Identification of Novel Carbonic Anhydrase IX Inhibitors Using High-Throughput Screening of Pooled Compound Libraries by DNA-Linked Inhibitor Antibody Assay (DIANA)

Abstract

The DNA-linked inhibitor antibody assay (DIANA) has been recently validated for ultrasensitive enzyme detection and for quantitative evaluation of enzyme inhibitor potency. Here we present its adaptation for high-throughput screening of human carbonic anhydrase IX (CAIX), a promising drug and diagnostic target. We tested DIANA’s performance by screening a unique compound collection of 2816 compounds consisting of lead-like small molecules synthesized at the Institute of Organic Chemistry and Biochemistry (IOCB) Prague (“IOCB library”). Additionally, to test the robustness of the assay and its potential for upscaling, we screened a pooled version of the IOCB library. The results from the pooled screening were in agreement with the initial nonpooled screen with no lost hits and no false positives, which shows DIANA’s potential to screen more than 100,000 compounds per day.

All DIANA screens showed a high signal-to-noise ratio with a Z′ factor of >0.89. The DIANA screen identified 13 compounds with K_i values equal to or better than 10 µM. All retested hits were active also in an orthogonal enzymatic assay showing zero false positives. However, further biophysical validation of identified hits revealed that the inhibition activity of several hits was caused by a single highly potent CAIX inhibitor, being present as a minor impurity. This finding eventually led us to the identification of three novel CAIX inhibitors from the screen. We confirmed the validity of these compounds by elucidating their mode of binding into the CAIX active site by x-ray crystallography.

Keywords

high-throughput screening carbonic anhydrase IX drug discovery DNA-linked inhibitor antibody assay

Introduction

Carbonic anhydrase IX (CAIX), 1 of 12 catalytically active human carbonic anhydrase (CA) isoforms, is a promising cancer target.¹ It is an integral type I transmembrane protein with a catalytically active domain facing to the extracellular space, which catalyzes the hydration of carbon dioxide to carbonate and thus helps cells to maintain their acid–base balance.² CAIX is upregulated by hypoxia and is thus overexpressed in a variety of solid tumors.^3,4 Through its enzymatic activity, CAIX is responsible for acidification of the extracellular space of tumors, which is associated with their aggressiveness. Simultaneously, it counteracts acidification of intracellular space, which is particularly important in tumors due to their metabolic acidification.^5,6 Therefore, it is not surprising that inhibition of CAIX has been shown to impair the growth of CAIX-expressing tumors in both in vitro and in vivo models.⁷ Furthermore, due to its rather specific expression in several solid tumors, CAIX is investigated as an address for specific delivery of cytotoxic drugs.^8,9

Novel CAIX inhibitors are needed to enable clinical exploitation of CAIX, but current assays for activity/inhibition testing of CAs are laborious and time- and enzyme-consuming.⁸ Currently, all inhibitors of CAs investigated for potential clinical use are derivatives of sulfonamides. Use of sulfonamide derivatives for clinical CAIX targeting is hindered by several obstacles. First, they have unfavorable pharmacokinetic properties, mainly due to high nonspecific adsorption to serum proteins.¹⁰ Second, it has been estimated that sulfonamide administration leads to allergic reactions in approximately 3%–6% of the population, which can evolve in a life-threatening state.^11,12 Finally, sulfonamides act nonselectively by binding to zinc ions contained in the active site of all human CAs and therefore show poor selectivity across the CA family.¹³ Given these issues, the identification of a new structural class of inhibitors could represent a breakthrough in the design of selective CAIX inhibitors and enable significant advances in both primary and translational research employing CAIX inhibitors as an anticancer therapy target. Even though methods like isothermal titration calorimetry (ITC) and fluorescent thermal shift assay (FTSA) were used for the assessment of large numbers of CAs inhibitors, they lack sufficient throughput and are quite demanding on sample consumption to represent effective platform for high-throughput screening (HTS) of CAIX inhibitors.¹⁴

The DNA-linked inhibitor antibody assay (DIANA) has been recently developed for ultrasensitive enzyme detection and inhibitor screening.¹⁵ In DIANA, the target protein is captured by an immobilized antibody and subsequently detected by a fully synthetic detection probe, which comprises a small-molecule ligand (i.e., competitive inhibitor) of a target enzyme covalently linked to a DNA oligonucleotide.

On top of the sensitive detection of protein targets, DIANA can also be used for the screening of enzyme inhibitors. The inhibition potency of the test compounds is determined by their ability to outcompete binding of the probe to the enzyme. DIANA has several unique advantages over state-of-the-art screening technologies, such as (1) a wide dynamic range enabling the determination of the inhibition potency (K_i value) from a single well measurement, (2) an extremely high signal-to-noise ratio, and (3) the ability to screen inhibitors with minuscule amounts of unpurified enzymes. The robustness of DIANA can be proven by Z′ values, which are commonly in the range of 0.85–0.90.¹⁵ The initially reported DIANA protocol was performed manually in a 96-well plate format and was not suitable for high throughput. However, adapting DIANA to automated 384-well format would represent significant improvement of its HTS capability. An assay such as DIANA would address several main drawbacks of current inhibitor screening assays, such as low accuracy, narrow dynamic range, and only qualitative output.

The IOCB library represents a unique proprietary well-curated collection of small molecules, which were synthetized at the IOCB of the Czech Academy of Sciences. It contains various unique chemical scaffolds typically not present in commercial libraries (i.e., natural compounds, steroids, peptides and peptidomimetics, and nucleosides and nucleotides), with the main emphasis on nucleoside and nucleotide analogs. Importantly, most compounds present within the library were prepared as a result of medicinal chemistry efforts to develop potential drugs against various pharmaceutically relevant targets, which ensures their drug-likeness. Indeed, there are already several FDA-approved drugs originating from this library, such as tenofovir, one of the most widely used inhibitors of HIV reverse transcriptase; adefovir, a potent virostatic used in the treatment of hepatitis B infection; and cidofovir, an effective anticytomegalovirus compound.¹⁶ The library currently contains more than 3000 compounds and it is still growing, as more compounds are continuously prepared at IOCB. Considering all the abovementioned properties, the IOCB library represents a very useful compound collection for the screening of pharmaceutically relevant enzymes such as CAIX.

