Abstract

   Time-restricted eating (TRE) has been shown to extent lifespans in
   drosophila and mouse models by affecting metabolic and
   anti-inflammatory activities. However, the effect of TRE on the human
   immune system, especially on immunosenescence, intestinal microbiome,
   and metabolism remains unclear. We conducted a 30-day 16:8 TRE
   single-arm clinical trial with 49 participants. Participants consumed
   daily meals from 9 a.m. to 5 p.m., provided by a nutrition canteen with
   a balanced, calorie-appropriate nutrition, which is designed by
   clinical nutritionists (ChiCTR2200058137). We monitored weight changes
   and weight-related parameters and focused on changes in the frequency
   of CD4^+ senescent T cells, immune repertoire from peripheral blood, as
   well as serum metabolites and gut microbiota. We found that up to
   95.9 % of subjects experienced sustained weight loss after TRE. The
   frequency of circulating senescent CD4^+ T cells was decreased, while
   the frequency of Th1, Treg, Tfh-like, and B cells was increased.
   Regarding the immune repertoire, the proportions of T cell receptor
   alpha and beta chains were increased, whereas B cell receptor kappa and
   lambda chains were reduced. In addition, a reduced class switch
   recombination from immunoglobulin M (IgM) to immunoglobulin A (IgA) was
   observed. TRE upregulated the levels of anti-inflammatory and
   anti-aging serum metabolites named sphingosine-1-phosphate and
   prostaglandin-1. Additionally, several anti-inflammatory bacteria and
   probiotics were increased, such as Akkermansia and Rikenellaceae, and
   the composition of the gut microbiota tended to be “younger”. Overall,
   TRE showed multiple anti-aging effects, which may help humans maintain
   a healthy lifestyle to stay “young”. Clinical Trial Registration URL:
   [59]https://www.chictr.org.cn/showproj.html?proj=159876.

   Keywords: Time-restricted eating, Immune cellular senescence,
   Intestinal microbiome, Immune repertoires, Metabolomic, Young

Graphical abstract

   Image 1
   [60]Open in a new tab
     __________________________________________________________________

   Time-restricted eating (TRE) reduces weight, senescent T cells, and
   alters immune repertoire, promoting a "younger" immune system.

1. Introduction

   Obesity has been identified as one of the most prevalent chronic
   disease worldwide [[61]1]. Multiple fundamental studies have revealed
   that obese people have low-grade, chronic inflammation [[62]2], which
   contributes to the development of insulin resistance [[63]3]. In
   addition, T cells in obesity-related visceral adipose tissue have been
   reported to undergo cellular senescence, potentially leading to
   accelerated immunosenescence [[64]4]. Therefore, except for
   metabolism-related diseases, obesity may contribute to
   immunosenescence-related disorders.

   To fight obesity, low-calorie diets and fasting have been proposed and
   have shown great effects on weight loss [[65]5]. To our surprise,
   low-calorie diets and fasting regimens also relieve the clinical
   phenotypes of autoimmune and inflammatory diseases, including multiple
   sclerosis [[66]5] and rheumatoid arthritis (RA) [[67]6], and even
   extend the lifespan of mice [[68]7]. Time-restricted eating (TRE) is a
   method of fasting that involves restricting daily food intake to a
   period of 6–8 h. TRE has been found to reduce blood glucose, body mass
   index and body fat, alleviate inflammation and increase the diversity
   of the gut microbiome [[69]8]. Some studies have suggested that TRE
   provides benefits by enhancing the process of autophagy [[70]9].
   However, the effect of TRE in human immunosenescence remains unclear.

   Senescence occurs in the aging status, and it also manifests in
   pathological conditions such as breast cancer [[71]10,[72]11], RA and
   acquired immune deficiency syndrome [[73]12]. Cellular senescence
   refers to a state in which a cell exhibits cell cycle arrest, apoptosis
   resistance, cell dysfunction and secretion of senescence-associated
   secretory phenotypes (SASP), such as IL-6 and TNF-alpha.
   Immunosenescence involves inflammation and the accumulation of
   senescent cells [[74]13]. Senescent immune cells, such as senescent T
   cells, play a crucial role in the development of age-related diseases
   [[75]14], including cancer, atherosclerosis and osteoarthritis.
   Although previous studies have reported that fasting on alternate days
   regulates aging-related metabolism in healthy adults [[76]15], it
   remains to be clarified how TRE affects immune system function and the
   immunosenescent state in humans. Therefore, we established a clinical
   trial, Effect of Time-Restricted Eating on the Immune System (TREI), in
   healthy subjects. This trial contributes to our better understanding of
   the benefits of TRE and may provide a candidate regimen for public
   health.

2. Results

2.1. Study design

   This study was a single-arm clinical trial with a TRE group (n = 49)
   and a control group (n = 10). The consort flow diagram is shown in
   [77]Fig. S1. Over a period of 30 consecutive days, the TRE participants
   consumed a balanced diet ([78]Table S1) and were restricted to eating
   between 9 a.m. and 5 p.m. All TRE subjects completed the trial without
   loss to follow-up or poor completion. [79]Fig. 1a illustrates the
   eating window, sampling timeline and specimen collection for the TRE
   group. The weight of TRE participants was recorded before the
   intervention, while their weight was recorded once a week during the
   dietary intervention. Peripheral blood was collected from TRE
   participants before the trial and at Day 0, Day 14 and Day 30 of the
   trial on their fasting state (8:00 a.m.), and samples of stool were
   collected before and after the trial. Day 90 after TRE ended, we
   reweighed participants and collected peripheral blood to observe the
   long-term effects of TRE. For the subjects in the control group, their
   diet structure was similar to that in the TRE group, and there was no
   intervention on food intake. Blood samples were collected and body
   weight and weight-related indicators were measured only on Day 0 and
   Day 30.

Fig. 1.

   [80]Fig. 1
   [81]Open in a new tab

   Experimental design and changes in body weight and related indexes
   before and after TRE. (a) Eating window, sampling timeline and specimen
   collection during TRE; (b) Proportion of participants who experienced
   weight loss after 30 days of TRE; (c) Statistics of weight loss based
   on different percentage ranges; (d) Continued weight loss observed in
   participants during 30-day TRE. (e) Statistical analysis of the change
   of body fat rate, body mass index (BMI), muscle mass, body moisture,
   visceral fat, basal metabolism, fat weight and skeletal muscle rate
   before and after a 30-day TRE. n = 45. Statistical analysis was
   performed using Mauchly's Test of Sphericity in [82]Fig. 1d and paired
   t-test at two different time points in [83]Fig. 1e and f. Data are
   mean ± s.e.m in [84]Fig. 1d.

   In the TRE group, we conducted the analysis of weight and
   weight-related indicators, circulating T and B cell frequencies,
   circulating CD4^+ senescent T cell frequencies, immune repertoire (IR)
   sequencing, liquid chromatography-mass spectrometry (LC-MS) metabolomic
   analysis and 16S rDNA microbiome sequencing. In the control group, data
   were analyzed from IR sequencing to body weight and weight-related
   indicators on Day 0 and 30 ([85]Fig. S1). Besides, new evidence shows
   that immunological aging doesn't align with biological aging, with
   notable age turning points at 34, 60, and 78 [[86]16]. In the Chinese
   population, pronounced changes in aging, metabolism, and immunity occur
   around age 30 [[87]17,[88]18]. Therefore, we divided participants into
   two subgroups: the under 30 years old TRE group (<30 y) and the over 30
   years old TRE group (>30 y), to better understand TRE's impact on
   immunosenescence. In this study, we hypothesized that 16:8 TRE for 30
   days in healthy participants would reduce the proportion of CD4^+
   senescent T cells in peripheral blood.

