Regulation of lipid metabolism by the unfolded protein response

Abstract The endoplasmic reticulum (ER) is the site of protein folding and secretion, Ca2+ storage and lipid synthesis in eukaryotic cells. Disruption to protein folding or Ca2+ homeostasis in the ER leads to the accumulation of unfolded proteins, a condition known as ER stress. This leads to activation of the unfolded protein response (UPR) pathway in order to restore protein homeostasis. Three ER membrane proteins, namely inositol‐requiring enzyme 1 (IRE1), protein kinase RNA‐like ER kinase (PERK) and activating transcription factor 6 (ATF6), sense the accumulation of unfolded/misfolded proteins and are activated, initiating an integrated transcriptional programme. Recent literature demonstrates that activation of these sensors can alter lipid enzymes, thus implicating the UPR in the regulation of lipid metabolism. Given the presence of ER stress and UPR activation in several diseases including cancer and neurodegenerative diseases, as well as the growing recognition of altered lipid metabolism in disease, it is timely to consider the role of the UPR in the regulation of lipid metabolism. This review provides an overview of the current knowledge on the impact of the three arms of the UPR on the synthesis, function and regulation of fatty acids, triglycerides, phospholipids and cholesterol.


| INTRODUC TI ON
The endoplasmic reticulum (ER) is one of the largest organelles in the cell. It is involved in multiple fundamental biological processes including protein folding and secretion, Ca 2+ storage and lipid synthesis. In response to stress (eg accumulation of misfolded proteins), the ER triggers the unfolded protein response (UPR), a complex and conserved signalling pathway that is mediated by three ER transmembrane sensor proteins: inositol-requiring enzyme 1 alpha (IRE1α), protein kinase RNA-like ER kinase (PERK) and activating transcription factor 6 (ATF6). Once active, these sensors induce an elaborate and integrated signalling network that either allows restoration of ER homeostasis or triggers cell death. 1 In addition to protein folding, the ER plays an important role in the regulation of lipid metabolism. Several enzymes involved in triglyceride (TG) and cholesterol biosynthesis, as well as enzymes participating in the regulation of membrane turnover and dynamics, are located in the ER. Consequently, activation of the UPR has important implications for lipid and sterol synthesis, some of which are yet to be unravelled. The purpose of this review is to describe the role of the UPR in the regulation of lipid metabolism, in order to introduce this complex field to a new audience. We have focussed on primary pathways including the synthesis, function and regulation of fatty acids (FAs), TGs, phospholipids (PLs) and cholesterol, and describe their regulation by the UPR.

| Role and activation
IRE1α, PERK and ATF6 share a common architecture, which con- However, these sensors can also be activated in response to ER membrane perturbations caused by changes in PL composition, 4 in cholesterol, 5 sterol 6 and inositol 7 levels and by changes in lipid accumulation 8 and saturation. 9 For example, in the absence of misfolded proteins, IRE1α and PERK can still respond to the lipid bilayer stress induced by increased lipid membrane saturation, 9 while ATF6 can be activated by specific species of sphingolipids. 4 More details on the mechanisms governing the activation of the UPR sensors are reviewed elsewhere. 1

| IRE1α
IRE1 is the most evolutionarily conserved UPR sensor. IRE1α is expressed ubiquitously while its paralog, IRE1β, is restricted to the lungs and gastrointestinal tract. 1 IRE1α possesses two distinct enzymatic activities that are mediated by cytosolic kinase and RNase domains. 10 Upon activation during ER stress, IRE1α forms homodimers and oligomers. This enables the trans-autophosphorylation of IRE1α's kinase domain leading to the allosteric activation of its RNase domain. 11 IRE1α's RNase domain catalyses the excision of 26 nucleotides from X-box binding protein 1 (XBP1) mRNA and produces a frameshift that allows the translation of a longer isoform called spliced XBP1 (XBP1s) (Figure 1). 12 XBP1s is a transcription factor that induces the expression of genes involved in lipid synthesis, chaperone protein synthesis and ER-associated degradation (ERAD) machinery, leading to increased ER size and capacity. 1 In contrast, unspliced XBP1 (XBP1u) lacks transcriptional activity. IRE1α's RNase activity also facilitates the degradation of various mRNAs, cleaving them at a defined consensus sequence through a process called regulated IRE1-dependent decay (RIDD) (Figure 1). 1 Identification of IRE1α mRNA targets revealed that RIDD activity can reduce the load of newly synthesized peptides entering into the ER, or promote apoptosis. 1,13 The kinase activity of IRE1α is associated with the activation of both the JNK and the NF-κB pathways, resulting in increased autophagy and apoptosis. 14,15

