Physiologic Responses to Dietary Sulfur Amino Acid Restriction in Mice Are Influenced by Atf4 Status and Biological Sex

ABSTRACT

Background

Dietary sulfur amino acid restriction (SAAR) improves body composition and metabolic health across several model organisms in part through induction of the integrated stress response (ISR).

Objective

We investigate the hypothesis that activating transcription factor 4 (ATF4) acts as a converging point in the ISR during SAAR.

Methods

Using liver-specific or global gene ablation strategies, in both female and male mice, we address the role of ATF4 during dietary SAAR.

Results

We show that ATF4 is dispensable in the chronic induction of the hepatokine fibroblast growth factor 21 while being essential for the sustained production of endogenous hydrogen sulfide. We also affirm that biological sex, independent of ATF4 status, is a determinant of the response to dietary SAAR.

Conclusions

Our results suggest that auxiliary components of the ISR, which are independent of ATF4, are critical for SAAR-mediated improvements in metabolic health in mice.

See corresponding commentary on page 749.

Introduction

Dietary limitation of methionine and cysteine, referred to as sulfur amino acid restriction (SAAR; also called methionine restriction), improves metabolic health and prolongs the life span of several model organisms (1–3). Dietary SAAR promotes phenotypical changes in metabolism such as attenuated body weight gain, reduced adiposity, improved insulin sensitivity, and altered redox capacity (4–6). Underlying mechanisms controlling these responses include enhanced sympathetic signaling (7, 8), increased mitochondrial uncoupling (9), and the sensing of amino acid deficiency (10, 11).

The evolutionarily conserved integrated stress response (ISR) (12) functions as a sentinel “first responder” to various types of environmental stressors, including amino acid insufficiency (13–15). The ISR features a family of 4 related protein kinases, each activated by different stress signals to phosphorylate the α subunit of eukaryotic initiation factor 2 (eIF2) at serine 51 (p-S51-eIF2α) (16). Phosphorylation of eIF2 slows bulk translation but also directs preferential translation of select mRNAs that are critical for proteostasis and ISR feedback control (17–22). Among the preferentially translated mRNAs, the basic leucine zipper (bZIP) transcription factor, activating transcription factor 4 (ATF4), is considered a core effector (17, 23) and master regulator of metabolism. Newly synthesized ATF4 heterodimerizes with other bZIP transcription factors and binds within promoter regions of target genes to facilitate metabolite transport, amino acid synthesis, redox and antioxidant metabolism, and cell cycle control (24–26). One of the ISR target genes that directs SAAR-mediated changes in physiology is the hepatokine, fibroblast growth factor 21 (FGF21), an endocrine factor involved in altering feeding behavior and body composition (27, 28).

Our previous studies reveal that dietary SAAR can promote ATF4 activity independent of p-S51-eIF2α, suggesting that additional mechanisms are regulating the transcriptional response (11). Based on this, in combination with findings suggesting that ATF4 was necessary for upregulation of FGF21 and other ISR target genes during dietary SAAR, we hypothesized that genetic loss of Atf4would impair these responses to SAAR, affecting whole-body physiology. To address the importance of the physiologic responses of the ISR and ATF4, animals with liver-specific or whole-body deletions in Atf4 were challenged with dietary SAAR.

Methods

Study descriptions

All animal experiments were approved by the Rutgers University Institutional Animal Care and Use Committee, which comply with the NIH guide for the care and use of laboratory animals (29). Mice were housed in conventional shoebox cages with soft bedding and enrichment and were kept in a temperature-controlled (23°C) and humidity-controlled (40–60%) facility with a 12:12-h light/dark cycle. Unless otherwise specified, all animals were provided unrestricted access to water and commercial diet (product code 5001; LabDiet) or specified experimental diet).

Genetic knockdown of Atf4 in liver was achieved by crossing C57Bl/6J mice expressing Cre-recombinase driven by the albumin promoter (AlbCre) with Atf4 floxed mice (AlbCreAtf4^fl/fl/lsAtf4KO), as previously described (30). AlbCre-negative littermates (Atf4^fl/fl) were used as controls. Whole-body Atf4 knockout (KO) mice (Atf4KO) were purchased from The Jackson Laboratory (Atf4^tm1Tow/J; stock no.: 013072). Whole-body Atf4 knockout mice were initially bred on a C57Bl/6J background but, following poor reproduction success, were crossed into a Swiss Webster background to generate B6J:SW F1 progeny for use in experiments in which littermates with intact Atf4 served as wild-type (WT) controls. Additional information, including the ages of the mice at the start of each study, are specified in the detailed description of animal experiments (Supplemental Methods).

Experimental diets

Four different types of semipurified experimental diets were used in this study (as detailed in Supplemental Table 1). Two of the diets, referred to as regular-fat (Reg-fat) control (Ctrl) (product code 510072; Dyets, Inc.) or Reg-fat SAAR (product code 510071; Dyets, Inc.), contained either sufficient (0.86% by weight) or restricted amounts (0.17%) of methionine (percentage methionine and cysteine is hereinafter reported as by weight, unless otherwise specified), no cysteine, and 8% fat (percentage fat is hereinafter reported as by weight, unless otherwise specified). These diets were used in experiments described as part of Figures 1–5. The other two diets, referred to as High-fat Ctrl (product code A11051306; Research Diets, Inc.) or High-fat SAAR (product code A11051305; Research Diets, Inc.), were analogous to those described but contained 35% fat, and the High-fat SAAR diet contained slightly lower concentrations of methionine compared with the Reg-fat SAAR diet (0.16% compared with 0.17%) (Supplemental Table 1). These diets were used in experiments described as part of Figures 5–8. All experimental diets were isocaloric and isonitrogenous, which was achieved by increasing the glutamic acid content when methionine content was reduced; amino acid content of all diets was in the form of free amino acids. Briefly, based on the observation that mice lacking Atf4 globally appear lighter and leaner (26, 31), we decided to bolster the fat content and energy density in the diets provided to mice lacking Atf4 globally and their control counterparts. These high-fat diets have been used previously by us and others (11, 32). All diets were provided in pelleted form.

