RNase P Inhibitors Identified as Aggregators

ABSTRACT

RNase P is an essential enzyme responsible for tRNA 5′-end maturation. In most bacteria, the enzyme is a ribonucleoprotein consisting of a catalytic RNA subunit and a small protein cofactor termed RnpA. Several studies have reported small-molecule inhibitors directed against bacterial RNase P that were identified by high-throughput screenings. Using the bacterial RNase P enzymes from Thermotoga maritima, Bacillus subtilis, and Staphylococcus aureus as model systems, we found that such compounds, including RNPA2000 (and its derivatives), iriginol hexaacetate, and purpurin, induce the formation of insoluble aggregates of RnpA rather than acting as specific inhibitors. In the case of RNPA2000, aggregation was induced by Mg²⁺ ions. These findings were deduced from solubility analyses by microscopy and high-performance liquid chromatography (HPLC), RnpA-inhibitor co-pulldown experiments, detergent addition, and RnpA titrations in enzyme activity assays. Finally, we used a B. subtilis RNase P depletion strain, whose lethal phenotype could be rescued by a protein-only RNase P of plant origin, for inhibition zone analyses on agar plates. These cell-based experiments argued against RNase P-specific inhibition of bacterial growth by RNPA2000. We were also unable to confirm the previously reported nonspecific RNase activity of S. aureus RnpA itself. Our results indicate that high-throughput screenings searching for bacterial RNase P inhibitors are prone to the identification of “false positives” that are also termed pan-assay interference compounds (PAINS).

INTRODUCTION

RNase P is responsible for the endonucleolytic removal of 5′ leader sequences from tRNA precursors (1, 2). The vast majority of known bacteria encode an RNA-based enzyme consisting of a catalytic RNA subunit, ∼300 to 400 nucleotides (nt) in length, and a single small protein cofactor of ∼13 kDa. Both subunits are essential in vivo (3–5). Human cells harbor an RNA-based enzyme with 10 protein subunits in the cell nucleus and a protein-only RNase P in mitochondria (6–8). Both human enzymes lack substantial structural similarities to the bacterial enzyme, a favorable precondition for the development of new antibiotics that selectively inhibit bacterial RNase P enzymes. Beyond this, RNase P fulfils the key condition of belonging to the ∼7% of bacterial genes that are indispensable for the survival of bacteria (9, 10).

Specific antisense oligonucleotides were demonstrated to inhibit the catalytic RNA subunit (P RNA) of bacterial RNase P (11–13), but their efficient uptake by bacterial cells remains a major hurdle for therapeutic applications. On the other hand, potential small-molecule inhibitors of P RNA, such as aminoglycosides (14–16) and phenothiazine derivatives (17), showed insufficient P RNA specificity. Apart from targeting the catalytic P RNA subunit, there have been attempts to address the protein subunit (termed RnpA or P protein) of bacterial RNase P by small-molecule inhibitors.

Fierke and coworkers identified iriginol hexaacetate (Ir6Ac; see Table S1 in the supplemental material) as an inhibitor of Bacillus subtilis RNase P with a 50% inhibitory concentration (IC₅₀) of ∼0.8 μM (18). As there was no evidence for binding of the compound to the precursor tRNA substrate, Ir6Ac appeared to be a more specific inhibitor of bacterial RNase P than aminoglycosides and phenothiazine derivatives (17, 19). However, it remained elusive whether the compound binds to RnpA or to P RNA. A thiosemicarbazide derivative, RNPA2000 (see Fig. S1 in the supplemental material), was reported to interact with RnpA and to inhibit in vitro processing by RNase P of Staphylococcus aureus with an IC₅₀ of ∼140 μM (20, 21). The compound was described to have bactericidal activity (MIC, 16 μg/ml) without eliciting human cell cytotoxicity after 24 h of incubation at 256 μg/ml. In addition, the authors described RNPA2000 as a dual-function antimicrobial compound that also inhibits a nonspecific RNA degradation activity reported to be associated with recombinant S. aureus RnpA protein at an IC₅₀ of 125 μM in vitro (20). In structure-activity relationship studies, RNase P inhibition was improved ∼100-fold for derivatives of RNPA2000 (with IC₅₀ values of 1 μM for the two most effective compounds), but antimicrobial activities were not improved at the same time (21). Recently, a 1,2,4-trihydroxyanthraquinone (purpurin) was identified as an RnpA binder in a screen using a fluorescence-based enzyme assay. Purpurin was reported to inhibit the Thermotoga maritima RNase P holoenzyme reaction in the micromolar range (IC₅₀ values of ∼96 μM and ∼13 μM, depending on the substrate [22]) and its binding to T. maritima RnpA was inferred from biolayer interferometry assays. Finally, a crystal structure of the T. maritima RnpA protein in complex with purpurin was reported (PDB entry 6MAX) (22).

During in vitro studies of RNase P inhibition by RNPA2000, we noticed solubility problems of the compound in Mg²⁺-containing buffers, suggesting that the compound may induce protein aggregation. In particular, hits from high-throughput screening and virtual screening campaigns have often been identified as “false positives” or pan-assay interference compounds (PAINS). These compounds act by a variety of mechanisms, including interference with fluorescent signals, metal ion complexation, covalent protein reactivity, redox cycling, and formation of aggregates (23). In particular, small molecule-induced aggregation is a major source of false-positive results in drug discovery campaigns. Although computational methods to predict potential aggregators have been established (24), experimental analyses are still indispensable for the detection of aggregators. As the three aforementioned RNase P inhibitors have been identified via high-throughput screening and the RnpA protein has been described as “intrinsically unstructured” (25), suggesting increased susceptibility to aggregation, we decided to perform appropriate control experiments using several experimental approaches, as follows: (i) addition of detergent to enzyme kinetic reactions (26), (ii) variation of RNase P holoenzyme/RnpA concentration (27), (iii) solubility determination of the small molecule under assay conditions (28), and (iv) cosedimentation analysis of RnpA protein and the small molecule (26), and, for RNPA2000, growth inhibition of a Bacillus subtilis strain whose RNase P activity is provided by a eukaryal protein-only RNase P. All three types of compounds, Ir6Ac, RNPA2000 (and derivatives), and purpurin, showed properties of protein aggregators. Finally, we could not reproduce the reported unspecific RNase activity of Staphylococcus aureus RnpA (29).

RESULTS

RNPA2000 testing using the T. maritima enzyme.

