A critical role of DDRGK1 in endoplasmic reticulum homoeostasis via regulation of IRE1α stability

Disturbance of endoplasmic reticulum (ER) homoeostasis induces ER stress and leads to activation of the unfolded protein response (UPR), which is an adaptive reaction that promotes cell survival or triggers apoptosis, when homoeostasis is not restored. DDRGK1 is an ER membrane protein and a critical component of the ubiquitin-fold modifier 1 (Ufm1) system. However, the functions and mechanisms of DDRGK1 in ER homoeostasis are largely unknown. Here, we show that depletion of DDRGK1 induces ER stress and enhances ER stress-induced apoptosis in both cancer cells and hematopoietic stem cells (HSCs). Depletion of DDRGK1 represses IRE1α-XBP1 signalling and activates the PERK-eIF2α-CHOP apoptotic pathway by targeting the ER-stress sensor IRE1α. We further demonstrate that DDRGK1 regulates IRE1α protein stability via its interaction with the kinase domain of IRE1α, which is dependent on its ufmylation modification. Altogether, our results provide evidence that DDRGK1 is essential for ER homoeostasis regulation.

T he endoplasmic reticulum (ER) is an essential organelle for multiple cellular functions, including the maintenance of calcium homoeostasis; the biosynthesis of proteins, lipids or sterols; and the transport of synthesized proteins 1,2 . Disturbances in redox homoeostasis, calcium regulation, glucose deprivation or viral infection can lead to ER stress and trigger the unfolded protein response (UPR) to alter transcriptional and translational programs in the stressed cell 1,3,4 . In eukaryotic cells, the UPR is an adaptive cellular response to the disturbance of normal ER functions that attenuates the aggregation of unfolded or misfolded proteins and promotes cell survival. However, during prolonged or overwhelming ER stress, the UPR fails to restore ER homoeostasis, and the apoptotic cascade is activated 5,6 . There are three branches of the UPR that are initiated by distinct ER stress sensors located on the ER membrane: inositol-requiring enzyme 1 (IRE1a), PKR-like endoplasmic reticulum kinase (PERK) and activating transcription factor 6 (ATF6) 1,7 . IRE1a has been considered a master regulator of cell fate determination under ER stress. In the early stage of the UPR, IRE1a splices the mRNA of the transcription factor X-box-binding protein 1 (XBP1) to induce the transcription of ER quality control components to restore ER homoeostasis and promote cell survival 7,8 . If the restoration of ER homoeostasis fails, IRE1a represses the adaptive responses and activates apoptosis through JNK signalling 9,10 . Similar to IRE1a, PERK directly phosphorylates the subunit of eukaryotic initiation factor 2 (eIF2a), which attenuates global mRNA translation to protect cells from ER stress-mediated apoptosis at the initial phase of the UPR (refs 3,11). In addition, eIF2a selectively translates activating transcription factor 4 (ATF4), which subsequently regulates several UPR target genes, including those involved in ER stress-mediated apoptosis, such as C/EBP homologous protein (CHOP) 5,9,11 . The ability to respond to perturbations in ER stress is a fundamental property of cells.
DDRGK domain-containing protein 1 (DDRGK1) is a newly identified ER-associated protein that is induced by ER stress 12 . The DDRGK1 protein contains a signal peptide for ER anchoring, a transmembrane helix and a proteasome component (PCI) domain. The PCI domain is a known protein-protein interaction mediator involved in several multi-protein complexes that regulates the protein life span 13 . Several studies indicate that DDRGK1 forms a large protein complex with UFL1/NLBP (Ufm1-specific ligase 1), the putative tumour suppressor C53/LZAP and Ufm1 (ubiquitin-fold modifier 1) and each of these proteins is involved in ufmylation 12,[14][15][16] . Ufmylation is a ubiquitin-like modification, and components of the ufmylation system are induced under ER stress 12,[17][18][19] . A recent study demonstrated that DDRGK1 is a critical component of the Ufm1 system, which targets ASC1 for oestrogen receptor (ERa) transactivation and breast cancer progression 20 . However, the biological functions of DDRGK1 are largely unknown.