Here, we present and describe the identification of novel CAIX inhibitors using DIANA adapted to HTS format combining several strategies for an increased throughput: (1) reformatting the assay from 96- to 384-well plate format, (2) substituting all manual pipetting steps by liquid handling platforms, and (3) pooling of test compounds for the screen. We used DIANA for screening more than 2800 compounds from the IOCB library, and thoroughly validated the obtained hits from the assay in terms of their purity, inhibition activity, and mode of binding toward CAIX using x-ray crystallography.

Materials and Methods

HTS of IOCB Library by DIANA against CAIX

The general DIANA protocol for CAIX HTS was performed as previously described with minor modifications.¹⁵ Briefly, 25 ng of M75 antibody was coated on a FrameStar 384 Well Skirted PCR Plate (4ti-0381; 4titude Ltd., Surrey, UK) in Tris-buffered saline (TBS) followed by addition of casein blocking agent diluted fivefold in TBS (CBC1; SDT, Baesweiler, Germany). After washing, 5 µg of HT-29 cell lysate, which endogenously expresses CAIX protein, was added to the wells, incubated, and washed away. The bivalent detection probe CAIX probe 2 (structure published previously¹⁵) was added to the wells together with the test compounds. If not stated otherwise, the final probe concentration was 500 pM and the concentration of test compound was 10 µM (during pooled testing each compound in the well was 9.1 µM so the final concentration of all test compounds within one well was 100 µM). After additional washes, the quantitative PCR (qPCR) mixture was added to the wells and the signal was detected using a LightCycler 480 II qPCR instrument (Roche, Basel, Switzerland). Threshold cycles (Cq) were obtained from the measured fluorescence curves using the method of maxima of the second derivative in the LightCycler 480 II Software (Roche). The acquired data were then analyzed and K_i values were calculated based on the previously described equations.¹⁵

For the washes between all the steps TBS + 0.05% Tween-20 using BlueWasher (BlueCatBio, Neudrossenfeld, Germany) was used. During all experiments using the 384-well plate format, the last two columns of the 384-well plate were used for control samples that contained dilution series (3–10,000 nM) of the known potent CAIX inhibitor acetazolamide (AZA) in duplicates and wells without any inhibitor in at least octaplicate. These controls served as intraplate and interplate controls throughout the screens and enabled the determination of the assay window and calculation of Z′ factors.

For all DIANA measurements, the K_i values of test compounds were calculated using the equation for bivalent probe:

K_{i} = (\frac{1 - 2 * R_{a f f}}{1 - R_{a f f} - \sqrt{{(R_{a f f})}^{2} + (1 - 2 * R_{a f f}) * 2^{- d C q}}} - 1) * (\frac{[I]}{1 + \frac{[P]}{K_{D (P)}}})

(1)

where R_aff is defined as a ratio of dissociation constants of bivalent to monovalent probes toward CAIX, which is equal to 0.022; dC_q is the difference between Cq values of wells with and without test compound; [I] is the working concentration of the test compound (variable); and [P] and K_D_(P) are the working concentration (500 pM) and dissociation constant (1.5 nM) of the bivalent probe, respectively. Detailed derivation of the equation is described in the supplementary information of Navrátil et al.,¹⁵ where it is listed as eq 15.

K_i Determination of Screen Hits by DIANA

To validate results from HTS, the K_i values for identified hits were determined again by DIANA in the 96-well plate format. The protocol was identical to HTS, except each test compound was measured in four wells with different concentrations of test compound ranging from 50 µM to 50 nM. The acquired data were then analyzed and K_i values were calculated based on the previously described equations.¹⁵

CA Inhibition Assay

A stop-flow instrument (Applied Photophysics, Surrey, UK) was used to measure the CA-catalyzed CO₂ hydration activity in the presence of inhibitors.¹⁷ The assay was performed as described previously with minor modification.¹⁸ The CA concentrations in the reaction mixture ranged from 5 to 19 nM. The apparent inhibition constants were obtained by nonlinear least-squares methods in GraphPad Prism version 8.0.0 for Windows (GraphPad Software, San Diego, CA; www.graphpad.com), which explore nonlinear fitting of the Williams–Morrison equation:¹⁹

\frac{v_{i}}{v_{0}} = 1 - \frac{\begin{array}{l} ([E] + [I] + K_{i}^{a p p}) - \\ \sqrt{{([E] + [I] + K_{i}^{a p p})}^{2} - 4 * [E] * [I]} \end{array}}{2 * [E]}

(2)

The K_i values were derived from the apparent K_i using the Cheng–Prusoff equation:¹⁹

K_{i} = \frac{K_{i}^{a p p}}{(1 + \frac{[S]}{K_{m}})}

(3)

FBSA Quantifi cation by LC-MS

All hits from the DIANA screen (compounds 1–13) were dissolved in a 50% mixture of acetonitrile (ACN) and water at a final concentration 10 µM. The dilution series of perfluoro-1-butane-sulfonamide (FBSA) was prepared identically, ranging from 10 nM to 10 µM. Under these concentrations, the integrated total ion current (TIC) signal showed linear dependence of the FBSA amount and was used as a calibration curve to calculate the amounts of FBSA within analyzed samples. Ten microliters of each compound was injected per sample and analyzed on an Agilent 6230 TOF LC/MS using a Waters BEH C18 column (100 × 2.1 mm). All samples were run during a single experiment to minimize the variability of the data.

Results

Adapting DIANA into HTS Format

To enable screening of large compound libraries via DIANA, we first needed to adapt the assay to the 384-well format and automate all liquid handling steps. A detailed scheme of the final DIANA protocol is described in the Materials and Methods section and summarized in Supplemental Table S1 . Briefly, the developed protocol consists of several incubation and washing steps, and its total duration is less than 3 h, when starting with precoated and blocked plates. The rate-limiting step in the protocol is qPCR, which takes less than 30 min. As a result, 16 plates per 8 h can be run on a single qPCR instrument resulting in a throughput of 5632 compounds per day having 352 compounds per plate.

