GCN2 drives diurnal patterns in the hepatic integrated stress response and maintains circadian rhythms in whole body metabolism during amino acid insufficiency

Abstract

Disruptions in circadian rhythms are associated with an increased risk of developing metabolic diseases. General control nonderepressible 2 (GCN2), a primary sensor of amino acid insufficiency and activator of the integrated stress response (ISR), has emerged as a conserved regulator of the circadian clock in multiple organisms. The objective of this study was to examine diurnal patterns in hepatic ISR activation in the liver and whole body rhythms in metabolism. We hypothesized that GCN2 activation cues hepatic ISR signaling over a natural 24-h feeding-fasting cycle. To address our objective, wild-type (WT) and whole body Gcn2 knockout (GCN2 KO) mice were housed in metabolic cages and provided free access to either a control or leucine-devoid diet (LeuD) for 8 days in total darkness. On the last day, blood and livers were collected at CT3 (CT = circadian time) and CT15. In livers of WT mice, GCN2 phosphorylation followed a diurnal pattern that was guided by intracellular branched-chain amino acid concentrations (r² = 0.93). Feeding LeuD to WT mice increased hepatic ISR activation at CT15 only. Diurnal oscillations in hepatic ISR signaling, the hepatic transcriptome including lipid metabolic genes, and triglyceride concentrations were substantially reduced or absent in GCN2 KO mice. Furthermore, mice lacking GCN2 were unable to maintain circadian rhythms in whole body energy expenditure, respiratory exchange ratio, and physical activity when fed LeuD. In conclusion, GCN2 activation functions to maintain diurnal ISR activation in the liver and has a vital role in the mechanisms by which nutrient stress affects whole body metabolism.

NEW & NOTEWORTHY This work reveals that the eIF2 kinase GCN2 functions to support diurnal patterns in the hepatic integrated stress response during natural feeding and is necessary to maintain circadian rhythms in energy expenditure, respiratory exchange ratio, and physical activity during amino acid stress.

INTRODUCTION

The evolution of life on our planet centers around the Earth’s rotation on its axis, creating an ∼24-h solar day. Daily changes in temperature, sunlight, and food intake are among the important environmental cues called zeitgebers that help entrain the body’s endogenous circadian clocks with cyclic changes in the environment, driving biological rhythms in physiology, metabolism, and behavior. In mammalian systems, the central clock is located in the suprachiasmatic nucleus (SCN) of the hypothalamus, whereas the peripheral clocks are found in almost all other tissues within the organism (1). Both the central and peripheral circadian clocks consist of intracellular genetic feedback circuits, known as core regulatory loops, which establish biochemical oscillations over a period of ∼24 h. Input signals from zeitgebers entrain core regulatory loops and subsequent physiological and behavioral outputs in a rhythmic pattern.

Although metabolic rhythms are regulated by the interactions between the SCN and the peripheral clocks, universal cues, such as food intake, also play a significant role in rhythmic metabolic regulation within nutrient-sensitive tissues such as the liver (2, 3). The liver is a highly metabolic peripheral tissue that is exquisitely sensitive to nutrients and especially amino acids. Changes in meal timing can disrupt the functions of the liver clock. Indeed, the timing of food intake serves as the primary regulator of circadian rhythms in the liver and other peripheral tissues (4). These findings have spurred further investigation into the specific macronutrients responsible for entraining the clocks within peripheral tissues and the signaling mechanisms involved. In this regard, the majority of available information is focused on lipid content and insulin signaling (5–9). A role for protein/amino acids is suggested (10), but this has yet to be tested directly.

The integrated stress response (ISR) plays a crucial role in maintaining proteostasis by relaying stress signals to the translational machinery via four distinct kinases (11). This cascade leads to the inhibition of translation initiation and general protein synthesis through the phosphorylation of eukaryotic initiation factor 2α (eIF2α) (12). Phosphorylation of eIF2α also promotes selective translation of stress response genes, including activating transcription factor 4 (ATF4), in an effort to re-establish homeostasis (13, 14). General control nonderepressible 2 (GCN2), an eIF2 kinase and sensor of amino acid insufficiency, is shown to play a key role within circadian translational regulation in multiple experimental models. Under conditions of amino acid insufficiency, GCN2 undergoes activation through phosphorylation, triggered by its direct sensing of either the accumulation of uncharged tRNA species or ribosomal stalling and collisions (15–22). Using Neurospora crassa as a model organism, researchers demonstrate that eIF2 phosphorylation follows a circadian pattern that is driven by daily oscillations in charged tRNA levels via the GCN2 ortholog, CPC-3 (23). Similarly, mouse studies show that ISR activation in the SCN follows a circadian pattern in a GCN2-dependent manner (24). Furthermore, this cyclic ISR activation is necessary to maintain circadian rhythm in Period2 (Per2) mRNA expression, a core circadian oscillatory gene, via rhythmic ATF4 translation in the SCN of mice (24). In the liver, GCN2 is a primary sensor of amino acid insufficiency, coordinating nutrient sensing upon dietary or drug-induced amino acid starvation with other signal transduction pathways (25–27). Robust activation of the ISR is reported in rodents fed diets lacking leucine, and mice lacking GCN2 show severe hepatic steatosis upon feeding on a leucine-devoid diet (28–30). Furthermore, GCN2 is required for increased ATF4 synthesis upon leucine deficiency and other forms of amino acid insufficiency (31, 32).