For evaluation and improvement of published RNase P inhibitors, we used the RNase P of T. maritima as a model enzyme because a high-resolution crystal structure of the T. maritima P protein (RnpA) is available (30), making in silico docking and crystal soaking experiments feasible. At the IC₅₀ of RNPA2000 (140 μM) determined previously for S. aureus RNase P, we observed ∼60% inhibition of pre-tRNA processing by T. maritima RNase P in our experimental setup (Fig. 1A). This effect was comparable to that reported for S. aureus RNase P (20, 31). Neomycin, an aminoglycoside and known RNase P inhibitor, was used as a control. Its IC₅₀ value of ∼40 μM, determined here (see Fig. S2A in the supplemental material), is also in line with data published for the Escherichia coli RNase P holoenzyme (50% inhibition at 60 μM [14]). Next, RNPA2000 was tested in the P RNA-alone reaction (without RnpA) to ensure that the compound exclusively acts on the protein subunit. Indeed, as for the S. aureus enzyme (20), no inhibitory effect was detectable in the RNA-alone reaction (Fig. 1B). As this reaction was performed at 100 mM Mg²⁺ to compensate for the absence of the RnpA protein, inhibition by neomycin was essentially abolished as well, which is attributable to the elevated Mg²⁺ concentration. Neomycin was shown to compete with Mg²⁺ ions for binding to functionally important metal ion binding sites in the catalytic RNA (14). We further tested processing activity by T. maritima RNase P in the absence versus presence of 2% and 10% dimethyl sulfoxide (DMSO). In our setup, activity reached 80% at 2% and 57% at 10% DMSO compared with the “no DMSO” control (Fig. S2B). This finding opened up the option to analyze RNase P inhibition in 10% DMSO at increased concentrations of compounds with too low solubility at 2% DMSO. We further showed concentration-dependent inhibition of T. maritima RNase P by RNPA2000 (see Fig. S3 in the supplemental material). At this point, we observed turbidity of a solution containing 500 μM RNPA2000 in 2% DMSO. Therefore, experiments addressing compound solubility were initiated.

FIG 1.

Inhibition of T. maritima RNase P by RNPA2000. (A) T. maritima RNase P holoenzyme activity in the presence of 140 μM RNPA2000 (green). Kinetics were performed with 50 nM holoenzyme and 500 nM pre-tRNA^Gly substrate in kinetic (KIN) buffer with 10 mM MgCl₂. Neomycin (Neo; pink) was used for all experiments (A to D) as a control at its 50% inhibitory concentration (IC₅₀) of ∼40 μM (see Fig. S2). Mean values (± standard error of the mean [SEM]), normalized to the 2% dimethyl sulfoxide (DMSO)-only control, are based on 6 biological replicates. (B) Effect of 140 μM RNPA2000 on the T. maritima P RNA-alone reaction, using KIN buffer with 100 mM MgCl₂, 50 nM P RNA, and 500 nM pre-tRNA^Gly. Mean values (±SEM), normalized to the 2% DMSO-only control, are based on 3 biological replicates. (C) Effect of addition of 0.01% Triton X-100. Relative activity of the T. maritima RNase P holoenzyme assayed in KIN buffer with 10 mM MgCl₂, 0.01% Triton X-100, and 2% DMSO. Experiments were performed with 50 nM holoenzyme (equal amounts of P RNA and RnpA) and 500 nM pre-tRNA^Gly. Mean values (±SEM), normalized to the 2% DMSO-only control, are based on 4 biological replicates. (D) Effect of a 10-fold increase in RnpA concentration. Activity of the T. maritima holoenzyme was assayed in KIN buffer with 10 mM MgCl₂ and containing 50 nM P RNA, 500 nM RnpA protein, and 500 nM pre-tRNA^Gly. Mean values (±SEM), normalized to the 2% DMSO-only control, are based on 3 biological replicates.

Solubility analysis of RNPA2000.

Microscopy was performed to evaluate compound solubility. Dunman and coworkers (31) observed solubility of RNPA2000 at 200 μM (in phosphate-buffered saline [PBS] with 2% DMSO), thus inferring the solubility to be >200 μM (20). Consistent with this, we observed RNPA2000 to be soluble at a concentration of 200 μM in three tested buffer systems (kinetic [KIN], Tris, and PBS; each with 2% DMSO) (Fig. 2A, top). However, upon addition of Mg²⁺, precipitates formed in all three buffer systems (Fig. 2A, middle and bottom). To measure more quantitatively if such precipitates also form at concentrations near the IC₅₀, solubility was determined using high-performance liquid chromatography (HPLC). In accordance with the microscopic findings, the solubility of RNPA2000 severely decreased upon addition of Mg²⁺ ions (Table 1; see also Fig. S4 in the supplemental material). For the Tris buffer system with 5 mM Mg²⁺, conditions that were previously used for kinetics with the S. aureus enzyme (20), a solubility of ∼25 μM was determined. This is far below the reported IC₅₀ of ∼140 μM, and even increasing the DMSO percentage to 10% had only a small impact (solubility, ∼40 μM; Table 1). For the KIN buffer system, similar results were obtained (∼30 μM for 2% DMSO and ∼60 μM for 10% DMSO in the presence of 10 mM MgCl₂ [the Mg²⁺ concentration used for enzyme kinetics]). For both buffers, the solubility of RNPA2000 decreased ∼10-fold after addition of MgCl₂, indicating metal ion complexation.

FIG 2.

Protein aggregation. (A) Microscopy of RNPA2000 in different buffers with various Mg²⁺ concentrations. Three different buffers were chosen, as follows: (i) KIN buffer, used for enzyme kinetics in the present study, (ii) 50 mM Tris buffer, previously used for kinetic experiments (20), and (iii) phosphate-buffered saline (PBS), used for solubility determinations (31). Solutions (200 μl) with 2% DMSO and 200 μM RNPA2000 were prepared either without or with Mg²⁺ at the indicated concentration. After incubation for 15 min at room temperature, solutions were applied to microscopy. (B) SDS-PAGE of a cosedimentation experiment with RNPA2000; s, supernatant; p, pellet. Lane 1, a 15-kDa marker (M); lane 2, 1 μg of T. maritima RnpA; lanes 3 and 4, 5 μg T. maritima RnpA only; lanes 5 and 6, 140 μM RNPA2000 only; lanes 7 and 8, 5 μg T. maritima RnpA and 140 μM RNPA2000. Proteins in lanes 1 and 2 were in double-distilled water (ddH₂O), samples in lanes 3 to 8 were in KIN buffer containing 10 mM MgCl₂ and 2% DMSO, and lanes 9 and 10 were derived from samples containing 5 μg T. maritima RnpA and 140 μM RNPA2000 in KIN buffer with 2% DMSO but lacking MgCl₂ (indicated by an asterisk [*]). (C) Model sketches of protein aggregation induced by RNPA2000. (Left) RNPA2000 is soluble in the absence of Mg²⁺ (depicted by the homogeneous greenish color); (middle) addition of Mg²⁺ induces the formation of insoluble compound aggregates (green spheres); (right) protein (in blue) adsorbs to green RNPA2000 aggregates.

