Common and distinct metabolomic markers related to immune aging in Western European and East African populations

In old age, impaired immunity causes high susceptibility to infections and cancer, higher morbidity and mortality, and poorer vaccination efficiency. Many factors, such as genetics, diet, and lifestyle, impact aging. This study aimed to investigate how immune responses change with age in healthy Dutch and Tanzanian individuals and identify common metabolites associated with an aged immune profile. We performed untargeted metab-olomics from plasma to identify age-associated metabolites, and we correlated their concentrations with ex-vivo cytokine production by immune cells, DNA methylation-based epigenetic aging, and telomere length. Innate immune responses were impacted differently by age in Dutch and Tanzanian cohorts. Age-related decline in steroid hormone precursors common in both populations was associated with higher systemic inflammation and lower cytokine responses. Hippurate and 2-phenylacetamide, commonly more abundant in older individuals, were negatively correlated with cytokine responses and telomere length and positively correlated with epigenetic aging. Lastly, we identified several metabolites that might contribute to the stronger decline in innate immunity with age in Tanzanians. The shared metabolomic signatures of the two cohorts suggest common mechanisms of immune aging, revealing metabolites with potential contributions. These findings also reflect genetic or environmental effects on circulating metabolites that modulate immune responses.


Introduction
Aging is a physiological process of decline in the function of different organs and systems of an organism, often associated with increased morbidity.Population aging will become an increasing health and economic issue in the coming decades, with individuals over 65 estimated to reach 1.5 billion by 2050 (Nations U, 2020).The aging process affects virtually all tissues and systems, including the immune system.An aged immune system is generally characterized by chronic subclinical systemic inflammation that can lead to tissue damage, a cellular senescence-associated pro-inflammatory secretory profile, impaired immune function causing vulnerability to infections, and suboptimal response to vaccination (Bulut et al., 2020;Pawelec, 2018).
Finding biomarkers of an aged immune system is essential to monitor the immune health of individuals and to identify potential targets to delay the aging process.Many biomarkers have been suggested, particularly since the increased availability of omics-based approaches, but only a few markers that represent age-related systemic inflammation have been identified, such as interleukin 6 (IL-6) and C-reactive protein (CRP) (Pawelec, 2020).Many genetic and non-genetic factors, such as diet and climate, affect the inflammatory response (Ter Horst et al., 2016).A deeper understanding of the factors influencing immune aging in various populations with different genetic backgrounds and lifestyles is needed.
Metabolic processes modulate immune function (Loftus and Finlay, 2016), are influenced by age, and influence the process of aging itself (Bulut et al., 2021).In addition, gut microbiota alters the body's metabolic status (Fan and Pedersen, 2021) and is also implicated in aging (Badal et al., 2020).The metabolites derived from the gut microbiota might help accelerate or slow down the aging of the immune system.This study hypothesizes that changes in the immune system in old age might be partly induced by changes in metabolism.The potential identification of metabolic pathways associated with immune aging would open the door for specific lifestyle and pharmacological interventions to slow the aging of the immune system.
Combining untargeted metabolomics with ex vivo cytokine responses against various pathogenic stimuli, the present study aimed to identify the circulating metabolites associated with age and their impact on immune responses mainly in two similar-sized healthy cohorts: one from East Africa (Tanzania) and one from Western Europe (Netherlands).Another healthy Dutch cohort was used for the validation of the findings.Understanding the similarities and differences of metabolicimmune interaction between geographically distinct populations may lead to better population-specific approaches.

