Development of a High-Throughput Assay to Identify Inhibitors of ENPP1

Abstract

The innate immune response to cancer is initiated by cytosolic DNA, where it binds to cGAS and triggers type I interferon (IFN) expression via the STING receptor, leading to activation of tumor-specific T cells. Ectonucleotide pyrophosphatase/phosphodiesterase 1 (ENPP1) has been identified as the primary enzyme responsible for degrading cGAMP, and therefore it is under intense investigation as a therapeutic target for cancer immunotherapy. ENPP1 hydrolyzes cGAMP to produce AMP and GMP, and hydrolyzes ATP and other nucleotides to monophosphates and pyrophosphate. We developed a robust, high-throughput screening (HTS)-compatible enzymatic assay method for ENPP1 using the Transcreener AMP²/GMP² Assay, a competitive fluorescence polarization (FP) immunoassay that enables direct detection of AMP and GMP in a homogenous format. The monoclonal antibody used in the Transcreener AMP²/GMP² Assay showed more than 104-fold selectivity for AMP and GMP versus cGAMP, and 3000-fold selectivity for AMP over ATP, indicating that the assay can be used for detection at initial velocity with either substrate. A working concentration of 100 pM ENPP1 was determined as optimal with a 60 min reaction period, enabling screening with very low quantities of enzyme. A Z′ value of 0.72 was determined using ATP as substrate, indicating a high-quality assay. Consistent with previous studies, we found that ENPP1 preferred ATP as a substrate when compared with other nucleotides like GTP, ADP, and GDP. ENPP1 showed a 20-fold selectivity for 2′3′cGAMP compared with 2′3′c-diGMP and showed no activity with 3′3′c-diAMP. The Transcreener AMP²/GMP² Assay should prove to be a valuable tool for the discovery of ENPP1 lead molecules.

Introduction

Pattern recognition receptors (PRRs) recognize distinct molecular patterns from pathogens or damage patterns from host cells and induce a signaling cascade ultimately resulting in the production of cytokines and type I interferons (IFNs). ¹ The primary PRR responsible for recognizing cytoplasmic DNA derived from external or internal sources is the enzyme cyclic GMP-AMP synthase (cGAS), which, upon binding DNA, dimerizes and produces 2′3′cyclic GMP-AMP (cGAMP) from ATP and GTP.^2,3 cGAMP binds to STING (STimulator of Interferon Genes), initiating the pathway for type I IFN induction ⁴ via phosphorylation of transcription factors IRF3 and NFκβ, followed by their transport into the nucleus. Activation of the cGAS/STING pathway by tumor cell DNA is required for induction of tumor immunity, ⁵ and STING agonists are under intense investigation for cancer immunotherapy.^4–7

Ectonucleotide pyrophosphatase/phosphodiesterase 1 (ENPP1) is a transmembrane glycoprotein that was first identified as a negative regulator of bone mineralization. ⁸ ENPP1 hydrolyzes ATP to AMP and pyrophosphate, which inhibits hydroxyapatite deposition, thereby negatively regulating bone mineralization. ⁸ ENPP1 is expressed in diverse tissues and plays a critical role in purinergic signaling, regulating neurological, immunological, hormonal, musculoskeletal, and cardiovascular functions in mammals.^9,10 ENPP1 preferentially hydrolyzes ATP, but it also hydrolyzes other nucleotides, including GTP, ADP, and GDP.^11,12

ENPP1 garnered lot of attention as a potential cancer drug target when it was found to be expressed in lymphoid tissue, where it contributed to the production of adenosine, an immunosuppressive signal in the tumor microenvironment.^7,11–13

An additional role of ENPP1 in tumor immunity was suggested when the enzyme was found to hydrolyze cGAMP,^12,13 attenuating the cGAS/STING-driven type I IFN response. Additionally, hydrolysis of cGAMP by ENPP1 produces AMP, which leads to immunosuppression following dephosphorylation to adenosine by CD73. Knocking out ENPP1 in mice resulted in a much longer half-life for cGAMP and increased activation of STING, and it has been reported to impart resistance to carcinogenesis and metastasis.^13,14 Therefore, blocking ENPP1 is an attractive strategy for cancer immunotherapy, and efforts to develop small-molecule antagonists are underway.^6–8,14 Though they suffer from hypercalcification, ENPP1^–/– mice are viable, suggesting that it is a reasonable anticancer target. ¹⁴