Figure 1.

Dietary sulfur amino acid restriction (SAAR) promotes leanness in liver-specific Atf4-knockdown mice in a sex-dimorphic manner. (A) Graphical overview showing activation of the integrated stress response in response to dietary SAAR and hypothesized eukaryotic initiation factor 2 (eIF2) phosphorylation-independent induction of ATF4. (B) Overview of the genetic mouse models and experimental diets (top) and illustrated timeline of study interventions (bottom). A 1-wk period of control (Ctrl) diet habituation (gray dashed line) was followed by a 10-wk experimental period (black solid line). Mice were placed in a comprehensive laboratory animal monitoring system (CLAMS) for 48 h at the start of week 5. (C) Average cumulative food intake (grams per gram body weight) in male (left) and female (right) mice over the complete 10-wk experimental period. (D) Average change in body weight (g) in male (left) and female (right) mice following the 10-wk experimental period. (E) Average change in fat mass (percentage points) in male (left) and female (right) mice following the 10-wk experimental period. (F) Average change in lean mass (percentage points) in male (left) and female (right) mice following the 10-wk experimental period. (G) Average energy expenditure (kcal/h) measured in the CLAMS: males during light photoperiod (far left), males during dark photoperiod (center left), females during light photoperiod (center right), and females during dark photoperiod (far right). Bar chart values are presented as mean ± SEM with individual data points overlaid. n = 5–12/group. *P < 0.05 with a main effect of diet. n.s. indicates no statistically significant differences at α = 0.05, as determined by two-factor ANOVA and ANCOVA using lean mass as a covariate in the case of energy expenditure.

Figure 5.

Dietary sulfur amino acid restriction (SAAR) alters protein synthesis and DNA synthesis in liver independent of dietary fat content. (A) Overview of the experimental groups (left) and study outline (right), designed to study the impact of dietary energy and fat content on dietary SAAR-mediated changes in liver protein synthesis and DNA synthesis. All mice were habituated to a control (Ctrl) diet for 1 wk (gray dashed line). At the start of the experimental (black solid line) period, mice were injected with deuterium oxide (D₂O) intraperitoneally and then allowed free access to D₂O in their drinking water. Blood and tissues were collected from 3 mice per group at each of the indicated days. (B) Protein synthesis rates (k), measured in cytosolic (left), structural/nuclear (center), and mitochondrial (right) fractions in liver. (C) DNA synthesis rates (k) in liver. n = 12–20/group. *P < 0.05 with a main effect of diet (SAAR, independent of fat content), as determined by two-factor ANOVA. Bar chart values are presented as mean ± SEM.

Figure 8.

Global loss of Atf4 disrupts endogenous hydrogen sulfide (H₂S) production during dietary sulfur amino acid restriction (SAAR) in male mice only. (A, B) Hepatic concentrations of endogenous H₂S production capacity in male (A) and female (B) wild-type (WT) and whole-body Atf4 knockout (Atf4KO) mice, fed either a control (Ctrl) diet or subjected to dietary SAAR, displayed as relative to the Ctrl-fed WT group for each biological sex. (C, D) Hepatic Nrf2 mRNA levels in male (C) and female (D) WT and Atf4KO mice, shown as relative to the respective Ctrl-fed WT group. (E, F) Hepatic mRNA expression levels of Cbs, Cth, Slc7a11, and Slc3a2 in male (E) and female (F) WT and Atf4KO mice, displayed as relative to the Ctrl-fed WT group for each biological sex. n = 3–11/group. *P < 0.05 with a main effect of diet. ^#P < 0.05 with a main effect of genotype. n.s. indicates no statistically significant differences at α = 0.05, as determined by two-factor ANOVA. In case of a statistical interaction effect (between diet and genotype), groups not sharing a common letter indicate a statistically significant difference between groups as determined by post hoc analysis using pairwise t tests with Bonferroni correction. Data shown as part of bar charts are presented as mean ± SEM, with individual data points overlaid. Boxplots indicate median values, with upper and lower hinges representing the 25th and 75th percentiles, respectively, and with upper and lower whiskers reaching to the largest and smallest values, respectively, while not extending further than 1.5 times the interquartile range, with individual data points overlaid.

Open-circuit calorimetry

Measurements of energy expenditure and food intake were collected in a comprehensive laboratory animal monitoring system (CLAMS) (OxyMax-CLAMS; Columbus Instruments International). Mice were maintained on their respective experimental diets and had unrestricted access to both food and water while in the CLAMS. Measurements were made over 48 to 72 h, of which the initial 24 h, classified as acclimatization, were excluded from the subsequent data analysis. Analysis of the exported raw data was partially conducted in CalR version 1.2 (33).

Serum amino acid concentrations

Serum samples for determination of amino acid concentrations were analyzed using HPLC, as previously described (34, 35). Briefly, between 30 and 60 μL serum was added to 90 to 180 μL 0.1% formic acid in methanol and vortexed prior to being filtered through a column to remove precipitated protein. From the resulting elute, 20 μL was collected, to which 5 μL of internal standard was added (norvaline and sarcosine), as well as derivatization agents (ortho-phthalaldehyde and fluorenylmethyloxycarbonyl chloride), and was subsequently analyzed using HPLC. Peak identification and quantification were performed in Agilent OpenLab (Agilent Technologies).

