Fasting-induced FOXO4 blunts human CD4+ T helper cell responsiveness

Abstract

Intermittent fasting blunts inflammation in asthma¹ and rheumatoid arthritis², suggesting that fasting may be exploited as an immune-modulatory intervention. However, the mechanisms underpinning the anti-inflammatory effects of fasting are poorly characterized^3-5. Here, we show that fasting in humans is sufficient to blunt CD4⁺ T helper cell responsiveness. RNA sequencing and flow cytometry immunophenotyping of peripheral blood mononuclear cells from volunteers subjected to overnight or 24-h fasting and 3 h of refeeding suggest that fasting blunts CD4⁺ T helper cell activation and differentiation. Transcriptomic analysis reveals that longer fasting has a more robust effect on CD4⁺ T-cell biology. Through bioinformatics analyses, we identify the transcription factor FOXO4 and its canonical target FK506-binding protein 5 (FKBP5) as a potential fasting-responsive regulatory axis. Genetic gain- or loss-of-function of FOXO4 and FKBP5 is sufficient to modulate T_H1 and T_H17 cytokine production. Moreover, we find that fasting-induced or genetic overexpression of FOXO4 and FKBP5 is sufficient to downregulate mammalian target of rapamycin complex 1 signalling and suppress signal transducer and activator of transcription 1/3 activation. Our results identify FOXO4-FKBP5 as a new fasting-induced, signal transducer and activator of transcription-mediated regulatory pathway to blunt human CD4⁺ T helper cell responsiveness.

Trial registration: ClinicalTrials.gov NCT02719899 NCT01143454 NCT00001846.

Extended Data Figure 1. Initial analysis of RNA-seq data acquired from the 3 nutritional-load conditions.

a, Individual points (symbols) and means±SEM (lines) of glucose, insulin and growth hormone levels following 24-hr. fasting and 3-hrs. following a fixed caloric meal. Dot plots represent mean±SEM with value of each subject (n=26–28 subjects). The values represent average of duplicates. Two-sided, paired Student’s t-tests. Sera glucose (n=28), Sera insulin (n=26), Sera growth hormone (n=26). b, Table showing number of DE genes (p<0.05) identified in the indicated comparisons. PBMCs’ RNA from 21 subjects used to generate RNA-seq data. c, Unsupervised principal component analysis (PCA) performed on DE genes (p<0.05) for indicated comparisons (n=21 subjects). d, Combined PCA analysis of the top 1000 DE genes (p<0.05) from all 3 groups. e, Top 10 pathways in which the DE genes (p<0.05) from the indicated comparisons. The q values (p values adjusted for false discovery rate (FDR)) from the enrichment result are represented by negative log10 scale (x axis). The most significant pathways predominantly align with lymphocyte and T cell differentiation and activation comparison with refed state. f-g, Box and whisker plots show range of fold change of the subset of 114 DE genes (p<0.05, RNA-seq analysis) that were either downregulated (f) or upregulated (g) to a greater degree following the 24-hr. fast vs. refeeding than baseline (overnight fast) vs. refeeding. The box and whiskers plots show median and upper/lower quartile of the relative gene expression. The whiskers show Turkey distribution and the outlier levels are shown as individual genes (n=21 subjects). h, Pathway enrichment analysis of 844 DE genes exclusively identified in the 24-hr. fasted versus refed state shown with q values (p value adjusted for FDR) for each pathway represented by negative log 10 scale (x axis). Statistical Source Data of Extended Data Fig. 1

Extended Data Figure 2. Flow cytometry using PBMCs exhibit differential fasting and refeeding cell-surface receptor expression levels.

