An NF-κB-Based High-Throughput Screen Identifies Piericidins as Inhibitors of the Yersinia pseudotuberculosis Type III Secretion System

The bacterial strains used in this paper are listed in Table 1. Y. pseudotuberculosis was grown in 2× yeast extract-tryptone (YT) medium at 26°C with shaking overnight. The cultures were back diluted into low-calcium medium (2× YT plus 20 mM sodium oxalate and 20 mM MgCl₂) to an optical density at 600 nm (OD₆₀₀) of 0.2 and grown for 1.5 h at 26°C with shaking, followed by 1.5 h at 37°C to induce Yop synthesis, as previously described (13).

HEK293T cells were maintained in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum (FBS) and 2 mM l-glutamine at 37°C in 5% CO₂. CHO-K1 cells were maintained in Ham's F-12 nutrient mixture with Kaighn's modification (F-12K) with 10% FBS and 2 mM l-glutamine at 37°C in 5% CO₂.

A screening campaign for T3SS inhibitors was carried out using a marine-natural-products library. This library was generated from environmental-sediment-derived marine microorganisms specifically from the class Actinomycetales, known for their prolific production of pharmacologically interesting secondary metabolites. Sediment samples were collected into sterile 15-ml Falcon tubes by scuba diving, mostly from the West Coast of the United States. The supernatant was removed, and sediment samples were plated onto Actinobacterium-specific isolation medium with added antifungal and Gram-negative antibacterial agents by radial stamping with sterile cotton swabs. Morphologically distinct colonies were picked and replated on Difco marine broth agar plates repeatedly until pure isolates were obtained. Isolated actinobacterial colonies were subjected to liquid medium culturing using our standard fermentation conditions (see below) and cryopreserved as glycerol stock solutions at −80°C.

Frozen stocks of environmental isolates were streaked onto fresh marine broth plates (Difco, USA) and incubated at 25°C until discrete colonies became visible. Selected colonies were inoculated into 7 ml of modified saline SYP (mSYP) medium (10 g starch, 4 g peptone, 2 g yeast extract, and 31.2 g Instant Ocean in 1 liter of distilled water), and the cultures were stepped up in stages by first inoculating 2.5 ml of 3-day-old 7-ml cell cultures into 60 ml of mSYP (medium scale), followed by inoculation of 40 ml of these 2-day-old medium-scale cell cultures into 1 liter of the same broth (large scale). All cultures were incubated at 26°C and shaken at 200 rpm.

Large-scale cultures were fermented for 7 days prior to chemical extraction. Twenty grams of prewashed Amberlite XAD-16 resin (CH₂Cl₂, methanol [MeOH], and water) was added to each large-scale culture and shaken for 2 h (200 rpm), and the resulting slurry was filtered under vacuum through a glass microfiber filter (Whatman). The cells, resins, and filter paper were extracted with 1:1 CH₂Cl₂-MeOH (250 ml), and the suspension was shaken at 200 rpm for 1 h. Organic extracts were filtered and concentrated to dryness in vacuo. The dried crude extracts were prefractionated by solid-phase extraction chromatography (5-g C₁₈ cartridge; Supelco, USA) using a stepwise MeOH-H₂O gradient: 40 ml of 10%, 20% (fraction A), 40% (fraction B), 60% (fraction C), 80% (fraction D), and 100% (fraction E) MeOH and then 100% ethyl acetate (EtOAc) (fraction F). Fractions A to F were concentrated to dryness in vacuo and then resuspended in dimethyl sulfoxide (DMSO) (1 ml), and aliquots of these DMSO stock solutions were reformatted into 384-well plates prior to screening.

From the primary screening of crude prefractions, active hits were selected for peak library screening to identify the active constituent(s) within a particular crude prefraction. A 45-μl aliquot of prefraction DMSO stock was lyophilized and fractionated by C₁₈ reversed-phase high-performance liquid chromatography (RP-HPLC) (Phenomenex Synergi Fusion-RP; 10- by 250-mm column; 2-ml min⁻¹ flow rate) using an MeOH-H₂O (0.02% formic acid) solvent system. Each prefraction was run on a gradient specifically tailored to produce the most highly resolved chromatography. The eluent was collected into deep-well 96-well plates using an automated time-based fraction collection method consisting of 1-min time slices and subsequently concentrated to dryness in vacuo. The dried plates were resolubilized (10 μl DMSO per well), sonicated to ensure homogeneity, reformatted into 384-well format, and subjected to secondary screening.

