Identification of TRIML2, a Novel p53 Target, that Enhances p53 SUMOylation and Regulates the Transactivation of Proapoptotic Genes Free

This article is featured in Highlights of This Issue, p. 209

The p53 tumor-suppressor gene (TP53) is inactivated by mutation in more than half of human cancers (1). Following stresses such as DNA damage or oncogene activation, p53 is activated to trigger a variety of biologic functions to suppress tumor development, including cell-cycle arrest, senescence, and apoptosis. This versatility in p53 outcomes lies primarily in its role as a transcription factor, and its ability to transcriptionally activate different classes of p53 target genes that play roles in different outcomes (2). For example, the p53 target genes CDKN1A (p21) and Gadd45 play key roles in p53-induced cell-cycle arrest, whereas the BH3-only encoding target genes BBC3 (PUMA) and PMAIP1 (NOXA) are critical players in p53-mediated apoptotic cell death (3). One of the challenges in p53 research has been to identify how this protein selectively induces different classes of target genes, and different cell fates, in response to different stresses. To date, protein–protein interactions and posttranslational modification have been linked to the ability of p53 to selectively transactivate cell-cycle arrest or proapoptotic target genes (4). In addition, the cell-of-origin, and the level and type of stress, are also implicated in the p53 decision between growth arrest and apoptosis (5). Finally, data from our group and others indicate that this decision can also be affected by a common polymorphism in the p53 gene at codon 72, encoding either proline (P72) or arginine (R72).

The codon 72 polymorphism in p53 is the most common coding region polymorphism in the TP53 gene (6). There is a distinct latitudinal bias in the frequencies of P72 and R72 alleles, with the P72 allele more common in populations near the equator (7). This latitudinal bias in codon 72 allele frequency has been suggested to be associated with either the level of UV exposure or winter temperature (8). The change from a proline to an arginine at amino acid 72 is predicted to result in a significant structural change of p53 (9), and several functional differences between these polymorphic variants have been described. Specifically, under the same DNA damage signals, the P72 variant preferentially promotes cell-cycle arrest, whereas the R72 variant shows superior ability to induce apoptosis (9, 10). At present, the underlying basis for the differences in growth arrest and apoptosis between these variants is incompletely understood. In this study, we undertook an unbiased approach toward this question, and identified a p53 target gene that is transactivated to a significantly greater extent by the R72 variant of p53, in multiple different cell lines containing endogenous or inducible p53. We show that this gene, TRIML2, encodes a protein that feeds back on p53 to bind to it and target it for SUMO-2 modification. We further show that cells with higher levels of TRIML2 show superior ability to transactivate a subset of p53 target genes that are associated with prolonged DNA damage and apoptosis, including PIDD (PIDD1), PIG3 (TP53I3), and PIG6 (PRODH). We posit that the R72 variant possesses enhanced apoptotic potential, at least in part, because of its superior ability to transactivate TRIML2; this leads to increased transactivation of a subset of p53 target genes that are associated with prolonged DNA damage and apoptosis.

Unless otherwise mentioned, all cell lines were obtained from the ATCC, and cells were used during early passages after being thawed from frozen stocks. Normal human fibroblast (NHF) cells were purchased from the Coriell Institute for Medical Research (Camden, NJ) and cultured in 15% FBS (Gembio), penicillin (100 IU/mL)–streptomycin (100 μg/mL; Pen/Strep; Cellgro, 30-002-CI), Nonessential amino acids (Cellgro; 25-030-CI), l-glutamine (Cellgro; 25-005-CI) in DMEM (Cellgro). Temperature-sensitive Saos2 cells containing P72 or R72 variant of p53 were described previously (11). These cells were maintained at 39°C in DMEM supplemented with 10% FBS and penicillin–streptomycin. H1299 cells containing Tet-inducible variants of p53 P72 and R72 were provided by Steven McMahon (Thomas Jefferson University; ref. 12) and cultured in DMEM supplemented with 10% Tet-approved system FBS (Clontech; 631106) and penicillin–streptomycin. Hct116 p53 WT and Hct116 p53^−/− cells were provided by Bert Vogelstein from Johns Hopkins School of Medicine (13). These cells were cultured in McCoy's 5A medium (Cellgro; 10-050-CV) supplemented with 10% FBS and penicillin–streptomycin. U2OS-Trex cells were provided by Pradip Raychaudhuri (University of Illinois, Chicago, IL) and maintained in DMEM supplemented with 10% Tet system–approved FBS and penicillin–streptomycin. Etoposide (Sigma; E1383) was used at a concentration of 20 to 100 μmol/L. 5-fluoruracil (5-FU; Sigma; 858471) was used at a concentration of 5 μmol/L. A concentration of 0.75 μg/mL doxycycline (BD Biosciences; 631311) was used to induce gene expression from tetracycline-inducible constructs. Ginkgolic acid (Sigma; 75741) was used as sumoylation inhibitor at a concentration of 100 μmol/L. TRIML2 cDNA was purchased in pCR4-TOPO vector from Open Biosystems (8327427) and subcloned into pCR2.1-TOPO vector through TOPO TA cloning according to the manufacturer's protocols (Invitrogen; K4575-01SC). TRIML2-overexpressing construct was created by subcloning TRIML2 to pcDNA3.1D/V5-His-TOPO vector through restriction enzyme digestions (Hind III and Xho I) and ligation. TRIML2 was subsequently subcloned into pcDNA4/TO vector through Hind III/Xho I digestions and ligation to generate tetracycline-inducible construct. Stable cells overexpressing pcDNA3.1-TRIML2 or pcDNA4/TO-TRIML2 were maintained under the selection using 400 μg/mL G418 and 100 μg/mL Zeocin, respectively. Expression constructs (all in pRK5 vector) of TRIM27 (Flag-tagged), PML (isoform IV, Flag-tagged), Ubiquitin (HA-tagged), SUMO1 (His-tagged), and SUMO2 (His-tagged) were obtained from Xiaolu Yang (University of Pennsylvania, Philadelphia, PA; ref. 14). Fugene 6 transfection reagent (Promega) was used for all transfection experiments.