Screen of the IOCB Library against CAIX Using DIANA

To evaluate DIANA screening capabilities, we performed a screen of 2816 compounds from the IOCB library against CAIX using DIANA, that is, eight 384-well plates with each plate comprising 352 wells with test compounds and 32 wells with controls. The screen revealed 13 hits and high signal-to-noise ratios expressed by the Z′ factor ranging between 0.91 and 0.95 for each of the eight plates. The results from the screen are summarized in Figure 1A .

Figure 1.

Results of nonpooled and pooled DIANA screen of IOCB library against CAIX. Each test compound is represented by a dot and the shown dCq values are calculated as Cq of the well without any test compound subtracted from Cq of the well with compound (Cq_cpd – Cq_blank). The empirical threshold (gray dashed line) shows the dCq value corresponding to 43% inhibition, that is, 10 µM K_i value (dCq = 1.57). Compounds with dCq higher than 1.57 (K_i < 10 µM) are shown as diamonds, and compounds below this threshold are shown as gray circles. The compound shown as a triangle (dCq = 1.42) represents a compound that was identified as a hit in the pooled screen but not in the nonpooled screen. DIANA was performed as described in Materials and Methods. Briefly, CAIX was captured on a 384-well plate via the M75 antibody. Subsequently, each test compound (A, final concentration 10 µM, 8 plates, 2816 data points) or mixture of 11 compounds (B, final concentration 9.1 µM, 1 plate, 256 data points) was premixed with the detection probe and added to the corresponding well. After washing, the qPCR chemicals were added, and the signal was obtained by qPCR.

We set the cutoff threshold to dCq 1.57, which corresponds to approximately 43% inhibition of CAIX under the assay conditions. In this way, we identified 13 positive hits (diamonds in Fig. 1A ) within the IOCB library, which corresponds to a hit rate of 0.46%. The chosen cutoff signal corresponds to a theoretical K_i value of 10 µM, which is equal to the final concentration of the test compound.

Given the very low chance of obtaining false-positive hits in DIANA (see Discussion for more information), we also investigated the ability of DIANA to screen pooled libraries. We pooled 11 compounds from the IOCB library per well and thus limited the number of tested wells from 2816 to 256. As shown in Figure 1B , we identified 10 positive wells from the screen using identical cutoff criteria as for the nonpooled screen. The Z′ factor for the pooled screen was 0.89. The general characteristics of both the nonpooled and pooled DIANA screens are summarized in Table 1 .

Table 1.

General Characteristics of DIANA CAIX Screen.

Parameter	DIANA Screening Assay
Assay container	384-well plate
Assay total time^a	<3 h
Assay window [dCq]	>10 (>1000-fold)
SD without compounds [Cq]	0.07–0.13^b
Assay Z′ factor	0.91–0.95 (nonpooled screen)^c 0.89 (pooled screen)
Assay throughput^d	5632 compounds/day (nonpooled) 61,952 compounds/day (pooled)

Assay time is calculated from the step of antigen addition.

Range of standard deviations (SD) from 16 blank wells for each of the nine tested 384-well plates (eight from nonpooled screen and one from pooled screen).

Range of Z′ factors from eight tested 384-well plates.

The throughput is calculated based on the assumption that 16 plates containing 352 test wells can be measured per 8 h on one qPCR instrument and that 11 compounds per well are pooled together within 1 well.

Comparison of Nonpooled and Pooled DIANA Screens

A closer look at the identified hits from both nonpooled and pooled DIANA screens revealed that all 13 hits identified during the nonpooled screen were also identified by the pooled screen with two pool wells (nos. 194 and 215) containing more than one hit. The determined dCq values for each hit are summarized in Table 2 , with each row comprising one well from the pooled screen and all identified hits from the nonpooled screen within that well. Additionally, one positive well in the pooled screen (no. 170) was found that did not contain a positive hit from the nonpooled screen. However, examining the signal of 11 compounds present within the pooled well (no. 170), there was one compound (no. 2143) that provided a signal very close to the cutoff threshold (highlighted as triangle in Fig. 1A ). This compound, responsible for the positive outcome of the pool, is a positive hit with weak inhibition potency (>10 µM) against CAIX.

Table 2.

Summary of Hits Identified during Nonpooled and Pooled DIANA Screen against CAIX.

Pooled Screen		Nonpooled Screen
Well No.	dCq	Well No.	dCq	Compound ID	K_i [nM]
178	11.49	1269	10.57	1	92
254	11.34	2260	10.82	6	79
194	10.82	1794	9.77	2	140
		1796	9.18	3	200
		1800	2.01	4	7200
		1792	10.29	8	110
215	5.43	2047	4.01	9	2300
		2067	3.06	11	3700
106	4.45	2101	2.81	5	4300
199	3.52	1973	1.81	7	8300
174	2.83	2245	3.03	12	3800
240	2.54	2803	3.30	10	3300
176	2.14	2772	1.99	13	7200
170	2.42	2143	1.42

For pooled and nonpooled screens each well that provided a signal above the threshold (dCq = 1.57) is shown together with its dCq value. Additionally, well 2143 from the nonpooled screen is also shown as it contained a hit that was close but did not surpass the set threshold. The wells shown in the same row of the nonpooled screen were present in the same well during the pooled screen. dCq was calculated as Cq of the well without any test compound subtracted from Cq of the well with compound (Cq_cpd – Cq_blank). All compounds showing potency better than 10 µM were assigned a unique number for easy referencing, and its K_i value was calculated from the dCq signal of the nonpooled DIANA screen.

Overall, the comparison of these two datasets confirmed that the pooled DIANA screen maintains its robustness (high Z′ factor) and that it has identified the identical set of hits as the nonpooled screen.

