. Overall, robust and stable detection of qualitative protein binding was demonstrated throughout the campaign. 240 primary hits (1.02% hit rate) were identified from the compound library and subjected to the hit confirmation workflow. Here, primary hits were tested in duplicates, and the amount of PTP_CA binding was quantified against high controls (eluates spiked with 7 µM compound) and low controls (BGG instead of PTP_CA). From the pool of 240 hits, 229 were rejected because of interfering low control signals. Two compounds did not reproduce their assay response among duplicates and were therefore also excluded from the hit pool. From the remaining nine substances, four had relative binding affinities lower than 5% and were not considered significant. This left us with five confirmed hits (0.021% final hit rate) exhibiting conclusive and reproducible target engagement ( Fig. 4B ). Comparing determined RBA values for two independent experiments further demonstrated good correlation (R² = 0.97) of assay results ( Suppl. Fig. S14 ). Among the identified hits, both reference compounds were found along with three new ligands [compounds (Cpd.) 1–3] for which the binding to PTP1B was also confirmed by NMR experiments ( Suppl. Fig. S15 ). Chemical shift perturbations similar to reference compounds ( Suppl. Fig. S8 ) indicated orthosteric protein binding for Cpd. 1 and Cpd. 2 and an allosteric binding mode for Cpd. 3. A summary of the validation study results can be found in Figure 4C . Performing dose–response experiments instead of single-concentration testing with identified PTP1B_CA ligands demonstrated the general feasibility of the assay to measure concentration-dependent protein binding ( Fig. 4D ). We have found, however, that due to the high protein concentration required and the resulting limited assay sensitivity, in the case of PTP1B, RBA values correlate much better with a functional readout measuring PTP1B enzymatic activity than EC₅₀ values from dose–response experiments (R² = 0.94 vs. 0.35, respectively; Fig. 4E ). Therefore, repetitive single-concentration testing with adequate controls was preferred for quantifying protein–ligand interaction in the current setup.

Figure 4.

Results from the matrix-assisted laser desorption/ionization time-of-flight (MALDI-TOF) affinity selection mass spectrometry (AS-MS) validation project. (A) Distribution of ratio values for positive controls [reference compound (Ref_Cpd.) 1] on 1536-well (4×384-well) plates. Every plate contained 32 positive controls (n = 1072). 1070 controls were detected (99.81%) with a mean coefficient of variation (CV) of 23%. (B) Results of the hit confirmation of 240 primary assay hits (n = 2). Mean relative binding affinity (RBA) is plotted against the RBA deviation between experiments. Blue dashed lines indicate deviations that lead to complete inactivity in one of the experiments. Compounds with considerable signals in low control measurements are depicted in gray. Reference compounds are highlighted in orange, and reproducible assay hits (RBA ≥ 5%) are marked in green. (C) Summary of assay results for the MALDI-TOF AS-MS validation study. (D) AS-MS dose–response curves of confirmed protein tyrosine phosphatase 1B (PTP1B) ligands. Compounds were quantified against high controls [100% percentage of controls (PoC)] and low controls (0% PoC) assuming linear signal response (n = 3). (E) Correlation of binding assay data with functional inhibition for identified PTP1B ligands. Plotted are RBA values against inhibitory activity (relative inhibition) from single-concentration experiments (left) and EC₅₀ values against IC₅₀ values from dose–response experiments (right). Binding data were generated by MALDI-TOF AS-MS, and inhibition data by the MALDI-TOF activity assay. RBA values (R² = 0.94) show better correlation to functional inhibition than dose–response data (R² = 0.35).

In summary, the validation study demonstrated automated screening of more than 23,000 compounds for PTP1B target engagement per day by MALDI-TOF AS-MS. Stable and robust separation of assay components was achieved by automated in-plate SEC on a regular unmodified automation platform, as indicated by monitoring QC controls and reference compounds. The so-identified PTP1B ligands exhibited reproducible target binding, determined by RBA values, which correlated with functional inhibition of PTP1B enzymatic activity. Screening single-compound incubation mixtures allowed for the convenient integration of the platform into our regular HTS environment. Together, we provide evidence that the presented assay platform is an HTS-compatible tool, built exclusively from commercially available components, for providing high-quality data on small-molecule target engagement. MALDI-TOF AS-MS is easy to adapt, as demonstrated for a number of diverse proteins, and does not require any knowledge about target function or endogenous ligands, which makes it especially suited for addressing targets for which no detection reagents are available or multiple binding epitopes could be captured. We aim to use the entire concept mainly for filtering large compound sets for their interaction with a selected target protein. The presented data suggest that ligand ranking by RBA might be possible, but detailed characterization by biophysical reference methods (e.g., ITC, SPR, and NMR spectroscopy) should follow the AS-MS concept for accurate ranking of binding affinities. As demonstrated, the hit rate of the AS-MS concept is quite low, so the biophysical methods can cope with the resulting ligand numbers.