A number of assay methods have been developed for ENPP1, but none of them allow direct detection of products formed from native substrates. Synthetic probe substrates have been developed for colorimetric assay (405 nm) and fluorescent assay (520 nm emission). ¹⁵ However, the use of nonnative substrates raises the question of whether inhibitors will exhibit the same potency as in the presence of native substrates. In addition, interference from optically active screening compounds can be problematic for colorimetric assays and fluorescent assays at wavelengths less than far red.^15–17 The AMP Glo method can be employed, but its limited sensitivity and the use of multiple coupling enzymes make it susceptible to interference. ⁷ Here, we describe a direct, high-throughput, sensitive, and robust ENPP1 assay based on the selective detection of AMP and GMP with fluorescent polarization readout^18,19 using the Transcreener AMP2/GMP2 Assay.

Materials and Methods

Enzymes, Substrates, and Reagents

Recombinant human ENPP1 protein was obtained from R&D Systems (Minneapolis, MN, cat. no. 6136-EN). Suramin was purchased from Sigma (St. Louis, MO, cat. no. 574625). Transcreener AMP²/GMP2 Assay kits were obtained from BellBrook Labs (Madison, WI). ATP and ADP were obtained from Sigma Aldrich (cat. nos. A2383 and A5285). 2′3′cGAMP, 3′3′c-di-AMP, and 2′3′c-di-GMP were obtained from InvivoGen (San Diego, CA, cat. nos. tlrl-nacga-23, tlrl-nacda-33, and tlrl-nacdg-23). ENPP1 Inhibitor C was purchased from Cayman Chemical (Ann Arbor, MI, cat. no. 29809).

Instrumentation and Analysis

A PHERAstar Plus plate reader (BMG Labtech, Cary, NC) was used for fluorescence polarization (FP) measurements in black, nonbinding, low-volume, 384-well plates (Corning, Corning, NY cat. no. 4514). After adding all the enzyme components, the reaction was mixed in an orbital shaker for 30 s. The FP assays used an excitation wavelength of 620 nm and an emission wavelength of 670 nm. EC₅₀ and EC₈₅ values, Hill slopes, and curves were generated by GraphPad Prism (La Jolla, CA). The raw data were analyzed using a four-parameter fit (minimal mP, highest enzyme concentration; maximal mP, no enzyme control; inflection point, EC₅₀; slope of the curve, Hill coefficient) using GraphPad Prism. The polarization data were converted to the amount of AMP formed using a standard curve mimicking conversion of substrates (ATP or cGAMP) to products (AMP and GMP); for example, ATP initially at 10 µM is reduced stepwise and AMP is added proportionately.

Assay Conditions

ENPP1 enzyme reactions contained 10 µM cGAMP or 10 µM ATP in 10 µL reactions in 25 mM Tris, 5 mM MgCl₂, 0.01% Brij-35, pH 7.5. Reactions were incubated for 60 min at room temperature followed by the addition of 10 µL of stop and detect mix containing 8 nM tracer and 5 µg/mL antibody and 10 µL of EDTA to quench reactions. The optimal AMP²/GMP² antibody concentration was determined by serially titrating the antibody in the presence of 10 µM cGAMP or ATP in assay buffer containing 4 nM tracer to determine the concentration that results in 85% of the maximal polarization signal (EC₈₅). The optimal enzyme concentration, determined by serial dilution, was selected to produce >100 mP shift under initial velocity conditions (<20% substrate consumption). Raw data (millipolarization units for FP) were converted to AMP using a standard curve run under similar conditions. Substrate K_m values were determined in reactions containing varying amounts of cGAMP or ATP, up to 200 µM, and 500 pM ENPP1. Staggered incubation periods of 5, 10, 20, 30, 45, or 60 min were used prior to the addition of stop and detect mix to allow for initial velocity measurements at all substrate concentrations. Raw data (millipolarization units for FP) were converted to AMP using separate standard curves for each ATP or cGAMP concentration. The initial velocity from each of those time points was fitted to a Michaelis–Menten curve using GraphPad Prism to calculate the apparent K_m values.