On day 1, 3.75 × 10⁶ HEK293T cells were plated onto three 100- by 20-mm tissue culture dishes (BD Falcon) and incubated at 37°C and 5% CO₂. On day 2, the HEK293T cells were transfected with an NF-κB luciferase reporter plasmid (Stratagene) using Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions. This plasmid contains an NF-κB binding site (five repeats of GGAAAGTCCCCAGC) upstream of the luciferase gene. On day 3, the transfected cells were pooled, and 5 × 10⁴ cells were transferred into each well of a 384-well plate. Each plate was centrifuged for 5 min at 290 × g. Y. pseudotuberculosis overnight cultures were back diluted into 2× YT medium to an OD₆₀₀ of 0.2 and grown in a shaking incubator at 26°C for 1.5 h. The cultures were then pelleted by centrifugation and resuspended in half the original volume of low-calcium 2× YT medium. The bacteria were transferred to a 384-well plate containing low-calcium medium plus prefractions from the natural-product library or plain DMSO and incubated for 1.5 h at 37°C. Immediately prior to infection, prefractions or DMSO vehicle control was added to the 384-well plate containing HEK293T cells by a pinning robot. A Janus MDT pinning robot (PerkinElmer) was next used to transfer Y. pseudotuberculosis from the bacterial 384-well plate to the HEK293T plate at a multiplicity of infection (MOI) of 7. After 4 h at 37°C and 5% CO₂, the medium was aspirated before adding a 1:1 Neolite–phosphate-buffered saline (PBS) solution. Plates were covered in foil and incubated for 5 min, and bioluminescence was measured using an EnVision plate reader (PerkinElmer).

We used Y. pseudotuberculosis lacking the six known T3SS effector proteins, YopHEMOJT (Δyop6), for this screen because YopHEMOJT are not required for T3SS-dependent NF-κB induction, and instead, several Yop proteins modulate NF-κB activation (8). A Y. pseudotuberculosis mutant lacking the T3SS translocon component YopB (Δyop6 ΔyopB) was used as a T3SS-negative control. Y. pseudotuberculosis Δyop6 in the absence of natural products triggered, on average, 12-fold-greater luminescence than the Y. pseudotuberculosis Δyop6 ΔyopB mutant (data not shown).

Aurodox was purchased from Enzo Life Sciences, MBX1641 from ChemBridge, C22 from TimTec, and C15 and C24 from Princeton Biomolecular Research.

Piericidin A1 and its analog Mer-A 2026B were isolated from an Actinomycetes strain, RL09-253-HVS-A, isolated from a marine sediment sample collected at Point Estero, CA, at a depth of 50 feet by scuba diving. This Actinomycetes strain was identified as a Streptomyces sp. based on the typical morphology of streptomycetes, which form dry powdery spores on agar plates, and by sequencing the 16S rRNA gene (data not shown). From a large-scale culture (4 liters) of the strain producing extract 1772, 0.33 g of prefraction D was obtained (using the prefractionation method described above). The active constituents were purified using C₁₈ RP-HPLC (with a gradient of 58% to 88% MeOH–0.02% formic acid-H₂O; 2 ml/min [Synergi 10μ Fusion-RP column; Phenomenex, USA]) (t_R [retention time] = 12.5 min for Mer-A 2026B and 30.5 min for piericidin A1) to give 1.6 mg of Mer-A 2026B and 2.7 mg of piericidin A1. Electrospray ionization-time of flight high-resolution mass spectrometry (ESI-TOF HRMS) analysis predicted the molecular formulae C₂₄H₃₅NO₃ and C₂₅H₃₇NO₄ for Mer-A 2026B and piericidin A1, respectively. One-dimensional ¹H nuclear magnetic resonance (NMR) spectra (see Fig. S2 and S3 in the supplemental material) were obtained using a Varian Unity Inova spectrometer at 600 MHz equipped with a 5-mm HCN triple-resonance cryoprobe. The spectra were referenced to residual solvent signals (δ_H [proton chemical shift] of 7.24 for d-CDCl₃ [deuterated chloroform]).