Determination of K_i Value from Single Well Measurement

One of the key features of DIANA is the fact that it is a quantitative method; that is, it enables determination of hit K_i values from a single well measurement (see Table 2 ). In this particular case, in which 10 µM compounds were tested, the assay window enabled precise determination of K_i values in the range of two orders of magnitude between 100 nM and 10 µM; this range depends on several factors, such as assay window, valency of the used probe, and test compound working concentration.¹⁵ Given these assay features for the CAIX DIANA, all K_i values close to 100 nM could be in theory underestimated and the actual inhibition potency of these compounds may have been better.

To confirm this assumption, we performed DIANA at four different concentrations of compounds 1–13 ranging from 50 nM to 50 µM. For clarity, we provide a graphical representation of the data as the correlation of K_i values determined from the nonpooled screen (DIANA single point) and from DIANA with a dilution series of test compounds (DIANA inh. titration; see Fig. 2A and raw data in Suppl. Table S2 ). As expected, the correlation is very good for compounds with K_i values above 100 nM (R² = 0.835; black dots), while for compounds showing K_i values around 100 nM in the screen the actual K_i values are lower (gray dots). These data confirm that the K_i calculation from the single well measurement is accurate but is limited by the assay window and can lead to underestimation of the inhibition potency of very potent hits.

Figure 2.

Comparison of determined K_i values of identified inhibitors against CAIX. (A) Comparison of the K_i values of 13 hits determined by DIANA at four different concentrations (50 nM–50 µM; DIANA inh. titration) and at a single 10 µM concentration in the nonpooled DIANA screen (DIANA single point). K_i values obtained from more than one dilution of the compound during the DIANA four-concentration experiment are shown as an average with SD. Compounds highlighted in gray are highly potent CAIX inhibitors whose inhibition potency could not be precisely determined by the nonpooled DIANA screen. (B) Comparison of K_i values of 10 potent hits determined by DIANA at four different concentrations (50 nM–50 µM; DIANA inh. titration) and orthogonal enzymatic inhibition assay (enzyme kinetics). Low-potency compounds 10, 11, and 13 are not included in the comparison since it was not possible to measure their K_i value by enzymatic inhibition assay due to high sample amount requirements. The raw data can be found in Supplemental Table S2. The K_i values from DIANA were calculated using a previously derived equation (eq 1), while K_i values from the enzymatic inhibition assay were fitted using GraphPad Prism software utilizing eqs 2 and 3 (for more details, see the Materials and Methods section).

To further validate the precision of K_i value determination of DIANA, we determined the K_i values of the 10 most potent hits by an established orthogonal enzymatic inhibition assay, which is based on the measurement of pH changes of CA-catalyzed CO₂ hydration using a spectrophotometric readout. As shown in Figure 2B , we observed a good correlation of K_i values obtained by DIANA (DIANA inh. titration) and enzymatic inhibition assay (enzyme kinetics; R² = 0.870), confirming the robustness and specificity of DIANA (see raw data in Suppl. Table S2 ).

Validation of the Hits from DIANA Screen against CAIX

After identification of the 13 hits by DIANA screen, we performed a thorough analysis of these compounds by liquid chromatography–mass spectrometry (LC-MS) to confirm their chemical purity and homogeneity. The LC-MS analysis provided several important findings. First, we found that compounds 2 and 8 are identical, being present in the library twice. Second, we observed that compounds 1–5 are partially hydrolyzed and that they have a common hydrolysis product. By inspection of the chemical structures of compounds 1–5 (shown in Fig. 3B ) and data from LC-MS we identified the most probable hydrolysis product to be FBSA (shown in Fig. 3A ). Lastly, LC-MS analysis of compounds 9–13 (shown in Fig. 3C ) also revealed the presence of FBSA even though they are structurally unrelated with compounds 1–5. Compounds 6 and 7 were shown to be homogenous as data from LC-MS confirmed the expected molecular mass of these compounds.

Figure 3.

Chemical structures of identified hits against CAIX by DIANA. (A) Structures and inhibition constants for validated hits against CAIX. Inhibition constants were measured by two orthogonal methods: DIANA and established enzymatic inhibition assay (enz.). For DIANA, the inhibition constants were determined from experiments at four different concentrations of each test compound. The final K_i is shown as an average with its SD. For the enzymatic assay, inhibition constants were obtained by nonlinear least-squares methods in GraphPad Prism (for more experimental details, see Materials and Methods and the inhibition dose–response curves used for fitting in Suppl. Fig. S2). (B) Compounds that were shown to hydrolyze and form FBSA. (C) Compounds where FBSA was identified by LC-MS analysis, but they do not contain the FBSA group within their structures.

To confirm the presence of FBSA in the analyzed compounds and to find out whether the observed inhibition activity against CAIX is attributed to the actual compounds or FBSA, we synthesized FBSA, analyzed it by LC-MS, and determined its inhibition potency against CAIX. Indeed, the chromatographic and mass spectral data of FBSA confirmed its presence in compounds 1–5 and 9–13 (data not shown). Using FBSA as the standard, we performed a semiquantitative analysis, integrating the TIC signal from MS data, to determine approximate amounts of FBSA in each compound. We found a correlation (R² = 0.662) between the amount of FBSA present in the test compounds and their apparent K_i values (summarized in Suppl. Fig. S1 with the raw data in Suppl. Table S2 ). These data suggest that the observed inhibition activity in compounds 1–5 and 9–13 was caused by FBSA, a novel highly potent low nanomolar inhibitor of CAIX (K_i = 17 ± 2 nM).

To conclude, the number of hits identified via the DIANA screen was upon validation reduced from 13 to 3 unique chemical entities, namely, FBSA, compound 6, and compound 7. All three compounds are sulfonamides derivatives with low nanomolar (compound 6 and FBSA) and low micromolar (compound 7) inhibition potencies against CAIX, respectively. The structures of compound 6 and FBSA together with their K_i values determined by both DIANA and orthogonal inhibition assay are shown in Figure 3A . Compound 7 is not shown since it is part of another study conducted at IOCB Prague and its structure cannot be disclosed at this point.