Challenges of the MALDI-TOF AS-MS platform are the potential occurrence of false-negative assay results due to the incompatibility of ~15% of analytes with MALDI and micromolar analyte concentrations required for stable hit detection, consequently resulting in a high protein demand of the assay. While the use of focused compound libraries with proven MALDI amenability would simply exclude potential false negatives, possible enhancements of detection rates and method sensitivity could be achieved by the further development of matrices and additives. In addition, the commercialization of matrix-independent MALDI target plates, which omit the occurrence of interfering background signals, holds great potential for improving small-molecule detection by MALDI-TOF MS.³⁷ Furthermore, the detectability of binding ligands is dependent on the k_off constant of the respective protein–ligand complex, which requires fast separation of assay components to minimize this bias. By demonstrating the transferability of the presented concept to other protein targets, we were able to show specific binding of known ligands with binding affinities in the low-double-digit micromolar range ( Suppl. Fig. S13 ), which represents the expected upper limit of detectable binding affinity in line with earlier reports using SEC.^7,9,11 For a detailed investigation of detection limits with respect to binding affinities and especially k_off constants, a set of characterized ligands would be required. With regard to the target protein, it is of note that, prior to screening, careful QC of the expression construct should be carried out when applicable to ensure correct protein folding and high fractional functionality, because dysfunctional protein would lower the effective protein concentration and therefore the sensitivity of the assay.

At the moment, MALDI-TOF AS-MS is a fast methodology to screen in high-throughput fashion for protein binders, whereas the suitability of ESI-based high-resolution MS coupled with emerging HTS-capable sampling technology as a readout strategy will be the subject of future investigations.^38–40 Higher method sensitivity would likely result in a concurrent reduction of protein demand, which could be further decreased by assay miniaturization beyond the 384-well format. For projects in which protein supply is limited, screening with less protein-intensive technologies followed by MALDI-TOF AS-MS-enabled hit confirmation is an attractive option for ensuring the identification of high-quality hits with proven target engagement, which is particularly applicable when addressing enzymatic targets.

Conclusively, MALDI-TOF AS-MS is a versatile expansion of our existing MALDI-TOF MS screening platform and will be helpful in upcoming drug discovery campaigns by providing the capability to perform automated single-incubation HTS for in vitro target engagement to native targets in a label-free fashion and embedded in regular HTS infrastructure.

Supplemental Material

Supplemental_Material_MALDI_TOF_ASMS_Simon_et_al – Supplemental material for MALDI-TOF-Based Affinity Selection Mass Spectrometry for Automated Screening of Protein–Ligand Interactions at High Throughput

Supplemental material, Supplemental_Material_MALDI_TOF_ASMS_Simon_et_al for MALDI-TOF-Based Affinity Selection Mass Spectrometry for Automated Screening of Protein–Ligand Interactions at High Throughput by Roman P. Simon, Martin Winter, Carola Kleiner, Lucie Wehrle, Michael Karnath, Robert Ries, Markus Zeeb, Gisela Schnapp, Dennis Fiegen, Tim T. Häbe, Frank Runge, Tom Bretschneider, Andreas H. Luippold, Daniel Bischoff, Wolfgang Reindl and Frank H. Büttner in SLAS Discovery

Footnotes

Acknowledgements

We thank Meike Hamester and Arndt Asperger from Bruker Daltonics (Bremen, Germany) for their valuable input and helpful discussion. Furthermore, we thank Margit Bauer for the preparation of the PTP1B catalytic domain and Christofer Tautermann for his valuable support.

Declaration of Conflicting Interests

The authors declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: All authors are employees of Boehringer Ingelheim Pharma GmbH & Co. KG and completed this work within the scope of their employment.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: All work was completed as work-for-hire for the employer Boehringer Ingelheim Pharma GmbH & Co. KG.

ORCID iDs

Dennis Fiegen

Tom Bretschneider

Supplemental material is available online with this article.

References

Schiebel

Radeva

Krimmer

S. G.

, et al. Six Biophysical Screening Methods Miss a Large Proportion of Crystallographically Discovered Fragment Hits: A Case Study. ACS Chem. Biol. 2016, 11, 1693–1701.

Glickman

J. F.

Schmid

Ferrand

Scintillation Proximity Assays in High-Throughput Screening. Assay Drug Dev. Technol. 2008, 6, 433–455.

