|Home | About | Journals | Submit | Contact Us | Français|
NF-κB activation may play an important role in the pathogenesis of cancer and also in resistance to treatment. Inactivation of the p53 tumor suppressor is a key component of the multi-step evolution of most cancers. Links between the NF-κB and p53 pathways are under intense investigation. In this study, we show that the receptor interacting protein (RIP, RIP1), a central component of the NF-κB signaling network, negatively regulates p53 tumor suppressor signaling. Loss of RIP1 from cells results in augmented induction of p53 in response to DNA damage, while increased RIP1 level leads to a complete shutdown of DNA-damage induced p53 induction by enhancing levels of cellular mdm2. The key signal generated by RIP1 to upregulate mdm2 and inhibit p53 is activation of NF-κB. The clinical implication of this finding is demonstrated in glioblastoma (GBM), the most common primary malignant brain tumor in adults. We show that RIP1 is commonly overexpressed in GBM but not in grade II-III glioma and increased expression of RIP1 confers a worse prognosis in GBM. Importantly, RIP1 levels correlate strongly with mdm2 levels in GBM. Our results demonstrate a key interaction between the NF-κB and p53 pathways that may have implications for the targeted treatment of GBM.
A link between chronic inflammation and cancer has been suspected for over a century (1). A major link between inflammation and cancer is mediated by activation of NF-κB (2-4). NF-κB activation is a strong pro-survival signal (5). Constitutive and deregulated activation of NF-κB is widespread in human cancer (3, 6) and promotes survival of tumor cells and resistance to treatment (7, 8). Furthermore, experimental models support a causal role for NF-κB activation in inflammation induced cancer (9, 10). Cross-talk between stress induced/inflammatory responses and oncogenic signaling pathways is likely to play an important role in cancer. A number of studies have linked components of the NF-κB signaling pathway to cell cycle progression and tumorigenesis (11-16).
An intriguing mechanism underlying the pathogenesis of inflammation induced cancer is negative regulation of tumor suppressor pathways by inflammatory and stress-induced signals. p53 is a key tumor suppressor altered in a broad range of human cancers including glioma and an important outcome of p53 activation is cell cycle arrest or apoptosis following DNA damage (17, 18). Previous studies have documented links between the NF-κB and p53 networks that have largely been reported to be antagonistic (19) but may also be synergistic (20). There is evidence that components of the NF-κB signaling network interact with p53 at multiple levels. For example, IKK2 (IKKβ) inhibits p53 induction in response to chemotherapeutic drugs via an upregulation of mdm2 (21), while IKK1 (IKKα) interferes with p53-mediated gene transcription by inducing CBP phosphorylation (22). Thus, NF-κB pathway-mediated inhibition of p53 function may promote the pathogenesis of cancer.
The death domain-containing kinase receptor interacting protein (RIP1, RIPK1) is an essential component of the signaling cascade that activates NF-κB in response to cellular stress and inflammation (23, 24). Thus, RIP1 is required for TNFα-induced and DNA damage-induced NF-κB activation (25-28). In addition, RIP1 is also a key component of innate immunity, essential for TLR-3 mediated activation of NF-κB (29). RIP1 is composed of kinase, intermediate, and death domains. RIP1 is involved in the activation of the IKKs via a kinase-independent mechanism (30). Thus, RIP1 appears to function as an adaptor, and it is the intermediate domain of RIP1 that is essential for NF-κB activation.
Glioblastoma (glioblastoma multiforme, GBM) is the most common primary malignant brain tumor in adults and is resistant to treatment (31). The median survival of GBM patients with radiation and chemotherapy was recently noted to be 14.6 months (32). The molecular pathogenesis of GBM includes genetic alterations in pathways mediating proliferation, apoptosis and cell cycle control (33, 34). Inflammatory responses can be readily detected in glioma in the form of infiltrating macrophages/microglia and lymphocytes, production of inflammatory cytokines, and activation of NF-κB (35, 36). Importantly, NF-κB activation may be linked to the resistance of glioblastoma cells to O6-alkylating agents (37, 38).
