TRIM21 directly ubiquitylates p62 at residue K7 to inhibit its oligomerization and sequestration function

The failure of the K7R mutant to dimerize suggests two opposite possible scenarios: 1) Modification of K7 by ubiquitylation promotes p62 dimerization via the ubiquitin chain; or 2) Unmodified K7 is essential for p62 dimerization and its modification by ubiquitylation leads to the loss of function of p62. Importantly, an X-ray crystallography analysis has demonstrated that K7 is a basic cluster residue that would form a hydrogen bond with its corresponding acidic cluster residue D69 (Wilson et al., 2003). This supports the latter option that the non-ubiquitinated form of K7 is critical for p62 dimerization. To further understand the nature of p62 K7 ubiquitylation, and to unequivocally demonstrate that p62 K7 ubiquitylation plays a negatively regulates p62 dimerization and aggregation, we sought to identify the ubiquitin E3 ligase that catalyzes the p62 K7 ubiquitylation. In an in vitro ubiquitylation assay, we noticed that the immunoprecipitated Flag-p62 samples from HEK293T cells led to p62 ubiquitylation in the presence of purified E1, E2, and ubiquitin without the addition of an exogenous E3 (Fig. 2A). This suggests that an E3 enzyme was co-immunoprecipitated in the Flag-p62 IP samples. Indeed, mass spectrometry of Flag-p62 IP samples identified two putative E3 ligases: TRIM21 and RNF138. Expression of TRIM21 but not RNF138 led to p62 ubiquitylation (Fig. 2B and 2C), suggesting that TRIM21 is the E3 ligase for p62. A ligase-dead (LD) mutant TRIM21(C16A, C31A, and H33W) (Wada and Kamitani, 2006) revealed that the E3 ligase activity of TRIM21 was required for p62 ubiquitylation (Fig. 2D). An in vitro ubiquitylation assay further revealed that TRIM21 directly ubiquitylates p62 (Fig. 2E) via the K63-linkage (Fig. 2F), which was abolished in the p62K7R mutant (Fig. 2G). Silencing of TRIM21 abolished p62 ubiquitylation both at the basal level and upon MG132 treatment (Fig. 2H). Therefore, TRIM21 is a direct ubiquitin E3 ligase for the p62 residue K7.

Open in a separate window

Figure 2

p62 is ubiquitylated by TRIM21

(A) Flag-p62 was expressed in HEK293T cells, and subjected to IP using the Flag antibody and eluted using Flag peptides. The eluates were incubated with in vitro ubiquitination assay system containing E1, E2, and ubiquitin, then subjected to IB. (B and C) His-Ub was transfected with vector and HA-TRIM21 (B) or HA-RNF138 (C) into HEK293T cells. 24 h later, cells were subjected to His pull-down followed by IB. (D) His-Ub was co-transfected with HA-TRIM21wt or HA-TRIM21-LD mutant in HEK293T cells. Cells were subjected to His pull-down and IB. (E) Purified Flag-p62 was incubated with IgG control, purified HA-TRIM21wt or LD mutant in the ubiquitination assaying mix, and probed for indicated proteins. (F) His-Ub (wt, K48R, and K63R) mutants were expressed with or without HA-TRIM21 in HEK293T cells. Cells were subjected to His pull-down and IB. (G) Flag- and His-doubly tagged p62wt or K7R mutant were co-expressed with vector or TRIM21 in HEK293T cells. Cells were subjected to His-pull down and IB. (H) MD-MBA-231 with stable expression of Flag-p62-His were infected with lentiviral non-targeting control (NTC) shRNA or shTRIM21. Cells were treated with MG132, then subjected to His pull-down and IB.