2.2. TRE reduced body weight and weight-related indicators

   95.9 % of TRE participants lost weight ([89]Fig. 1b). Most of them lost
   2–4% of their body weight ([90]Fig. 1c). Weight losing among
   participants continued steadily throughout the 30-day TRE period
   ([91]Fig. 1d). Meanwhile, changes were also observed in weight-related
   indicators before and after the 30-day TRE. The value of body fat rate,
   body mass index (BMI), muscle mass, visceral fat, basal metabolism and
   fat weight decreased significantly ([92]Fig. 1e). On the other hand,
   the value of body moisture and skeletal muscle rate was increased
   significantly after a 30-day TRE ([93]Fig. 1f).

   According to previous studies, the aging-related and metabolism-related
   changes in Chinese people show a dramatic peak at the age of 30^17.
   Interestingly, similar trends were observed in the age subgroup
   analysis of TRE ([94]Figs. S2a and b). In order to exclude the
   influence of confounding factors such as seasons, live conditions,
   temperature change and diet pattern on body weight and weight-related
   indicators, we also recruited 10 healthy volunteers with similar living
   conditions (from the same research center), diet patterns to TRE group,
   without eating time restriction, and measured body weight and
   weight-related indicators on Day0 and Day30. There was no difference in
   body weight and weight-related indicators in this control group
   ([95]Fig. S2c), indicating that the changes might be attributed to TRE.

2.3. TRE increased the proportion of circulating Th1, Tfh-like and Treg cells
while reduced the proportion of circulating CD8^+ T cells

   The TRE-induced changes in lymphocytes were investigated in 40 healthy
   participants by analyzing circulating T and B cell subsets on Day 0,
   Day 14, Day 30 of TRE and Day 90 after TRE ended. The identified
   markers of each cell subset are listed in [96]Table 2.

Table 2.

   Identify maker of cell subsets for flow cytometry.
      Cell subset         Identify marker
   CD3^+ T           CD3^+
   CD4^+ T           CD4^+
   CD8^+ T           CD8^+
   CD4^+ naïve T     CD45RO^−CD45RA^+CD4^+
   CD4^+ effector T  CD45RO^+CD45RA^−CD4^+
   Th1               CXCR3^+CCR6^−CCR4^−CD4^+
   Th2               CXCR3^−CCR6^−CCR4^+CD4^+
   Th17              CXCR3^−CCR6^+CCR4^+CD4^+
   Tfh-like          CXCR5^+PD-1^+CD4^+
   Treg              CD127^loCD25^hiCD4^+
   CD45RA^+CD8^+ T   CD45RA^+CD8^+
   CD45RA^−CD8^+ T   CD45RA^−CD8^+
   Senescent CD4^+ T CD27^−CD28^−CD4^+
   Total CD19^+ B    CD19^+
   Memory B          CD38^−IgD^-CD27^+CD19^+
   Naive B           CD38^−IgD^+IgM^+CD19^+
   Plasma cell       CD38^+CD20^−CD19^+
   [97]Open in a new tab

   Interestingly, after 30 days of TRE, the frequencies of lymphocytes in
   the participants were changed. The frequency of total CD3^+ T cells was
   decreased ([98]Fig. S3a). The frequency of CD3^+ T cells decreased at
   Day 30 but increased to baseline levels after the end of TRE in the
   <30 y group ([99]Fig. S4a). The frequency of total CD4^+ T cells
   remained stable during TRE ([100]Fig. S3a, Fig. S4a). For the subsets
   of CD4^+ T cells, we found that the frequency of CD4^+ naive T cells
   was decreased while the frequency of CD4^+ effector T cells was higher
   than that on Day 0 ([101]Fig. S3b), while no significant differences
   were observed in the age subgroup analysis ([102]Fig. S4a). We further
   analyzed the effect of TRE on CD4^+ helper T cells. The frequency of
   Th1 cells increased, while Th2 cells and Th17 cells decreased after
   30-day TRE ([103]Fig. 2a). The frequency of Tfh-like cells and Treg
   cells increased on Day 30 and then decreased to baseline levels at Day
   90 after the end of TRE ([104]Fig. 2b). In addition, we found the
   changes in the frequencies of Th1, Th2, Th17, Tfh-like and Treg cells
   in different age group were in accordance with the general trend of the
   ungrouped results ([105]Fig. S4b). The frequency of total CD8^+ T cells
   was decreased ([106]Fig. S3a). For the subsets of CD8^+ T cells, the
   frequency of CD45RA^+CD8^+ T cells was decreased ([107]Fig. S3c) and
   that of CD45RA^−CD8^+ T cells was increased after 30-day TRE ([108]Fig.
   S3c). [109]Fig. S4a shows the subgroup analysis of CD8^+ T lymphocytes
   based on age stratification at 30 years old. There were no significant
   differences in the changing trend of CD8^+ T lymphocyte frequencies
   between the two age subgroups during TRE.

Fig. 2.

   [110]Fig. 2
   [111]Open in a new tab

   Frequencies of CD4^+ T subsets from PBMCs before and after TRE.

   Statistical analysis and cytometry gates of (a) CD4^+Th[1], Th2 and
   Th17 cells; (b) CD4^+Tfh-like and T[reg] cells for each T-cell subset
   from PBMCs of the TRE Day 0 (n = 41), Day 14 (n = 41), Day 30 (n = 40)
   and Day 90 after TRE groups (n = 19). (c) Statistical analysis and
   cytometry gates of CD4^+ senescent cells (CD27^−CD28^−) in the TRE Day
   0 (n = 22), Day 14 (n = 38), Day 30 (n = 37) and Day 90 after TRE
   groups (n = 21). Statistical analysis was performed using paired t-test
   at two different time points in [112]Fig. 2a, b and c. Data are
   mean ± s.e.m.

2.4. TRE reduced the proportion of circulating senescent CD4^+ T cells

   To evaluate the levels of immunosenescence, we further assessed the
   impact of TRE on circulating senescent CD4^+ T (CD4^+CD27^−CD28^−)
   cells. We found that the frequency of senescent cells in total CD4^+ T
   cells dramatically reduced during TRE (Day14, 30) and continuously
   decreased on Day 90 after TRE ending ([113]Fig. 2c). Further, the
   frequency of senescent cells in total CD4^+ T cells in >30y group was
   higher than that in <30y group in baseline, as expected. Interestingly,
   the frequency of CD4^+ senescent T cells was reduced in both the >30y
   group and <30y group, and the downward trend was maintained 90 days
   after TRE in the >30y group ([114]Fig. S4c).

2.5. TRE increased the proportion of circulating total B cells in healthy
individuals

   In addition, we observed that the frequency of total CD19^+ B cells
   increased continuously from Day 0 to Day 30 but dropped to baseline
   level at Day 90 after TRE ([115]Fig. S3d). For the B-cell subsets,
   there were no significant changes in classical memory B cells or naive
   B cells. Meaningfully, the ratio of plasma cells (PCs) remained stable
   during TRE ([116]Fig. S3d).

   There was no difference in the frequency of total CD19^+ and classical
   memory B cells between age subgroups at baseline. However, the
   frequency of naive B cells was higher and plasma cells was lower in the
   >30y group at baseline. After conducting subgroup analysis by age, we
   observed the same change of total CD19^+ cells both in total and the
   two age subgroups ([117]Fig. S5).

2.6. TRE increased the frequency of TCR alpha and beta chains and decreased
BCR-kappa and lambda chains

   T cell receptor (TCR) and B cell receptor (BCR) are essential elements
   of adaptive immunity and constitute IRs [[118]19]. The diversity and
   distribution of immune receptor clones, reflecting clonal selection and
   expansion, can uniquely indicate an individual's immunological state
   (e.g., healthy, infected, vaccinated) and be utilized for
   immunodiagnostic purposes. IR sequencing was conducted before and after
   30-day TRE to evaluate its effects on the immune system on multiple
   dimensions.