| PERK
Similar to IRE1α, PERK also oligomerises and trans-autophosphorylates following activation by ER stress. Once activated, the cytosolic kinase domain of PERK phosphorylates and inactivates eIF2α, an essential component of the 43S pre-initiation complex necessary for the initiation of cap-dependent protein translation ( Figure 1). This leads to a global arrest in protein translation. 1 At the same time, eIF2α phosphorylation allows translation of a specific set of mRNAs that carry one or more upstream open reading frames in their 5′ untranslated regions. 16 Thus, activation of the PERK pathway has a dual function: decreasing the entry of newly synthesized peptides into the ER to alleviate ER stress while simultaneously stimulating the production of proteins that are critical for stress adaptation. 1,16 The latter process is exemplified by the specific translation of ATF4, an important transcription factor that plays key roles in autophagy, antioxidant response, amino acid metabolism and the synthesis of stress-induced proteins. 16

| ATF6
Unlike IRE1α and PERK, ATF6 activation does not involve phosphorylation. Release of ATF6 from BiP exposes Golgi-localization sequences present on the luminal domain of ATF6. 17 Once transported into the Golgi, site-1 (S1P) and site-2 proteases (S2P) cleave ATF6 and release a cytosolic fragment containing a basic leucine zipper (bZIP) transcription factor called ATF6f ( Figure 1). 18,19 ATF6f induces expression of XBP1 as well as genes involved in protein folding, ERAD machinery, ER homeostasis and ER and Golgi biogenesis. 20,21 ATF6f and XBP1s can form heterodimers, whose association induces expression of ERAD proteins. 22

| Overview of lipid metabolism
Lipid molecule in cells can be subdivided into two large groups: the long hydrocarbon chain-containing FAs, including TGs and PLs, and the ring-structured sterols. FAs are composed of a carboxyl group linked to a long aliphatic hydrocarbon chain that is either saturated or unsaturated. De novo FA synthesis occurs through a cellular process called lipogenesis ( Figure 2) and FAs provide the building blocks for the formation of TGs and PLs. Energy stored in TGs can be released in a catabolic process called β-oxidation. TGs usually coalesce into lipid droplets before being secreted into the blood as very low-density lipoproteins (VLDL), usually by hepatocytes ( Figure 2).
VLDLs can then either enter adipocytes, where TGs will be stored, or be transported to other cell types, where they support energy production. 23 In contrast, PLs are the main components of cell membranes and can fulfil structural (eg phosphatidylethanolamine Cholesterol can also act intracellularly as a precursor for steroid hormones, oxysterols and bile acids. 26 To avoid high levels of circulating free cholesterol in the blood, the latter is  29 and fatty acid synthase (FASN) (Figure 2). The product of this reaction is palmitic acid, a 16-carbon saturated FA, which can be elongated to produce very long chain FAs (VLCFA). 30 Addition of FAs to coenzyme A produces FA-CoA molecules, which are used to generate both glycero-and phospholipids ( Figure 2). Glycerol-P acyltransferase (GPAT) catalyses the attachment of the first FA-CoA to a glycerol-3 phosphate (G3P) backbone, producing a monoacylglycerol molecule also called lysophosphatidic acid (LPA). Acylglycerol-P acyltransferase (AGPAT) adds a second FA-CoA to LPA, converting it into phosphatidic acid F I G U R E 1 The Unfolded Protein Response (UPR) is controlled by three endoplasmic reticulum (ER) stress sensors: inositol-requiring enzyme 1 (IRE1), activating transcription factor 6 (ATF6) and PKR-like ER kinase (PERK). Upon activation, IRE1 splices x-box binding protein 1 (XBP1) mRNA, which is then ligated by RTCB and translated into XBP1s. IRE1 also cleaves cytosolic RNA in the process called regulated IRE1 dependent decay (RIDD), which reduces levels of target transcripts. Activated ATF6 is translocated to the Golgi, where its N-terminal fragment is released by proteases and translocated to the nucleus. PERK-mediated phosphorylation of eIF2α inhibits a global expression of genes, like SREBP activity-regulating Insig, and prompts selective translation of ATF4. Despite also relying on acetyl-CoA, cholesterol synthesis is synthesized through a different multistep metabolic pathway termed mevalonate pathway, involving more than 15 enzymes and 30 different reactions. 34 Here, we have focussed on the limiting steps of that pathway such as the rate of cholesterol synthesis, which is mediated by the activities of the HMG-CoA synthase (HMG-CS) and the HMG-CoA reductase (HMG-CR), two enzymes whose expressions are tightly regulated by lipid metabolism (Figure 2). 34 The SREBP family is composed of three proteins SREBP1a, SREBP1c and SREBP2, encoded by two different genes: SREBP1 and SREBP2.