As the p62K7R mutant was unable to dimerize (Fig. 1D) and to promote protein aggregation (Fig. 1F), we tested whether TRIM21 plays a positive or negative role in p62 dimerization and aggregation using three approaches. First, we used the bimolecular fluorescence complementation (BiFC) assay (Huang et al., 2010). p62wt or the K7R mutant was fused to the split Venus proteins, Venus N(1–173) and Venus C(155–239), and expressed in HEK293T cells (Fig. 3A). BiFC was observed when p62wt-VN and p62wt-VC were co-expressed (Suppl. Fig. S2A and Fig. 3B), but not with the K7R mutants (Fig. 3B). Induction of wild-type TRIM21 using the estrogen receptor inducing system led to reduced BiFC, whereas the ubiquitin ligase-dead (LD) mutant TRIM21-LD failed to do so (Suppl. Fig. S2B and Fig. 3C), indicating that TRIM21 decreases p62 oligomerization in an E3 ligase activity-dependent manner.

Open in a separate window

Figure 3

TRIM21 inhibits p62 dimerization and its sequestration function

(A) Schematic for the strategy constructing p62 fused with Venus N-terminus (p62-VN) or C-terminus (p62-VC). p62-VN and p62-VC were expressed individually or simultaneously in HEK293T cells. The expression was determined by IB. (B) Wild-type or K7R p62-VN/p62-VC were expressed in HEK293T cells. Representative fluorescence microscopic images are shown. (C) p62-VN or p62-VC were expressed with estrogen receptor (ER)-fused TRIM21wt or LD mutant in HEK293T cells. 16 h later, cells were treated with 4-OHT for indicated times to induce TRIM21 expression (top panel). Venus fluorescence was determined by flow cytometry and quantified (bottom panel). (D) HeLa cells stably expressing both Flag- and HA-p62 were infected with lentivirus shNTC and shTRIM21. Cells were used for Flag IP, and IB for HA. The amount of HA-p62 was quantified by ImageJ and is shown in the right panel. *p<0.05. (E) HeLa with stable shTRIM21 were reconstituted with shRNA-resistant TRIM21wt or TRIM21 LD. Cells treated with MG132, then crosslinked with DSP, and lysed in IP lysis buffer with 1% SDS. The cells lysates mixed with loading buffer containing or not containing β-mercaptoethanol were subjected to IB. (F) TRIM21^−/− MEFs reconstituted with TRIM21wt or LD were treated with MG132 (0.25 µM) for 12 h. Cells were used for IF using anti-TRIM21 (green) and anti-p62 (red) antibodies, and observed under deconvolution microscope. Cells with aggregates were quantified by blind counting. Shown on the right is the mean plus S.D. of three countings. *p<0.05, **p<0.01. Cells with high expression of TRIM21 wt or LD are marked with an asterisk.

Alternatively, we examined p62 oligomerization by co-expressing p62 tagged with either Flag or HA epitopes. Since a significant amount of p62 resides in protein aggregates, especially under the condition of proteotoxic stress such as treatment of MG132, we used the reversible crosslinking agent dithiobis succinimidyl propionate (DSP) to treat the cells then lysed the cells in strong detergent (1% SDS) to solubilize all proteins. The lysates were subjected to co-IP upon diluting the detergent to 0.1% SDS. Indeed, TRIM21 silencing led to increased p62 oligomerization at both basal condition and upon MG132 treatment (Fig. 3D). Moreover, since p62 oligomerization is important for its sequestration function, we determined whether p62-mediated aggregation is regulated by TRIM21. To this end, we detected the presence of p62 in aggregates by treating cells with the crosslinking agent DSP and running the cell lysates under either reducing or non-reducing condition (Suppl. Fig. 2C). Under the reducing condition, total p62 would be detected with a migration of 62 kDa, whereas under the non-reducing condition, only p62 monomers would be detected at 62 kDa and the oligomers or those conjugated with other proteins would display such a slow migration that it would not run through the gel (illustrated in Suppl. Fig. S2C). Therefore, higher ratio of the 62 kDa band in the non-reducing vs reducing conditions would indicate reduced p62 oligomerization or sequestration. MG132 treatment led to less p62 monomer in the non-reducing condition (Fig. 3E), indicating p62 aggregation and sequestration. TRIM21 silencing led to further decreased p62 monomer hence increased aggregation and sequestration (Fig. 3E). In TRIM21 knockdown cells, reconstitution of TRIM21wt but not the LD mutant led to increased p62 monomer indicating decreased p62 aggregation and sequestration (Fig. 3E).

