Excessive ER-phagy mediated by the autophagy receptor FAM134B results in ER stress, the unfolded protein response, and cell death in HeLa cells

Autophagy is typically a prosurvival cellular process that promotes the turnover of long-lived proteins and damaged organelles, but it can also induce cell death. We have previously reported that the small molecule Z36 induces autophagy along with autophagic cell death in HeLa cells. In this study, we analyzed differential gene expression in Z36-treated HeLa cells and found that Z36-induced endoplasmic reticulum–specific autophagy (ER-phagy) results in ER stress and the unfolded protein response (UPR). This result is in contrast to the common notion that autophagy is generally activated in response to ER stress and the UPR. We demonstrate that Z36 up-regulates the expression levels of FAM134B, LC3, and Atg9, which together mediate excessive ER-phagy, characterized by forming increased numbers of autophagosomes with larger sizes. We noted that the excessive ER-phagy accelerates ER degradation and impairs ER homeostasis and thereby triggers ER stress and the UPR as well as ER-phagy–dependent cell death. Interestingly, overexpression of FAM134B alone in HeLa cells is sufficient to impair ER homeostasis and cause ER stress and cell death. These findings suggest a mechanism involving FAM134B activity for ER-phagy to promote cell death.

, peroxisomes (pexophagy) (9), etc. The specificity of autophagy for each target is determined by the target-specific autophagy receptors for autophagosome protein LC3. Receptors for mitophagy, such as BNIP3, NIX, and FUNDC1, have been studied most extensively, and they all bind to LC3 in a similar fashion through their short LC3-interacting region (LIR) 2 motifs, but also with subtle differences (10). FAM134B is the first identified receptor for endoplasmic reticulum-specific autophagy (ER-phagy) in mammalian cells (11), followed by SEC62, RTN3, and CCPG1 (12). They function in mediating ER turnover while maintaining ER homeostasis.
Autophagy is usually considered as an essential prosurvival mechanism for cells during starvation or stress conditions (13,14). Nevertheless, in the past decade, much evidence has suggested that enforced overactivation of autophagy will lead to cell death in certain contexts. Autophagy is involved mechanistically in the death of developmental cells during the salivary gland destruction of Drosophila (15). In human ovarian epithelial cells expressing the oncogene H-RAS, high levels of autophagy can lead to caspase-independent autophagic cell death (16), whereas in cells lacking an intact apoptosis pathway, autophagy can also contribute to cell death in the myeloma cells (17) or the murine embryonic fibroblasts (18). Autosis is a novel described form of autophagy-mediated cell death, which is characterized by its unique morphological features and depends on the cellular Na ϩ /K ϩ -ATPase (19). Although there are many reports showing the regulatory role of autophagy during cell death (20 -22), the mechanisms of autophagic cell death remain to be elucidated.
We have previously reported that a small molecule, Z36, can induce Beclin1-dependent autophagy and autophagic cell death in HeLa cells (23). Z36-induced cell death showed no characteristic feature of apoptosis or necrotic-like cell death, and it can be inhibited by the autophagy inhibitors 3-methyladenine (3-MA) and chloroquine (CQ) or the knockdown of BECN1, Atg5, and Atg12 genes (23,24). In this study, we show that Z36 treatment significantly up-regulates FAM134B and LC3 in HeLa cells, which induces extensive ER-phagy that accelerates ER degradation and in turn causes ER damage. Consequently, it triggers ER stress and the UPR, which further result in cell death. We also show that the overexpression of FAM134B alone has similar consequences for ER and causes cell death. These findings provide a novel mechanism for autophagy to result in cell death and establish a new relationship between autophagy and ER stress, in contrast to the common perception of autophagy as the consequence of ER stress.