Hupki (human p53 knock-in) P72 and R72 mice were described previously (12). All studies with mice complied with all federal and institutional guidelines as per Institutional Animal Care and Use Committee protocols. Mice were housed in plastic cages with ad libitum diet and maintained at 22°C with a 12-hour dark/12-hour light cycle. Primary mouse embryonic fibroblasts (MEF) obtained from 13.5-day-old Hupki mouse containing either homozygous P72 or R72 p53 were grown in DMEM supplemented with 10% FBS and 1% penicillin—streptomycin. For irradiation experiments, mice were exposed to a cesium-137 gamma source (The Wistar Institute) and tissues harvested were subjected to the RNA Extraction Using RNeasy Mini Kit (Qiagen; 74104).

NHF cells expressing homozygous P72 or R72 forms of p53 as well as cells expressing an shRNA against p53 (shp53) were treated with 5 Gy of gamma radiation. RNA was isolated from the cells using TRizol (Invitrogen; 15596-026) before being amplified and labeled using the Agilent Quick Amp Labeling Kit. Amplified cDNAs were hybridized onto human gene expression 4 × 44 K v2 arrays (Agilent; G4845A) according to the Agilent protocol. Hybridized slides were scanned at a 5-μm resolution on an Agilent scanner, and fluorescence intensities of hybridization signals were extracted using Agilent Feature Extraction software. Raw expression data obtained from Agilent microarrays were background corrected and quantile normalized across the experimental conditions (15). The LIMMA (Linear Models for Microarray Data) methodology was applied to the log₂-transformed expression data to identify differentially expressed genes in each comparison. The LIMMA module in the Open Source R/Bioconductor package was used in the computations (16). Differentially expressed genes were identified on the basis of statistical significance (P < 0.01) as well as biologic significance using fold change cutoff. Genes identified through microarray were analyzed through the use of Ingenuity Pathway Analysis (IPA) (Ingenuity Systems, www.ingenuity.com) for their associated functions and diseases. Gene-expression data were deposited into the GEO database with accession number GSE61124.

Stable cell lines for shRNA knockdowns were generated by infection with the lentiviral vector pLKO.1-puro carrying an shRNA sequence against TRIML2: sh-A(TCCAATGTTAAATGTCTCTGG) TRCN0000150366, sh-B(TTTAGCTGCTTCAAGTTTCTC) TRCN0000150766, and sh-C(AAATCCAATCTTTCTGGGTTG) TRCN0000150389 (Open Biosystems). VSVG-pseudotyped lentivirus was generated by cotransfection of 293-FT cells with shRNA constructs and packaging vectors according to the manufacturer's protocols (Invitrogen; K4960-00). Lentivirus was added to cells with Polybrene (6 μg/mL) for maximum viral transduction. Stable cells were selected using puromycin (1 μg/mL), and gene knockdown was confirmed by quantitative reverse-transcription PCR (qRT-PCR) and Western blot analysis.

For Western blot analysis, 25 to 50 μg of protein was resolved over SDSPAGE using precast NuPAGE Bis-Tris gels (Life Technologies) and transferred onto polyvinylidene difluoride membranes (Bio-Rad). Primary antibodies used in this study include p53 (Ab6, Calbiochem, and OP43), p53 Ser-15-P (Cell Signaling Technology; 9284), p53 Ser-46-P (Cell Signaling Technology; 2521), TRIML2 (Sigma; HPA043838), TRIML2 (Abcam; ab87292), FLAG-Tag (M2, Sigma; F3165), GAPDH (14C10; Cell Signaling Technology; 2118), Actin (AC-15, Sigma; A5441), p21 (Ab6, Calbiochem; OP79), MDM2 (Ab1-OP46, Ab2-OP115; Calbiochem), cleaved lamin A (Cell Signaling Technology, 2035), cleaved caspase-3 (Cell Signaling Technology; 9061), PIDD (Anto-1, Novus Biologicals; NBP1-97595), caspase-2 (Cell Signaling Technology; 2224), PIG3 (Ab1, Oncogene; PC268), PML (Santa Cruz Biotechnology; sc-5621), PARP (46D11, Cell Signaling Technology; 9532), SUMO2/3 (Cell Signaling Technology; 4971), SUMO1 (Cell Signaling Technology; 4930), ubiquitin (Cell Signaling Technology; 3933). Secondary antibodies conjugated to horseradish peroxidase were used at a dilution of 1:10,000 (Jackson Immunochemicals). ECL (Amersham; RPN2232) was then applied to blots and protein levels were detected using autoradiography. Densitometry quantification of protein signals was performed using ImageJ software (NIH). For immunoprecipitation, a total of 500 to 2,000 μg of whole-cell lysate was precleared by protein G-agarose beads (Millipore; 16266) and incubated with 1 μg of antibody overnight at 4°C. Protein G-agarose beads were then added for 1 hour, followed by washes using lysis buffer. Pulled-down proteins were eluted using 2× Laemmli sample buffer (62.5 mmol/L Tris-HCl, pH 6.8, 25% glycerol, 2% SDS, 0.01% Bromophenol Blue) and subjected to Western blot analysis as described earlier. Horseradish peroxidase–conjugated light-chain–specific secondary antibody was used (Jackson Immunochemicals).