Binding Modes of Novel CAIX Inhibitors

To elucidate the binding mode of the identified compounds, we co-crystallized FBSA and compound 6 with a CAIX mimic, the variant that is widely used in structural studies to resemble CAIX. The CAIX mimic contains seven amino acid substitutions (A65S, N67Q, E69T, I91L, F130V, K169A, and L203A) in the CAII active site to mimic the active site of CAIX, while it retains good crystallization properties.^20,21 The co-crystal structure with FBSA (PDB code 6RZX) and compound 6 (PDB code 6S03) was determined at resolutions of 1.0 and 1.4 Å, respectively (see Suppl. Table S3 for crystal parameters, data collection, and refinement statistics).

The high quality of electron density maps allowed for unambiguous modeling of compounds in the CAIX active site (see Fig. 4 ). Both compounds contain the sulfonamide moiety, which is deeply buried in the conical-shaped cavity of the active site and makes several polar interactions with three conserved histidine residues (His94, His96, and His119) coordinating the zinc ion located at the bottom of the active site cavity. Such binding interactions are typical for other sulfamide- and sulfonamide-containing inhibitors, which were successfully crystallized in the past.^22,23

Figure 4.

Structures of FBSA and compound 6 bound to the CAIX active site (colored figure). Compounds are depicted in stick representation with different colors of carbon atoms, FBSA in yellow (PDB code 6RZX) and compound 6 in magenta (PDB code 6S03); fluorine atoms are colored light blue, and oxygen, nitrogen, and sulfur in standard colors as blue, red, and yellow, respectively. The zinc ion is depicted as a gray sphere. (A) A 2Fo-Fc map of electron density is shown and contoured at 1 σ. Both compounds were modeled in one conformation with full occupancy. (B) Top view of the CAIX active site, which is represented by its solvent accessible surface colored by electrostatic potential (red for negative, blue for positive). The position of interacting residues on the surface is highlighted. (C) Overview of interatomic interaction of compounds with CAIX. The protein is shown in the green cartoon representation with interacting residues highlighted as sticks. Polar interactions are represented by dashed lines.

Furthermore, FBSA forms an additional 29 interatomic contacts with Gln92, Val121, Leu140, Val142, Leu197, Thr198, Thr199, and Trp208, thus filling the bottom of the active site cavity. In contrast, much bulkier compound 6 makes an additional 45 interatomic contacts with CAIX active site residues. Compound 6 occupies the bottom of the catalytic cavity with its phenyl sulfonamide moiety in a similar manner to whole FBSA and interacts with Leu91, Gln92, Val121, Val130, Gly131, Val134, Leu140, Val142, Ser196, Leu197, Thr198–199, Pro201, and Trp208. The hydrophobic aromatic rings of compound 6, distal from the sulfonamide, interact with a hydrophobic pocket formed by Leu91, Val130, Val134, Leu140, and Val142. Extensive interaction of compound 6 with a hydrophobic pocket containing Val130, one of the key residues modifying CA specify, might explain the high inhibitory potency of compound 6 toward the CAIX isoform.

Discussion

In this work, we present the development of a high-throughput screen of CAIX inhibitors from a compound library and structural characterization of their binding mode to the enzyme. For the screening, we adapted the recently published DIANA method and used the proprietary in-house compound library prepared at IOCB Prague.

DIANA Development and Validation

First, we analyzed if the HTS format does not compromise DIANA’s advantageous features. Compared with the original DIANA, the current assay setup was miniaturized to the automated 384-well plate format while it maintained an identical assay window and consumption of target enzyme (approximately 400 pg of CAIX present in 5 µg of the HT-29 cell lysate), thus confirming that the assay keeps its minimal requirements on target amount and purity. Furthermore, implementation of the liquid handling platforms minimizes human errors and increases assay reproducibility. This fact together with the large assay window contributed to excellent Z′ factors (>0.89) for the assay in all conducted experiments, demonstrating its overall robustness.

Additional advantageous features of DIANA are its low false-positive and false-negative rates. While the determination of false positives of the assay is straightforward, using orthogonal inhibition assays after the initial screen, the determination of the false negatives is challenging since these compounds are not identified during the initial screen as hits. However, during our validation of the pooled DIANA screen, we collected two independent datasets (from pooled and nonpooled screen) that enabled us to at least estimate the false-negative rates of the assay to be very low. As summarized in Table 2 , we identified all the hits from the nonpooled screen that were also in the pooled screen. Moreover, we identified one extra positive well in the pooled screen that did not contain a hit from the individual screen. However, this pooled well contained one compound that provided a dCq signal of 1.42, which is barely below the set threshold (dCq = 1.57). Regarding the false-positive rate of the assay, we did not identify a single compound that would be positive in DIANA and not show the inhibition activity in the orthogonal inhibition assay ( Fig. 2B ). Overall, the presented data indicate DIANA’s extremely low false-negative and false-positive rates and its suitability for pooling strategies.

As stated in the Results section, we intentionally set the cutoff threshold in the DIANA screen to a dCq value of 1.57, which corresponds to the K_i value of the compound concentration in the screen (10 µM). Given the very low standard deviation (SD) of DIANA and the large assay window, our cutoff threshold was much more stringent (eight times SD above blank wells without inhibitors) than the generally accepted statistical threshold (three to five times SD above blank). This assay feature ensures a very low probability of obtaining false-positive hits; in theory, the statistical chance of false-positive hit occurrence assuming normal data distribution would be lower than 10^–14. If the threshold would be calculated as a signal corresponding to blank measurement with five times of its SD added (dCq = 0.92), it would correspond to only 28% inhibition of CAIX under the assay condition and still provide a stochastic false-positive rate lower than 0.001%. Such a threshold would enable us to identify compounds with a K_i value lower than 20 µM while using only the 10 µM compound concentration during the screen.

Yet another beneficial feature of DIANA is its ability to calculate the compound inhibition potency from a single well measurement and thus rank and quantify the inhibition potency of the hits right after the initial screen. This is enabled by the sensitivity of the detection and by the possibility to accurately determine inhibition in a wide range. Since qPCR has a dynamic range of up to nine orders of magnitude and it is a logarithmic readout measuring the fold change rather than absolute change, and at the same time DIANA is detecting the amount of free enzyme unoccupied by the inhibitor, virtually any inhibition above 25% can be measured with high precision. Specifically, the relative coefficient of variation of determined inhibition is usually lower than 10% of the remaining free enzyme; for example, 90% inhibition is determined with an error of less than 1%, while 99.9% inhibition has an error of less than 0.01%. These properties translate into the ability of DIANA to determine K_i values from a single well measurement.

