IRE1A Stimulates Hepatocyte-derived Extracellular Vesicles That Promote Inflammation in Mice With Steatohepatitis

Abstract

Background & Aims:

Endoplasmic reticulum to nucleus signaling 1 (ERN1, also called IRE1A) is a sensor of the unfolded protein response that is activated in livers of patients with nonalcoholic steatohepatitis (NASH). Hepatocytes release ceramide-enriched inflammatory extracellular vesicles (EVs) following activation of IRE1A. We studied the effects of inhibiting IRE1A on release of inflammatory EVs in mice with diet-induced steatohepatitis.

Methods:

C57BL/6J mice and mice with hepatocyte-specific disruption of Ire1a (IRE1a^Δhep) mice were fed a diet high in fat, fructose, and cholesterol (FFC) to induce development of steatohepatitis or a standard chow diet (controls). Some mice were given intraperitoneal injections of the IRE1A inhibitor 4μ8C. Mouse liver and primary hepatocytes were transduced with adenovirus or adeno-associated virus that expressed IRE1A. Livers were collected from mice and analyzed by quantitative PCR and chromatin immunoprecipitation assays; plasma samples were analyzed by ELISA. EVs were derived from hepatocytes and injected intravenously into mice. Plasma EVs were characterized by nanoparticle-tracking analysis, electron microscopy, immunoblots, and nanoscale flow cytometry; we used a membrane-tagged reporter mouse to detect hepatocyte-derived EVs. Plasma and liver tissues from patients with NASH and without NASH (controls) were analyzed for EV concentration and by RNAscope and gene expression analyses.

Results:

Disruption of Ire1a in hepatocytes or inhibition of IRE1A reduced release of EVs and liver injury, inflammation, and accumulation of macrophages in mice on the FFC diet. Activation of IRE1A, in livers of mice, stimulated release of hepatocyte-derived EVs, and also from cultured primary hepatocytes. Mice given intravenous injections of IRE1A-stimulated, hepatocyte-derived EVs accumulated monocyte-derived macrophages in liver. IRE1A-stimulated EVs were enriched in ceramides. Chromatin immunoprecipitation showed that IRE1A activated X-box binding protein 1 (XBP1) to increase transcription of serine palmitoyltransferase genes, which encode the rate-limiting enzyme for ceramide biosynthesis. Administration of a pharmacological inhibitor of serine palmitoyltransferase to mice reduced the release of EVs. Levels of XBP1 and serine palmitoyltransferase were increased in liver tissues, and numbers of EVs were increased in plasma, from patients with NASH compared with controls and correlated with the histologic features of inflammation.

Conclusions:

In mouse hepatocytes, activated IRE1A promotes transcription of serine palmitoyltransferase genes via XBP1, resulting in ceramide biosynthesis and release of EVs. The EVs recruit monocyte-derived macrophages to liver, resulting in inflammation and injury in mice with diet-induced steatohepatitis. Levels of XBP1, serine palmitoyltransferase, and EVs are all increased in liver tissues from patients with NASH. Strategies to block this pathway might be developed to reduce liver inflammation in patients with NASH.

Graphical Abstract

INTRODUCTION

Nonalcoholic fatty liver disease (NAFLD) includes a histologic spectrum of manifestations ranging from asymptomatic and nonprogressive steatosis to steatohepatitis (NASH) with progressive fibrosis leading to cirrhosis.¹ NASH is characterized by hepatocellular injury, recognized as ballooning and hepatic inflammation, which together beget fibrosis. Apart from weight loss, there are no regulatory agency-approved therapies for NASH.² The lack of therapies and the suboptimal efficacy of recent lead drugs underscore the incomplete understanding of NASH pathogenesis. The recently proposed multi-parallel hypothesis suggests that NASH is the result of interactions between numerous signaling pathways including genetic predisposition, gut dysbiosis, dietary intake, and abnormal lipid metabolism that lead to lipotoxicity and inflammation.³ At a cellular and molecular level, lipotoxicity is well-defined. Saturated free fatty acids, such as palmitate, can induce sublethal and lethal hepatocyte toxicity.⁴ The importance of sublethal hepatocyte toxicity was realized with the insight that apoptosis, a terminal event, may not fully represent the perturbations that occur during sublethal stress signaling in hepatocytes. It also became clear that while the majority of hepatocytes are steatotic and several develop ballooning, apoptosis remains a rare event. Therefore, we and others have been examining sublethal hepatocyte stress responses.

Endoplasmic reticulum (ER) stress is a prominent manifestation of lipotoxic sublethal stress. Lipotoxic ER stress activates inositol-requiring enzyme-1A (IRE1A) in a manner distinct from canonical activation of IRE1A by misfolded proteins.⁵ Regardless of how IRE1A is activated, the endoribonuclease activity of IRE1A cleaves XBP1 (X-box binding protein 1) mRNA to generate a spliced variant, XBP1s, mRNA. The protein encoded by XBP1s mRNA is commonly referred to as XBP1 and acts as a transcription factor. Several studies have linked IRE1A-XBP1 signaling to insulin resistance, dyslipidemia, hepatic steatosis, and inflammation in NAFLD.⁶ In investigating the role of IRE1A in sublethal lipotoxic stress, our laboratory previously demonstrated that IRE1A is necessary for the release of extracellular vesicles (EVs) from palmitate-treated hepatocytes⁷. EVs are nanoparticles released by cells under both basal and stress conditions that can mediate cell-to-cell communication by transporting bioactive cargo molecules. However, the role of IRE1A-stimulated EVs in NASH remains unexplored.

Our objective was to investigate the impact of IRE1A-stimulated hepatocyte-derived EVs in NASH pathogenesis with a specific focus on immune cell responses. Here we report that a pharmacological small molecule inhibitor, or genetic deletion, of IRE1A in hepatocytes can mitigate IRE1A-stimulated EV release and concomitant inflammation in NASH. IRE1A-stimulated EVs attract proinflammatory monocyte-derived macrophages into the liver. IRE1A upregulates de novo ceramide biosynthesis leading to the release of hepatocyte-derived EVs; these pathways are upregulated in human NASH, and potentially amenable to therapeutic targeting.