   After 30-day TRE, the proportion of 7 chains in volunteer peripheral
   blood mononuclear cells (PBMCs) in the >30y TRE, <30y TRE and control
   groups changed apparently ([119]Fig. 3a). Compared with Day 0, the
   percentage of TCR alpha chains and beta chains increased ([120]Fig. 3b)
   and the percentage of BCR heavy chains among the seven sequenced chains
   declined ([121]Fig. 3b). Moreover, these changes were predominantly
   observed in >30y group ([122]Figs. S6a and b). Notably, there was no
   discernible difference in diversity (represented by d50) between TRE
   and its predecessors ([123]Fig. S6c).

Fig. 3.

   [124]Fig. 3
   [125]Open in a new tab

   Immune repertoire changes before and after 30 days of TRE. (a)
   Proportion of 7 chains in volunteer PBMCs before and after TRE by
   counting unique CDR3; (b) Statistical analysis of the percentage
   changes in the proportions of BCR and TCR chains across time points;
   (c) The representative treemaps showing the 7-chain repertoire in PBMCs
   before and after the 30-day TRE. (d) The proportion of the 5 Ig
   isotypes in >30 y TRE, <30 y TRE and Ctrl groups by counting reads; (e)
   The CSR index values of five Ig before and after the 30-day TRE for the
   >30 y TRE, <30 y TRE groups and in total. The arrow indicates the
   direction of class switching, and the thickness of the arrow connecting
   lines indicates the displacement of the CSR index. The spheres
   represent different Ig isotypes. The graph displays IgM and IgD as a
   whole. > 30 y TRE, n = 5, <30 y TRE, n = 5, Ctrl, n = 5. Statistical
   analysis was performed using a paired t-test shown in [126]Fig. 3b.
   Data are mean ± s.e.m in [127]Fig. 3b.

2.7. TRE decreased BCR large clones and reduced class switch recombination in
PBMCs

   The treemaps in [128]Fig. 3c depict the seven-chain repertoire of
   typical participants before and after a 30-day TRE. In the treemaps,
   the area corresponds to the proportion of TCR alpha, beta, delta,
   gamma, BCR heavy, kappa and lambda chains, arranged from up to down and
   left to right. Larger color blocks indicate larger clones. The treemaps
   show the same changes in the proportions of the seven chains before and
   after TRE as in [129]Fig. 3a and b and S6a and b. Besides, it can be
   observed that after 30 days of TRE, the four chains of TCR were
   compressed by the BCR chain, while the large clones of heavy, kappa and
   lambda BCR chains were significantly reduced, thus the distribution of
   TCR and BCR became more balanced after TRE.

   Additionally, [130]Fig. 3d revealed variations in the proportions of
   immunoglobulins among the six isotypes of BCR before and after
   intervention. After 30 days of TRE, the proportion of immunoglobulin A
   (IgA) decreased ([131]Fig. S7a), especially in the >30 y group
   ([132]Fig. S7b), while the proportion of immunoglobulin M (IgM)
   increased ([133]Fig. S7a). However, there was no difference in serum
   secreted immunoglobulin levels both in the TRE group and the control
   group. Moreover, age subgroup analyses of the TRE group also did not
   show significant differences ([134]Figs. S7c and d). In
   antigen-antibody immune responses, the mature naive B cells produce IgM
   and immunoglobulin D (IgD) and undergo the class switch recombination
   (CSR) to produce isotype-switched high-affinity antibody immunoglobulin
   G (IgG), IgA or immunoglobulin E (IgE) that can best protect against
   the pathogen [[135]20,[136]21]. CSR is generally considered a hallmark
   of germinal centers and it also occurs readily in extrafollicular
   responses [[137]22]. Active CSR represents active immune response and
   the “CSR index” was utilized to quantify the immune response before and
   after 30 days of TRE [[138]23]. The thickness of the arrow represents
   the percentage of CSR between the immuneglobulins (Igs). [139]Fig. 3e
   shows a dramatic decrease in the CSR index values for the switch from
   IgM and IgD to IgG, IgE, or IgA expression after TRE, observed in both
   the <30 y and >30 y groups. The decrease in CSR index values indicates
   TRE may reduce the activation ability of B cells.

2.8. TRE attenuated serum inflammation by altering the abundance of
inflammation-related metabolites

   Considering the lack of changes in weight and immune system in the
   control group, the serum metabolomics and microbiome samples of the
   control group were not sequenced. Dietary restriction has metabolic
   benefits. A number of studies have reported that dietary restriction
   can affect the serum metabolites of healthy individuals, thus inducing
   anti-aging effects [[140]24] and improving metabolic syndrome
   [[141]25], etc. In this research, we focused on alteration in serum
   metabolomics before and after TRE.

   The principal component analysis (PCA) plot illustrates a distinct
   separation between Day0 and Day30 samples, suggesting a significant
   change in the metabolic phenotype of volunteers induced by TRE
   ([142]Fig. 4a). According to variable importance in projection
   (VIP) > 1 in orthogonal partial least squares discriminant analysis
   (OPLS-DA) model and P < 0.05 in T-Test, we screened 421 differential
   metabolites, with 219 up-regulated and 202 down-regulated in Day 30
   than in Day 0 ([143]Fig. 4b). And the heatmap illustrates 110
   endogenous metabolites changes in metabolite abundance ([144]Fig. 4c).

Fig. 4.

   [145]Fig. 4
   [146]Open in a new tab

   Metabolic differences before and after 30-day TRE. (a) Score scatter
   plot of the OPLS-DA model for metabolites on Day 0 and Day 30 of TRE;
   (b) Volcano plot of differential metabolites on Day 0 and Day 30 (VIP
   >1 and P < 0.05 in paired t-test); (c) Heatmap showing the metabolite
   abundances in participants before and after TRE. The 110 differential
   metabolites (d) and 71 subclasses of lipid and lipid-like molecules (e)
   after TRE. (f) KEGG metabolic pathway analysis in participants before
   and after TRE; (g) Analysis of pathway-based differential metabolites
   after TRE; (h) and (i): correlation analysis between CD4^+CD27^−CD28^−
   cell frequency and metabolite abundances; Abbreviations:
   #PC(20:6/22:5):
   PC(20:6(4Z,7Z,10Z,13Z,16Z,19Z)/22:5(4Z,7Z,10Z,13Z,16Z));
   #PS(22:6/22:2): PS(22:6(4Z,7Z,10Z,13Z,16Z,19Z)/22:2(13Z,16Z)); TRE Day
   0 group, n = 20; TRE Day 30 group, n = 20, these two groups used the
   same subject's samples before and after TRE. Statistical analysis was
   performed using Spearman's correlation coefficients in [147]Fig. 4i.

   Furthermore, metabolite subclass analysis was performed to further
   explore their functional classifications ([148]Fig. 4d). Enrichment
   analysis revealed that a significant portion of these differential
   metabolites are categorized under lipids (64.55 %) and proteins and
   amino acids (11.82 %) ([149]Fig. 4d). Among lipids,
   glycerophospholipids (52.11 %), fatty Acids (18.31 %) and acyl groups
   (16.90 %) were the top three subgroups ([150]Fig. 4e). The KEGG
   metabolic pathway analysis reveals that these differential metabolites
   were predominantly enriched in the glycerophospholipid metabolic
   pathway, cysteine and methionine metabolism and sphingolipid metabolism
   ([151]Fig. 4f and g). Further, we analyzed the correlations between the
   CD4^+CD27^−CD28^− cell frequency and differential metabolites
   ([152]Fig. 4h). The results suggested that, the increased abundance of
   some anti-inflammatory lipids including N4-Acetylcytidine,
   PS(22:6(4Z,7Z,10Z,13Z,16Z,19Z)/22:2(13Z,16Z)), LysoPE(15:0/0:0),
   PC(16:2e/3:0) and LysoPC(22:4(7Z,10Z,13Z,16Z)) were negatively
   associated with a decrease CD4^+CD27^−CD28^− cell frequency ([153]Fig.
   4i, [154]Fig.S9). β-Tyrosine and 7a-Hydroxyandrost-4-ene-3,17-dione, of
   which the abundance reduced after 30-day TRE, were positively
   associated with a decrease in the frequency of CD4^+CD27^−CD28^− cells
   ([155]Fig. S8e, Fig.S9). In addition, we also observed significant
   correlations between differential metabolites and Treg cell frequency.
   The increased abundance of PA(16:0/18:2(9Z,12Z),
   PC(20:4(5Z,8Z,11Z,14Z)/14:0), PI(20:1(11Z)/18:2(9Z,12Z)) and
   sphingosine-1-phosphate (S1P) was negatively correlated with the
   increased frequency of Treg cell, whereas decreased abundance of
   Farnesylcysteine was positively correlated ([156]Fig. S9).