Determination of Screening Parameters

DMSO tolerance was determined by running optimized ENPP1 reactions at 10%, 5%, 2%, 1%, 0.5%, and no DMSO with 10 µM ATP. (This experiment was performed only with ATP, and the optimal concentration was applied to cGAMP as well.) Screening conditions were further validated by running 16 replicates of ENPP1 in optimized reactions with 10 µM cGAMP or 10 µM ATP. Controls in the absence of nucleotides were also run under similar conditions. Z′ measurements were calculated using the following formula:

$$\mathrm{Z}\prime =1-\left(\begin{array}{l}3*{\mathrm{SD}}_{[\mathrm{ENPP}1+\mathrm{substrate}]}+3*{\mathrm{SD}}_{[\mathrm{ENPP}1-\mathrm{substrate}]}/\\ \left({\mathrm{Avg}}_{[\mathrm{ENPP}1-\mathrm{substrate}]}-{\mathrm{Avg}}_{[\mathrm{ENPP}1+\mathrm{Sub}]}\right)\end{array}\right)$$

Dose–response curves for the ENPP1 inhibitors were conducted with both substrates to determine IC₅₀ values and correlate the experimental values with those reported in the literature.^13,16,17

Results and Discussion

The Transcrener AMP²/GMP² assay relies on the selective detection of AMP and GMP, which are both products of ENPP1 hydrolysis of cGAMP, using a competitive FP immunoassay: AMP or GMP generated by the target enzyme ( Fig. 1A ) displaces the tracer from antibody, resulting in lower polarization due to increased rotational mobility.^18,19 Optimization for a specific enzyme involves optimization of the antibody and enzyme concentrations to produce a polarization shift of at least 100 mP under initial velocity conditions, with tracer concentration held constant at 4 nM. To determine the feasibility of using the Transcreener AMP²/GMP² Assay for measuring ENPP1 activity, an antibody/tracer binding analysis was performed ( Fig. 1B ) in the presence of 10 µM ATP or cGAMP to determine the EC₈₅, the concentration of antibody required for near-saturation binding of the tracer. An EC₈₅ of 5 µg/mL was determined with either ATP and cGAMP as substrate, indicating that these conditions provide a favorable balance between sensitivity and signal magnitude.

Figure 1.

Optimization of AMP²/GMP2 antibody for developing a selective and sensitive assay for ENPP1. (A) Hydrolysis of ATP by ENPP1 produces a molecule of AMP and pyrophosphate; hydrolysis of cGAMP by ENPP1 produces a molecule of AMP and GMP. (B) AMP²/GMP2 antibody was titrated in the presence of either 10 μM ATP or cGAMP, in buffer containing 50 mM Tris, 5 mM MgCl₂, 1% DMSO, 0.01% Brij, pH 7.5, with 2 nM tracer. The plate was mixed well, incubated for 1 h, and read in PHERAstar Plus. The optimal antibody (EC₈₅) concentration was determined to be 10 μg/mL for both substrates. (C) Nucleotides ATP, GTP, ADP, AMP, GMP, and cGAMP were titrated in buffer containing 50 mM Tris, 5 mM MgCl₂, 1% DMSO, 0.01% Brij, pH 7.5, with 2 nM tracer and 10 μg/mL AMP²/GMP² antibody. (D) Table shows the selectivity for the AMP²/GMP2 antibody for the nucleotides with respect to either ATP or cGAMP. The antibody is 3000-fold more selective for AMP over ATP and >15,000-fold more selective for both AMP and GMP over cGAMP, enabling the detection of hydrolysis products of ENPP1 under initial velocity conditions.

We next confirmed that the antibody had sufficient selectivity for ENPP1 products to allow initial velocity activity measurements, that is, in the presence of excess substrate. ENPP1 hydrolyzes ATP to AMP and pyrophosphate, and it hydrolyzes cGAMP into AMP and GMP; the Transcreener AMP²/GMP² monoclonal antibody has equal affinity for AMP and GMP. We have previously shown that the antibody has very low cross-reactivity with other purine nucleotides, including ATP, GTP, ADP, and GDP, as well as cAMP and cGMP (BellBrook Labs, unpublished data.) We confirmed these results and extended them to cGAMP using competition binding curves ( Fig. 1C ). The antibody showed almost 3000-fold selectivity for AMP compared with ATP and >20,000-fold selectivity for AMP and GMP relative to cGAMP ( Fig. 1D ).

ENPP1 was titrated in the presence of 10 µM ATP or cGAMP to determine the optimal enzyme concentration ( Fig. 2A ). At a concentration of 100 pM, ENPP1 produced about 0.6 µM AMP (6% conversion) with ATP as substrate and 1.5 µM AMP and GMP (15% conversion) with cGAMP ( Fig. 2B ) in 60 min reactions. These results indicate that the enzyme is functioning at a similar velocity with the two substrates, which have similar K_m values (see below). A linear correlation between the concentration of ENPP1 and product formation is shown in Figure 2B , demonstrating initial velocity conditions.

