Coupling of COPII vesicle trafficking to nutrient availability by the IRE1α-XBP1s axis

Significance

One-third of the mammalian proteome is transported by the cytoplasmic coat protein complex-II (COPII) secretory vesicles. However, how this core coat machinery is regulated to meet the metabolic demand in response to alterations of the nutritional state remains largely unexplored. Here, we show that COPII vesicle trafficking is highly dynamic and responsive to nutrient availability fluctuations. We also uncover that the nutrient-sensing inositol-requiring enzyme 1α (IRE1α)–transcription factor X-box binding protein 1 (XBP1s) axis links COPII-mediated trafficking to nutrient availability. Furthermore, restoration of XBP1s in mice lacking hepatic IRE1α activates COPII-dependent lipoprotein traffic and reverses hepatosteatosis and hypolipidemia. Hence, we reveal a mechanism for the orchestration of COPII vesicle trafficking in response to nutrient availability.

Abstract

The cytoplasmic coat protein complex-II (COPII) is evolutionarily conserved machinery that is essential for efficient trafficking of protein and lipid cargos. How the COPII machinery is regulated to meet the metabolic demand in response to alterations of the nutritional state remains largely unexplored, however. Here, we show that dynamic changes of COPII vesicle trafficking parallel the activation of transcription factor X-box binding protein 1 (XBP1s), a critical transcription factor in handling cellular endoplasmic reticulum (ER) stress in both live cells and mouse livers upon physiological fluctuations of nutrient availability. Using live-cell imaging approaches, we demonstrate that XBP1s is sufficient to promote COPII-dependent trafficking, mediating the nutrient stimulatory effects. Chromatin immunoprecipitation (ChIP) coupled with high-throughput DNA sequencing (ChIP-seq) and RNA-sequencing analyses reveal that nutritional signals induce dynamic XBP1s occupancy of promoters of COPII traffic-related genes, thereby driving the COPII-mediated trafficking process. Liver-specific disruption of the inositol-requiring enzyme 1α (IRE1α)–XBP1s signaling branch results in diminished COPII vesicle trafficking. Reactivation of XBP1s in mice lacking hepatic IRE1α restores COPII-mediated lipoprotein secretion and reverses the fatty liver and hypolipidemia phenotypes. Thus, our results demonstrate a previously unappreciated mechanism in the metabolic control of liver protein and lipid trafficking: The IRE1α-XBP1s axis functions as a nutrient-sensing regulatory nexus that integrates nutritional states and the COPII vesicle trafficking.

The cytoplasmic coat protein complex-II (COPII) is evolutionarily conserved secretory machinery that is essential for cellular protein and lipid trafficking through cargo sorting and vesicle formation at the endoplasmic reticulum (ER) (1–4). The vast majority of proteins and lipids exported from the ER require the COPII secretory machinery. The assembly of COPII-coated vesicles for facilitating the transport of cellular cargos has been demonstrated to be a highly complex process (1–6). Activated small GTPase SAR1 localizes to the specialized ER exit sites and initiates the COPII coat assembly, by first recruiting the inner coat formed by the heterodimer SEC23/SEC24, followed by the outer coat heterotetramer SEC13/SEC31, to deform the ER membrane and eventually produce carrier vesicles (2, 4, 7–9). Mutations in COPII components or accessory factors have been implicated in several human genetic diseases, including chylomicron retention disease, congenital dyserythropoietic anemia type II, and cranio-lenticulo-sutural dysplasia (10–14). However, it remains largely unexplored how the COPII machinery is regulated to meet the cellular secretory demand in response to various physiological stimuli.

As a metabolically active tissue, the liver possesses a remarkable adaptive capacity to secrete lipids and proteins according to physiological fluctuations of nutrient availability. COPII vesicle trafficking in the liver is of particular importance in systemic metabolic homeostasis. Bulk lipid transport from the liver relies on the apolipoprotein B (ApoB)-containing very low-density lipoproteins (VLDLs) under conditions of excess liver cholesterol and fatty acid synthesis or upon increased peripheral lipid demands. The assembly of VLDLs starts in the ER, where ApoB is cotranslationally translocated into the ER lumen and acquires triglycerides and cholesterol when these lipids are supplied from biosynthesis or storage places such as the lipid droplets (15, 16). ApoB exists in two predominant forms, the full-length ∼500-kDa ApoB-100 and the truncated ∼240-kDa ApoB-48 due to mRNA editing (17). Lipoproteins represent a unique class of secretory cargos for the COPII complex; in particular, it remains unclear how the COPII machinery adapts to the fluctuation of VLDLs in response to nutrient signals (18). Nevertheless, numerous studies in mice, and in humans, have shown a close association between the derangements in liver ApoB-containing VLDL secretion and the development of liver steatosis (19–23).

Changes in nutrient availability elicit metabolic adaptations that require coordinated alterations in the profile of COPII cargos in the liver. Intracellular nutrient-sensing and COPII-mediated trafficking machinery needs to be precisely coupled for maintaining hepatic metabolic homeostasis. The ER can be viewed as a critical metabolic-sensing site that orchestrates glucose, lipid, and protein metabolism (24, 25). In mammals, increases in the protein folding demand or perturbations in lipid compositions at the ER trigger the activation of unfolded protein response (UPR) signaling (26–28). The inositol-requiring enzyme 1α (IRE1α) is an ER-localized transmembrane Ser/Thr protein kinase and endoribonuclease, and serves as a key signal transducer of the UPR (26–29). Activation of IRE1α results in the production of the spliced active form of the transcription factor X-box binding protein 1 (XBP1s) through nonconventional splicing of its mRNA, thereby initiating a major UPR program (30, 31). Many recent studies have revealed multifaceted, context-dependent functions of the IRE1α pathways during metabolic ER stress, which has been critically linked to obesity, insulin resistance, and hepatic steatosis (32–40). While the IRE1α-XBP1s pathway has been shown to have important roles in maintaining ER homeostasis in professional secretory cells (27), it has yet to be unraveled whether this UPR signaling arm is implicated in controlling cargo sorting and vesicle formation at the ER.