Study cohorts
The cohorts were recruited as part of the Human Functional Genomics Project (see www.humanfunctionalgenomics.org) and consisted of healthy individuals without chronic illnesses or medication use, with the exception of oral contraceptives.The East African cohort consists of 323 healthy Tanzanian individuals between the ages of 18-65.Enrollment took place at the Kilimanjaro Christian Medical Center and Lucy Lameck Research Center between March and December 2017.The demographics of the cohort were previously described in detail (Temba et al., 2021).The Western European cohort includes 324 healthy individuals between the ages of 18 and 71 enrolled at Radboud University Medical Center (Radboudumc) between April 2017 and June 2018.The second Western European cohort used for validation consisted of 534 healthy participants between 18 and 75 years old, also recruited at Radboudumc, between August 2013 and December 2014.General characteristics of the European cohorts were previously published (Ter Horst et al., 2016;Mourits et al., 2020).Characteristics of the cohorts investigated in the current analyses are provided in Supplementary Table 1.
The European cohort studies were approved by the Arnhem-Nijmegen Medical Ethical Committee (NL42561.091.12 and NL58553.091.16).The cohort study in Tanzania was approved by the ethical committees of Kilimanjaro Christian Medical University College (2443) and the National Institute for Medical Research in Tanzania (NIMR/HQ/R.8a/Vol.IX/2290 and NIMR/HQ/R.8a/Vol.IX/3318).All participants signed a written informed consent prior to sample collection, and the study procedures were conducted in accordance with the Declaration of Helsinki.

Plasma metabolome measurement and analysis
Untargeted metabolomics measurements in plasma were performed by high-throughput flow injection-time-of-flight mass spectrometry (Fuhrer et al., 2011).The platform employed an Agilent (CA, USA) 6520 Series Quadrupole Time-of-flight mass spectrometer and Agilent Series 1100 LC pump coupled to a Gerstel MPS2 autosampler.The metabolites were matched and annotated with HMDB (www.hmdb.ca),KEGG (www .genome.jp/kegg/) and ChEBI (www.ebi.ac.uk/chebi/) identifiers.
The MetaboAnalyst v.4.0 platform (www.metaboanalyst.ca)was employed to perform comprehensive data analysis using the peak intensity table of annotated metabolites (Chong et al., 2019).Log transformation and Pareto scaling were applied before data analysis.Significant differences between the metabolome of different age groups were calculated with t-tests and visualized in volcano plots.Pathway enrichment analysis of identified metabolite lists was performed using the pathway analysis function of MetaboAnalyst, which combines network topology analysis and functional enrichment analysis.KEGG library was selected as the reference pathway library.

Measurement of circulating protein concentrations
Cytokine and chemokine concentrations in plasma were measured with targeted proteomics using the Olink Inflammation Panel consisting of 92 markers (Olink Biosciences, Sweden) as previously described (Koeken et al., 2020).This method employs proximity extension assay and provides relative protein quantification expressed as normalized protein expression (NPX) values (Assarsson et al., 2014).

Whole blood (WB) or peripheral blood mononuclear cell (PBMC) stimulations
Depending on the cohort, WB and/or PBMC stimulations were performed to assess cytokine production.Details of the stimuli used for each cohort are provided in Supplementary Table 2.

Telomere length analysis
DNA was isolated from whole blood samples of the European cohort, and average telomere length was determined with the Absolute Human Telomere Length Quantification qPCR Assay Kit (ScienCell, CA, USA) following the supplier's instructions.This qPCR-based assay includes primers for the telomere sequence and a single copy reference gene for normalizing genome numbers.A reference genomic DNA sample with a known telomere length is used to calculate the telomere length of the samples.

Epigenetic aging analysis
The DNA methylation profile of the European cohort was measured using the Illumina MethylationEPIC array.DNA methylation data were pre-processed in R with the Bioconductor package Minfi (Aryee et al., 2014), using the original IDAT files extracted from the HiScanSQ scanner.Quality control was performed to filter bad quality probes with a detection P-value > 0.01, cross-reactive probes, polymorphic probes, and probes in the sex chromosome.We subsequently implemented stratified quantile normalization.Epigenetic age acceleration (EAA) was generated from methylation profiles using Horvath's online DNA Methylation Age Calculator (https://dnamage.genetics.ucla.edu/).The EAA corresponded to the residuals (difference between the actual epigenetic age and the predicted value) resulting from the linear regression of each epigenetic age estimator and chronological age.EAA was correlated with candidate metabolites using Spearman's correlation.