MATERIALS and METHODS

Animals.

C57BL/6J male mice were fed rodent chow diet (CD) or at 12 weeks of age switched to a high fat, fructose, and cholesterol (FFC) diet for 24 weeks to induce NASH.⁸ Littermate Ire1a^loxP/loxP male mice, gifted by Professor Randal J Kaufman,⁹ were fed with FFC diet for 20 weeks and transduced intravenously with 1.5 x 10¹¹ viral copies adeno-associated virus (AAV) expressing Cre-recombinase under the control of hepatocyte-specific thyroxin binding globulin promoter (AAV8-TBG-Cre) to generate hepatocyte-specific Ire1a deleted (Ire1a^Δhep) mice or AAV8-TBG-Null for control mice (Ire1a^loxP/loxP). Gt(ROSA)26Sortm4(ACTB-tdTomato,-EGFP)Luo/SjJ (mT/mG) male and female mice,¹⁰ gifted by Dr. Samar Ibrahim, were transduced intravenously with 1.5 x 10¹¹ viral copies of AAV8-TBG-Cre to generate hepatocyte reporter mice. For adenoviral transduction, male C57BL/6J mice received 1 x 10¹¹ viral copies per mouse via tail vein injection.

Human samples.

Human liver and plasma samples were procured from the biospecimens core of the Center for Signaling in Gastroenterology, Mayo Clinic, Rochester.

Isolation and transduction of primary mouse hepatocytes (PMH) and bone marrow-derived macrophages (BMDM) chemotaxis assay.

PMH were isolated from C57BL/6J mice¹¹ and transduced with AdIRE1A or AdB-gal (2 x 10⁸ viral copies/cm²) for 24 hours. BMDMs were isolated from the hind legs of C57BL/6J mice and utilized for measuring chemotaxis in Microfluidic 2D Cell Migration Chambers as previously described.¹²

EV isolation, characterization, and gradient purification.

EVs were isolated from platelet poor plasma or cell culture supernatants by differential ultracentrifugation (UTC) and characterized by nanoparticle tracking analysis (NTA).⁷ Iodixanol density gradient separation was performed¹³ and purified pooled fractions 3, 5 and 6 were used for immunogold staining.

EV transplantation.

EVs isolated from equal numbers (3 x 10⁶ cells) of AdB-gal or AdIRE1A transduced PMH were suspended in 100 μL of PBS each and injected intravenously into C57BL/6J mice. Mice were sacrificed 48 hours after injection.

EV bio-distribution.

EVs were labeled with Zirconium-89 using radiolabeling synthon ([⁸⁹Zr]Zr-DBN)¹⁴ and the resulting [⁸⁹Zr]Zr-DBN-EVs were separated from unconjugated [⁸⁹Zr]Zr-DBN) with size exclusion chromatography. Following intravenous injection into C57BL/6J mice (~4 x 10⁹ EVs/ mouse), single plane coronal PET scan images were acquired at 0 hour and 4 hours with Inveon PET/CT scanner.

Sphingolipidomics.

Ceramides were measured using tandem mass spectrometry at the Mayo Clinic Metabolomics Resource Core as previously described.⁷

Pharmacological inhibitor administration.

IRE1A inhibitor 4μ8C (3.3 mg/kg) or DMSO (in 16% vol/vol Cremophor EL, Sigma-Aldrich) were administered by daily intraperitoneal injections.¹⁵ SPT inhibitior myriocin (1 mg/kg, Enzo Life Sciences) or 50 mM NaOH (in 0.9% normal saline) were administered by daily intraperitoneal injections.¹⁶

Electron microscopy and immunogold staining.

EVs were characterized by electron microscopy (EM) and immunogold staining using two hepatocyte specific antisera ASGPR1 and CYP2E1 with gold-protein A particles (10 nm and 15 nm, respectively). Objects were observed with JEOL 1400 electron microscope.

Western blotting.

Equal amounts of proteins were resolved and transferred to PVDF membrane. EVs, from equal volume of plasma, were suspended in RIPA buffer and Instant-Bands Pre-Staining Protein Sample Loading Buffer (EZ-Biolab) for western blotting.¹³ Primary antibodies (Supplementary Table 1) and horseradish peroxidase-conjugated secondary antibodies (Santa Cruz Biotechnology) were used for chemiluminescent detection (Amersham). Band intensity was measured using ImageJ software.

Enzyme-linked immunosorbent assay (ELISA).

Plasma levels of tumor necrosis factor-a (TNF-α), interleukin 1β (IL1β), and interleukin 10 (IL10) were measured by ELISA (R&D Systems).

Reverse transcription PCR and quantitative real time PCR.

Liver total RNA was isolated by Quick-RNA MiniPrep kit (Zymo Research) and reverse-transcribed into complementary DNA (Bio-Rad Laboratories). Quantitative real time PCR (qPCR) was performed using gene specific primers (Supplementary Table 2 and 3) and SYBR® Green PCR Master Mix (Roche Diagnostics). All datasets are expressed as fold change relative to control as delta delta Ct utilizing Hprt or 18s as housekeeping genes.¹³

Chromatin immunoprecipitation (ChIP assay).

ChIP assay was performed in cryopreserved liver tissue.¹⁷ Sheared chromatin was immunoprecipitated with specific antibody (Supplementary Table 1) coated Protein-G beads. Immunoprecipitated chromatin was eluted and purified by QIAquick PCR purification kit followed by qPCR to estimate the abundance of XBP1 at the promoter region of target genes. Data are interpreted as a ratio of precipitated DNA: total input of genomic DNA.

Histology, immunohistochemistry, and TUNEL assay.

H&E staining was performed by the Mayo Clinic Histology Core. For immunohistochemistry (IHC), formalin-fixed paraffin embedded (FFPE) tissue sections were stained with specific antibody for Galectin-3 (Supplementary Table 1) and quantified using the NIS Elements AR software (Nikon).⁸ In Situ Cell Death Detection Kit Fluorescein (Roche) was used to detect DNA strand breaks by the TUNEL assay. Hepatocyte apoptosis was quantified by the measuring the percentage of TUNEL positive nuclei in total nuclei per high power field for at least 10 fields per sample.