TABLE 1.

HPLC solubility determination results^a

Buffer	MgCl2 (mM)	DMSO	Solubility (μM)
			RNPA2000	NL20	NL48
KIN		2%	∼200	∼75	∼35
KIN	5	2%	∼25
KIN	10	2%	∼30
KIN	10	10%	∼60	∼210	∼13
Tris		2%	∼200
Tris	5	2%	∼25
Tris	5	10%	∼40
PBS		2%	∼200

^aRNPA2000 solubility was determined for different buffers with various amounts of MgCl₂ and dimethyl sulfoxide (DMSO). For NL20 and NL48, solubility was only analyzed in kinetic (KIN) buffer at 2% DMSO and without Mg²⁺, and under the conditions used for enzyme kinetic assays. PBS, phosphate-buffered saline. For each measurement, at least six solutions with increasing concentration were subjected to HPLC. After plotting area under the curve (AUC) values for the corresponding compound concentrations, a linear correlation was obtained. The solubility was determined as the highest concentration at which there was still a linear correlation to be detected. Solubility limits are approximate, as the concentration variation was performed with a resolution of 5 or 10 μM steps. Evaluation for RNPA2000 is exemplarily shown in Fig. S4. Each measurement is based on at least 3 replicates.

Protein aggregation.

The solubility results suggested that RNPA2000 was not completely dissolved during enzyme kinetics, raising the question of whether the observed inhibition effects were due to aggregation. Small-molecule aggregation has been extensively described as a source of false-positive hits in high-throughput screenings (27, 28, 32, 33). Mechanistically, it was proposed that proteins of interest adsorb to small-molecule aggregates, leading to unspecific inhibition effects (34). Partial protein denaturation induced by aggregation was also suggested as a cause of enzyme inhibition (35). Although the mechanistic details are not completely understood in all cases, there are some characteristics of aggregators studied so far, which can help in identifying such compounds.

As colloidal aggregates can be removed from solution by centrifugation (26, 32, 34), we performed cosedimentation (pulldown) experiments to reveal direct interaction of aggregates and enzymes. Indeed, in the presence of RNPA2000 aggregates, the T. maritima RnpA protein was found to be enriched in the centrifugation pellet (Fig. 2B, lane 8). In contrast, this was not seen when the same experiment was carried out in KIN buffer lacking Mg²⁺ ions (Fig. 2B, lane 10). This further corroborated the involvement of magnesium ion complexation in the aggregation process. This is schematically illustrated in Fig. 2C.

Addition of nonionic detergents as surface-active agents can prevent aggregation (illustrated in Fig. 3, compare left and right sketches), which manifests as a reduction or loss of the inhibitory effect. This approach has been widely used to detect aggregators in early drug discovery campaigns (23, 33, 36–38). Indeed, when T. maritima RNase P processing assays were conducted as in Fig. 1A, but additionally containing 0.01% Triton X-100, inhibition by RNPA2000 vanished almost completely, while inhibition strength by neomycin remained essentially the same (Fig. 1C).

FIG 3.

Model of aggregator effects on RNase P. (Left) After aggregate formation of the small molecule (green spheres), RnpA protein (blue) adsorbs to such aggregates, leading to the depletion of soluble protein that is available for enzyme function. (Middle) Increasing the RnpA concentration (10-fold in the sketch) saturates the contact sites on the aggregates, leaving a fraction of soluble protein, which manifests as attenuated enzyme inhibition. (Right) Triton X-100 interferes with aggregate formation or disrupts aggregates, thereby restoring protein solubility.

Finally, a strong sensitivity of aggregator inhibition to enzyme concentration has been described (28, 33, 39–41). Although the molar ratio of inhibitor to enzyme is roughly 3,000:1 in our experiments, the ratio of aggregate complexes to enzyme molecules will be much lower. As a consequence, increasing the enzyme concentration (e.g., 10-fold) may easily saturate adsorption of enzyme molecules to aggregates and increase the fraction of unaffected enzyme (illustrated in Fig. 3, compare central and left sketches), thus mitigating inhibition. In standard assays, we used 50 nM each T. maritima P RNA and RnpA protein because a molar excess of RnpA did not enhance but even reduced holoenzyme activity. To evaluate concentration sensitivity, we increased the amount of RnpA protein to 500 nM without changing the P RNA concentration (50 nM). In this setup, enzyme inhibition in the presence of 140 μM RNPA2000 decreased from ∼60% to ∼10%, while inhibition by neomycin remained unaffected (see Fig. 1A and D).

Taken together, the observed solubility problems, the co-pulldown of RNPA2000 and RnpA protein, sensitivity of enzyme inhibition to the presence of Triton X-100, and mitigated inhibition at elevated RnpA concentration are consistent with the notion that effects of RNPA2000 on bacterial RNase P activity are attributable to Mg²⁺-dependent aggregate formation (Fig. 2C). Mechanistically, we propose that Mg²⁺ ions induce the formation of RNPA2000 aggregates, which then adsorb RnpA protein molecules to their surface.

Derivatives of RNPA2000.