The role of TRIM21 in negatively regulating p62 oligomerization was further supported using immunofluorescence analysis and fractionation assay, which shows that reconstitution of TRIM21wt but not the LD mutant led to diminishment of MG132-induced p62 aggregates (Fig. 3F and Suppl. Fig. S2D). Consistent with these findings, TRIM21 silencing led to increased p62 binding to ubiquitylated proteins (Suppl. Fig. S2E), and reconstitution of TRIM21wt but not the LD mutant led to diminished ubiquitin aggregates in TRIM21^−/− cells (Suppl. Fig. S2F).

In addition to the proteasome inhibitor MG132, we also examined other conditions that are known to induce p62 aggregation including genetic ablation of autophagy-essential proteins Atg7 and Vps34. In Atg7^flox/flox (Komatsu et al., 2005) and Vps34^flox/flox (Jaber et al., 2012) MEFs, ablation of respective genes by Cre expression led to decreased p62 monomer in the non-reducing condition as a result of accumulated p62 aggregation due to the failure of autophagic clearance of p62 (Suppl. Fig. S3A and S3B). Overexpression of wild-type TRIM21, but not the LD mutant, led to increased amount of p62 monomer (Suppl. Fig. S3A and S3B), indicating that TRIM21 prevents p62 aggregation under the condition of genetic inhibition of autophagy. Consistently, the increased amount of insoluble polyubiquitylated proteins in Atg7^−/− cells was reduced by the expression of TRIM21wt but not the LD mutant (Suppl. Fig. S3C). Taken together, these data indicate that residue K7 is critical for p62 dimerization, aggregation, and sequestration function, and that TRIM21 negatively regulates these functions by ubiquitylating p62 at residue K7.

TRIM21 interaction with p62 is via the PRYSPRY domain and is enhanced by proteotoxic stress

We next sought to characterize the molecular nature of the interaction between p62 and TRIM21. Two p62 deletion mutants Δ3–117 and Δ117–200 showed significantly decreased interaction with TRIM21 (Fig. 4A). Further deletion mutation revealed that the two domains a.a.3–61 and 163–200 had lost TRIM21 interaction (Fig. 4B). In TRIM21, deletion of a.a.252–476, which spans the PRYSPRY domain that is conserved among numerous TRIM family members and is important for their anti-viral function (Nisole et al., 2005; Song et al., 2005), led to a near complete loss of p62 interaction (Fig. 4C). It has been shown that residues D355, W381, W383, and F450 within the TRIM21 PRYSPRY domain are clustered at the center of the interface that interacts with IgG Fc (James et al., 2007). We reasoned that these residues may also be critical for TRIM21 interaction with other proteins. Indeed, the TRIM21W381/383A mutant abolished p62 interaction (Fig. 4D), p62 ubiquitylation (Fig. 4E), and the suppression of p62 aggregation (Fig. 4F). The importance of the TRIM21 ligase activity was also determined by the autophagic degradation of p62. Wild-type p62 was degraded by starvation-induced autophagy, but p62K7R, which is deficient in oligomerization (Fig. 1C), was not (Suppl. Fig. S4A). This is consistent with the notion that p62 autophagic degradation is dependent on p62 dimerization and aggregation (Itakura and Mizushima, 2011). Importantly, wild-type TRIM21 but not the W381/383A mutant suppressed starvation-induced p62 autophagic degradation (Suppl. Fig. S4B). These data indicate that W381 and W383 are required for TRIM21 interaction with p62 and for its function to ubiquitylate p62 and suppress p62 oligomerization.