Flow cytometry.

Livers were digested with liver dissociation enzymes mix using the gentleMACS dissociator (Miltenyi Biotec). Intrahepatic leukocytes were isolated from the interphase between 25% and 50% percoll (Millipore Sigma)¹⁸ and were labeled with fluorochrome-conjugated specific antibodies (Supplementary Table 1). Data were acquired by multicolor flow cytometry on MACSQUANT Analyzer 10 (Miltenyi Biotec) using appropriate compensation setting and fluorescence minus one control. Data were analyzed on FlowJo software version 10.6.1.

RNAscope in-situ hybridization (ISH) of human liver tissue.

FFPE liver sections were used for RNAscope ISH using base scope probes against spliced XBP1 (715181-C2) and SPTLC1 (721831) by Advanced Cell Diagnostics (ACD), BaseScope®. Duplex Human Control Pack consisting of PPIB-C1 and POLR2A-C2 was included as housekeeping controls. For quantification, signal dots per high power field were manually counted for each color channel in 25 random fields for each sample. The data were expressed as the average number of dots per field normalized to the geometric mean of the two housekeeping genes.

Statistics.

All data represent mean ± SEM from three or more experiments. Each dot represents data from one mouse or biological replicate. Two-tailed Student’s t-test was used for statistical analyses in GraphPad Prism 8, *denotes p<0.05, **p<0.01, and ***p<0.001. In rare instances a one-tailed t-test was performed; indicated by #p<0.05. Pearson correlation analysis was employed for analyzing the strength and direction of the linear relationship between two variables and represented as R² value (square of the Pearson correlation coefficient R).

RESULTS

Pharmacological IRE1A inhibition ameliorates NASH.

Human and dietary models of NASH demonstrate activation of a lipotoxic ER stress response.⁸ In cultured hepatocytes we have demonstrated that lipotoxic ER stress activates IRE1A leading to EV release⁷ and FFC diet-induced steatohepatitis in mice is associated with an increase in circulating proinflammatory EVs.¹⁹ Therefore, to demonstrate a mechanistic link between IRE1A activation-induced EV release and hepatic inflammation, we examined the effect of inhibiting IRE1A on EV release, liver injury, and inflammation in the FFC-fed mouse NASH model. NASH was induced by FFC diet feeding for 22 weeks followed by 4μ8C administration for two weeks (Figure 1A). FFC diet-induced increase in spliced Xbp1 mRNA abundance and circulating EVs were significantly attenuated in 4μ8C-treated mice (Figure 1B,C and Supplementary Figure 1A). FFC-fed 4μ8C-treated mice demonstrated reduction in hepatic inflammatory foci by histological assessment, alanine aminotransferase (ALT), macrophage accumulation by galectin-3 IHC, and cell death by TUNEL assay (Figure 1D-G).

Figure 1. Pharmacological IRE1A inhibition attenuates NASH.

A) Schematic depicting 4μ8C or DMSO treatment in FFC-fed mice. B) Relative mRNA expression of spliced Xbp1; C) EV counts in plasma; D) Representative H&E stained images (black arrows indicate inflammatory foci; E) Serum ALT; F) Representative images of IHC for galectin-3 and its quantification; G) Representative images of TUNEL stained liver sections and its quantification; relative mRNA expression of H) Cd68, I) Lgals3, and J) Ly6c in CD- and FFC-fed mice treated with 4μ8C (n=5 each). K) Flow cytometry data showing CD11b⁺F4/80⁺Ly6C⁺cells in livers and Ly6C⁺ cells (%) in CD11b⁺F4/80⁺ population in livers (n=3 each). Plasma levels of L) TNF-α, M) IL1β, and N) IL10 (n=5 each) in 4μ8C or DMSO treated FFC-fed mice. Two-tailed Student’s t-test was used for statistical analyses. Scale bar=50 μm.

Macrophage-mediated inflammation plays an essential role in NASH.³ Hepatic macrophages may be yolk sac-derived Kupffer cells or recruited from Ly6C-positive proinflammatory monocytes, referred to as monocyte-derived macrophages henceforth, following the nomenclature by Guilliams et al.²⁰ Recruitment of monocyte-derived macrophages is a cardinal feature of the sterile inflammatory response observed in NASH;³ therefore, we next examined if inhibition of IRE1A-induced EV release impacted the recruitment of these cells by gene expression analysis and flow cytometry of intrahepatic leukocytes. 4μ8C-treatment led to a reduction in mRNA abundance of Cd68, Lgals3, and Ly6c (Figure 1H-J) and Ly6C positive monocyte-derived macrophages (CD11b⁺F4/80⁺Ly6C⁺) (Figure 1K). Corresponding to a reduction in inflammation, a significant reduction in the plasma levels of proinflammatory cytokines TNF-α and IL1β (Figure 1L,M) was observed whereas the anti-inflammatory cytokine IL10 (Figure 1N) remained unchanged. Fibrosis is a sequela of chronic liver injury and inflammation; therefore, we next measured collagen deposition by picro-sirius red staining. FFC diet led to a modest increase in pericellular fibrosis, which was significantly reduced in the livers of 4μ8C-treated mice (Supplementary Figure 1B). This was further confirmed by SHG microscopy (Supplementary Figure 1C) and reduced mRNA expression of fibrogenic genes (Supplementary Figure 1D). Lipotoxic EVs activated cultured hepatic stellate cells (HSC) (Supplementary Figure 1E); therefore we cannot exclude a direct effect of lipotoxic EVs on HSC in vivo.²¹

Perturbations in IRE1A may impact hepatic steatosis and the two additional proximal unfolded protein response (UPR) sensors, RNA-dependent protein kinase (PKR)-like ER kinase (PERK) and activating transcription factor 6 alpha (ATF6a). No change was observed in steatosis measured by label-free CARS microscopy and hepatic triglyceride content (Supplementary Figure 1F-G). No significant change was observed in the activation of PERK and ATF6a pathways demonstrated by comparable protein levels of activating transcription factor 4 (ATF4), CCAAT-enhancer-binding protein homologous protein (CHOP), phospho and total eukaryotic initiation factor 2a (eIF2a), and mRNA abundance of ATF6a target genes ERp57(Pdia3) and ERp72 (Pdia4)²² (Supplementary Figure 1H, I). All together, these findings indicate that pharmacological inhibition of IRE1A reduces EV release, liver injury and inflammation in mice with diet induced steatohepatitis.