To confirm that the results obtained during the DIANA screen at one compound concentration can be used for K_i determination, we compared the K_i values from the nonpooled screen with K_i values from both DIANA run at four different compound concentrations and an established orthogonal CAIX enzymatic inhibition assay. As shown in Figure 2 , data from both assays show very good correlation with data obtained from the DIANA screen apart from compounds that had high inhibition potencies (K_i < 100 nM). This observation can be explained by the fact that under the given assay conditions (such as the assay window, the concentration of test compound, and usage of bivalent probe), the range of K_i values that can be reliably determined in the presented CAIX DIANA screen at a 10 µM concentration of the test compounds is 100 nM to 10 µM. The compounds with a K_i lower than 100 nM are able to outcompete almost all DIANA probes, and thus all of them appear as 100 nM inhibitors, while the compounds with K_i higher than 10 µM are unable to outcompete a sufficient amount of probe to provide a reliable signal change compared with the signal from the noninhibited reaction. This phenomenon can be demonstrated on the known potent CAIX inhibitor AZA that was used as a control for all DIANA experiments. From the DIANA screens the K_i value of AZA calculated from the 10 µM concentration was 99 ± 10 nM (average of 10 measurements with SD; data not shown), while if calculated from serial dilutions within the DIANA range the K_i value was 7 ± 1 nM (the value is an average from eight measurements with SD; data not shown), a value that is in agreement with previously reported data.¹³ Nevertheless, this feature does not represent a significant disadvantage of the assay, since all highly potent compounds will still be ranked as the most potent hits within the screen and there is no danger that they would be omitted from further analysis. The only limitation will lie in the fact that within this group of potent hits the ranking will be less precise.

The range of inhibitory potencies that can be accurately determined in DIANA is very broad thanks to the possibility to calculate a K_i value from a single point measurement. The lowest detectable potency (highest K_i) approximately equals the highest tested inhibitor concentration. The highest accurately measurable potency (lowest K_i) is limited by the assay window and by the amount of enzyme used for screening, as the inhibitor should not be assayed below the enzyme concentration. In this assay using endogenous CAIX, it is possible to calculate K_i values in the range of 0.01-fold to 1-fold of the assayed inhibitor concentration; that is, the lowest measurable K_i is ~100-fold below the used enzyme concentration. Specifically, CAIX concentration in the assay is approximately 2 nM, which translates into the lowest precisely measurable K_i of 20 pM, while inhibitors with lower K_i would appear as 20 pM as well. A higher assay window, which can be gained by using recombinant and not endogenous CAIX, would enable precise measurement of even lower K_i values. As shown for prostate-specific membrane antigen, the DIANA window can be as high as six orders of magnitude, resulting in the lowest measurable K_i value of 1 fM.¹⁵

In DIANA, any compound capable of outcompeting the probe out of the active site is identified as a hit, and thus any competitive inhibitor should be identified. Even though the probe binds to the active site, the inhibitor does not necessarily need to bind to the same site. For example, allosteric inhibitors bind to the exosites and cause structural changes within the enzyme active site and thus prevent the substrate/probe binding. Indeed, such an allosteric inhibitor has been identified during an internal DIANA screen against another enzyme (prostate-specific membrane antigen) at our institute and its allosteric mode of action was confirmed by x-ray crystallography (unpublished data). However, if the enzyme contained two completely independent binding sites, such as cofactor and peptide binding site, and inhibitors of both sites were of interest, then the screening could be run with two probes, each targeting different binding sites. Even though this is associated with extra effort to develop two probes, the extra time and cost associated with such a screen could be minimized by multiplex qPCR, and thus screening both probes in the same well. Use of multiple probes would bring the possibility to decipher the binding site of the hits.

Even though only a small compound library (~3000 compounds) was used for validation of DIANA, the performed analysis confirms that the assay is highly robust with very low false-positive and false-negative rates. Notably, the presented pooling approach significantly reduces the time of the screen as it first tests mixtures of compounds in the initial pooled screen, followed by deconvolution testing of only positive pooled wells. The number of compounds pooled together must be appropriately chosen taking into the account the expected positive hit rate; for the majority of random libraries the expected positive hit rate ranges from 0.2% to 0.5 %. Additionally, it is important to have a reasonable estimate of the false-positive rate of the used screening assay; since we did not identify a single false-positive hit, this rate will be <0.04% for DIANA. Based on the presented data, we deduce that the ideal number of pooled compounds will be in the range of 10–20 compounds per well. In line with these findings, we have pooled 11 compounds per well. Considering the expected ultra-low false-positive rate of DIANA, this number could be even bigger for difficult targets with low hit rates.

Finally, it should be noted that DIANA’s throughput is in the range of 50,000 to 100,000 test compounds per day on one qPCR instrument when 10–20 compounds are pooled per well. Since the current bottleneck of the assay is the qPCR readout, the utilization of more than one qPCR instrument enables further upscaling of the method to significantly more than 100,000 test compounds per day.

Hit Validation and Characterization

From the initial screen we identified 13 hits with inhibition potency below 10 µM. During assay validation we repeatedly determined the inhibition potency of the hits not only from solutions stored within the library but also from freshly prepared solutions of the stored powders. During this analysis we identified an unexpected discrepancy for compounds 1 and 3 between the inhibition potency of their solutions retested from the library, which inhibited as expected, and fresh solutions prepared from their solids, which showed negligible inhibition activity. To address these differences, we performed LC-MS analysis of both solutions for these two compounds. The analysis revealed that solutions retested from the library are partially hydrolyzed, while solutions from powders show one compound with expected mass (data not shown). In hydrolyzed samples, there was one common product for both compounds. This degradation product did not absorb at UV 210 nm and was visible only in the negative mode of MS analysis with a molecular mass of 297.96 Da. Upon inspection of the chemical formula of compounds 1 and 3, we derived, and by further experiments confirmed, that the contaminant is FBSA, which acts as a potent low nanomolar inhibitor of CAIX.