As part of structure-activity relationship (SAR) studies, derivatives of RNPA2000 with enhanced potency against bacterial RNase P were identified (21, 31). Among those were compounds NL20 (IC₅₀, 75 μM) (31) with a semicarbazide core instead of the RNPA2000 thiosemicarbazide core, and NL48 differing from RNPA2000 by the presence of a benzofuran substituent instead of the furan ring (see Fig. S1), which was reported to be one of the most active compounds against S. aureus RNase P (IC₅₀, 1 μM) (21, 31). For the two compounds, solubilities (in PBS containing 2% DMSO) were determined to be ∼41 (NL20) and 49 μM (NL48) (31). We determined the solubilities of NL20 and NL48 to be ∼75 μM and ∼35 μM, respectively, in KIN buffer with 2% DMSO (Table 1), which are in a similar range as the solubilities determined before using a different method (31). To maximize the likelihood that solubilities of the two compounds upon addition of Mg²⁺ are higher than their aforementioned IC₅₀ values for RNase P inhibition, we analyzed their solubilities at the highest feasible DMSO concentration (10%) under our standard assay conditions (KIN buffer, 10 mM Mg²⁺). We determined solubilities of ∼210 μM (NL20) and ∼13 μM (NL48) (Table 1), well above the reported IC₅₀ values of 75 μM (31) and 1 μM (21), respectively. Effects of NL20 and NL48 on the processing reaction catalyzed by T. maritima RNase P were then analyzed at concentrations of 175 μM (NL20) and 5 μM (NL48). As a control, we included RNPA2000 at 50 μM, which is just below the solubility limit under these conditions according to Table 1. However, the inhibitory effects published before (20, 21, 31) were not observable under our assay conditions (Fig. 4A), indicating that there is no inhibitory effect when NL20, NL48, or RNPA2000 is fully dissolved. As available information suggests that the compounds were previously tested at only 2% DMSO and in the presence of 5 mM Mg²⁺ (20, 31), it is very possible that fractions of compound NL20 and especially of NL48 were insoluble and induced aggregate formation in those studies (20, 21, 31).

FIG 4.

Effect of RNPA2000 and its derivatives NL20 and NL48 on RNase P cleavage. (A) Relative activity of the T. maritima RNase P holoenzyme in the presence of the three compounds (chemical structures shown in Fig. S1). Experiments were performed with 50 nM T. maritima holoenzyme and 500 nM pre-tRNA^Gly substrate using KIN buffer, 10 mM MgCl₂, and 10% DMSO. DMSO (10%) was chosen to maximize compound solubility while maintaining substantial enzyme activity (see Fig. S2). The compound concentrations used were selected based on the solubility data shown in Table 1; for NL20 and NL48, compound concentrations were above the IC₅₀ values reported for inhibition of S. aureus RNase P (75 μM for NL20 [31] and 1 μM for NL48 [21]). Mean cleavage rate constants (± SEM), derived from 3 biological replicates each, were normalized to reaction mixtures containing 10% DMSO without added compound. (B) Relative S. aureus RNase P holoenzyme activity in the absence or presence of RNPA2000 assayed in KIN buffer, 10 mM MgCl₂, and 2% DMSO. P RNA (50 nM) and substrate (500 nM) were used at the same concentrations as for kinetic assays with T. maritima RNase P (Fig. 1A). Only the S. aureus RnpA concentration was increased to 100 nM for optimal enzyme activity. The RNPA2000 concentration (140 μM) corresponds to the previously determined IC₅₀ value for the same enzyme reaction (20, 31). Mean values (±SEM), normalized to the 2% DMSO control, are based on 4 biological replicates. (C) S. aureus RNase P activity assayed in KIN buffer, 10 mM MgCl₂, and 10% DMSO, using the same enzyme and substrate concentrations as in panel B. Compound concentrations were identical to those used in panel A for the T. maritima enzyme. Mean values (±SEM), normalized to 10% DMSO, are based on 4 biological replicates.

S. aureus RNase P.

Dunman and coworkers identified RNPA2000 as an inhibitor of the RNase P of the human pathogen S. aureus (20). As we observed the solubility problems of RNPA2000 and derivatives in both buffer systems, the Tris buffer used in the previous study (20) and the KIN buffer used here (Table 1), we could essentially exclude our KIN buffer system as the cause of RnpA aggregation. Thus, we considered the possibility that the T. maritima RnpA protein used here and the one from S. aureus might have different propensities to induce protein aggregation. However, we assumed this to be unlikely, taking into account that RnpA proteins share a very similar overall structure (30, 42, 43) and previously identified aggregators inhibited a variety of enzymes via aggregation (27). Indeed, in the presence of 140 μM RNPA2000 in KIN buffer containing 2% DMSO, we saw ∼60% inhibition of S. aureus RNase P (Fig. 4B), in line with the previously described inhibition effects (20). However, when performing the enzyme assays at 10% DMSO to combine solubility with compound concentrations as close as possible to or above the reported IC₅₀ values (20, 21, 31), neither RNPA2000 nor NL20 and NL48 caused substantial inhibition effects (Fig. 4C). These findings support the notion that the inhibition effects reported for S. aureus RNase P resulted from protein aggregation as well.

Analysis of RNPA2000 specificity.

Although protein aggregation is described in literature as an unspecific effect, reported antibacterial effects without human cytotoxicity might indicate a certain specificity (20). To further investigate whether other proteins than RnpA are also affected by aggregation, cosedimentation experiments were performed with nucleic acid binding proteins such as the human helicase eIF4A and T7 RNA polymerase. Similarly as for T. maritima RnpA, the presence of 140 μM RNPA2000 enhanced the amount of protein in the pellet (see Fig. S5A and B in the supplemental material). This finding indicates that protein aggregation by RNPA2000 is rather unspecific. Next, the B. subtilis d7 strain was used to evaluate whether the observed antibacterial effects are RnpA specific. In the B. subtilis d7 strain, expression of the essential rnpA gene is under the control of a xylose-inducible promoter (4). Under nonpermissive conditions, RnpA levels are critically depleted, and survival of cells becomes dependent on a functional plasmid-encoded RNase P activity. Complementation experiments revealed that the protein-only RNase P PRORP3 from the plant Arabidopsis thaliana is able to restore B. subtilis growth under conditions of an otherwise lethal depletion of endogenous RnpA and RNase P holoenzyme activity, although growth was retarded relative to that of d7 cells complemented with the same plasmid but expressing B. subtilis RnpA (see Fig. S6 in the supplemental material). Using strain d7, we analyzed the antibacterial effects of RNPA2000 under permissive (in the presence of xylose) versus nonpermissive (in the presence of glucose) conditions. As RnpA expression is negligible in the presence of glucose as a carbon source and PRORP3 instead executes the basic RNase P activity in this strain, one would expect that a specific RnpA inhibitor exerts a reduced antibacterial effect on this strain. First of all, we observed different sensitivities of our test strains to ampicillin and RNPA2000 (Table 2): strain YB886, the parental strain used for the construction of strain d7, was the least sensitive one (smallest inhibition zone diameters); d7 strains harboring the pDG148 plasmid vector (d7 vector, d7 RnpA, and d7 PRORP3) tended to be more sensitive than strain d7 under xylose conditions (Table 2 and Fig. S7 in the supplemental material); this can be explained by exposure of strains harboring pDG148 plasmids to dual antibiotic (chloramphenicol and kanamycin) stress during growth. Comparison of strains d7 RnpA and d7 PRORP3 (xylose conditions) revealed that the latter strain was more sensitive to ampicillin as well as to 0.25 mM RNPA2000, which can be attributed to PRORP3 acting with lower efficiency than endogenous RNase P, resulting in lower overall fitness. Yet, under glucose conditions, one would have expected a stronger RNPA2000 inhibition effect on strain d7 RnpA than on d7 PRORP3 if RNPA2000 were a specific RnpA inhibitor. However, strain d7 PRORP3 even showed a trend toward stronger inhibition (larger diameter of inhibition zone) than d7 RnpA under nonpermissive glucose conditions (Table 2 and Fig. S7). This finding argues against RnpA-specific inhibition of RNPA2000, but rather points to unspecific or toxic effects.