Genetic Ire1a knockout ameliorates NASH.

We confirmed the hepatoprotective effect of inhibiting IRE1A in NASH in a hepatocyte-specific Ire1a knock out (Ire1a^Δhep) mouse model. Ire1a was deleted in hepatocytes after 20 weeks of FFC feeding (Figure 2A). Ire1a^Δhep mice had a lower abundance of spliced Xbp1 mRNA (Figure 2B, Supplementary Figure 2A) along with reduction in circulating EVs, inflammatory foci, ALT, macrophage accumulation, and cell death 4 weeks after Ire1a deletion (Figures 2C-G). Hepatic expression of macrophage markers, Cd68, Lgasl3, and Ly6c was significantly lower (Figure 2H) and fewer CD11b⁺F4/80⁺Ly6C⁺ monocyte-derived macrophages were observed in FFC-fed Ire1a^Δhep livers (Figure 2I). Plasma levels of TNF-α and IL1β were lower and IL10 remained unchanged in FFC-fed Ire1a^Δhep mice (Figure 2J-L). Pericellular fibrosis was reduced in Ire1a^Δhep mice compared to control Ire1a^loxP/loxP mice (Supplementary Figure 2B,C) and was associated with reduced expression of profibrogenic genes (Supplementary Figure 2D). Thus, the fibrosis reduction occurred secondary to loss of hepatocyte IRE1A signaling with improvements in liver injury and inflammation. Steatosis and triglyceride content were similar in Ire1a^Δhep and Ire1a^loxP/loxP mice (Supplementary Figure 2E,F). There was no activation of PERK and ATF6a pathways following deletion of Ire1a in FFC-fed mice (Supplementary Figure 2G,H). Thus, using a complementary pharmacological-genetic approach we demonstrate that NASH attenuation following hepatocyte IRE1A inhibition is associated with lower circulating EVs and decreased hepatic monocyte-derived macrophages without any changes in steatosis and triglyceride content. We next dissected the mechanistic relationship between IRE1A activation and release of EVs from hepatocytes.

Figure 2. Hepatocyte-specific Ire1a knock out attenuates NASH.

A) Schematic depicting FFC-fed hepatocyte-specific Ire1a knock out (Ire1a^Δhep) or wildtype control (Ire1aloxP/loxP) mice. B) Relative mRNA expression of spliced Xbp1; C) Plasma EV counts; D) Representative H&E stained images (black arrows indicate inflammatory foci); E) Serum ALT; F) Representative images of galectin-3 IHC and its quantification; G) Representative images of TUNEL stained liver sections and its quantification; H) Relative mRNA expressions of Cd68, Lgals3, and Ly6c (n=5 each). I) Flow cytometry data showing CD11b⁺F4/80⁺Ly6C⁺ cells in livers and Ly6C+cells (%) in total CD11b⁺F4/80⁺ cell population (n=3 each). Plasma levels of J) TNF-α, K) IL1β, and L) IL10 (n=5 each) in FFC-fed Ire1a^Δhep or Ire1a^loxP/loxP mice. Two-tailed Student’s t-test was used for statistical analyses. Scale bar=50 μm.

IRE1A-induced EVs originate from hepatocytes.

In FFC-fed mice IRE1A inhibition lowered circulating EVs suggesting that IREa may play a role in EV release from hepatocytes; therefore, we next tested if IRE1A-activation was sufficient to lead to EV release from hepatocytes. C57BL/6J mice received either adenovirus expressing IRE1A (AdIRE1A) or B-gal (AdB-gal) as control (Supplementary Figure 3A). IRE1A overexpression (Supplementary Figure 3B) led to its activation as confirmed by Xbp1 splicing, which was significantly higher in AdIRE1A transduced livers (Supplementary Figure 3C, D). Circulating EVs were significantly higher in AdIRE1A than AdB-gal transduced mice (Figure 3A), and this increase was attenuated by 4μ8C (Figure 3B). 4μ8C treatment significantly reduced spliced Xbp1 abundance (Figure 3C, Supplementary Figure 3E) confirming inhibition of IRE1A RNase activity. IRE1a-stimulated EVs were comparable in size (Figure 3D) and morphology (data not shown) to control EVs. Immunogold electron microscopy with hepatocyte markers CYP2E1 and ASGPR1²³ further substantiated the hepatocellular origin of circulating EVs by confirming the presence of EVs that express both protein markers (Figure 3E). Purified EVs, from equal volumes of plasma, expressed characteristic markers including Tsg101, Alix, CD81 and CD9¹³ confirmed by western blotting (Figure 3F). Alix, Tsg101, CD81, CD9 and ASGPR1 expression was higher in AdIRE1A-stimulated EVs than AdB-gal in keeping with higher EV count (Figure 3G). To further substantiate the hepatocellular origin of IRE1A-stimulated EVs we used ROSAmT/mG mice. These mice express a cell membrane-targeted, two-color fluorescent Cre-reporter allele that expresses membrane-targeted tandem dimer Tomato (mT) prior to Cre-mediated excision and membrane-targeted green fluorescent protein (mG) after excision. After the administration of AAV8-TBG-Cre, Cre recombinase expressing hepatocytes would have cell membrane-localized EGFP fluorescence and only hepatocyte-derived EVs would express green-fluorescence. After administering AdIRE1A or AdB-gal, total number of EVs was significantly higher in Ire1a transduced mice by nanoflowcytometry (Figure 3H). This increase was due to the release of EVs from hepatocytes as demonstrated by the increase in EGFP positive EVs in Ire1a transduced mice. We further confirmed this key finding utilizing hepatocyte-specific AAV8-TBG-IRE1A (Supplementary Figure 4A). IRE1A overexpression and activation following transduction with AAV8-TBG-IRE1A led to an increase in circulating EVs (Supplementary Figure 4B-E). Altogether, utilizing a gain-of-function approach and hepatocyte reporter mice, we demonstrate a significant increase in the release of hepatocyte-derived EVs in mice following activation of IRE1A.

