RKO, SW480 and HCT116 (RRID: CVCL_0291) colon cancer cells are purchased from American Type Culture Collection (Manassas, VA) and validation on selected cell lines were determined by biosynthesis (Lewisville, TX). Cells are cultured in DMEM medium with 10% FBS at 37 °C in the presence of 5% CO2. The details of antibody using for Western blotting and ChIP assays are shown in Supplementary Table S1.1, 1-Bis(3′-indolyl)-1-(3,5-dichlorophenyl)methane (DIM-3,5-CI2) was synthesized by coupling indole and 3,5-dichlorobenzaldehyde as described (Lee et al., 2014a) and piperlongumine was purchased from MilliporeSigma (Burlington, MA).

The recombinant LBD of NR4A1 (0.5 μmol/L) in 1.0 mL of phosphate buffered saline (pH 7.4) was incubated for 3 min at 25 °C in a temperature-controlled fluorescence spectrometer (Varian Cary Eclipse). Fluorescence was measured using an excitation wavelength of 285 nm (excitation slit width = 5 nm) and an emission wavelength ranging from 300 to 420 nm (emission slit width = 5 nm). Aliquots of piperlongumine in DMSO were added, incubated at 25 °C for 3 min. The loss of fluorescence curve was measured, and K_D values were determined from the net fluorescence curve (background ligand fluorescence is subtracted). The fluorescence spectrum curve was derived using the internal filter correction and subtractions of the background fluorescence of the compound alone.

Isothermal titration calorimetry (ITC) was used to determine the ligand binding constant (K_D) to NR4A1 utilizing an Affinity ITC (TA Instruments, New Castle, DE). Briefly, the experimental setup was as follows. The ITC sample cell contained 250 μL of NR4A1 protein (ligand binding domain, LBD) at a concentration of 20 μmol/L in buffer containing 20 mmol sodium phosphate/L (pH 7.4), 5% glycerol, and 1.0% ethanol. The ligand titrant was prepared in the same buffer as above at a ligand concentration of 100 μmol/L. The initial ligand stock solution was prepared at a final concentration of 20 mmol ligand/L DMSO prior to preparation of the ligand titrant. The ligand titration into protein was performed at 25 °C with a stir rate of 125 rpm. Each ligand injection volume was 2.5 μL followed by up to 300 s to measure the total heat flow required to maintain constant temperature. A total of thirty injections were performed for each ligand/NR4A1 combination. In a separate set of injections, the same ligand was injected into buffer only (no protein) in order to determine heat flow as a result of ligand dilution into buffer. The ligand/buffer values were subtracted from the ligand/protein values prior to data analysis using the Affinity ITC manufacturer-supplied data analysis software package. The resulting data are plotted as heat flow (μJ) versus the molar ratio of injected ligand to NR4A1 in the sample cell.

Molecular modeling studies were conducted using Maestro (Schrödinger Release 2020–1, Schrödinger, LLC, New York, NY, 2020). The version of Maestro used for these studies is licensed to the Laboratory for Molecular Simulation (LMS), a Texas A&M University core user facility for molecular modeling and is associated with the Texas A&M University High Performance Research Computing (HPRC) facility (College Station, TX, 77,843). All Maestro-associated applications were accessed via the graphical user interface (GUI) VNC interactive application through the HPRC Ada OnDemand portal. The crystal structure coordinates for human orphan nuclear receptor NR4A1 ligand binding domain (LBD) (Zhan et al., 2012) were downloaded from the Protein Data Bank (https://www.rcsb.org; PDB ID 3V3Q). The human NR4A1 LBD crystal structure was prepared for ligand docking utilizing the Maestro Protein Preparation Wizard; restrained minimization of the protein structure was performed utilizing the OPLS3e force field. Each ligand (piperlongumine or DIM-3,5-Cl2) three-dimensional structure was prepared for docking utilizing the Maestro LigPrep, again using the OPLS3e force field. Maestro Glide (Friesner et al., 2004; Halgren et al., 2004; Friesner et al., 2006) was utilized with the default settings to dock each prepared ligand to the prepared protein, predict the lowest energy ligand binding orientation, and calculate the predicted binding energy in units of kcal/mol.