In this study, we hypothesized that diurnal patterns in the hepatic ISR are directed by GCN2 activation. Indeed, we observed that hepatic GCN2 activity inversely correlates with intracellular branched-chain amino acid (BCAA) concentrations and drives diurnal ISR activation and transcriptional execution in the liver during ad libitum feeding. We also found that sensing of dietary amino acid insufficiency by GCN2 functions to maintain diurnal patterns in lipid metabolism and circadian rhythms in energy expenditure during nutrient stress. These results expand the role of GCN2 as a circadian regulator to the peripheral circadian clock in mammals and highlight its pivotal role in maintaining metabolic rhythms under nutrient stress.

MATERIALS AND METHODS

Animal Model

Animal procedures and experimental protocols were approved by the Institutional Animal Care and Use Committee at Rutgers, The State University of New Jersey. This study used male C57BL/6J wild-type (WT) mice and whole body Gcn2^–/– (GCN2 KO) mice backcrossed onto the C57BL/6J genetic background for at least 10 generations (28, 29). These mice were between 16 and 29 wk old at the beginning of the experiment, representing adulthood. Activation of the ISR across this age range responds similarly in multiple models of amino acid insufficiency (33, 34). However, we previously reported differences in the ISR according to biological sex with feeding a diet restricted in sulfur amino acids (35). In that work, males were more responsive to activation of the ISR to dietary amino acid restriction. Based on those observations, we decided to conduct this experiment in males. All mice were bred and maintained at the Bartlett animal facility on the Rutgers University Cook campus. Mice were housed in a climate-controlled room with temperature maintained at ∼23°C and humidity between 40% and 60%. Prior to the study, all mice had free access to commercial pelleted food (5001 Laboratory Rodent Diet; LabDiet) and purified water and were maintained on a 12-h light-dark cycle with same-sex littermates until experimental group assignment, wherein mice were housed in individual plastic cages with soft bedding and environmental enrichment.

Group Assignment and Blinding

WT and GCN2 KO mice were first assessed for body weight and then randomly assigned to experimental treatment groups. After group assignments, the animals were deidentified using alphanumerical coding.

Experimental Diets

Over the course of the study, mice were provided either a purified amino acid-replete diet (control) containing 1.11% leucine by weight or an isocaloric and isonitrogenous diet devoid of leucine (LeuD). The diet compositions and catalog numbers are detailed in Supplemental Table S1 (Dyets, Inc., Bethlehem, PA). The experimental diets were formulated to match those used previously by this group (29). Mice had unrestricted access to food and purified water during the experimental feeding period.

Experimental Design

This study was conducted using four independent experimental cohorts of male mice (n = 12 mice per experimental cohort for a study total of n = 48 mice). Each experimental cohort contained six WT and six GCN2 KO mice. Two of the four experimental cohorts were placed within an environmentally controlled comprehensive laboratory animal monitoring system (CLAMS; Columbus Instruments, Columbus, OH) and provided free access to a control diet and purified water while kept on their normal 12-h light-dark cycle. Following a 2-day adjustment period, half the mice within each genotype (n = 3 WT and n = 3 GCN2 KO) were switched to a LeuD diet to represent amino acid insufficiency, whereas the other half received a fresh control diet (n = 3 WT and n = 3 GCN2 KO). Twenty-four hours later, the lights were turned off and remained off for the remaining 8 days of the experiment (Figs. 1A and 3A). At the end of the experiment, mice were killed by decapitation under red light at one of two timepoints: either CT3 (CT = circadian time), which represented the resting phase, or CT15, which represented the active phase. Circadian feeding patterns over the 8-day darkness period were used to select these tissue collection timepoints. The CT15 timepoint allowed mice to consume the majority of their food intake within their active phase feeding window, and the CT3 timepoint represented 12 h later. This same experimental design was conducted twice (n = 12 per cohort, containing 6 WT and 6 GCN2 KO mice) using two environmentally controlled cabinets (I-30L; Percival Scientific Inc., Perry, IA) to control the light:dark cycle as described earlier.

Figure 1.

Circadian rhythms in whole body metabolism are maintained in WT mice independent of changes in leucine intake. A: visual representation of the experimental design. B: hourly energy expenditure measures normalized to total body weight in control and LeuD-fed WT mice over the study period. C: hourly energy expenditure over the 8-day total darkness period separated into two 12-h time periods according to the time of day the measurements were taken. D: hourly wheel-running activity over the study period. E: hourly wheel-running activity over the 8-day total darkness period separated into two 12-h time periods according to the time of day the measurements were taken. F: hourly food intake normalized to total body weight over the study period. G: hourly food intake over the 8-day total darkness period separated into two 12-h time periods according to the time of day the measurements were taken. The yellow backdrop represents the time at which the lights were on in the CLAMS. Black dotted line represents the time which half the mice were switched to a leucine-devoid diet. Hourly CLAMS data are expressed as the mean values. Bar graph values are presented as means ± SE. Bar graphs were analyzed using a three-factor ANOVA with *#P < 0.05 signifying a diet:time interaction. Groups not sharing a common letter indicate P < 0.05 determined by a Tukey’s post hoc test (n = 5 or 6 mice per group). Figure panel A created with BioRender.com. CLAMS, comprehensive laboratory animal monitoring system; CT, circadian time; DD, total darkness; LeuD, leucine-devoid diet; WT, wild type.

Figure 3.

