Optimization of a non-radioactive high-throughput assay for decarboxylase enzymes

Abstract

Herein, we describe the optimization of a linked enzyme assay suitable for high-throughput screening of decarboxylases, a target family whose activity has historically been difficult to quantify. Our approach uses a commercially available bicarbonate detection reagent to measure decarboxylase activity. The assay is performed in a fully enclosed automated screening system under inert nitrogen atmosphere to minimize perturbation by exogenous CO2. Receiver operating characteristic (ROC) analysis following a pilot screen of a small library of approximately 3,600 unique molecules for inhibitors of Trypanosoma brucei ornithine decarboxylase quantitatively demonstrates that the assay has excellent discriminatory power (area under the curve = 0.90 with 95% confidence interval between 0.82 and 0.97).

Scheme 1.

Linked assay mechanism. ornithine decarboxylase (ODC) catalyzes the decarboxylation of ornithine, releasing CO₂, which is then captured by the basic buffer (pH 8.05) as bicarbonate. Phosphoenolpyruvate carboxylase (PEPC) uses this bicarbonate to generate oxaloacetate from phosphoenolpyruvate. The oxaloacetate is then reduced by malate dehydrogenase (MDH) to malate in a NADH-dependent fashion. There is a 1:1 relationship between the amount of CO₂ produced by ODC and the amount of NADH oxidized by MDH, allowing kinetic parameters for ODC to be calculated from observing NADH levels as a function of time. The assay system has been optimized such that ODC is the rate-limiting step.

Fig. 1.

Optimization of assay conditions. (A) Increase in background signal under normal atmosphere. The signal seen here in green is solely from atmospheric CO₂ (orange = nitrogen atmosphere, no ornithine; green = standard atmosphere, no ornithine; blue = nitrogen atmosphere, 625 μM ornithine; red = standard atmosphere, 625 μM ornithine). Quadruplicate data were collected in a 384-well microplate as described in the presence of the indicated concentrations of ornithine. All other reagent concentrations were identical to the optimized conditions described in Materials and Methods. The nitrogen atmosphere was maintained as described with O₂ levels under 2.5%. (B) Optimization of enzyme levels. ▾ = Z′ values, = reaction rates. Dashed line indicates standard Z′ cutoff value of 0.5. Data were collected in a 384-well microplate using optimized assay conditions with varied final ornithine decarboxylase (ODC) concentrations. Z′ Values were calculated using 8 positive (1 mM difluoromethylornithine (DFMO)) and 8 negative (DMSO) controls. The signal window at 150 nM ODC is ∼5-fold. (C) Determination of ODC K_m at 150 nM ODC and 60 μM PLP. Data were collected in a 384-well plate as described and fit to the Michaelis–Menten equation. (D) Optimization of Infinity CO₂ percentage ( = 60% Infinity CO₂, ▴ = 45% Infinity CO₂, ▾ = 30% Infinity CO₂, ♦ = 15% Infinity CO₂, • = 7.5% Infinity CO₂). Data were collected in 384-well plates as described at optimized assay conditions with varied Infinity CO₂. Data points were taken every 15 s in quadruplicate. The plot of AU₃₄₀ vs. time at 60% Infinity CO₂ represents a typical data set under optimized assay conditions with a ΔAU₃₄₀ of ∼0.4 AU.

Fig. 2.

Optimization of PEPC-MDH assay system. Data were collected under nitrogen as described in Materials and Methods. All data were collected under nitrogen in 384-well microplates. (A) Determination of the K_m apparent with respect to bicarbonate. The value obtained is consistent with literature values. Data were fit to the Michaelis–Menten equation. (B) Determination of assay linear time ( = 0.75 mM HCO₃, ▴ = no HCO₃). Under optimized assay conditions, the bicarbonate signal was linear for ∼7 min.

Fig. 3.