Figure 2.

Development and optimization of ENPP1 activity assay for measurement of AMP production. (A) ENPP1 was titrated in the presence of 10 μM ATP or cGAMP in buffer containing 50 mM Tris, 5 mM MgCl₂, 1% DMSO, 0.01% Brij, pH 7.5. The plate was mixed well and incubated for 1 h. The enzyme reaction was quenched with an equal volume of EDTA containing stop buffer with 4 nM tracer and 10 μg/mL antibody. The plate was mixed well and read in PHERAstar Plus. (B) Raw data and polarization values were converted into AMP demonstrating initial velocity conditions. ENPP1 (500 pM) was used to determine the substrate K_m values in reactions containing varying amounts of (C) ATP or (D) cGAMP. Staggered incubation periods of 5, 10, 20, 30, 45, and 60 min were used prior to addition of stop and detect mix to allow for initial velocity measurements at all substrate concentrations. Raw data (millipolarization units for FP) were converted to AMP using separate standard curves for each ATP or cGAMP concentration. The initial velocity from each of those time points was fitted to a Michaelis–Menten curve using GraphPad Prism to calculate the apparent K_m values. (E) DMSO tolerance up to 10% for the ENPP1 enzyme reaction. (F) Z′ measurements using optimized ENPP1 reaction conditions indicate a robust assay.

Most screening is performed using substrates at or near their K_m concentrations to ensure detection of competitive inhibitors. We initially selected 10 µM ATP and cGAMP concentrations based on reported K_m values of 20 and 15 µM, respectively. ¹² However, apparent kinetic parameters can vary considerably depending on the assay method, so we determined K_m values for both substrates using the Transcreener AMP²/GMP² Assay. Reactions were run with varying concentrations of ATP or cGAMP, and ΔmP values were converted into product formation using separate standard curves for each substrate concentration and then fit using linear regression. The slope (velocity/min) determined for each of the substrates was plotted against the concentrations of substrate and fitted using the Michaelis–Menten equation to determine the K_m and V_max. The K_m for ATP was 7.7 µM, and the V_max was 0.0045 µM/min ( Fig. 2C ); the K_m for cGAMP was 4.7 µM, and the V_max was 0.0038 µM/min ( Fig. 2D ), confirming that 10 µM is a reasonable concentration to use for screening purposes.

To ensure compatibility with typical screening protocols, we tested the tolerance of the ENPP1 assay to DMSO using ATP as substrate. Note that the Transcreener AMP²/GMP² detection reagents are unaffected by DMSO concentrations as high as 10% (BellBrook Labs, unpublished data). There was no impact on AMP formation by ENPP1 at any of the DMSO concentrations tested, with the maximum being 10%, indicating that the enzyme has a high tolerance for this solvent ( Fig. 2E ).

To assess assay suitability for high-throughput screening (HTS), we determined Z′ values in assay buffer containing 1% DMSO; values of 0.8 and 0.74 were obtained with 10 µM ATP or cGAMP, respectively, as substrates ( Fig. 2F ), indicating robust assay performance.

We used four known ENPP1 inhibitors to generate dose–response curves and determine IC₅₀ values using the Transcreener AMP²/GMP² Assay. Suramin is a polyanionic compound (polysulfonated napthylurea) used to treat African trypanosomiasis with an unknown mechanism of action. ¹⁹ ENPP1 Inhibitor C (6-[(3-aminophenyl)methyl]-N,N,5-trimethyl-[1,2,4]triazolo[1,5-a]pyrimidin-7-amine) was identified as a selective inhibitor with an IC₅₀ of 10 µM. ¹⁷ POM-1 and PSB-069 are members of the polyoxometalate family of compounds, inorganic cluster metal complexes, that have been shown to inhibit ENPP1^18,19 as well as NTPDase 1 (CD39). The four compounds were tested first using subsaturating substrate concentrations (10 µM cGAMP or ATP). Under these conditions, ENPP1 Inhibitor C showed an IC₅₀ of 2.3 µM with cGAMP and 70 µM with ATP ( Fig. 3A,B ), a 35-fold difference depending on the substrate used. Suramin had a reverse effect in comparison to Inhibitor C: it was fourfold more potent with ATP as substrate (80 nM) compared with cGAMP (0.34 µM) ( Fig. 3A,B ). The polyoxometalate compounds, PSB-069 and POM-1, had IC₅₀ values ranging from 1 µM to 0.7 nM with no significant substrate dependence ( Fig. 3A,B ). More detailed analysis with varying substrate concentrations would be required to fully characterize these inhibitors, but our results suggest that they function to some degree with different mechanisms ( Fig. 3C ).