a, Nutrient-load dependent CD45⁺ PBMC flow cytometry distribution. Schematic representation of cytometric labelling to distinguish cell type distributions at baseline, following 24-hr. fasting and refeeding (n=19 subjects). b, Cytometric plots and gating strategies to measure cell surface markers on T helper cells. c-j, Dot and line plots show cell populations of each subject (Wilcoxon two-sided, paired analysis to compare groups, n=19 subjects). c, Flow plots of activated classical monocytes from representative subject comparing 3 nutrient conditions. The plots show relative cell population frequencies of classical monocytes (CD14^highCD16⁻) and median fluorescence intensity (MFI) of activated classical monocytes (HLADR⁺ in CD14^highCD16⁻). d, Representative flow plots showing gating and quantitation of activated CD8⁺ T cells. Plots show expression of activated CD8⁺ T cells (CD38⁺HLADR⁺ in CD8⁺) showing significant increases in refed samples compared to baseline and fasting. e, Dot and line plots showing significant blunting in refed samples compared to baseline and fasting in activated DCs (HLADR⁺ in CD16⁻CD56⁻), myeloid DCs (CD11c⁺CD123⁻) and plasmacytoid cells (CD11c⁻CD123⁺). f, Representative flow plots showing gating and quantitation of regulatory T cells and plots show no difference in Treg cells (FOXP3⁺). g, Follicular helper T cell (CD4⁺CXCR5⁺) levels show no difference in three caloricload conditions. h, Plots showing no difference in activated NK cells (HLADR⁺ in CD16⁻CD56⁺, CD16⁺CD56⁺, and CD16⁺CD56⁻). i, Plots showing no difference in immature B cells (CD19⁺CD20⁻) and mature B cells (CD19⁺CD20⁺). j, Quantifying relative cell population frequencies with specific CD8⁺ T (Tc) markers. Cell populations of Tc1 (CXCR3⁺CCR6⁻) and Tc17 (CXCR3⁻CCR6⁺) cells show no change in three caloric-load conditions. The antibody information of BD lyoplate and 18-color panel and gating strategy of flow cytometry is shown in Supplementary Tables 1 and 2 and Supplementary Data3.

Extended Data Figure 3. Weighted Gene Co-expression Network Analysis (WGCNA) identifies distinct and coordinate gene expression patterns in the fasting and refed states.

a, Modules of genes with correlated expression patterns were clustered using WGCNA. The modules, with their distinct color designations using the fasted and refed data are shown on the x and y axes. Numerical assignment aligned to module colors represent number of genes per module. The correlated modules from fasting and refed were aligned to determine gene overlap (significance of the overlap determined by Fisher’s exact test where red shading denotes significance - darker shade > significance). b, Representative cell type enrichment analysis (CTen) result of a fasting cluster showing enrichment of CD8⁺ and CD4⁺ T cells encoding genes (p-value depicted by red geometric plot extending from the center of the figure towards enriched cell types – representing −log10 Benjamini-Hochberg adjusted P. c, GeneMANIA derived protein-protein interaction (PPI) networks. Significant DE genes fold change information where blue circles represent increased, and red circles decreased expression during fasting compared to refeeding. d, qRT-PCR validation of selected network genes. Bar graph represent mean±SEM with value of each subject (n=14 subjects following 3–4 replicates using two-sided paired Student’s t-test). e, Relative FOXO4 RNA expression in CD4⁺ T cells in response to siRNA (Bar graph represent mean±SEM, n=6 health volunteers, two-sided, paired Student’s t-test). f, Representative protein blot show FOXO4 expression in FOXO4-siRNA treated CD4⁺ T cells from healthy subjects (n=3), 3 days following TCR activation (+). g-h, Representative immunoblots showing expression levels of (g) FOXO4- and (h) FKBP5-overexpression in CD4⁺ T cells from healthy subjects (n=3). Open arrowheads - endogenous protein bands, overexpressed tagged-proteins - solid arrowheads. Source Data Extended Data Fig.3: Unprocessed immunoblots.

Extended Data Figure 4. Fasting and refed modules identified by WGCNA analysis.

*CTen, Cell type enrichment; †PPI, Protein-protein interactions; §TF, Transcription factors

Extended Data Figure 5. Evaluating FKBP5 effect on T cell activation.

a, Top 5 variable importance in prediction (VIP) genes by partial least squares discrimination analysis (PLS-DA) of RNA-seq data from three caloric-load conditions. b, Representative protein blots and quantitative changes normalized by Actin of canonical TFs of Th1 (TBX21), Th2 (GATA3), and Th17 (RORC) in FKBP5-overexpression (OE) in CD4⁺ T cells isolated from healthy volunteers, 3 days following TCR activation. Bar graph represent mean±SEM with data point of each health volunteer (n=8 subjects, Wilcoxon two-sided, paired analysis). c-d, CD4⁺ T cells were isolated from healthy volunteers and transfected with FKBP5 siRNA and scrambled controls. c, Relative FKBP5 RNA expression in CD4⁺ T cells (Bar graph represent mean±SEM with normalized value to scrambled control, n=6 subjects, values represent average of quadruplicates, two-sided, paired Student’s t-test). d, Cytokine release of IFNγ, IL-4 and IL-17 following FKBP5 knockdown (KD) in CD4⁺ T cells isolated from healthy volunteers, 3 days following TCR activation. The dot and line graphs represent mean value of each subject (n=6 biologically independent subjects, values represent average of duplicates, two-sided, ratio paired Student’s t-tests). Source Data Extended Data Fig. 5: Unprocessed immunoblots.