Z36 up-regulates the expression of genes related to autophagy, ER stress, and the UPR
To better understand the mechanism for Z36 to induce autophagy and cell death, we first compared the morphology of autophagosomes induced by Z36 with those induced by rapamycin (Rapa), which results in canonical protective autophagy (25). Under the fluorescence microscope, GFP-LC3 puncta in Z36-treated HeLa cells appeared to be obviously agglomerated, whereas the GFP-LC3 puncta were mainly small dots in the cells treated with Rapa (Fig. 1A). Using transmission EM (TEM), we found that there are many more autophagosomes in Z36-treated HeLa cells, compared with those of Rapa. In Z36treated cells, the estimated average number of autophagosomes per cell was 30, whereas there were 17 and 6 autophagosomes per cell in Rapa-or DMSO (as control)-treated cells, respectively (Fig. 1B). Meanwhile, Z36 treatment led to a much larger size for autophagosomes, as the average maximal cross-sectional diameters of autophagosomes in Z36-, Rapa-, and DMSO-treated cells were 1.20, 0.74, and 0.70 m, respectively ( Fig. 1, C and D). It is reported that Atg8 (homologue of LC3 in yeast) regulates the size of autophagosomes (26), whereas the level of Atg9 determines the number of autophagosomes (27,28). We have analyzed the expression of these two genes at both D, estimation of maximal autophagosome diameters from TEM images (n ϭ 30 for each sample of three replicate experimental sets). ns, not significant; ***, p Ͻ 0.001. Shown are Western blotting (E) and quantitative real-time PCR analysis (F) of LC3 and Atg9 expression levels in HeLa cells treated with DMSO, 1 M Rapa, or 13 M Z36 for 10 h. The intensities of the respective protein bands in E were quantified using ImageJ, relative to ␤-actin, and then normalized to control. Data represent values of three independent experiments. ns, not significant; **, p Ͻ 0.01; ***, p Ͻ 0.001. Error bars, S.D. mRNA and protein levels. Indeed, the expression levels of both LC3 and Atg9 were significantly higher in Z36-treated cells, than those of Rapa (Fig. 1, E and F).
To gain further insights into the regulation of autophagy in cell death, high-throughput RNA-Seq and differential gene expression analysis were performed on HeLa cells treated with DMSO (as control), Z36, and Rapa. The sequencing generated more than 30 million reads for each sample. The majority of the reads (ϳ95% for all samples) were aligned to the human genome, and over 80% of all the sequence reads were assembled against human genes (Table 1).
Differentially expressed gene (DEG) analysis showed that there are 3588 DEGs with over 2-fold changes (͉log 2 (-fold change)͉ Ͼ 1 and p value Ͻ 0.05) in Z36-treated cells versus those of DMSO, with 1654 genes up-regulated with the largest log 2 scale -fold changes of 5.9, and 1934 genes downregulated with the largest log 2 scale -fold change of 4.9 (Table 2 and Sheet S1). On the contrary, there were only 58 DEGs for cells treated with Rapa, with the highest log 2 scale -fold changes less than 3 (Table 2 and Sheet S2). It is noteworthy that expression levels of autophagic genes were significantly changed in Z36-treated cells, and the change pattern was different from that in Rapa-treated cells ( Fig. S1A and Sheet S3). Eight of the ATG genes were up-regulated Ͼ2-fold (log 2 Ͼ 1) after Z36 treatment, whereas Rapa only caused small changes for the ATG genes, with the highest log 2 change of 0.7 (Fig. S1B). These data indicate that Z36 treatment leads to significant modulation of a large number of genes at the transcriptional level.
Gene ontology (GO) enrichment analysis of Z36-resulted DEGs showed that they are mainly involved in pathways associated with ER stress and the unfolded protein responses, as well as lipid biosynthesis, starvation responses, etc. (Table 3). Most of these pathways were related to ER stress and the UPR, with a majority of ER stress and the UPR genes affected by Z36 treatment (Fig. S2A and Sheet S4 and Table 4). The ER transmembrane proteins, IRE1 (ERN1), PERK (EIF2AK3), and ATF6, acting as ER stress sensors to activate the UPR signaling (29) were up-regulated 4.7-, 3.3-, and 1.5-fold by Z36, respectively. The genes of some prominent proteins involved in the UPR, including multifunctional transcription factor CHOP (DDIT3) and ER chaperone proteins GRP78 (HSPA5 or Bip) were also highly up-regulated in Z36-treated cells (Table 4 and Fig. S2 (B and C)).
All of these data indicate that Z36 stimulates strong responses of multiple cellular processes, different from the typical autophagy inducer Rapa. Especially, ER stress and the UPR pathways are highly activated in Z36-treated cells.

