Substrate-Dependence of Competitive Nucleotide Pyrophosphatase/Phosphodiesterase1 (NPP1) Inhibitors

Nucleotide pyrophosphatase/phosphodiesterase type 1 (NPP1) is a membrane glycoprotein involved in the hydrolysis of extracellular nucleotides. Its major substrate is ATP which is converted to AMP and diphosphate. NPP1 was proposed as a new therapeutic target in brain cancer and immuno-oncology. Several NPP1 inhibitors have been reported to date, most of which were evaluated vs. the artificial substrate p-nitrophenyl 5′-thymidine monophosphate (p-Nph-5′-TMP). Recently, we observed large discrepancies in inhibitory potencies for a class of competitive NPP1 inhibitors when tested vs. the artificial substrate p-Nph-5′-TMP as compared to the natural substrate ATP. Therefore, the goal of the present study was to investigate whether inhibitors of human NPP1 generally display substrate-dependent inhibitory potency. Systematic evaluation of nucleotidic as well as non-nucleotidic NPP1 inhibitors revealed significant differences in determined Ki values for competitive, but not for non- and un-competitive inhibitors when tested vs. the frequently used artificial substrate p-Nph-5′-TMP as compared to ATP. Allosteric modulation of NPP1 by p-Nph-5′-TMP may explain these discrepancies. Results obtained using the AMP derivative p-nitrophenyl 5′-adenosine monophosphate (p-Nph-5′-AMP) as an alternative artificial substrate correlated much better with those employing the natural substrate ATP.

p-Nitrophenyl 5 ′ -thymidine monophosphate (p-Nph-5 ′ -TMP, Figure 1) is frequently used as a synthetic substrate for NPP1 to perform kinetic and inhibition assays since the monitoring of enzymatic reactions using natural substrates is much more timeconsuming and expensive requiring chromatographic separation or antibodies. The artificial substrate allows colorimetric monitoring of the enzymatic reaction through the formation of the intensively yellow-colored p-nitrophenolate, which absorbs at 400 nm (Henz et al., 2007;Buffon et al., 2010). The use of this artificial substrate is popular since it is straightforward and allows high-throughput screening of compound libraries.
Non-nucleotide-derived NPP1 inhibitors have also been described (see Figure 3). A series of quinazoline-4-piperidine-4methylsulfamides was reported as potent NPP1 inhibitors tested against ATP as a substrate, the most potent derivative being SAR 03004 (12) with an IC 50 -value of 0.036 µM (Patel et al., 2009). The mechanism of inhibition was not determined in that study. Very recently, 12 was further investigated as an NPP1 inhibitor using a colorimetric assay with p-Nph-5 ′ -TMP as a substrate. The study confirmed that the quinazoline derivative possesses high inhibitory potency with a K i -value of 0.059 µM (Shayhidin et al., 2015). However, that class of compounds also showed high affinity binding to hERG potassium channels, which precluded its further development as a drug, since QT prolongation was to be expected as a side-effect (Patel et al., 2009;Shayhidin et al., 2015). The non-selective purine P2 receptor antagonists reactive blue 2 (13) and suramin (14) were reported to be relatively potent NPP1 inhibitors vs. ATP as a substrate, displaying K i -values of 0.52 and 0.26 µM, respectively (Iqbal et al., 2008). Their mechanism of inhibition has not been reported. Recently, a series of thioacetamide derivatives was developed as potent competitive NPP1 inhibitors vs. p-Nph-5 ′ -TMP as a substrate, the most potent derivative PZB08513136A (15) displaying a K ivalue of 0.00500 µM (Chang et al., 2014). Moreover, inorganic polyoxometalates, e.g., [TiW 11 CoO 40 ] 8− (PSB-POM141, 16), were discovered to be potent and selective non-competitive NPP1 inhibitors, the best compound showing a K i -value of 0.00146 µM vs. ATP as a substrate. This compound represents the most potent inhibitor of human NPP1 described to date (Lee et al., 2015).
The majority of reported NPP1 inhibitors has only been investigated under completely unnatural conditions using spectrophotometric assays with p-Nph-5 ′ -TMP as an artificial substrate. However, in several studies substrate-dependent inhibitory potencies of various competitive enzyme inhibitors had been observed (Michaud et al., 1997;Hosoda et al., 1999;Schiemann et al., 2012;Ben Henda et al., 2013;Chang et al., 2014;Lee and Müller, 2014). For example, captopril, a competitive inhibitor of angiotensin-I converting enzyme (ACE), was significantly more potent when the synthetic substrates N-[3-(2-furyl)acryloyl]-Phe-Gly-Gly (FAPGG) or N-hippuryl-His-Leu hydrate (HHL) were used instead of the natural substrate angiotensin-I (Michaud et al., 1997;Ben Henda et al., 2013). Such a discrepancy in inhibitory potencies depending on the substrate was also observed for NPP2 (autotaxin) (Schiemann et al., 2012), an enzyme that is closely related to NPP1 but prefers phospholipids rather than nucleotides as substrates (Aoki et al., 2008). In our laboratory, we recently found that the inhibitory potencies of several nucleotidic inhibitors of NPP1 were significantly lower when tested vs. ATP as compared to the commonly used artificial substrate p-Nph-5 ′ -TMP (Lee and Müller, 2014). The non-nucleotide-derived thioacetamides (e.g., compound 15, Figure 3) displayed a particularly large discrepancy being much more potent vs. p-Nph-5 ′ -TMP as a substrate than vs. ATP (more than 100-fold difference for 15) (Chang et al., 2014). Thus, the goal of the present study was (i) to fundamentally investigate whether substrate-dependent potency of NPP1 inhibitors was a common phenomenon, and (ii) to find a possible explanation for these observations. To this end, we evaluated a wide range of structurally and mechanistically diverse NPP1 inhibitors vs. both, the artificial substrate p-Nph-5 ′ -TMP and the natural substrate ATP, considering both, nucleotidic and non-nucleotidic structures. The results were compared and correlation coefficients were calculated. Moreover, we synthesized and evaluated a new artificial substrate, p-nitrophenyl 5 ′ -adenosine monophosphate (p-Nph-5 ′ -AMP), which is structurally more similar to ATP than the standard artificial substrate. The results of this study will be highly relevant FIGURE 3 | Potent, non-nucleotide-derived NPP1 inhibitors.
Frontiers in Pharmacology | www.frontiersin.org with respect to in vivo studies with NPP1 inhibitors and for translational research aimed at drug development.