TABLE 2.

Disk diffusion test results^a

Strain	Carbon source	Diameter of inhibition zone (mm) forc:
		Amp	0.1 mM RNPA2000	0.25 mM RNPA2000
YB886	Xylose	15/16/16	7/8/8	9/9/9/10
	Glucose	15/15/15	7/8/7	8/9/9/10
d7b	Xylose	18/15/17	8/8/9	10/10/10
	Glucose	ND	ND	ND
d7 RnpA	Xylose	18/20/20	9/9/9	10/11/13
	Glucose	19/18/21	8/9/9	10/13/12
d7 vectorb	Xylose	21/21/24	9/9/9	12/12/13
	Glucose	ND	ND	ND
d7 PRORP3	Xylose	22/25/22	9/9/9/10	13/13/20/12
	Glucose	25/19/21	10/9/10/9	16/12/13/13

^aSee Fig. S7.

^bFor the d7 strain and the d7 strain transformed with the empty pDG148 vector (d7 vector), no bacterial growth was observed in the presence of glucose (nonpermissive condition); thus, no zone of inhibition could be determined. ND, nondeterminable.

^cThe measured diameters (in mm) for 3 or 4 biological replicates each are given (lower limit of measurement precision, 1 mm). The filters themselves had a diameter of 6 mm. Ampicillin (Amp; 10 μg) loaded onto the center of the filter was used as a control.

Nonspecific RNase activity of S. aureus RnpA.

Beyond inhibiting RNase P-catalyzed tRNA 5′-end maturation, RNPA2000 was also described to inhibit RnpA-mediated, nonspecific RNA degradation, which was considered a dual-function antimicrobial activity of RNPA2000 (20). Dunman and coworkers further reported evidence to suggest that RnpA is a component of the S. aureus mRNA degradosome and observed degradation of two staphylococcal model RNA substrates, spa mRNA and rRNA, in the presence of recombinant S. aureus RnpA in vitro (29). To reproduce this nonspecific RNase activity of S. aureus RnpA, we recombinantly expressed N-terminally tagged S. aureus RnpA by two chromatographic steps, namely Ni-nitrilotriacetic acid (NTA) chromatography under denaturing conditions followed by cation exchange chromatography (see the supplemental material). In contrast, in the previous study, N-terminally tagged S. aureus RnpA was purified in a single Ni-NTA affinity chromatography step under nondenaturing conditions (29). It is of note that we were not able to purify recombinant S. aureus RnpA with satisfactory purity by Ni-NTA chromatography under native conditions (see Fig. S8 in the supplemental material). We then incubated our recombinant S. aureus RnpA protein with the pre-tRNA^Gly substrate that carries single-stranded extensions (14 nt at the 5′ end and 6 nt at the 3′ end) that mimic unstructured RNAs, while the tRNA moiety is highly structured. After 1 h of incubation at 37°C in the same manner as carried out before (29), no substantial RnpA-mediated degradation of pre-tRNA^Gly was observed; bovine serum albumin (BSA) served as negative control and RNase T1 as positive control (Fig. 5).

FIG 5.

Assay for the detection of S. aureus RnpA-mediated RNA degradation. Incubation of 5′-[³²P]-end-labeled T. thermophilus pre-tRNA^Gly (<5 nM) with 50 pmol S. aureus RnpA for 60 min. Buffer (storage buffer of S. aureus RnpA; 50 mM Tris-HCl [pH 7.0], 100 mM NaCl, and 10% glycerin) only and bovine serum albumin (BSA) were used as negative controls and incubation with 1 U RNase T1 was used as a positive control. Samples were analyzed by 8% denaturing PAGE, and radioactive bands were visualized by phosphorimaging.

Iriginol hexaacetate.

At 1 nM B. subtilis RNase P, the presence of 10 μM iriginol hexaacetate (Ir6Ac) showed a reduction in the rate constant of pre-tRNA^Gly cleavage and also in the maximum fraction of cleavable substrate (Fig. 6A). Yet, the inhibitory effect was rather mild, considering that almost complete inhibition was reported at 10 μM Ir6Ac in the previous study (18). There might be several reasons for this discrepancy, such as different enzyme concentrations (1 nM P RNA and 1 nM RnpA versus 0.1 nM P RNA and 2 nM RnpA [18]) or the different substrate (concentration) used.

FIG 6.

Iriginol hexaacetate (Ir6Ac) testing. (A to D) Single exponential fittings (see the supplemental material) for pre-tRNA^Gly processing by the B. subtilis holoenzyme or P RNA alone in the presence of 10 μM Ir6Ac (red curves). DMSO (1%) was used as a control (black dashed curves). Experiments were performed in KIN buffer with 1% DMSO and 5 mM (A to C) or 100 mM (D) MgCl₂. Concentrations of P RNA, RnpA, and pre-tRNA^Gly substrate were varied. All data are based on at least three biological replicates each. (A) Holoenzyme (1 nM; equal amounts of P RNA and RnpA protein) and 10 nM pre-tRNA^Gly substrate; (B) 1 nM P RNA, 10 nM RnpA, and 10 nM pre-tRNA^Gly substrate; (C) 10 nM holoenzyme (equal amounts of P RNA and RnpA) and 100 nM pre-tRNA^Gly substrate. (D) RNA-alone activity using 1 nM P RNA and 10 nM substrate and assayed at 100 mM Mg²⁺.