Z36-induced ER stress and the UPR result in cell death, but not autophagy
To verify ER stress onset, typical markers of ER stress and the UPR, including PERK, IRE1␣, eIF2␣, and CHOP, were examined by Western blotting. The results showed that the phosphorylation of PERK, IRE1␣, and eIF2␣ and the level of CHOP are increased after Z36 treatment ( Fig. 2A), confirming that ER stress is indeed activated by Z36 treatment. We then used ER stress inhibitor 4-phenylbutyric acid (4-PBA), a chemical chaperon known to stabilize protein conformation and improve protein folding capacity of ER (30), to inhibit Z36-induced ER stress. Western blotting results showed that the levels of ER stress markers are lowered in Z36-treated cells in the presence of 4-PBA, indicating that the degree of ER stress is reduced ( Fig.  2A). Z36-induced cell death was also reduced from 57 to 37% with 4-PBA treatment (Fig. 2B). However, 4-PBA treatment    2A). Consistently, there was no significant change for the GFP-LC3 punctate distribution in Z36-treated HeLa cells with or without 4-PBA (Fig. 2C). Obviously, attenuating ER stress by 4-PBA reduces Z36-induced cell death but does not affect Z36induced autophagy. Among the three UPR pathways, the mRNA expression of IRE1 and PERK were highly affected due to Z36 treatment (Table 4), which promoted us to investigate the roles of these two pathways in Z36-induced autophagy and cell death. GSK2656157, a selective catalytic PERK inhibitor, was used to inhibit PERK pathway of the UPR (31). Western blot analysis showed that the phosphorylation of eIF2␣ and level of CHOP are reduced by the treatment of GSK2656157 in Z36-treated cells (Fig. 3A), and the cell death rate was also decreased from 53 to 32% (Fig. 3C). This indicates that the inhibition of the PERK pathway with GSK2656157 can inhibit Z36-induced cell death. However, GSK2656157 treatment had no effect on autophagy, as there were no significant changes in the conversion of LC3 or GFP-LC3 puncta distribution (Fig. 3, A and B). We also used shRNA to knock down PERK in HeLa cells (Fig.  3D). As expected, the levels of p-eIF2␣ and CHOP were reduced, and the cell death rate was reduced from 61 to 36% in Z36-treated cells with PERK knockdown. Again, there were no changes in LC3 conversion and GFP-LC3 puncta distribution (Fig. 3, D-F). Meanwhile, re-expression of PERK in PERK knockdown cells restored the PERK pathway signaling and significantly increased the Z36-induced cell death rate to 86%, whereas rescuing PERK still had no effect on LC3 (Fig. 3, D-F). CHOP is known to be pro-death in the PERK-mediated eIF2␣ phosphorylation pathway (32,33), and the mRNA level of CHOP was increased by 10-fold due to Z36 treatment (Table 4). We used RNAi to reduce the expression of CHOP, and the cell death rate was reduced from 57 to 37% (Fig. 3, G and H). These results indicate that PERK-CHOP pathway of the UPR promotes cell death for Z36-induced ER stress. We next used STF083010, a small-molecule inhibitor of IRE1, to inhibit the endonuclease and mRNA splicing activity of IRE1 (34). Z36 treatment resulted in the up-regulation of XBP1 splicing (XBP1s), whereas the splicing was reduced with the addition of STF083010 (Fig. 4, A and B). Interestingly, we found that the Z36-induced cell death rate is increased from 51 to 66% due to the inhibition of IRE1 endonuclease activity (Fig. 4D), implying that the IRE1 pathway plays a prosurvival role in Z36induced ER stress. Meanwhile, Western blotting showed that the level of CHOP is further up-regulated due to STF083010 treatment ( Fig. 4B), which may explain why Z36-induced cell death is increased when the endonuclease activity of IRE1 is inhibited by STF083010. Nevertheless, inhibition of the mRNA-splicing activity of IRE1 had no effect on the conversion of LC3 or GFP-LC3 puncta distribution either (Fig. 4, B and C). We also used shRNA to knock down IRE1 in HeLa cells, which resulted in the level of XBP1s being reduced and CHOP increased (Fig. 4E), whereas the Z36-induced cell death rate was increased from 60 to 73% (Fig. 4G). Again, there was no change in the LC3 conversion and GFP-LC3 dot distribution as well (Fig. 4, E and F). Rescuing IRE1 in IRE1 knockdown cells restored the expression levels of XBP1s and CHOP. Consistently, the Z36-induced cell death rate was also reduced. Re-expression of IRE1 still had no effect on LC3 either (Fig. 4, E-G).
Taken together, these data reveal that ER stress and the UPR pathways are directly involved in the cell death induced by Z36, whereas the PERK arm promotes cell death through CHOP, and the IRE1 branch is prosurvival. This is in line with the previous reports that the IRE1 branch is generally associated with prosurvival pathways in response to ER stress through XBP1s' regulation of chaperones (35) and that the PERK pathway plays a crucial role in cell death through the PERK-ATF4-CHOP route (35,36). Importantly, Z36-induced autophagy is not affected by any of the perturbations of ER stress and the UPR pathways.

ER-phagy results in ER stress, UPR, and cell death Z36-induced autophagy results in ER stress and cell death
To further study the relationship between Z36-induced autophagy and ER stress, autophagy inhibitor 3-MA was used to pretreat HeLa cells before Z36 treatment. Western blot analysis showed that 3-MA treatment inhibits the LC3 conversion (Fig.  5A), and the GFP-LC3 dot areas are also shrunk in HeLa cells treated with Z36 (Fig. S3A). Interestingly, we found that the phosphorylation of IRE1␣, PERK, and eIF2␣ and the level of CHOP are also decreased by 3-MA (Fig. 5A), indicating that Z36-induced ER stress is diminished when autophagy is inhibited. Meanwhile, the cell death rate was reduced from 63 to 34% (Fig. 5B). We also tried to inhibit autophagy with shRNA to knock down Atg5, which facilitates the conversion of LC3I to LC3II (37) (Fig. 5C). Western blotting showed that the conversion of LC3I to LC3II is decreased, consistent with the reduced GFP-LC3 punctate distribution in Atg5 knockdown cells treated with Z36 ( Fig. 5C and Fig. S3B). Meanwhile, the levels of p-IRE1␣, p-PERK, p-eIF2␣, and CHOP were reduced due to Atg5 knockdown (Fig. 5C), indicating that Z36-induced ER stress and the UPR are repressed due to the inhibition of autophagy. Consistently, the cell death rate was decreased from 61 to 42%, presumably because of reduced ER stress and the UPR (Fig. 5D). When we re-expressed Atg5 in Atg5 knockdown cells, the levels of Z36-induced LC3 conversion and GFP-LC3 dot distribution were restored ( Fig. 5C and Fig. S3B). Recue of Atg5 also enhanced the ER stress-UPR signaling and Z36-induced cell death. (Fig. 5, C and D).
Furthermore, we inhibited Z36-induced autophagic flux with CQ, which neutralizes the lysosomal pH and prevents both