Determination of Kinetic Parameters of Artificial Substrates
Enzyme kinetic parameters were measured for p-Nph-5 ′ -TMP and p-Nph-5 ′ -AMP, both being artificial substrates of human NPP1. Solutions with different concentrations of both substrates (ranging from 1.0 to 500 µM) were prepared in 10 mM CHES buffer (in mM: 1 MgCl 2 , 2 CaCl 2 , 10 CHES, pH 9.0) and added in a final volume of 100 µl to 96-well-plates. The enzyme reactions were initiated by the addition of 20 ng of human NPP1 (for p-Nph-5 ′ -TMP), or 75 ng of human NPP1 (for p-Nph-5 ′ -AMP). The mixture was incubated at 37 • C for 15 min (p-Nph-5 ′ -TMP), or 30 min (for p-Nph-5 ′ -AMP), respectively, Frontiers in Pharmacology | www.frontiersin.org and subsequently terminated by the addition of 20 µl of 1.0 N aqueous NaOH solution. The amounts of p-nitrophenolate liberated were measured at 400 nm. Each analysis was repeated twice in three separate experiments.
For the natural substrate ATP, the enzyme inhibition assays were performed in 10 mM CHES buffer containing 400 µM of substrate along with different inhibitor concentrations. Incubation and operation conditions were the same as described above with artificial substrates. The analysis was performed by CE and the amounts of AMP produced were quantified by their UV absorption at 260 nm. Each analysis was repeated twice in three separate experiments.