FIGURE 1.. RNAseq Analysis Shows Distinct Separation of the 3 Caloric-load Conditions.

a, Schematic of the protocol showing intervals between fixed caloric meals and temporal research blood draws. The clinical protocol was established to perform immune cell profiling in 28 healthy human subjects. b, Volcano plots of all the genes in the indicated comparison is shown with DE genes. (p value threshold <0.001, colored as blue dots (Source Data: Statistical Source Data Fig. 1)). RNA sequencing was performed on PBMCs from 21 subjects following an overnight (baseline) and following a 24-hr. fast and 3 hrs. after refeeding. c, Venn diagram shows the number of overlapping versus distinct DE genes from baseline and 24-hr. fasting to refed comparisons. d, Top 10 pathways (q value<0.05, which depict p values adjusted for false discovery rates) from the pathway enrichment analysis of the 846 DE genes overlapping in the two comparisons (see Fig. 1C.). e, T cell specific differential pathway enrichment maps by overlaying the results of 24-hr. fasting vs. refed (green nodes) on top of the baseline vs. refed (red nodes) comparison. Each node represents distinct T cell processes or a pathway, and a single-colored node represents a pathway that was exclusively enriched by DE genes from one of the two comparisons. If common genes are annotated within two biological processes, then an edge connecting the two nodes is shown. The red and the teal-colored nodes represent pathways exclusively regulated by baseline or fasting respectively compared to refeeding. The bicolored nodes are regulated by both baseline and fasting vs. refeeding.

FIGURE 2.. Fasting/Feeding Differentially Regulate CD4⁺ T Cell Activation and Differentiation.

a-c, The dot-line plots show relative cell populations (percentiles, n=19 biologically independent subjects, Wilcoxon two-sided paired analysis). a, Representative flow plots of activated CD4⁺ T cells (CD38⁺HLADR⁺) comparing PBMCs (baseline, 24-hr. fasting, and refed states). b, Comparing relative cell population frequencies with specific CD4⁺ T cell surface markers. The plots show increased Th1 (CXCR3⁺CCR6⁻), Th17 (CXCR3⁻CCR6⁺), and Th2 (CD294⁺) surface markers on refed PBMCs. c, Comparison of intracellular cytokine markers. The plots show increased (percentiles) of IFNγ⁺ and IL-17⁺ in refed cells (n=19 biologically independent subjects). d-e, ELISA measurement of cytokine release after CD4⁺ T cell activation. Bar graphs (mean ± SEM with value of each subject, n=13, duplicate experiments, two-sided, Wilcoxon paired analysis after normalization). d, Release of IFN, IL-17, IL-22, e, IL-5 and IL-13 from baseline, 24-hr. fasted and refed CD4⁺ T cells. f, Following Th-subpopulation differentiation IFNγ, IL-5 and IL-17 secretion were measured from Th1, Th2 cells and Th17 cells respectively. Bar graphs (mean ± SEM with data points for each subject, n=5 duplicate experiments, two-sided, ratio paired Student’s t-test). g, Relative mRNA levels of Th1 (TBX21), Th2 (GATA3) and Th17 (RORC) canonical TFs in fasted vs. refed CD4⁺ T cells. Bar graphs represent mean ± SEM with each value 3–4 replicates (n=10 biological distinct subjects, two-sided paired Student’s t-test). Supplementary Data1for Fig.2 shows flow cytometry gating strategy.