Serum FGF21 concentrations

Serum concentrations of FGF21 were measured using sandwich ELISA, according to the manufacturer's instructions (RD291108200R; BioVendor LLC). Samples were assayed in duplicate on a 96-well plate and the absorbance measured spectrophotometrically (SpectraMax M2; Molecular Devices LLC) at 450 nm, with 630 nm as the reference wavelength. The average coefficient of variation for technical duplicates in this analysis was ∼12.7%. Concentrations were derived using the resulting 4-parameter logistic standard curve, constructed using the “dcr” package in R (36, 37).

Hepatic glutathione concentrations

Hepatic concentrations of total glutathione (GSH) were measured using a commercial kit (Bioxytech GSH/GSSG-412; OxisResearch, Aoxre LLC) with minor modifications, as previously described (10, 11). Briefly, 30–50 mg of pulverized, frozen liver was rinsed twice in ice-cold PBS and subsequently centrifuged at 100 × g at 4°C for 10 min, after which the retained pellet was diluted in 7 volumes of chilled 5% (w/v) metaphosphoric acid (diluted in water) and vortexed. Samples were then clarified by centrifugation at 20,800 × g at 4°C for 10 min, after which the supernatant was collected and subjected to the remaining sample preparation steps described in the abovementioned kit protocol prior to measuring the absorbance, in triplicates, at 412 nm spectrophotometrically (SpectraMax M2; Molecular Devices LLC) over a 3-min period with kinetic read settings. Concentrations of measured GSH signal are expressed relative to the respective control groups, as specified in figure legends.

Hepatic hydrogen sulfide production capacity

Endogenous hydrogen sulfide (H₂S) production capacities from liver lysates prepared in nondenaturing passive lysis buffer (Promega) were determined using the lead acetate/lead sulfide method for headspace H₂S detection, as described previously (38, 39). Endogenous H₂S production concentrations, as quantified by the ImageJ (version 1.52p; NIH) (40) integrated density analysis function of the lead sulfide dots, are expressed relative to the respective control groups, as specified in figure legends.

Liver protein and DNA synthesis rates

We used previously described procedures to separate liver into subcellular protein fractions (cytosolic, structural/nuclear, and mitochondrial) and to isolate DNA and prepare plasma (11, 41, 42). For protein synthesis, pulverized samples were homogenized in a bullet blender in mitochondrial isolation buffer (100 mM KCl, 40 mM Tris HCl, 10 mM Tris Base, 5 mM MgCl₂, 1 mM EDTA, 1 mM ATP, pH 7.5) as previously described (43, 44). Following isolation of subcellular fractions, hydrolyzed aliquots were cation exchanged and derivatized. Dried derivatives were resuspended in ethyl acetate and analyzed on a 7890A gas chromatograph coupled to a 5977A mass spectrometer with a DB-5MS GC column (Agilent Technologies). Deuterium enrichment in the total body water pool was used to estimate the alanine precursor enrichment using mass isotopomer distribution analysis (45). Body water enrichment was measured in evaporated plasma samples as previously described (43, 44). Briefly, 125 μL plasma was pipetted onto inverted plastic microcentrifuge caps with a rubber O-ring, sealed to the tube, and heated at 80°C upside down overnight to capture evaporated water. Captured samples were proton exchanged with 10 M NaOH and acetone overnight. Proton-exchanged samples were extracted in hexane into anhydrous sodium sulfate and analyzed on a 7890A gas chromatograph coupled to a 5975C mass spectrometer using a DB-17MS column (Agilent Technologies). Using the changes in fraction new protein over time, we calculated the kinetic parameter k (1/d) using a 1-phase nonlinear curve fit. The fraction new at each time point was calculated by dividing product enrichment by precursor enrichment. A 1-phase association was used to calculate the rate of the rise of fraction new over time to determine the rate of synthesis (k, 1/d) for protein.

For DNA synthesis, total DNA was extracted using a commercial kit (QiAamp DNA mini kit, product 51306; Qiagen) from liver samples and from bone marrow as described previously (43, 46). Briefly, DNA was suspended into 200 μL nuclease-free water and hydrolyzed to free deoxyribonucleic acids by overnight incubation at 37°C with nuclease S1 and potato acid phosphatase. Hydrolysates were reacted with pentafluorobenzyl hydroxylamine and acetic acid, then acetylated with acetic anhydride and 1-methylimidazole. Dichloromethane extracts were dried and resuspended in ethyl acetate. Incorporation of deuterium into purine deoxyribose in the DNA samples was analyzed on a 7890A gas chromatograph coupled to a 5975C mass spectrometer on a DB-17 column (Agilent Technologies) (47). The fraction of new DNA synthesized in liver samples was calculated using a comparison to bone marrow DNA synthesis in the same animal, representing a fully turned-over population of cells (43, 46). Using the changes in fraction new DNA over time, we calculated the kinetic parameter k (1/d) using a 1-phase nonlinear curve fit.

Immunoblot analyses

Frozen liver samples were processed and analyzed by immunoblotting as previously described (35), with minor modifications as detailed in the Supplemental Methods and Supplemental Table 2.

Quantitative real-time PCR

Analyses of hepatic gene expression were conducted as previously described (35), with minor modifications as detailed in the Supplemental Methods and Supplemental Table 3.