In this study, we investigated the nutrient-sensing function of the IRE1α-XBP1s pathway in association with COPII vesicle trafficking upon physiological nutrient fluctuations. We found that in response to nutrient availability, COPII-mediated vesicle trafficking dynamically parallels the XBP1s activity in both live cells and mouse livers. Using live-cell imaging approaches, together with gain- and loss-of-function strategies in mice, we demonstrate a crucial role for the IRE1α-XBP1s axis in the coordinated control of the COPII vesicle trafficking in the liver.

Results

COPII Vesicle Trafficking Dynamically Parallels XBP1s Activity upon Physiological Nutrient Fluctuations.

As an initial step, we explored whether the COPII vesicle trafficking is responsive to nutritional cues with a COPII-mediated trafficking reporter system in live cells. The retention using selective hooks (RUSH) COPII reporter system [C-terminal ER retention signal (Lys-Asp-Glu-Leu; KDEL) hook and ManII fused to green fluorescent protein (GFP)] (41) was used for real-time monitoring of the trafficking of COPII-coated vesicles from the ER to the Golgi at a physiological temperature (Fig. 1A). Consistent with previously published data (41), live-cell time-lapse imaging revealed a synchronous release of ManII-GFP reporter from the ER into a perinuclear locale, the Golgi complex, upon application of biotin (Fig. 1 B and C). At ∼18 min after biotin addition, the majority (83%) of the ManII-GFP signal was clearly present in the Golgi complex, whereas the GFP signal in the ER strongly decreased, indicating that the COPII reporters trafficked from the ER to the Golgi (Fig. 1 B and C). We then analyzed COPII secretion in live cells that were subjected to nutrient deprivation for 12 h. As shown in Fig. 1 B and C, quantitative microscopy of the ManII-GFP trafficking kinetic revealed that the ER-to-Golgi traffic rates were significantly lower in starved cells compared with nonstarved controls, with a large amount of GFP signal detected in the ER in starved cells at 18 min and remaining up to 30 min after biotin application (Fig. 1 B and C).

Fig. 1.

COPII vesicle trafficking dynamically parallels XBP1s activity upon nutrient availability fluctuations. (A, Top) Construction map of the RUSH COPII reporter system [the KDEL hook and the ManII fused to GFP reporter were expressed under a shared cytomegalovirus promoter (pCMV)]. (A, Bottom) Schematic showing the use of the RUSH reporter system to monitor the trafficking of the COPII-coated vesicles in live cells. Bio, biotin; Str, streptavidin; SBP, streptavidin binding protein. Under basal conditions, the reporter is retained in the ER by the hook. Application of biotin induces a synchronous release of the reporter from the ER into the Golgi apparatus. (B) Time-lapse images of COS7 cells expressing the KDEL-ManII-GFP reporter at different time intervals after biotin (40 μM) addition. At 24 h posttransfection, cells were treated for 12 h with complete culture medium (Top), starvation (Starv) medium (low glucose, DMEM only) (Middle), or Starv medium plus 200 μM oleic acid (OA) (Bottom). (Scale bars, 50 μm.) (C) Quantitative analysis of the GFP fluorescence intensity in the Golgi region in B at the indicated time point, after correction for background and normalization to the maximum value. Curves depict the measurement of 10–12 cells from four independent experiments. AU, arbitrary units. (D) RT-qPCR analysis of the indicated genes in COS7 cells. Cells were starved or treated with Starv medium plus 200 μM OA for 12 h (n = 4 independent experiments). (E, Left) In vitro COPII budding assay using liver cytosol extracted by ultracentrifugation from WT mice subjected to fasting or refeeding and membrane extracts from HEK293A cells. The resulting vesicle fractions and permeabilized cell membrane (input) were subjected to immunoblotting. Mice were fasted at 6:00 PM on day 1 and harvested at 10:00 AM on day 2, or followed by 24 h of refeeding and harvested at 10:00 AM on day 3. Representative immunoblots are shown. (E, Right) Quantification of LMAN1, SEC22B, and SEC24A packaged into the COPII-coated vesicles after normalization to the fed control (n = 5 independent experiments). (F) RT-qPCR of liver gene expression of WT mice subjected to a 16-h fast or 2-h refeeding. Mice were fasted at 6:00 PM on day 1, followed by 2 h of refeeding at 6:00 PM, and harvested at 8:00 PM on day 2 (n = 5–6 mice per group). Data are presented as mean ± SEM. *P < 0.05 vs. corresponding controls; ^‡P < 0.05 vs. Starv or fasted.

To identify the nutrient signals involved in the regulation of COPII traffic, we supplemented the starved cells with fatty acids and then examined the kinetic release of the ManII-GFP reporter. Interestingly, treatment with fatty acids significantly enhanced COPII traffic and markedly diminished the suppressive effects of nutrient deprivation (Fig. 1 B and C). These results demonstrate that COPII vesicle trafficking is a dynamic process that is tightly coupled to cellular nutrient availability.

The ER can be viewed as a key integrator of intracellular metabolic responses during nutrient fluctuations. Among the regulatory circuits controlling the metabolic capacity of the ER, the IRE1α-XBP1s pathway of the UPR has been shown to fine-tune many cellular metabolic processes in response to nutrient status. Indeed, we observed a dynamic correlation between COPII-mediated trafficking with the expression of Xbp1s upon nutrient stimuli in cultured cells. The mRNA expression of Xbp1s was markedly repressed in starved cells, which could be significantly reversed by fatty acid supplementation (Fig. 1D). By contrast, the Atf6 mRNA levels were induced in starved cells, which were not affected by fatty acid treatment (Fig. 1D).

We next performed an in vitro reconstitution assay to test directly the packaging capacity of the COPII coats in liver cytosol prepared from mice under different feeding conditions. When liver cytosolic fractions from ad libitum-fed mice were applied to semi-intact 293A cells in the presence of ATP regeneration system (ATPr) and GTP, COPII cargos such as LMAN1 and SEC22B were recovered from the vesicle fractions, along with the COPII subunit SEC24A (Fig. 1E and SI Appendix, Fig. S1 A and B), suggesting efficient cargo packaging. However, this packaging process was markedly repressed when using liver extracts prepared from fasted mice (Fig. 1E and SI Appendix, Fig. S1 A and B). Moreover, liver cytosolic fractions from refed mice corrected such deficiency in vesicle budding to produce the highest level of packaging of COPII cargos (Fig. 1E and SI Appendix, Fig. S1 A and B). Thus, the reconstitution data revealed that hepatic COPII packaging capacity is tightly regulated by the feeding status. Notably, expression of COPII component SEC23B protein, but not SEC24A protein, was significantly down-regulated in fasted liver cytosol compared with fed controls (SI Appendix, Fig. S1 A and C). In contrast to that of Atf6, the expression of Xbp1s, as well as its transcriptional target gene Erdj4, was also suppressed in fasted liver, which was significantly reversed upon refeeding (Fig. 1F). Together, these results indicate that COPII vesicle trafficking could dynamically parallel XBP1s activity during physiological fluctuations of nutrient availability.