LeuD feeding increases hepatic ISR activation in a time-dependent manner. A and B: representative immunoblots and quantified ratios of phosphorylated GCN2 and eIF2α to their respective total forms in control and LeuD-fed WT mice killed at CT3 and CT15. C: linear regression analysis of hepatic GCN2 phosphorylation in relation to total intracellular BCAA concentrations in control-fed WT mice. D: volcano plot of differentially expressed transcripts in LeuD-fed WT mice killed at CT3 relative to control-fed WT mice killed at CT3. E: volcano plot of differentially expressed transcripts in LeuD-fed WT mice killed at CT15 relative to control-fed WT mice killed at CT15. Transcripts are defined as differentially expressed if adjusted P value is <0.05 and absolute (log2 fold change) is >1 and are represented in volcano plots as blue and red dots if they are decreased or increased, respectively. F: hepatic triglyceride levels in control and LeuD-fed WT mice killed at CT3 or CT15. Bar graph values are presented as means ± SE with individual datapoints overlayed. Western blot data are expressed relative to control-fed WT mice killed at CT3. Western blot and triglyceride data were analyzed using a two- or three-factor ANOVA using time, diet, and genotype as variables when appropriate with *P < 0.05 signifying a main effect of diet, $P < 0.05 signifying a main effect of genotype and #$P < 0.05 signifying a time:genotype interaction. Groups not sharing a common letter indicate P < 0.05 determined by a Tukey’s post hoc test (n = 3–6 mice per group). BCAA, branched-chain amino acid; LeuD, leucine-devoid diet; ISR, integrated stress response; WT, wild type.

Environmentally Controlled Chambers

The CLAMS was located in the Bartlett animal facility and featured 12 metabolic cages with running wheels placed within an enclosed cabinet with the ability to control light, temperature, and access to food. Each cage provided real-time monitoring of horizontal and vertical activity, wheel activity, feeding and drinking, oxygen consumption, and CO₂ production every 13 min. Measurements were taken over an 11-day period, of which the first 24 h, considered an acclimation period, were removed from subsequent analyses (36). Percival chambers were located in the Bartlett animal facility. The Percival chambers were used to control the light:dark cycle only. The two Percival chambers used in this study each housed six mice individually in their home cages. Within a single experimental cohort, each Percival chamber contained n = 3 WT and n = 3 GCN2 KO.

Whole Body Calorimetry and Behavior Measurements

Following the study, the CLAMS data were placed into the CalR application (36), allowing for energy expenditure, respiratory exchange ratio, food intake, and wheel-running activity data to be compiled into hourly measurements, which were used for subsequent analyses of diurnal rhythms in whole body metabolism and physical activity. To ensure rigor and reproducibility of the behavioral and whole body metabolic data analyses, CLAMS measurements taken over the study duration were inspected to ensure data continuity and validity during the darkness period. Individual cage data displaying missing food intake measures for >24 h or an implausible rapid decline in oxygen consumption to <50 mL/h were excluded from subsequent food intake and calorimetry analyses, respectively. These exclusions resulted in n = 5 rather than n = 6 for 1) GCN2 KO control, GCN2 KO LeuD, and WT LeuD energy expenditure and respiratory exchange ratio measures and 2) GCN2 KO control and GCN2 KO LeuD food intake measures.

Actogram Generation and Periodogram Analysis

Actograms were generated using the average wheel-running activity for each group over the full duration of the experiment. Actograms and periodograms were generated using the Actogram J plugin (37) within Fiji. Subsequent periodograms were generated using a Fourier analysis within the Actogram J plugin.

RNA Sequencing

Isolated RNA was quantified using a Qubit 2.0 Fluorometer (ThermoFisher Scientific), and RNA integrity was verified using the TapeStation System (Agilent Technologies). RNA sequencing libraries were prepared using the NEBNext Ultra II RNA Library Prep Kit for Illumina (New England Biolabs) following the manufacturer’s instructions. Briefly, mRNAs were initially enriched with Oligod(T) beads. Enriched mRNAs were fragmented for 15 min at 94°C. First-strand and second-strand cDNA were subsequently synthesized. cDNA fragments were end-repaired and adenylated at 3′-ends, and universal adapters were ligated to cDNA fragments, followed by index addition and library enrichment by PCR with limited cycles. The sequencing libraries were validated with the TapeStation System (Agilent Technologies) and quantified by using Qubit 2.0 Fluorometer (ThermoFisher Scientific). Further validation of the libraries was done by quantitative PCR (KAPA Biosystems) of selected genes.

The sequencing libraries were multiplexed and clustered onto a flowcell. After clustering, the flowcell was loaded onto the Illumina HiSeq instrument according to the manufacturer’s instructions. The samples were sequenced using a 2 × 150 bp paired end configuration. Image analysis and base calling were conducted by the HiSeq Control Software. Raw sequence data (.bcl files) generated from Illumina HiSeq was converted into fastq files and demultiplexed using Illumina bcl2fastq 2.20 software. One mismatch was allowed for index sequence identification. After investigating the quality of the raw data, sequence reads were trimmed to remove possible adapter sequences and nucleotides with poor quality using Trimmomatic v.0.36. The trimmed reads were mapped to the Mus musculus reference genome available on ENSEMBL using the STAR aligner v.2.5.2b. BAM files were generated by this step. Unique gene hit counts were calculated by using feature counts from the Subread package v.1.5.2. Only unique reads that fell within exon regions were counted. After extraction of gene hit counts, the gene hit counts table was used for downstream differential expression analysis.