As high-throughput screenings are prone to the identification of aggregators, we consulted the Aggregator Advisor tool (http://advisor.bkslab.org/; provided by the Shoichet Laboratory, UCSF, San Francisco, CA) (24). Aggregation Advisor predicted a high structural similarity of Ir6Ac to an aggregator previously reported in the literature (see Table S1 in the supplemental material). In addition, compounds sharing the isoflavone skeleton are known to act as aggregators (32, 44). To validate whether this is also the case for Ir6Ac, we varied the enzyme concentration. Increasing the RnpA concentration from 1 to 10 nM at 1 nM or 10 nM P RNA completely abolished inhibition (Fig. 6B and C), indicating concentration sensitivity typical for aggregators. Although an interaction of Ir6Ac with nucleic acids has been discussed previously (18), the inhibition seems to be protein dependent because there was no inhibition seen for the P RNA-alone reaction at 10 μM Ir6Ac and 100 mM Mg²⁺ (Fig. 6D). Taken together, our results provide evidence that the inhibitory effect of Ir6Ac is primarily due to aggregation.

Purpurin.

Using a real-time fluorescence-based assay to monitor RNase P activity, Torres-Larios and coworkers identified four potential RNase P inhibitors in a screen of a library with ∼2,600 compounds, the most promising candidate being purpurin (see Fig. S9A in the supplemental material). Processing of a minihelix substrate by T. maritima RNase P was reported to be inhibited by purpurin with an IC₅₀ value of ∼100 μM at 100 mM MgCl₂ (the minimal substrate was not cleavable at 10 mM Mg²⁺) and in the presence of 10% DMSO. Analysis of cleavage of a pre-tRNA-like substrate in the presence of 10 mM MgCl₂ and 10% polyethylene glycol 200 (PEG200) instead of DMSO for solubilizing purpurin gave an IC₅₀ of 13 μM (22). Execution of a gel-based cleavage assay with a nonfluorescent minihelix substrate and canonical pre-tRNA at 500 μM purpurin excluded an interaction with the fluorescent dye. After confirming purpurin binding to the RnpA protein via biolayer interferometry measurement, a crystal structure of T. maritima in complex with purpurin was obtained. The structure suggested binding of purpurin to a hydrophobic patch where the compound interferes with binding of the pre-tRNA 5′ leader to RnpA (22). When we analyzed purpurin, we noticed solubility problems. At a concentration of 200 μM, precipitation was observed under the microscope in two buffer systems previously used to characterize the interaction with RNase P (22) (Fig. S9B). We also performed a co-pulldown experiment with 100 μM purpurin (a concentration that resembled the IC₅₀ for cleavage of the aforementioned minihelix substrate) in buffer H containing 10% DMSO (Fig. S9C). In the presence (lane 8) and absence (lane 10) of Mg²⁺ ions, T. maritima RnpA was detected in the pellet, indicating protein aggregation.

DISCUSSION

In the present study, three published RNase P inhibitors were identified as protein aggregators. Protein aggregation was described as one mechanism leading to promiscuous inhibition. Much effort has been made in the field of medicinal chemistry to identify such compounds and to rid the literature of such inhibitors by either automated filters or experimental procedures. For two of the three compounds analyzed here, similar compounds were already described in the literature as promiscuous inhibitors. Compounds with the same isoflavone core as Ir6Ac, such as genistein, were reported to act as protein aggregators (32, 44). Not only isoflavones but also flavonoids in general have been categorized as false positives, with quercetin being the most promiscuous, with over 50 reported targets studied in hundreds of publications (32, 36, 45), and for which protein aggregation was identified as the cause of promiscuity (27). Besides such polyhydroxylated natural phytochemicals, quinones resemble one of the most problematic and readily identifiable PAINS (23, 46, 47). Mechanistically, quinones are highly reactive electrophiles that react with protein thiol groups, and hence are termed covalent modifiers (46, 48, 49). Purpurin, an anthraquinone derivative with hydroxyl groups in 1-, 2- and 4-position, might be involved in metal ion complexation, as the two carbonyl functions are in perfect distance to hydroxyl groups for such interactions. However, in our study, general solubility problems of purpurin seemed to be the cause of protein aggregation (see Fig. S9). Nonetheless, purpurin could be a border case, as the compound was soaked into crystals of T. maritima RnpA, where it localized to the 5′ leader binding cavity (22). Thus, purpurin might give rise to RnpA-specific inhibitors, provided that derivatives with improved aqueous solubility and that retain binding specificity could be developed. The third compound analyzed here, RNPA2000, has a thiosemicarbazide core that, to our knowledge, has not previously been linked with protein aggregation. For RNPA2000, we could show that protein aggregation is induced by metal ion complexation (Fig. 2), and we provided evidence that the reported antibacterial effects are not RnpA-specific but likely are caused by aggregation of multiple proteins. Cell-based evidence (Table 2) also contradicts the claim that RNPA2000 and related compounds specifically target bacterial RnpA.

Promiscuous inhibitors, although they interfere with in vitro enzyme-based assays, may well produce a desired cellular readout in initial experiments (50), thus looking quite promising at first sight. One example is tetrahydroquinolines that are described as PAINS in the literature (51). Iodinated tetrahydro-3H-cyclopenta[c]quinolines exhibited high affinity and selectivity toward the G protein-coupled estrogen receptor GPR30 but not toward the classical estrogen receptors ERα and ERβ, which are nuclear receptor transcription factors. Nevertheless, during in vivo experiments, issues concerning pharmacokinetics were observed, such as rapid metabolism, problematic plasma protein binding, and poor targeting characteristics (50). Notably, 3 to 6% of the drugs on the market harbor such PAINS structural elements, causing artifactual inhibition and non-drug-like behavior (51). Clotrimazole, for instance, a well-known azole antifungal that was not discovered by a target-based in vitro assay, was later identified as a protein aggregator. The drug inhibited three different enzymes (β-lactamase, chymotrypsin, and malate dehydrogenase), was sensitive to detergent, and formed particles detectable by light scattering (52). Protein aggregation by clotrimazole was observed long after drug approval when the compound was tested against the aforementioned enzymes at concentrations (10 to 400 μM) largely above the IC₅₀ value of 0.1 μM against the clinical target, Candida albicans 14α-demethylase (53). This indicates that inhibition by protein aggregation in the micromolar concentration range does not exclude specific target inhibition by clotrimazole in the nanomolar range. Likewise, other approved drugs may also induce nonspecific protein aggregation under certain in vitro conditions, which are unrelated to their specific medical indication. Nonetheless, this is a balancing act, and the approval of clotrimazole is restricted to topical applications because of pharmacokinetic problems. Conversely, it is unlikely that a compound hit discovered by high-throughput screening that is active through aggregation may be a specific ligand for the investigated target (52). This is illustrated by the experience that none of the compounds with PAINS structural elements that have been identified in target-based high-throughput campaigns has ever become an approved drug (33). Moreover, such structural elements in already approved drugs are associated with pharmacokinetic and/or toxicological problems (46). Hence, the presence of PAINS structural elements is considered an exclusion criterion from further development, and PAINS should be identified as early as possible during drug discovery to save time and costs.