   TRE may alter the sphingolipid metabolism pathway, in turn plays
   anti-inflammatory and immune-enhancing effects. The results showed the
   serum level of S1P and l-serine, enriched in the sphingolipid
   metabolism pathway, were significantly increased ([157]Fig. S8a).
   Subgroup analysis revealed lower baseline l-serine levels in the >30y
   group, which remained stable after TRE, while S1P levels showed no
   difference ([158]Fig. S8b). Specifically, S1P, a potent bioactive lipid
   [[159]26], could delay senescence [[160]27] and enhance immune cell
   functions [[161]28]. l-serine can not only suppress inflammation and
   oxidative stress by inhibiting the hypoxia-inducible factor α pathway
   [[162]29], but also maintain neuronal development and brain function,
   restore cognitive function and inhibit inflammation [[163]30].
   Increased levels of S1P and l-serine suggest TRE may exhibit potential
   anti-aging and essential processes regulating effects ([164]Fig. S8a).

   After TRE, the serum level of 5′-Methylthioadenosine (5′-MTA) decreased
   and 2-ketobutyric acid (α-KB) increased both in total and the age
   subgroup (Fig. S8a and b ). They are the meaningful differential
   metabolites in the cysteine and methionine metabolism pathway, which
   are characterized as anti-inflammatory factors. 5′-MTA inhibits TNF-α
   expression in macrophages and can alleviate inflammation-induced colon
   cancer [[165]31]. While α-KB inhibits anti-inflammatory responses by
   altering glycine metabolism, reduction in α-KB was found to impair
   anti-inflammatory effects [[166]32]. And supplementation with α-KB
   reduces mammalian cellular senescence and prolongs lifespan by
   promoting peroxisome function in Cryptomeria hidradii nematodes
   [[167]33].

   TRE can remodel lipid metabolism, in turn inhibiting inflammation and
   anti-aging effects. 30-day TRE increased the serum levels of
   PE(14:1/14:0), PC(22:6/22:5), LysoPC(15:0) and PA(16:0/18:2) ([168]Fig.
   S8c), which were enriched in glycerophospholipid metabolism pathway. In
   the age subgroup analysis, PE(14:1/14:0) and PC(22:6/22:5) levels were
   significantly increased in the <30 y group and had a comparatively
   higher baseline in the >30 y group, whereas LysoPC(15:0) and
   PA(16:0/18:2) levels were elevated in serum from both age groups
   ([169]Fig. S8d). Studies have revealed that the normal levels of
   phosphatidyl ethanolamine (PE) can decline with age in the brain,
   acting as a protective factor against aging-related diseases [[170]34].
   Particularly, introducing PE into Caenorhabditis elegans has been shown
   to enhance their anti-oxidative stress defense activity and effectively
   counteract the decline in muscle function due to aging. In addition,
   apoptotic cells release lysophosphatidylcholine (LysoPC) to recruit
   macrophages to clear themselves [[171]35], which prevents inflammatory
   responses caused by cellular debris. PE and phosphatidylcholine (PC)
   have been found to prolong the organism's lifespan by downregulation of
   the insulin/IGF-1-like signaling pathway [[172]36].

2.9. TRE shifted gut microbiota toward youthful composition by increasing
alpha diversity and abundance of Akkermansia and Rikenellaceae

   The gut microbiota shows varying characteristics with aging and is
   directly linked to the host's health and longevity. Studies have found
   that the intestinal microecological structure of aged individuals
   differs from that of young individuals [[173]37], while centenarians
   and young people share a similar gut microbiota composition [[174]38].
   Therefore, we performed 16S DNA amplicon sequencing to investigate the
   effects of 16:8 TRE on the gut microbiota and how these effects relate
   to aging.

   In microbiology, quantifying biodiversity is a vital aspect of
   research. In this study, we assessed the diversity of the gut
   microbiome before and after TRE by analyzing alpha diversity, beta
   diversity, Chao's index and Amplicon sequence variables (ASVS). Alpha
   diversity can show the richness of species in a community and the
   evenness of the number and distribution of each species. Beta diversity
   describes the species diversity between two communities or ecosystems.
   The Chao index estimates the total species richness of a community by
   calculating rare species. ASVS groups similar DNA sequences into one
   ASV via high-throughput sequencing, allowing for the analysis of
   microbial community diversity, composition, and structure. In this
   research, the Chao1 diversity index was higher on Day 30 than on Day 0
   ([175]Fig. 5a) with a more pronounced increase in the <30y group
   ([176]Fig. S10a). However, there is no significant difference in
   Shannon index and beta diversity ([177]Fig. 5b, [178]Fig. S10 a&f). The
   ASVS for the gut microbiota was higher on Day 30 ([179]Fig. 5c) with a
   more pronounced increase in the >30y group ([180]Fig. S10c). These
   results suggest a significant elevation in microbiome alpha diversity.

Fig. 5.

   [181]Fig. 5
   [182]Open in a new tab

   Gut microbiome differences before and after 30-day TRE. (a) The Chao1
   community richness index, (b) Shannon index and (c) the total number of
   amplicon sequence variants (ASVs) observed before and after TRE; (d)
   Changes in the phylum distribution of gut microbiomes before and after
   TRE; (e) Changes in abundance of Akkermansia and Rikenellaceae before
   and after TRE; (f) Changes in the genus distribution of gut microbiomes
   before and after TRE, only the top 30 predominant taxa are shown. (g)
   Correlation analysis between genus and differential metabolites on Day
   0 and Day 30. (h) Correlation analysis between genus and the
   frequencies of Sene, Th1, Th2, Th17 and Treg cells. TRE Day 0 group,
   n = 12; TRE Day 30 group, n = 12. Abbreviations: ∗un: unclassified;
   Clostridiales_Family_XIV:
   ∗Clostridiales_Family_XIV._Incertae_Sedis_unclassified. Statistical
   analysis was performed using the paired Wilcoxon test at two different
   time points, as shown in [183]Fig. 5a, and the paired t-test was
   performed in [184]Fig. 5a,b&e. Data are the mean ± s.e.m.

   At the phylum level, after a 30-day TRE, the relative abundances of
   Firmicutes and Proteobacteria decreased, while those of Bacteroidetes
   and Actinobacteria showed an increasing trend ([185]Fig. 5d). Age
   subgroup analyses showed that these changes were more pronounced in the
   >30 year group ([186]Fig. S10b). The gut microbiome seems to be more
   significantly impacted by TRE in >30y individuals.

   At the genus level, our study observed a notable increase in the
   relative abundances of the genera Akkermansia and Rikenellaceae
   ([187]Fig. 5e). While we found no statistical significance in age
   subgroup analysis ([188]Fig. S10d). Additionally, we detected changes
   in the relative abundances of several other flora after TRE, as shown
   in [189]Fig. 5f and [190]Fig. S10e. These included an increase in
   certain butyric acid-producing bacteria, such as Clostridiales _
   unclassified, Eubacterium, Coprococcus, and Clostridiales, as well as
   other genera like Blautia, Phascolarctobacterium, Bifidobacterium, and
   Roseburia. On the basis of increased α-diversity and a rise in the
   number of beneficial bacteria, we also observed that the microbial
   community structure of the two age groups became more consistent after
   TRE ([191]Fig. S10e). However, these variations did not reach
   statistical significance. In general, while only the alterations in
   Akkermansia and Rikenellaceae were statistically significant, the
   overall trend in the gut microbiome composition indicated a shift
   toward patterns observed in centenarians and young adults.