Cell proliferation was investigated using XTT Cell Viability Kit (Cell Signaling Biotechnology) according to the manufacturer’s instructions. Cells (1.5 × 10⁴/well) were plated in 100 μL of plating medium (as above) in 96-well plates and allowed to attach for 24 h. The medium was then changed to DMEM containing 2.5% charcoal-stripped FBS, and either vehicle (dimethyl sulfoxide (DMSO)) or designed concentrations of compounds in DMSO were added. After 24 and 48 h of culture, 35 μL of XTT reaction solution (sodium 3′-[1-(phenyl-aminocarbonyl)-3,4-tetrazolium]-bis(4-methoxy-6-nitro) benzenesulfonic acid hydrate and N-methyl dibenzopyrazine methyl sulfate; mixed in proportion 50:1) was added to the each well. The optical density was read at 450 nm wavelength in a plate reader after 4 h of incubation. All determinations were replicated in at least three separate experiments.

Cells were plated on 12-well plates at 5 × 10⁴/well in DMEM medium supplemented with 2.5% charcoal-stripped FBS. After 24 h growth, various amounts of plasmid DNA [i.e., UASx5-Luc (400 ng), GAL4-NR4A1 (250 ng) and β-gal (250 ng)] were cotransfected into each well by GeneJuice Transfection reagent (Millipore Sigma, Darmstadt, Germany) according to the manufacturer’s protocol. After 6 h of transfection, cells were treated with plating media (as indicated above) containing either solvent (DMSO) or the indicated concentration of compound (in DMSO) for 18 h. Cells were then lysed using a freeze–thaw protocol and 30 μL of cell extract was used for luciferase and β-gal assays. LumiCount (Packard, Meriden, CT) was used to quantify luciferase and β-gal activities. Luciferase activity values were normalized against corresponding β-gal activity values as well as protein concentrations determined by Bradford assay. The rescue assay used FLAG-NR4A1 expression vector to partially inhibit downregulation of TXNDC5 by piperlongumine.

Cell death was assessed using terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL). In brief, 5.0 × 10⁴ cells were seeded on 24-well plates that attached with glass coverslips and allowed 24 h to attach. After treating with the compound for 24 h, cells were fixed in 4% paraformaldehyde and then stained using the TUNEL kit (Elabscience, Wuhan, China) for 3 h at 37 °C. Cell nuclei were counterstained with DAPI for 5 min. Cells were imaged using ImageXpress Confocal HT. ai High-Content Imaging System (Molecular Devices, San Jose, CA).

Attached cells (2.0 × 10⁵) were treated with DMSO or with different concentration of piperlongumine in DMEM medium supplemented with 2.5% charcoal stripped FBS for 24 h. For Boyden chamber invasion assay, 1.0 × 10⁵ cells from each treatment condition were allowed to invade through the Boyden Chamber for 48 h. Cells that invaded into the Boyden Chamber were fixed using formaldehyde, stained, and then counted. For the scratch migration assay, cells were grown to 90% confluency in 6-well plates then scratched with a 200 μL sterile pipette tip and washed with PBS to remove detached cells from the plates. Cells were maintained in an incubator with DMSO or indicated treatments for 48 h and cells were then fixed with 4% formaldehyde and stained with crystal violate solution. The wound gap was observed under AMG EVOS fl microscope. At least 3 replicates were performed for each treatment group.

Cells (3.0 × 10⁵) were seeded on 6-well plate and after various treatments, whole cell lysates were obtained by treating them with high salt lysis buffer RIPA (Thermo Scientific, Waltham, MA) that contained protease and phosphatase inhibitors (GenDEPOT, Baker, TX). The total protein in the lysates was quantified by Bradford assay. Equal amounts of protein from each lysate were then loaded on SDS polyacrylamide gel; 35 µg of whole cell lysate were run in 12% of SDS page gels for survivin and c-caspase-3 proteins; 35 µg of whole cell lysate were run in 8% of SDS page gels for G9a, mTOR and p-mTOR proteins and 25 µg of whole cell lysate were run in 10% of SDS page gels for the remaining proteins. The proteins from the gel were transferred to a PVDF membrane, then blocked for 1 hour using 5% skimmed milk. The membranes were incubated with primary antibody for 12 h at 4 °C, then washed with Tris-buffered saline and Polysorbate 20 (TBST) and incubated with HRP-linked secondary antibody for 1 h at 20 °C. The membranes were further washed with TBST and treated with Immobilon western chemiluminescence HRP-substrates to detect the protein bands using Kodak 4000 MM Pro image station (Molecular Bioimaging, Bend, OR, United States).

