Flag-ATF3 and GST-ATF3, HA-ubiquitin, pCG-ATF3 were kindly provided by Dr. Chunhong Yan (Augusta University) and Dr. Tsonwin Hai (Ohio State University). GST-PXR, GST-PXR (1-107), and GST-PXR (107-434) were kindly presented by Dr. Yanan Tian (Texas A&M University). pCDNA3-MDM2 (Plasmid #16233) and Myc3-HDM2 (Plasmid #20935) were purchased from Addgene (Watertown, MA, USA). The primers used to develop GST-ATF3 truncated protein upon request. PXR protein was purchased from ProteinOne (Rockville, MD, USA). All site-directed mutations of PXR or ATF3 constructs were generated using Quick Change Site-Directed Mutagenesis Kit (Agilent, Santa Clara, CA, USA).

Colon cancer LS180 cells and liver cancer HepG2-C3A cells were obtained from ATCC and cultured in EMEM supplemented with 10% FBS, 100 units/ml penicillin, and 100 µg/ml streptomycin. Cells were verified to be mycoplasma-free by using Mycoplasma Detection Kit (R&D Systems). To generate PXR-knockout (KO) cell line, PXR CRISPR/Cas9 KO plasmid (sc-400824-KO-2) and PXR HDR plasmid (sc-400824-HDR-2) were co-transfected using Lipofectamine 2000 (Invitrogen) according to the manufacturer’s instruction. Here, 2 μg/ml puromycin was used to screen the PXR-KO single clones. For colony formation assays, 1,000 cells were seeded in 6-well plates and were exposed to 0, 10, and 30 Gy of IR, and surviving colonies were fixed with 10% formalin and stained with 0.1% crystal violet 8 days later. Stained colonies were dissolved with buffer containing SDS, and absorbance was recorded at 490 nm using a microplate reader.

Radiated cells were trypsinized, centrifuged for 2 min at 300×g, and resuspended in PBS at 2 × 10⁴/ml. Cell suspensions were then mixed with 1% of low-melting point agarose at 1:3 (V/V) and layered on a microscope slide precoated with 1% agarose gel. Cells in gels were lysed with the neutral lysis buffer (2 M NaCl, 30 mM EDTA, 10 mM Tris-HCl, pH 8.5, 1% N-laurylsarcosine, 1% Triton X-100) at 4°C for 1 h. After rinsing 3 times with the TBE buffer (90 mM Tris, 90 mM boric acid, 2 mM EDTA, pH 8.5) at room temperature for 30 min, slides were submerged in TBE for electrophoresis at 20 V for 25 min. Slides were then rinsed, neutralized in distilled water, dipped in ethanol, and dried. Slides were stained using 2.5 µg/ml of ethidium bromide in distilled water for 20 min and observed under a fluorescence microscope. At least 100 comet images from each slide were analyzed using the CASP software (http://casplab.com/).

Cells were lysed in modified radioimmunoprecipitation assay (RIPA) containing 50 mm Tris-HCl (pH 7.4), 150 mm NaCl, 1% Nonidet P-40, 0.1% SDS, 0.5% sodium deoxycholate, 1 mm EDTA, and proteinase inhibitor mixtures. The proteins were resolved in 10%–12% SDS-PAGE gels and detected using the desired antibodies. The antibodies and the dilution ratios are as follows: ATF3 (sc-44C3a, 1:1000), PXR (sc-48340, 1:1000), and HA antibody (sc-805, 1:200) were purchased from Santa Cruz Biotechnology, Inc. (Dallas, TX, USA). MDM2 (4B11, 1:500) was purchased from EMD Millipore (Billerica, MA, USA); anti-mouse IgG VeriBlot for IP secondary antibody (HRP) (ab131368) and anti-Phosphor-(Ser/Thr) Phe antibody (ab17464, 1:1000) were purchased from Abcam (Cambridge, MA, USA). FLAG (F3165, 1:4000) and actin (A5441, 1:5000) were purchased from MilliporeSigma (St. Louis, MO, USA). The p-Ser1981 ATM (#5883, 1:1,000) and p-H2AX (#2557, 1:1000) were purchased from Cell Signaling (Danvers, MA, USA).