In the present study, we provide evidence that DDRGK1 plays an essential role in the maintenance of ER homoeostasis. Our data show that DDRGK1 regulates IRE1a protein stability through an ufmylation-dependent protein-protein interaction. These results underscore the physiological significance of DDRGK1 and its ufmylation modification in ER homoeostasis.

Results
DDRGK1 is required for ER homoeostasis. To understand the cellular function of DDRGK1, we examined the effects of DDRGK1 knockdown using a mixture of two siRNAs as described previously 21 . We found that knockdown of DDRGK1 induced apoptotic cell death in both MCF7 and HepG2 cells, characterized by Annexin V/PI staining (Fig. 1a), caspase-3 activation and PARP cleavage (Fig. 1b). It should be noted that we can detect the capspase-3 cleavage in HepG2 cells induced by knockdown of DDRGK1 but not in MCF7 cells, because MCF7 cells are caspase-3 deficient 22 . Quantitative real-time PCR (Q-PCR) analysis showed that knockdown of DDRGK1 increased the expression of the pro-apoptotic genes BAX, BAK, NOXA, DR5 and Bid, whereas the expression of the anti-apoptotic gene Bcl-2 was decreased (Fig. 1c,d). Importantly, we found that knockdown of DDRGK1 in MCF7 and HepG2 cells induced ER stress responses, with increased ER stress-specific gene expression of BiP, HSPA8 and CHOP (Fig. 1e,f).
To further confirm that DDRGK1 is involved in the ER stress response, we determined the effect of DDRGK1 on cell survival after treatment with the ER stress inducers thapsigargin (Tg) and tunicamycin (Tm). Knocking down DDRGK1 enhanced ER stress-induced apoptosis by Tg treatment in both HepG2 and MCF7 cells, as assessed by Annexin V/PI staining and cleavage of caspase-3 and PARP (Fig. 2a-c and Supplementary  Fig. 1A-C). In contrast, over-expression of DDRGK1 reduced ER stress-induced cell death in HepG2 cells (Fig. 2d-f). In addition, the viability of MCF7 cells was assessed by MTT assay, and the results showed that knockdown of DDRGK1 rendered the cells more sensitive to Tg or Tm treatment, with reduced cell viability ( Supplementary Fig. 1D), whereas over-expression of DDRGK1 promoted cell survival after Tg or Tm treatment ( Supplementary Fig. 1E). Taken together, our results suggest that DDRGK1 plays a vital role in regulating ER homoeostasis.
DDRGK1 modulates the UPR. To understand how DDRGK1 regulates ER homoeostasis, we first analysed changes in the protein levels of three UPR sensors and their downstream targets in DDRGK1-depleted cells. The results showed that knockdown of DDRGK1 in MCF7 and HepG2 cells significantly reduced the levels of total IRE1a but weakly decreased phosphorylated IRE1a (p-IRE1a), which may be because of the reduced levels of total IRE1a (Fig. 3a). The levels of ATF6 and cleaved ATF6 were not affected (Fig. 3a). Interestingly, knockdown of DDRGK1 did not affect the levels of total PERK, but the levels of phosphorylated PERK (p-PERK) and BiP were significantly increased (Fig. 3a). We then examined the effects of DDRGK1 depletion on the IRE1a substrate XBP1. The results showed that XBP1s, a spliced form of XBP1, was significantly decreased in DDRGK1-knockdown MCF7 cells in a time-dependent manner (Fig. 3b), which was correlated with changes in the IRE1a protein levels (Supplementary Fig. 2A). In addition, depletion of DDRGK1 increased the levels of eIF2a phosphorylation (p-eIF2a) and CHOP expression in both MCF7 and HepG2 cells (Supplementary Fig. 2B). These results indicate that depletion of DDRGK1 represses UPR-IRE1a signalling and activates the UPR-PERK apoptotic pathway, which consequently triggers apoptosis, as observed in Figs 1 and 2. However, the level of IRE1a was increased in DDRGK1 over-expressing cells in a dose-dependent manner (Fig. 3c), while the levels of p-PERK, PERK, ATF6, p-IRE1a, BiP and the spliced form of XBP1 were not significantly changed ( Fig. 3c; Supplementary Fig. 2C and D). To explore the functional relationship between DDRGK1 and the UPR, we assessed the dynamic changes of IRE1a and PERK in DDRGK1-knockdown MCF7 cells and in control cells following Tg treatment at different time points. Treatment with Tg increased the levels of IRE1a and p-IRE1a, p-PERK and BiP in control cells, as expected. However, the total IRE1a levels were significantly decreased (0-24 h), while the p-IRE1a levels were modestly decreased (4-24 h) in DDRGK1-depleted cells compared with control cells (Fig. 3d). Concomitantly the level of XBP1s was decreased in DDRGK1-depleted cells compared with control cells (4-12 h) (Fig. 3e). The levels of total PERK were not changed, but p-PERK was significantly increased in DDRGK1-depleted cells (0-12 h) (Fig. 3d). Similar results were obtained in HepG2 cells (Supplementary Fig. 2E and F). Taken together, our results suggest that depletion of DDRGK1 causes IRE1a degradation and PERK activation.