Determination of Inhibition Constants and Mechanism of Inhibition
The inhibition mechanisms of the nucleotidic and nonnucleotidic inhibitors were determined using different concentrations of each substrate (from 10 to 1500 µM), and three different concentrations (0, ∼0.5-and ∼2-fold of IC 50 -value) of each test compound. The instrumentations and operation conditions for the experiments were the same as those described in the Sections Determination of Kinetic Parameters of Artificial Substrates and Determination of Concentration-Dependent Inhibition Curves. Each analysis was performed in three separate experiments. The inhibition type of each inhibitor was then evaluated graphically from the Hanes-Woolf plots. For the determination of the (α)K i -values the slope of the reciprocal lines from the Hanes-Woolf plot were plotted as a function of inhibitor concentrations using Prism 5.0.

Molecular Docking of Artificial Substrates
The generated homology model of human NPP1 described in Namasivayam et al. (2017). was used for the docking procedure using AutoDock 4.2 (Morris et al., 2009). The AutoDockTools package was employed to generate the docking input files and to analyze the docking results (Sanner, 1999). The search algorithm Lamarkian genetic algorithm (LGA) and the default scoring function, a hybrid scoring function (semi-empirical and free-energy) was employed for docking calculations. Threedimensional energy scoring grids for a box of 60 × 60 × 60 points with a spacing of 0.375 Å were computed. The grids were centered based on the co-crystallized ligand, which was transformed into the homology model. A total of 50 runs with a maximum of 250,000 energy evaluations were performed with the default parameters for the genetic algorithm (GA) and Solis-Wet local search, a method that facilitates random moving around the binding pose identified through the GA. High scoring binding poses (of lowest energy) or more populated poses were selected for the analysis on the basis of visual inspection.

Statistical Analyses
Statistical data analyses of pK i -values were performed using Prism 5.0 software. The pK i -values [-log 10 (α)K i -values] were calculated from the obtained (α)K i -values in each assay. Data were tested for statistical significance by one-way ANOVA as appropriate. When significant differences were observed, Tukey multiple comparison tests were performed. A value of p < 0.05 was considered significant.

Calculation of Correlation Coefficients between Assays with Different Substrates
The correlation coefficients (R 2 ) were evaluated by comparing pK i -values of one assay to those of another assay using Prism 5.0.

RESULTS AND DISCUSSION
The main natural substrate of NPP1, ATP (Figure 1A), and the generally used artificial NPP1 substrate, p-Nph-5 ′ -TMP ( Figure 1B; Laketa et al., 2010;Lee et al., 2012), differ not only in the phosphoric ester part (triphosphate vs. p-nitrophenyl phosphate), but also with respect to their nucleoside partial structure (adenosine vs. thymidine). In order to investigate potential substrate-dependence of various NPP1 inhibitors, we intended to test selected antagonists vs. both substrates. But, in addition, we decided to additionally evaluate them on a second artificial substrate, which is structurally more closely related to ATP, namely p-nitrophenyl 5 ′ -adenosine monophosphate (p-Nph-5 ′ -AMP, 19). Since 19 was not commercially available we decided to synthesize the compound.

Synthesis of a New Artificial Substrate of NPP1
p-Nph-5 ′ -AMP had been previously synthesized (Borden and Smith, 1966;Ivanovskaya et al., 1987), but no detailed characterization of the compound has been published. Initially we tried to prepare 19 directly from adenosine by reaction with p-nitrophenyl phosphorodichloridate. However, the reagent is toxic and difficult to handle. Moreover, tedious separation and purification procedures were required to obtain the desired product in sufficient purity, and the yield was only about 20%. Therefore, we decided to utilize commercially available AMP (17) for the preparation of 19. In a one-step reaction AMP was condensed with p-nitrophenol in the presence of N,N ′dicyclohexylcarbodiimide yielding product 19 in 62% yield after purification (see Figure 4).
In order to gain insights into the molecular determinants involved in the formation of the enzyme-substrate complex, the new artificial substrate p-Nph-5 ′ -AMP was docked into a homology model of the human NPP1 (Namasivayam et al., 2017), which was generated based on the recently solved crystal structure of the mouse NPP1 (Kato et al., 2012). The observed interactions were compared to those of ATP and p-Nph-5 ′ -TMP. As shown in Figure 5A, the α-phosphate group of the substrate ATP is bound between the two zinc ions, and the two other phosphate groups form hydrogen bond interactions with the following amino acid residues: Lys255, Thr256, Asn277, His380, and His535. Tyr340 forms a hydrogen bond with the ribose moiety, and the adenine ring of ATP is stacked between Phe257 and Tyr340 (Namasivayam et al., 2017). Similarly, both artificial substrates form complexes with the zinc ions of the enzyme with their phosphate groups as shown in Figures 5B,C. Because the p-nitrophenylphosphate group of both artificial substrates interacts in the same way with the zinc ions in the active site of the enzyme, the ground state of binding interactions may be comparable, which explains the similar K m -values determined for both artificial substrates. While p-Nph-5 ′ -AMP binding is stabilized through π-π interactions of the adenine base mainly with Tyr340 of the enzyme, this interaction is expected to be weaker for the artificial substrate p-Nph-5 ′ -TMP due to the exchange of adenine for thymine. Furthermore, a 2 ′ -hydroxyl group at the ribose moiety is lacking in p-Nph-5 ′ -TMP, but not in p-Nph-5 ′ -AMP, and therefore, the interaction between that OH group and the side chain of Tyr340 is missing in p-Nph-5 ′ -TMP. Overall, p-Nph-5 ′ -TMP has less interactions than p-Nph-5 ′ -AMP and therefore, it may be hydrolyzed faster via the transition state than the new artificial substrate. This may explain why p-Nph-5 ′ -TMP showed a significantly higher k cat -value than p-Nph-5 ′ -AMP.

