Molecularly defined unfolded protein response subclasses have distinct correlations with fatty liver disease in zebrafish

Abstract

The unfolded protein response (UPR) is a complex network of sensors and target genes that ensure efficient folding of secretory proteins in the endoplasmic reticulum (ER). UPR activation is mediated by three main sensors, which regulate the expression of hundreds of targets. UPR activation can result in outcomes ranging from enhanced cellular function to cell dysfunction and cell death. How this pathway causes such different outcomes is unknown. Fatty liver disease (steatosis) is associated with markers of UPR activation and robust UPR induction can cause steatosis; however, in other cases, UPR activation can protect against this disease. By assessing the magnitude of activation of UPR sensors and target genes in the liver of zebrafish larvae exposed to three commonly used ER stressors (tunicamycin, thapsigargin and Brefeldin A), we have identified distinct combinations of UPR sensors and targets (i.e. subclasses) activated by each stressor. We found that only the UPR subclass characterized by maximal induction of UPR target genes, which we term a stressed-UPR, induced steatosis. Principal component analysis demonstrated a significant positive association between UPR target gene induction and steatosis. The same principal component analysis showed significant correlation with steatosis in samples from patients with fatty liver disease. We demonstrate that an adaptive UPR induced by a short exposure to thapsigargin prior to challenging with tunicamycin reduced both the induction of a stressed UPR and steatosis incidence. We conclude that a stressed UPR causes steatosis and an adaptive UPR prevents it, demonstrating that this pathway plays dichotomous roles in fatty liver disease.

Keywords: ER stress; Fatty liver disease; Steatosis; Thapsigargin; Tunicamycin; Unfolded protein response; Zebrafish.

Fig. 1.

Tm is the only ER stressor that induces fatty liver. (A,B) Representative images of 5-dpf larvae treated for 48 hours with DMSO, BFA, Tg or Tm at the indicated concentrations. (A) Live transgenic Tg(fabp10:dsRed) larvae were imaged. (B) Fixed larvae were stained for neutral lipids with Oil Red O. Livers are circled. (C) Larvae treated with DMSO, 1 and 2 μg/ml BFA, 0.25, 0.5 and 0.75 μM Tg or 1 μg/ml Tm from 3–5 dpf were stained with Oil Red O and scored for the presence (black bars) or absence (white bars) of steatosis. The total number of larvae scored and the number of clutches are indicated. **P<0.01, ***P<0.0001, calculated using Fisher’s exact test. Scale bars: 200 μm (A), 1 mm (B).

Fig. 2.

BFA, Tg and Tm differentially induce UPR sensors in the liver. (A,B) Standard PCR (A) and qPCR (B) analysis of xbp1 splicing in the livers of fish treated from 3 to 5 dpf with DMSO, 1 μg/ml BFA, 0.75 μM Tg or 1 μg/ml Tm using primers that amplify both unspliced (-u) and spliced (-s) xbp1. The average ratio of xbp1-second to xbp1-t is indicated (n=2). (C) Protein extracts from livers dissected from 5-dpf larvae treated as in A were blotted with anti-P-Eif2a and anti-tubulin as a loading control. Quantification and normalization to tubulin and DMSO controls is shown (n=2). (D) Analysis of atf4 and atf6 expression in livers of fish treated as in A. In B and D, target gene expression was normalized to rpp0 and fold changes compared to DMSO are plotted; n=13 for DMSO, 10 for Tm, 8 for Tg and 6 for BFA. Bars represent standard error in all graphs. *P<0.05, **P<0.01, calculated using paired Wilcoxon test.

Fig. 3.

BFA, Tg and Tm differentially induce UPR target genes in the liver. (A) qPCR analysis of bip, dnajc3, edem1 and ddit3 in cDNA from dissected livers of 5-dpf larvae treated with DMSO, 1 μg/ml BFA, 0.75 μM Tg and 1 μg/ml Tm. The fold change relative to DMSO is plotted with error bars indicating the standard error; n=13 for DMSO, 10 for Tm, 8 for Tg and 6 for BFA. (B) Liver protein lysates from larvae treated as in A were immunoblotted with anti-Bip and anti-tubulin as a loading control. Band intensities were assessed, normalized to the loading control and plotted with error bars indicating the standard error; n=2. (C) Heat map of UPR target gene expression in the livers of larvae treated as in A with genes in columns and individual clutches of fish in rows. Red correlates with +Log 2 fold changes and blue correlates with −Log 2 fold changes, which are normalized across each column. Unsupervised clustering revealed three major UPR subclasses termed homeostatic, intermediate and stressed. Brackets reflect the results of unsupervised clustering of the genes and samples. *P<0.05, **P<0.01, ***P<0.001, calculated using paired Wilcoxon test.

Fig. 4.