Performance of known inhibitors in the ODC-PEPC-MDH and PEPC-MDH assays ( = 60% Infinity CO₂, ▴ = 45% Infinity CO₂, ▾ = 30% Infinity CO₂, ♦ = 15% Infinity CO₂). Data were collected under optimized screening conditions as described unless otherwise noted. Compounds were allowed to equilibrate with enzymes for 20 min prior to addition of substrate. Percent inhibition is defined as (1 − v_i/v_o) × 100. At Infinity, CO₂ percentages lower than 45% reaction rates were calculated from the first 10 min of the reaction rather than the normal 20 min used at higher percentages due to limitations in assay linear time (see Fig. 1D). For PEPC-MDH assay results, rates were determined from the first 5 min (see Fig. 2B). (A) Performance of difluoromethylornithine (DFMO), a known ornithine decarboxylase (ODC) inhibitor, in the ODC-PEPC-MDH assay with varying amounts of linking enzymes. Note that linking enzyme concentrations do not effect the performance of DFMO. (B) Performance of baicalein, a known inhibitor of PEPC, at several different final linking mix concentrations in the ODC-PEPC-MDH assay. At normal screening conditions (60% Infinity CO₂) no inhibition is seen, even from this submicromolar inhibitor. However, upon dilution of the linking mix, inhibition becomes apparent. (C) Performance of isoquinoline, a known inhibitor of MDH in the ODC-PEPC-MDH assay at several Infinity CO₂ percentages. No inhibition was seen in any condition tested. Reaction rates at Infinity CO₂ concentrations lower than 45% were calculated from the first 10 min of data due to limitations in assay linear time (see Fig. 1D). (D) Performance of DFMO in the PEPC-MDH assay system at varying levels of Infinity CO₂. (E) Performance of baicalein in the PEPC-MDH assay system at varying levels of Infinity CO₂. As with the ODC-PEPC-MDH system, a dependency on the amount of linking enzymes present greatly effects the amount of inhibition seen. (F) Performance of isoquinoline in the PEPC-MDH assay. No inhibition was observed under any conditions tested. *Literature value, **Literature value, ***Literature value.

Fig. 4.

ODC-PEPC-MDH-linked plate geographic effects and high-throughput data scatter plot. Assays were performed in 384-well plates as described and rates were calculated from the final 6 data points from a 20-min observation. = High signal (DMSO), ▴ = mid signal (200 μM difluoromethylornithine (DFMO)), ▾ = low signal (1 mM DFMO). (A) Column effects resulting in a slight increase in signal on the outer columns. (B) Data from 3A arranged by row, showing that signal across rows is constant. The signal window for the assay is also apparent in these figures and is ∼4.5. (C) Data from the small-scale high-throughput assay of the bioactive compound collection at St. Jude Children’s Research Hospital. Green circles = positive control (1 mM DFMO), red circles = negative control (DMSO), blue circles = hit compound, black circles = inactive compound. Magenta line = 99th percentile cutoff, orange line = 95th percentile cutoff. Percent activity was calculated by normalizing kinetic reaction rates (20-min data collection with the final 6 data points fit as described in Materials and Methods) to positive and negative controls. The 8,832 total data points were collected from ∼3,600 unique compounds. Absolute replicate number for each compound varied from 1 to 7, depending on vendor library composition and plating. The assay took 8.7 h to complete using our automation system.

Fig. 5.

Effect of varying curve fitting methods on average Z′ values for scaling screen. ▸ = Median, ◂ = mean, solid line = standard Z′ cutoff of 0.5. Data were collected as described in Materials and Methods under optimized assay conditions. Time points were taken every minute for 20 min. Note that the first 5 to 10 min are poorly behaved and that eliminating them dramatically increases the average Z′ for the run. This is primarily a result of a decrease in scatter rather than an increase in signal, since the average rates calculated were consistent for all methods tested.

Fig. 6.

Receiver operating characteristic (ROC) curve for the ornithine decarboxylase (ODC) assay. The ROC area under the curve (AUC) is 0.90 (0.82–0.97, 95% confidence interval). Bootstrap simulation curves are in light gray. The dotted line is the curve for a random assay.