Our study is the first that identifies reported inhibitors of bacterial RNase P as protein aggregators, demonstrating that the bacterial RNase P protein subunit is highly susceptible to aggregation. RNase P protein subunits (termed RnpA or P proteins) adopt an αβββαβα fold that is also found in other RNA-binding proteins (42). RnpA proteins have pI values of around 10, owing to many positively charged amino acid side chains involved in binding to the RNase P RNA subunit and to the 5′ leader sequence of pre-tRNA substrates. A central cleft formed by the N-terminal α-helix and the central β-sheet binds the 5′ leader, and α-helix 2, which contains the RNR motif (representing the region of highest amino acid conservation among RnpA proteins), is the primary interaction site with the catalytic RNA subunit. Investigation of the biophysical properties of recombinantly expressed and purified B. subtilis RnpA (25) revealed that the protein behaves as an “intrinsically unstructured” or “natively unfolded” protein at low ionic strength (10 mM sodium cacodylate, pH 7). Based on circular dichroism (CD) measurements, the protein was further inferred to possess features of a “molten globule” species, being hydrophobically collapsed and consistent with the formation of an ensemble of conformations with residual extents of β-sheet formation. Addition of anions induces formation of the native fold of the protein in a one-step conformational transition. Analysis of pyrophosphate binding to B. subtilis RnpA indicated the binding of two anions per protein molecule. Based on X-ray and nuclear magnetic resonance (NMR) data, one of the two anion binding sites is located between the terminus and α-helix 2. Pyrophosphate or (ribo)nucleoside triphosphates showed high affinity for RnpA, while phosphate, sulfate, or nucleoside monophosphates bound with intermediate affinity (dissociation constant [K_d] values of 10 to 150 μM). The osmolyte trimethylamine N-oxide (TMAO) was also able to increase the fraction of RnpA in the native state, but in this case the primary reason for protein stabilization by TMAO is thought to be an unfavorable interaction with the peptide backbone that increases the energy of the denatured state (54). TMAO also causes contraction of denatured ensembles (55), thereby lowering the conformational entropy of folding by bringing aliphatic side chains in closer vicinity to each other to favor hydrophobic collapses of proteins. As TMAO did not compete with anions for binding to RnpA, previous work concluded that TMAO favors population of a metastable native state, while occupation of the two anion binding sites results in a more stable native fold of the protein (25). Biologically, the “intrinsically unstructured” state of RnpA may increase the specificity of its interaction with the cognate RNase P RNA subunit. On the other hand, the protein’s capacity to fold in the presence of anions, polyphosphates, or nucleotides may mitigate its degradation by cellular proteases when being unfolded. Overall, the findings demonstrate that the folding state of RnpA proteins is highly sensitive to interaction with small molecules, consistent with the aggregation effects observed in our investigation. This malleability of the protein’s folding state is expected to aggravate the prospects of finding specific inhibitors of RnpA, particularly when the protein is present in its free form.

Besides being part of the holoenzyme responsible for 5′ processing of pre-tRNAs, the RnpA of S. aureus was described to mediate nonspecific RNA degradation (20), a finding we were unable to confirm. We purified recombinantly expressed RnpA from S. aureus and tested it for RNase activity in the absence of the RNase P RNA subunit. No degradation of a pre-tRNA substrate was observed after 1 h of incubation with S. aureus RnpA (Fig. 5). In contrast, a previous study reported vigorous degradation of rRNA and an mRNA substrate by the same protein (29). One reason for this discrepancy could be the different procedure of protein purification. As we were not able to purify the RnpA protein without nucleic acid contamination by the procedure described in (29), we instead performed the first purification step under denaturing conditions to separate the protein from associated nucleic acids, followed by a second purification step under native conditions. Nucleic acid contamination after native affinity chromatography has been described as the reason for aggregation of T. maritima and B. subtilis RnpA (56, 57). Nucleic acid contamination may also entail copurification of traces of RNases, which could explain the observed RNA degradation activity of recombinant S. aureus RnpA, although Dunman and coworkers ruled this out based on matrix-assisted laser desorption ionization (MALDI) analysis (29). In general, the role of S. aureus RnpA in cellular RNA degradation is still under debate. A bacterial two-hybrid study identified the DEAD box RNA helicase CshA as an interaction partner of RnpA within the degradosome (58), but this interaction was not observed in a tandem affinity purification approach (59). In vitro pulldown assays then detected an interaction of S. aureus RnpA with RNase J1/J2, enolase, and PNPase instead of binding to CshA (60).

MATERIALS AND METHODS

Cloning and expression of S. aureus rnpA and rnpB genes.

Cloning of S. aureus rnpA and rnpB genes, T7 in vitro transcription, and preparation of RNase P RNA were performed essentially as described previously (61, 62). For further details, see the supplemental material. Dephosphorylation and 5′-[³²P]-end labeling of pre-tRNA^Gly was performed as described previously (63, 64). Protocols for the recombinant expression and purification of RnpA proteins from S. aureus, B. subtilis, and T. maritima are detailed in the supplemental material.

RNase P kinetics.

RNase P processing assays were performed as described previously (61, 65). For details, see the supplemental material.

Co-pulldown experiments.