In this study, we demonstrate that RIP1 negatively regulates p53 induction in response to DNA damage. RIP1 regulates p53 via the upregulation of mdm2 levels. We also elucidate a key role for NF-κB activation in RIP1 mediated regulation of mdm2 and p53 downregulation. Analysis of glioma patient samples demonstrated that RIP1 is overexpressed in about 30% of GBM (grade IV), but not in grade II-III glioma. Importantly, there was a striking correlation between RIP1 and mdm2 levels in GBM, consistent with our mechanistic data in vitro. Finally, high RIP1 level is an independent negative prognostic indicator in GBM.
RIP1 plasmid was obtained from Dr. Brian Seed (Boston MA) and cloned into pcDNA3.1 with a C-terminal FLAG tag using standard molecular techniques. rip1+/+ and rip1-/-murine embryo fibroblasts were provided by Dr. Michelle Kelliher (27). In this study we used primary MEFs at passage 3-4. A p21-LUC promoter was provided by Dr. Bert Vogelstein.
p53 (DO1, sc-126), IκBα (sc-371), p65 (sc-109) and ERK2 (sc-154) antibodies were obtained from Santa Cruz Biotechnology (Santa Cruz, CA). RIP (#610459) antibody were purchased from BD Biosciences (San Diego, CA) and was used for both IHC and Western blotting. mdm2 (2A10) antibodies were purchased from Calbiochem (San Diego, CA). p53 (1C12), p21(#2947) and FLAG (#2368) were purchased from Cell Signaling Technology (Danvers, MA). We used p53 DO-1 antibodies for detection of p53 in (human) glioma cell lines while p53 1C12 antibodies were used for detection of (mouse) p53 in MEFs. For tumor samples, 30 μg of protein was loaded on to 7.5% polyacrylamide gels and Western blotting was performed according to standard protocols.
U87MG cells (70K) were plated in 24 well dishes followed by transfection with transfected with either p21-LUC along with RIP plasmid or empty vector, using calcium phosphate. A dual-luciferase reporter assay system was used according to the instructions of the manufacturer (Promega, Madison WI). Firefly luciferase activity was measured in a luminometer and normalized on the basis of Renilla luciferase activity. Experiments were done in triplicate and repeated twice.
Preparation of adenovirus expressing RIP1 has been described previously (39). An MOI of 10-100 was used in the experiments. Adenovirus-null or Adenovirus-IκBαM were obtained from Vector Biolabs (Philadelphia, PA) and used at an MOI of 50.
All studies were conducted according to IRB approved protocols at UTSW. Frozen tissue specimens of human gliomas were stored following resection. Mirror tissue was processed for paraffin sections and examined to ensure that the frozen sample is tumor. Frozen samples available were analyzed by Western blotting. Paraffin-embedded sections were cut at four-micrometer thickness and deparaffinized. The sections were incubated in 3% H2O2 for 10 min to block endogenous peroxidase activity. For antigen retrieval, the sections were placed in a Pascal programmable pressure cooker (DakoCytomation, Carpinteria, CA) containing EDTA solution (Biocare Medical, Walnut Creek, CA), with the target temperature and time set to 125° C for 30 sec. The sections were exposed to the mouse monoclonal RIP antibody (#610458; BD Biosciences, San Diego, CA) at a dilution of 1:50 for 30 min. The detection system was Envision Dual Link (DakoCytomation). The sections were counterstained lightly with hematoxylin. The IHC staining intensity for RIP1 was scored semiquantitatively as 0, + (1), ++ (2), and +++ (3). The final IHC score was derived from IHC staining intensity multiplied by the percentage of tumor cells showing positive staining.
RIP1 siRNA were obtained from Dharmacon (Chicago, IL). Three sequences directed against the RIP1 mRNA were used (D-004445-03, RIPK1, D-004445-04, RIPK1, and D-004445-06, RIPK1). A negative control siRNA (non sequence, Silencer Negative Control #1 siRNA, AM4611) was obtained from Ambion (Austin, TX). siRNA was performed according to the manufacterer's protocol using Lipofectamine 2000 reagent (Invitrogen). Lysates were prepared 72h after siRNA for Western blot.