Total RNA was extracted from cells using TRIzol (Invitrogen, Carlsbad, CA, USA), reverse-transcribed using the SuperScript^® VILO™ cDNA Synthesis Kit (Thermo Fisher Scientific Inc. NYSE: TMO), and subjected to real-time PCR assays using SYBR Green reagents (Qiagen, Venlo, Netherlands). The sequences of the primers were as follows: GAPDH, 5’-AACGGATTTGGTCGTATTGGG-3’ and 5’-CCTGGAAGATGGTGATGGGATT-3’; ATF3, 5’-GTGCCGAAACAAGAAGAAGG-3’ and 5’-TCTGAGCCTTCAGTTCAGCA-3’; PXR, 5’-GGCCACTGGCTATCACTTCAA-3’ and 5’-TTCATGGCCCTCCTGAAAA-3’.

The PXR and MDM2 siRNA SMARTpool were purchased from Dharmacon (Thermo Fisher Scientific Inc. NYSE: TMO). Double knockdown of PXR and MDM2 or ATF3 expression was carried out using RNAiMAX (Invitrogen, Carlsbad, CA, USA) following the manufacturer’s protocols. Briefly, HepG2-C3A and LS180 cells were transfected with 100 pM SiPXR, Si MDM2, SiATF3, or SiLuc as control for 48 h and aliquoted for further run of knockdown. After 72 h, cells were harvested and subjected to Western blotting assays.

In this study, 1 μg GST or GST fusion proteins immobilized on glutathione-agarose from MilliporeSigma (St. Louis, MO, USA) were incubated with in vitro-translated proteins (5 μl) (TnT^® Quick Coupled Transcription/Translation Systems, L1170, Promega, Madison, WI, USA) or recombinant proteins (50–100 ng) (ProteinOne, Rockville, MD, USA) in a buffer containing 20 mM Tris-HCl, pH 8.0, 100 mM NaCl, 2 mM EDTA, 2 mM DTT, 5% glycerol, and 0.4% NP-40 at 4°C overnight, followed by extensive washes with a similar buffer containing 150 mM NaCl. Bound proteins in the beads were eluted by boiling in the SDS-PAGE loading buffer and detected by Western blotting or fluorography using a desired antibody.

Cell lysates (1–2 mg) were incubated with 20 μl protein A/protein G agarose beads (Cat #: IP05, Millipore) in RIPA buffer (50 mM Tris-HCl, pH 7.4, 100 mM NaCl, 1% NP-40, 0.1% SDS, 0.5% sodium deoxycholate, and 1 mM EDTA) at 4°C overnight for precleaning. After centrifugation, the supernatants were transferred to incubate with another 20 μl of protein A/protein G agarose beads (Millipore) together with 1 μg corresponding antibody in RIPA buffer at 4°C for 5 h. After extensive washes with RIPA buffer containing 150 mM NaCl, bound proteins in the beads were eluted by boiling in the SDS-PAGE loading buffer and detected by Western blotting or fluorography using a desired antibody.

The in vivo ubiquitination assay was performed according to the published method (10). Briefly, H1299 cells were transfected with FLAG-ATF3 and HA-ubiquitin with or without MDM2 or PXR and treated with 5 μM of MG132 overnight and then lysed in the FLAG lysis buffer (50 mM Tris-HCl, pH 7.9, 137 mM NaCl, 10 mM NaF, 1 mM EDTA, 1% Triton X-100, 0.2% sarkosyl, and 10% glycerol). Cell lysates (1–2 mg) were incubated with 20 μl of anti-FLAG M2 affinity gel (Sigma) at 4°C overnight. After extensive washes, agarose gels were loaded on spin columns (Affymetrix), and bound FLAG-ATF3 was eluted with 20 μl of FLAG peptide at a final concentration of 0.1 mg/ml. ATF3 ubiquitination was determined by Western blotting using the hemagglutinin (HA) antibody.

The in vitro ubiquitination assay was performed using a human MDM2/HDM2 Ubiquitin Ligase Kit (K-200B) by following the manufacturer’s instructions with modifications. Briefly, in vitro translated ATF3 (TnT^® Quick Coupled Transcription/Translation Systems, L1170) was incubated with E1, E2, and GST-MDM2, ubiquitin with buffer containing 50 mM Tris-HCl, pH 8.0, 50 mM NaCl, 1 mM EDTA, 5% glycerol, 10 mM fresh DTT on ice for 2 h and then add with or without PXR recombinant protein (ProteinOne, Rockville, MD) at 37°C for another 2 h. Polyubiquitinated proteins were detected using ATF3 antibody or FK-1 (ENZO) after resolving in SDS-PAGE.