   To identify all genera that were associated with serum metabolites, we
   investigated the associations of genus-level abundances ([192]Fig. 5g,
   [193]Fig. S11) with serum metabolites, and Spearman's correlation
   coefficients were applied. We observed a shift in the association of
   microbial species with metabolites from Day0 to Day30 of TRE,
   suggesting that TRE may influence the dynamic interactions between gut
   microbes and serum metabolites ([194]Fig. 5g). Specifically, we found
   that the serum level of N4-acetylcytidine, PC (16:2e/3:0) and
   phosphatidylserine (PS) (14:0/15:0) were positively correlated with
   Akkermansia. N4-Acetylcytidine, a conservative chemical modification in
   eukaryotic prokaryotes, is negatively related to inflammation
   [[195]39]. Threonic acid and cholesterol sulfate, which are positively
   associated with metabolic diseases and stress, were negatively
   correlated with Rikenellaceae [[196]40]. The results showed that TRE
   affects material metabolism in part through the gut microbiota.
   Further, we also analyzed the correlation between gut microbial species
   and immune cell subsets ([197]Fig. 5h). Decreased abundance of
   disease-associated Burkholderia and Thiohalomonas was positively
   correlated with a decrease in senescent T-cells, whereas decreased
   abundance of Burkholderia was also positively correlated with a
   decrease in Th2 cells. Increased abundance of
   Clostridiales_unclassified was negatively correlated with decreased
   frequencies of Th2 and Th17 cells, while increased abundance of
   Eubacterium was negatively correlated with decreased frequency of
   Th17 cells. Increased abundance of Phascolarctobacterium was negatively
   correlated with increased Th1 cell frequency. Together, these results
   highlight the potential role of gut microbes in modulating host immune
   responses ([198]Fig. 5h).

3. Discussion

   TRE is commonly used to fight obesity. It could be an effective,
   economical, safe and simple form of treatment for reducing weight,
   improving metabolism, and even fighting aging. The senescence of immune
   cells is the core factor of body senescence, and serum metabolomics and
   intestinal microbiota are associated with the status of immune cells.
   Therefore, we provided the concept that TRE is not just for weight loss
   but can achieve anti-aging effects and improve metabolism by
   “refreshing” the immune system and intestinal microecology.

   After the 30-dayTRE, a decrease in BMI within the normal range was
   observed. Meanwhile, parameters including body fat rate, visceral fat,
   and fat weight decreased, while body moisture and skeletal muscle rate
   increased. Body moisture usually refers to the water inside the body,
   but it also decreases with age and decreases with obesity. Also, most
   participants expressed that they get reduced physical burden,
   gastrointestinal comfort, good mental outlook and happy mood during the
   TRE. These data were subjective feelings from the questionnaire and
   were hard for statistical analysis. To sum up, TRE is a relatively
   healthy way to lose weight.

   However, muscle mass and basal metabolic rate decreased after TRE, but
   dieting inherently causes muscle loss and reduces lean muscle tissue,
   further reducing basal metabolism, which is consistent with our
   observations. Exercise prevents lean muscle tissue loss and increases
   the amount of lean muscle tissue, which is the most effective way to
   promote basal metabolic rate [[199]41]. Thus, TRE combined with proper
   exercise and calorie intake adjustment may be an even better way to
   lose weight.

   We evaluated the effect of TRE on immune system function in terms of
   (1) the frequency of immune cell subsets in PBMCs and (2) the
   composition of the IR in PBMCs before and after 30 days of TRE.

   After 30-day TRE, the differences were mainly manifested in CD4^+
   T-cell subsets. Previous studies also showed that CD4^+ T cells
   mediated obesity memory and promoted weight regain after weight loss
   [[200]42]. Obesity and metabolic syndrome have potential connections
   with immune diseases [[201]43,[202]44]. In detail, our results showed
   that the frequency of Th2 and Th17 cells decreased. Additionally, the
   frequencies of Treg and Tfh-like cells increased after TRE. Increased
   frequency of circulating Tfh-like cells was observed, suggesting higher
   humoral immune responses after TRE. However, when we detected the Ig
   levels after TRE, there was no significant difference. Circulating
   Tfh-like cells in humans might not really reflect the germinal center
   responses and Ig production.

   Senescent cells are key drivers of aging, and their elimination can
   alleviate many age-related diseases [[203]45]. In the process of aging,
   many cells will exhibit senescence phenotypes, but immune cell
   senescence is the most harmful. Studies have shown that immune cell
   senescence can exacerbate the whole body's aging [[204]46]. Our
   findings suggest that 30-day TRE treatment can reduce CD4^+ senescent T
   cells and may serve as an anti-CD4^+ T-cell senescence therapy.
   According to Li et al., the 30s and 50s are transitional periods during
   Chinese women's aging [[205]17]. Therefore, we further stratified the
   analysis of data from subjects >30y and <30y. We were pleasantly
   surprised to see that the decrease in >30y became more pronounced.

   Regarding the IR, the normal body constantly encounters a variety of
   antigens, both endogenous and exogenous. The detection of monoclonal
   rearrangement of the BCR gene indicates pathological status, such as
   infection [[206]47] and autoimmune diseases [[207]48]. The
   significantly reduced large clones of heavy, kappa and lambda BCR
   chains ([208]Fig. 3b) showed that TRE was able to improve immune system
   function in healthy individuals. Despite changes in the proportion of
   immune cell subsets, there was no difference in IR diversity between
   TRE and its predecessors ([209]Figs. S6b and c). The higher the IR
   diversity is, the more robust the immune system. Thus, the 30-day TRE
   was able to eliminate immune-senescent cells without affecting the
   functions of immune cells. However, TRE decreased all CSR index values,
   indicating that B-cell activation ability was reduced and that TRE
   therapy may impair humoral immune responses, which still needs to be
   confirmed by further experiments.

   Changes in serum metabolites showed a clear anti-inflammatory effect
   after 30-day TRE, which is consistent with previous studies
   [[210]49,[211]50]. In short, the abundance of some metabolites with
   anti-inflammatory effects (PE, PC, LysoPC, S1P, α-KB, l-serine and
   N4-Acetylcytidine) increased. Notably, the shifts in these metabolite
   levels are strongly linked to the reduction of senescent T cells and
   the rise in Treg cells, indicating that TRE might enhance health by
   regulating immune responses. Among them, S1P is an important bioactive
   lipid, in inflammatory or infectious situations, its concentration
   gradient plays a key role in the migration of T cells from the lymph
   nodes (relatively lower concentration) to the tissues and eventually
   into the blood (relatively higher concentration) [[212]51]. In
   addition, the duration of T cells in the lymph node is thought to be an
   important factor in regulating their differentiation [[213]52]. It has
   been shown that a decrease in S1P concentration contributes to the
   maintenance of the central memory T cell and hinders the
   differentiation of Naïve T cells to Treg by inhibiting the PPARγ
   pathway [[214]53]. After 30-day TRE, an observed increase in Treg cell
   frequency and a corresponding decrease in Th17 and Th2 cells in PBMCs
   may suggest a regulatory influence of S1P on T cell differentiation.
   Although a negative correlation was found between S1P levels and Treg
   cell frequency, the precise regulatory mechanisms of S1P on T cell
   differentiation require further study. These findings provide new
   perspectives for understanding the role of S1P in immune regulation.
   Additionally, other anti-inflammatory metabolites, which showed a
   similar negative correlation with Treg cell frequency, merit further
   investigation into their relationship with T cell regulation.