Error bars represent the means ± standard deviations of three independent experiments. The RIP1 level was dichotomized into high and low groups, based on a RIP1 value of 1.2 which represents a 50% increase over the levels found in normal brain. To investigate whether RIP1 level (Ratio between RIP1 and ERK2) is associated with the clinical outcome, we applied a proportional hazard model using both overall survival time and progression free survival time as the outcomes (censored by the last known number of follow up months). The multivariate proportional hazard model was used to analyze the effect of additional prognosis factors, such as age and Karnofsky scores, with the RIP1 level. The survival curves were fitted using the Kaplan-Meier method. The log-rank test was used to compare different survival curves. Fisher's exact test was used to calculate the p-value for association between RIP1 and mdm2 in GBM. The survival analysis was performed using survival package in SAS software.
Previous studies have documented interactions between components of the NF-κB signaling network and p53. Thus, we investigated whether the cellular level of RIP1 could influence p53 induction in response to DNA damage. Exposure of RIP1-/- primary mouse embryo fibroblasts (MEFs) to ionizing radiation (IR) resulted in a significantly greater induction of p53 compared to wild type cells (Figure 1A). Similarly, exposure of RIP1-/- MEFs to UV radiation elicited a greater induction of p53 (Supplemental Figure 1A). Next, we investigated the effect of increased RIP1 expression in the human glioblastoma cell line U87MG that expresses wild type p53. To overexpress RIP1, we used a tetracycline-inducible adenoviral expression vector (tet-off). An empty adenoviral vector (Ad-null) was used as an additional control (Supplemental Figure 1C). Cells were exposed to IR (10 Gy) followed by Western blot. While p53 and p21 are induced in response to IR in control cells, overexpression of RIP1 completely blocked IR-mediated induction of p53 and p21 in these cells (Figure 1B). Indeed, increased expression of RIP1 also downregulates the steady state levels of p53 in these cells. Similarly, in A172 cells, a second glioblastoma cell line with wild type p53, increased RIP1 expression inhibited basal and IR-mediated induction of p53 and p21 (Supplemental Figure 1B). RIP1 also inhibited p53 induction in response to UV radiation (Supplemental Figure 1D).
siRNA knockdown of RIP1 resulted in increased basal level of p53 and p21 levels (Figure 1C). Furthermore, RIP1 silencing results in an increase in p53 induction in response to IR (Supplemental Figure 1E). We used three different siRNA sequences directed against different parts of the RIP1 mRNA with similar results. Similarly, siRNA knockdown of RIP1 in A172 cells increased p53 and p21 levels (Figure 2D).
Next, we investigated whether the kinase activity of RIP1 was required to regulate p53. We introduced wild type RIP1 and mutant RIP1 with either the kinase (DKD) or intermediate domain (DID) deleted into U87MG cells. We find that the DKD mutant efficiently downregulated p53 while the DID mutant did not (Figure 1D). Loss of the intermediate domain increases the levels of the RIP1 protein dramatically and RIP1 DID fails to downregulate p53 even though it is expressed at higher levels. When a lower MOI is used for the DID to equalize expression with DKD, DID fails again to downregulate p53 (Supplemental Figure 2A).
Since RIP1 is commonly overexpressed in human glioblastoma (see below), it is important to note that the levels of RIP1 overexpression in these experiments were similar to that detected in human GBM (Figure 4D).
We also tested the effect of RIP1 on the transcriptional activity of p53 in U87MG cells. In this experiment we tested the ability of IR to activate a p53-responsive promoter (p21-LUC). We find that co-expression of RIP1 with p21-LUC completely blocked the basal and the IR-induced p53 induced transcriptional activity of the p21 promoter (Supplemental Figure 1F).
mdm2 is a major regulator of p53 and plays a key role in promoting ubiquitination and degradation of p53. We investigated whether RIP1 regulated levels of cellular mdm2. Indeed, increased expression of RIP1 resulted in the upregulation of mdm2 in U87MG cells (Figure 2A). Conversely, silencing RIP1 in U87MG cells resulted in a decrease in levels of cellular mdm2 (Figure 2B). Similarly, increased expression of RIP1 in A172 cells resulted in increased mdm2 levels (Figure 2C) while silencing RIP1 downregulated mdm2 levels (Figure 2D).