ER-phagy results in ER stress, UPR, and cell death
fusion of autophagosome with lysosome and lysosomal protein degradation. As expected, CQ treatment led to more GFP-LC3 puncta visible in Z36-treated cells, and the level of LC3II was elevated ( Fig. S3C and Fig. 5E). Meanwhile, the inhibition of autophagy flux resulted in reduced levels of p-PERK, p-eIF2␣, and CHOP in Z36-treated cells (Fig. 5E). Unexpectedly, the level of p-IRE1␣ was increased due to the block of autophagic flux. As a result, the cell death rate was dramatically reduced from 66 to 19% due to CQ treatment (Fig. 5F), which is consistent with the above findings that the PERK arm promotes cell death and that the IRE1 pathway is prosurvival for Z36-induced ER stress and the UPR. Taken together, our data clearly demonstrate that Z36-induced autophagy results in ER stress and the UPR, in contrast to the fact that autophagy is commonly perceived as a consequence of ER stress and the UPR (38, 39). on the UPR markers and LC3 conversion in Z36-treated HeLa cells. HeLa cells were treated with 13 M Z36 for 10 h, with or without the pretreatment of 1 M GSK2656157, and DMSO was used as a control. The intensities of respective protein bands were quantified using ImageJ, relative to ␤-actin, and then normalized to control. Data represent values of three independent experiments. ns, not significant; **, p Ͻ 0.01; ***, p Ͻ 0.001. B, punctate distribution of GFP-LC3 in Z36-or DMSO-treated HeLa cells with or without GSK2656157. Cells were transfected with GFP-LC3 plasmid for 24 h and then treated with DMSO or 13 M Z36 for 10 h. Scale bars, 20 m. The GFP-LC3 aggregate area present in cells was quantified using ImageJ, and the LC3 dot area per cell was calculated (n ϭ 50 cells for each sample of three replicate experiments). ns, not significant. C, effect of GSK2656157 on Z36-induced cell death. HeLa cells were treated with 13 M Z36 for 30 h, with or without the pretreatment of 1 M GSK2656157, and DMSO was used as control. ns, not significant; ***, p Ͻ 0.001. D, Western blot analysis of the effects of PERK knockdown on the UPR markers and LC3 conversion in HeLa cells treated with Z36. Cells were transfected with CON207, shPERK, or both shPERK and PERK plasmids for 24 h and then treated with DMSO or 13 M Z36 for 10 h. CON207 was used as control plasmid. The intensities of the respective protein bands were quantified using ImageJ, relative to ␤-actin, and then normalized to control. Data represent values of three independent experiments. ns, not significant; *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001. E, punctate distribution of GFP-LC3 in Z36-treated HeLa cells treated as in D. Scale bars, 20 m. The areas of green fluorescent puncta present in cells were quantified using ImageJ, and the LC3 dot area per cell was calculated (n ϭ 50 cells for each sample of three replicate experiments). ns, not significant. F, effect of PERK knockdown on Z36-induced cell death. HeLa cells were transfected with the respective plasmids as in D for 24 h and then treated with DMSO or 13 M Z36 for 30 h. CON207 was a control plasmid. ns, not significant; ***, p Ͻ 0.001. G, efficiency for shRNA knockdown of CHOP determined with Western blotting. The intensity of the CHOP band was quantified using ImageJ, relative to ␤-actin, and then normalized to control. Data represent values of three independent experiments. **, p Ͻ 0.01. H, effect of CHOP knockdown on Z36-induced cell death. HeLa cells were transfected with shRNA plasmid for 24 h and then treated with DMSO or 13 M Z36 for 30 h. CON207 was a control plasmid. ns, not significant; ***, p Ͻ 0.001. Error bars, S.D.

ER-phagy results in ER stress, UPR, and cell death FAM134B-mediated ER-phagy is a prerequisite of ER stress and cell death induced by Z36
In our RNA-Seq data, FAM134B was significantly up-regulated (6-fold) after Z36 treatment (Table 4), which was confirmed by real-time PCR analysis (Fig. S4). FAM134B is a receptor for ER-phagy, and it has been demonstrated that overexpression of FAM134B or its yeast counterpart Atg40 directly promotes ER-phagy and causes ER fragmentation (11,40). We examined the ER morphology of Z36-treated HeLa cells using TEM and found that there are indeed pronounced morphological changes and fragmentation of ER (Fig. 6A). ER whorls were also observed inside autolysosomes (Fig. 6B). These are the characteristic morphology reported for FAM134B-mediated ER-phagy (11).
We used shRNA to mediate the knockdown of FAM134B in HeLa cells (Fig. 6C) and found that it led to attenuated LC3 conversion and reduced GFP-LC3 puncta in Z36-treated cells, indicating that Z36-induced autophagy is suppressed (Fig. 6,  C and D). Western blot analysis also showed that the levels of ER stress and the UPR markers are reduced, suggesting that the Z36-induced ER stress is also attenuated due to the knockdown of FAM134B (Fig. 6C). Consistently, the Z36induced cell death rate was dropped as well, from 57 to 39% (Fig. 6E). It is noticed that the FAM134B protein is reduced by 35% due to shRNA knockdown, whereas it is only elevated by ϳ20% due to Z36 treatment for HeLa cells with or without transfection of shRNA, significantly lower than the elevation at mRNA level. The intensities of the respective protein bands were quantified using ImageJ, relative to ␤-actin, and then normalized to control. Data represent values of three independent experiments. ns, not significant; *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001. C, punctate distribution of GFP-LC3 in Z36 or DMSO treated HeLa cells with or without STF083010. Scale bars, 20 m. The GFP-LC3 aggregate areas present in cells were quantified using ImageJ, and the LC3 dot area per cell was calculated (n ϭ 50 cells for each sample of three replicate experiments). ns, not significant. D, effect of STF083010 on Z36-induced cell death. HeLa cells were treated with 13 M Z36 for 30 h, with or without the pretreatment 80 M STF083010, and DMSO was used as control. ns, not significant; ***, p Ͻ 0.001. E, Western blot analysis of the effects of IRE1 knockdown on the UPR markers and LC3 conversion in HeLa cells treated with Z36. Cells were transfected with CON207, shIRE1␣, or both shIRE1␣ and IRE1␣ encoding plasmids for 24 h and then treated with DMSO or 13 M Z36 for 10 h. CON207 was used as a control plasmid. The intensities of the respective protein bands were quantified using ImageJ, relative to ␤-actin, and then normalized to control. Data represent values of three independent experiments. ns, not significant; *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001. F, punctate distribution of GFP-LC3 in HeLa cells treated as in E. Scale bars, 20 m. The areas of green fluorescent puncta present in cells were quantified using ImageJ, and the LC3 dot area per cell was calculated (n ϭ 50 cells for each sample of three replicate experiments). ns, not significant. G, effect of IRE1␣ knockdown on Z36-induced cell death. HeLa cells were transfected with respective plasmids as in E for 24 h and then treated with DMSO or 13 M Z36 for 30 h. CON207 was a control plasmid. ns, not significant; ***, p Ͻ 0.001. Error bars, S.D.