The IRE1α-XBP1s Pathway Controls COPII Vesicle Trafficking.

To test whether the IRE1α-XBP1s pathway is functionally connected to COPII-mediated trafficking, we examined the effect of XBP1s overexpression upon the kinetic COPII trafficking using the ManII-mCherry reporter in live cells. As shown in Fig. 2 A and B, XBP1s overexpression resulted in dramatically accelerated ManII-mCherry export (threefold faster) out of the ER. Remarkably, in cells overexpressing an active XBP1s, the ManII-mCherry started reaching the Golgi complex within 6 min after biotin addition, and at 12 min, nearly all ManII-mCherry reporters were clearly present in the Golgi complex (Fig. 2 A and B and Movie S1), whereas in control cells, relocalization of ManII-mCherry from the ER to Golgi was observed at 21 min after biotin addition (Fig. 2 A and B and Movie S2). Moreover, overexpression of XBP1s was sufficient to block the suppressive effects of nutrient deprivation on COPII traffic, as the trafficking rate of ManII-mCherry reporter from the ER to the Golgi remained high in XBP1s-overexpressing cells even under the starved state (Fig. 2 A and C). Together, these data demonstrate that overexpression of XBP1s is sufficient to promote COPII traffic and blunts its suppression during nutrient deprivation.

Fig. 2.

XBP1s directly controls COPII vesicle trafficking. (A) Time-lapse images of COS7 cells with the KDEL-ManII-mCherry reporter at the indicated time after biotin (80 μM) treatment. Cells were mock-transfected or transfected with the XBP1s-GFP plasmid. At 24 h posttransfection, cells were treated for 12 h with complete culture medium (Control) or starvation (Starv) medium (low glucose, DMEM only). (Scale bars, 50 μm.) (B and C) Quantitative analysis of the red fluorescence intensity in the Golgi region in A after correction for background and normalization to the maximum value. Curves depict the measurement of eight to 12 cells from three independent experiments. AU, arbitrary units. (D) In vitro COPII budding assay using cytosol extracted by ultracentrifugation from WT and IRE1α LKO livers and membrane fractions from HEK293A cells. The resulting vesicle fractions and permeabilized cell membrane (input) were analyzed by immunoblotting, and representative results are shown. (E) Quantification of LMAN1, SEC22B, and SEC24A packaged into the COPII-coated vesicles in D after normalization to the WT control (n = 4 independent experiments). Data are presented as mean ± SEM. *P < 0.05 vs. corresponding controls; ^‡P < 0.05 vs. control + vector.

To further affirm the dependence on XBP1s of COPII trafficking from the ER, we performed an in vitro reconstitution assay with semi-intact HEK293 cells using liver cytosols prepared from wild-type (WT) versus liver-specific IRE1α knockout (IRE1α LKO) mice. In comparison to its WT counterpart, the packaging of LMAN1 and SEC22B into in vitro-synthesized COPII vesicles was significantly reduced when using IRE1α LKO hepatic cytosol (Fig. 2 D and E and SI Appendix, Fig. S1 D and E). A similar reduction was also observed for SEC24A (Fig. 2 D and E and SI Appendix, Fig. S1 D and E), suggesting a general reduction in the packaging capacity of the COPII complex upon loss of the IRE1α-XBP1s cascade. This is consistent with the XBP1s gain-of-function results using the in vitro COPII-mediated trafficking assay. These data further demonstrate that XBP1s is necessary for maintaining the COPII traffic capacity in the liver.

Dynamic XBP1s Occupancy in Metabolic Regulation of COPII Secretory Genes.

To determine whether hepatic XBP1s is directly involved in the metabolic control of COPII traffic and has a requisite role in liver metabolic reprogramming during nutrient deficiency, we examined the genome-wide XBP1s occupancy on mouse liver chromatin by chromatin immunoprecipitation (ChIP) coupled with high-throughput DNA sequencing (ChIP-seq) (SI Appendix, Fig. S2A). We prepared liver chromatin from mice under ad libitum feeding or subjected to a 16-h fast. Oil red staining showed a marked increase in liver triglyceride (TG) accumulation, representing a typical fasting response in the liver (SI Appendix, Fig. S2B). Sequencing of XBP1s chromatin-immunoprecipitated DNA by an Illumina HiSeq 2500 system generated a dataset of 20–50 million reads in each sample (SI Appendix, Fig. S2 C and D). To evaluate the quality of genomic XBP1s-binding peaks, we conducted a peak distribution analysis relative to a known transcriptional start site (TSS) from the UCSC Genome Browser and observed a significant enrichment of XBP1s-binding sites around the TSS under both fed and fasted states (SI Appendix, Fig. S2E), while no overall changes were detected in XBP1s preferential binding between fed and fasted livers (SI Appendix, Fig. S2F).

Comparison of XBP1s-binding peaks revealed considerable differences in XBP1s genomic occupancy in the fed and fasted livers, as shown in Fig. 3A. Of a total 7,923 XBP1s-binding regions, 6,029 (76%) “Unchanged” XBP1s sites were found (Fig. 3B). However, 1,473 (19%) “Fed-high” and 421 (5%) “Fasted-high” XBP1s sites were identified to be significantly altered by different feeding conditions (Fig. 3B), suggesting the dynamic XBP1s occupancy in response to nutrient availability. The search for enriched motifs within the 1,473 “Fed-high” XBP1s sites using Homer revealed a sequence element, 5′-CGTCACGT-3′, as the highest score motif with P = 1e⁻¹⁰³ (Fig. 3C), whereas that within the top 1,000 “Unchanged” sites revealed 5′-GTCACGTC-3′ as the highest score motif, with P = 1e⁻⁴⁷ (Fig. 3C). By contrast, no specific enriched motif was found within the 421 “Fasted-high” XBP1s regions. Interestingly, both the “Fed-high” and “Unchanged” motifs closely corresponded to a consensus site that has been previously known as the XBP1s-binding site (42).