Previous studies have documented that increased expression of RIP1 is sufficient to activate NF-κB in various cell types (40, 41). First, we confirmed that NF-κB was activated by increased RIP1 expression in U87MG cells, as shown in other cells. We tested the ability of increased RIP1 to downregulate levels of IKBα. As expected, increased expression of RIP1 in U87MG cells resulted in a complete and persistent downregulation of IKBα as shown in Figure 3A. Furthermore, increased expression of RIP1 was sufficient to increase the transcriptional activity of NF-κB in reporter assays in U87MG cells (Figure 3B).
To test whether NF-κB activation played a role in RIP1-mediated regulation of p53, we investigated whether introduction of a dominant-negative IκBα mutant (IκBαM, S32/S36) that inhibits NF-kappa B activation would rescue RIP1 mediated suppression of p53 induction. Expression of IκBαM resulted in a substantial reversal of the RIP1 mediated block of p53 induction following ionizing radiation (Figure 3C). A similar result was found in A172 cells (Supplemental Figure 2B). Next, we investigated whether RIP1 mediated upregulation of mdm2 required NF-κB activation, we expressed IκBαM together with RIP1 in U87MG cells. Inhibition of NF-κB activity also blocked the ability of RIP1 to upregulate mdm2 levels in U87MG and A172 cells (Figures 3D and Supplemental Figure 2C). As expected, RIP1 did not influence levels of the p65 subunit of NF-κB.
Previous studies have demonstrated that NF-κB is activated in GBM. Since RIP1 plays a key role in NF-κB activation, we investigated RIP1 levels in glioma. The RIP1 antibody we used has been utilized in previous studies (30, 42-45). Gliomas are graded I-IV in order of increasing malignancy and GBM (grade IV) is the most malignant. We examined 23 gliomas graded II-III, 70 glioblastomas (grade IV, GBM) and one non-tumor brain sample. About 30% of GBMs show increased levels of RIP1, while RIP1 level among the gliomas graded II-III is low or absent (Figure 4A and in Supplemental Table 1 and 2). Lysates were made directly from resected brain tumors, RIP1 level was detected by Western blots, quantitated by densitometry, and values normalized for loading using total ERK2 levels. ERK2 is expressed ubiquitously and we find its level to be stable in glioma. RIP1 levels in GBMs were 1.01 ± 0.15, and levels of RIP1 in Non GBMs (grade II-III gliomas) was 0.42 ± 0.07. The log-transformation was applied to the RIP1 level and the student t-test showed that the RIP1 level in the GBM group was significantly higher than the non GBM group (p=0.03) (Figures 4A and C). We also conducted immunohistochemistry (IHC) in randomly selected tumors with the same RIP1 monoclonal antibody to test the correlation between Western blot and IHC. The IHC score was significantly associated with the RIP1 expression by Western blotting (Supplemental Table 3). Figure 4B shows levels of RIP1 in normal brain (NB), an anaplastic astrocytoma (AA), and intense RIP1 staining in a GBM. Thus, RIP1 levels correlated well with increased malignancy in glioma. RIP1 is expressed mostly in the cytoplasm of tumor cells but we also detect expression in endothelial cells (Fig. 4B and Supplemental Figure 3A-B).
In addition, we examined RIP1 levels in four matched pairs of primary low grade glioma that progressed to secondary GBM. In each case, we found that RIP1 was low in the primary tumor and significantly increased in the secondary GBM (Supplemental Figure 3C and Supplemental table 4).
Next, we investigated whether there is a correlation between RIP1 and mdm2 levels in GBM. We tested the level of mdm2 by Western blot in 31 GBMs as shown in Figure 5. mdm2 was detectable by Western blot in 11/31 GBM cases. The detection rate of mdm2 in high RIP1 group was 73% and 0% in low RIP1 group (p value less than 0.0001). Thus RIP1 levels correlate strongly with mdm2 in GBM, suggesting that increased RIP1 may upregulate mdm2 levels in this disease. We further explored the association between p53 and mdm2 levels. The detection rate of mdm2 in low p53 group was 75% and 6% in high p53 group (p-value=0.0002). Thus, p53 levels correlate inversely with mdm2 levels in our sample of GBMs.