ER-phagy results in ER stress, UPR, and cell death
Next, we performed transient overexpression of mCherry-FAM134B in HeLa cells, together with GFP-LC3, and it was observed that both mCherry and GFP puncta are formed, and they are mostly co-localized, consistent with the previous report that the overexpression of FAM134B alone can promote ER-phagy. Z36 treatment led to further accumulation larger GFP puncta that are predominantly co-located with mCherry puncta, along with the appearance of mCherry-only puncta (Fig. 7A). As mCherry-FAM134B resides on ER, it is possible that the mCherry-only puncta are engulfed ER fragments from autophagosomes bound by the endogenous LC3 or have resulted from the extinction of GFP fluorescence owing to the acidic milieu inside autolysosomes. In addition, Z36 treatment promoted the colocalization of LC3-positive autophagosomal puncta with Rtn4-positive puncta from tubular ER or Climp-63-positive puncta from sheetlike cisternal ER, based on im- Figure 5. Inhibitions of autophagy machinery suppress both ER stress and the UPR as well as cell death. A, Western blot analysis of the effects of 3-MA treatment on ER stress and the UPR markers and LC3 conversion in Z36-treated HeLa cells. HeLa cells were treated with 13 M Z36 for 10 h, with or without the pretreatment of 1 mM 3-MA, and DMSO was used as a control. The intensities of the respective protein bands were quantified using ImageJ, relative to ␤-actin, and then normalized to control. Data represent values of three independent experiments. *, p Ͻ 0.05; ***, p Ͻ 0.001. B, effect of 3-MA on Z36-induced cell death. HeLa cells were treated with 13 M Z36 for 30 h, with or without the pretreatment of 1 mM 3-MA, and DMSO was used as control. ns, not significant; ***, p Ͻ 0.001. C, Western blot analysis of the effect of Atg5 knockdown on ER stress and the UPR markers and LC3 conversion. Cells were transfected with CON207, shAtg5, or both shAtg5 and Atg5 encoding plasmids for 24 h and then treated with DMSO or 13 M Z36 for 10 h. CON207 was used as a control plasmid. The intensities of the respective protein bands were quantified using ImageJ, relative to ␤-actin, and then normalized to control. Data represent values of three independent experiments. ns, not significant; *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001. D, effect of knockdown or rescue of Atg5 on Z36-induced cell death. HeLa cells were transfected with respective plasmids as in C for 24 h and then treated with DMSO or 13 M Z36 for 30 h. CON207 was a control plasmid. ns, not significant; ***, p Ͻ 0.001. E, Western blot analysis of the effects of CQ treatment on ER stress and the UPR markers and LC3 conversion in Z36-treated HeLa cells. HeLa cells were treated with 13 M Z36 for 10 h, with or without the pretreatment of 20 M CQ, and DMSO was used as control. The intensities of the respective protein bands were quantified using ImageJ, relative to ␤-actin, and then normalized to control. Data represent values of three independent experiments. *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001. F, effect of CQ treatment on Z36-induced cell death. HeLa cells were treated with 13 M Z36 for 30 h, with or without pretreatment with 20 M CQ. DMSO was used as a control. ns, not significant; ***, p Ͻ 0.001. Error bars, S.D.