RNA Sequencing and Gene Ontology Analysis

A comparison of gene expression between the groups of samples was performed using DESeq (38). The Wald test (39) was used to generate P values and log2 fold changes. To ascertain the relative number of transcripts demonstrating discernible changes between our CT3 and CT15 timepoints within each experimental treatment group, we used a threshold of an adjusted P value (q) of <0.05 to filter our gene list and divided the number of resulting genes by the total number of transcripts identified. Meanwhile, genes with q < 0.05 and absolute (log2 fold changes) >1 were deemed to be differentially expressed and used in subsequent Venn diagram and Gene Ontology (GO) analyses. Venn diagrams were generated using VENNY (40). Gene Ontology analyses were performed using the Gene Ontology (GO): Biological Process tool (41). In short, the filtered list of differentially expressed genes was sorted by direction of change and subsequently submitted to the program. Clustered heatmaps were generated using the pheatmap package in R.

SDS-PAGE and Immunoblotting

Protein lysates were prepared from tissues as previously described (32). Briefly, ∼20 mg of frozen liver tissue was pulverized and suspended in ice-cold RIPA lysis solution consisting of 25 mM HEPES, 2 mM EDTA, 10 mM DTT, 50 mM sodium fluoride, 50 mM β-glycerophosphate pentahydrate, 3 mM benzamidine, 1 mM sodium orthovanadate, 0.5% (wt/vol) sodium deoxycholate, 1% (wt/vol) SDS, 1× protease inhibitor cocktail (P8340; Millipore-Sigma), and 5 nM microcystin (33893; Millipore-Sigma). The homogenized lysates were subjected to centrifugation for clarification at 10,000 g for 10 min at 4°C. Equal amounts of total protein were mixed 1:1 (vol/vol) with a 2× sample buffer solution [20% (vol/vol) glycerol, 60 mM Tris (pH 6.8), 2% (wt/vol) SDS, 0.01% (wt/vol) bromophenol blue, and 5% (vol/vol) β-mercaptoethanol], after which samples were heated at 95°C for 4 min and subsequently stored at −80°C until further use. Gel electrophoresis was performed by separating equal amounts of protein (as determined using Pierce BCA Protein Assay, 23227; ThermoFisher Scientific) by electrophoresis in SDS-polyacrylamide gels, and the separated proteins were then transferred onto PVDF membranes. Protein-bound membranes were blocked with a 5% milk solution for 1 h at room temperature before incubation with primary antibodies overnight. All primary antibodies used in the study are commercially available, have been validated by their respective manufacturers, and can be found in Supplemental Table S2. Immunoreactive bands were visualized by first incubating the membranes for 1 h at room temperature with secondary antibody, followed by briefly incubating the membranes with enhanced chemiluminescence (ECL) solution (RPN2235, Cytiva Amersham ECL Select Western Blotting Detection Agent; Cytiva, Marlborough, MA) to image the targeted proteins (FluorChem M; ProteinSimple, San Jose, CA). Densitometry was performed using ImageJ (Fiji, v. 1.0) (35). Values were normalized to the total of the respective proteins phosphorylated and unphosphorylated species and expressed as fold change compared with the WT control-fed animals killed at CT3.

Liver and Serum Amino Acid Concentrations

Serum and hepatic amino acid concentrations were analyzed using HPLC as previously described (26, 35). Subsequent peak identification and area under the curve calculations were performed using Agilent OpenLab (Agilent Technologies, Santa Clara, CA).

Liver Triglyceride Content

Liver triglyceride content was measured using the Triglyceride Assay Kit (ab65336) on frozen liver samples following the manufacturer’s instructions.

Statistical Analysis

Analyses and visualization were performed using R version 4.2.2 (42) and RStudio. The following packages were performed in R to visualize and analyze the data: tidyverse, ggpubr, DescTools, and pheatmap. A Shapiro–Wilk test was used to check for normality. Data that did not follow a normal distribution were normalized using a log2 transformation. Measurements of period length and amplitude of wheel-running activity were calculated using a Fourier analysis. The resulting period and amplitude values were analyzed using a two-factor ANOVA followed by a post hoc analysis using a Tukey’s honestly significant difference (HSD) correction for multiple comparisons to determine the impact of diet and genotype. Diurnal patterns in the hourly calorimetry and behavioral data were analyzed using a three-factor ANOVA using time of day (CT0–12 and CT12–0), genotype, and diet as variables. Similarly, liver triglyceride, serum, and intracellular amino acid levels data were compared using a three-factor ANOVA using time (CT3 and CT15), genotype, and diet as variables. Western blot data were analyzed using either a two- or three-factor ANOVA for GCN2 and eIF2 phosphorylation, respectively. Both two- and three-factor ANOVAs were followed by post hoc analyses using a Tukey’s HSD correction for multiple comparisons. Linear regression modeling was used to generate regression line equations, R, R², and P values to look at the correlations between hepatic intracellular amino acid levels and relative GCN2 phosphorylation. Outputs of the statistical analyses can be found in Supplemental Tables S3–S5, S7, and S9.