Open in a separate window

Figure 4

TRIM21 interacts with p62 via the PRYSPRY domain

(A–C) Flag-tagged TRIM21 and HA-tagged p62wt or their truncation mutants were expressed in indicated combinations in HEK293T cells for 24 h. Cells were subjected to Flag IP and IB. p62 lost interaction with TRIM21 Δ213–117 and Δ117–200 (A). TRIM21 lost interaction with p62 Δ3–61, Δ117–162, and Δ163–200 (B). p62 lost interaction with TRIM21 Δ252–476 (C). (D) Flag-tagged TRIM21wt or W381/383A mutant and HA-tagged p62 were expressed in HEK293T cells for 24 h. Cells were subjected to Flag IP and IB. (E) HA-tagged TRIM21wt or W381/383A mutant and His-Ub were expressed in HEK293T cells. Cells were subjected to His pull-down and IB. (F) TRIM21^−/− MEFs reconstituted with vector, TRIM21wt, and TRIM21 W381/383A mutant were treated with DSP and lysed in IP lysis buffer with 1% SDS. Lysates were mixed with loading buffer with/without β-mercaptoethanol and subjected to IB. (G) HA-tagged p62 and Flag-tagged TRIM21 were expressed in HEK293T cells for 24 h. Cells were treated with MG132 (0.5 µM) for 16 h, rapamycin (50 nM) for 16 h, tunicamycin (0.5 µg/ml) for 16 h, or cultured in serum-free medium for 6 h. Cells were subjected to Flag IP and IB. (H) A vector control or Flag tag was introduced into MEFs to fuse with the endogenous TRIM21 by CRISPR. Cells were treated with MG132 (0.25 µM) for 12 h, then subjected to Flag IP and IB.

We then studied how TRIM21-mediated p62 ubiquitylation might be regulated. In a co-overexpression system, treatment of MG132 but not serum starvation, rapamycin, or tunicamycin led to increased interaction between TRIM21 and p62 (Fig. 4G). Since the available TRIM21 and p62 antibodies were not suitable for immunoprecipitation, we introduced a Flag tag to the endogenous TRIM21 gene via CRISPR gene editing in MEFs (Suppl. Fig. S4C), so that Flag-tagged TRIM21 is expressed at the endogenous level. Co-IP analysis using a Flag antibody showed that MG132 induced the interaction between TRIM21 and p62 (Fig. 4H). These results are consistent with our observation that p62 ubiquitylation is enhanced upon MG132 treatment, and suggest that the TRIM21-mediated p62 ubiquitylation is a cell stress adaptation mechanism in response to proteasome inhibition in order to maintain a hemostatic level of functional p62.

TRIM21 negatively regulates p62-mediated sequestration of Keap1 and antioxidant response