The chromatin immunoprecipitation (ChIP) assay was performed using the ChIP-IT Express magnetic chromatin immunoprecipitation kit (Active Motif, Carlsbad, CA) according to the manufacturer’s protocol. Cells (3 × 10⁷) were treated with DMSO or indicated concertation of piperlongumine for 24 h. Cells were then fixed with 1% formaldehyde, and the cross-linking reaction was stopped by addition of 0.125 M glycine. After washing twice with phosphate-buffered saline, cells were scraped and pelleted. Collected cells were hypotonically lysed, and nuclei were collected. Nuclei were then sonicated to the desired chromatin length (200–1,500 bp). The sonicated chromatin was immunoprecipitated with 3 µg of normal IgG (abcam), NR4A1 (Abcam), Sp1 (Abcam), or RNA polymerase II (pol II; Abcam) antibodies and protein G-conjugated magnetic beads at 4 °C for overnight. After the magnetic beads were extensively washed, protein-DNA cross-links were reversed and eluted. Reversed cross-link DNA was prepared by proteinase K digestion followed by Chromatin IP DNA purification (Active Motif). Purified DNA products were then analyzed by quantitative real-time PCR using amfiSure qGreen Q-PCR master mix (genDEPOT) using the manufacturer’s protocol. The primers for detection of the G9a promoter region were F: 5′- CAG​ATG​GGG​ACA​GAG​ACG​C -3′, R: 5′- CCCGGAGCATTGCACG -3’.

Each assay was performed in triplicate and the results were presented as means with standard deviation (SD). The statistical significance of differences between the treatment groups was determined by Dunnett’s multiple comparison test in ordinary one-way ANOVA. Gel analysis of Western blotting was done using ImageJ (1.53K) software (RRID:SCR_003070). GraphPad Prism 8 (Version 8.4.3) software (RRID:SCR_002798) was used for analysis of variance and determine statistical significance. Data with a p-value of less than 0.05 were considered statistically significant and indicated (*) in the figures.

Based on the functional similarities between ROS-inducing anticancer agents and NR4A1 inverse agonists in cancer cell lines we initiated studies on the potential activity of piperlongumine as an NR4A1 ligand. Figure 1A illustrates the direct binding of piperlongumine to the ligand binding domain (LBD) of NR4A1 using an assay which measures the loss of fluorescence of a Trp residue in the LBD (Lee et al., 2014a). The results show that piperlongumine binds NR4A1 with a K_D value of 7.1 µM and as a positive control for this assay we observed a similar binding curve for the known NR4A1 ligand celastrol with a K_D value of 2.3 µM (data not shown). Figure 1B illustrates the interaction of piperlongumine and NR4A1 as determined in the ITC assay and the K_D and ΔG values were 4.97 μmol/L and −30.9 kj/mol respectively. The lower K_D value in the ITC assay may represent interactions of piperlongumine not only within the binding pocket but also other sites in the LBD or NR4A1.

Figure 1C illustrates Maestro/Schrodinger modeling of piperlongumine with the LBD of NR4A1 and shows interactions of piperlongumine with hydrophobic (Pro266, Thr264, Phe261, Leu108, Glu109, Ser110), polar (Ala111, Phe112, Leu113, Glu114, Leu239), and negatively charged (Thr236, Cys235, Arg232, Val179 and Arg184) amino acids. These results (Figures 1A–C) clearly demonstrate that piperlongumine is an NR4A1 ligand and this is further confirmed in NR4A1-dependent transactivation studies in cells transfected with a GAL4-NR4A1 chimera and a GAL4-dependent reporter gene (UAS-luc). The results show that piperlongumine decreased luciferase activity in RKO (Figure 1D), SW480 (Figure 1E), and HCT116 (Figure 1F) colon cancer cells demonstrating the NR4A1 inverse agonist activity of piperlongumine for this transactivation response.

In cancer cells inverse NR4A1 agonists decrease cancer cell growth, induce apoptosis and inhibit migration/invasion (Lee et al., 2014a; Safe et al., 2021) and treatment of RKO, SW480 and HCT116 cells with 7.5 and 15 µM piperlongumine inhibited cell growth (Figures 2A–C) as determined in an XTT assay. Treatment with 7.5 and 15 µM piperlongumine caused the morphology of colon cancer cells to visibly shrink and change into a rounded shape (Figure 2D). Moreover, this same treatment protocol was also used to show that piperlongumine induced apoptosis using the TUNEL assay (Figure 2E).