Statistical analysis

Statistical tests and graphs were conducted and generated in R (36) through RStudio (48) using the following packages: “tidyverse” (49), “car” (50), “ggpubr” (51), and “rstatix” (52). Statistical comparisons were performed using a 2-factor ANOVA or ANCOVA with genotype and diet as independent variables and, in the case of ANCOVA, lean mass as a covariate, unless otherwise specified. If a significant interaction effect was detected in the ANOVA, post hoc analysis was done by pairwise t tests with Bonferroni correction for multiple comparisons. The assumptions of the ANOVA were tested using the Shapiro–Wilk test as well as the Levene tests for normality and homoscedasticity, respectively. If assumptions were not met, data were log transformed to approach adherence of test assumptions. The significance level (α) was set to α = 0.05. Unless otherwise specified, data are presented as bar charts with means ± SEMs and individual values overlaid as dots.

Results

Dietary SAAR promotes leanness in liver-specific Atf4-knockdown mice in a sex-dimorphic manner

To address the role of hepatic Atf4 in the physiologic responses to dietary SAAR (Figure 1A), we assigned Atf4^fl/fl or lsAtf4KO mice to one of two Reg-fat, zero-cysteine diets that contained either 0.86% methionine (Ctrl) or 0.17% methionine (SAAR) for ∼10 wk (Figure 1B and Supplemental Table 1). One of the first phenotypic hallmarks noted during dietary SAAR is increased food intake, which was observed in all mice fed SAAR regardless of Atf4 status and biological sex (Figure 1C). Another hallmark of SAAR is attenuation of weight gain; this outcome was identified only in males independent of Atf4 (Figure 1D). In contrast, all females showed similar changes in body weight regardless of diet or Atf4 function (Figure 1D). The attenuation of weight gain in males by dietary SAAR coincided with a reduction in fat mass and retention of lean mass, independent of genotype (Figure 1E, F). By comparison, females showed no effects of diet or genotype on fat mass or lean mass (see Figure 1E, F and Supplemental Table 4 for specific information on tissue weights). Finally, dietary SAAR increased energy expenditure during the dark photoperiod in males only, independent of genotype (Figure 1G). Taken together, we observed that hepatic Atf4 is not necessary for dietary SAAR to evoke body composition changes in males.

Loss of hepatic Atf4 alters the ISR in liver

Dietary SAAR increased p-S51-eIF2α independent of Atf4 status in both male and female mice (Figure 2A, B). Hepatic mRNA levels of Atf4 were substantially reduced in both male and female lsAtf4KO mice, confirming loss of Atf4 in liver (Figure 2C, D). Loss of Atf4 prevented induction of asparagine synthetase (Asns), an ATF4 target gene, during dietary SAAR (Figure 2C, D). In contrast, dietary SAAR in lsAtf4KO mice increased hepatic gene expression levels of other ATF4 target genes, namely, Atf3, Atf5, and CCAAT/enhancer binding protein homologous protein (Chop/Ddit3), independent of biological sex (Figure 2C, D). These results indicate that hepatic loss of Atf4 does not block the transcriptional execution of the ISR but rather triggers an auxiliary transcriptional response to dietary SAAR.

Figure 2.

Loss of hepatic Atf4 alters the integrated stress response in liver. (A, B) Average ratios of hepatic phosphorylated eukaryotic initiation factor 2α (eIF2α) (p-S51) to total eIF2α in male (A) and female (B) mice without or with liver-specific knockdown of Atf4 (Atf4^fl/fl and lsAtf4KO, respectively). Mice were fed either a control (Ctrl) diet or subjected to dietary sulfur amino acid restriction (SAAR) for 10-wk. Representative immunoblots are shown above summary bar charts. Ratios are expressed as relative to the respective Ctrl-fed Atf4^fl/fl groups for male and female mice. (C, D) Hepatic mRNA concentrations of Atf3, Atf4, Atf5, Asns, and Chop (Ddit3), relative to the respective Atf4^fl/fl Ctrl-fed group for each specific gene, in male (C) and female (D) mice following the 10-wk experimental period. n = 5–12/group. *P < 0.05 with a main effect of diet. ^#P < 0.05 with a main effect of genotype. n.s. indicates no statistically significant differences at α = 0.05, as determined by two-factor ANOVA. In case of a statistical interaction effect (between diet and genotype), groups not sharing a common letter indicate a statistically significant difference between groups as determined by post hoc analysis using pairwise t tests with Bonferroni correction. Bar chart values are presented as mean ± SEM with individual data points overlaid. Boxplots indicate median values, with upper and lower hinges representing the 25th and 75th percentiles, respectively, and with upper and lower whiskers reaching to the largest and smallest values, respectively, while not extending further than 1.5 times the interquartile range, with individual data points overlaid.

Dietary SAAR increases circulating FGF21 independent of hepatic Atf4

To delineate acute compared with chronic effects of Atf4 loss on FGF21 production during dietary SAAR, we measured serum FGF21 at 12 h and 10 wk in both male and female Atf4^fl/fl and lsAtf4KO mice. We noted that serum FGF21 was elevated by dietary SAAR at 12 h in Atf4^fl/fl but not lsAtf4KO males (Figure 3A). In contrast, both Atf4^fl/fl and lsAtf4KO male mice showed elevations in serum FGF21 at 10 wk, with lsAtf4KO males exhibiting even higher concentrations (Figure 3A). Female mice fed dietary SAAR showed increased serum FGF21 concentrations at both 12 h and 10 wk compared with Ctrl-fed females that were independent of Atf4 status; however, the magnitude of change was less compared with their male counterparts (Figure 3B). Overall, 10 wk of dietary SAAR increased hepatic Fgf21 mRNA expression levels in both biological sexes, independent of Atf4 status (Figure 3C, D).

Figure 3.

