A systems biology approach utilizing a mouse diversity panel identifies genetic differences influencing isoniazid-induced microvesicular steatosis

Abstract

Isoniazid (INH), the mainstay therapeutic for tuberculosis infection, has been associated with rare but serious hepatotoxicity in the clinic. However, the mechanisms underlying inter-individual variability in the response to this drug have remained elusive. A genetically diverse mouse population model in combination with a systems biology approach was utilized to identify transcriptional changes, INH-responsive metabolites, and gene variants that contribute to the liver response in genetically sensitive individuals. Sensitive mouse strains developed severe microvesicular steatosis compared with corresponding vehicle control mice following 3 days of oral treatment with INH. Genes involved in mitochondrial dysfunction were enriched among liver transcripts altered with INH treatment. Those associated with INH treatment and susceptibility to INH-induced steatosis in the liver included apolipoprotein A-IV, lysosomal-associated membrane protein 1, and choline phosphotransferase 1. These alterations were accompanied by metabolomic changes including reduced levels of glutathione and the choline metabolites betaine and phosphocholine, suggesting that oxidative stress and reduced lipid export may additionally contribute to INH-induced steatosis. Finally, genome-wide association mapping revealed that polymorphisms in perilipin 2 were linked to increased triglyceride levels following INH treatment, implicating a role for inter-individual differences in lipid packaging in the susceptibility to INH-induced steatosis. Taken together, our data suggest that INH-induced steatosis is caused by not one, but multiple events involving lipid retention in the livers of genetically sensitive individuals. This work also highlights the value of using a mouse diversity panel to investigate drug-induced responses across a diverse population.

FIG. 1.

INH induces inter-strain variations in hepatocytic microvesicular steatosis and triglyceride accumulation. (A) Microvesicular steatosis severity scores for vehicle- (white circles) and INH-treated (black squares) animals. Data are represented as mean ± SE for N = 4 mice per treatment and strain, except for strains noted in the Materials and Methods section. Strains are ordered from left to right by increasing average severity score for the INH-treated animals. *p < 0.05 indicates the difference between vehicle- and INH-treated animals. +/- identifies strains sensitive and resistant to INH-induced steatosis as determined by significance following a Mann-Whitney U-test. (B) Representative photomicrographs (400X) of liver tissue from vehicle- and INH-treated animals in the NON/ShiLtJ and LG/J strains. (C) Liver triglyceride (TG) levels for vehicle- (white circles) and INH-treated (black squares) animals. Data are represented as mean ± SE for N = 4 mice per treatment and strain, except for strains noted in the Materials and Methods section. Strains are ordered from left to right by increasing average microvesicular steatosis severity scores for the INH-treated animals. *p < 0.05 indicates the difference between vehicle- and INH-treated animals. Significance was evaluated by a Student's t-test. (D) Triglyceride concentrations for vehicle- and INH-treated mice were binned according to microvesicular steatosis severity score. Values represent mean ± SE of liver triglyceride concentrations. Individual means across all strains for each bin were statistically significant (one-way ANOVA). Post hoc statistical significance was determined using a Tukey's multiple comparison test. *p < 0.001 indicates the difference relative to severity scores of 0.

FIG. 2.

Gene expression analysis identifies mitochondrial transcript changes associated with INH treatment. mRNA expression analysis was performed in liver tissue derived from all vehicle- and INH-treated animals in the study. An ANOVA model using the fixed categorical factors strains (resistant vs sensitive to INH-induced microveiscular steatosis) and treatment (vehicle vs. INH) was then applied to the intensity data to determine which probesets were significantly altered by each experimental factor individually or in concert. (A) The number of probesets significantly altered by sensitivity to steatosis and treatment is depicted in a Venn diagram. Canonical pathways enriched among genes significantly associated with treatment (p < 0.05) but not strains (p > 0.05) were identified using the Tox Analysis feature in IPA. The top 10 pathways are represented. Relative gene expression changes in the top pathway, mitochondrial dysfunction (bold), are represented in a heat map. (B) A heat map representing relative gene expression changes the 26 probesets significantly altered by both factors: strains and treatment. The color scales for both heat maps represent the relative ratio of log₂ (intensity).

FIG. 3.

Liver cholesterol and serum LDL/VLDL concentrations in select strains binned by microvesicular steatosis severity scores. (A) Liver cholesterol and (B) serum LDL/VLDL evaluated from two drug-resistant strains (steatosis severity score in INH-treated animals of 0–1; NON/ShiLtJ, C57BR/cdJ) and three drug-sensitive strains (microvesicular steatosis severity score in INH-treated animals of 2–5; NOR/LtJ, DBA/J, and LG/J). All animals were binned according to low (0–1) or high (2–5) severity scores as done for analysis of metabolomics data. Data are represented as mean ± SE. *p < 0.05 indicates the difference between groups as assessed by a Student's t-test.

FIG. 4.

GWA mapping using average fold change in liver triglyceride identifies QTLs associated with sensitivity to INH-induced microvesicular steatosis. (A) Manhattan plot of haplotype associations. Black line indicates the threshold for significance, −log₁₀(p) = 3.5. Circled numbers identify the QTL peaks. (B) QTL regions identified by GWA mapping. Plin2 (bold) was selected for further analysis.

FIG. 5.

Polymorphisms in Plin2 are associated with INH-induced lipid accumulation in the liver. (A) Average INH-induced liver triglyceride (TG) fold change and (B) average INH-induced Plin2 gene expression fold change for each strain are plotted by genotype. (C) Severity scores for adipophilin (Plin2) staining evaluated in liver tissue for vehicle-treated (white bars) and INH-treated (black bars) mice are plotted by genotype. Data are represented as mean ± SE. *p < 0.05 indicates the difference between groups as assessed by a Student's t-test.

FIG. 6.

Schematic for INH-induced lipid accumulation in genetically sensitive mouse strains. Transcriptomic analysis identified that transcriptional changes in Apoa4 and Lamp1 are associated with INH-induced microvesicular steatosis in sensitive strains. As a result, triglyceride levels are increased due to increased lipoprotein uptake and decreased lipid export. Triglycerides and sterols are then directed toward packaging in lipoproteins and lipid droplets. Transcriptomic and metabolomic analysis indicated decreased phosphatidylcholine biosynthesis due to lower levels of Chpt1, reducing the capacity for lipoprotein packing and favoring lipid droplet formation. QTL analysis identified polymorphisms in Plin2 that may inhibit lipid droplet hydrolysis resulting in lipid droplet accumulation and steatosis. Points at which INH may affect these processes by increasing (red) or decreasing (blue) key metabolites and transcripts are highlighted.