An important function of p62 is to retain proteins in aggregates (sequestosomes) for subsequent autophagic degradation. One of the client proteins is Keap1, which negatively regulates the antioxidant response by interacting and preventing the nuclear accumulation of the transcriptional factor Nrf2, the master regulator to activate antioxidant gene transcription (Komatsu et al., 2010). As our results thus far indicate that TRIM21 negatively regulates p62-mediated protein sequestration by ubiquitylating p62 at K7, we sought to study whether this mechanism has an impact on cell’s antioxidant response. To this end, we used sodium arsenate (As(III)), which induces a detrimental level of intracellular reactive oxygen species (Barchowsky et al., 1996; Lynn et al., 1998). We first used p62^−/− MEFs reconstituted with either p62wt or p62K7R. While As(III) treatment did not induce Keap1 aggregation in p62^−/− MEFs, reconstitution of p62wt but not p62K7R led to aggregation of both p62 and Keap1 and their accumulation in the insoluble fraction (Fig. 5A and 5B). The ability of p62wt to sequester Keap1 was also indicated by increased Nrf2 and NQO1 expression (Fig. 5C and 5D). This As(III)-induced p62 sequestration is likely to be negatively regulated by TRIM21 since As(III) induced p62-TRIM21 interaction at the endogenous level in the TRIM21-Flag knock-in cells (Fig. 5E). This As(III)-induced p62 aggregation is suppressed by TRIM21wt but not the W381/383A mutant (Fig. 5F and Suppl. Fig. S5A). TRIM21 reconstitution led to decreased p62 and Keap1 in the insoluble fraction (Fig. 5G), decreased Nrf2 in the nucleus (Fig. 5H), and decreased Nrf2 stability (Fig. 5I and Suppl. Fig. S5B). Consistently, wild-type TRIM21 showed low level of NQO1 promoter activation (Fig. 5J) and increased oxidative stress (Fig. 5K). Together, these data indicate that p62 plays an antioxidant role by sequestrating Keap1 in response to As(III), and that TRIM21 negatively regulates this function by ubiquitylating p62.

Open in a separate window

Figure 5

Trim21 negatively regulates p62-mediated Keap1 sequestration and antioxidant response

(A) p62^−/− MEFs reconstituted with vector, Flag-p62wt, or Flag-p62K7R mutant were untreated or treated with sodium arsenate (As(III), 10 µM) for 6 h. Cells were subjected to IF with Flag and Keap1 antibodies, and observed under deconvolution microscope. Cells with Keap1 aggregates were counted blindly. Data shown are the averages plus S.D. of at least three countings with over 200 cells. **p<0.01, ***p<0.001. (B) p62^−/− MEFs reconstituted with vector, Flag-p62wt, and Flag-p62K7R were treated with As(III) (10 µM) for 2.5 h. Cells were lysed in IP lysis buffer with 1% Triton X-100. The insoluble fraction was lysed in IP lysis buffer with 1% SDS. 30 µg soluble proteins and corresponding volume of insoluble proteins were used for IB. (C) p62^−/− MEFs reconstituted with vector, Flag-p62wt, and Flag-p62K7R were left untreated or treated with As(III) (10 µM) for 4 h. Cells were treated with cycloheximide (CHX) for indicated times. Cells were lysed in RIPA (1% SDS) and probed for Nrf2 and tubulin. (D) p62^−/− MEFs reconstituted with vector, Flag-p62wt, and Flag-p62K7R were transfected with NQO1-ARE-luc and pRL-TK (Renilla Luciferase Control Reporter Vector, Promega) plasmids. 16 h post transfection, cells were treated with As(III) and assayed for luciferase activity. (E) MEFs without/with CRISPR Flag fused into TRIM21 gene were treated with As(III) (10 µM) for 4 h, and subjected to Flag IP and IB. (F) TRIM21^−/− MEFs reconstituted with vector, TRIM21wt, and TRIM21W381/383A mutant were treated with the crosslinking agent DSP and lyse in IP lysis buffer with 1% SDS. The lysates were mixed with loading buffer with/without β-mercaptoethanol and subjected to SDS-PAGE and IB. (G) TRIM21^−/− MEFs reconstituted with vector and HA-TRIM21wt were treated with As(III). Cells were lysed in IP lysis buffer containing 1% Triton X-100. The insoluble fraction was dissolved in RIPA buffer containing 1% SDS. Both the Triton X-100 soluble and insoluble fractions were subjected to IB. (H) TRIM21^−/− MEFs reconstituted with vector and HA-TRIM21wt were treated with As(III) (10 µM). Cell were fractionated to obtain the cytosol and nuclear fractions, and probed for indicated proteins. (I) TRIM21^−/− MEFs reconstituted with vector, HA-TRIM21wt, or HATRIM21W381/ 383A mutants were treated with As(III) (10 µM) for 4 h. Cells were lysed in RIPA buffer containing 1% SDS and probed for indicated proteins. (J) TRIM21^−/− MEFs reconstituted with vector, HA-TRIM21wt, HA-TRIM21-LD, or HA-TRIM21W381/383A were transfected with NQO1-ARE-luc and renilla luciferase pRL-TK constructs. 16 h later, cells were treated with As(III) (10 µM), and assayed for luciferase activity. (K) TRIM21^−/− MEFs reconstituted with vector, HA-TRIM21wt, HA-TRIM21-LD, or HA-TRIM21W381/383A mutants were treated with As(III) (10 µM) for 4 h. Cells were stained with H₂DCFDA and analyzed by flow cytometry. Quantification of relative H₂DCFDA intensity (Geo mean plus S.D. of four repeats) is shown on the right. ***p<0.001, ****p<0.0001.