Treated cells were lysed using the QIAshredder columns (Qiagen, 79656) and total RNA was isolated using the RNeasy Mini Kit (Qiagen; 74104), including on-column DNase digestion (Qiagen; 79254) following the manufacturer's protocol. Equal amounts of RNA from samples were used to create cDNA with the High Capacity cDNA Reverse Transcription Kit (Applied Biosciences; 4368814). Quantitative PCR (qPCR) was performed using the Brilliant III Ultra-Fast SYBR Green QPCR Mix Kit (Agilent Technologies; 600882) on the Stratagene Mx3005P device (Agilent Technologies), according to the manufacturer's protocol. Data analysis was done using the MxPro program (Stratagene). Messenger RNA expression levels were normalized to cyclophilin A. Primers used in qRT-PCR analyses were designed on the basis of published sequence information (Ensembl genome browser) and to prevent amplication of genomic DNA of target genes (Supplementary Table S2).

Chromatin immunoprecipitation (ChIP; ref. 16) analysis was performed according to the manufacturer's protocol (EZ-ChIP; Millipore) with slight modifications. Briefly, cells were cultivated in 100-mm plates to 90% confluence and scrape harvested. Cells were then fixed for 10 minutes in freshly made fixation solution (1% formaldehyde, 10 mmol/L NaCl, 0.1 mmol/L EDTA, 5 mmol/L HEPES, pH7.9). Fixation was stopped by 0.125 mol/L glycine, followed by PBS wash, and cells were lysed in SDS lysis buffer (1% SDS, 10 mmol/L EDTA, 50 mmol/L Tris pH8.1) containing protease inhibitors. Chromatin was sheared by sonication to an average size of approximately 200 to 500 bp, clarified, and precleared for 1 hour at 4°C with salmon sperm DNA-saturated protein A-Sepharose beads (Millipore; 16-157). The supernatant was incubated with normal mouse IgG or anti-p53 (Ab6, Calbiochem; OP43) and rotated overnight at 4°C. Lysates were immunoprecipitated with salmon sperm DNA-saturated protein A-Sepharose beads for 1 hour at 4°C and washed extensively according to the manufacturer's instructions. Input and immunoprecipitated protein/DNA complexes were eluted at room temperature, and the cross-linking was reversed overnight at 65°C in the presence of 200 mmol/L NaCl. After RNase A (37°C for 30 minutes) and proteinase K (45°C for 2 hours) treatment, sample DNAs were purified through DNA columns for further analysis. Conventional PCR of ChIP products was performed with GoTaq Hot Start Polymerase (Promega; M5001). qPCR was performed as described above. p53 responsive elements (RE) on TRIML2 were predicted through MatInspector software of Genomatix using consensus the p53-binding sequence RRRCWWGYYY (R = A or G; W = A or T; Y = C or T), allowing no more than two mismatches (17). Primer pairs used for different ChIP target sequences are listed in Supplementary Table S2.

p53-null H1299 cells seeded in 6-well plates were cotransfected with pcDNA3-TRIML2 and p53 (in pRc/CMV vector) expression constructs for 48 hours, followed by 4% paraformaldehyde fixation and 0.2% Triton X-100 permeabilization. Transfected cells were stained with antibodies to p53 (Ab6, Calbiochem; OP43) and TRIML2 (Sigma; HPA043838) and then incubated with fluorescence-conjugated secondary antibodies (Jackson Immunoresearch Laboratories; Rhodamine Red-X RRX anti-rabbit, 111-295-144; Alexa Fluor 488 anti-mouse, 115-545-146). Counterstaining with DAPI was applied to stain the cell nucleus. The samples were imaged using a Leica TCS SP5 II scanning laser confocal system (Wistar Institute Imaging Facility). Annexin V staining to detect apoptotic cells and cell viability assay were performed using Guava Nexin Reagent (Millipore; 4500-0450) and ViaCount Reagent (Millipore; 4000-0041) on the Guava easyCyte HT system according to the manufacturer's protocols, respectively. For statistical analysis, data were analyzed by the two-sided unpaired Student t test using a GraphPad software package. Data were considered significant if P < 0.05.