Fig. 3.

Dynamic liver XBP1s occupancy upon COPII traffic-related genes during changes of feeding states. (A) Identification of XBP1s binding peaks in livers through XBP1s antibody-based ChIP-seq. A tag heatmap shows the XBP1s binding signal intensity for fed or fasted mouse livers (4 kb wide, centered on the XBP1s peak). Liver chromatin was prepared from fed mice or after a 16-h fast. (B) Schematic of identification of “Fed-high,” “Unchanged,” or “Fasted-high” XBP1s sites with the cutoff criteria of a fed/fasted fold change greater than 1.5 (either direction) and a significant P < 0.0001. (C) De novo motif analysis of liver XBP1s binding regions using Homer. Shown is the top-scoring motif present in the 1,473 “Fed-high” (Top) and the top 1,000 “Unchanged” (Bottom) XBP1s sites. The motifs are displayed in a sequence logo which shows how well nucleotides are conserved at each position. (D) KEGG functional enrichment analysis of the “Fed-high,” “Unchanged,” or “Fasted-high” gene regions, with the top six pathways shown in the “Fed-high” and “Unchanged” datasets. Notably, the “Protein export” and “SNARE interactions in vesicular transport” terms are not significantly enriched in the “Unchanged” dataset (with statistically significant defined as a value of P < 0.01). (E) XBP1s binds the promoters of genes involved in almost every stage of the COPII-coated vesicle production, including markers of the ER exit sites and components of the inner/outer coats. (F) Representative XBP1s ChIP-seq peaks. XBP1s binding peaks were mapped onto the mm10 mouse genome and visualized using the Integrative Genomics Viewer tool. Input, genomic DNA from liver; WT-Fed, WT mice under ad libitum feeding; WT-Fasted, WT mice subjected to a 16-h fast; IRE1α LKO, mice with liver-specific IRE1α ablation (fed). The gray box indicates the high-confidence XBP1s binding regions corresponding to the Sec16a, Sec23b, Sec24c, and Sec31a genes.

Gene ontology (GO) analysis further revealed that the Kyoto Encyclopedia of Genes and Genomes (KEGG) term “protein processing in ER” was found in both “Fed-high” and “Unchanged” XBP1s-binding sites, but not in “Fasted-high” regions (Fig. 3D). XBP1s was found to bind to many known target genes involved in protein-folding homeostasis (SI Appendix, Fig. S3A). More interestingly, “Fed-high” XBP1s-binding regions also revealed genes associated with “protein export” and “SNARE vesicular transport,” which were not found in the “Unchanged” or “Fasted-high” regions (Fig. 3D). Furthermore, we found that XBP1s could bind to the promoter of genes involved in almost every stage of the COPII-coated vesicle production, including markers of the ER exit sites and components of the inner/outer coats (e.g., Sec16a, Sec23b, Sec24c, Sec31a, Sec24a, Sec24b, Sec24d, Sec31b, Sec13, Sec23a) (Fig. 3 E and F and SI Appendix, Fig. S3B). Compared with that in fed livers, we observed a significant decrease of XBP1s binding at the promoter of COPII genes (e.g., Sec16a, Sec23b, Sec24c, Sec31a, Sec24a, Sec13, Sec23a) in fasted livers (Fig. 3F and SI Appendix, Fig. S3B). As expected, XBP1s binding on these promoter regions was markedly diminished in the IRE1α LKO livers (Fig. 3F and SI Appendix, Fig. S3B). These results suggest that dynamic hepatic XBP1s occupancy controls the COPII secretory gene program in response to feeding conditions. Similar dynamic profiles of XBP1s binding were also observed for a broad array of other COPII-related early secretory and translocon component genes (SI Appendix, Fig. S3C). Notably, the significant binding of XBP1s in the fasted liver was verified by ChIP-seq with chromatin from IRE1α LKO liver (SI Appendix, Fig. S4 A–E). GO analysis confirmed that the top KEGG term of the “Fasted” XBP1s binding region was “protein processing in ER” (SI Appendix, Fig. S4E). No difference in XBP1s binding was observed in many “Unchanged” regions (SI Appendix, Fig. S4F), while the XBP1s binding on many “Fasted-high” regions was increased in fasted liver compared with fed controls (SI Appendix, Fig. S4G and Table S1). Taken together, these data suggest that fasting only alters liver XBP1s occupancy on select nutrient-responsive targets such as COPII genes, rather than an overall reduction of XBP1s occupancy across the genome.

We then affirmed the nutrient-responsive changes of the XBP1s-activated secretory gene program by qRT-PCR analysis. The mRNA expression of Sec23b, a key component of COPII-coated vesicle (8, 9), was robustly down-regulated in fasted livers (SI Appendix, Fig. S5A), as were the mRNA levels of other early secretory genes (e.g., Bet1, Stx5a) and translocon subunits (e.g., Sec61a1, Sec61b, Sec61g, Spcs2), as well as those known targets of Xbp1s (SI Appendix, Fig. S5A). Restoration of the expression of XBP1s in fasted liver reversed the expression of many COPII-related genes (SI Appendix, Fig. S5B). Next, we also assessed the relationship between the COPII-related early secretory gene expression and XBP1s activity during refeeding. Upon refeeding for 2 h after a 24-h fast, a significant increase in hepatic Xbp1s mRNA abundance was observed, and the mRNA expression of Sec23b and other ER protein transport genes, such as Spcs2, accompanied the up-regulated Xbp1s expression (SI Appendix, Fig. S5C). Liver-specific knockout of IRE1α further confirmed that the IRE1α-XBP1s axis is required for the full induction of the expression of Sec23b upon refeeding (SI Appendix, Fig. S5D). Thus, these results further link XBP1s-mediated transcriptional regulation of the COPII secretion program to nutrient availability in vivo.

Transcriptional Control of COPII Secretion Program by the IRE1α-XBP1 Pathway.