Lea

W. A.

Simeonov

Fluorescence Polarization Assays in Small Molecule Screening. Expert Opin. Drug Discov. 2011, 6, 17–32.

Rossato

Miralles

M’Kadmi

, et al. Quantitative MALDI-MS Binding Assays: An Alternative to Radiolabeling. ChemMedChem 2016, 11, 2582–2587.

Pantoliano

M. W.

Petrella

E. C.

Kwasnoski

J. D.

, et al. High-Density Miniaturized Thermal Shift Assays as a General Strategy for Drug Discovery. J. Biomol. Screen. 2001, 6, 429–440.

Cimmperman

Baranauskiene

Jachimoviciute

, et al. A Quantitative Model of Thermal Stabilization and Destabilization of Proteins by Ligands. Biophys. J. 2008, 95, 3222–3231.

O’Connell

T. N.

Ramsay

Rieth

S. F.

, et al. Solution-Based Indirect Affinity Selection Mass Spectrometry—A General Tool for High-Throughput Screening of Pharmaceutical Compound Libraries. Anal. Chem. 2014, 86, 7413–7420.

Annis

D. A.

Nickbarg

Yang

, et al. Affinity Selection-Mass Spectrometry Screening Techniques for Small Molecule Drug Discovery. Curr. Opin. Chem. Biol. 2007, 11, 518–526.

Annis

A. D.

Athanasopoulos

Curran

P. J.

, et al. An Affinity Selection–Mass Spectrometry Method for the Identification of Small Molecule Ligands from Self-Encoded Combinatorial Libraries: Discovery of a Novel Antagonist of E. coli Dihydrofolate Reductase. Int. J. Mass Spectrom. 2004, 238, 77–83.

10.

Adam

G. C.

Meng

Athanasopoulos

, et al. Affinity-Based Ranking of Ligands for DPP-4 from Mixtures. Bioorg. Med. Chem. Lett. 2007, 17, 2404–2407.

11.

Zehender

Le Goff

Lehmann

, et al. SpeedScreen: The “Missing Link” between Genomics and Lead Discovery. J. Biomol. Screen. 2004, 9, 498–505.

12.

Bretschneider

Ozbal

Holstein

, et al. RapidFire BLAZE-Mode Is Boosting ESI-MS toward High-Throughput-Screening. SLAS Technol. 2019, 24, 386–393.

13.

Heap

R. E.

Hope

A. G.

Pearson

L. A.

, et al. Identifying Inhibitors of Inflammation: A Novel High-Throughput MALDI-TOF Screening Assay for Salt-Inducible Kinases (SIKs). SLAS Discov. 2017, 22, 1193–1202.

14.

Gurard-Levin

Z. A.

Scholle

M. D.

Eisenberg

A. H.

, et al. High-Throughput Screening of Small Molecule Libraries Using SAMDI Mass Spectrometry. ACS Comb. Sci. 2011, 13, 347–350.

15.

De Cesare

Johnson

Barlow

, et al. The MALDI-TOF E2/E3 Ligase Assay as Universal Tool for Drug Discovery in the Ubiquitin Pathway. Cell Chem. Biol. 2018, 25, 1117–1127 e4.

16.

Winter

Bretschneider

Thamm

, et al. Chemical Derivatization Enables MALDI-TOF-Based High-Throughput Screening for Microbial Trimethylamine (TMA)-Lyase Inhibitors. SLAS Discov. 2019, 24, 766–777.

17.

Simon

R. P.

Winter

Kleiner

, et al. MALDI-TOF Mass Spectrometry-Based High-Throughput Screening for Inhibitors of the Cytosolic DNA Sensor cGAS. SLAS Discov. 2020, 25, 372–383.

18.

Hofner

Merkel

Wanner

K. T.

MS Binding Assays with MALDI toward High Throughput. ChemMedChem 2009, 4, 1523–1528.

19.

Hannewald

Maunit

Muller

J. F.

Tubulin-Binding Drug Screening by MALDI-TOF MS. Anal. Chem. 2006, 78, 4390–4397.

20.

VanderPorten

E. C.

Scholle

M. D.

Sherrill

, et al. Identification of Small-Molecule Noncovalent Binders Utilizing SAMDI Technology. SLAS Discov. 2017, 22, 1211–1217.

21.

Winter

Ries

Kleiner

, et al. Automated MALDI Target Preparation Concept: Providing Ultra-High-Throughput Mass Spectrometry-Based Screening for Drug Discovery. SLAS Technol. 2019, 24, 209–221.

22.

Calvano

C. D.

Monopoli

Cataldi

T. R. I.

, et al. MALDI Matrices for Low Molecular Weight Compounds: An Endless Story? Anal. Bioanal. Chem. 2018, 410, 4015–4038.