DDRGK1 regulates the UPR by targeting IRE1a. The UPR is triggered by ER stress sensors upon ER stress to regulate ER homoeostasis 8 . It has been reported that hepatocyte-specific deletion of IRE1a in mice resulted in activation of the UPR-PERK pathway 23 . To address whether the induction of p-PERK observed in DDRGK1-depleted cells was mediated by decreased levels of IRE1a, we examined the levels of p-PERK, PERK and its downstream targets in IRE1a-depleted cells. The results  ARTICLE showed that knockdown of IRE1a significantly increased the protein levels of p-PERK and p-eIF2a in both MCF7 and HepG2 cells (Fig. 4a), similar to those observed in DDRGK1depleted cells ( Fig. 3a; Supplementary Fig. 2B). These results indicate that DDRGK1 regulates the UPR by targeting IRE1a. We then investigated how DDRGK1 regulates IRE1a. Our results showed that IRE1a protein levels, but not mRNA levels ( Supplementary Fig. 3A), were dramatically decreased in DDRGK1-knockdown cells (Fig. 3a). Treatment of the cells with the proteasome inhibitor MG132 blocked the DDRGK1mediated reduction of IRE1a protein ( Fig. 4b; Supplementary  Fig. 3B). Furthermore, we observed that depletion of DDRGK1 dramatically reduced the half-life of IRE1a in a cycloheximide chase assay (Fig. 4c). In accordance with these observations, we found that ectopic expression of IRE1a improved cell survival in both DDRGK1-depleted MCF7 and HepG2 cells compared with control cells, characterized by Annexin V/PI staining and caspase-3 activation (Fig. 4d,e; Supplementary Fig. 3C and D).
Altogether, these results suggest that DDRGK1 regulates the stability of IRE1a and thereby regulates the UPR.
DDRGK1 interacts with IRE1a. To understand how DDRGK1 regulates IRE1a stability, we tested whether DDRGK1 interacts with IRE1a via a reciprocal immunoprecipitation (IP) assay in HEK293T cells. The results showed that exogenously expressed Flag-IRE1a specifically interacted with endogenous DDRGK1 and BiP (Fig. 5a). Consistently, exogenously expressed Flag-DDRGK1 specifically interacted with endogenous IRE1a and the known DDRGK1-associated protein C53 (ref. 14). However, we were unable to detect BiP in the Flag-DDRGK1 immunoprecipitates (Fig. 5b), which suggests that DDRGK1 may not directly interact with BiP. IRE1a, as an endogenous ER-associated degradation (ERAD) substrate, is degraded through the association between IRE1a and the Sel1L-Hrd1 ERAD complex in a BiP-dependent manner 24 . Therefore, we examined whether BiP is involved in the interaction between DDRGK1 and IRE1a. The results showed that the interaction between DDRGK1 and IRE1a was not affected by knockdown of BiP ( Supplementary Fig. 4). To further confirm the interaction of DDRGK1 with IRE1a, we performed immunofluorescence staining experiments. The results showed that these two proteins co-localized in the ER (Fig. 5c). To map the region involved in the interaction of IRE1a with DDRGK1, we co-transfected Flag-IRE1a wild-type and truncation mutants along with DDRGK1 into HEK293T cells, and the IP results showed that the cytoplasmic kinase domain of IRE1a was necessary for its interaction with DDRGK1 ( Fig. 5d). To understand the dynamic changes in the interaction between DDRGK1 and IRE1a under ER stress conditions, HEK293T cells were transfected with Flag-DDRGK1 and treated with Tg for different time points. As expected, we observed that total IRE1a, p-IRE1a and BiP were significantly increased under ER stress. Interestingly, we found that DDRGK1 specifically interacted with unphosphorylated IRE1a, but not with p-IRE1a, in the immunoprecipitates (Fig. 5e). These results suggest that DDRGK1 interacts with and maintains the stability of non-phosphorylated IRE1a.