Dietary sulfur amino acid restriction (SAAR) increases circulating fibroblast growth factor 21 (FGF21) independent of hepatic Atf4. (A, B) Serum concentrations of FGF21 (ng/mL) at 12 h after introduction of experimental diets (left) or at the end of the 10-wk experimental period (right) in male (A) and female (B) mice without or with liver-specific knockdown of Atf4 (Atf4^fl/fl and lsAtf4KO, respectively), fed either a control (Ctrl) diet or subjected to dietary SAAR for the indicated time. (C, D) Hepatic Fgf21 mRNA expression at the end of the 10-wk experimental period in male (C) and female (D) mice, displayed as relative to the Atf4^fl/fl control-fed group in each biological sex. (E) Serum concentrations of FGF21 (ng/mL) at the end of a 12-h experimental period during which male mice were allowed to consume the respective experimental diets. (F) Hepatic Fgf21 mRNA expression, relative to the Atf4^fl/fl control-fed group, at the end of a 12-h experimental period. n = 5–12/group. *P < 0.05 with a main effect of diet. n.s. indicates no statistically significant differences at α = 0.05, as determined by two-factor ANOVA. In case of a significant interaction effect (between diet and genotype), groups not sharing a common letter indicate a statistically significant difference between groups as determined by post hoc analysis using pairwise t tests with Bonferroni correction. Bar chart values are presented as mean ± SEM with individual data points overlaid. Boxplots indicate median values, with upper and lower hinges representing the 25th and 75th percentiles, respectively, and with upper and lower whiskers reaching to the largest and smallest values, respectively, while not extending further than 1.5 times the interquartile range, with individual data points overlaid.

To further investigate the early induction of FGF21 during SAAR in males, we collected blood and tissues from a second cohort of Atf4^fl/fl and lsAtf4KO mice at 12 h of experimental diet consumption. The circulating FGF21 concentrations in lsAtf4KO males fed dietary SAAR were reduced by half compared with their littermates expressing hepatic Atf4 (Figure 3E). In addition, the mean increase in hepatic Fgf21 mRNA expression levels in lsAtf4KO males fed SAAR was also diminished relative to Atf4^fl/fl mice, with mean mRNA expression levels ∼35% of Atf4^fl/fl males fed SAAR (Figure 3F). We conclude that loss of hepatic Atf4 delays early production of FGF21 during SAAR but ultimately is not necessary for sustained production in male mice and that male mice show greater increases in FGF21 to dietary SAAR compared with female mice.

Dietary SAAR alters antioxidant response in liver

We next turned our attention to use of methionine and cysteine in redox metabolism via endogenous production of GSH and H₂S given that key enzymatic steps in the synthesis of these products are ATF4 target genes (Figure 4A). Measurements of circulating amino acids showed reduced serum methionine in both male and female SAAR-fed mice (Figure 4B, C), whereas serum cystine was reduced in only male mice fed SAAR (see Supplemental Table 5 for serum amino acid profiles). In agreement with our previous publication (11), total hepatic GSH in male and female mice was reduced by dietary SAAR compared with control-fed counterparts (Figure 4D, E). Conversely, hepatic production of another product of the transsulfuration pathway, H₂S, was increased by dietary SAAR in males, regardless of Atf4 status (Figure 4F, G). In contrast to male lsAtf4KO mice, female lsAtf4KO mice showed no discernable pattern in H₂S production capacity in liver (Figure 4G). These differences between the biological sexes were substantiated by measuring relative mRNA levels of enzymes that are involved in cytosolic H₂S production: cystathionine-β-synthase (Cbs) and cystathionine-γ-lyase (Cth). Even though Cth is reported to, in part, be an ATF4 target gene (53), we noted that dietary SAAR increased Cth in males independent of Atf4 status (Figure 4H). We also examined the glutamine/cystine antiporter system x_c⁻, a major contributor to intracellular cysteine concentrations, and observed that genes encoding both the light chain, Slc7a11 (also referred to as xCT), and heavy chain, Slc3a2, were increased in all males fed the SAAR diet (Figure 4H), and lsAtf4KO males fed the SAAR diet also displayed the largest relative increases in Slc3a2. Male lsAtf4KO mice also showed increased hepatic gene expression of another transcriptional regulator of antioxidant defenses, nuclear factor erythroid 2–related factor 2 (Nrf2, also referred to as Nfe2l2), independent of diet (Figure 4H). In contrast, no significant changes in relative levels of Slc7a11, Slc3a2, or Nrf2 mRNA were observed in female mice (Figure 4I). These results indicate that dietary SAAR alters the hepatic antioxidant response specifically in male mice by mechanisms largely independent of Atf4 in hepatocytes.

Figure 4.

Dietary sulfur amino acid restriction (SAAR) alters antioxidant responses in liver. (A) Graphical overview of hepatic production of the antioxidants glutathione (GSH) and hydrogen sulfide (H₂S), either via the transsulfuration pathway, involving cystathionine-β-synthase (CBS) and cystathionine-γ-lyase (CTH), or through the cystine/glutamate antiporter (comprising SLC7A11 and SLC3A2). (B, C) Serum concentrations (μM) of methionine (left) and cystine (right) in male (B) and female (C) mice without or with liver-specific knockdown of Atf4 (Atf4^fl/fl and lsAtf4KO, respectively), at the end of the 10-wk experimental period. (D, E) Hepatic concentrations of total GSH in male (D) and female (E) mice, relative to the Atf4^fl/fl control (Ctrl)-fed group in each biological sex. (F, G) Hepatic concentrations of endogenous H₂S production capacity in male (F) and female (G) mice, displayed as relative to the Atf4^fl/fl Ctrl-fed group in each biological sex. (H, I) Hepatic mRNA expression levels of Cbs, Cth, Slc7a11, Slc3a2, and Nrf2 in male (H) and female (I) mice, displayed as relative to the Atf4^fl/fl Ctrl-fed group for each gene, in each biological sex. n = 5–12/group. *P < 0.05 with a main effect of diet. ^#P < 0.05 with a main effect of genotype. n.s. indicates no differences at α = 0.05, as determined by two-factor ANOVA. In case of a statistical interaction effect (between diet and genotype), groups not sharing a common letter indicate a statistically significant difference between groups as determined by post hoc analysis using pairwise t tests with Bonferroni correction. Bar chart values are presented as mean ± SEM with individual data points overlaid. Boxplots indicate median values, with upper and lower hinges representing the 25th and 75th percentiles, respectively, and with upper and lower whiskers reaching to the largest and smallest values, respectively, while not extending further than 1.5 times the interquartile range, with individual data points overlaid.