HeLa cell staining was performed as previously described (14). Briefly, HeLa cells were incubated with microbial extracts for 19 h and stained with Hoechst stain to visualize individual nuclei. The 10% of natural-product fractions that most reduced HeLa nuclear counts were classified as cytotoxic to mammalian cells and excluded from follow-up. This top 10% of nucleus reduction correlated strongly with the effects of previously characterized cytotoxic compounds within the training set used by Schulze et al. (14). For unpurified natural-product fractions (including 1772D), the mammalian cytotoxicity data were generated by Schulze et al. (14). The cytotoxicity data for purified piericidin A1 and Mer-A 2026B at ≤250 μM were obtained specifically for this study.

Overnight cultures of wild-type (WT) Y. pseudotuberculosis IP2666 were back diluted to an OD₆₀₀ of 0.2, and 100 μl was added to each well of a 96-well plate. A total of 0.3 μl of DMSO or purified natural products was added to each well. Bacteria were grown at 23°C in 2× YT medium or at 37°C in low-calcium medium (T3SS-inducing conditions), and the OD₆₀₀ of the culture was measured every 15 min for 5 or 6 h using a VersaMax Tunable Microplate Reader (Molecular Devices). The 96-well plates were continuously shaken throughout the experiment. Additional growth curves were carried out by taking samples of bacterial cultures at 0, 3, 6, and 24 h of growth; serially diluting them; and plating them for CFU (see Fig. S5 in the supplemental material). Starting inocula of 4 × 10³ or 1 × 10⁶ CFU/ml were used. The 26°C growth curves were carried out in 500 μl of 2× YT medium with continuous shaking, while the 37°C growth curves were performed in high-calcium medium (2× YT plus 5 mM CaCl₂) to prevent induction of the T3SS and associated growth restriction. All growth curves used DMSO at 0.3%, piericidins at ≤143 μM, or kanamycin at 50 μg/ml.

A total of 6 × 10³ CHO-K1 cells were plated in each well of a 384-well plate in 70 μl of F-12K medium plus 10% FBS and incubated overnight. The following day, Y. pseudotuberculosis YopM–β-lactamase (YopM-Bla) reporter strain overnight cultures were back diluted into low-calcium 2× YT medium to an OD₆₀₀ of 0.2 and grown in a shaking incubator at 26°C for 1.5 h. The cultures were then transferred to a 384-well plate containing low-calcium medium and purified compounds or DMSO and incubated for 1.5 h at 37°C. Immediately prior to infection, the purified compounds or DMSO was added to the 384-well plate containing CHO-K1 cells by a pinning robot (Janus MDT; PerkinElmer). The pinning robot was next used to transfer Y. pseudotuberculosis from the bacterial 384-well plate to the CHO-K1 plate at an MOI of 6. Five minutes after this transfer, the plate was centrifuged at 290 × g for 5 min to initiate bacterium-host cell contact and incubated for 1 h at 37°C and 5% CO₂. Thirty minutes prior to the end of the infection, CCF2-AM (Invitrogen) was added to each well, and the plate was covered in foil and incubated at room temperature. At the end of the infection, the medium was aspirated, and 4% paraformaldehyde was added to each well for 20 min to fix the cells. The paraformaldehyde was then aspirated, and the DNA dye DRAQ5 (Cell Signaling Technology) in PBS was added to each well. The monolayers were incubated at room temperature for 10 min, washed once with PBS, and visualized using an ImageXpress^MICRO automated microscope and MetaXpress analysis software (Molecular Devices). The number of YopM-Bla-positive cells was calculated by dividing the number of blue (CCF2-cleaved) cells by the number of green (total CCF2⁺) cells. Data from three separate wells were averaged for each experiment.

Visualization of T3SS cargo secreted in broth culture was performed as previously described (13). Y. pseudotuberculosis low-calcium medium cultures were grown for 1.5 h at 26°C. Purified compounds or DMSO was added, and the cultures were switched to 37°C for another 2 h. The cultures were spun down at 13,200 rpm for 10 min at room temperature, and the supernatants were transferred to a new Eppendorf tube. Ten percent (final concentration) trichloroacetic acid was added, and the mixture was vortexed vigorously. Samples were incubated on ice for 20 min and then spun down at 13,200 rpm for 15 min at 4°C. The pellet was resuspended in final sample buffer (FSB) plus 20% dithiothreitol (DTT). Samples were boiled for 5 min prior to running on a 12.5% SDS-PAGE gel. Sample loading was normalized for the bacterial density (OD₆₀₀) of each sample. Densitometric quantification of the bands was done using Image Lab software (Bio-Rad), setting the first DMSO-treated WT Y. pseudotuberculosis YopE band to 1.00.