RESULTS

Circadian Rhythms in Whole Body Metabolism Are Maintained in WT Mice Independent of Changes in Leucine Intake

To determine if prolonged amino acid insufficiency feeding impacts circadian behavioral and metabolic rhythms, WT mice were placed within an environmentally controlled comprehensive laboratory animal monitoring system and provided free access to a control diet and kept on their normal 12-h light-dark cycle. Following a 2-day adjustment period, half the mice were switched to a LeuD diet, whereas the other half received a fresh control diet. Twenty-four hours later, the lights were turned off and remained off for the remaining 8 days of the experiment for all animals (Fig. 1A). When switched to a LeuD diet, WT mice exhibited increased energy expenditure (EE) and wheel-running activity and reduced food intake between CT12 and CT0. In addition, they showed an overall decrease in respiratory exchange ratio (RER) and, by the end of the experiment, a reduction in body weight (Fig. 1, B–G; Supplemental Fig. S1, A–C). Despite these changes, LeuD feeding did not impact wheel-running period length or amplitude (Fig. 2, A–E). We conclude that in WT mice, whole body rhythms in energy metabolism are insensitive to reductions in dietary protein quality and associated energy intake.

Figure 2.

Circadian wheel-running behavior is maintained under LeuD feeding in WT mice. A: average Actogram of wheel-running activity in WT control-fed mice over the entire study duration. B: average Actogram for wheel-running activity in WT LeuD-fed mice over the entire study duration. The original light-dark (LD) schedule is denoted by the yellow and black overhead bar. The mice were placed in total darkness (DD) beginning on line 4. The time at which mice were switched from the control to LeuD diet or given fresh control diet is denoted by the red and blue arrows, respectively. C: periodogram analysis of the average wheel-running activity under total darkness in WT control and LeuD-fed mice. Period length is denoted by the vertical black line. D: period length of wheel-running activity in control and LeuD-fed WT mice under total darkness. E: period amplitude of wheel-running activity in control and LeuD-fed WT mice under total darkness. Bar graphs are represented as means ± SE with individual datapoints overlayed. Periodogram analysis outputs were analyzed using a two-factor ANOVA using genotype and diet as variables (n = 6 mice per group). LeuD, leucine-devoid diet; WT, wild type.

LeuD Feeding Increases Hepatic ISR Activation in a Time-Dependent Manner

At the end of the 8-day experimental dark period, serum and liver tissues were collected from WT mice at CT3 (resting) and CT15 (active) timepoints. Circulating amino acid concentrations (Supplemental Tables S6 and S7) and intracellular liver amino acid concentrations (Supplemental Tables S8 and S9) were measured at both timepoints alongside ISR activation. Similar to observations reported in the SCN (24), phosphorylation of hepatic GCN2, a measure of its activity, and its substrate eIF2 displayed a diurnal pattern in the WT mice fed a control diet, with a higher proportion of each protein phosphorylated at CT3 relative to CT15 (Fig. 3, A and B). In WT mice fed a control diet, both intracellular and serum branched-chain amino acids (BCAAs) were significantly elevated at CT15 relative to CT3. Furthermore, GCN2 phosphorylation displayed an inverse relationship with intracellular BCAA concentrations and especially leucine (R² = 0.97) (Fig. 3C and Supplemental Fig. S2). Following 8 days of LeuD feeding, both serum and intracellular leucine concentrations decreased relative to control-fed WT mice at CT15 only. These reductions corresponded with increased phosphorylation of GCN2 and eIF2α phosphorylation at CT15 (Fig. 3, A and B). No other amino acid demonstrated a significant linear relationship with GCN2 activation.

In WT mice, LeuD feeding impacted the hepatic transcriptome at both timepoints, but the magnitude of change and the identity of differentially altered transcripts were substantially influenced by the time of day the tissues were collected. Volcano plots illustrate differential gene expression between WT mice fed control and LeuD at the CT3 and CT15 timepoints (Fig. 3, D and E). In line with the eIF2 phosphorylation patterns observed, LeuD feeding increased the transcription of several ISR target genes, including Ddit3, Erp27, and Herpud1, at CT15 only. These results highlight that although leucine deprivation impacts the hepatic transcriptome, the extent of impact is different in the resting versus active phases. To delve deeper into the impact of LeuD feeding on transcriptional patterns in the liver, we compared the quantity and characteristics of transcripts that showed differential expression between resting and active phases in WT mice fed either the control or LeuD diet. Using a false discovery rate (FDR) cutoff (q < 0.05), ∼6.1% of all identified transcripts showed some level of alteration between the CT3 and CT15 timepoints in control-fed WT mice (Supplemental Fig. S3A). Applying the same FDR criteria (q < 0.05), LeuD feeding slightly amplified the overall proportion of genes undergoing changes between the resting and active phases (7.5%) (Supplemental Fig. S3B). Furthermore, the composition of these transcripts was notably different. A Venny analysis showed minimal overlap between the identity of transcripts differentially expressed over time in our control and LeuD-fed mice, with only 37 downregulated and 96 upregulated gene transcripts shared between WT treatment groups (Supplemental Fig. S3C).

When classifying the transcripts upregulated over time in control and LeuD-fed WT mice, Gene Ontology (GO) analysis revealed that both groups exhibited heightened expression of transcripts involved in lipid metabolism from the CT3 to CT15 timepoints. However, the specific identity of these transcripts differed, with LeuD-fed mice showing a transcriptional signature more indicative of changes in triglyceride biosynthesis (Supplemental Fig. S3, A and B). Regardless of these transcriptional changes, diurnal patterns in hepatic triglyceride content were observed in both the WT control and WT LeuD groups (Fig. 3F). These findings demonstrate that diurnal patterns in ISR signaling are present within the liver and are responsive to changes in dietary leucine intake in a time-dependent manner.