Statistical analyses
Apart from the ones conducted on the MetaboAnalyst v.4.0 platform, statistical analyses were performed using R 3.6.1 (www.R-project.org) and GraphPad Prism 8 (GraphPad Software Inc.).Correlation heatmaps were built by the R packages 'corrplot' or 'gplots' upon Spearman's rank correlation and Benjamini-Hochberg correction using the 'corr.test'function.False discovery rate (FDR) values smaller than 0.05 after multiple testing correction were considered statistically significant.'hclust' function was used for hierarchical clustering in dendrograms.R function 'PCA' and package 'ggpubr' were utilized for the principle component analyses after scaling the data.To compare two groups for a single variable in box plots, the Mann-Whitney test or Wilcoxon matched-pairs signed-rank test was used for unpaired and paired conditions, respectively.P-values below 0.05 were considered statistically significant.

Aging leads to distinct profiles of cytokine production capacity in East Africans and Western Europeans
In order to determine the shared and distinct immune characteristics in the African and European cohorts, we investigated their ex-vivo cytokine production.We correlated the age of participants with their corresponding cytokine production capacity in response to various microbial stimuli (Supplementary Table 2) in both cohorts.Data was corrected for sex to detect generalizable patterns.In the African cohort, IFNγ production upon stimulation with E. coli, S. aureus, S. pneumoniae, and C. albicans was negatively correlated with age (Fig. 1A).Similarly, IL-6 and TNFα production also decreased with age in Africans.In contrast, IL-1β production was not significantly affected by age, except for the C. albicans-induced response.
In contrast, we could not observe significant correlations of proinflammatory cytokine production with age in the European cohort (Fig. 1B).Of note, the stimulations in the African cohort were performed with whole blood, whereas PBMCs were used for the European cohort.To test the robustness of the results, we validated them in a larger European cohort of similar demographics (Ter Horst et al., 2016), for which both types of stimulations (whole blood and PBMCs) were performed.While we observed a consistent decline in IFNγ production by PBMCs with age, similar to the African cohort, there was no effect on IL-6 or TNFα responses (Fig. 1C).Whole blood stimulations of this cohort also did not reveal a significant effect of age on IL-6 and TNFα responses (Fig. 1D).
Overall, these data suggest that a decline in T cell function with age is common in various populations, while the production of proinflammatory cytokines mainly released by innate immune cells is affected differently in African and European populations.

Aging influences the metabolomes of Africans and Europeans
To move forward with analyses comparing young and older individuals, 50 was selected as the age cut-off, although the elderly are usually defined as 65 and older.The main reason was the low number of individuals over 65 years of age in all cohorts (Fig. 1E): importantly, however, various immunological markers of aging were already significantly different between people below or above 50 years of age (Supplementary Figure 1).
We used an untargeted approach to identify the circulating metabolites influencing age-related immunological changes.1376 metabolites could be detected and annotated.Principal component analyses revealed that geographical location and genetic background exert a more substantial effect than age on the circulating metabolites (Fig. 2A-B).In this respect, while the circulating metabolomes of people over and younger people were significantly different in both cohorts (Fig. 2A), there were much greater differences between the two geographically different cohorts in both younger and older individuals (Fig. 2B).
In the European cohort, 246 metabolites were significantly more abundant, and 265 were less abundant in people over 50 compared to younger individuals (Fig. 2C).In the African cohort, the number of metabolites that were more and less abundant in older individuals were 132 and 141, respectively.A complete list of these metabolites and the relevant metrics are provided in Supplementary File 1. 61 of the metabolites with higher circulating concentrations in individuals over were shared between the cohorts, while 81 metabolites were commonly less abundant (Supplementary Table 3).The top 5 commonly abundant metabolites with the highest fold changes were quinic acid, 15-Keto-13,14-dihydroprostaglandin A2, hexonic acid, D-Glycero-D-galactoheptitol, and Ferulic acid 4-sulfate.The top 5 commonly less abundant metabolites in older individuals were dehydroepiandrosterone sulfate, 2-oxoglutarate (2-), niazicinin A, curcumin, and androsterone sulfate.These data indicate that metabolomes of different demographics differ in young and old individuals, but they also display significant common age-related changes.