Next, we investigated whether RIP1 expression influenced prognosis in GBM. The clinical outcomes measured were overall survival and progression-free survival (n=70). Survival time was defined as the time from diagnosis to death (or disease progression), which was censored by the last known number of follow up days. We dichotomized the RIP1 level based on a RIP1/ERK2 level of 1.2 which represents a 50% increase over the levels found in normal brain (figure 4B, densitometry data not shown). By this criterion none of the grade II-III tumors overexpressed RIP1 (Figure 4A and Supplemental Table 2).
In the high RIP1 group, there were 10 out of 20 patients dead; in the low RIP1 group, there were 18 out of 50 patients dead within the follow up months. Figure 6A shows the Kaplan-Meier overall survival curves for the high and low RIP1 groups, which shows that patients with high RIP1 levels have significantly shorter survival than patients with low RIP1 levels (p=0.014, log-rank test); the median survival time was 12.4 months for the high RIP1 group and 26.3 months for the low RIP1 group. The univariate proportional hazard model showed the dichotomized RIP1 was significantly associated with overall survival time (p=0.018, Cox survival model). The hazard ratio for the dichotomized RIP1 was 2.6, indicating that the risk of death for a patient with a high RIP1 value would increase about 1.6 fold, compared with a patient with a low RIP1 value. The dichotomized RIP1 was also significantly associated with progression free survival time (p=0.023, Cox survival model), the hazard ratio was 2.3, the progression free median survival time was 7.7 months for the high and 23.2 months for low RIP1 group, and the survival curves was shown in Figure 6B (p=0.02, log-rank test). Progression was defined as the time to detection of recurrent disease on MRI imaging.
There was no significant association between diagnosis age and RIP1 level (p=0.44, Wilcoxon rank test), and the median ages for high and low RIP groups were 48 and 50 years old respectively. The multivariate proportional hazard survival model was fitted with RIP1, age and Karnofsky performance score as covariates. After adjusting for the effects of age and Karnofsky scores, high RIP1 level was still significantly associated with short survival time (hazard ratio = 2.3, and p=0.045) and progression free survival time (hazard ratio=2.2, and p=0.039). Almost all patients underwent surgery, radiation and chemotherapy with temozolomide (Supplemental Table 1).
The major finding of this study is that RIP1, an essential component of inflammation and NF-κB signaling, plays an important role in regulating p53. Our findings indicate that RIP1 activates NF-κB, resulting in upregulation of mdm2 and a complete shutdown of the p53 tumor suppressor signaling network. We show that RIP1 is overexpressed in human glioblastoma (GBM), the most common adult malignant brain tumor, but not in lower grade glioma, and confers a worse prognosis in this disease. Furthermore, our data suggest that the RIP1 may regulate mdm2, lowering p53 function in GBM.
Stressful stimuli, such as DNA damage, results in activation of both p53 and NF-κB pathways. Cross-talk between the NF-κB and p53 signaling pathways is well documented and may play important roles in the pathogenesis of stress/inflammation induced cancer and in resistance to treatment. However, specific mechanisms by which NF-κB and p53 cross-talk are still under intense study. Altered regulation of both the NF-κB and p53 pathways is established in GBM. p53 function is frequently altered in glioma either by direct mutation or changes in regulatory signals, due to mdm2 gene amplification or loss of p14ARF. Our data suggest that increased expression of RIP1 may be an important additional mechanism of regulating p53 in GBM. It is important to note that mdm2 is known to downregulate both wild type as well as mutant p53 which may result in complex biological outcomes (46).
Previous studies have demonstrated that IKK2 or Bcl-3, a protein related to the IκB family of NF-κB inhibitors, regulate p53 via augmentation of mdm2 levels (21, 47). In the case of IKK2, upregulation of mdm2 is mediated by activation of NF-κB. Our data show that inhibition of NF-κB activation using a dominant-negative IκBαM results in a block of RIP1 mediated mdm2 upregulation and rescues RIP1-mediated p53 inhibition. Also, a RIP1 mutant lacking the intermediate domain known to be deficient in NF-κB activation failed to inhibit p53 induction. Thus, the negative regulation of p53 by NF-κB occurs at multiple nodes and may be important in the pathogenesis of cancer. An increase in IKK2 levels has not been reported in GBM, and our data suggest that RIP1 may be the key player conducting the NF-κB-p53 cross-talk in GBM.