Figure 3. IRE1A induces the release of EVs from hepatocytes.

A) Plasma EV counts of AdB-gal or AdIRE1A transduced mice (n=10 AdB-gal, n=14 AdIRE1A). B) Plasma EV counts and C) Relative mRNA expression of spliced Xbp1 in liver of mice transduced with AdIRE1A and treated with 4μ8C or DMSO (n=6 each). D) EV size and E) Hepatocyte-origins of EVs by immuno-electron microscopy using ASGPR1 (10 nm beads) and CYP2E1 (15 nm beads) antibody-coated beads (scale bar=200 nm); F) Western blotting for exosome markers (Alix, Tsg101, CD81 and CD9) and hepatocyte marker ASGPR1 and G) their quantification in AdIRE1A or AdB-gal transduced mice (n=3 each). H) Plasma EV counts in ROSAmT/mG mice, transduced with AAV8-TBG-Cre and AdIRE1A or AdB-gal (n=4 each). Two-tailed Student’s t-test was used for statistical analyses.

IRE1A activation is associated with macrophage accumulation and liver injury.

Having demonstrated an increase in circulating hepatocyte-derived EVs, we asked whether IRE1A activation causes macrophage accumulation and liver injury. We observed an accumulation of galectin-3 positive macrophages in AdIRE1A transduced livers (Supplementary Figure 3F). Gene expression analysis demonstrated increased mRNA abundance of Cd68, Lgals3, and Ly6c (Supplementary Figure 3G) but no increase in markers of T cells, NK cells, NKT cells or eosinophils (Supplementary Figure 3H). The accumulation of monocyte-derived macrophages was further confirmed by an increase in the CD11b⁺F4/80⁺Ly6C⁺ population by flow cytometry (Supplementary Figure 3I). We found an increase in the proinflammatory cytokine TNF-α with no change in the plasma levels of IL1β or IL10 (Supplementary Figure 3J-L).

We further examined the cellular heterogeneity of intrahepatic leukocytes following IRE1A activation by CyTOF. These data demonstrated a significant enrichment in monocyte-derived macrophages. Intrahepatic leukocytes separated into 26 clusters, of which 8 clusters (4, 11, 12, 16, 17, 23, 25 and 26) were significantly differentially expressed in AdIRE1A transduced livers (Supplementary Figure 5A-D). Among these 8 clusters, distribution of cell frequencies, defined on the basis of the intensity of each cell surface marker, demonstrated an increase in CD11b⁺F4/80⁺Ly6C⁺CCR2⁺ cells, which comprise clusters 12, 16, and 23, thus, confirming recruitment of monocyte-derived macrophages in AdIRE1A-transduced mouse livers. Cluster 11 was comprised of Ly6C+ cells that also expressed LGALS3 (Galectin-3), suggesting recruited monocytes, potentially on the path to differentiate into macrophages (Supplementary Figure 5E-H). On the other hand, cluster 17, which contains CD206 positive macrophages, was significantly lower in AdIRE1A transduced livers, suggesting a reduction in alternatively-activated M2-like, restorative macrophages (Supplementary Figure 5I). Clusters 4, 25 and 26 were comprised of heterogeneous cell populations (data not shown). Thus, these high dimensional data highlight the importance of monocyte-derived macrophages in the liver injury response to IRE1A activation.

Along with accumulation of monocyte-derived macrophages in the liver we observed inflammatory foci on H&E staining, elevated ALT, and increased hepatocyte cell death by TUNEL assay without any changes in hepatic triglyceride content (Supplementary Figure 6A-D). Gene expression of key lipogenic genes remained unaltered in AdIRE1A-transduced livers (Supplementary Figure 6E). Overexpression of IRE1A modestly increased phosphorylation of eIF2a without significant increase in downstream ATF4 or CHOP levels (Supplementary Figure 6F). There was no increase in the expression of ATF6a target genes ERp57 and ERp72 (Supplementary Figure 6G). Therefore, IRE1A activation was sufficient to lead to EV release and an increase in monocyte-derived macrophages in the liver. To dissect the specific contribution of IRE1A-stimulated hepatocyte-derived EVs in attracting monocyte-derived macrophages into the liver, we performed EV transplantation experiments.

IRE1A-stimulated hepatocyte-derived EVs recruit macrophages into the liver.

We isolated IRE1A-stimulated EVs from PMH (IRE1A EVs) and administered these via tail vein injection into adult mice (Figure 4A). This approach allowed us to confirm that IRE1A transduction of PMH stimulates the release of EVs, and we avoided the confounding effects of EVs derived from other cell types and also from circulating factors. We confirmed the almost total and immediate hepatic uptake of transplanted B-gal EVs or IRE1A EVs after intravenous injection (Figure 4A). Consistent with our hypothesis of bioactive effects in the hepatic microenvironment, we found an increase in macrophage accumulation in response to IRE1A-stimulated EV injection, by galectin-3 IHC (Figure 4B). The expression of macrophage markers Cd68, Lgals3, and monocyte-derived macrophage marker Ly6c was significantly higher in IRE1a-EV transplanted mouse livers (Figure 4C). Flow cytometry confirmed that monocyte-derived macrophages (CD11b⁺F4/80⁺Ly6C⁺) were significantly increased in the livers of IRE1A EV transplanted mice (Figure 4D). Plasma levels of the cytokines TNF-α, IL1β, and IL10 remained unaltered in IRE1A EV transplanted mice (Figure 4E-G) denoting a lack of systemic inflammation and supporting a role for EVs in attracting monocyte-derived macrophages within the hepatic microenvironment. We confirmed the chemotactic effect of IRE1A EVs in vitro. BMDM migrated significantly more toward IRE1A EVs (Figure 4H-J). Thus, in summary, our data demonstrate that hepatocyte-derived IRE1A EVs are sufficient to attract monocyte-derived macrophages into the liver. In conjunction with the data above on IRE1A inhibition in a mouse NASH model, our findings suggest a proinflammatory role for IRE1A activation in NASH by stimulating the release of monocyte-attracting EVs.