To carry out our HTS (Fig. 1), we transiently transfected HEK293T cells with a plasmid carrying an NF-κB-inducible luciferase reporter gene and infected the cells with Y. pseudotuberculosis carrying a functional T3SS in the presence of marine-derived natural products or DMSO vehicle control. The bioluminescence intensity produced by the cultures was used as a readout of T3SS activity. The in-house natural-products library used consisted of over 5,000 partially purified “prefractions” containing 2 to 20 actinobacterium-derived small molecules per prefraction (14, 15). This library has been shown to contain compounds with specific antimicrobial activities (15) but has not previously been screened for T3SS inhibitors.

We performed the screen described above on 2,560 prefractions in duplicate. We identified 355 prefractions that reduced NF-κB-driven luminescence by 2 standard deviations below the mean of the DMSO-treated control (see Dataset S1 in the supplemental material). This pool of prefractions may contain T3SS inhibitors but may also contain compounds toxic to eukaryotic cells, killing the HEK293T cells before robust NF-κB activation can be induced. For instance, the known cytotoxins staurosporine and gliotoxin lowered NF-κB-driven luminescence to 7- and 11-fold less than that induced by the Δyop6 ΔyopB T3SS-deficient strain, respectively (Fig. 2A). To exclude prefractions containing potently cytotoxic compounds, we eliminated 217 prefractions that reduced T3SS-driven luminescence to at least 1 standard deviation below the average luminescence produced by the Δyop6 ΔyopB-infected cells (see Dataset S1 in the supplemental material). A total of 92% of these eliminated prefractions also showed HeLa cell cytotoxicity in work by Schulze et al. (14), validating this approach.

Of the remaining 138 prefractions, we hypothesized that some may contain compounds with general antibiotic activity, blocking Y. pseudotuberculosis replication. In a separate study, Wong et al. carried out antibiotic mode-of-action profiling on the same natural-products library using a panel of bacteria. Based on their findings, we eliminated four additional prefractions that inhibited Y. pseudotuberculosis growth in broth culture (15) (see Dataset S1 in the supplemental material).

Finally, we eliminated 48 additional prefractions that caused HeLa cell cytoxicity in the study carried out by Schulze et al. (14) but did not reduce NF-κB activity in HEK293T cells to the very low levels seen with staurosporine or gliotoxin (see Dataset S1 in the supplemental material). This discrepancy in toxicity may be due to the cell types used (HeLa versus HEK293T) or the time frame of the experiments (19 h versus 4 h). However, T3SS inhibitors that are also cytotoxic toward mammalian cells would be significantly less useful as T3SS research probes and pretherapeutics. Therefore, we eliminated prefractions displaying such dual activity, leaving 86 prefractions in our final pool.

Estimating an average of 11 compounds per prefraction, and one compound per bioactive prefraction responsible for T3SS-inhibitory activity, the hit rate per compound is ∼0.3%. This hit rate is within the range of previously published T3SS inhibitor screens (16–21).

We selected 21 prefractions for further investigation by separating the small molecules within each prefraction using liquid chromatography-mass spectrometry (LC-MS) to generate “one compound-one well” peak libraries for secondary screening. This approach provides mass spectrometric, UV absorbance, and retention time data for all active constituents and permits direct identification of bioactive compounds from active fractions. We then used these peak libraries to repeat the experiment described above (Fig. 1) and identified individual constituents able to inhibit T3SS-driven NF-κB activation.