ER-phagy results in ER stress, UPR, and cell death
munofluorescence analysis (Fig. S5). These findings are consistent with the previous report that the overexpression of FAM134B initiates ER-phagy, and it affects both Climp-63positive sheets and Rtn4-enriched tubules/edges of sheets in mouse embryo fibroblasts (11). Taken together, it appears that Z36-induced autophagy is predominantly ER-phagy in nature.
To study the effect of FAM134B-mediated ER-phagy on ER stress and cell death, we constructed a FAM134B LIR-motif mutant (FAM134B-mut) in which the LIR sequence FELL was replaced with GEGG, so that it can no longer bind to LC3. It was observed that there are almost no mCherry and GFP dots in HeLa cells overexpressing mCherry-FAM134B-mut and GFP-LC3, whereas only GFP puncta can be observed with Z36 treatment (Fig. 7A), indicating that the ability of FAM134B to induce ER-phagy was abolished by the mutation. Western blot analysis showed that the overexpression of FAM134B further increases the conversion of LC3I to LC3II in Z36-treated cells, whereas the levels of p-IRE1␣, p-PERK, p-eIF2␣, and CHOP are increased as well. On the contrary, no obvious differences were observed for the LC3 conversion and ER stress-UPR markers in Z36-treated cells, with or without the overexpression of FAM134B-mut (Fig. 7, B and C). Cell death analysis showed that the overexpression of FAM134B boosted the Z36-induced cell death rate from 65 to 85%, whereas the mutant protein resulted in no change in cell death (Fig. 7D).
It is interesting to notice that the overexpression of FAM134B alone in HeLa cells can result in enhanced ER stress and cell death, as well as LC3 conversion and GFP-LC3 puncta distribution (Fig. 7, B-D). We further investigated the effects of the overexpression of FAM134B and its mutant on cells without Z36 treatment. Overexpression of FAM134B alone indeed increased the cell death rate from 15 to 29%, which could be suppressed by 3-MA and 4-PBA, whereas the overexpression of FAM134B-mut had no obvious effect on cell death (Fig. 8A). Similar to Z36 treatment, the overexpression of FAM134B alone up-regulated the mRNA levels of LC3 and ATG9, which are genes regulating the size and number of autophagosomes, respectively. The mRNA levels of the UPR genes XBP1s, IRE1␣, and CHOP were also up-regulated (Fig. 8B), in line with the data of Z36-treated cells (Fig. 1F and Table 4). As expected, the expression of these genes was not affected by the overexpression of FAM134B-mut, except that the mRNA of CHOP was slightly higher (Fig. 8B). Although the mRNA of PERK was not affected by the overexpression of WT FAM134B (Fig. 8B), Western blot analysis revealed that the phosphorylated PERK is up-regulated, along with the UPR markers p-IRE1␣, p-eIF2␣,   A, effect of FAM134B and its LIR motif mutant FAM134B-mut overexpression on cell death. HeLa cells were first transfected with empty vector, mCherry-FAM134B, and mCherry-FAM134B-mut plasmids for 24 h. The mCherry-FAM134B-overexpressing cells were also treated with 1 mM 3-MA or 2 mM 4-PBA. Cell death was assayed after an additional 30 h. ns, not significant; **, p Ͻ 0.01; ***, p Ͻ 0.001. B, quantitative real-time PCR analysis of mRNA expression levels for LC3, ATG9, XBP1s, PERK, IRE1, and CHOP in FAM134B-and FAM134B-mut-overexpressing HeLa cells. HeLa cells were first transfected with respective plasmid for 24 h, and the cells were collected for analysis after an additional 10 h. ns, not significant; **, p Ͻ 0.01; ***, p Ͻ 0.001. C, Western blot analysis of ER stress and the UPR markers and LC3 conversion in FAM134B-and FAM134B-mut-overexpressing HeLa cells. The FAM134B overexpression cells were also treated with 1 mM 3-MA or 2 mM 4-PBA. HeLa cells were first transfected with empty vector, mCherry-FAM134B, and mCherry-FAM134B-mut plasmid for 24 h, and the cells were collected for analysis after an additional 10 h. The intensities of the respective protein bands were quantified using ImageJ, relative to ␤-actin, and then normalized to control. Data represent values of three independent experiments. ns, not significant; *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001. D and E, Western blot analysis of ER structural proteins Rtn4 and Climp-63 degradation in HeLa cells. D, cells were treated with 13 M Z36, or together with 1 mM 3-MA or 2 mM 4-PBA, DMSO was used as a control. Time 0 is the time for adding Z36 to the cells. E, cells were transfected with mCherry-FAM134B or mCherry-FAM134B-mut and empty vector as a control. The effects of 1 mM 3-MA or 2 mM 4-PBA treatments on mCherry-FAM134B overexpression cells were also assayed. Time 0 is set to be 24 h after transfection. The intensities of Rtn4 and Climp-63 were quantified using ImageJ, relative to ␤-actin, and then normalized to 0 h. Data represent mean Ϯ S.D. (error bars) of three independent experiments. ns, not significant; **, p Ͻ 0.01; ***, p Ͻ 0.001.

ER-phagy results in ER stress, UPR, and cell death
and CHOP, and the conversion of LC3I to LC3II is enhanced as well (Fig. 8C). Again, no significant changes were observed for these proteins in the FAM134B-mut-transfected cells (Fig.  8C). In accordance with the results of Z36 treatment, the upregulation of ER stress and the UPR markers due to FAM134B overexpression could be inhibited by autophagy inhibitor 3-MA or ER stress inhibitor 4-PBA, whereas 4-PBA did not affect the conversion of LC3 (Fig. 8C). All of these data further indicate that Z36-induced autophagy should be principally FAM134B-mediated ER-phagy, which is directly involved in Z36-induced ER stress, the UPR, and cell death.
Furthermore, we investigated the ER turnover by monitoring the degradation of ER structural proteins Rtn4 and Climp-63. The results showed that Z36 treatment leads to progressive degradation of Rtn4 and Climp-63 with time, whereas they are relatively unchanged for control cells. Inhibition of autophagy by 3-MA significantly suppressed the degradation rates, whereas ER stress inhibitor 4-PBA had no effect (Fig. 8D). Similarly, increasing degradation of Rtn4 and Climp-63 was also observed in cells overexpressing WT FAM134B without Z36 treatment, but not for FAM134B-mut. Autophagy inhibitor 3-MA could inhibit the degradation of Rtn4 and Climp-63, whereas ER stress inhibitor 4-PBA failed (Fig. 8E), in cells overexpressing FAM134B.
Taken together, it is apparent that Z36 treatment up-regulates of FAM134B expression and the conversion of LC3I to LC3II, which results in massive selective ER-phagy. The excessive ER-phagy degrades ER and impairs ER homeostasis and then causes ER stress and the unfolded protein responses and cell death. Therefore, it is ER-phagy mediated by FAM134B that is central to Z36-induced ER stress, the UPR, and cell death.