   Enrichment analysis of differential metabolites identified two
   potential pathways after TRE intervention: glycerophospholipid
   metabolic pathway and cysteine and methionine metabolism pathway. A
   study in RA patients revealed a significant negative correlation
   between glycerophospholipids and inflammatory markers like erythrocyte
   sedimentation rate (ESR), C-reactive protein (CRP), and IL-6, as well
   as age, and also a negative correlation with leukocyte counts. This
   indicates a potential anti-inflammatory role of the glycerophospholipid
   pathway [[215]54]. Cysteine and methionine are important amino acids,
   and their metabolic pathways are important for the body's antioxidant
   and free radical elimination functions. The cysteine methionine pathway
   has been found to help enhance resistance to stressful environments
   [[216]55]. The enrichment of differential metabolites in this pathway
   suggests that TRE may augment the body's antioxidant defenses to a
   certain degree.

   As for gut microbiota, it tightly links to the health and longevity. On
   the one hand, high microbial diversity supports gut ecosystem stability
   and resilience, with a healthy microbiome marked by abundant
   Bacteroidetes and Firmicutes, high levels of probiotics like
   Bifidobacterium, and the production of short-chain fatty acids (SCFAs)
   like acetic and butyric acid [[217]56]. On the other hand, the
   composition of microbiota changes with age, stabilizing by age 3–5 and
   varying due to lifestyle, diet and so on, with major changes in old age
   (>65). Notably, the microbiota of healthy elderly or centenarians
   resembles that of the 30s [[218][57], [219][58], [220][59]].

   We also found some meaningful results regarding the gut microbiota in
   this research. After 30-TRE, a significant increase in the alpha
   diversity of gut microbes was observed, along with an increase in the
   abundance of key microbial populations such as Akkermansia and
   Rikenellaceae. Akkermansia, a beneficial gut probiotic, has been shown
   to prevent obesity, improve glucose tolerance, extend lifespan
   [[221]60] and possess anti-inflammatory properties [[222]61].
   Interestingly, although its relative abundance decreased with age, it
   was comparable to that of young adults in centenarians [[223]37]. The
   genus Rikenellaceae spp. is more abundant in lean individuals and is
   considered a potential indicator of healthy aging and longevity
   [[224]62].

   In addition, the relative abundance of some butyrate-producing bacteria
   increased with no significant statistical differences
   (Clostridiales_unclassified, Roseburia, Clostridiales, Parabacteroide,
   Blautia and Eubacterium).These bacterial genera, which tend to decrease
   in abundance in the elderly, show similar levels in centenarians and
   young individuals [[225]57,[226]63,[227]64], and possess
   anti-inflammatory [[228]65] and anti-obesity potential [[229]66].
   Studies also indicate that an increase in butyrate-producing bacteria
   is associated with reduced frailty and inflammation, as well as
   improved cognitive function [[230]67]. Moreover, the relative abundance
   of β-proteobacteria, which decreases in centenarians [[231]39], also
   declined after TRE. These results suggest a trend towards a "younger"
   and "healthier" composition of the gut microbiota.

   In the correlation analysis, we found that serum metabolomics and gut
   microecology of healthy participants showed significant
   anti-inflammatory trends through a 30-day TRE intervention.
   Specifically, metabolites negatively associated with inflammation, such
   as N4-acetylcytidine, showed a positive correlation with Akkermansia
   abundance [[232]39]. Meanwhile, metabolites positively associated with
   metabolic diseases and stress, such as threonine and cholesterol,
   showed a negative correlation with Rikenellaceae abundance
   [[233]40,[234]68]. The association analysis results marked an
   anti-inflammatory trend in serum metabolomics and intestinal
   microecology in healthy participants after a 30-day TRE intervention.
   Moreover, three genera, Clostridiales_unclassified, Eubacterium and
   Phascolarctobacterium, produce SCFAs in the gut. Their increased
   abundance is negatively correlated with changes in Th cell subsets.
   SCFAs promote the release of the anti-inflammatory cytokine IL-10,
   inhibit pro-inflammatory factors, and maintain intestinal immune
   homeostasis through the activation of G protein-coupled receptors (GPR)
   such as GPR43 and GPR41 [[235]69,[236]70]. In addition, SCFAs play an
   anti-inflammatory role by regulating T cell differentiation
   [[237]71,[238]72], which is consistent with the negative correlation we
   observed. These results suggest that TRE remodels the intestinal
   microecology and helps reduce intestinal inflammation.

   Undoubtedly, this study is subject to the limitations inherent in
   clinical trial design. Due to these constraints, our capacity to enroll
   a sufficiently large cohort of age-matched control subjects was
   compromised, and it was not viable to subject the control group samples
   to a comprehensive panel of biomarker assessments. Consequently,
   individuals around the age of 50 were less willing to participate in
   our TRE clinical trial, which posed a considerable challenge in
   volunteer recruitment. Next, the analytical strategy employed in this
   study predominantly relied on comparisons of the participants'
   biomarker profiles before and after TRE. Furthermore, there was a
   significant loss of follow-up among subjects in the extended
   post-intervention period, which may impact the robustness of our
   long-term findings. Similarly, in aging-related studies, subjects may
   not be old enough to effectively assess the effects of TRE on aging.
   Despite these challenges, our findings indicate a pronounced
   improvement in immune senescence metrics among female participants over
   the age of 30, suggesting a heightened susceptibility of older
   individuals to the beneficial effects of TRE.

   The multi-omics and correlation analyses have revealed the mutual
   connections between the frequency of immune senescent cells, serum
   metabolites, and the gut microbiome. ([239]Fig. 4h–i, [240]Fig. 5h).
   Overall, the reduction in the frequency of immune senescent cells is
   the most compelling evidence of the effectiveness of TRE intervention.
   Concurrently, we observed an upregulation of beneficial metabolites
   (such as PE, PC, LysoPC, S1P, α-KB, l-serine, and N4-acetylcytidine)
   and an increase in the abundance of microbiota associated with
   longevity (such as Akkermansia and Rikenellaceae). These data suggest a
   comprehensive shift towards a "younger" state post-TRE.

   In conclusion, all the evidence suggests that TRE, as a dietary
   intervention, is not just beneficial for weight loss but also reveals a
   “younger” immune system and intestinal microbiome.

4. Methods

4.1. Study registration

   The study was carried out with the agreement of the Second Xiangya
   Hospital's ethical committee and in compliance with good clinical
   practice and the International Declaration of Helsinki (GCP). The
   clinical examinations were conducted at the Hunan Key Laboratory of
   Medical Epigenomics at the Second Xiangya Hospital, China. The study
   was registered at chictr.org.cn (ChiCTR2200058137).

4.2. Participants

   We recruited a total of 59 participants from the area around the Second
   Xiangya Hospital (<3 km), Hunan Province, through posters and social
   media and they were all available to participate in the clinical trial
   in the laboratory. 49 participants formed the TRE group and 10
   participants formed the control group. The trial recruited participants
   who consumed food over 8 h per day and had not fasted in the last year.
   All participants should meet the inclusion criteria and not the
   exclusion criteria. The consort diagram can be found in [241]Fig. S1,
   and baseline information for each group is presented in [242]Table 1.

Table 1.

   Characteristics of baseline.
      Characteristic     TRE (Total) TRE (<30 y) TRE (>30y)    Control
   No.                   49          20          29          10
   Age, mean (SD), years 33.2 (8.50) 25.5 (2.11) 38.5 (6.85) 25.1 (1.60)
   Female                44 (89.8 %) 18 (90.0 %) 26 (89.7 %) 7 (70 %)
   Weight, mean (SD), kg 63.2 (10.1) 61.6 (6.3)  64.3 (11.8) 56.55 (11.9)
   BMI, mean (SD)        24.0 (3.9)  22.7 (2.4)  24.9 (4.3)  20.84 (2.6)
   [243]Open in a new tab

   BMI: body mass index (BMI = kg/m^2), TRE: time-restricted eating, Data
   are mean (SD).

   This study focused on dietary timing adjustments and used a single-arm
   design without implementing blinding or randomization procedures
   Participants who met the criteria and volunteered to participate in TRE
   were included in the study. Volunteers who lived in the same area and
   had similar living habits as TRE volunteers were included in the
   control group, which was mainly used to reduce errors in the analysis
   of IR.