Figure 4. IRE1A-stimulated EVs recruit monocyte-derived macrophages into the liver.

A) Schematic depicting the experimental design for EV transplantation. EVs from equal number of AdB-gal or AdIRE1A-transduced primary mouse hepatocytes (PMH) were transplanted into mice. Representative images and quantification of labelled EV-transplanted mice (n=3 each) at 0 hour and 4 hours to show the hepatic uptake of EVs. The images are scaled to show standardized uptake values (SUV). B) Representative images of galectin-3 IHC (scale bar equals 50 μm) and its quantification; C) Relative mRNA expressions of Cd68, Lgals3, and Ly6c (n=6 for B-gal EV transplanted mice, n=7 for IRE1A EV transplanted mice). D) Flow cytometry data showing CD11b⁺F4/80⁺Ly6C⁺cells in livers and Ly6C+cells (%) in total CD11b⁺F4/80⁺ cell population (n=3 each). Plasma levels of E) TNF-α, F) IL1β, and G) IL10 in both groups of EV-transplanted mice (n=5 each, cytokine levels were below detection level in some samples from each group). H) Representative BMDM migration plots in EV gradients at 2 hours. Cells highlighted in red are migrating up toward the positive end of the EV gradient. I) Migration velocity of BMDMs exposed to B-gal EVs or IRE1A EVs gradient and J) Accumulated distance (total path length) by BMDMs migrating in B-gal EVs or IRE1A EV gradient (n=3 each). Two-tailed Student’s t-test was used for statistical analyses.

IRE1A-stimulated EV release is mediated by transcriptional upregulation of de novo ceramide synthesis.

We previously demonstrated that in isolated hepatocytes the release of lipotoxic EVs depends on the de novo synthesis of ceramide.⁷ Therefore, we examined the contribution of this pathway to the biogenesis of IRE1A-stimulated EVs in vivo. Sphingosine, sphinganine, sphingosine 1-phosphate, C16:0-, C18:1-, C20:0-, C22:0-, C24:1- and C24:0- ceramides were found to be significantly enriched in IRE1A-transduced livers (Figure 5A). Sphinganine, sphingosine 1-phosphate, C20:0-, C22:0-, C24:1- and C24:0-ceramides were significantly enriched in IRE1A-stimulated EVs (Figure 5B). The rate-limiting step in de novo ceramide synthesis is the formation of 3-ketosphinganine which is catalyzed by serine palmitoyltransferase (SPT). In humans and mice, SPT is comprised of 2 of 3 paralogous subunits Sptlc1, Sptlc2 or Sptlc3, of which Sptlc1 and Sptlc2 are expressed in the liver.²⁴ To further examine the genetic regulation of SPT by IRE1A, we focused on XBP1. In silico analyses determined the presence of putative XBP1 binding sites in the promoter regions of Sptlc1 and Sptlc2 (Figure 6A). Sptlc1 and Sptlc2 mRNA and protein expression was induced by IRE1A activation (Figure 6B,C). To test XBP1 binding to these putative regulatory sites we performed chromatin immunoprecipitation (ChIP) assays. PCR amplification of the chromatin from ChIP revealed that XBP1 occupancy at the promoters of Sptlc1 and Sptlc2 was significantly greater in IRE1A-transduced mouse livers (Figure 6D,E). ERdj4, a canonical XBP1 target, was included to demonstrate XBP1 activation (Figure 6F). Thus, we confirmed increased expression of Sptlc1 and Sptlc2 in the liver in an IRE1A-XBP1 dependent manner leading to an increase in hepatic sphingolipids.

Figure 5. IRE1A induces de novo ceramide biosynthesis.

Sphingolipid content of A) Livers (n=8 each); B) EVs (n=3 each) from AdB-gal or AdIRE1A transduced mice. Two-tailed Student’s t-test was used for statistical analyses.

Figure 6. IRE1A-stimulated EV release is mediated by the transcriptional activation of SPT.

A) Cartoon of putative XBP1 binding sites in the promoter of Sptlc1 and Sptlc2 genes by in silico analysis. B) Relative mRNA expression of Sptlc1and Sptlc2 (n=5 each) and C) Western blotting for SPTLC1 and SPTLC2 (n=3 each) in livers of AdIRE1A or AdB-gal transduced mice. ChIP assay demonstrating XBP1 promoter occupancy of D) Sptlc1, E) Sptlc2, and F) ERdj4 (n=5 each). G) Schematic for myriocin treatment of AdIRE1A-transduced mice. H) Relative mRNA expression of spliced Xbpl; I) Sphingolipid contents of livers; J) EV counts in plasma; K) Sphingolipid content of EVs (n=4 each) from myriocin-treated mice. Two-tailed Student’s t-test was used for statistical analyses.

De novo ceramide synthesis is necessary for IRE1A-induced EV release.

Given the upregulation of de novo ceramide synthesis in IRE1A-transduced livers, we reasoned that the inhibition of ceramide synthesis would prevent the release of EVs. We approached this by administering myriocin, a pharmacological inhibitor of SPT, to mice (Figure 6G). Equal IRE1A expression (data not shown) and lack of inhibition of Xbp1 processing was confirmed in these livers (Figure 6H). By tandem mass spectrometry, we observed a significant reduction in the hepatic levels of sphingomyelin, sphinganine, C14:0, C16:0-, C18:1, C18:0-, C20:0-, C22:0-, C24:1- and C24:0- ceramides in myriocin-treated mice, consistent with inhibition of SPT-mediated de novo ceramide synthesis by myriocin (Figure 6I). Myriocin treatment suppressed IREa-stimulated EV release (Figure 6J). Additionally, EV ceramide content was also reduced in myriocin-treated mice (Figure 6K). Thus, these findings confirm that IRE1A-induced EV release requires SPT-mediated de novo ceramide synthesis.

Increased SPTLC1 and XBP1 abundance are features of human NASH.