Ablation of TRIM21 led to protection against arsenic-induced cell death and liver injury

We next examined the effect of TRIM21 on As(III)-induced cell death. Acute As(III) treatment induces ROS, DNA damage, and activation of poly(ADP-ribose) polymerase (PARP) as well as depletion of NAD and ATP, leading to bioenergetic catastrophe and cell death (Bernstam and Nriagu, 2000; Chou et al., 2004; Kang et al., 2004). Indeed, in wild-type cells, As(III) treatment led to DNA damage indicated by increased H2A.X phosphorylation (Fig. 6A), PARP activation indicated by increased poly(ADP) ribosylation (PAR) (Fig. 6B), and cell death in a manner that was inhibitable by ROS scavengers N-acetylcysteine (NAC) and 2,2,6,6-tetramethylpiperidine-N-oxyl (TEMPO) (Suppl. Fig. S6A) and by PARP inhibitors 3,4-Dihydro-5-[4-(1-piperidinyl)butoxyl]-1(2H)-isoquinolinone (DPQ) and 1,5-dihydroxyisoquinilone (DHIQ) (Suppl. Fig. S6B). Importantly, As(III)-induced DNA damage, PARP activation, and cell death were markedly suppressed in TRIM21^−/− cells (Fig. 6A–D). Reconstitution of TRIM21wt but not the LD mutant re-sensitized cells to As(III) (Fig. 6E). Similarly, silencing of p62 led to increased sensitivity to As(III) (Suppl. Fig. S6C). Similar to TRIM21^−/− MEFs, silencing of TRIM21 in MDA-MB-231 and MCF-7 cells also led to protection against As(III) toxicity (Suppl. Fig. S6D and S6E). The As(III)-induced cell death was not inhibited by necrostatin-1 (Nec-1) (Suppl. Fig. S6F). Interestingly, while TRIM21-deficient cells were refractory to As(III)-induced cell death, they were more susceptible to cell death induced by MNNG (Suppl. Fig. S6G) that was blocked by PARP inhibitor but not Nec-1 (Suppl. Fig. S6H), and to cell death induced by TNFα/cycloheximide/z-VAD (TCZ) (Suppl. Fig. S6I) that was blocked by Nec-1 (Suppl. Fig. S6J), suggesting that TRIM21 may play a protective role in response to harmful conditions other than oxidative stress. Nonetheless, the above data indicate that TRIM21 inhibition leads to protection against As(III)-induced oxidative stress.