Tissue array that incorporated 50 pairs of human colon adenocarcinoma tissue and adjacent non-tumor tissue sample (D100Co01) from Bioaitech (China) was used to stain the protein expression of PXR (sc-48340, 1: 100) and MDM2 (4B11, 1: 100). Briefly, the tissue array was permeabilized with 0.1% Triton X-100 in PBS solution, and then the array was incubated with the indicated antibody overnight at 4°C, followed by incubation with immunofluorescence second antibody (A32728 and A32742, respectively, Invitrogen) at room temperature for 1 h. The array was washed three times with PBS and sealed with mounting media containing DAPI (Vector Labs, Burlingame, CA, USA; H-2000). The photographs were scanned with 3DHISTECH (PANNORAMIC DESK/MIDI/250/1000, Hungary) and analyzed by ImageJ software.

Data were analyzed using Prism Software 5.0 (GraphPad Software, Inc.). The statistical significance (p < 0.05) was evaluated by Student’s t test and one-way ANOVA.

Given that ATF3 is quite critical in regulating Tip60 and p53-mediated DDR (7, 10), we want to explore whether PXR binds to ATF3. GST-pulldown assays showed that immobilized GST-ATF3 directly pulled down PXR protein in vitro. This was specific since GST alone did not pull down PXR ( Figure 1A ). In order to investigate whether the interaction between PXR and ATF3 occurs endogenously, we performed in vivo reciprocal co-immunoprecipitation (co-IP) assays using H1299 cells transiently co-transfected with PXR and/or ATF3 plasmid. The results showed that ATF3 antibody but not IgG precipitated PXR in H1299 cells transfected with both ATF3 and PXR ( Figure 1B ). Since PXR protein was only expressed in the liver cancer HepG2-C3A cells and colon adenocarcinoma LS180 cells ( Supplementary Figure S1A ), we then generated PXR KO of HepG2-C3A and LS180 cells. We found that PXR antibody precipitated ATF3 in wild-type LS180 cells but not in PXR-KO LS180 cells ( Figure 1C ). These results suggested that PXR physically interacted with ATF3. Because ATF3 expression alters largely under DNA damage treatments (7), we also determined whether the DNA damage condition affects the interaction between ATF3 and PXR. We then subjected LS180 cells to γ-IR and incubated cell lysates with GST-ATF3 and carried out the endogenous GST-pulldown assay. The results showed that PXR also interacted with ATF3 upon IR ( Figure 1D , Supplementary Figure S1B ); furthermore, like the ATF3–Tip60 interaction (7), the binding between PXR and ATF3 was altered neither by time extension after IR treatment nor by DNA damage treatment, such as IR, doxorubicin (DOX), or camptothecin (CPT) ( Figure 1D and Supplementary Figure S1B ).

ATF3 protein has a very short half-life, and its protein stability can be altered (17). We then want to know whether PXR can affect the protein stability of ATF3. Results of co-expression with a fixed amount of ATF3 and different doses of PXR expression plasmids showed that the ATF3 protein levels were increased in association with increments of the doses of PXR plasmid ( Figure 2A ). KO of PXR decreased ATF3 expression ( Figure 2B ). The regulation of ATF3 by PXR does not likely occur in the transcriptional level, as there were no significant differences in the ATF3 mRNA levels between the wild-type and the screened PXR-KO clones ( Figure 2B ). Furthermore, this effect is not clone specific, as we transfected HepG2-C3A and LS180 wild-type cells with siRNAs targeting PXR, and the results confirmed that knockdown of PXR resulted in downregulation of ATF3 ( Supplementary Figure S2B ). PXR indeed extended the half-life of endogenous ATF3 protein ( Figure 2C ). These results have suggested that PXR stabilizes ATF3 at the protein level. It has been shown that ATF3 is susceptible to undergo ubiquitination, then proteolytic degradation (17, 26, 27); we next investigate whether ubiquitination of ATF3 is regulated by PXR. Since ATF3 can work as a ubiquitin trapper and FLAG antibody can precipitate ubiquitinated ATF3 (8), we transfected FLAG-ATF3, HA-ubiquitin, with or without PXR, and found that PXR indeed decreased ubiquitination of ATF3 in a dose-dependent manner ( Figure 2D , lane 5 vs. lane 4). And this effect was proteasome dependent, as ATF3 decreased when ubiquitin was added ( Supplementary Figure S2B ). Taken together, PXR increased ATF3 stability and extended ATF3 protein half-life by blocking its ubiquitination and degradation.