To ascertain the direct involvement of IRE1α, the upstream UPR sensor and regulator of XBP1, we carried out genome-wide mRNA expression analysis in livers of WT and IRE1α LKO mice. GO term KEGG pathway analysis showed that the global gene expression profiles affected by IRE1α abrogation were highly correlated with XBP1s-dependent transcriptional programs identified from XBP1 ChIP-seq, with the prominently down-regulated genes involved in the maintenance of ER function and protein export, the top two terms in “Fed-high” XBP1s binding genes (Fig. 4A). RT-qPCR analyses further confirmed significantly lower mRNA expression levels for an array of genes encoding COPII traffic-related (e.g., Sec16a, Sec23b, Sec24d, Sec22b, Bet1, Stx5a) and ER protein (e.g., Sec61a1, Sec61b, Sec61g, Sec11a, Sec11c, Spcs2) translocation in IRE1α LΚΟ liver (Fig. 4B). Similar reductions were also observed for the mRNAs encoding the XBP1s-regulated ER stress target genes (Fig. 4B). Consistently, decreased levels of SEC23B and SEC24D proteins, as well as SEC61B and LMAN1 proteins, were also detected in IRE1α LKO livers (Fig. 4 C and D), whereas SEC22B protein remained unchanged. To further ascertain the role of the COPII subunits in IRE1α-XBP1s–regulated ER export, we determined whether supplementing the IRE1α LKO liver cytosol with the inner coat SEC23/SEC24 can restore the COPII packaging capacity. SEC23A was recently shown to function interchangeably with SEC23B (43), and supplementing the IRE1α LKO liver cytosol with purified SEC23A/SEC24D significantly restored LMAN1 packaging into the COPII vesicles (SI Appendix, Fig. S6 A and B). In addition, when liver cytosol prepared from fasted mice was supplemented with purified SEC23A/SEC24D, an increase in the packaging efficiency of LMAN1 was observed (SI Appendix, Fig. S6C).

Fig. 4.

IRE1α-XBP1 pathway regulates the COPII secretory gene expression program. (A) GO term KEGG pathway analysis of gene transcripts down-regulated in IRE1α LKO livers identified a number of terms related to protein processing in the ER and protein export. RNA-seq data (n = 2 independent pools) were generated using livers of 8-wk-old male IRE1α LKO mice and their littermate controls (WT). (B) RT-qPCR analysis of the indicated genes in the livers (n = 5 mice per group). AU, arbitrary units. (C) Representative immunoblot analysis of liver lysates from the indicated mice using SEC23B, SEC24D, SEC22B, SEC61B, LMAN1, and GAPDH (control) antibodies. (D) Quantification of the SEC23B/GAPDH, SEC24D/GAPDH, SEC22B/GAPDH, SEC61B/GAPDH, and LMAN1/GAPDH in C after normalization to the WT control (n = 5–6 mice per group). Data are presented as mean ± SEM. *P < 0.05 vs. corresponding controls.

We further conducted quantitative XBP1s-ChIP analysis by PCR to affirm the XBP1s binding of the promoters of the core COPII genes as well as ER protein transport genes. XBP1s robustly bound to the promoter regions of Sec23b, Sec61a1, and Sec61b, which were markedly reduced in IRE1α LKO livers (SI Appendix, Fig. S7A). DNA sequence analysis of the proximal promoter regions of Sec23b, Sec61a1, and Sec61b identified several conserved putative XBP1s-binding elements (SI Appendix, Fig. S7 B and C). Similar XBP1s-binding sites were also found in a number of other COPII secretion-related target genes (SI Appendix, Fig. S7C). To evaluate the functionality of the identified XBP1s-binding sites, we utilized cell-based luciferase reporter assays with promoter constructs for the mSec23b, mSec61a1, and mSec61b genes. Indeed, overexpression of XBP1s significantly activated the three promoters, and the XBP1s-mediated activation of the mSec61a1 promoter was dramatically attenuated when the proximal XBP1s site was mutated (SI Appendix, Fig. S7D). These results demonstrate that these COPII traffic-related genes are direct transcriptional targets of XBP1s.

Recently, another basic leucine zipper (bZIP) transcription factor, CREB3L2, was shown to regulate COPII gene expression (44–46). We also sought to determine whether XBP1s can cooperate with CREB3L2 in the regulation of the COPII secretion program. Interestingly, we found that XBP1s could occupy the promoters of both the Creb3l1 and Creb3l2 genes, and that fasting resulted in reduced XBP1s occupancy on Creb3l1/2 genes (SI Appendix, Fig. S8A), suggesting that Creb3l1 and Creb3l2 are direct targets of XBP1s. In vivo manipulation of the IRE1α-XBP1s axis in mouse liver further confirmed a critical role for XBP1s in regulating Creb3l1/2 gene expression (SI Appendix, Fig. S8 B and C). Moreover, Creb3l1/2 mRNA levels were significantly reduced in fasted liver, a pattern that mirrors the changes of XBP1s mRNA (SI Appendix, Fig. S8D). Reactivation of XBP1s in IRE1α LKO liver can rescue the expression of Creb3l1/2 genes even under the fasted state (SI Appendix, Fig. S8D). We also conducted coimmunoprecipitation studies to determine whether XBP1s directly interacts with CREB3L2. The full-length form of CREB3L2 is cleaved upon stress to generate an active NH₂-terminal fragment (N) of CREB3L2, thereby initiating transcription of target genes (45). HEK293T cells were cotransfected with expression vectors for Flag-XBP1s and the hemagglutinin (HA)-N form of CREB3L2. Anti-HA was found to coimmunoprecipitate XBP1s and the N form of CREB3L2 (SI Appendix, Fig. S8E). Together, these data suggest that XBP1s serves to cooperate with CREB3L2 to regulate COPII gene expression in addition to activating Creb3l2 transcription.