We previously demonstrated an increase in circulating EVs in human subjects with NASH.⁷ Here we extended those observations by demonstrating an increase in the abundance of spliced XBP1 and SPTLC1 mRNA in human liver samples from NASH subjects in comparison with normal controls (Figure 7). Using RNAscope we found an increase in XBP1 variant 2 (spliced XBP1) (Figure 7A and B) and SPTLC1 (Figure 7A and C) in NASH livers. We further confirmed that spliced XBP1 and SPTLC1 abundance demonstrated significant collinearity (Figure 7D), consistent with our previously demonstrated transcriptional dependency of SPTLC1 on IRE1A-XBP1 axis (Figure 6). The RNAscope signals for spliced XBP1 and SPTLC1 correlated with the NAFLD activity score (NAS),²⁵ histologic grade of inflammation, and ALT (Supplementary Figure 7A-F). We further confirmed these findings in a smaller number of subjects that had available mRNA and NAS, demonstrating a significant correlation between the abundance of spliced XBP1 and SPTLC1 (Figure 7E), and their correlation with histologic grade of inflammation (Supplementary Figure 7G, H). Finally, we correlated plasma EVs counts with the NAS and grade of inflammation to demonstrate that EVs correlated both with NAS and inflammation (Figure 7F, G). Taken together, these findings illustrate the activation of IRE1A-XBP1 signaling, the expression of XBP1 transcriptional target SPTLC1, and their correlation with NAS and EV count, and an increase in EV counts with increasing grade of inflammation in NASH.

Figure 7. Spliced XBP1 and SPTLC1 are increased in human NASH.

A) Representative images of RNAScope ISH using specific probes against spliced XBP1 (red dots, black arrow head) and SPTLC1 (green dots, black arrow) in human liver sections (normal and NASH); positive control is stained for POLR2A-C2 (red dots) and PPIB-C1 (green dots), scale bar=50 μm; B) quantification of spliced XBP1 signal and C) SPTLC1 signal in high power fields (n=7 in normal liver, n=10 in NASH). Graph showing collinearity between D) Spliced XBP1 signal and SPTLC1 signal by RNAscope (n=17); E) Relative mRNA expression of spliced XBP1 and SPTLC1 (n=13); F) EV count and NAFLD activity score (NAS) (n=27); G) EV count and grade of inflammation (n=27). Two-tailed Student’s t-test was used for statistical analyses. Pearson correlation analysis was performed in Microsoft excel.

DISCUSSION

The activation of IRE1A is a feature of sterile inflammatory liver diseases including NASH where it has been implicated in adaptive and maladaptive processes.^{26, 27} Here we explored the link between IRE1A activation and inflammation in NASH and report how IRE1A activation promotes the release of proinflammatory EVs. We demonstrate that inhibition of IRE1A leads to a reduction in circulating EVs, macrophage accumulation in the liver, injury and inflammation. Furthermore, we confirm that the activation of IRE1A leads to the release of EVs from hepatocytes. These EVs are sufficient to attract monocyte-derived macrophages into the liver. IRE1A-XBP1 activation transcriptionally upregulates ceramide biosynthesis with concomitant increases in hepatic and EV ceramides. IRE1A-stimulated EV biogenesis requires ceramide synthesis. Altogether, these data establish a novel link between ER stress and proinflammatory signaling in the liver mediated by IRE1A-induced release of proinflammatory EVs.

IRE1A is a type-1 transmembrane endoplasmic reticulum resident protein with cytosolic kinase and endoribonuclease (RNase) activities.²⁸ IRE1A activation results in generation of XBP1, a critical component of the transcriptional output of IRE1A signaling, regulated IRE1A-dependent decay (RIDD), and phosphorylation-induced activation of the stress kinase c-Jun N-terminal kinase (JNK). Due to these diverse outputs, IRE1A activation can result in divergent cell fates during ER stress and is also linked to inflammatory signaling. IRE1A activation can occur canonically under proteotoxic ER stress or noncanonically under lipotoxic conditions.^{29, 30} In keeping with the later observations, here we show that IRE1A is activated in a dietary mouse model of NASH and in human NASH livers. We examined the pathophysiological relevance of IRE1A activation using a dietary model of NASH and pharmacological and genetic approaches. We demonstrate a reduction in circulating EVs and associated macrophage accumulation, inflammation, and fibrosis in the liver. In this work we focused on macrophage responses, though we cannot exclude a direct effect of IRE1A-stimulated EVs on HSC. The beneficial effect of inhibiting IRE1A is in keeping with other reports, for example, pharmacologic inhibition of IRE1A reduced atherosclerotic plaque size in mouse models.¹⁵ Constraining IRE1A activation by restoring the hepatic expression levels of Bax inhibitor 1 (BI-1) improved insulin resistance.³¹ BI-1 knockout mice developed exaggerated XBP1 processing and inflammation.³² Though ceramides and circulating EVs were not measured in these studies, they could have mediated liver macrophage accumulation and inflammation.

EVs are membrane defined nanoparticles that perform cell-to-cell communication, both in healthy and disease states.³³ Healthy EVs are implicated in the maintenance or restoration of homeostasis.³⁴ On the other hand, EVs from stressed cells can activate myriad responses in recipient cells such as inflammation and pro-fibrogenic signaling.^4,35 Recent studies have demonstrated an increase in circulating EVs in sterile inflammatory diseases including mouse models of NASH and in humans with NASH.^7,33 However, how EVs are formed and released from stressed cells and concomitant changes in bioactive cargoes contained within EVs are not well understood. EV biogenesis may occur in an ESCRT-dependent, or ESCRT-independent ceramide-dependent manner. Here, utilizing a reporter mouse, we demonstrate that IRE1A activation specifically induces EV release from hepatocytes in vivo. The release of IRE1A-stimulated EVs is dependent on the transcriptional activity of XBP1 and occurs via an increase in de novo ceramide biosynthesis. Our data demonstrate that XBP1 transcriptionally upregulates SPT, the rate limiting enzyme of de novo ceramide biosynthesis. Matching livers and EVs demonstrate a parallel increase in ceramides. Furthermore, pharmacological inhibition of SPT prevents hepatic ceramide enrichment and the release of EVs without impacting Xbp1 splicing, placing SPT downstream of XBP1 in this pathway. Thus, we have elucidated the mechanism by which IRE1A activation stimulates the release of EVs from hepatocytes in a ceramide-dependent manner, likely leading to an increase in the number of EVs in the hepatic microenvironment. The IRE1A-XBP1 pathway regulates many aspects of lipid synthesis which is necessary for ER expansion in ER stress and also hepatic lipid metabolism.³⁶ The later include VLDL assembly and secretion and the prevention of ER stress-induced steatosis via regulation of C/ebpβ and Pparγ.^{37, 38} Recently, IRE1A-XBP1 signaling was shown to transcriptionally upregulate cyclooxygenase 2 and prostaglandin E2 which mediate inflammation-associated pain.³⁹ Our findings expand the known regulation of various lipid classes by IRE1A-XBP1 signaling.