Dietary SAAR alters protein synthesis and DNA synthesis in liver independent of dietary fat content

To determine if the diets provided to the Atf4 floxed mice (Atf4^fl/fl and lsAtf4KO) affected protein synthesis differently compared with our previously published findings that restricted methionine to 0.16% on a high-fat background (11), we compared male C57BL/6J mice fed the 4 different diets: Reg-fat Ctrl, Reg-fat SAAR, High-fat Ctrl, and High-fat SAAR (Figure 5A and Supplemental Table 1). Four important conclusions were drawn from this dietary study: first, the different energy densities of the High-fat Ctrl compared with Reg-fat Ctrl diets (Supplemental Table 1; 5.3 kcal/g compared with 3.9 kcal/g, respectively) did not appreciably change the rates of protein synthesis or DNA synthesis (Figure 5B, C). Second, the Reg-fat SAAR and High-fat SAAR diets reduced protein synthesis similarly in both cytosolic and structural/nuclear fractions compared with their respective control diets (Figure 5B). Third, in agreement with our previous report (11), mitochondrial protein synthesis rates were not reduced by dietary SAAR, suggesting that the proteomic integrity of this cellular fraction was prioritized in the hepatic tissues (Figure 5B). Fourth and finally, both SAAR diets reduced liver DNA synthesis rates similarly (Figure 5C). These experiments showed that dietary SAAR in the range of 0.16–0.17% methionine with 0% cysteine elicits similar effects on liver protein synthesis and DNA synthesis regardless of dietary energy density and fat content.

Global loss of Atf4 does not alter metabolic responses to dietary SAAR

We next turned our attention to studying physiologic responses to dietary SAAR in whole-body Atf4KO mice. In agreement with published reports (26, 31), we noted that both male and female Atf4KO mice were lighter and leaner than WT littermates (Supplemental Figure 1). To support Atf4KO growth, all mice were fed the High-fat Ctrl and SAAR diets for 5 wk (Figure 6A) similar to our previous study (11). Male mice fed the High-fat SAAR diet showed increased food intake compared with the High-fat Ctrl diet, independent of genotype (Figure 6B), whereas female Atf4KO mice showed increased food intake compared with WT mice, independent of diet (Figure 6B). Interestingly, both male and female mice, independent of Atf4 function, demonstrated attenuated weight gain by dietary SAAR (Figure 6C), but the magnitude of change in females was less pronounced. Nevertheless, only male mice demonstrated relative increases and decreases in lean and fat mass, respectively, independent of Atf4 genotype (Figure 6D, E). Finally, energy expenditure appeared to be induced during the dark photoperiod in only male mice fed SAAR (Figure 6F).

Figure 6.

Global loss of Atf4 does not alter metabolic responses to dietary sulfur amino acid restriction (SAAR). (A) Overview of the genetic mouse models and experimental diets (top) and study outline (bottom) indicating the length of habituation (gray dashed line) and experimental (black solid line) periods in weeks as well as timing of interventions. Mice were placed in a comprehensive laboratory animal monitoring system (CLAMS) for 24 h during the fourth week of the experimental period. (B) Average cumulative food intake (grams per gram body weight) in male (left) and female (right) wild-type (WT) and Atf4 knockout (Atf4KO) mice over the 5-wk experimental period. (C) Average body weight change (g) in male (left) and female (right) mice over the 5-wk experimental period. (D) Average change in lean mass (percentage points) in male (left) and female (right) mice over the 5-wk experimental period. (E) Average change in fat mass (percentage points) in male (left) and female (right) mice over the 5-wk experimental period. (F) Average energy expenditure (kcal/h) measured in a CLAMS: males during dark photoperiod (left) and females during dark photoperiod (right). n = 3–11/group. *P < 0.05 with a main effect of diet. n.s. indicates no statistically significant differences at α = 0.05, as determined by two-factor ANOVA and ANCOVA using lean mass as covariate, in the case of energy expenditure. Bar chart values are presented as mean ± SEM.