To identify p53 target genes that are differentially regulated by the P72 and R72 variants of p53, we purchased and analyzed over two dozen NHF cell lines from the Coriell Institute by genotyping them for the codon 72 polymorphism. This led to the identification of four NHF cell lines, two of which were homozygous for P72, and two of which were homozygous for R72; these were all early passage and showed similar doubling times of approximately 24 hours (not shown). Two of these lines were selected for further analysis. We first set out to show that these cell lines, which are from nonrelated individuals, respond similarly to DNA damage. In response to etoposide, these P72 and R72 NHF lines showed nearly identical p53 induction, phosphorylation, and upregulation of MDM2 and CDKN1A (p21; Fig. 1A).

Two of the P72 and R72 NHF lines, which we denote P72-1 and R72-1, were used to create subclones expressing a p53 short hairpin (shp53), and were confirmed to express significantly reduced p53 levels (data not shown). Subsequently all six cell lines (P72-1, P72-1shp53, P72-2, R72-1, R72-1shp53, and R72-2) were subjected to 5 Gy gamma radiation and RNA was isolated after 0, 2, and 4 hours for microarray analysis on Agilent arrays (Supplementary Fig. S1A); gamma irradiation was chosen as the stressor so that these findings could be compared with a previous analysis from our group of P72- and R72-induced genes in the mouse (12). This analysis was performed in triplicate on independent samples; for data analysis, we focused on genes whose expression level was changed more than 50% after irradiation in all four P72 and R72 cell lines, and whose upregulation was absent or significantly reduced in NHFs expressing the short hairpin to p53 (Supplementary Fig. S1B). This analysis revealed more than 200 genes regulated in a p53-dependent manner; the overwhelming majority of these genes were regulated similarly in P72 and R72 NHFs. A significant number of these “commonly regulated” genes were previously identified p53 target genes whose protein products are associated with known p53-mediated functions, including cell death, cell cycle, and DNA repair (Supplementary Fig. S1C). This analysis also revealed approximately three dozen genes that were differentially regulated between P72 and R72 NHFs; this subset included genes that were induced only in P72, or only in R72 (Supplementary Table S1 and Supplementary Fig. S1D). Interestingly, the majority of the “differentially regulated” genes could not be linked to classic p53-associated functions (Supplementary Fig. S1C, bottom).

Several of the genes showing greater upregulation by the P72 variant were identical to those previously identified by our group as showing preferential upregulation by this variant in mouse cells (12). Conversely, there was only one gene identified as being specifically upregulated in R72 NHFs; this was the gene encoding TRIML2 (identified as FLJ25801 in the microarray gene list), which was upregulated in the R72 NHFs, but not P72, and whose upregulation was lost in shp53-expressing cells (Supplementary Fig. S2A). Western blot analysis confirmed the increased expression of TRIML2 protein following gamma radiation in R72 but not P72 NHFs (Fig. 1B). Importantly, qRT-PCR validated that this gene was significantly upregulated in both sets of R72 NHFs, but was not induced in either the P72 NHF cell line following gamma radiation (P < 0.05 comparing P72 vs. R72, Fig. 1C).

To validate the above findings, we next tested three other cell line model systems for the preferential induction of TRIML2 by the R72 variant of p53. One of these model systems consisted of H1299 (p53 null non–small cell lung carcinoma, NSCLC) cells containing a doxycycline-inducible version of p53 encoding either the P72 or R72 variant (11). In these cells, the P72 variant showed some ability to upregulate TRIML2 following etoposide treatment, but the R72 variant reproducibly showed significantly increased ability to induce TRIML2 (P < 0.05 comparing P72 vs. R72, Fig. 1D). This difference was evident at the protein level as well (Fig. 1D, bottom). Similar findings were made in Saos2 (human osteosarcoma) cells containing temperature-sensitive versions of the P72 and R72 alleles (Supplementary Fig. S2B). Moreover, in mouse embryonic fibroblasts (MEF) from the P72 and R72 Hupki mouse, the upregulation of Triml2 in response to etoposide was significantly greater in R72 MEFs compared with P72 (P < 0.05, Fig. 1E). Finally, increased induction of Triml2 by the R72 variant was also seen in the small intestine from Hupki mice treated with gamma radiation (P < 0.05, Fig. 1F); in this tissue, we have previously seen increased apoptosis in R72 mice, compared to P72 (18). The combined data indicate that in five different cell line systems, three with endogenous p53 and two with inducible versions, the TRIML2 gene is consistently preferentially transactivated by the R72 variant.