To test whether the metabolic regulation by the IRE1α-XBP1 pathway of the COPII secretory program is related to conventional ER stress, we treated mice with the chemical ER stressor tunicamycin (TM). In accordance with previously documented studies (40), TM caused a significant increase in hepatic lipid accumulation (SI Appendix, Fig. S9A), while stimulating the expression of XBP1s and classical ER stress genes, such as Erdj4, Erdj5, Hspa5, Pdia6, Herpud1, Grp94, Dnajc3, PDI, and Gale, in the livers (SI Appendix, Fig. S9 B and C). However, this ER stress-induced activation of the IRE1α-XBP1 pathway did not result in higher expression of the COPII-related secretory genes; rather, suppressed mRNA levels were detected for Sec24a, Sec24b, Sec24c, and Sec31b in TM-treated livers (SI Appendix, Fig. S9D). These data suggest that metabolic activation of the IRE1α-XBP1 pathway possesses functionality distinct from that under experimental ER stress, which can sense nutrient availability to regulate the COPII secretory program. We next sought to determine whether the major nutrient-sensing pathways, such as autophagy and mammalian target of rapamycin (mTOR), are involved in IRE1α-XBP1–mediated regulation of COPII. Disruption of liver IRE1α resulted in increased conversion of LC3-I to LC3-II at the fed state but no further intensified fasting-mediated stimulatory effect (SI Appendix, Fig. S10A), indicating that autophagy is activated in IRE1α LKO liver in the absence of fasting. Levels of phosphorylated ribosomal S6 protein (p-S6) kinase, p-S6, and EIF4E-binding protein 1 (p-4EBP1), downstream products of mTORC1 activation, were not different in WT and IRE1α LKO liver under the fed or fasted state (SI Appendix, Fig. S10B). Interestingly, we found that autophagy inhibition using wortmannin failed to affect liver Sec23b and Erdj4 expression during fasting, despite reduction of fasting-induced conversion of LC3-I to LC3-II (SI Appendix, Fig. S10 C and D), while inhibition of mTORC1 activity using rapamycin resulted in reduced expression of Xbp1s, Sec23b, and Erdj4 upon refeeding (SI Appendix, Fig. S10 E and F). Together, these results suggest an autophagy-independent regulation of COPII by fasting and that the mTORC1 signaling could act upstream of XBP1s-COPII axis.

COPII-Mediated Lipoprotein Secretion Requires the IRE1α-XBP1s Pathway.

To identify candidate COPII secretory cargo proteins that were affected by disruption of the liver IRE1α-XBP1 pathway, we examined serum protein profiles from fasted WT or IRE1α LKO mice (Fig. 5A). Sodium dodecyl sulfate polyacrylamide gel electrophoresis analysis followed by silver staining detected no obvious changes in most abundant serum proteins (Fig. 5A), but revealed two proteins (∼250 kDa and ∼20 kDa) that were markedly reduced in IRE1α LKO serum compared with WT controls (Fig. 5A). Mass spectrometry analysis identified the ∼250-kDa protein as ApoB and the ∼20-kDa protein as ApoC4, both the structural components of lipoproteins. We previously showed that lipoproteins are preferred cargos delivered by the COPII machinery in vivo (7). Immunoblotting analysis revealed significant decreases in ApoB-48 and ApoB-100 levels in IRE1α LKO serum relative to WT controls under both fed and fasted states (Fig. 5 B and C and SI Appendix, Fig. S11A). Conversely, a significant accumulation of ApoB-48 was found in IRE1α LKO livers (Fig. 5 B and C). Moreover, ApoB-48 levels were increased in fasted livers compared with fed controls in WT mice, which was accompanied by a significant reduction in the expression of SEC23B protein (Fig. 5D). The accumulation of ApoB-48 protein that paralleled reduced expression levels of SEC23B was also observed in IRE1α LKO livers under both fed and fasted states (Fig. 5D). Together, these results indicate an essential role for the IRE1α-XBP1 pathway in COPII-mediated lipoprotein secretion in the liver.

Fig. 5.

COPII-mediated lipoprotein secretion requires the IRE1α-XBP1 pathway. (A, Left) Strategy for the identification by mass spectrometry of COPII cargo proteins that were affected by loss of liver IRE1α. KO, knockout; SDS-PAGE, sodium dodecyl sulfate polyacrylamide gel electrophoresis. (A, Right) Serum proteins were subjected to SDS/PAGE followed by silver staining. Arrows indicate the reduced abundance of serum proteins from IRE1α LKO mice relative to WT controls, with the ∼250-kDa band identified as ApoB and the ∼20-kDa band identified as ApoC4. (B) Representative immunoblot analysis of serum (Top) and liver lysates (Bottom) from the indicated mice after a 16-h fast using ApoB, IgG, or α-Tubulin (control) antibodies. Two predominant forms of ApoB protein, ApoB-100 and ApoB-48, are indicated. (C) Quantification of ApoB-48/IgG (Top) in serum samples and ApoB-48/tubulin (Bottom) in liver lysates in B (n = 4–6 mice per group). AU, arbitrary units. (D, Left) Representative immunoblots for liver protein extracts from the indicated mice after a 16-h fast using ApoB, SEC23B, and α-Tubulin (control) antibodies. (D, Right) Quantification of ApoB-48/tubulin and SEC23B/tubulin. Values represent mean (±SEM) AU normalized (=1.0) to the value of WT fed controls (n = 5–7 mice per group). *P < 0.05 vs. WT controls; ^‡P < 0.05 vs. WT-fasted.

Disruption of the IRE1α-XBP1 Pathway Leads to Defective COPII-Mediated Lipoprotein Secretion, Hepatosteatosis, and Hypolipidemia in IRE1α LKO Mice.

We then examined the metabolic significance of the regulation by the IRE1α-XBP1 pathway of COPII-mediated lipoprotein secretion. Following a 16-h fast, we observed a significant reduction (65%) in serum TG levels, but an ∼2.5-fold increase in hepatic TG accumulation in IRE1α LKO mice relative to their WT controls (SI Appendix, Fig. S11B). Consistently, transmission electron microscopy (TEM) analysis also showed increased numbers and sizes of lipid droplets in livers of fasted IRE1α LKO mice (SI Appendix, Fig. S11C). These results suggest that defective secretion of hepatic lipids contributes to the occurrence of liver steatosis and hypolipidemia in fasted IRE1α LKO mice.

Next, we determined whether restoration of XBP1s could correct the defective lipid secretion in IRE1α LKO liver. Adenovirus-mediated overexpression of XBP1s markedly decreased liver TG content (Fig. 6A) and increased serum TG levels (Fig. 6B) in IRE1α LKO mice to similar levels seen in WT control mice. This was accompanied by complete normalization of serum ApoB-48 and ApoB-100 levels that were decreased in fasted IRE1α LKO mice (Fig. 6C). In agreement with the critical role of XBP1s regulation of COPII secretory gene program, restored expression of XBP1s resulted in significantly increased expression of an array of COPII genes in IRE1α LKO livers (SI Appendix, Fig. S11D). Immunoblot analyses also showed that the decreased expression of SEC23B, SEC24D, and LMAN1, as well as the increased accumulation of ApoB-48 protein, was reversed by XBP1s expression in IRE1α LKO livers (Fig. 6 D and E). To further test the effect of increasing the COPII machinery on liver ApoB secretion, we determined whether overexpression of SEC23B can promote the COPII-mediated ApoB secretion in IRE1α LKO livers. Liver-specific overexpression of SEC23B using adeno-associated virus resulted in a significant increase in ApoB secretion in IRE1α LKO mice (SI Appendix, Fig. S12).