Open in a separate window

Figure 6

Ablation of TRIM21 led to the protection from arsenic toxicity

(A and B) TRIM21^+/+ and ^−/− MEFs were treated with As(III) (10 µM) for 12 h (A), or for the indicated times (B). Cells were collected and the lysates were subjected to IB. (C and D) MEFs were treated with As(III) (10 µM). (C) Cells were observed under phase-contrast microscope 16 h after the treatment. Representative micrographs are shown. (D) Cell viability was determined by Trypan Blue. Data shown are the mean plus S.D. of three countings. **p<0.01. (E) TRIM21^−/− MEFs reconstituted with vector, HA-TRIM21wt, or HA-TRIM21W381/383A mutant were treated with As(III) (10 µM). Cell viability was determined by Trypan Blue. Shown are the mean plus S.D. of three countings. **p<0.01. (F–I) Nine pairs of litter mates of TRIM21^+/+ and −/− mice were gavaged with 30 mg/kg As₂O₃. (F) Body weight was measured before and after 24 h treatment, and loss of body weight was calculated and plotted. *p<0.05. (G) Livers from water or As₂O₃-treated mice were collected after 24 h. Tissues were processed for H&E staining and IHC. Representative images are shown. (H) TRIM21^+/+ and TRIM21^−/− mice were treated with As₂O₃ for 24 h. Liver tissues were homogenized and separated into detergent (1% Triton X-100)-soluble “S” and insoluble “I” fractions, and subjected to IB. The ratio of insoluble/soluble band intensities was normalized to the corresponding tubulin band, and is shown on the right. *p<0.05, **p<0.01. (I) Liver tissue was lysed in RIPA buffer containing 1% SDS and subjected to IB.

TRIM21 is ubiquitously expressed in all tissues (Suppl. Fig. S7A). To test the physiological relevance of TRIM21, we tested arsenic trioxide (As₂O₃)-induced liver injury using the whole body TRIM21 knockout mouse (Yoshimi et al., 2009). Sublethal dosage of As₂O₃ (30 mg/kg) was given to wild-type or TRIM21^−/− littermates for 24 h. A significant body weight loss was observed in wild-type mice, which was significantly alleviated in TRIM21^−/− mice (Fig. 6F). Hematoxylin/eosin (H&E) staining revealed moderate level of necrosis near the central vein in the wild-type mice, which was blocked in TRIM21^−/− mice (Fig. 6G). Immunohistochemistry (IHC) analysis revealed that TRIM21^−/− mice displayed more profound aggregation of p62 and ubiquitylated proteins, as well as reduced oxidative DNA damage indicated by 8-oxodG staining (Fig. 6G). Cell fractionation revealed that As₂O₃ treatment led to increased insoluble p62 and Keap1, which was markedly further enhanced and correlated with the higher NQO1 expression in TRIM21^−/− livers (Fig. 6H and 6I). This was consistent with the increased p62-Keap1 colocalization (Suppl. Fig. S7B) as well as increased expression of a number of antioxidant genes upon As(III) treatment (Suppl. Fig. S7C). Together, these data indicate that ablation of TRIM21 confers protection against As₂O₃-induced oxidative burst, cell death, and liver injury.

Lentiviral and retroviral infection

Lentiviruses and retroviruses were generated in HEK293T cells. The infection was performed as previously described (Pan et al., 2011).

Mouse experiments

All mouse experiments were done in compliance with the Institutional Animal Care and Use Committee guidelines at Stony Brook University and Rutgers University.

Statistics

Student’s t test was used to compare the differences between two groups. One-way ANOVA with Tukey’s post-hoc test was used for comparisons between more than two groups. Results were considered significant when p<0.05.

We thank Drs. Michael Frohman (Stony Brook University), Terje Johanson (University of Tromso, Norway), Huikuan Lin (M.D. Anderson Cancer Center), Noboru Mizushima (University of Tokyo), and Zhenyu Yue (Mt Sinai School of Medicine) for reagents. We thank Drs. Yanxiang Zhao (Hong Kong Polytechnic University) and David Goldfarb (University of Rochester) for insightful comments. This work was supported by grants from NIH (R01CA129536 and R01GM97355 to WXZ, R01CA172025 and R01DK108743 to JM, R01CA192642 to MTDM, 5P30CA030199 to MTDM and JM), Department of Veterans Affairs Merit Review Program (to RZL), and a key project (12KG107) from the Municipal Healthy Bureau of Tianjin, China (to BY).