To determine whether TRIML2 is a direct p53 target gene, Hct116 human colorectal cancer cells, and the somatic cell knockout counterpart Hct116 p53^−/− were treated with etoposide for 24 hours, and the RNA and protein level of TRIML2 was assessed. This analysis revealed that TRIML2 is induced by etoposide at both the RNA and protein levels, and that this induction is significantly (but not completely) reduced in p53^−/− cells (Fig. 2A). We also analyzed TRIML2 induction in response to other stresses. p53-dependent induction of TRIML2 is evident in Hct116 cells in response to 5-FU treatment and following UV irradiation (Supplementary Fig. S2C). Analysis of the TRIML2 promoter and intronic regions revealed the presence of three consensus p53-binding sites (Fig. 2B). To test the possibility that p53 could bind to these, we performed ChIP analysis using p53 antibody in Hct116 cells treated with 5-FU. As positive controls for p53 binding, we analyzed the known p53 consensus elements in the CDKN1A (p21) and MDM2 target genes. ChIP analysis of the three p53 consensus elements in the TRIML2 gene revealed that two of these sites, both located in the TRIML2 promoter (−1,439 and −2,855, relative to the start site of transcription, +1), were consistently immunoprecipitated with p53 antibody, whereas the third site was not (Fig. 2B). We next used ChIP to test the ability of P72 and R72 proteins to bind to the TRIML2 promoter. For this experiment, we used H1299 cells expressing doxycycline-inducible versions of P72 and R72, as these express equivalent levels of both p53 variants (Fig. 2C). As controls, we analyzed the interaction of these variants with the CDKN1A (p21) promoter, which we and others showed demonstrates enhanced binding by P72 protein (19), and the p53 consensus element in the BTG2 gene, which according to our microarray was induced to identical levels in P72 and R72 cells (Supplementary Fig. S1B and data not shown). As expected, these analyses revealed that the P72 and R72 protein bound identically to the BTG2 target gene, and that the P72 variant bound to greater extent to the CDKN1A (p21) consensus site (Fig. 2C). Interestingly, however, the R72 protein immunoprecipitated both RE1 and RE2 of the TRIML2 promoter significantly more efficiently than P72 (Fig. 2C). These data suggest that the R72 variant preferentially transactivates the TRIML2 gene, at least in part, due to enhanced binding of the R72 protein to the promoter elements in this gene.

To assess the impact of TRIML2 on downstream p53 functions, we next sought to silence this gene in cells with WT p53, and assess the impact of this silencing on p53-mediated programmed cell death and transactivation. Toward this goal, we first tested three different short hairpins for their ability to silence TRIML2. We identified one (sh-A, see Materials and Methods) that routinely led to the most efficient silencing (Fig. 3A). Lentiviral transduction and selection for this short hairpin into Hct116 cells led to the isolation of several clones with silenced TRIML2, and this silencing did not appear to markedly affect cell behavior (data not shown). As expected, treatment of these clones with etoposide led to decreased induction of TRIML2 protein, compared with vector-transduced controls; interestingly, however, this also led to decreased steady-state levels of p53 (Fig. 3A, bottom). These data suggest that TRIML2 positively regulates p53 protein levels.

We next extended these analyses to include an assessment of programmed cell death in TRIML2 silenced cells. In Hct116 cells treated with 5-FU, knockdown of TRIML2 led to decreased steady-state level of p53, as well as decreased levels of cleaved lamin A and cleaved caspase-3 at late time points (72 hours), indicating that apoptosis was impaired (Fig. 3B). Similar findings were made when a different short hairpin (sh-B) to TRIML2 was used, indicating that this was not an off-target effect (Supplementary Fig. S2D). A similar decrease in apoptosis was evident in U2OS cells with TRIML2 silenced (Supplementary Fig. S2E). Annexin V staining and cell viability assays using flow cytometry confirmed that TRIML2 knockdown results in significant inhibition of 5-FU–induced apoptosis and increased viability of Hct116 cells after 72 hours (Fig. 3B, top and bottom right). We next sought to test the impact of overexpression of TRIML2 on p53 stabilization and apoptosis. Toward this end, we created a doxycycline-inducible TRIML2 cell line, in the background of U2OS cells, which contain WT p53. Doxycycline treatment of these but not parental cells followed by etoposide revealed that overexpression of TRIML2 protein led to increased levels of p53, along with increased induction of the apoptosis markers cleaved lamin A and cleaved caspase-3 (Fig. 3C). This increase in apoptosis was particularly evident when apoptosis was assayed by Annexin V staining (Fig. 3C, right). Similar findings were made in Hct116 cells stably transfected with TRIML2 (Supplementary Fig. S2F). The combined data indicate that TRIML2 positively regulates p53 levels, and can also enhance p53-mediated apoptosis.

We next sought to test the hypothesis that the increased induction of TRIML2 in cells containing the R72 variant of p53 might contribute to the higher apoptotic potential of this variant, relative to P72. To test this hypothesis, we generated subclones of tetracycline-inducible P72 and R72 cells, each stably transfected with parental vector or a CMV (cytomegalovirus) promoter–driven construct of TRIML2. Analyses of these cells revealed that overexpression of TRIML2 in tetracycline-inducible P72 cells resulted in increased apoptosis (indicated by cleaved PARP and cleaved caspase-3), to levels seen in R72 control cells (Supplementary Fig. S3A). Similarly, we found that silencing TRIML2 in R72-inducible cells led to decreased apoptosis (indicated by cleaved PARP and cleaved lamin A), to levels nearly identical to that in P72 cells infected with the short hairpin control (Supplementary Fig. S3B). These findings support the premise that the increased upregulation of TRIML2 in R72 cells contributes significantly to the increased apoptotic potential of this variant, relative to P72.