For co-pulldown experiments, we prepared 1-ml solutions containing 5 μg T. maritima RnpA and 140 μM RNPA2000 in kinetic (KIN) buffer also used for the kinetic experiments (20 mM HEPES-KOH [pH 7.4], 150 mM NH₄CH₃COO, 2 mM spermidine, 50 μM spermine, and 4 mM β-mercaptoethanol) supplemented with 10 mM MgCl₂ and 2% DMSO. A sample containing 5 μg T. maritima RnpA and 140 μM RNPA2000 but without MgCl₂ was prepared to control for Mg²⁺-dependent effects. After incubation at room temperature (RT) for 5 min, samples were centrifuged for 30 min at 16,200 × g and 20°C. The supernatant was withdrawn and the pellet (only visible in the sample with Mg²⁺, RnpA, and RNPA2000) was resuspended in 15 μl TuS buffer (50 mM Tris-HCl [pH 7.4], 100 mM NaCl, and 8 M urea) to which 1 μl 1% Triton X-100 was added. Then, 10 μl of the supernatant and the entire redissolved pellet were loaded onto a 15% SDS gel. After SDS-PAGE, protein bands were detected by staining with Coomassie brilliant blue R-250. RnpA in the supernatant was too dilute (∼50 ng/10 μl supernatant) to be visible after Coomassie staining.

HPLC solubility determination.

Solubility of inhibitory compounds was determined by HPLC. A Hitachi (Tokyo, Japan) Primaide system was used, consisting of a 1110 pump, a 1210 auto sampler, a 1310 column oven, and a 1430 diode array detector. The stationary phase was a Machery Nagel 250/4.6 Nucleodur 100-5 C₁₈ ec (endcapping) column with a CC 8/4 Nucleodur 100-5 C₁₈ ec precolumn. A mixture of acetonitrile and ultrapure water with 0.1% trifluoroacetic acid (TFA) was used as the mobile phase. The flow rate was adjusted to 1 ml/min, and the acetonitrile content was increased by 1% per min. The injection volume was 99 μl each, and the chromatographic separation was performed at 30°C. Detection profiles were recorded at 220 nm and evaluated using EZ Chrom Elite (version 3.3.2 SP2).

Samples of 200 μl were prepared. Before addition of the compound, buffer, DMSO, MgCl₂, and H₂O were vigorously mixed. Three different buffers were used, as follows: (i) KIN buffer (see above), (ii) Tris buffer (50 mM Tris-HCl [pH 8.0]), and (iii) phosphate-buffered saline (PBS [pH 7.4]). After addition of the compound and mixing, samples were incubated for 20 min at RT and applied to the auto sampler.

For each individual experiment, samples with at least 6 increasing concentrations of the test compound were consecutively injected into the HPLC system (with column washing between injections). For each run, the area under the curve (AUC) at the peak of the 220-nm absorption profile was determined. AUC values were plotted against the compound concentration, resulting in an ascending calibration line that descends at concentrations beyond the point of maximum solubility. The inflection point of this type of curve is taken as the solubility limit (see examples in Fig. S4). At least three independent experiments (as described above), starting from independent stock solutions of the test compound, were conducted and used to assess the solubility limit.

Testing for nonspecific RNase activity of S. aureus RnpA.

Trace amounts of 5′-[³²P]-end-labeled pre-tRNA^Gly (<5 nM) were preincubated for 5 min at 55°C in 1× reaction buffer (2 mM NaCl, 2 mM MgCl₂, and 50 mM Tris-HCl [pH 6.0]), followed by addition of 2.5 μM S. aureus RnpA in protein storage buffer (50 mM Tris-HCl [pH 7.0], 100 mM NaCl, and 10% glycerol), or by adding the same volume of storage buffer without protein (control), or by adding 2.5 μM BSA (UltraPure BSA; Thermo Fisher Scientific), or by adding 1 U RNase T1 (Thermo Fisher Scientific) as positive control. Samples were incubated for 60 min at 37°C in 1× reaction buffer (total volume, 20 μl); for RNase T1, a 4-μl aliquot was already withdrawn after 2 min. Reaction conditions were similar to those applied previously to demonstrate nonspecific RNase activity of S. aureus RnpA (29). After incubation, 4-μl aliquots of the reactions were analyzed by 8% denaturing PAGE followed by phosphorimaging.

Complementation experiments with B. subtilis strain d7.

In the B. subtilis d7 strain, endogenous RnpA expression is under the control of a xylose-inducible promoter (4). B. subtilis cells were streaked from frozen glycerol stocks onto LB agar plates (supplemented with 2% xylose, 20 μg/ml kanamycin, and 5 μg/ml chloramphenicol) that were incubated overnight at 37°C. A single colony from such a plate was used to inoculate a 3-ml liquid culture that was grown overnight at 37°C under shaking. On the next day, 40 ml prewarmed LB medium, again supplemented with 2% xylose, 20 μg/ml kanamycin, and 5 μg/ml chloramphenicol, was inoculated with overnight culture to a final optical density at 600 nm (OD₆₀₀) of 0.1; the culture was grown at 37°C (with shaking) to the mid-exponential phase (OD₆₀₀ of 0.6 to 0.8). Cells were harvested (5 min at 2000 × g and 22°C) and washed twice with LB medium (resuspension of cells in 40 ml LB followed by centrifugation at 2000 × g and 22°C for 5 min) to remove xylose and finally redissolved in 40 ml LB medium. After dilution of the cell suspension with LB medium in three steps to a final ratio of 1:25,000, 100-μl aliquots were plated on agar plates supplemented with 20 μg/ml kanamycin, 5 μg/ml chloramphenicol, and either 2% xylose or 2% glucose, followed by incubation at 37°C for 48 h.

Disk diffusion tests.

For disk diffusion tests, cells were cultivated as described above. Except for strain YB886, cultures were supplemented with the following antibiotics: 5 μg/ml chloramphenicol for strain d7 and 5 μg/ml chloramphenicol and 20 μg/ml kanamycin for strain d7 harboring the plasmid vector pDG148. After harvest and washing, cells were resuspended in the original culture volume (40 ml), then diluted 1:10 with LB medium, and 100-μl aliquots of the dilution were plated on agar plates supplemented with the antibiotics specified above and either 2% xylose or 2% glucose; homogeneous plating was achieved by the use of glass beads. For evaluation of antibacterial effects, four filter papers (each with a diameter of 6 mm) were positioned in the plate quarters; after 5 min at RT, 10 μl of either DMSO, ampicillin, or RNPA2000 was cautiously pipetted onto the filter paper. Plates were incubated for 15 min at RT for prediffusion of the compounds, followed by incubation for 19 to 21 h at 37°C and measurement of the inhibition zone diameter with a ruler.

Compound synthesis.

RNPA2000 and its derivatives NL48 and NL20 were synthesized in a three-step procedure. For details, see the supplemental material. Iriginol hexaacetate was obtained from Cfm Oskar Tropitzsch GmbH, and neomycin sulfate and purpurin were from Merck Sigma-Aldrich.