Ufm1 is required for the interaction of DDRGK1 and IRE1a.
A recent study demonstrated that DDRGK1 is an ufmylation substrate and is required for ASC1 ufmylation 15,20 . To investigate whether ufmylation is involved in the regulation of IRE1a stability by DDRGK1, we examined whether Ufm1 depletion affects IRE1a protein expression. Similar to the depletion of DDRGK1, we observed that knockdown of Ufm1 expression dramatically reduced the levels of IRE1a protein (Fig. 6a), indicating that DDRGK1 regulates IRE1a through ufmylation. Consistently, we found that over-expression of DDRGK1 failed to stabilize the IRE1a protein in Ufm1-knockdown MCF7 cells compared with control cells (Fig. 6b). In addition, IRE1a protein levels were dramatically decreased in Ufm1-knockdown MCF7 cells in both the absence and presence of Tg (Fig. 6c), which suggests that ufmylation is required for the regulation of IRE1a protein stability by DDRGK1. We then wondered whether IRE1a is subject to ufmylation modification. To address this question, we detected the ufmylation of IRE1a in cells expressing DDRGK1 and Ufm1 as well as UBA5 (E1), UFC1 (E2) and UFL1 (E3), as previously described 20 . However, we were unable to detect any ufmylation of the IRE1a protein, although ufmylation of the DDRGK1 protein was readily detectable, as previously described (data not shown) 15,20,25 , suggesting that IRE1a might not be a target of ufmylation. Interestingly, we found that the association between DDRGK1 and IRE1a was significantly decreased in Ufm1-knockdown HEK293T cells compared with control cells (Fig. 6d). Consistent with this observation, we found that  ARTICLE the interaction of DDRGK1 K267R (a DDRGK1 mutant deficient in ufmylation, in which the lysine residue 267 was substituted with an arginine residue) 15 with IRE1a was dramatically decreased compared with the wild-type DDRGK1 (Fig. 6e), indicating that ufmylation of DDRGK1 is required for the association between DDRGK1 and IRE1a. Furthermore, DDRGK1 K267R failed to stabilize the IRE1a protein compared with wild-type DDRGK1 (Fig. 6f). Taken together, these results suggest that ufmylation of DDRGK1 at K267 is required for its interaction with IRE1a and the regulation of IRE1a protein stability.

DDRGK1 is required for HSC reconstitution ability in mice.