This finding would suggest that FBSA is responsible for the inhibition activity of compounds 1 and 3, whereas the expected compounds may have negligible inhibition potency. Therefore, we decided to analyze all identified hits from our screen by the LC-MS method to check for the presence of FBSA. As summarized in Supplemental Table S2 , FBSA was identified in all compounds that contained its moiety within their structures (compounds 1–5). Interestingly, we also detected low amounts of FBSA in compounds 9–13 that did not contain FBSA moiety (see Fig. 3B,C for structures of compounds 1–5 and 9–13). Correlation of FBSA amounts and compound inhibition potencies (see Suppl. Fig. S1 ) strongly suggests that FBSA is responsible for most of the observed inhibition activity of test compounds. FBSA in compounds 1–5 originates most probably from partial hydrolytic cleavage of these hemiaminals, such as the aminobenzotetrahydrofuran (compound 1).²⁴ These hemiaminals were synthetically in situ transformed to more stable compounds 2–5, which could, however, still loose FBSA by elimination reactions. This fact also explains the difference between inhibition potencies of solutions retested from the library and solutions freshly prepared from isolated precursors—the former solution being more prone to hydrolysis or elimination reactions in solution over time, thus containing more FBSA and showing significantly higher potency than the latter. On the other hand, the presence of FBSA in compounds 9–13 can be rationalized by the fact that the diazene bridge in these bicyclic compounds was introduced by a diazo transfer reaction using nonafluorobutanesulfoyl azide, generating FBSA as a stoichiometric byproduct in the reactions (the synthesis of compounds 9–13 is described elsewhere²⁵). FBSA can be separated from the products by column chromatography, but apparently very small amounts remained, which were not detectable by the common nuclear magnetic resonance (NMR) techniques and electrospray ionization–mass spectrometry (ESI-MS) routinely used for mass spectrometric analysis of compounds 9–13. Altogether, the FBSA was most likely the sole molecule responsible for the inhibition activity of compounds 1–5 and 9–13.

We identified three novel inhibitors (FBSA and compounds 6 and 7) from screening the IOCB library against CAIX. FBSA and compound 6 are low nanomolar inhibitors, while compound 7 is a micromolar inhibitor of CAIX. All these compounds contain a sulfonamide moiety in their structure, which is a known structural feature of CA inhibitors that binds to the zinc ion within the enzyme active site. To confirm the mode of inhibition of these compounds, we co-crystallized FBSA and compound 6 with CAIX. Indeed, the inhibitors bind to the zinc ion in CAIX active via sulfonamide moiety, as expected.

Since sulfonamide derivatives are generally nonspecific CA inhibitors, it can be expected that the identified inhibitors will also inhibit other CA isoforms. Considering the extensive interatomic interaction of compound 6 with CAIX, this inhibitor may show inhibition specificity toward CAIX over other CAs. On the other hand, the FBSA makes most interatomic interaction with CAIX at the bottom of its active site, which is highly conserved within CAs, and thus will most probably show very broad inhibition potency over the CA isoforms. However, the inhibition specificity of these novel CAIX inhibitors within the family of human CAs remains to be determined.

In conclusion, we present data from a screen of the IOCB medicinal compound library against CAIX using the novel HTS method DIANA. We showed that DIANA represents a very robust and efficient HTS platform for the screening of enzyme inhibitors that requires a minimal amount of nonpurified target enzyme, has negligible false-positive and false-negative rates, and enables screening in compound mixtures. Moreover, DIANA can be easily adapted to various enzymatic and nonenzymatic (e.g., protein–protein interactions) targets. From the screen, we identified three novel inhibitors of CAIX within the IOCB library. They are all sulfonamide-based compounds, and thus we have not identified any structurally unexplored scaffolds within the IOCB library. Nevertheless, the presented DIANA is the first assay suitable for HTS of CAIX inhibitors, and its use on larger libraries may lead to the identification of completely novel scaffolds in the future.

Supplemental Material

DISC918836-Supplement – Supplemental material for Identification of Novel Carbonic Anhydrase IX Inhibitors Using High-Throughput Screening of Pooled Compound Libraries by DNA-Linked Inhibitor Antibody Assay (DIANA)

Supplemental material, DISC918836-Supplement for Identification of Novel Carbonic Anhydrase IX Inhibitors Using High-Throughput Screening of Pooled Compound Libraries by DNA-Linked Inhibitor Antibody Assay (DIANA) by Jan Tykvart, Václav Navrátil, Michael Kugler, Pavel Šácha, Jiří Schimer, Anna Hlaváčková, Lukáš Tenora, Jitka Zemanová, Milan Dejmek, Vlastimil Král, Milan Potáček, Pavel Majer, Ullrich Jahn, Jiří Brynda, Pavlína Řezáčová and Jan Konvalinka in SLAS Discovery

Footnotes

Acknowledgements

The use of the beamline MX14.1 operated by the Helmholtz-Zentrum Berlin at the BESSY II electron storage ring (Berlin-Adlershof, Germany) to collect diffraction data is gratefully acknowledged. We would also like to thank Milan Fábry for preparation of expression plasmid for the CAIX mimic and Radko Souček for his excellent assistance with the quantification of FBSA by LC-MS.

Supplemental material is available online with this article.

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: Patent applications describing the DIANA technological platform have been filed. V.N., J.S., J.K., and P.Š. are co-owners of a company commercializing the technology.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported in part by the Grant Agency of the Czech Republic, projects 19-10280S and 17-14510S; in part by the European Regional Development Fund, OP RDE (no. CZ.02.1.01/0.0/0.0/16_019/0000729); and in part by the Ministry of Education of the Czech Republic (program “NPU I”), project LO1304.

ORCID iD

Jan Tykvart

References

Pastorekova

Gillies

R. J.

The Role of Carbonic Anhydrase IX in Cancer Development: Links to Hypoxia, Acidosis, and Beyond. Cancer Metastasis Rev. 2019, 38, 65–77.