23.

McCombie

Knochenmuss

Small-Molecule MALDI Using the Matrix Suppression Effect to Reduce or Eliminate Matrix Background Interferences. Anal. Chem. 2004, 76, 4990–4997.

24.

Persike

Karas

Rapid Simultaneous Quantitative Determination of Different Small Pharmaceutical Drugs Using a Conventional Matrix-Assisted Laser Desorption/Ionization Time-of-Flight Mass Spectrometry System. Rapid Commun. Mass Spectrom. 2009, 23, 3555–3562.

25.

Zhu

Shen

Wang

, et al. An Auxiliary Matrix for Routine Analysis of Small Molecules and Biological Macromolecules Using Matrix-Assisted Laser Desorption Ionization Mass Spectrometry. Anal. Bioanal. Chem. 2019, 411, 1041–1052.

26.

Roth

M. J.

Maresh

E. M.

Plymire

D. A.

, et al. Surface Preparation Strategies for Improved Parallelization and Reproducible MALDI-TOF MS Ligand Binding Assays. ACS Appl. Mater. Interfaces 2013, 5, 6–10.

27.

Winter

Bretschneider

Kleiner

, et al. Establishing MALDI-TOF as Versatile Drug Discovery Readout to Dissect the PTP1B Enzymatic Reaction. SLAS Discov. 2018, 23, 561–573.

28.

Han

Belley

Bayly

C. I.

, et al. Discovery of [(3-bromo-7-cyano-2-naphthyl)(difluoro)methyl]phosphonic Acid, a Potent and Orally Active Small Molecule PTP1B Inhibitor. Bioorg. Med. Chem. Lett. 2008, 18, 3200–3205.

29.

Wiesmann

Barr

K. J.

Kung

, et al. Allosteric Inhibition of Protein Tyrosine Phosphatase 1B. Nat. Struct. Mol. Biol. 2004, 11, 730–737.

30.

Pervushin

K. V.

Wider

Wuthrich

Single Transition-to-Single Transition Polarization Transfer (ST2-PT) in [15N,1H]-TROSY. J. Biomol. NMR 1998, 12, 345–348.

31.

Fukuda

Ohta

Sakata

, et al. Pharmacological Profiles of a Novel Protein Tyrosine Phosphatase 1B Inhibitor, JTT-551. Diabetes Obes. Metab. 2010, 12, 299–306.

32.

Zhang

, et al. The Mechanism of Allosteric Inhibition of Protein Tyrosine Phosphatase 1B. PLoS One 2014, 9, e97668.

33.

Hall

Brault

Vincent

, et al. Discovery of PF-06928215 as a High Affinity Inhibitor of cGAS Enabled by a Novel Fluorescence Polarization Assay. PLoS One 2017, 12, e0184843.

34.

Iyer

Barrese

A. A.

3rd Parakh

, et al. Inhibition Profiling of Human Carbonic Anhydrase II by High-Throughput Screening of Structurally Diverse, Biologically Active Compounds. J. Biomol. Screen. 2006, 11, 782–791.

35.

Hasegawa

Hara

Hata

Binding of Dorzolamide and Its Metabolite, N-Deethylated Dorzolamide, to Human Erythrocytes In Vitro. Drug Metab. Dispos. 1994, 22, 377–382.

36.

Bogdan

Pirnau

Floare

, et al. Binding Interaction of Indomethacin with Human Serum Albumin. J. Pharm. Biomed. Anal. 2008, 47, 981–984.

37.

Yang

, et al. Nanomaterials as Assisted Matrix of Laser Desorption/Ionization Time-of-Flight Mass Spectrometry for the Analysis of Small Molecules. Nanomaterials (Basel) 2017, 7, 87.

38.

Sinclair

Bachman

Addison

, et al. Acoustic Mist Ionization Platform for Direct and Contactless Ultrahigh-Throughput Mass Spectrometry Analysis of Liquid Samples. Anal. Chem. 2019, 91, 3790–3794.

39.

Haebe

Liu

Covey

, et al. Ultrahigh-Throughput ESI-MS: Sampling Pushed to Six Samples per Second by Acoustic Ejection Mass Spectrometry. chemRxiv 2020. doi:10.26434/chemrxiv.12093918.v1.

40.

DiRico

K. J.

Hua

Liu

, et al. Ultra-High-Throughput Acoustic Droplet Ejection-Open Port Interface-Mass Spectrometry for Parallel Medicinal Chemistry. ACS Med. Chem. Lett. 2020, 11, 1101–1110.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

5.18 MB