Figure 3.

Dose–response curves for ENPP1 inhibitors using ATP or cGAMP as substrates. (A) Dose–response curves for ENPP1 probe compounds with ATP. (B) Dose–response curves for ENPP1 probe compounds with cGAMP. (C) Table showing the IC₅₀ values obtained with varying substrate concentrations—ATP and cGAMP at 10 and 100 µM, respectively. (D) Specificity of ENPP1 toward cyclic dinucleotides. (E) Specificity of ENPP1 toward di- and trinucleotides. Data are plotted as the absolute magnitude of the polarization shift (ΔmP).

We next tested the substrate specificity of ENPP1 for cyclic nucleotides 2′3′cGAMP, 2′3′c-diGMP, and 3′3′c-diAMP and nucleoside di- and triphosphates; all nucleotides were titrated in the presence of 100 pM ENPP1. 2′3′cGAMP is unique to metazoans, whereas the other two cyclic nucleotides are produced only by bacteria; all three cyclic nucleotides effectively agonize STING. ²⁰ Consistent with previous results, ENPP1 was unable to hydrolyze 3′3′c-diAMP, and it hydrolyzed 2′3′c-diGMP at a 20-fold lower rate than 2′3′cGAMP ( Fig. 3D ), supporting the idea that ENPP1 specifically suppresses intrinsic activation of STING by 2′3′cGAMP, but not activation by microbial pathogens.^6,12–14

Next, we tested ENPP1 substrate specificity with purine nucleotides. Consistent with prior reports, ENPP1 preferentially hydrolyzed ATP (EC₅₀ ~ 0.1 µM) over other nucleotides^10,17 ( Fig. 3E ), which provides further validation of the assay method. GTP was hydrolyzed by ENPP1 with an EC₅₀ of 0.42 µM, a fourfold lower affinity when compared with ATP. (Note that these EC₅₀ values are a function of the dynamic range of the assay under the conditions used for this experiment and are therefore not equivalent to K_m values.) ENPP1 hydrolyzed the dinucleotides far less efficiently; the EC₅₀ values for ADP and GDP were 5 µM and 3.4 µM, respectively. The overall order of preference was ATP > GTP >> GDP > ADP.

There is great potential and excitement in developing new ENPP1 inhibitors to stimulate tumor immunogenicity via the cGAS-STING pathway. As this and other pathophysiological roles of ENPP1 begin to emerge, there is considerable untapped potential for drug discovery strategies to identify novel inhibitors. Structural insights from crystal structures of ENPP1 with 2′3′cGAMP can now enable the development of small-molecule inhibitors for ENPP1. Existing inhibitors of ENPP1 include ATP analogs and non-nucleotide-derived inhibitors such as dyes bearing sulfonate groups (e.g., suramin and related compounds) and polyoxometalates, which are inorganic, negatively charged metal complexes. In general, such compounds are intrinsically useful for biochemical studies but lack utility as pharmacological agents due to limitations such as lack of specificity, high molecular weight, low potency, and unfavorable pharmacokinetic profiles. While further investigation is needed to determine whether ENPP1 inhibitors could be viable drug candidates, success will ultimately require highly specific and sensitive HTS methods.

In this study, we show that the Transcreener AMP²/GMP² Assay serves as a robust HTS method for ENPP1. The assay enables quantitative detection of ENPP1 reaction products with either cGAMP or ATP as substrate without the use of coupling enzymes and has the robustness required for HTS. Because the assay is highly sensitive, low concentrations of substrate and enzyme can be used, which is advantageous from a cost perspective when scaling for large screens or extended discovery campaigns. Use of the assay for dose–response measurements yielded inhibitor potencies similar to literature values, ¹⁴ as did its use for determining kinetic properties of the enzyme. ¹² The Transcreener AMP²/GMP² Assay should prove to be a valuable tool for developing probes or lead molecules targeting ENPP1 for cancer immunotherapy and other indications.