Alterio

Hilvo

Fiore

A. D.

; et al. Crystal Structure of the Catalytic Domain of the Tumor-Associated Human Carbonic Anhydrase IX. Proc. Natl. Acad. Sci. U.S.A. 2009, 106, 16233–16238.

Wykoff

C. C.

Beasley

N. J. P.

Watson

P. H.

; et al. Hypoxia-Inducible Expression of Tumor-Associated Carbonic Anhydrases. Cancer Res. 2000, 60, 7075–7083.

Závada

Závadová

Zat’ovičová

; et al. Soluble Form of Carbonic Anhydrase IX (CA IX) in the Serum and Urine of Renal Carcinoma Patients. Br. J. Cancer 2003, 89, 1067–1071.

Chiche

Ilc

Laferrière

; et al. Hypoxia-Inducible Carbonic Anhydrase IX and XII Promote Tumor Cell Growth by Counteracting Acidosis through the Regulation of the Intracellular PH. Cancer Res. 2009, 69, 358–368.

Thiry

Dogné

J.-M.

Masereel

; et al. Targeting Tumor-Associated Carbonic Anhydrase IX in Cancer Therapy. Trends Pharmacol. Sci. 2006, 27, 566–573.

Pacchiano

Carta

McDonald

P. C.

; et al. Ureido-Substituted Benzenesulfonamides Potently Inhibit Carbonic Anhydrase IX and Show Antimetastatic Activity in a Model of Breast Cancer Metastasis. J. Med. Chem. 2011, 54, 1896–1902.

Kazokaitė

Aspatwar

Parkkila

; et al. An Update on Anticancer Drug Development and Delivery Targeting Carbonic Anhydrase IX. PeerJ 2017, 5, e4068.

Krall

Pretto

Neri

A Bivalent Small Molecule-Drug Conjugate Directed against Carbonic Anhydrase IX Can Elicit Complete Tumour Regression in Mice. Chem. Sci. 2014, 5, 3640–3644.

10.

Hsu

P.-L.

J. K. H.

Jun

H. W.

; et al. Structure Relationship for Binding of Sulfonamides and Penicillins to Bovine Serum Albumin by Fluorescence Probe Technique. J. Pharm. Sci. 1974, 63, 27–31.

11.

Kelly

T. E.

Hackett

P. H.

Acetazolamide and Sulfonamide Allergy: A Not So Simple Story. High Alt. Med. Biol. 2010, 11, 319–323.

12.

Wulf

N. R.

Matuszewski

K. A.

Sulfonamide Cross-Reactivity: Is There Evidence to Support Broad Cross-Allergenicity?

Am. J. Health. Syst. Pharm. 2013, 70, 1483–1494.

13.

Supuran

C. T.

Carbonic Anhydrases: Novel Therapeutic Applications for Inhibitors and Activators. Nat. Rev. Drug Discov. 2008, 7, 168–181.

14.

Linkuvienė

Zubrienė

Manakova

; et al. Thermodynamic, Kinetic, and Structural Parameterization of Human Carbonic Anhydrase Interactions toward Enhanced Inhibitor Design. Q. Rev. Biophys. 2018, 51, e10.

15.

Navrátil

Schimer

Tykvart

; et al. DNA-Linked Inhibitor Antibody Assay (DIANA) for Sensitive and Selective Enzyme Detection and Inhibitor Screening. Nucleic Acids Res. 2017, 45, e10.

16.

Clercq

E. D.

Holý

Acyclic Nucleoside Phosphonates: A Key Class of Antiviral Drugs. Nat. Rev. Drug Discov. 2005, 4, 928–940.

17.

Khalifah

R. G.

The Carbon Dioxide Hydration Activity of Carbonic Anhydrase I. Stop-Flow Kinetic Studies on the Native Human Isoenzymes B and C. J. Biol. Chem. 1971, 246, 2561–2573.

18.

Pospíšilová

Knedlík

Šácha

; et al. Inhibitor–Polymer Conjugates as a Versatile Tool for Detection and Visualization of Cancer-Associated Carbonic Anhydrase Isoforms. ACS Omega 2019, 4, 6746–6756.

19.

Copeland

R. A.

Enzymes. A Practical Introduction to Structure, Mechanism, and Data Analysis, 2nd ed.; Wiley-VCH: New York, 2000.

20.

Mujumdar

Teruya

Tonissen

K. F.

; et al. An Unusual Natural Product Primary Sulfonamide: Synthesis, Carbonic Anhydrase Inhibition, and Protein X-Ray Structures of Psammaplin C. J. Med. Chem. 2016, 59, 5462–5470.

21.

Pinard

M. A.

Boone

C. D.

Rife

B. D.

; et al. Structural Study of Interaction between Brinzolamide and Dorzolamide Inhibition of Human Carbonic Anhydrases. Bioorg. Med. Chem. 2013, 21, 7210–7215.

22.

Eriksson

A. E.

Kylsten

P. M.

Jones

T. A.

; et al. Crystallographic Studies of Inhibitor Binding Sites in Human Carbonic Anhydrase II: A Pentacoordinated Binding of the SCN– Ion to the Zinc at High pH. Proteins 1988, 4, 283–293.

23.

Nocentini

Trallori

Singh

; et al. 4-Hydroxy-3-Nitro-5-Ureido-Benzenesulfonamides Selectively Target the Tumor-Associated Carbonic Anhydrase Isoforms IX and XII Showing Hypoxia-Enhanced Antiproliferative Profiles. J. Med. Chem. 2018, 61, 10860–10874.

24.

Hernández-Guerra

Hlavačková

Pramthaisong

; et al. Photochemical C–H Amination of Ethers and Geminal Difunctionalization Reactions in One Pot. Angew. Chem. Int. Ed. 2019, 58, 12440–12445.

25.

Kapras

Pohl

Císařová

; et al. Asymmetric Domino Aza-Michael Addition/[3 + 2] Cycloaddition Reactions as a Versatile Approach to α,β,γ-Triamino Acid Derivatives. Org. Lett. 2014, 16, 1088–1091.

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

0.49 MB