GCN2 Is Required to Maintain Circadian Rhythms in Whole Body Energy Expenditure and Food Intake during LeuD Feeding

Given that GCN2 activation follows a diurnal pattern and is responsive to dietary leucine depletion, we next sought to directly address the role of GCN2 in the maintenance of circadian behavioral and metabolic rhythms under basal conditions as well as under nutrient stress. Using the same experimental design used earlier, we compared behavioral and whole body energy metabolism rhythms in GCN2 KO mice fed a control or LeuD diet (Fig. 4A). Under control-fed conditions (Supplemental Fig. S4A), GCN2 KO mice displayed reduced EE during the resting phase as compared with WT mice, whereas circadian rhythms in RER and food intake were unaltered (Supplemental Fig. S4, B–G). In stark contrast to the WT, GCN2 KO mice fed a LeuD diet showed an immediate reduction in energy expenditure, RER, and wheel-running activity (Fig. 4, B–E) (Supplemental Fig. S5, A and B). In addition, LeuD-fed GCN2 KO mice shifted their food intake relative to their control-fed counterparts by decreasing and increasing their food intakes during the active and inactive phases, respectively (Fig. 4, F and G). GCN2 KO mice fed LeuD displayed lowered period amplitude in their wheel-running activity as compared with the control-fed GCN2 KO mice without altering period length (Fig. 5, A–E). GCN2 KO mice fed LeuD lost significant body weight, but the amount was similar to WT mice fed LeuD (Supplemental Fig. S5C). Overall, these results reveal that GCN2 is required for the maintenance of circadian whole body energy expenditure and behavioral rhythms during dietary amino acid insufficiency.

Figure 4.

GCN2 is required to maintain circadian rhythms in whole body metabolism under LeuD feeding. A: visual representation of the experimental design. B: hourly energy expenditure measures normalized to total body weight in control and LeuD-fed GCN2 KO mice over the study period. C: hourly energy expenditure over the 8-day total darkness period separated into two 12-h time periods according to the time of day the measurements were taken. D: hourly wheel-running activity over the study period. E: hourly wheel-running activity over the 8-day total darkness period separated into two 12-h time periods according to the time of day the measurements were taken. F: hourly food intake normalized to total body weight over the study period. G: hourly food intake over the 8-day total darkness period separated into two 12-h time periods according to the time of day the measurements were taken. The yellow backdrop represents the time at which the lights were on in the CLAMS. Black dotted line represents the time which half the mice were switched to a leucine-devoid diet. Hourly CLAMS data are expressed as the mean values. Bar graph values are presented as means ± SE. Bar graphs were analyzed using a three-factor ANOVA with *#P < 0.05 signifying a diet:time interaction. Groups not sharing a common letter indicate P < 0.05 determined by a Tukey’s post hoc test (n = 5 or 6 mice per group). Figure panel A created with BioRender.com. CLAMS, comprehensive laboratory animal monitoring system; CT, circadian time; DD, total darkness; KO, knockout; LeuD, leucine-devoid diet.

Figure 5.

GCN2 is required to maintain circadian wheel-running behavior under LeuD feeding. A: average Actogram for wheel-running activity in GCN2 KO control-fed mice over the entire study duration. B: average Actogram for wheel-running activity in GCN2 KO LeuD-fed mice over the entire study duration. The original light-dark (LD) schedule is denoted by the yellow and black overhead bar. The mice were placed in total darkness (DD) beginning on line 4. The time at which mice were switched from the control to LeuD diet or given fresh control diet is denoted by the red and blue arrows, respectively. C: periodogram analysis of the average wheel-running activity under total darkness in GCN2 KO control and LeuD-fed mice. Period length is denoted by the vertical black line. D: period length of wheel-running activity in control and LeuD-fed GCN2 KO mice under total darkness. E: period amplitude of wheel-running activity in control and LeuD-fed GCN2 KO mice under total darkness. Bar chart values are presented as means ± SE with individual datapoints overlayed. Periodogram analysis outputs were analyzed using a two-factor ANOVA using genotype and diet as variables. Groups not sharing a common letter indicate P < 0.05 determined by a Tukey’s post hoc test (n = 6 mice per group). KO, knockout; LeuD, leucine-devoid diet.

GCN2 Is Required to Maintain Homeostatic Patterns in Lipid Metabolism

We then examined the contribution of GCN2 under both control and LeuD feeding on diurnal patterns in hepatic ISR activation. Regardless of diet, livers from GCN2 KO mice failed to show diurnal changes in eIF2 phosphorylation (Fig. 6A) even though circulating and liver intracellular amino acid concentrations fluctuated similarly to WT mice (Supplemental Tables S6–S9). RNA sequencing conducted on the livers of GCN2 KO mice fed either a control or LeuD diet and killed at CT3 and CT15 revealed that only 2.4% and 5.6% of total transcripts altered their expression between the CT3 and CT15 timepoints in GCN2 KO mice fed either a control or LeuD diet, respectively (Fig. 6B), reflecting a substantial dampening in diurnal changes in the hepatic transcriptome compared with WT mice. Genes that were differentially expressed between CT3 and CT15 in control-fed WT were compared with the list of genes that were differentially expressed between CT3 and CT15 in control-fed GCN2 KO mice to identify unique and shared transcripts. The list of genes that were differentially expressed between CT3 and CT15 in control-fed WT mice but not control-fed GCN2 KO mice was analyzed using Gene Ontology (GO) analysis. The pathways found to be the most impacted by GCN2 status were involved in alcohol, sterol, lipid, and fatty acid biosynthesis. Upon closer inspection, most of the GO terms included the genes encoding Acyl-CoA thioesterases, which function to regulate hepatic triglyceride content over feeding-fasting cycles (43). Indeed, a diurnal pattern in the expression of these genes was impeded in GCN2 KO livers (Fig. 6C). In alignment with these dampened patterns, GCN2 KO mice displayed impeded diurnal changes in hepatic triglyceride content regardless of diet (Fig. 6D). Overall, these data indicate that during a circadian day, diurnal activation of GCN2 guides changes in the liver transcriptome and triglyceride content.