Discussion
In this study, we demonstrate that Z36 up-regulates the expression levels of ER-phagy receptor protein FAM134B, along with key autophagic proteins LC3 and Atg9, which are responsible for the size and number of autophagosomes, respectively. Z36 also increases the conversion of LC3I to LC3II. As a result, Z36 treatment induces excessive ER-phagy in HeLa cells, which further leads to ER stress and the UPR. This is contrary to the common perception that generally regards autophagy, including ER-phagy, as the consequence of ER stress and the UPR (41,42).
We have clearly showed that inhibitions of ER stress and the UPR have no effect on Z36-induced ER-phagy, whereas perturbations of autophagy machinery significantly impact ER stress and the UPR. Inhibitions of autophagy at early stage by 3-MA or Atg5 knockdown can suppress both the IRE1 and PERK pathways of the UPR (Fig. 5, A and C), whereas inhibition of autophagic flux with CQ only attenuates the PERK pathway, with the IRE1 pathway further enhanced (Fig. 5E). This may suggest that the intact autophagic flux machinery of Z36-induced ER-phagy should have multiple roles in regulating ER stress and the UPR.
Interestingly, we found that the up-regulation of PERK pathway of the UPR promotes Z36-induced cell death (Fig. 3), whereas the up-regulation of the IRE1 arm of the UPR is pro-survival (Fig. 4). Therefore, the PERK pathway and IRE1 pathway play opposite roles in regulating Z36-induced cell death. When ER stress and the UPR or autophagy are suppressed with some shRNAs or inhibitors, both PERK and IRE1 pathways are normally down-regulated; thus, only modest effects on cell death are observed. Consistently, the overexpression of FAM134B results in a moderate increase in cell death, as both the PERK and IRE1 pathways are up-regulated (Fig. 7, C and D). However, when IRE1 is knocked down with shRNA or its endonuclease activity is inhibited with STF083010, it is found that the expression of CHOP in the PERK pathway is up-regulated (Fig. 4, B and E), and the cell death rates are increased (Fig. 4, D  and G). On the other hand, when the autophagic flux is inhibited with CQ, the pro-death PERK pathway is suppressed, but the prosurvival IRE1 pathway is enhanced (Fig. 5E). As a result, the Z36-induced cell death rate is reduced most dramatically (Fig. 5F). This is different from the effects of inhibiting autophagy with 3-MA or Atg5 shRNA, which moderately reduces the cell death rates, as both the PERK and IRE1 pathways are down-regulated (Fig. 5, A-D).
Selective ER-phagy is suggested to play key roles in maintaining ER homeostasis to limit stress-induced ER expansion by delivering ER fragments to lysosome for degradation (11,41). However, for Z36-induced ER-phagy, we found accumulation of autophagosomes characterized by much larger size and number than those due to rapamycin treatment. These are consistent with the up-regulation of FAM134B, LC3, and Atg9 expression (Fig. S4 and Fig. 1F). Meanwhile, the rate for ER degradation is very much accelerated, as indicated by the progressive degradation of ER marker proteins Rtn4 and Climp-63 (Fig. 8D). As a result, the overly activated ER-phagy indiscriminately degrades ER until it can no longer function properly and impairs the ER homeostasis. Apparently, Z36-induced ER stress and the UPR originate from the "overeating" of Z36-induced excessive ER-phagy and further cause cell death. Therefore, the cell death induced by Z36 not only results from ER stress and the UPR, but is also a direct consequence of ERphagy, and thus it is an ER-phagy-dependent cell death according to the definition by Klionsky et al. (25) and the Nomenclature Committee on Cell Death (NCCD) (42), because FAM134B-mediated ER-phagy is a prerequisite for ER stress and the UPR.
It has been reported previously that the overexpression of FAM134B in U2OS cells or its functional counterpart Atg40 in yeast can directly activate ER-phagy (11,40). With Z36 treatment, we observed a dramatic increase (ϳ4-fold) of FAM134B expression at the mRNA level (Fig. S4). However, Z36 treatment only increased FAM134B protein expression by ϳ20% from Western blot analysis (Fig. 6C). But this should not reflect the actual situation, because FAM134B is key to fragment and sequester ER into autophagosomes, and it is subjected to continuous degradation due to Z36-induced ER-phagy. Interestingly, we find that ER-phagy can be triggered by the overexpression of FAM134B alone in HeLa cells, which also displays similar effects as Z36-induced ER-phagy (e.g. it can up-regulate the mRNA of LC3 and ATG9 and enhance the conversion of LC3 as well as impair ER homeostasis through the degradation of ER and activate the ER stress-UPR response) (Fig. 8, B-E).