FIGURE 3.. Fasting-induced FOXO4 Effects.

a, FOXO4 transcription network and its targets identified by GeneMANIA analysis. The 24-hr. fasting DE genes (circular nodes) show fold increased (blue circles) or decreased (red circles) expression relative to refeeding levels. Validation of target relative genes transcript levels by qRT-PCR in PBMCs (bar plots mean ± SEM with mean values of 3–4 replicates from n=14 subjects, two-sided, paired Student’s t-test). b, Concurrent analysis using the Find Individual Motif Occurrences (FIMO) tool identified the FOXO4 binding motif in the promoter regions of a significant proportion of fasted DE upregulated genes (fisher exact test for FOXO4 motif enrichment significance, p=0.049). c, Representative protein blots from 3 subjects (left panel) and quantitative changes (right panel) of FOXO4 levels in CD4⁺ T cells from the 3 caloric-load states. (Bar graph mean ± SEM, n=7 subjects, two-sided, paired Student’s t-test). d-e, Cytokine release of IFNγ, IL-4 and IL-17 following FOXO4 knockdown (KD) and overexpression in activated CD4⁺ T cells. The cytokine release of KD cells measured (n=5 separate experiment (d)) or from FOXO4-overexpressed cells (n=10 separate experiments (e)) using two-sided, paired Student’s t-test analysis. The dot and line plots show the mean value of each subject and the values represent the average of duplicate experiments. Statistical analysis using two-sided, paired Wilcoxon test. f-g, Representative protein blot (left panel) and quantification (right panel) comparing transduction of control vs. FOXO4 lentivirus (OE) with (+) or without (−) CD4⁺ T cell activation (two-sided, paired Student’s t-test). f, mTOR signaling pathway phosphorylation and total protein kinase levels (Bar graphs - mean ± SEM of n=6 biologically independent subjects). g, STAT signaling proteins (Bar graphs - mean ± SEM of 6–8 biologically independent subjects). h-I, Flow cytometric plots of cell surface; Th1 (CXCR3⁺CCR6⁻), Th17 (CXCR3⁻CCR6⁺), and Th2 (CD294⁺) and intracellular; IFNγ, IL-17, and IL-4, markers from healthy volunteer CD4⁺ T cells transduced with control or FOXO4 lentivirus (Bar graphs - mean ± SEM from n=8 subjects, two-sided, paired Wilcoxon test). Supplementary Data2 for Fig.3 shows flow cytometry gating strategy. Source Data Fig.3 Unprocessed immunoblots.

FIGURE 4.. FKBP5 Induction Mimics FOXO4 overexpression and Fasting effects on Th Cell Activation.

a, RT-PCR analysis of FOXO4 target genes following FOXO4 transduction in CD4⁺ T cells (Bar graphs - mean ± SEM from duplicate experiments in n=6 biologically independent subjects, two-sided, paired Student’s t-test). b, Representative protein blot of FKBP5 (top panel) and quantification (bottom panel) comparing transduction of control vs. FOXO4 lentivirus (OE) with (+) and without (−) CD4⁺ T cell activation (Bar graphs - mean ± SEM of n=6 subjects, two-sided, paired Student’s t-test). c, Cytokine release of IFNγ, IL-4 and IL-17 following FKBP5-overexpression in activated CD4⁺ T cells from n=9 subjects). The dot and line plots show the mean value of each subject. The values represent the average of duplicate experiments. Statistical analysis using two-sided, paired Wilcoxon test. d-e, Representative protein blot (left panel) and quantification (right panel) comparing transduction of CD4⁺ T cells with control vs. FKBP5 lentivirus (OE) with (+) or without (−) TCR activation (Bar graph - mean ± SEM value of subjects. d, mTOR signaling pathway phosphorylated and total protein kinase levels are shown (n=6 biologically independent experiments, two-sided, paired Student’s t-test compared to vector controls). e, STAT signaling proteins (n=8–9 biologically independent experiments, two-sided, paired Student’s t-test compared to vector controls). pSTAT1/STAT1 (n=8); pSTAT3/STAT3 (n=9); pSTAT6/STAT6 (n=9). f, Representative protein blot (top panel) and quantitative changes (bottom panel) in mouse CD4⁺ T cells obtained from fed- or 48-hr. fasted-mice. Protein level of FOXO4 and FKBP5 in mouse CD4⁺ T cells after 24-hr. TCR activation (Bar graph - mean ± SEM of n=11 mice per group, two-sided, unpaired Student’s t-test). g, The cytokine release in fed or fasted mouse CD4⁺ T cells. The histogram shows mean ± SEM levels from n=11 mice from duplicate assays (two-sided, unpaired Student’s t-test). Source Data Fig.4 Unprocessed immunoblots.