Correlation
Inhibitory potencies of non-nucleotidic and nucleotidic inhibitors vs. different substrates were compared (Figure 7). Data analysis revealed substrate-dependent inhibitory potencies of competitive inhibitors, but not of non-or uncompetitive inhibitors. When K i -values obtained with the natural substrate ATP were compared with those obtained with the artificial substrate p-Nph-5 ′ -TMP the competitive inhibitors were 3-3600-fold more potent against p-Nph-5 ′ -TMP than vs. ATP [p < 0.05 for SAR 03004 (12), p < 0.01 for α,β-metATP (4) and 2-MeSATP (6), and p < 0.001 for α,β-metADP (3), 2-MeSADP (5) and PZB08513136A (15)]. Differences were also dependent on the structure of the competitive antagonists, e.g., it was particularly high for the thioacetamide derivative PZB08513136A (15), but less pronounced for the quinazoline derivative SAR 03004 (12). In contrast, results obtained vs. the new artificial substrate p-Nph-5 ′ -AMP were similar to those obtained vs. the natural substrate ATP. As opposed to competitive inhibitors, substrate-dependent inhibition was not observed for non-competitive and un-competitive inhibitors,  (Chang et al., 2014), expressed as means ± SEM. d K i -values from the literature (Lee et al., 2015), expressed as means ± SEM.
FIGURE 7 | Comparison of inhibitory potency of nucleotidic (A) and non-nucleotidic (B) inhibitors using different substrates. Data are means ± SDs of pK i -values. The bars in gray represent pK i -values of inhibitors vs. the natural substrate ATP, those in red are the pK i -values of inhibitors vs. p-Nph-5 ′ -TMP and those in blue are pK i values vs. p-Nph-5 ′ -AMP. *p < 0.05, **p < 0.01, and ***p < 0.001. and very similar inhibition constants were obtained vs. all investigated substrates. Correlation analyses of pK i -values obtained vs. one substrate with those measured vs. another substrate were performed. Considering the competitive inhibitors, a low correlation of data obtained with p-Nph-5 ′ -TMP as a substrate with those obtained with the natural substrate ATP was obtained [correlation coefficient (R 2 ) = 0.5722, see Figure 8A], whereas a high correlation between the results obtained with p-Nph-5 ′ -AMP as a substrate and those determined with ATP was observed (R 2 = 0.9578). Moreover, Figure 8A (left) showed that the data points were shifted to the right of the ideal correlation line [dotted line in Figure 8A (left)]. This indicates that the competitive NPP1 inhibitors were generally more potent vs. p-Nph-5 ′ -TMP than vs. ATP as a substrate. Contrary to this, the non-and un-competitive inhibitors showed high correlations, no matter which substrates were used for comparison (R 2 = 0.9742 for competitive inhibitors; R 2 = 0.9900 for non-and un-competitive inhibitors), see Figure 8B.