We next analyzed the impact of TRIML2 on p53-mediated transactivation of target gene expression. Toward this goal, we used a qRT-PCR approach to analyze the expression levels of p53 target genes associated with growth arrest and apoptosis in Hct116 cells expressing vector alone or the short hairpin for TRIML2, in the presence and absence of 5-FU. Analysis of over two dozen p53 target genes revealed that, despite causing a decrease in steady state p53 levels, silencing of TRIML2 did not affect the transactivation of the majority of p53 target genes analyzed, and actually led to increased ability of p53 to transactivate the growth arrest target genes CDKN1A (p21), GADD45, and CYCLIN G1 (Fig. 4A). Conversely, at the time point analyzed (5-FU treatment for 72 hours), there was no impact of TRIML2 silencing on the transactivation of the proapoptotic genes BBC3 (PUMA), BAX, DRAM1, or EI24, with the exception of PMAIP1 (NOXA; Fig. 4A). Interestingly, however, TRIML2 silencing significantly impaired the transactivation of a small subset of p53 proapoptotic target genes, consisting of the prodeath genes PIDD (PIDD1), PIG3 (TP53I3), and PIG6 (PRODH). Taken together, these findings implicate TRIML2 in the growth arrest versus cell death decision by p53, as TRIML2 silencing resulted in upregulation of growth arrest genes (CDKN1A, GADD45, and CYCLIN G1), and decreased expression of three p53 proapoptotic genes (PIDD, PIG3, and PIG6).

To extend our findings, we expanded our qRT-PCR analysis to include several dozen p53 target genes (Supplementary Fig. S4A and S4B). This analysis was performed using RNA isolated from Hct116 cells with TRIML2 silenced compared with vector control, in the absence or presence of 5-FU. This analysis confirmed that TRIML2 silencing led to reduced expression of not only proapoptotic p53 target genes PIDD, PIG3, and PIG6, but also several p53 metabolism target genes, such as GLS2, GLUT1 (SLC2A1), and TIGAR (C12orf5), as well as two genes that require the p53 transactivation domain II (TA II) for transactivation, CRIP2 and KANK3 (Supplementary Fig. S4C–S4E; ref. 20). We next sought to confirm these findings by analyzing the expression of these p53 target genes in the U2OS cell line containing a tetracycline-inducible version of TRIML2. qRT-PCR analysis confirmed that the p53 proapoptotic genes PIDD, PIG3, and PIG6 showed increased expression in etoposide-treated cells when TRIML2 was induced with doxycycline (Fig. 4B). Similar increases in gene expression were also seen in Hct116 cells stably transfected with TRIML2 following 5-FU treatment (Supplementary Fig. S5A). In sum, in two different cell types (U2OS and Hct116), using TRIML2 overexpression and silencing, we were able to detect the same set of genes affected by TRIML2.

The protein product of the p53 target gene PIDD regulates p53-mediated apoptosis through the formation of the PIDDosome. The PIDDosome consists of PIDD, RAIDD, and caspase-2; this complex controls the cleavage and activation of caspase-2 en route to apoptotic cell death (21). We reasoned that altered transactivation of PIDD following TRIML2 overexpression or knockdown might be accompanied by differences in the level of cleaved caspase-2. To test this premise, we analyzed Hct116 cells infected with short hairpin control or TRIML2 short hairpin for the presence of activated (cleaved) caspase-2 following p53 induction. We also analyzed cleaved caspase-2 levels in U2OS cells that overexpress TRIML2. We found that silencing of TRIML2 in Hct116 cells led to significantly decreased level of cleaved caspase-2 following 5-FU treatment (Fig. 4C). Furthermore, overexpression of TRIML2 in doxycycline-inducible U2OS cells led to significantly increased PIDD protein, along with increased cleaved caspase-2 upon etoposide treatment (Fig. 4D). These data support the premise that the alterations of PIDD level following silencing or overexpression of TRIML2 could lead to the observed effects on apoptosis, as manifested by cleavage and activation of caspase-2.

The impact of TRIML2 on the expression of target genes like PIDD suggests that some of these target genes might show increased expression in cells expressing the R72 variant of p53. To test this premise, we performed qRT-PCR on PIDD and PIG3, compared with BAX and BBC3 (PUMA), in H1299 cells that express tetracycline-inducible versions of P72 and R72 forms of p53. These analyses revealed similar transactivation of BAX and BBC3 (PUMA) in P72 and R72 tet-inducible cells, but significantly increased transactivation of both PIG3 and PIDD in R72 cells (Fig. 4E; PIG6 was not expressed in this cell line). Increased PIDD level was also evident in temperature-sensitive Saos2-R72 cells, compared with P72 (Supplementary Fig. S5B). The combined data suggest that the R72 variant of p53 possesses increased apoptotic potential due in part to increased ability to transactivate TRIML2, along with enhanced transactivation of genes like PIG3 and PIDD.

Our data indicate that TRIML2 knockdown reduces apoptosis following p53 induction, but that this was evident only following prolonged periods of p53 induction (72 hours, Fig. 3B). To further explore this finding, we performed a time course analysis of apoptosis and p53 target gene expression in Hct116 cells infected with control vector or TRIML2 short hairpin; in this case, we used treatment with 5-FU as the genotoxic stress, as this agent has a longer half-life in cell culture medium compared with other cytotoxic agents (22). qRT-PCR analysis revealed that cells with TRIML2 silenced showed a significant decrease in PIG3, PIG6, and PIDD, compared with cells infected with vector alone, as early as 24 hours following 5-FU treatment (Fig. 5A and B). However, the changes in PIG3, PIG6, and PIDD following TRIML2 silencing were most evident at later time points (72 and 96 hours, Fig. 5B). Similarly, cells expressing TRIML2 short hairpin showed decreased apoptosis at 24 and 72 hours, but this difference was most pronounced at 96 hours following DNA damage (Fig. 5C). In contrast, the cell-cycle arrest genes CDKN1A (p21) and CYCLIN G1 are not affected at 48 hours and become inversely correlated with TRIML2 expression only at later time points (72 and 96 hours, Fig. 5A). These combined data all support the premise that TRIML2 has the most pronounced effects on p53 target expression at late time points of p53 induction.