A recent study suggested that hematopoietic stem cells (HSCs) are extremely sensitive to ER stress 26 . This led us to speculate that DDRGK1 might be critical in HSC function. We first determined the expression level of DDRGK1 across multiple hematopoietic lineages from 2-month-old wild-type mice. Q-PCR revealed a significantly higher level of DDRGK1 expression in HSCs, including long-term HSCs (LT-HSCs), short-term HSCs (ST-HSCs) and multipotent progenitors (MPPs) (Fig. 7a). In addition, the protein level of DDRGK1 was highly expressed in HSC-enriched Lineagec-Kit þ bone marrow (BM) cells compared with Lineage þ c-Kit -BM cells (Fig. 7b). These observations suggest an important role of DDRGK1 in stem cells. To examine the role of DDRGK1 in HSC function, competitive transplantation experiments were performed with HSCs following lentiviral knockdown of DDRGK1 (Fig. 7c). Lineage -Sca1 þ c-Kit þ (LSK) cells from CD45.1 donor mice were transduced with the control or DDRGK1knockdown lentivirus and transplanted into lethally irradiated CD45.2 recipient mice. Flow cytometric analysis showed that the percentage of donor-derived peripheral blood (PB) cells was significantly reduced in the DDRGK1-knockdown group compared with the control, indicating that knockdown of DDRGK1 significantly impaired the competitive reconstitution ability of HSCs (Fig. 7d). While in a non-competitive transplantation experiment, the recipient mice transplanted with DDRGK1-knockdown HSCs were still alive 3 months after transplantation, this suggests that knockdown of DDRGK1 was merely compromising the competitive reconstitution capacity of HSCs. To exclude off-target effects, we designed another shRNA targeting DDRGK1 and observed similar results ( Supplementary Fig. 5A). Three months after transplantation, the recipient mice were euthanized for BM analysis and in vitro colony-forming assays. Flow cytometric analysis showed that knockdown of DDRGK1 dramatically impaired the maintenance of transduced HSCs without affecting their lineage potential (that is, myeloid, B cell and T cell lineages), as indicated by a decreased frequency of donor-derived hematopoietic stem and progenitor cells, as well as differentiated hematopoietic cells in the BM and PB (Fig. 7e). A similar result was observed in an independent experiment using another shRNA targeting DDRGK1 ( Supplementary Fig. 5B). In addition, we isolated donor-derived CD34 -Flt3 -LSK cells (LT-HSCs) from the recipient mice and performed a single-cell colony-forming assay.
The results showed that knockdown of DDRGK1 impaired the ability of HSCs to form intermediate and large colonies (Fig. 7f). Altogether, these data suggest that DDRGK1 is required for the maintenance of HSC function.
To examine whether impaired HSC function was because of reduced homing efficiency, we labelled control or sh-DDRGK1 lentiviral transduced LSK cells purified from 2-month-old mice with fluorescent dye (violet), and injected them into lethally irradiated recipients. The percentage of labelled LSK cells present  Fig. 6A and B). To determine whether knockdown of DDRGK1 had an effect on the cell cycle status of LSK cells, we stained donor-derived control and DDRGK1-knockdown LSK cells with the proliferation marker Ki67 and DAPI and found no significant difference in the cell cycle profiles between DDRGK1-knockdown HSCs and control HSCs (Supplementary Fig. 7A). We then examined the effect of DDRGK1-knockdown on HSCs apoptosis using Annexin V staining and found a significantly higher frequency of apoptosis in donor-derived DDRGK1-knockdown LSK cells compared with control LSK cells; this phenomenon was observed with both of the shRNAs targeting DDRGK1 (Fig. 8a; Supplementary Fig. 7B). Next, we examined the levels of reactive oxygen species (ROS) in the transduced HSCs. The results showed that the ROS level was comparable between the control and DDRGK1-knockdown groups (Supplementary Fig. 7C).
On the basis of above observations, we concluded that knockdown of DDRGK1 in HSCs induced apoptotic cell death, thereby impairing HSC function.
To investigate whether the impaired HSC function was because of activation of the ER stress response in DDRGK1-knockdown HSCs, we analysed the expression levels of BiP and CHOP by Q-PCR in control and DDRGK-knockdown engrafted HSCs, the results showed that the mRNA levels of BiP and CHOP were significantly increased in DDRGK1-knockdown HSCs (Fig. 8b). In addition, we examined the protein levels of ER stress sensors in the control and DDRGK-knockdown Lineagec-Kit þ BM cells. In accordance with the observation in cancer cell lines, knockdown of DDRGK1 showed decreased protein levels of IRE1a and p-IRE1a and an increased level of p-PERK, whereas the ATF6 levels did not differ in the control and DDRGK1-knockdown groups (Fig. 8c). Consistently, the level of XBP1s was decreased in donor-derived DDRGK1knockdown LSK cells (Fig. 8d). Taken together, these results indicate that knockdown of DDRGK1 impaired IRE1a signalling but activated the PERK pathway and further support that DDRGK1 is critical for the proper maintenance of ER homoeostasis and thereby essential for the survival and maintenance of HSCs.