Hepatic macrophage infiltration is a key lesion in the sterile inflammatory response in progressive NASH; therefore, we examined the role of IRE1A-stimulated EVs in regulating hepatic macrophage populations. Using flow cytometry we found that IRE1A-stimulated hepatocyte-derived EVs are sufficient to increase the accumulation of monocyte-derived macrophages in the liver. IRE1A inhibition in mice with dietary NASH showed a corresponding reduction in monocyte-derived macrophages. High dimensional interrogation of intrahepatic leukocytes following IRE1A activation confirmed the key role of monocyte-derived macrophages in the inflammatory response to this sterile inflammation. Our previous observations support a role for lipid mediators, specifically ceramide-derived sphingosine 1-phosphate, in myeloid cell chemotaxis and in liver inflammation in a dietary NASH model. ^{7, 8, 12} In the current study IRE1A activation upregulates ceramide biosynthesis with evident increases in hepatic and EV ceramides. Therefore, we propose that ceramide-derived sphingosine 1-phosphate is the bioactive cargo in IRE1A-stimulated EVs in terms of macrophage chemotaxis, and indeed we demonstrate macrophage chemotaxis toward IRE1A-stimulated EVs in our microfluidic chemotaxis assay. EV number and cargo can both influence biological responses. As IRE1A stimulates the release of EVs, we propose that an increase in EV numbers in the hepatic microenvironment, which indirectly increases the local availability of signaling sphingolipids, plays a role in attracting macrophages. These observations suggest that EVs transmit hepatocellular ER stress to macrophages thus, promoting liver inflammation. Inhibition of the enzyme dihydroceramide desaturase 1 ameliorated hepatic steatosis and insulin resistance,⁴⁰ and we have shown that pharmacological inhibition of ceramide-derived sphingosine 1-phosphate signaling ameliorates NASH.⁸ Indeed, this is associated with lower circulating EVs suggesting that inhibition of EV release may be a potential therapeutic strategy for NASH.

Lastly, we correlate our mechanistic observations from mice with steatohepatitis to human NASH liver and plasma samples. We demonstrate that hepatic XBP1 and SPTLC1 are upregulated in human NASH. Furthermore, we demonstrate a significant correlation between expression of XBP1, SPTLC1 and circulating EVs with NAS and grade of inflammation. These observations will need further validation in larger sample sizes, given the small numbers available for inclusion in our study. Others have demonstrated an increase in hepatic ceramides and sphingomyelin levels in NAFLD and reported them to be higher in NASH in comparison with isolated steatosis.^{41, 42}

In conclusion, we have identified a novel role for IRE1A-XBP1 signaling in driving the biogenesis and release of proinflammatory EVs mediated by the de novo synthesis of ceramide. Our studies establish EVs as transmitters of a cellular stress response, i.e., IRE1A-activation, to macrophages. Our previous studies demonstrate a role for sphingosine 1-phosphate in mediating the biologic effects of EVs on macrophages. Since the functional consequences of EVs can depend on the number or cargo, our future studies will address the possibility of finding the existence and function of other IRE1A-dependent cargoes sorting into the EVs in NASH. Lastly, these studies identify a pathway that could be potentially targeted for the treatment of human NASH.

What you need to know:

Background and Context: Endoplasmic reticulum to nucleus signaling 1 (ERN1, also called IRE1A) is activated, as measured by increased mRNA levels of the gene it regulates, spliced X-box binding protein 1 (XBP1), in liver cells of patients with nonalcoholic steatohepatitis (NASH). Hepatocytes release ceramide-enriched inflammatory extracellular vesicles (EVs) following activation of IRE1A

New Findings: In mouse hepatocytes, activated IRE1A (via XBP1) promotes transcription of the serine palmitoyltransferase genes, resulting in ceramide biosynthesis and release of EVs. The EVs recruit monocyte-derived macrophages to liver, resulting in inflammation and injury in mice with diet-induced steatohepatitis. Liver XBP1 and serine palmitoyltransferase and plasma EVs are all increased in patients with NASH.

Limitations: This study was performed in mice and human cells and tissues; further studies are needed in humans.

Impact: Strategies to block this pathway might be developed to reduce liver inflammation in patients with NASH

Lay Summary: Cells in fatty liver release extracellular vesicles, which communicate with white blood cells called macrophages. This contributes to a pathway by which the liver becomes inflamed in patients with nonalcoholic steatohepatitis.

Abbreviations:

AAV

Adeno-associated virus

ALT

Alanine aminotransferase

BMDM

Bone-marrow derived macrophage

ChIP

Chromatin immunoprecipitation

EM

Electron microscopy

ER

Endoplasmic reticulum

EVs

Extracellular vesicles

FFC

Fat, fructose, and cholesterol

HSC

Hepatic stellate cell

IRE1A

Inositol-requiring enzyme-1a

NAFLD

Nonalcoholic fatty liver disease

NASH

nonalcoholic steatohepatitis

NTA

Nanoparticle tracking analysis

PMH

Primary mouse hepatocyte

PET/CT

Positron Emission Tomography/ Computed Tomography

SPT

Serine palmitoyltransferase

t-SNE

t-Distributed Stochastic Neighbor Embedding

XBP1

X-box binding protein1