Possible Explanation for Substrate-Dependence of Competitive NPP1 Inhibitors
The observation of significantly different potencies of competitive enzyme inhibitors when determined vs. different substrates is puzzling, and an explanation for this phenomenon is not straightforward. The different assay conditions are clearly not the reason for the observed discrepancies because the same operating conditions (e.g., same stock solutions of inhibitors, same assay buffer) were applied for the enzyme inhibition assays with different substrates. A rational explanation for the different results between assays obtained with p-Nph-5 ′ -TMP and the natural substrate ATP could be an allosteric modulatory effect by p-Nph-5 ′ -TMP on the enzyme, in addition to acting as a substrate (Figure 9). Such allosteric binding of the substrate has previously been reported for another nucleotide-metabolizing enzyme, bacterial UDP-Nacetylglucosamine 2-epimerase, which is allosterically modulated by its substrate UDP-N-acetylglucosamine (Velloso et al., 2008). The binding of p-Nph-5 ′ -TMP to its allosteric binding site, which may be close or even distant from the active site, could induce a conformational change of the substrate binding site. This would modulate the interaction of competitive inhibitors with the substrate binding site, and could therefore explain the increased affinity of the investigated competitive inhibitors (Figure 9). This hypothesis is supported by the fact that the affinity increase depends on the structure of the inhibitors, e.g. some competitive inhibitors (e.g., 15) being much more strongly affected than others (see also Table 1 and Figure 9).
This hypothesis also provides a straightforward explanation for the finding that p-Nph-5 ′ -TMP is a much better NPP1 substrate than p-Nph-5 ′ -AMP despite the fact that-based on docking studies-the AMP derivative should have stronger interactions with the substrate binding site. p-Nph-5 ′ -TMP may FIGURE 8 | Correlation analyses between the results (A) for competitive inhibitors, and for (B) non-and un-competitive inhibitors obtained with different substrates. Determined correlation coefficients (R 2 ) were calculated by fitting pK i -values obtained with one substrate vs. those obtained with another substrate using the software Prism 5.0; red points, test compounds; solid line, the best fit line of the linear regression; the dotted line in (A) represents the ideal correlation (R 2 = 1.00).
additionally bind to an allosteric site and thereby act as a positive allosteric modulator which increases its binding affinity to the substrate binding site and accelerates its hydrolysis.
Further, investigations to corroborate this hypothesis of allosteric modulation of the active site by p-Nph-5 ′ -TMP are warranted.

CONCLUSIONS
In conclusion, we observed substrate-dependence of the inhibitory potency of NPP1 inhibitors, competitive inhibitors being (much) more potent vs. the artificial substrate p-Nph-5 ′ -TMP than vs. the natural nucleotide substrate ATP. In contrast, data obtained using the new artificial substrate p-Nph-5 ′ -AMP correlated well with those determined vs. the natural substrate ATP indicating that the nucleoside part of the artificial substrate was responsible for the observed effects. No significant differences in inhibitory potencies were observed for non-or un-competitive inhibitors. The most likely explanation for the FIGURE 9 | Possible explanation for the discrepancies observed for competitive inhibitors vs. the artificial substrate p-Nph-5 ′ -TMP (higher affinity observed for competitive antagonists) as compared to natural substrates assays (lower affinity for competitive inhibitors). p-Nph-5 ′ -TMP may not only act as a substrate, but also as an allosteric modulator.
observed phenomenon is an allosteric modulation of NPP1 by the artificial substrate p-Nph-5 ′ -TMP, but not by p-Nph-5 ′ -AMP. Therefore, we recommend to use p-Nph-5 ′ -AMP instead of p-Nph-5 ′ -TMP for high-throughput screening of NPP1 using colorimetric detection. Further, investigations to explain the discrepancy between results with the commonly used artificial substrate p-Nph-5 ′ -TMP and the natural substrate ATP are in progress.

AUTHOR CONTRIBUTIONS
SL performed the pharmacological experiments, analyzed the data and contributed to writing of the manuscript. SS, SB, SD, HS, PH, and AE synthesized, purified and analyzed compounds. VN performed the molecular modeling studies and contributed to writing of the manuscript. CM designed and supervised the study and wrote the manuscript.