We next want to figure out whether interactions between PXR and ATF3 are required for PXR-mediated regulation of ATF3. In order to gain this aim, we built several truncated constructs of ATF3 fused with GST agarose beads, incubated with in vitro translated PXR, and performed GST-pulldown assay. We found that full-length ATF3 except for leucine zipper (Zip) domain bound with PXR as a fragment of aa 1-79, aa 80-100, aa 141-181 was required for ATF3 to interact with PXR ( Figure 3A ). We therefore want to know whether there is an ATF3 binding-deficient lysine (K) residue that loses the ability to be ubiquitinated by PXR. Normally, lysine residue on the protein structure can be substrate to attach ubiquitination cascade (14), then we mutated 6 lysine residues on the binding region of ATF3 to arginine (R), performed FLAG-IP to test which mutant will lose the binding with PXR, and further abolished PXR-mediated regulation. Indeed, K42R mutation failed to pull down PXR ( Figure 3B , lane 3 vs. lane 2). Furthermore, the FLAG-tagged ATF3 K42R mutation abolished PXR-mediated reduction of ubiquitination of ATF3 ( Figure 3C , lane 4 vs. lane 3). The results suggested that the ATF3 K42 in a fragment of aa 1-79 was essential for PXR-inhibited ubiquitination of ATF3.

On the other hand, GST-pulldown assay revealed that partial ligand-binding domain of PXR (aa 300-434) was required for PXR to interact with ATF3 as a truncated fragment of aa 1-434, aa 1-420 but not aa 1-180, or aa1-300 pulled down ATF3 ( Figure 3D ). PXR protein structure is enriched with serine (S)/threonine (T) amino acid and defined as a phosphoprotein (28). Indeed, we observed that IR induced significant increases in the phosphorylation of S/T in PXR ( Figure 3E ). In order to identify which S/T sites are essential for PXR–ATF3 interaction, we mutated 12 S/T sites to alanine (A), which is located on aa 300-434 of PXR-binding region, and we found that the T432A mutant loses the ability to interact with ATF3 ( Figure 3F , lane 13 vs. lane 2). In order to confirm that the phosphorylation of T432 is critical for ATF3-mediated ATM activation upon IR, we transfected the wild-type PXR or the PXR T432A mutant plasmid into the PXR-KO LS180 cells and then subjected cells to IR. The results demonstrated that the p-ATM level was significantly decreased in the PXR-KO LS180 cells transfected with the PXR T432A mutant plasmid compared with that of the wild-type PXR transfection ( Figure 3G , lane 6 vs. lane 4), suggesting that the phosphorylation of PXR T432 was pivotal for ATF3-mediated ATM activation upon IR in colon cancer cells.

Besides p53, ATF3 is a well-known substrate of E3 ligase MDM2, and MDM2 is responsible for the ubiquitination and degradation of ATF3 (17). This knowledge leads us to investigate whether PXR disrupts ATF3–MDM2 interaction, thus abrogating MDM2-mediated ubiquitination and stabilizing ATF3. To gain this aim, we incubated in vitro translated MDM2 with glutathione agarose beads conjugated with ATF3 overnight, added recombinant PXR to the system, and performed GST-pulldown assay. The results showed that ATF3 pulled down MDM2 as expected, and the amount of MDM2 bound with ATF3 was decreased by addition of recombinant PXR in a dose-dependent manner, suggesting that PXR indeed disrupted ATF3–MDM2 interaction in vitro ( Figure 4A ). Next, we want to clarify whether PXR increases ATF3 stability by blocking MDM2-catalyzed ATF3 ubiquitination. By applying FLAG-IP, we found that the ubiquitination levels of ATF3 were increased by MDM2, whereas PXR abolished MDM2-mediated ubiquitination ( Figure 4B , lanes 5 vs. lane 4). By applying Western blotting, we found that the ATF3 protein levels were decreased by MDM2, while these were increased by addition of PXR ( Figure 4C , lane 3 vs. lane 2, lane 5 vs. lane 4). Furthermore, we also performed in vitro ubiquitination assay, and the results demonstrated that the ATF3 ubiquitination levels were enhanced by MDM2 in a dose-dependent manner ( Figure 4D , lane 2 and lane 4) as expected. More importantly, addition of PXR reduced the MDM2-mediated ATF3 ubiquitination ( Figure 4D , lane 3 vs. lane 2 and lane 5 vs. lane 4). However, PXR is not an E3 ligase for ATF3, and it has no capability to induce ubiquitination of ATF3 ( Supplementary Figure S4E , lane 6 vs. lane 3). Collectively, our results indicated that PXR suppressed MDM2-catalyzed ubiquitination of ATF3 in vivo and in vitro.