Figure 3 illustrates that after treatment with piperlongumine for 24 h there was a decrease in the growth-promoting oncogene cMyc and several markers of apoptosis, namely, decreased bcl-2 and survivin and increased cleaved caspase 3 and PARP in the 3 colon cancer cell lines (Figures 3A–C). Figure 3D illustrates and quantitates the effects of piperlongumine on migration of RKO, SW480 and HCT116 colon cancer cell lines. A time 0 the migration lane was created and after 48 h there was significant migration of cells into this lane and the relative amount of migrated cells was set at 100%. Treatment of cells with 1.875—15 µM significantly inhibited cell migration after 48 h, however, at the higher doses this was also due, in part, to the cytotoxicity of piperlongumine. In contrast, at the low concentrations of piperlongumine (1.875 and 3.75 µM) cell migration was still significantly inhibited in RKO and SW480 cells and cytotoxicity was minimal. In HCT116 cells there was also some inhibition of migration at the two lower doses, but it was not significant. Results of a Boyden chamber assay on cell invasion showed that 3.75 and 7.5 µM piperlongumine also inhibited invasion of colon cancer cells (Figure 3E). These results (Figures 3D,E) demonstrate that piperlongumine affects functional responses in colon cancer cells that are consistent with their activity as inverse NR4A1 agonists and effects of NR4A1 knockdown (Safe et al., 2021). We also show that in contrast with results observed for bis-indole derived inverse NR4A1 agonists and resveratrol that piperlongumine did not affect expression of β1-integrin and other integrins (Supplementary Figure S1D) (Zhang et al., 2022).

One of the hallmarks of NR4A1 inverse agonists is the downregulation of pro-reductant genes such as TXNDC5 and IDH-1 which results in increased reactive oxygen species and induction of the oxygen sensor sestrin2 (SESN2) and SESN 2-dependent activation of AMPK (Lee et al., 2014a; Lee et al., 2014b; Hedrick et al., 2015; Lacey et al., 2016; Mohankumar et al., 2019; Shrestha et al., 2020). Treatment of the colon cancer cells with piperlongumine resulted in downregulation of IDH1 in RKO (Figure 4A), SW480 (Figure 4B), and HCT116 (not significant) (Figure 4C) cells and selective downregulation of TXNDC5 in only SW480 cells. Supplementary Figure S1E illustrates that overexpression of FLAG-NR4A1 partially rescues the effects of piperlongumine downregulation of TXNDC5 in SW480 cells. Higher levels of FLAG-NR4A1 were less effective. Induction of sestrin2 and activation (increased phosphorylation) of AMPK was observed in all 3 cell lines. NR4A1 regulates multiple reductant genes (Lee et al., 2014b) and their selective regulation by bis-indole-derived CDIMs and piperlongumine is consistent with their activity as selective NR4A1 modulators.

Results illustrated in Figure 5 summarize effects of piperlongumine alone on induction of ROS and sestrin2 and effects of piperlongumine in combination with the antioxidant glutathione. Treatment of colon cancer cells with piperlongumine increases ROS-dependent fluorescence in all 3 cell which is clue to oxidation of the cell permeant DCFDA into its fluorescent metabolite (Figures 5A–C). In addition, this was accompanied by induction sestrin2 (Figures 5D–F) and the magnitude of the fluorescent and sestrin2 induction responses were decreased after cotreatment with glutathione. Sestrin2-dependent activation of AMPKα inhibits mTOR signaling and results in Figures 6A–C show that piperlongumine significantly decreased levels of phosphor-mTOR in the 3 colon cancer cell lines. Moreover, inhibition of mTOR is accompanied by significantly decreased phosphorylation of 4 E-BP1 and p70S6 in the 3 colon cancer cell lines.

Previous studies have demonstrated that the histone methyltransferase G9a gene is regulated by NR4A1/Sp1 where NR4A1 acts as a ligand-dependent cofactor of Sp1 which in turn is bound to the GC-rich promoter of the G9a gene (Shrestha et al., 2020). Results in Figures 7A–C show that piperlongumine downregulates G9a expression in RKO, SW480 and HCT116 cells and this parallels effects observed for bis-indole derived CDIM compound, quercetin and kaempferol which also bind NR4A1 and act as inverse NR4A1 agonists (Shrestha et al., 2020; Shrestha et al., 2021). Chromatin immunoprecipitation and QPCR were used to quantitatively detect interactions of NR4A1 and Sp1 with the GC-rich G9a promoter (Figure 7D) in SW480 cells. Treatment with piperlongumine had minimal effects on Sp1 binding but significantly decreased NR4A1 interactions in this region of the G9a promoter and similar results were previously observed using bis-indole derived NR4A1 ligands in Rh30 cells (Shrestha et al., 2020). This data is consistent with interaction of piperlongumine and NR4A1 with DNA bound Sp1 on the G9a promoter as previously observed for other inverse NR4A1 agonists that inhibit NR4A1/Sp-regulated gene expression (Lee et al., 2010; Hedrick et al., 2016; Shrestha et al., 2020).