Tm dose-response data reveals a nonlinear effect on hepatic protein glycosylation, Bip protein accumulation, xbp1 splicing and steatosis incidence. (A,B) Protein extracts from individual transgenic Tg(l-fabp:Gc-EGFP) larvae exposed to a range of Tm concentrations from 3 to 5 dpf were loaded in each lane and immunoblotted using anti-GFP (A) or anti-Bip (B). In A, PNGase treatment of extracts from DMSO-treated larvae revealed a faster migrating band corresponding to hypoglycosylated Gc-EGFP. Band intensities were plotted as the percentage of total Gc-EGFP that was either hypoglycosylated or fully glycosylated (n=3). One-way ANOVA was used to determine significance (P<0.05), where all samples marked with a letter are significantly different from DMSO controls, and those with different letters are significantly different from each other. In B, Bip protein levels in Tm-treated samples were compared with DMSO control and expressed as a fold change (n=2). (C) PCR analysis of unspliced (xbp1-u) and spliced (xbp1-s) xbp1 mRNA in livers of 5-dpf larvae treated as in A. The percentage of xbp1-second from total xbp1 is plotted (n=2). (D) Zebrafish larvae were treated as in A, stained with Oil Red O and scored for steatosis. Estimated log odds ratios of percentage steatosis between different concentrations of Tm and DMSO controls are plotted. *P<0.05, **P<0.01, ***P<0.005, calculated using paired Wilcoxon test.

Fig. 5.

UPR activation positively correlates with steatosis incidence. (A) Heat map of UPR target gene expression in the liver, with genes in columns and individual clutches in rows. Blue and red indicates below and above average, respectively, for each column. Supervised clustering was performed according to Tm concentration (first) and steatosis incidence (second) and was unsupervised for gene order. Brackets reflect the results of unsupervised clustering of the genes. The colored bar on the left represents steatosis incidence (see legend: dark blue, 0%; white, 50%; dark red, 100%). The raw data are listed in supplementary material Table S2. (B) First principal component (PC1) of the UPR gene expression for each clutch was plotted according to Tm concentration, with the box indicating the 75th and 25th quartile of the data, the whiskers indicating the 90th and 10th percentile and the cross line indicating the median. (C) Scatter plot representing correlation between PC1 and steatosis incidence in larvae treated with different doses of Tm. Each dot represents a single clutch and is assigned a different color corresponding to UPR subclass (blue, white and red characterize homeostatic, intermediate and stressed UPR, respectively). A least squares linear model fit (blue line) with 95% prediction error bands (gray overlay) was superimposed. (D) Plot of PC1 for DMSO (control; blue), 1 μg/ml BFA (green), 0.75 μM Tg (orange) and 1 μg/ml Tm (red). (E) Scatter plot representing correlation between PC1 and steatosis incidence in larvae treated with DMSO (blue), 0.75 μM Tg (orange) and 0.25–1 μg/ml Tm (red) using the same model as in C. The correlation between PC1 and steatosis has P<0.001, determined by the Spearman test. (F) Kernel density estimate of the distribution of the UPR-PC1 signature in three sets of clinical samples. Per-group average is marked with a thick black vertical line. *P<0.1, **P<0.01, ***P<0.005, calculated using paired Wilcoxon test.

Fig. 6.

UPR target gene activation precedes Tm-induced steatosis. (A–C) Larvae were treated starting on 3 dpf with DMSO and 0.25 μg/ml Tm and collected at the indicated time points. At least 15 larvae were stained with Oil Red O and scored for steatosis (A) and at least 5 livers were dissected for RNA extraction (B,C). (A) The log odds ratio of steatosis was plotted versus time (see Materials and Methods and supplementary material Table S3). (B) UPR target gene analysis measured by qPCR in liver cDNAs from the same cohorts assessed for steatosis in A. Expression of each gene is plotted as log fold change (log FC) over time (see Materials and Methods and supplementary material Table S7). (C) PCR analysis of xbp1 splicing with the percentage of spliced/total xbp1 for each sample plotted (n=2). *P<0.05, **P<0.01, ***P<0.001.

Fig. 7.

An adaptive UPR protects against Tm-induced FLD. (A) Experimental design for the adaptation protocol. Larvae were treated starting at 72 hpf with 0.75 μM Tg (adapted) or DMSO (naïve) for 24 hours, and the compounds were then washed out (WO) for 1 hour and the larvae were challenged with exposure to 0.25 μg/ml Tm or DMSO (control). Livers were dissected after 4 hours of exposure for qPCR analysis (B), for PC1 analysis (C) and after 24 hours of exposure fish were stained with Oil Red O (D). (B) ΔCt values of UPR genes in the livers of larvae treated as in A and collected at 101 hpf, after 4 hours of Tm challenge. Black lines indicate the median. (C) PC1 generated from qPCR analysis in B from samples treated as in A were plotted against the different treatments. See also supplementary material Table S8. (D) Larvae were treated with DMSO or Tg and then challenged with Tm and collected at 5 dpf as shown in A, stained with Oil Red O and scored for steatosis. (E) Model depicting the mechanism whereby Tg pretreatment induces an adaptive UPR that protects against Tm induction of a stressed UPR and steatosis. ***P<0.005, calculated using Fisher’s exact test.