TRIML2 is a member of the tripartite-motif (TRIM) family of proteins, the majority of which are reported to possess ubiquitin and/or SUMO-ligase activities (23). As p53 can be modified by ubiquitin, SUMO-1, and SUMO-2/3, we next sought to test the hypothesis that TRIML2 might control p53 function by specifically interacting with, and/or posttranslationally modifying p53. Because of the lack of availability of a TRIML2 antibody that works in immunoprecipitation, we performed these assays in Hct116 cells engineered to stably express a V5-tagged form of TRIML2, and we monitored a TRIML2–p53 interaction in these cells following exposure to etoposide. p53 was readily immunoprecipitated with TRIML2 by V5 antibody following DNA damage in stably transfected cells with endogenous p53 (Fig. 6A). We found no evidence for the ubiquitin ligase MDM2 in this complex. We next monitored the subcellular localization of p53 and TRIML2. Toward this end, we performed immunofluorescence of p53 and TRIML2 in transfected cells, and found that TRIML2 appears to localize to both the cytoplasm and nucleus of cells, and that the nuclear-localized TRIML2 showed significant colocalization with p53 (Supplementary Fig. S6A). These combined findings indicate that p53 and TRIML2 can be associated in a complex together, and that TRIML2 might posttranslationally modify p53.

We next chose to determine whether TRIML2 might enhance the ubiquitylation or SUMOylation of p53. To test this possibility, we transfected TRIML2 along with WT p53 and either tagged ubiquitin or tagged SUMO-1 into H1299 cells. We then analyzed the ubiquitylation and SUMOylation patterns of p53 in these cells; as positive controls, we analyzed p53 ubiquitylation and SUMOylation following transfection with TRIM27 and PML, which are known to regulate these processes (14). In these experiments, we failed to detect evidence for increased ubiquitylation or SUMOylation of p53 in TRIML2-transfected cells (Supplementary Fig. S6B and S6C). As p53 can also be modified with SUMO-2/3 (24, 25), and because some TRIM proteins are reported to be SUMO-2/3 ligases (14), we next tested whether TRIML2 might modify p53 with SUMO-2/3. Toward this end, we transfected H1299 cells with p53, TRIML2, and SUMO-2, and analyzed p53 by Western blot analysis, looking for increases in the higher molecular weight species that would be indicative of modification with SUMO-2. In these experiments Western blot analysis for p53 revealed a readily apparent upward shift in p53 mobility following transfection with TRIML2, similar to what was seen following transfection with the positive controls TRIM27 and PML; these findings were consistent with modification of p53 with SUMO-2 (Fig. 6B). To further confirm TRIML2-mediated SUMO-2 sumoylation of p53 in H1299 cells, we used an inhibitor of sumoylation, ginkgolic acid, to specifically inhibit this process (26). In this experiment, the higher molecular weight species of p53 that were induced by TRIML2 expression were decreased by addition of the inhibitor ginkgolic acid, supporting that they are likely SUMO-2–modified versions of p53 (Fig. 6C). We next sought to confirm this by immunoprecipitating p53 in TRIML2-transfected cells, followed by Western blot analysis using SUMO-2 antibody. This experiment confirmed that the approximately 70 and 130 kDa species of p53 induced by TRIML2 transfection were in fact p53 modified by SUMO-2 (Fig. 6D). These data also suggest that TRIML2 may induce poly-SUMOylation of p53 with SUMO-2; this feature is distinct from SUMO-1, which can be used only to mono-SUMOylate p53 (27, 28).

This report marks the first attempt to identify codon 72 allele-specific target genes in normal, nontransformed cells. We identified nearly a 100 genes whose expression was altered significantly in a p53-dependent manner after gamma-irradiation treatment of P72 or R72 NHFs (Supplementary Fig. S1). A small subset of these showed differential regulation by the P72 and R72 variants. The majority of these differentially regulated genes were preferentially transactivated by the P72 variant, and several of these genes were previously reported by our group (12), including CSF-1, which we showed participates in a feed forward loop to enhance p53-mediated growth arrest in P72 cells (29). The only gene identified in this study as being preferentially induced by R72 is TRIML2. We show that this gene is preferentially induced in cells containing the R72 variant in five different cell line systems, of both mouse and human origin. Notably, in a recent study of hemizygous p53-knockout RKO colon carcinoma cells containing either R72 or P72 alleles, several codon 72 allele-specific transcripts were identified following etoposide treatment; this group also reported TRIML2 as an R72 preferentially induced p53 target gene (30). We show here-in that this preferential transactivation by R72 is the result of increased ability of this variant to bind to the TRIML2 promoter. How the codon 72 polymorphism, which is located proximal but not within the DNA-binding domain of p53, affects the ability of p53 to bind to different response elements in promoters remains to be determined.