ER-phagy results in ER stress, UPR, and cell death
Meanwhile, the overexpression of FAM134B also promotes the death of HeLa cells, although the cell death rate is lower than that of Z36 treatment (Fig. 8A). Considering the fact that Z36 treatment regulates the mRNA expression of 3588 genes from our RNA-Seq results (Table 2), the mechanism underlying Z36induced cell death should be more complicated than just the overexpression of FAM134B, although FAM134B should play a critical role in this process.
Previously reported clinical observations indicate that FAM134B exhibits a tumor-suppressive function, as lower expression of FAM134B is associated with worse pathological outcomes, such as larger tumor size, more advanced stages of cancer, higher rates of cancer recurrence, and lower survival rates (43). More future studies are needed to reveal the detail mechanisms for Z36-induced cell death and the expression regulation of FAM134B in cancer cells, which may provide new strategies for antitumor therapy development.

Cell culture conditions and antibodies
HeLa cells were cultured in DMEM (Solarbio, catalog no. 12100-500) supplemented with 10% fetal bovine serum (HyClone, catalog no. SV30087) under 5% CO 2 in a humidified incubator at 37°C. To induce autophagy, cells were treated with rapamycin (Sigma, catalog no.

RNA extraction and cDNA synthesis
HeLa cells were collected and resuspended with PBS into a 1.5-ml microcentrifuge tube. Samples were centrifuged at 1000 rpm for 3 min to pellet the cells at room temperature. Total RNAs were extracted with the RNeasy Mini kit (Qiagen, catalog no. 74104). The complementary DNA was reverse-transcribed with 1 g of total RNA and oligo(dT) using the GoScript TM reverse transcription system (Promega, catalog no. 0000202447).

RNA-Seq and data analysis
cDNA library preparation and Illumina high-throughput sequencing (Illumina Hiseq2000) was performed by BIOPIC at Peking University. Each sample had two repeats. Demultiplexed and quality-filtered mRNA-Seq reads were aligned to the GRCh37/hg19 human genome using Subjunc program (http://bioinf.wehi.edu.au/subread/) (44), 3 and then Ensembl gene annotation version 75 was used for gene-level quantification. The raw count data of the expressed genes were normalized for RNA composition using the TMM method (45) from the EdgeR package (46,47) and then transformed to log2CPM values using the voom method (48) from the R Limma package (49). Next, a linear model was built for each comparison using the Limma package, and statistics for differential expression analysis were computed. To filter for differential expression, 2-fold change with FDR Յ 0.05 was used as the cutoff.

Western blotting
HeLa cells were lysed with lysis buffer (Beyotime, catalog no. P0013). Proteins were separated with 4 -20% gradient SDS-PAGE and transferred to polyvinylidene difluoride membranes. The membranes were first blocked with 5% low-fat milk in TBS buffer (20 mM Tris, 150 mM NaCl, 0.02% Tween 20, pH 7.4) and then incubated with the indicated primary antibodies overnight at 4°C. Then HRP-conjugated secondary antibodies were used. Band detection was performed with a Chemiluminescent Substrate kit (Thermo, catalog no. 34077) and analyzed with a Tanon 5200 system.

EM assay
For TEM, WT HeLa cells were initially fixed using 2.5% glutaraldehyde in 0.1 M sodium phosphate buffer (pH 7.4) for 1 h at 37°C and then post-fixed in 2% OsO 4 for 1 h at room temperature. After being dehydrated in a graded series of ethanol, cells were embedded into Spurr's resin. Then the samples were sliced into 70-nm sections using an ultramicrotome (Leica Microsystem). Ultrathin sections were stained with uranyl acetate and lead citrate and examined with a transmission electron microscope Tecnai G 2 20 TWIN (FEI) at 120 kV. The size of autophagosomes was calculated from the diameters of the largest 1-3 autophagosome cross-sections in each cell in TEM images. The diameter values were using the formula, diameter ϭ 2 ϫ radius. The radius of each autophagosome was measured as described previously (26).

Immunofluorescence microscopy assay
HeLa cells were grown to 60% confluence on a coverslip. After treatments, cells were washed three times with PBS and fixed with freshly prepared methyl alcohol at Ϫ20°C for 20 min. Cells were incubated with primary antibodies overnight at 4°C and, after washing with PBS, stained with fluorescent secondary antibodies for 1 h at room temperature. After they were rinsed with PBS, cells were stained with 4Ј,6-diamidino-2-phenylindole for 15 min at room temperature. The cells were then further washed three times with PBS. Cell images were captured using a Nikon A1RSi confocal microscope.

Cell viability assay
Cell viability was measured using a trypan blue dye exclusion assay as described previously (23). After treatment for 30 h, HeLa cells were trypsinized and suspended with DMEM and then stained with trypan blue (Solarbio, catalog no. C0040) for 3 min. The cell death rate was then counted using a Countess II Automated Cell Counter (Thermo Fisher Scientific).

Statistical analysis
For quantitative analysis, values were obtained from three independent experiments, and data were presented as points and S.D. Statistical analyses were performed using Student's t test or two-way analysis of variance, with a p value Ͻ 0.05 considered significant.