Execution of an auxiliary ISR to dietary SAAR delays but does not block induction of FGF21

Dietary SAAR increased eIF2α phosphorylation in WT liver, but global loss of Atf4 increased p-S51-eIF2α basally, partially masking elevations by dietary SAAR in Atf4KO mice (Figure 7A, B). Hepatic relative levels of Atf4 mRNA were not appreciable in all Atf4KO mice, irrespective of diet (Figure 7C, D). In agreement with the lsAtf4KO cohort, hepatic Asns mRNA expression levels during SAAR were blunted in whole-body Atf4KO mice while mean hepatic mRNA expression levels of Atf3, Atf5, and Chop were highest in Atf4KO mice fed SAAR (Figure 7C, D). By comparison, circulating FGF21 was increased by dietary SAAR in both WT and Atf4KO mice independent of biological sex (Figure 7E, F), similar to hepatic Fgf21 mRNA levels (Supplemental Figure 2). To confirm the time-dependent involvement of ATF4 in the induction of hepatic Fgf21 and serum FGF21 during SAAR, we performed a 12-h study, similar to the liver-specific cohort. Similar in pattern to the Atf4 floxed cohorts, relative hepatic Fgf21 mRNA levels increased in both WT and KO mice fed the SAAR diet (Figure 7G), whereas mean serum FGF21 concentrations in the Atf4KO mice were blunted compared with WT mice fed a SAAR diet (Figure 7H). These data confirm a time-dependent dispensability of Atf4 in the induction of FGF21 during SAAR.

Figure 7.

Execution of an auxiliary integrated stress response to dietary sulfur amino acid restriction (SAAR) delays but does not block induction of fibroblast growth factor 21 (FGF21). (A, B) Relative ratios of phosphorylated (p-S51) eukaryotic initiation factor 2α (eIF2α) to eIF2α (bottom) and representative lanes (top) for male (A) and female (B) wild-type (WT) and Atf4 whole-body knockout (Atf4KO) mice fed either a control (Ctrl) diet or subjeted to dietary SAAR for 5 wk. Results are expressed as relative to WT Ctrl for each biological sex. (C, D) Hepatic mRNA levels of Atf3, Atf4, Atf5, Asns, and Chop in male (C) and female (D) mice at the end of the 5-wk experimental period, relative to WT Ctrl for each biological sex. (E, F) Serum concentrations of FGF21 (ng/mL) at the end of the 5-wk experimental period in male (E) and female (F) WT and Atf4KO mice. (G) Hepatic expression levels of Fgf21 mRNA in male and female WT and Atf4KO mice after a 12-h experimental period, displayed as relative to the control-fed WT group. (H) Serum concentrations of FGF21 (ng/mL) in male and female WT and Atf4KO mice after a 12-h experimental period. n = 2–11/group. *P < 0.05 with a main effect of diet. ^#P < 0.05 with a main effect of genotype. n.s. indicates no statistically significant differences at α = 0.05, as determined by two-factor ANOVA. In case of a statistical interaction effect (between diet and genotype), groups not sharing a common letter indicate a statistically significant difference between groups as determined by post hoc analysis using pairwise t tests with Bonferroni correction. Data shown as part of bar charts are presented as mean ± SEM. Boxplots indicate median values, with upper and lower hinges representing the 25th and 75th percentiles, respectively, and with upper and lower whiskers reaching to the largest and smallest values, respectively, while not extending further than 1.5 times the interquartile range with individual data points overlaid.

Global loss of Atf4 disrupts endogenous hydrogen sulfide production during dietary SAAR in male mice only

To further examine the physiologic relevance of ATF4 activity during dietary SAAR, we measured H₂S production capacity in the livers of both male and female mice and noted that male Atf4KO mice had reduced endogenous H₂S production capacity relative to WT males (Figure 8A). Females did not differ in terms of endogenous H₂S production capacity, independent of either genotype or diet (Figure 8B). In contrast to observations made in the Atf4 floxed mice, we did not detect any differences in hepatic Nrf2 mRNA expression between WT and Atf4KO mice in either males or females (Figure 8C, D). Messenger RNA expression levels of the two cytosolic H₂S-producing enzymes, Cth and Cbs, were similar across diet groups and genotypes in both sexes (Figure 8E, F). In contrast to male WT mice, male Atf4KO mice failed to upregulate hepatic Slc7a11 mRNA expression levels in response to dietary SAAR (Figure 8E), whereas female Atf4KO mice showed smaller increases in Slc7a11 gene expression during SAAR (Figure 8F). Dietary SAAR increased relative hepatic levels of Slc3a2 mRNA in both WT and Atf4KO male mice (Figure 8E) but not in females (Figure 8F). These results suggest that ATF4 activity contributes to hepatic H₂S production in males.

Discussion

Dietary SAAR improves metabolic health (54, 55), promotes protection from surgical stress (56), and extends life span (57, 58) of experimental mice. The role of the ISR in these outcomes is unclear because in a previous collection of experiments, we discovered that hepatic increases in ISR gene targets can occur independent of eIF2α phosphorylation (11). Our in vivo findings corroborated prior in vitro studies showing no p-S51-eIF2α requirement for ATF4 synthesis in response to lowered SAA availability (53, 59, 60). Therefore, in the current study, we addressed the requirement for Atf4 in the hepatic response to dietary SAAR. Our results reveal that many of the physiologic responses to dietary SAAR, including the attenuation of body weight gain and the shift in body composition toward leanness, do not require Atf4. Instead, Atf4 is needed to bolster redox homeostasis, especially via the transsulfuration pathway, leading to sustained endogenous H₂S production, consistent with previous findings (61, 62).