Interestingly, we noticed a diminished band of MDM2 when we co-transfected MDM2 with PXR plasmid together ( Figure 4B , lane 5 vs. lane 4), which indicated that PXR might affect MDM2 expression level. To further understand this regulation, we analyzed 151 samples from The Cancer Genome Atlas rectum adenocarcinoma (TCGA-READ) database (portal.gdc.cancer.gov/) and found that the mRNA expression level of PXR negatively correlated with mRNA expression of MDM2 (p < 0.05) ( Figure 5A ). Immunofluorescence staining has revealed that PXR protein level is ubiquitously higher in human colon adenocarcinoma tissues than that in adjacent non-tumor tissues, whereas MDM2 protein expression has the opposite trend with PXR in tumor tissue vs. that in adjacent non-tumor tissues ( Figure 5B ). Indeed, PXR has physical binding with MDM2. By performing GST-pulldown assay, we found that MDM2 bound with PXR in the DNA-binding domain (DBD) (aa 1-107) ( Figure 5C and Supplementary Figure S4A ). Conversely, PXR binds with MDM2 on the RING domain (aa 431-491) ( Figure 5D and Supplementary Figure S4B ). The RING domain of MDM2 is essential for its E3 ligase activity. Coincidently, ATF3 also binds with MDM2 on this RING domain. The structure binding analysis provides the explanation that PXR structurally can interrupt the interaction between ATF3 and MDM2. In order to further address PXR-inhibited MDM2 expression, we detected the half-life of MDM2 and ubiquitination of MDM2. Strikingly, the result showed that PXR reduced the MDM2 half-life in colon cancer cells ( Figure 5E ), and PXR increased endogenous ubiquitination of MDM2 ( Figure 5F ). MDM2 is an oncogene, and its auto-ubiquitination enhances its substrate ubiquitin ligase activity (29). In order to further confirm that PXR increases auto-ubiquitination of MDM2, we performed in vitro ubiquitination assays, incubated E1 and E2, and purified GST-MDM2 and recombinant PXR in vitro together. FK-1 antibody was used to detect polyubiquitination (7). The results showed that PXR indeed increased auto-ubiquitination of MDM2 in a dose-dependent manner ( Figure 5G ). In order to further confirm that PXR is an upstream regulator for MDM2, we transfected siRNA targeting PXR, and the results showed that PXR knockdown resulted in downregulation of ATF3 and upregulation of MDM2 ( Figure 5H , lane 2 vs. lane 1, Supplementary Figure S4C ). More importantly, it seems that PXR expression level is not altered by MDM2 ( Figure 5G , lane 3 vs. lane 1), whereas MDM2 expression is dependent on PXR regulation ( Figure 5H , lane 4 vs. lane 2). As a consequence, like ATF3, tumor suppressor p53, both of which are substrates of MDM2, also can be upregulated by PXR ( Supplementary Figure S4D ).

Previously, we proved that ATF3 promoted ATM activation upon IR treatment via USP14-Tip60 axis. Consistent with these results, IR-induced phosphorylation of ATM was dramatically suppressed in the PXR-KO cells where the ATF3 expression level was downregulated ( Figure 6A , lanes 6–10 vs. lanes 1–5). Similarly, IR-induced phosphorylation of ATM substrates including H2AX was also repressed in the PXR-KO cells ( Figure 6A ). The suppression of ATM activation by PXR KO was not limited to LS180 cells, as similar observations were obtained in HepG2-C3A PXR-KO cells ( Supplementary Figure S5A ). Meanwhile, upregulated MDM2 followed PXR KO ( Figure 6A ). In line with this result, there are more colonies in the wild-type LS180 and HepG2-C3A cells after IR than those in PXR-KO cells ( Figure 6B ). Besides, the olive tail moments of the wild-type LS180 cells were much shorter than those of the PXR-KO LS180 cells ( Figure 6C ). These results suggested that PXR promoted cell viability and conferred IR resistance through facilitating ATM activation of DNA damage response and repair in liver and colon cancer cells. In order to confirm that PXR promotes ATM signaling through regulating ATF3, we knocked down ATF3 expression by siRNA in both wild-type and PXR-KO LS180 cells and then determined IR-induced ATM activation. Consistent with our previous report, knockdown of ATF3 obviously compromised ATM activation ( Figure 6D and Supplementary Figure S5B , lane 4 vs. lane 2), and KO of PXR indeed downregulated ATF3. Furthermore, KO of PXR induced much more suppression of ATM activation where ATF3 expression was further inhibited ( Figure 6D and Supplementary Figure S5B , lane 8 vs. lane 4). Surprisingly, the basal level of phosphorylated ATM was almost abolished when ATF3 expression was knocked down in PXR-KO HepG2-C3A cells ( Supplementary Figure S5B ).