Figure 6.

GCN2 is required to maintain homeostatic patterns in lipid metabolism. A: representative immunoblots and quantified ratios of phosphorylated eIF2α to its respective total forms in control and LeuD-fed GCN2 KO mice killed at CT3 and CT15. Western blot data are expressed relative to control-fed WT mice killed at CT3. B: heat map of hepatic transcripts within WT and GCN2KO mice fed a control or LeuD diet and killed at CT3 or CT15. C: heatmap of transcripts that are differentially expressed in control-fed WT mice between CT3 and CT15 and compose the GO term “Fatty Acid Metabolic Process.” Each row and column represents an individual gene and mouse, respectively. Bar colors represent the Z-score of the normalized read counts for each row. D: hepatic triglyceride levels in control and LeuD-fed GCN2 KO mice killed at CT3 or CT15. Bar graph values are presented as means ± SE with individual datapoints overlayed. Western blot and liver triglyceride data were analyzed using a three-factor ANOVA using time genotype and diet as variables with *P < 0.05 signifying a main effect of diet, $P < 0.05 signifying a main effect of genotype and #$P < 0.05 signifying a genotype:time interaction (n = 3–6 mice per group). CT, circadian time; KO, knockout; LeuD, leucine-devoid diet; WT, wild type.

DISCUSSION

Our study provides novel insights into the multifaceted role of GCN2 in metabolic regulation and diurnal rhythms. Our investigation reveals that GCN2 acts as a crucial sensor of dietary BCAA intake, orchestrating metabolic responses to fluctuations in amino acid availability. Specifically, our analysis of diurnal patterns in liver ISR activation and gene expression demonstrates the dynamic responses of GCN2 to changes in amino acid levels, particularly leucine. This highlights the intricate interplay between nutrient-sensing mechanisms and circadian rhythms, with GCN2 serving as a primary mediator in coordinating metabolic adaptations to both dietary protein quality and availability. Furthermore, our study elucidates the essential role of GCN2 in maintaining diurnal patterns in liver lipid metabolism. We observed significant disruptions in diurnal changes of genes involved in triglyceride biosynthesis and fatty acid metabolism in GCN2 KO mice as well as liver triglyceride concentrations. This underscores the critical involvement of GCN2 in regulating the temporal dynamics of hepatic lipid homeostasis. In addition, our results highlight the indispensable contribution of GCN2 to sustaining circadian rhythms under stress conditions, as evidenced by disruptions in both energy expenditure and feeding behavior in GCN2 KO mice subjected to dietary amino acid insufficiency. Together, these results deepen our understanding of the intricate molecular mechanisms governing metabolic homeostasis and circadian physiology, with implications for therapeutic interventions targeting metabolic disorders and circadian dysfunction.

Over the past 2 decades, a significant body of research has emerged, shedding light on the influence of dietary food intake on entraining circadian rhythms within peripheral tissues, notably the liver (4). However, there remains a gap in our understanding regarding the signaling pathways responsible for transmitting changes in nutrient abundance to the molecular clock. These data demonstrate that GCN2-mediated ISR activation serves as one of these signaling mechanisms relaying changes in dietary amino acid intake to the liver clock. In addition to the ISR, changes in amino acid intake are also sensed by or relayed to AMP-activated kinase (AMPK) and mammalian target of rapamycin complex 1 (mTORC1), both of which have strong connections to the circadian clock (44–46). In addition to GCN2, there is evidence that AMPK, an important cellular metabolite-sensing enzyme, can sense changes in cysteine content (47) as well as changes in dietary protein intake within the liver (48, 49). Furthermore, amino acid insufficiency suppresses mTORC1 signaling via GCN2 activation (29, 50–52). The literature collectively establishes a complex landscape wherein alterations in amino acid levels are presumably communicated to the circadian clock through the interplay of three networks, with GCN2 and AMPK signaling being most active during the fasting cycle and mTORC1 signaling reaching its peak immediately upon feeding. Hence, although our data reveal significant alterations in diurnal changes within the hepatic transcriptome due to the genetic loss of GCN2, further investigation is necessary to fully understand the intricate interactions among these signaling pathways throughout the circadian cycle and their role in conveying changes in nutrient abundance to the clock.

This work reveals that similar to previous results in the SCN, ISR activation within the liver also follows a diurnal pattern that is dependent on GCN2 activity (24). These diurnal changes in GCN2 activity are independent of light and instead likely guided by changes in intracellular BCAA amino acid levels due to internal feeding rhythms. The BCAAs are essential amino acids, but unlike other essential amino acids, the liver does not utilize BCAAs in first-pass metabolism. Therefore, given our findings, BCAAs function in the liver as a proxy for food quantity and protein quality to drive time-of-day changes in ISR activity. A collection of observations across model systems and experimental designs point to GCN2 as a critical regulator of circadian rhythms. Earlier work reported that in mice, GCN2 modulates the period and rhythmicity of the circadian clock of the SCN via the rhythmic phosphorylation of eIF2α and subsequent increase in the transcription of Per2 via ATF4 (24). Similarly, circadian changes in charged tRNA species abundance are sensed in Neurospora via the activation of Cpc3, the homolog of GCN2 in yeast and mammals, driving cyclical changes in eIF2α phosphorylation and rhythmic translation initiation (23). In alignment with these reports, GCN2 appears to function in the liver as a conserved zeitgeber sensor, relaying changes in intracellular amino acids over the course of a feeding-fasting cycle. To expand on these findings, we tested if manipulating ISR activation through dietary intervention could also alter circadian rhythms in whole body metabolism. In agreement with previous studies, energy expenditure was slightly elevated by LeuD feeding (53, 54). However, the period length was sustained. These conclusions fall in line with the current understanding of circadian biology. When studying the impact of specialized diets on circadian behavior and clock outputs from the SCN, researchers continue to demonstrate the resistance of the central clock to changes in nutrient status (4, 6, 8, 10, 55).