Fig. 6.

Restoration of XBP1s activates COPII-mediated ApoB secretion and reverses hepatosteatosis and hypolipidemia in IRE1α LKO mice. Total hepatic TG (A) and serum TG (B) from male WT or IRE1α LKO mice with adenoviral expression of EGFP or XBP1s (n = 6–8 mice per group) are shown. (C, Top) Representative immunoblot analysis of serum proteins from the indicated mice after a 16-h fast using ApoB or IgG (control) antibodies. (C, Bottom) Quantification of ApoB-48/IgG and ApoB-100/IgG (n = 5–7 mice per group). AU, arbitrary units. (D) Immunoblot analysis of liver lysates from the indicated mice using SEC23B, SEC24D, LMAN1, ApoB, and GAPDH (control) antibodies. (E) Quantification of SEC23B/GAPDH, SEC24D/GAPDH, LMAN1/GAPDH, and ApoB-48/GAPDH in D after normalization to the WT control (n = 3–6 mice per group). (F) Transmission electron micrographs of livers of WT and IRE1α LKO mice upon a 4-h fast. (Bottom) Magnified areas are shown. (Scale bars, 500 nm.) Lipoproteins are indicated by red arrows in the ER lumen for IRE1α LKO livers. M, mitochondrion. Representative images from five mice per group are shown. (G) Schematic depicts the proposed model for the nutrient-sensing IRE1α-XBP1 axis that links COPII-mediated lipoprotein secretion to nutrient availability. Data are presented as mean ± SEM. *P < 0.05 vs. WT controls; ^‡P < 0.05 vs. LKO.

To further affirm the defective hepatic COPII-mediated lipoprotein secretion in IRE1α LKO animals, we examined the liver specimens by TEM using imidazole staining to highlight lipoproteins. Few lipoprotein particles were detected in the WT liver sections showing the ER with normal morphology (Fig. 6F). By contrast, a large number of lipoprotein particles trapped within the lumen of dilated ER could be found in IRE1α LKO livers (Fig. 6F), indicating defective ER export of the lipid carriers. Conversely, adenoviral overexpression of XBP1s markedly reduced the number of lipoprotein particles in the ER lumen in IRE1α LKO livers (Fig. 6F). Together, these results demonstrate that the IRE1α-XBP1 pathway has an important role in the regulation of lipid homeostasis through promoting COPII-mediated lipoprotein traffic in the liver.

Discussion

One-third of the mammalian proteome is predicted to be sorted into the COPII secretory vesicles. The assembly of COPII-coated vesicles to facilitate the transport of lipids and proteins from the ER to the Golgi has been well described (1–6), but it remains poorly delineated how this process is regulated in response to changes of nutrient states. In this study, we found that the IRE1α-XBP1 branch of the cellular UPR is implicated in coordinating COPII vesicle trafficking during its dynamic regulation in response to nutrient availability. Our results indicate that the IRE1α-XBP1 pathway not only functions as an ER quality control mechanism in coping with protein folding stress but also coordinates the vesicle trafficking process from the ER in meeting the nutrient demand of the cell.

Our results from the global XBP1s-directed transcriptional analysis indicate that the IRE1α-XBP1 branch of the UPR selectively couples the nutrient status to the anterograde transport pathway. Within the secretory machinery, changes in the profile of COPII cargo components at the ER must be coordinated for the vesicle formation and ER export. Transcriptional regulation by XBP1s may enable such coordinated metabolic regulation of COPII-mediated trafficking according to the cellular nutritional state. As such, liver-specific disruption of the IRE1α-XBP1 pathway resulted in impairment of COPII-mediated lipoprotein secretion, thus contributing to liver steatosis and hypolipidemia in mice during fasting. Herein, we propose a model for the dynamic regulation of COPII vesicle trafficking by the IRE1α-XBP1 nutrient-sensing pathway at the ER (Fig. 6G), which may represent a general paradigm for coordinating nutrient status and transcriptional reprogramming of cellular traffic under a diversity of physiological circumstances. It remains to be dissected, however, whether the COPII components regulated by the IRE1α-XBP1 axis are involved in selective sorting and assembly of particular vesicles containing protein/lipid species that are associated with ER stress.

In response to changes in nutrient intake, alterations in the turnover of proteins and lipids need be coordinated with cellular traffic machinery to maintain cellular homeostasis. For instance, dynamic COPII-mediated lipoprotein secretion accompanies increased lipid intake in liver cells (24). Secretion of VLDLs can be viewed as a mechanism not only for hepatocytes to clear excess lipid intake but also for delivering lipid fuels to other peripheral tissues. Our data, both in vitro and in vivo, demonstrate that trafficking of the COPII-coated vesicles is highly dynamic in response to the availability of nutrients such as lipids, and that the IRE1α-XBP1s axis can promote COPII-mediated lipoprotein secretion. This is consistent with the fact that mild fatty acid treatment could stimulate the assembly and secretion of VLDLs and triglycerides (15, 21, 47). Our data suggest that COPII trafficking is directly involved in mediating IRE1α-XBP1s regulation of ApoB secretion. We found that restoration of SEC23B partially restored COPII-mediated ApoB secretion in IRE1α LKO mice. These results were intriguing, given the recent discovery that the COPII SEC23 subunit, which is the most robustly regulated COPII subunit in our study, has been shown to be crucial not only for coat assembly but also for vesicle tethering and transport along microtubules (8, 48–51). Recently, SEC23A, which is paralogous to SEC23B, was shown to function interchangeably with SEC23B in COPII-dependent secretion (43). It is also possible that increasing SEC23A expression can promote ApoB secretion in IRE1α LKO mice. However, we cannot exclude the possibility that XBP1s may also act through other target genes (in addition to COPII components) to regulate COPII trafficking. Interestingly, in contrast to SEC23B, SEC24A protein levels were not lower in fasted or IRE1α LKO liver cytosol, despite the fact that its transcriptional expression is also subjected to XBP1s regulation. The reason for this difference is not clear. This could reflect a COPII subunit-specific posttranscriptional regulatory effect. Indeed, regulatory mechanisms, including but not limited to miRNA regulation and protein degradation, have been shown to affect COPII protein levels. It is tempting to speculate that such mechanisms are active in liver such that the steady-state SEC24A protein levels are not decreased in fasted or IRE1α LKO liver.