Discussion
In the present study, we provide evidence that DDRGK1 is involved in regulating ER homoeostasis both in vitro and in vivo. We demonstrate that depletion of DDRGK1 induces ER stress and enhances ER stress-induced apoptosis in both cancer cells and HSCs. Our results indicate that DDRGK1 controls IRE1a protein stability via its interaction with the kinase domain of IRE1a, which is dependent on ufmylation at K267 of DDRGK1. Loss of DDRGK1 accelerates IRE1a protein degradation and promotes PERK activation, thereby inducing cell death. Thus, these findings establish the essential role of DDRGK1 and Ufm1 modification in maintaining ER homoeostasis and provide critical insight into the regulatory mechanisms of ER homoeostasis, as illuminated in Fig. 9.
Different physiological and pathological perturbations interfere with protein folding processes, leading to ER stress. There is substantial evidence for the involvement of chronic ER stress in many diseases, including diverse forms of cancer, neurodegeneration, diabetes and proinflammatory conditions 27 . The UPR is an adaptive response that facilitates protein folding, processing, export and degradation during ER stress. Although the initial phase of the UPR serves as a cytoprotective mechanism to rebalance ER homoeostasis, persistent activation of the UPR triggers apoptosis 2,28,29 . One of our important findings is that DDRGK1 regulates the steady-state levels of the key ER stress sensor IRE1a. IRE1a has long been considered  ARTICLE a positive regulator of cell survival, and its activities are attenuated by persistent ER stress; however, PERK signalling, including translational inhibition by activation of eIF2a and induction of the proapoptotic transcription regulator CHOP, is maintained 7 . Interestingly, we observed that depletion of DDRGK1 resulted in the activation of PERK signalling ( Fig. 3; Supplementary Fig. 2), which has similar effects on the UPR as IRE1a depletion. Furthermore, ectopic expression of IRE1a rescued DDRGK1-depleted cells from cell death (Fig. 4). These results imply that DDRGK1 depletion induces IRE1a degradation, which leads to activation of the UPR and apoptotic cell death. Ufm1 is a recently identified protein modifier 25 , and the targets and physiological functions of the Ufm1 system are largely unknown. Our results show that ufmylation participates in the regulation of IRE1a protein stability by DDRGK1. Although   ufmylation might not be responsible for IRE1a degradation, it is required for the interaction of DDRGK1 with IRE1a. Because the ufmylation of DDRGK1 on K267 is required for the association between DDRGK1 and IRE1a, substitution of a lysine residue with an arginine residue at position 267 (DDRGK1 K267R) impaired the interaction between DDRGK1 and IRE1a. The remaining question is how this interaction affects IRE1a degradation. One possibility could be that the interaction of DDRGK1 with IRE1a counteracts ubiquitylation and subsequently inhibits the ERAD-mediated degradation of IRE1a (ref. 23). It is also possible that the interaction of DDRGK1 with IRE1a might affect its endoribonuclease and kinase activity. However, the detailed mechanisms for the regulation of IRE1a protein stability by DDRGK1 and ufmylation require further investigation. Nonetheless, our results demonstrate that DDRGK1 plays a critical role in the maintenance of ER homoeostasis by modulating IRE1a protein stability. Consistent with in vitro studies in cancer cell lines, our in vivo results indicate that DDRGK1 is critical for the survival of HSCs, as knockdown of DDRGK1 leads to elevated ER stress and apoptosis in HSCs. It has been reported that HSCs and other progenitor cells exhibit distinct cellular responses to extracellular ER stressors, and the UPR plays a critical role in governing the integrity of the HSC pool during stress 26 . Previous studies suggested that the Ufm1 conjugation system is essential for murine embryonic erythropoiesis 30,31 . A recent study showed that UFBP1 (DDRGK1) knockout mice exhibited defective erythroid development and somatic ablation of UFBP1 impaired adult hematopoiesis 32 , which is consistent with our observation that depletion of DDRGK1 impaired HSC function via increased ER stress and activation of the UPR. Importantly, our results from both in vitro and in vivo studies provide mechanistic evidence that DDRGK1 plays a critical role in ER homoeostasis. We demonstrate that DDRGK1 regulates IRE1a protein stability via its interaction with IRE1a, which is dependent on ufmylation. Given that DDRGK1 and the Ufm1 system are mostly associated with subcellular membranes and are subject to ufmylation, it is conceivable that DDRGK1-mediated ufmylation may constitute a novel protein network that is broadly involved in cellular homoeostasis. Further studies will elucidate the regulatory mechanisms and cellular functions of these large protein complexes and their implications in human diseases.