4.3. Outcomes

   The primary and secondary outcome measures are as follows [Time window:
   30 days (from baseline to 30 days)]: (A) Primary outcome metric:
   CD4^+CD27^−CD28^− senescent T cell frequencies in PBMCs. Flow cytometry
   was used to detect changes in the frequencies of senescent T cells in
   PBMCs. (B) Secondary outcome indicators: (1) weight changes and other
   weight-related index: body moisture, skeletal muscle, body fat rate,
   BMI, muscle mass, visceral fat, bone mass, basal metabolism and fat
   weight. (2) Frequencies of each immunocyte and immunosenescent cell in
   PBMC. Flow cytometry detection of changes in the frequencies of the
   following immunocytes: (a) each subset of T cells (CD4^+ T cells, CD4^+
   Naive T cells, CD4^+ Effector T cells, Th1, Th2, Th17, Treg, Tfh-like,
   CD8^+ T cells, CD8^+ Naive T cells, CD8^+ Memory T cells). (b) B cell
   subpopulation (total CD19^+ B cells, Naive B cells, Memory B cells, and
   plasma cells). (c) CD27^−CD28^− immunosenescent cells in each subset.
   (3) Serum metabolomics: LC-MS metabolomic analysis was applied to
   detect changes in serum metabolites. (4) Immune repertoire (IR): IR
   refers to the sum of all functionally versatility T cells and B cells
   in the circulatory system of individuals. (5) Gut microorganism: 16S
   rDNA sequencing analysis was applied to detect changes in intestinal
   microecology.

4.4. Inclusion criteria and exclusion criteria

   The inclusion criteria were as follows: (1) 18–60 years old, both
   inclusive; (2) weight stability (change < ±10 % from current body
   weight) for Day90 before trial (3) BMI over 18.0 kg/m^2, both
   inclusive; (4) no cardiovascular or metabolic diseases; (5) no acute or
   chronic inflammation; (6) no drugs; (7) no alcohol abuse; (8) no
   dietary restrictions (e.g., vegetarian); (9) no; gastrointestinal
   diseases. The exclusion criteria were as follows: (1) eating disorder
   history; (2) malignancy confirmed by biopsy; (3) pregnancy,
   breastfeeding or trying to conceive; (4) hormone supplements or
   anti-fertility hormone drugs (for over 2 months); (5) received
   antidepressant medication within the past 6 months; (6) intake of
   nonsteroidal anti-inflammatory drugs (NSAIDs).

4.5. Grouping dietary interventions and intervention supervision

   The TRE group ate during the daily period from 9 a.m. to 5 p.m. with a
   24-h cycle. Participants in the TRE group were provided nutritious
   meals by the nutrition canteen of the Second Xiangya Hospital from 9
   a.m. to 10 a.m., 12 a.m.-1 pm, and 4 p.m.–5 pm every day for 30 days.
   The recipes were carefully designed by clinical nutritionists based on
   The Chinese Dietary Guidelines (2016 edition) to ensure a balanced
   nutritional profile. The daily diet is shown in [244]Table S1. The
   clinical nutritionist calculated each volunteer's basal metabolic rate
   by referring to the Mifflin-St. Jeor equation [[245]73] and then
   multiplied it by the daily exercise (an appropriate factor), resulting
   in a nutritious meal with slightly more calories than the daily energy
   requirement per meal (A. infrequent exercise, a lot of sitting work:
   1.1; B. moderate exercise, a lot of sitting work: 1.3; C. frequent
   exercise, a lot of physically active work: 1.5; D. a lot of exercise, a
   lot of physically active work: 1.7). Participants in the control group
   did not undergo any specific interventions since their daily diets are
   quite similar to nutritious meals provided by nutrition canteen. As a
   result, the daily calories and nutrients consumed by the two groups
   were roughly similar, except for the inconsistent eating times.

   Participants were expected to take pictures of their meal while eating
   it as well as after they completed it and to submit them to the
   investigators using WeChat as part of the compliance criteria. If the
   TRE participants failed to post a meal photo, they were reminded in an
   hour. It was considered a failure to complete the trial if participants
   missed uploading those photos nine times (or three days). The uploaded
   photos were checked daily to ensure that participants consumed enough
   nutrition.

4.6. Samples collection

4.6.1. Human peripheral blood sample collection

   A sterile vacuum blood collection tube was used for each group to
   collect 15 ml of fresh EDTA-anticoagulated peripheral venous blood and
   3 ml of procoagulant at Day 0, Day 14, and Day 30. We also took blood
   samples from participants on Day 90 after TRE to understand the
   long-term effects. [246]Table 1 presents the detailed baseline
   characteristics of the TRE participants.

4.6.2. Stool sample collection

   We collected stool samples from each subject on Day 0 and Day 30,
   transferred them into liquid nitrogen for quick freezing for 2 min
   within 1 h, and preserved them at −80 °C to send for 16S rDNA
   microbiome sequencing.

4.7. Body weight and weight-related indicators analysis

   The weight and weight-related indicators of all participants were
   measured using the same multifunctional body mass measuring instrument.
   The instrument uses an electrical resistance method to measure body fat
   percentage and body moisture. The instrument can measure indicators
   including body weight, body fat rate, BMI, muscle mass, body moisture,
   visceral fat, bone mass, basal metabolism, fat weight and skeletal
   muscle rate. On Day 0, Day 7, Day 14, Day 21 and Day 30 of this trial,
   the body weight and other weight-related indicators of the participants
   in their fasting were measured at 8 a.m. Notably, in order to rule out
   post-eating disturbances, the participants were also asked not to eat
   breakfast and drink too much water at baseline weight measurements.

4.8. Flow cytometry analysis

   PBMCs were separated from fresh EDTA-anticoagulated peripheral venous
   blood using Ficoll-Paque™ PLUS (cytiva, 17144003). Approximately two
   million cells per tube suspended in phosphate-buffered saline (PBS)
   were mixed with labeled surface antibodies for 45 min in the dark at
   4 °C to fully bind. FlowJo software (V10.5.3) was used to analyze flow
   cytometry data. All flow cytometry data were acquired by a DxP Athena™
   flow cytometer (Cytek, US). The flow cytometry antibodies and kits used
   are listed in the [247]Supplementary Table S2.

4.9. Immune repertoire analysis

   The PBMC samples of the top five subjects with the largest reduction in
   the proportion of immunosenescent cells in the >30 y group and the
   <30 y group were selected for sequencing, and five control samples were
   tested at the same time. IR analysis was supported by Nanjing ARP
   Biotechnology Co., Ltd., and the extracted total RNA was amplified
   using the commercially available iR-RepSeq-plus 7-Chain Cassette
   (iRepertoire Inc., US) covering all 7 chains of the human TCR and BCR
   through a reverse transcription technique that incorporates unique
   molecular identifiers. Using one disposable cassette with preloaded
   reagents for amplification and purification, human TCR and BCR
   libraries were constructed, and RT was performed using OneStep RT‒PCR
   mix (Qiagen). Amplification of the library was accomplished using
   primers engineered specifically for communal site primers. SPRIselect
   bead selection (Beckman Coulter) was used to remove unused primers from
   first-strand cDNA. V-gene primer mix was then used for the second
   binding and extension round. Using a 600-cycle kit, amplified libraries
   were multiplexed, pooled, and sequenced on the Illumina MiSeq platform.
   The iRmap tool was used to analyze the raw sequencing data [[248]74].
   Details of this IR methodology have been described previously
   [[249]75]. Sequence analysis was finished in GitHub, and the data were
   exported to Microsoft Excel (Microsoft 365).