In addition to its role in the central clock, our findings highlight a physiological function for GCN2 in preserving diurnal patterns within the liver and establish that GCN2 is required for diurnal changes in hepatic ISR signaling. Furthermore, our work also begins to uncover the physiological function of these diurnal changes within the liver. The transcriptional control of hepatic gene expression is essential for the maintenance of metabolic health. It is believed that ∼16% of protein coding genes show some degree of circadian oscillation within the liver (56). Under control feeding, not only do GCN2 KO mice display a ∼60% reduction in the total number of deferentially expressed transcripts between our two timepoints, but the gene transcripts themselves are largely unique. Although not well characterized, the relationship between GCN2 and hepatic lipid metabolism has been previously shown. Under other models of dietary leucine deprivation, researchers have found that GCN2 is required for the repression of sterol regulatory element binding protein one (SREBP-1), a transcription factor and regulator of lipid metabolism, which is strongly influenced by the circadian clock and nutrient status within the liver (30, 57). Given that GCN2 phosphorylation status is tightly correlated with changes in intracellular leucine content, further work is necessary to identify if GCN2 is required for the circadian regulation of SREBP-1, which may be contributing to this physiological outcome.

Prior research in mice demonstrated that the absence of Gcn2 disrupts the central clock, leading to prolonged periods and diminished rhythmicity in wheel-running activity (24). Consistent with this, GCN2 KO mice fed the control diet failed to maintain energy expenditure during the fasting period. Furthermore, under LeuD feeding, GCN2 KO mice failed to maintain the period amplitude in energy expenditure and wheel-running behavior, which is opposite of WT mice that showed increased EE and wheel-running during LeuD feeding despite similar reductions in food intake and body weight. Moreover, the reductions in EE, RER, and physical activity in LeuD-fed GCN2 KO mice manifested within the first 2 days of feeding, so they cannot be the result of chronic health decline. These findings suggest that GCN2 functions to ensure the robustness of circadian metabolic rhythms under amino acid insufficiency and are supported by the literature using Neurospora models showing through genome-wide transcriptional analysis that the GCN2 signaling pathway maintains robust rhythmic expression of metabolic genes under amino acid starvation (53). Taken together, these data demonstrate that loss of GCN2 alters circadian rhythms in whole body energy expenditure, which is exacerbated under amino acid stress.

Our findings offer novel insights into the multifaceted role of GCN2 in diurnal patterns in the liver and circadian rhythms in metabolism. Although these findings highlight the global importance of GCN2, future work using cell type-specific knockouts should be considered to discern the contribution of GCN2 signaling in each cell type. Furthermore, although we observe the impact of these variables on our liver measures, our data are constrained to two timepoints and a single sex, preventing us from concluding whether the rhythms are blunted or shifted over a full 24-h cycle and if these relationships can also be found in females. Previous research has demonstrated that in both male and female mice, Gcn2 deletion blunts circadian changes in PER2 protein levels, whereas manipulation of ISR signaling via pharmacological inhibition or activation lengthens or shortens the free-running period of the circadian clock, respectively (24). Although we can leverage this information to guide the interpretation of our data, further work is needed to characterize how these biological rhythms are altered using additional timepoints and including female mice. In addition, given the strong and consistent association observed in our data between alterations in diurnal GCN2 signaling patterns and diurnal changes in hepatic lipid metabolism, further studies are warranted to elucidate the underlying mechanisms driving this connection.

DATA AVAILABILITY

RNA sequencing data are deposited in GEO, Accession No. GSE262319.

SUPPLEMENTAL MATERIAL

Supplemental Figs. S1–S5 and Supplemental Tables S1–S9: https://doi.org/10.7282/00000411.

GRANTS

This work was supported by grants from the National Institutes of Health DK109714 (to T.G.A., R.C.W.), R35GM136331 (to R.C.W.), and NS118026 and GM143260 (to R.C.).

DISCLOSURES

R.C.W. is a member of the advisory board of HiberCell. T.G.A. has consulted for HiberCell. All other authors report no conflicts of interest.

AUTHOR CONTRIBUTIONS

J.L.L., R.C. and T.G.A. conceived and designed research; J.L.L., E.T.M., E.M.R., M.J.T., B.A.Z., and T.A.R. performed experiments; J.L.L. analyzed data; J.L.L. interpreted results of experiments; J.L.L. prepared figures; J.L.L. drafted manuscript; J.L.L., M.J.T., B.A.Z., T.A.R., R.C.W., R.C., and T.G.A. edited and revised manuscript; J.L.L., E.T.M., E.M.R., M.J.T., B.A.Z., T.A.R., R.C.W., R.C., and T.G.A. approved final version of manuscript.