| Classical known regulators of lipid metabolism
SREBP1a is highly expressed in intestinal epithelial cells, cardiomyocytes, macrophages and bone marrow dendritic cells, and has a high potency for stimulating both lipogenic and cholesterogenic gene expression. 43 In contrast, SREBP1c is predominant in most tissues and acts mainly by controlling the expression of lipogenic genes. 44 Thus, appropriate SREBP1c activity is critical for the regulation of FAs and TGs in lipogenic cells such as hepatocytes and adipocytes. 45 SREBP2 expression has been confirmed in a large variety of tissues. SREBP2 mainly mediates sterol regulation and is therefore complementary to SREBP1c. 36 Regulation of SREBP activation is dependent on SREBP cleavage-activating protein (SCAP) and Insulin Induced Gene (Insig).
SCAP is an escort protein that allows SREBPs to enter into the Golgitargeted COPII coated vesicles through to its MELADL motif. 46 Insig, which is directly bound to SCAP, maintains the SREBP-SCAP protein complex in the ER membrane ( Figure 1). SCAP has the ability to detect the presence of cholesterol, 47 while Insig, which can be induced by insulin in the liver, 48 is able to sense oxysterols. Hence, the presence of cholesterol and oxysterol promotes the binding of SCAP and Insig, ultimately inhibiting the SREBP pathway. 46,49,50 Carbohydrate-responsive element-binding protein (ChREBP), tissues like the liver, the heart and muscles, and is described to control FA catabolism. 56,57 Finally, PPARδ is ubiquitously expressed and is involved in various functions such as wound healing and VLDL signalling in macrophages. 58,59 C/EBPα is a transcription factor, which is also highly expressed in liver and adipose tissues. 60

| Regulation of lipid metabolism by IRE1α
Both the XBP1s and RIDD pathways modulate the expression of genes involved in lipid synthesis, thus supporting the critical role of IRE1α in the regulation of lipid metabolism. A transcriptional study highlighted a set of lipid-related genes activated in response to overexpression of human XBP1s in mouse fibroblasts. 65 Lipin genes (Lpin1 and Lpin3) were among the most up-regulated metabolic genes together with the Osbp gene, which encodes a sterol-sensing protein that modulates SREBP activity in response to sterol, 66 Pecr, an enzyme involved in FA elongation, 67 Lss, which catalyses the formation of lanosterol from squalene 68 and Gpat4, an enzyme which adds a FA to glycerol during lipogenesis (Figure 2). 69 The gene encoding FA elongase 4, Elovl4, was down-regulated. Another study demonstrated that conditional postnatal KO of Xbp1 in mice liver led to impairment of both FA and sterol synthesis in hepatocytes. 70 Impaired lipogenesis was associated with down-regulation of Scd1, Dgat2 and Acacb, with no significant changes in SREBP targets. Specific binding of XBP1s to the regulatory sequences of these genes was confirmed by ChIP in liver nuclear extracts from mice injected with tunicamycin or fed a high-fructose diet. However, XBP1s overexpression in primary mouse hepatocytes was insufficient to re-activate Scd1 expression, in contrast to Dgat2 and Acacb. 70 The IRE1α/XBP1s axis has also been studied in the context of ER expansion in plasmocytes, 71  However, inducible genetic deletion of XBP1 has been linked to hyperactivation of IRE1α, thus making it difficult to uncouple the roles of XBP1 and RIDD. 70 Thus, some observations made upon genetic manipulation of XBP1 could be due to increased RIDD activity. The contribution of XBP1 and RIDD activity to hepatic lipid metabolism was examined in mice carrying liver-specific deletions of Ern1a or Xbp1. Impairment of RIDD activity, through IRE1α silencing, partially restored TG and cholesterol levels in the liver of Xbp1 LKO mice. Partial reversion of the phenotype was accompanied by an increase in Dgat2 and Acacb mRNA expression, suggesting these genes are potential RIDD targets. 82 In addition, the mRNA for proprotein convertase subtilisin/kexin type 9, an enzyme involved in the clearance of the LDL receptor, and for Angptl3, an inhibitor of TG hydrolysis, were also proposed as potential RIDD targets. 82 RIDD targets also include miRNAs. For example,miR-34 and miR-200, two miRNAs that are known to down-regulate PPARα and SIRT1 expression in mouse liver, 83 have also been identified as RIDD targets. 56,84 Furthermore, a recent paper demonstrating unconventional IRE1- with both normal and high-carbohydrate diets. The protective effect of ATF4 deletion against diet-induced liver steatosis was proposed to be partially linked to Scd1 loss. Indeed, liver-specific overexpression of ATF4 in mice increased SCD1 protein expression, while oral supplementation with oleate, the main product of SCD1 activity, increased hepatic lipid accumulation and liver weight in Atf4 KO mice. 92 Similar observations were reported in Atf4 KO mice fed a high-fructose diet, 93 where expression of lipogenic enzymes (PPARγ, FAS and ACC) were reduced in the liver, while the expression of proteins involved in FA oxidation (ACOX1, CPT-1) was unaffected.
Hence, the authors concluded that ATF4 deficiency decreased hepatic lipogenesis but did not affect TG secretion and FA oxidation, thus protecting Atf4-deficient mice from fructose-induced hepatic hypertriglyceridemia. 93