Our finding suggests that the IRE1α-XBP1 branch of the UPR may serve as a “rheostat” at the ER, linking the nutrient status to COPII-mediated lipoprotein secretion. While such XBP1s-mediated transcriptional regulation of the COPII vesicle trafficking pathway is most likely distinct from the perturbation in the cellular secretory pathway during the rapid ER stress-induced clearance process (52), dysregulation of this COPII secretory program under metabolic ER stress may contribute to impairment in lipoprotein secretion. The precise mechanisms that link nutrient availability to the IRE1α-XBP1s network defined here are unclear. Evidence has emerged that IRE1α deactivation may occur through mechanisms that involve its dephosphorylation (35), ER-associated degradation (53), or disassembly of the IRE1α signaling platform at the ER. It is also possible that the dynamic activation/deactivation of IRE1α can be an even more complex, and most likely physiological context-dependent, process. Future studies will be necessary to further delineate the IRE1α-XBP1s pathway in response to nutrient availability.

The liver is a critical organ that manages the metabolic adaptation in response to changes in nutrient availability through coordinating a variety of gene expression programs. As the downstream effector in exerting IRE1α-initiated gene expression programs, XBP1s has been shown to regulate many hepatic fuel metabolism mechanisms (30, 32, 37, 39, 54). However, the precise mechanisms by which XBP1 exerts its distinct reprogramming actions on liver metabolism during feeding versus fasting have yet to be defined. Through a comprehensive mapping analysis of the chromatin occupancy by XBP1s in the liver during feeding versus fasting, we identified XBP1s directly regulating the COPII secretory program in response to nutrient availability. Notably, we found no differences in the recognition motifs between “Fed-high” and “Unchanged” XBP1s-binding sites, suggesting that fed-up–regulated XBP1s binding might not act through distinct XBP1s binding motifs, but possibly through other mechanisms such as cofactor recruitment. Indeed, our results suggest that the XBP1s-driven liver COPII program acts in cooperation with another bZIP transcriptional CREB3L2, which has also been shown to regulate COPII gene expression (44–46). We found that XBP1s directly activates Creb3l2 transcription, and this likely adds another regulatory layer downstream of the nutrient-sensing XBP1s cascade to regulate the COPII machinery. More interestingly, XBP1s also directly interacts with CREB3L2. It would seem likely that XBP1s works in concert with CREB3L2 to enhance the mechanism describe here and boosts the high-level COPII trafficking capacity. The IRE1α-XBP1 signaling axis has a key role in coping with ER protein folding stress. In professional secretory cells, such as B cells and pancreatic islet cells, the IRE1α-XBP1 pathway acts to handle ER stress as a result of higher cellular demand for protein synthesis and secretion (25–28, 55, 56). Herein, our findings reveal a metabolic role of the IRE1α-XBP1 pathway in nutrient (e.g., fatty acid) sensing and COPII-mediated trafficking (e.g., lipoproteins), processes not directly related to the protein-folding status. Notably, a previous study showed little alteration in ApoB secretion in another mouse model with liver-specific IRE1α knockout (39). This discrepancy is likely due to the use of IRE1α^+/− animals as the control group or to the different genetic deletion strategy (disruption within the RNase domain of the IRE1α protein).

Our findings that the COPII genes are more robustly up-regulated in XBP1s binding sites in the fed state revealed an important physiological aspect of the IRE1α pathway in metabolic control, and suggest a critical role of the IRE1α pathway in exerting the nutrient-sensing and -handling functions at the ER. This implicates the IRE1α-XBP1 pathway in the control of cellular processes beyond protein folding stress, which is in accordance with its direct involvement in fuel metabolism upon its metabolic activation as previously demonstrated (30, 32, 37, 39, 54). XBP1s has been shown to regulate genes related to ER-to-Golgi vesicular trafficking in cultured cells (42), and the Yip1A/IRE1α pathway has also been shown to regulate COPII genes during Brucella infection (57). Interestingly, upon acute ER stress, we observed no activation, but rather selective suppression, of liver COPII traffic-related genes, even in the face of the activated IRE1α-XBP1s pathway. Along the same line, it was documented that pharmacological inducers of ER stress could lead to reductions in ApoB-mediated lipoprotein secretion (21, 23, 58). Therefore, XBP1s may possess a distinct activation mode in the liver in response to changes in nutrient availability relative to that during typical ER stress states, likely through its regulatory modifications or cooperation with other regulatory factors. Nevertheless, our results also suggest that in the context of hypernutrition, dysregulation of the IRE1α-XBP1 pathway may cause disruption of COPII-mediated trafficking during metabolic ER stress, contributing to the pathogenic promotion of metabolic disorders.

In summary, our study illustrates a previously unappreciated nutrient-sensing function of the IRE1α-XBP1s pathway that contributes to the orchestration of the liver COPII secretory program in response to nutrient availability. Elucidation of the UPR-related regulatory pathways involved in liver COPII-mediated lipoprotein secretion could help us further understand the molecular links between ER stress and metabolic dysfunctions in liver steatosis.

Materials and Methods

All animal studies were conducted in strict accordance with the institutional guidelines for the humane treatment of animals and were approved by the Institutional Animal Care and Use Committees at the Model Animal Research Center of Nanjing University. Extended methods and information about animal studies, cell culture, DNA constructs and transfection, live-cell imaging, ChIP-seq and data processing, RNA-seq analyses, COPII budding assay, immunoblotting, virus infection, immunoprecipitation, histological analyses, mass spectrometry, blood and tissue chemistry, and statistical analyses are described in SI Appendix, Supplementary Materials and Methods. The ChIP-seq and RNA-seq datasets have been deposited in the National Center for Biotechnology Information Gene Expression Omnibus and are accessible through GEO accession nos. GSE101202 and GSE100358, respectively.