TRIML2 is a member of a large TRIM protein family that has more than 70 members in humans to date (23). TRIM protein family members are involved in diverse biologic functions, including IFN-induced antiviral and antimicrobial activities (31); recent evidence also implicates TRIM proteins in transcriptional regulation, apoptosis, and cancer (23). There are no published studies on the TRIM family member described here, TRIML2. However, a recent study showed that many TRIM proteins function as ligases for SUMO-1, -2, or -3 (the latter two are 95% identical, so are often referred to as SUMO-2/3; ref. 14). In addition, a recent study identified TRIML2 as part of the SUMO-2–containing complex in the nucleus of HeLa cells (32). TRIML2 is not the first p53-induced TRIM, nor is it the first TRIM that affects p53 modification and function. Four other TRIM genes are known to be transactivated by p53, including TRIM8, TRIM19 (PML), TRIM22, and TRIM24 (33–36). qRT-PCR data in our cell lines did not indicate that any of these other TRIMs show clear preferential transactivation by codon 72 variants of p53 (Supplementary Fig. S6D). Interestingly, whereas we found that TRIML2 enhances p53′s proapoptotic gene expression and apoptosis and diminishes transactivation of growth arrest genes, TRIM8 does the opposite: It facilitates p53-mediated cell-cycle arrest by selectively upregulating expression of CDKN1A(p21) and GADD45, and decreases transactivation of apoptosis-associated p53 targets (34). These data suggest that TRIMs may be used by the cell to fine-tune the growth arrest versus cell death decision by p53, and indeed that different TRIMs, some of which can heterooligomerize, may be able to antagonize each other to achieve this function.

Other TRIMs also have been shown to posttranslationally modify p53. TRIM8, TRIM19 (PML), and TRIM13 can increase p53 stability and function by inhibiting MDM2-mediated p53 degradation (34, 37, 38). TRIM24, TRIM28 and TRIM29 (also known as ataxia-telangiectasia group D complementing, or ATDC), and TRIM39 all inhibit p53 activity by increasing p53 degradation and/or nuclear export, largely by controlling p53 ubiquitylation (39–42). Finally, a recent study showed that TRIM21 inhibits p53 functions indirectly by blocking the formation of the deubiquitylation complex GMPS–USP7 (43). In our hands, TRIML2 did not appear to alter ubiquitylation of p53, but rather caused increases in p53 modification with SUMO-2/3. Little is known about modification of p53 with SUMO-2/3. In general, modification of p53 with SUMO-1 is believed to inhibit the DNA-binding activity and transcriptional function of p53 (44), but the available data suggest that modification with SUMO-2/3 is different. Specifically, conjugation of p53 with SUMO-2/3 has been seen on lysine residue (K386), and shown to selectively inhibit the transactivation of a subset of p53 transcriptional targets, and possibly to activate other p53 target genes (24, 25). Our findings support the hypothesis that SUMO-2 modification may differentially affect different p53 target genes: We show that TRIML2 increases the transactivation of a subset of proapoptotic target genes (PIG3, PIG6, and PIDD) and at the same time decreases the transactivation of growth arrest genes (CDKN1A, GADD45, and CYCLIN G1). How modification with SUMO-2/3 affects p53 target gene transactivation remains to be determined. The combined data support a model wherein the R72 variant, through increased transactivation of TRIML2, then shows enriched modification with SUMO-2/3 and increased ability to transactivate a subset of proapoptotic p53 target genes, thus leading to increased ability to induce apoptosis (Fig. 7).

Our data indicate that TRIML2 should be added to the growing list of proteins that control the growth arrest versus cell death decision by p53. Our data also support the hypothesis that increased ability of the R72 variant to transactivate TRIML2 explains part of the increased ability of this variant to induce programmed cell death, relative to P72. Interestingly, the collected cancer microarray data from the ONCOMINE database reveals that TRIML2 is frequently downregulated in a variety of cancers, including gastric, colorectal, lung, and liver cancers (data not shown), suggesting that TRIML2 may have anticancer or antiproliferative function. As the P72 variant of p53 has been identified as a risk factor for these same cancers (45–48), these findings provide a potential link between the failure of the P72 variant to induce TRIML2 and its reduced ability to suppress tumor development in these tissues.

No potential conflicts of interest were disclosed.

Conception and design: C.-P. Kung, M.E. Murphy

Development of methodology: C.-P. Kung

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): C.-P. Kung, M. Jennis

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): C.-P. Kung, M. Jennis, Y. Zhou, M.E. Murphy

Writing, review, and/or revision of the manuscript: C.-P. Kung, S. Khaku, M. Jennis, M.E. Murphy

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): C.-P. Kung, S. Khaku, Y. Zhou, M.E. Murphy

Study supervision: M.E. Murphy

The authors thank members of the Murphy laboratory for helpful discussions and review of the article. The authors thank Frederick Keeney and James Hayden of the Imaging Facility (The Wistar Institute) and Yue-Sheng Li of the Microarray Facility at Fox Chase Cancer Center.

This research was funded by R01 CA102184 (M.E. Murphy).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.