| Regulation of lipid metabolism by ATF6
The contribution of ATF6 to the regulation of lipid metabolism is evi-  96 Mechanistically, ATF6 physically interacts with PPARα/RXRα (retinoid X receptor alpha) heterodimers and is required for their transcriptional activity in mouse liver. 97 In line with these data, liver-specific overexpression of dnATF6 blocked PPARα/RXRα transcriptional activity in a luciferase reporter assay.
Expression of ATF6 in mice liver also correlated with PPARα expression during fasting, a scenario that enhances TG levels and promotes increased lipid deposition in the liver. Hence, liver-specific overexpression of dnATF6 in fasted mice caused hepatic steatosis and reduced levels of serum β-hydroxybutyrate (a marker of β-oxidation).
By contrast, liver-specific overexpression of ATF6f improved hepatic condition in steatosis-induced mice fed a high-fat high-sucrose diet.
In this study, ATF6 did not affect FA synthesis, suggesting that in the liver ATF6 activity might be more important for FA oxidation than Another study, performed in vitro in a human kidney cell line (HK-2) overexpressing either dnATF6 or ATF6f, also indicated that ATF6 function in kidney is opposite to that observed in liver. 99

| CON CLUS ION
Besides its well-characterized role in protein homeostasis, the contributions of the UPR to lipid metabolism are beginning to be appreciated. Multiple studies support the involvement of each of the three branches of the UPR in the modulation of lipid metabolism (Figure 3), and the implications of this in understanding the role of lipid metabolism in restoring ER proteostasis merit further investigation.
As the UPR is able to promote both lipogenesis and lipolysis in response to particular cellular contexts and stimuli, its impact on lipid  109 We anticipate that for cancer, and indeed other diseases, investigation into the interplay between the UPR, lipid metabolism and disease will be a fruitful area of research. Moreover, since UPR activation can lead to cell type-specific responses, organ-specific studies (eg in liver, pancreas, kidneys and white adipose tissue) will be informative in developing therapeutic approaches based on modulation of UPR sensors. To that end, new in vivo models such as the recently developed KINGS Ins2 +/G32S mouse model of human diabetes will be invaluable. 113 Taken together, this review highlights a crucial role for UPR in the coordination of lipid metabolism and metabolic reprogramming and suggests this is an important area for further research in relevant diseases.

AS and AG are co-founders and directors of Cell Stress Discoveries
Ltd. funding acquisition (equal); writing-review and editing (equal).

DATA AVA I L A B I L I T Y S TAT E M E N T
Data sharing is not applicable to this article as no new data were created or analysed in this study.