A major finding of this study is that RIP1 is overexpressed in GBM but not in lower grade glioma and confers a worse prognosis in GBM. Increased expression of RIP1 is common in GBM, with about 30% of tumors showing increases in RIP1, and frequently the increase is substantial. Furthermore, increased expression of RIP1 is uncommon in grade II-III gliomas, demonstrating a correlation of RIP1 with increased malignancy. Importantly, in matched pairs of primary low grade glioma that progressed to secondary GBM, RIP1 is usually low in the primary low grade glioma and increased in the secondary GBM. In a study of 70 GBMs we find that increased RIP1 is an independent negative prognostic indicator in GBM. There were no differences in the age, Karnofsky performance status, or treatment in the low versus high RIP1 groups. These findings imply that increased expression of RIP1 promotes a more malignant clinical phenotype in GBM. It should be noted that since most patients in our study had a complete resection, our study does not address whether low RIP1 level would confer a survival advantage in those patients with partial resection or biopsy alone.
RIP1 may contribute to the pathogenesis of GBM by multiple mechanisms. Firstly, increased expression of RIP1 is sufficient to activate NF-κB as shown in Figure 3 and reported previously (40, 41). Thus, increased RIP1 level in cancer is likely to lead to constitutive and deregulated NF-κB activation. In addition, RIP1 has also been reported to have a role in PI3K-Akt activation (39, 48). Thus, an augmented cellular RIP1 level in glioma cells appears sufficient to induce sustained activation of at least two pro-survival signaling pathways of central importance in cancer. In this study, we show that RIP1-mediated NF-κB activation leads to upregulation of mdm2 and inhibition of p53 pathways. Thus, cells with increased RIP with activated NF-κB and Akt, and downregulated p53, would favor oncogenic signaling, resistance to DNA damage, and chemotherapy-induced apoptosis, all favoring a more malignant phenotype.
Paradoxically, RIP1 is also involved in cell death when ectopically expressed, in response to inflammatory cytokines, or other forms of cellular stress (24). However, apoptosis and proliferation are closely linked, and a number of key oncogenic proteins such as Ras, c-Myc and E2F1 can also induce apoptosis or growth arrest (49). The RIP1 knockout phenotype includes failure to thrive and an early death with substantial apoptosis in lymphoid and adipose tissues (27), suggesting an important role for RIP1 in survival signaling. RIP1 knockout MEFs are more sensitive to TNF induced cell death and RIP1 protects thymocytes from TNFR-2 induced cell death (50). The data in this study also support an oncogenic and prosurvival role for RIP1 in GBM, but the effect of RIP1 signaling could be quite complex in the heterogenous tumor populations.
Our data suggest that increased expression of RIP1 identifies a group of glioblastoma patients who have a significantly worse prognosis. The clinical utility of this study may lie in the identification of a subgroup of patients with GBM that have high RIP1 levels, worse prognosis, and are resistant to standard chemotherapy. We propose that these patients may respond better to drugs targeting the NF-κB signaling network.
We thank Brian Seed for RIP plasmids and Michelle Kelliher for RIP knockout MEFs. We thank Chan Foong, MS, and Ping Shang for expert technical assistance. We thank Vicki Rankin for assistance in collecting clinical data and Daisi Tucker for help in preparing the manuscript. The Research Repository of Human Brain Tumors and Brain Tissue is supported by the Division of Neuropathology and Annette G. Strauss Center of Neurooncology, UTSW. This is paper CSCN038 and used the flow cytometry and biostatistics cores of the Simmons Cancer Center. This work was supported in part by DOE grant #DE-FG02-06ER64186 to D.A.B. D.S. a recipient of the Flight Attendant Medical Research Institute Clinical Scientist Award.