Cell line authentication. STR DNA profiling analysis was used for cell line authentication. MCF7 and HepG2 cell lines were found 100% matched with MCF7 and HepG2 cell lines in the ATCC and DSMZ databases, no crosscontamination of other human cells was found. The mycoplasma status of cells was tested and no mycoplasma contamination was found.
Plasmid constructs and transfection. The mammalian expression vector p3 Â FLAG-CMV (Sigma) was used to express Flag-tagged DDRGK1. A lysine-to-arginine mutation in DDRGK1 was generated by site-directed mutagenesis at residue Lys-267. pCMV-Tag4B-IRE1a wild-type and IRE1a deletion mutants were kindly provided by Dr Yong Liu (Key Laboratory of Nutrition and Metabolism, Institute for Nutritional Sciences, Shanghai Institutes for Biological Sciences, Chinese Academy of Sciences). Lentiviral vectors expressing specific shRNAs were constructed using the SF-LV-EGFP vector. The following shRNAs were used: mDDRGK1 shRNA-1: 5 0 -GGTGTTAGCGAAACCAT GACT-3 0 ; mDDRGK1 shRNA-2: 5 0 -GGCGGGAGCACGAGGAGTACC-3 0 . All of the constructs were verified by DNA sequencing. Plasmid transfection was performed with Turbofect (Fermentas), and RNA interference was performed with Lipofectamine 2,000 transfection reagent (Invitrogen), according to the manufacturer's instructions.
Lentivirus production. Lentiviruses were produced in HEK293T cells after the transfection of lenti-shRNA constructs with pspAX2 and pMD2G. The viruses were concentrated by centrifugation at 25,000 r.p.m. for 2.5 h at 4°C, and the virus pellet was subsequently suspended in SFEM medium.
Q-PCR and reverse transcription PCR (RT-PCR). Total RNA was extracted from cells using the TRIzol reagent/RNeasy Micro kit (Invitrogen) according to the manufacturer's instructions, followed by cDNA preparation using M-MLV reverse transcriptase (Promega). Q-PCR was performed in duplicate using SYBR Green Supermix (Bio-Rad). RT-PCR was performed using KOD-Plus-Neo Polymerase mix (Toyobo Life Science). The RT-PCR products were separated by electrophoresis in 3% or 1% agarose gels and were visualized using ethidium bromide. The sequences of the primers used for Q-PCR/RT-PCR are described in Supplementary  Table 1.
Single-cell colony-forming assay. Freshly isolated CD34 -Flt3 -LSK (LT-HSCs) cells were individually sorted into 96-well plates and cultured for 14 days in liquid medium. The colonies were counted and photographed.
HSC reconstitution assay. LSK cells were purified from donor mice and transduced with sh-DDRGK1 or sh-Vector lentivirus. The reconstitution analysis was performed by transplanting the infected cells into lethally irradiated recipient mice every 4 weeks after transplantation. PB was collected from the recipient mice and analysed for chimerism and lineage distribution. The BM of the recipient mice was analysed 3 months after transplantation.
Statistical analysis. Experiments were repeated at least three times. Statistical analyses were performed using a two-tailed, paired Student's t-test. Po0.05 was considered significant. *Po0.05, **Po0.01, and ***Po0.001. NATURE COMMUNICATIONS | DOI: 10.1038/ncomms14186 ARTICLE Data availability. The data that support the conclusions of this study are available from the corresponding author on request.