Loss of Atf4 does not cause ISR transcriptional failure during dietary SAAR but alternatively triggers an auxiliary transcriptional landscape that delays but does not block FGF21 production. This was initially surprising because ATF4 is described as an ISR core effector (12) and transcriptional regulator of FGF21 (63). Genetic loss of the eIF2 kinase general control nonderepressible 2 delays ATF4 binding to the Fgf21 promoter, in turn delaying the induction of circulating FGF21 in response to dietary protein restriction (64–66). The limited early dependency on Atf4 to trigger FGF21 production during dietary SAAR leaves unresolved which auxiliary transcription factors are driving hepatic expression of Fgf21. Based on the dimerization preferences within the bZIP family (67, 68), a multitude of auxiliary combinations may be available to compensate for loss of Atf4 during SAAR. This idea fits with the evolutionary complexity and long existence of the bZIP regulatory network (69), which in part functions to protect and provide organismal resilience during nutrient stress. In considering both dimerization specificity as well as tissue-specific expression patterns, ATF5, CHOP, and nuclear factor erythroid 2-related factor 2 (Nrf2) protein are worth special consideration. ATF5 expression is enriched in human and mouse livers (70) and is translationally induced by p-S51-eIF2α (18). ATF5 is furthermore classified as an ATF4 paralog based on sequence similarity, and its binding properties are suggested to be promiscuous (71, 72), leading us to speculate that under certain contexts, it may share some ATF4 gene targets. Another candidate is CHOP, which is also translationally regulated via eIF2α phosphorylation (73). Previous studies from our laboratory and others show that CHOP is increased in livers from Atf4KO mice challenged with amino acid starvation (30, 74) and that both FGF21 (75) and ATF5 (76) are CHOP target genes. Finally, we previously reported activation of a noncanonical PERK-Nrf2 antioxidant pathway by dietary SAAR (10). In certain diabetic mouse models, both chemical and genetic induction of Nrf2 can contribute to upregulation of FGF21 (77). These findings, together with the current data, suggest that Nrf2 may also contribute to FGF21 production in the absence of Atf4 during dietary SAAR.

During dietary SAAR, WT males showed a prominent hepatic ISR signature alongside large changes in metabolism and body composition, whereas females showed moderate to minimal changes in most measured variables. These findings are in agreement with other models of dietary restriction that show metabolic outcomes to be influenced by biological sex (78–81). Furthermore, evidence for androgen receptor involvement in nutrient stress signaling is reported in certain types of cancer (82, 83). There is experimental support for equalization of the sex-dimorphic response to SAAR in mice that have reached physical maturation, but this variable was not directly assessed in these experimental trials (84). In total, our observations support a role for sex hormones in activation of the ISR and warrant further exploration in other models of dietary restriction.

Although Atf4 is dispensable for FGF21-mediated changes in body composition, Atf4 appears partially necessary in whole-body Atf4KO males to fully maintain H₂S production during dietary SAAR, particularly under a high-fat diet background. We interpret these data to be in accordance with earlier reports identifying suppression of H₂S production under high-fat diet (85) and ATF4-mediated expression of Cth and induced angiogenesis via augmented H₂S under SAAR (53, 61, 62). AlbCre-mediated deletion of floxed Atf4 is limited to hepatocytes, whereas whole-body Atf4KO targets all cell populations. Based on this, it is possible that another cell population in liver aside from hepatocytes, such as the stellate cells, which express Cth (86, 87), may produce H₂S during dietary SAAR via ATF4 activity.

The range of dietary SAAR necessary to invoke an improved metabolic phenotype reportedly spans 0.12–0.17% methionine when cysteine is kept to zero. Restriction up to 0.25% methionine is partially effective, whereas reductions below 0.12% methionine cause food aversion and weight loss (88). Our previous study used a 0.16% methionine restriction level on a high-fat background, resulting in reduced fat mass while allowing for retention of lean mass (11). To determine if differing levels of restriction alter proteostasis differently, we compared the 2 diets directly, measuring deuterium oxide incorporation into alanine to calculate the fractional synthesis rate of cytosolic, structural/nuclear, and mitochondrial proteins. Overall, our data confirm that dietary SAAR in the range of 0.16–0.17% methionine plus zero cysteine reduces protein and DNA synthesis in liver similarly. The resistance of the mitochondrial fraction to show reduced protein synthesis may reflect heightened requirements for of mitochondrial proteostasis maintenance, a feature shared with models of slowed aging (89). Whether or not mitochondrial proteostasis is improved by dietary SAAR requires a fuller examination of mitochondrial function and quality control (90). Future work examining the molecular basis for differences in protein synthesis in the different cellular fractions as well as differences in protein turnover are warranted to more fully understand how dietary SAAR affects proteostasis.

Conclusions

In conclusion, we identify a time-dependent dispensability of Atf4 during dietary SAAR and prompt further investigation into the compensatory and auxiliary transcriptional networks that respond to dietary amino acid restriction. We further ascertain that many physiologic responses to dietary SAAR, including ISR activation, are influenced by biological sex, highlighting the potential importance to consider it as a factor when studying and implementing dietary interventions aimed at improving metabolic health.

Notes

Funding: Supported by the following research grants: NIH DK109714 (TGA, RCW), R01DK096311 (TWG), DK105032 (CDM), T32 DK064584 (MAL) and AG050777 and HL148352 (CH). This work was also supported by the USDA National Institute of Food and Agriculture, Hatch Multistate project 1184/NJ14240 (TGA).

Author disclosures: RCW, received grant support from Eli Lilly and Company and is a scientific advisor to HiberCell. All other authors report no conflicts of interest.

Supplemental Figures 1 and 2, Supplemental Tables 1–5, Supplemental Methods, and Supplemental References are available from the “Supplementary data” link in the online posting of the article and from the same link in the online table of contents at https://academic.oup.com/ajcn/.

Abbreviations used: ATF4, activating transcription factor 4; bZIP, basic leucine zipper; CLAMS, comprehensive laboratory animal monitoring system; Ctrl, control diet; eIF2, eukaryotic initiation factor 2; FGF21, fibroblast growth factor 21; GSH, glutathione; H₂S, hydrogen sulfide; High-fat, high-fat diet; ISR, integrated stress response; KO, knockout; Reg-fat, regular-fat diet; SAAR, sulfur amino acid restriction; WT, wild-type.