We focused on prefraction 1772D, which caused a 3.5-fold decrease in T3SS-driven NF-κB activation (Fig. 2B). In comparison to staurosporine, prefraction 1772D did not cause gross changes in HeLa cell morphology in the absence of bacteria (Fig. 2C), indicating that the compounds in the prefraction are not grossly cytotoxic to mammalian cells. Upon further separation, prefraction 1772D yielded four fractions, corresponding to 31, 32, 43, and 44 min of retention time on the HPLC, that exhibited significant inhibition of T3SS-driven NF-κB activation (Fig. 3A) and displayed tractable chromatography for compound isolation (Fig. 3B). We regrew Streptomyces sp. strain RL09-253-HVS-A, which produced prefraction 1772D, and reisolated and purified the bioactive compounds. The structures of two related compounds found in 1772D fractions 31, 32, 43, and 44 were determined using a combination of NMR and MS experiments (Fig. 3C; see Fig. S2 and S3 in the supplemental material). We identified one compound as the piericidin derivative Mer-A 2026B and the other as piericidin A1 (22–24). Piericidins have previously characterized activity as insecticides, vasodilators, and inhibitors of the mitochondrial NADH dehydrogenase and as general antibiotics against certain bacteria (22, 23, 25–27).

To confirm that the piericidins did not affect bacterial replication, we performed growth curves of Y. pseudotuberculosis in the presence of the purified compounds at 26°C and 37°C and monitored bacterial growth by optical density (Fig. 4). Piericidin-treated Y. pseudotuberculosis grew as well as or better than DMSO-treated bacteria at all tested concentrations up to 143 μM, in contrast to the known antibiotic kanamycin. We also performed more sensitive 24-h growth curves by serially diluting and plating cultures after 0, 3, 6, and 24 h of growth (see Fig. S4 in the supplemental material). We observed no difference in bacterial replication between DMSO- and 143 μM piericidin-treated Y. pseudotuberculosis at all time points. As expected, we could not recover any CFU from kanamycin-treated cultures at 3, 6, or 24 h of growth.

We evaluated the ability of Y. pseudotuberculosis to secrete effector Yop proteins into broth culture in the presence of the piericidins or five previously identified, commercially available T3SS inhibitors (see Fig. S5 in the supplemental material). MBX-1641 and aurodox were shown to reduce in vitro type III secretion by Yersinia pestis and Escherichia coli, respectively, and were chosen as positive controls. In contrast, C15, C22, and C24 were shown to inhibit translocation of effector proteins inside host cells, but not Yop secretion in vitro, and were chosen as negative controls (16). We grew Y. pseudotuberculosis in the presence of purified compounds or DMSO for 2 h at 37°C in the absence of calcium (T3SS-inducing conditions). We then precipitated the secreted proteins from the supernatant and analyzed relative protein abundance using SDS-PAGE analysis.

Mer-A 2026B at a concentration of 71 μM reduced secretion of the T3SS effector YopE by 45% (P < 0.02), while lower concentrations of inhibitor demonstrated a dose-dependent decrease in inhibition (Fig. 5). Piericidin A1 blocked type III secretion by 65% (Fig. 5B). MBX 1641 (17) and C15 (16) at a concentration of 71 μM also significantly reduced YopE secretion, although this inhibition was only 22 to 33% (P < 0.05 and P < 0.04, respectively). In our hands, C22 and C24 did not significantly reduce T3S, which was consistent with the original report (16). However, C15 did block in vitro secretion. Aurodox (21) did not significantly inhibit YopE secretion at the highest concentration used, 12.5 μM (Fig. 5B). We chose not to test aurodox at 71 μM, as Kimura et al. found concentrations above 12.5 μM cytotoxic to E. coli.

To analyze the ability of the piericidins to block translocation of Y. pseudotuberculosis T3SS effector proteins, we measured the translocation of a plasmid-encoded YopM-Bla reporter protein inside CHO cells using the fluorescent β-lactamase substrate CCF2-AM (28). In this assay, the CHO cells are loaded with the CCF2-AM dye, which normally fluoresces green. If the YopM–β-lactamase chimeric fusion is translocated into these cells, the dye is cleaved and the cells fluoresce blue, providing a quantifiable readout of T3SS-mediated translocation.

The piericidin derivative Mer-A 2026B significantly reduced YopM translocation into CHO cells at all concentrations (Fig. 6), ranging from 9 μM to 143 μM. The 71 μM concentration displayed the most robust T3SS inhibition, 75% (P < 0.005). Piericidin A1 also significantly diminished YopM–β-lactamase translocation at concentrations of 36 μM and greater (Fig. 6C). These results validate the notion that the piericidins identified through our screen inhibit T3SS effector translocation into eukaryotic cells.