4.10. LC-MS metabolomic analysis

   Serum samples were selected from the top 20 participants in the TRE
   group, who displayed a decreased proportion of immunosenescent cells,
   for untargeted metabolomics analysis. Samples were obtained on Day 0
   and Day 30, all during their fasting state at around 8 a.m.
   Metabolomics analysis was accomplished in the laboratory or with the
   assistance of Shanghai Biotree Biotech Co., Ltd. A total of 12,241
   peaks were extracted from the original data package, and a single peak
   of the original data was filtered to remove noise, filtered and filled
   with missing values, and normalized with an internal standard (IS), in
   which Student's t-test was applied. For the multivariate statistical
   analysis, SIMCA software (V16.0.2) was used [[250]76,[251]77]. The data
   were analyzed by PCA. OPLS-DA of the metabolites removed orthogonal
   variables not relating to categorical variables. To obtain more
   reliable information on metabolite group differences and their
   relevance to the experiments, we analyzed the orthogonal variables
   separately from the nonorthogonal variables. We first analyzed the
   OPLS-DA model on the first principal component based on logarithmically
   transformed and UV-formatted data. The quality of the model was checked
   using sevenfold cross-validation. The validity of the model was then
   assessed using the cross-validation results. A permutation test was
   used to further examine the validity of the model. An additional
   assessment of the robustness was performed on the OPLS-DA model with a
   permutation test. Using the combined findings of statistical analysis
   of unit variables and multivariate variables, the differential
   metabolites were filtered out following the aforementioned analysis.
   The chi-square value of multivariate statistical analysis has the
   following two criteria.
     * 1.
       The p-value of Student's t-test was less than 0.05; 2. The VIP
       value is greater than 1. Hierarchical clustering was used to
       analyze the results of the screened differential metabolites and
       the functional similarity/complementarity features.

   We identified metabolites by searching the KEGG database and HMDB
   database ([252]https://hmdb.ca/). Pathway enrichment analysis was
   performed using MetaboAnalyst 5.0[[253]78].

4.11. 16S rDNA microbiome sequencing

   16S DNA amplicon sequencing was performed on fecal samples from 12
   healthy participants. Microbiome sequencing analysis was accomplished
   with assistance from Shanghai Biotree Biotech Co., Ltd. On Day 0 and
   Day 30 of the study, stool samples were collected from the relevant
   groups. The total DNA of the stool microbiome was extracted with the
   E.Z.N.A.®Stool DNA Kit [[254]79]. The hypervariable V3–V4 region of
   bacterial 16S rDNA was sequenced on the NovaSeq PE250 platform
   [[255]80,[256]81], in which the primers 805R and 341F
   (5′-GACTACHVGGGTATCTAATCC-3′ and 5′-CCTACGGGNGGCWGCAG-3′, respectively)
   were applied. The manufacturer's instructions were followed for
   sequencing samples on the Illumina NovaSeq platform. Sample-specific
   barcodes were used to assign paired-end sequences to each sample, and
   primer sequences and barcodes added during library creation were
   removed. Matching end readings were combined using FLASH. Fqtrim
   (v0.94) claims that to produce high-quality clean labels, the raw read
   data were quality-filtered under particular filtering circumstances.
   Vsearch software was used to screen out chimeric sequences (v2.3.4)
   [[257]82]. The feature table and feature sequence were retrieved using
   DADA2. By normalizing to the same random sequence, diversity was
   computed. Each sample's relative abundance was used to normalize
   feature abundance, which was determined by the SILVA classifier. Chao1
   and alpha diversity were used to detect the species diversity of the
   samples. BLAST was employed for sequence alignment, and then we
   annotated unique sequences from the SILVA database on each
   representative sequence. We also computed Spearman correlations between
   the relative abundances of genera and serum metabonomics. The graphs
   were produced with Prism 8 and R (v3.5.2).

4.12. Statistical analysis

   IBM SPSS® software and GraphPad Prism v.8.0.2 performed statistical
   calculations. For the data of the proportion of immune cell subsets in
   TRE group participants, if Mauchly's test of sphericity sig >0.5, tests
   of within-subjects effects were used for data analysis. If Mauchly's
   test of sphericity sig <0.5, the Greenhouse-Geisser was used for data
   analysis. In addition, the proportions of immune cell subsets at
   different time points were compared by paired t-test. For the IR
   sequence data, two-way ANOVA was used for statistical analysis. For
   statistical analysis of data before and after the trial, such as
   metabolomic analysis and microbiome sequencing, a two-tailed paired
   t-test and paired Wilcoxon test were used for statistical analysis.
   Data are presented as the mean ± s.e.m. ns = not significant. The
   sample size calculation was done for the primary outcome only. The
   methodological principles of the statistical analyses for the age
   subgroups of the supplementary are the same as above. Sellami Maha et
   al. studied the changes in the number of immune cell subsets in human
   PBMCs after 12 weeks of intermittent fasting [[258]83]. In Sellami
   Maha's study, the mean ± standard deviation of CD4^+ T cells in the
   experimental group and the control group were 1206.4 ± 225.41 mm^3 and
   985.6 ± 216.14 mm^3. The α value was 0.05, the proportion of lost
   follow-up was 0.1, and the confidence was 0.95. Comparing the two
   independent sample means to estimate the sample size of the
   experimental group and the control group, the calculation results
   showed that 17 participants were required for the TRE group. Sample
   size calculations were performed using the MedSci Sample Size tools
   (MSST). All the statistical analyses were reviewed by Professor Irene
   Xy Wu, a statistical expert at Xiangya School of Public Health, Central
   South University, China.

CRediT authorship contribution statement

   Yiran Chen: Writing – review & editing, Writing – original draft,
   Visualization, Project administration, Methodology, Investigation,
   Formal analysis. Xi Li: Writing – review & editing, Writing – original
   draft, Visualization, Project administration, Investigation, Formal
   analysis. Ming Yang: Methodology, Funding acquisition, Formal analysis.
   Chen Jia: Investigation, Formal analysis. Zhenghao He: Visualization,
   Formal analysis. Suqing Zhou: Visualization. Pinglang Ruan:
   Visualization, Data curation. Yikun Wang: Data curation. Congli Tang:
   Formal analysis. Wenjing Pan: Formal analysis. Hai Long: Methodology,
   Funding acquisition. Ming Zhao: Methodology. Liwei Lu: Formal analysis.
   Weijun Peng: Formal analysis. Arne Akbar: Formal analysis. Irene XY.
   Wu: Data curation. Song Li: Formal analysis. Haijing Wu: Writing –
   review & editing, Writing – original draft, Validation, Supervision,
   Project administration, Methodology, Funding acquisition,
   Conceptualization. Qianjin Lu: Writing – review & editing, Supervision,
   Methodology, Funding acquisition.

Data and code availability

   In accordance with the participant's privacy agreement, personally
   identifying information is not released in this article but may be
   shared by the corresponding author in special circumstances.

   The paper does not report the original code. Any additional information
   required to reanalyze the data reported in this paper is available from
   the lead contact upon request.

Declaration of generative AI in scientific writing

   The authors declare that the ChatGPT 4.0 was used to improve
   readability and language of this manuscript. The authors remain fully
   responsible and accountable for the content of the work.

Funding

   This work is supported by the National Key R&D Program of China, No.
   2022YFC3601800 (Q. Lu), funding from the Ministry of Science and
   Technology of the People's Republic of China, China. The Special
   Program of National Natural Science Foundation of China, No. 32141004
   (Q. Lu), funding from the National Natural Science Foundation of China
   (NSFC), China. CAMS Innovation Fund for Medical Sciences (CIFMS) No.
   2021-I2M-1–059 (Q. Lu), Non-profit Central Research Institute Fund of
   Chinese Academy of Medical Sciences No. 2021-RC320-001 (Q. Lu),
   National Natural Science Foundation of China (NSFC) No. 82173425 (H.
   Wu), No. 82373488 (H. Wu), No. 82273540 (H. Long), No. 82473530 (M.
   Yang), Natural Science Foundation of Hunan Province of China No.
   2024JJ4077 (M. Yang).

Declaration of competing interest

   The authors declare that they have no known competing financial
   interests or personal relationships that could have appeared to
   influence the work reported in this paper.

Acknowledgements