Age-related decline in steroid hormone precursors is associated with increased systemic inflammation and lower cytokine response in both Africans and Europeans
Next, we performed pathway analysis using the 61 more and 81 less abundant metabolites in older individuals shared in both cohorts.Additionally, using the information provided by the Human Metabolome Database, we could identify whether these metabolites are endogenous or exogenous (e.g., food-derived).
While most of the 61 upregulated metabolites in the older individuals were endogenous or both endogenous and food-derived, 16.4% were strictly derived from food or beverages (Fig. 3A).The significantly overrepresented pathways in this metabolite list were caffeine metabolism and the pentose phosphate pathway (PPP).PPP is known to shape immune responses (Jansen et al., 2021).On the other hand, the 81 metabolites less abundant in older individuals included 9.9% exclusively food-derived metabolites (Fig. 3B).Pathway analysis revealed that the steroid hormone biosynthesis pathway was overrepresented among the 81 metabolites with decreased concentrations in older individuals.The declining metabolites belonging to that pathway included dehydroepiandrosterone (DHEA), DHEA sulfate (DHEA-S), and androsterone glucuronide (ADT-G).Their concentrations were much lower in people older than 50 than younger people in both cohorts (Fig. 3C).ADT-G was also lower in older individuals of the European validation cohort, while DHEA was not significantly different, and DHEA-S measurements were not available (Supplementary Figure 2).
To investigate whether the age-related decline in the metabolites involved in steroid hormone biosynthesis might have immunological implications, we correlated the concentrations of DHEA and DHEA-S with circulating inflammatory markers and cytokine responses upon microbial stimulation after correcting for the effect of age.DHEA-S, the dominant form of DHEA found in circulation, and ADT-G concentrations were significantly negatively correlated with serum IL-6 concentration in the African cohort (Fig. 3D).In contrast, DHEA-S was significantly positively correlated with TNFα and IFNγ production upon stimulation with S. pneumoniae.Similarly, ADT-G levels positively correlated with TNFα secretion after S. pneumoniae or S. aureus stimulation.In the European cohort, no association was found between these metabolites and ex-vivo cytokine production.However, DHEA-S significantly negatively correlated to circulating IL-6, IL-8, and IL-18 concentrations (Supplementary Figure 3).ADT-G was also associated with lower IL-6 circulating concentrations.Collectively, these data imply that the decline of steroid hormone precursors might contribute to age-related low-grade systemic inflammation on the one hand and defective cytokine response against pathogens on the other.This seems to be a general feature in populations of both African and European ancestry.

Hippurate and 2-phenylacetamide are aging-related metabolites associated with low cytokine production in both cohorts
We next tested if the 61 metabolites with higher concentration in people over 50 years of age in both populations were associated with the ex-vivo cytokine production capacity.An overall negative correlation pattern between these metabolites and IFNγ, IL-6, and especially TNFα production was identified in the African cohort after correcting for the effect of age (Supplementary Figure 4 and Fig. 4A).Among those, hippurate and 2-phenylacetamide were also negatively correlated with IL-6 and IL-1β production in the European cohort (Fig. 4B).These two metabolites were also more abundant in individuals over 50 in the European validation cohort (Supplementary Figure 2).
Although the observations remained broadly similar for innate cytokines when men and women were analyzed separately, some intriguing sex differences were observed for IFNγ production in the African cohort (Supplementary Figure 5 A).Contrary to the general trend, age-related metabolites were positively correlated with IFNγ production against viral stimulus poly(I:C) in women, but did not have a significant association in men.Moreover, a strong negative correlation was observed with the IFNγ response to C. burnetii in women, with no association in men.In the European cohort, negative correlations with  hippurate and 2-phenylacetamide could only be seen in women when sex-specific assessment was performed (Supplementary Figure 6 C).
The 81 metabolites with lower concentrations in older individuals were positively correlated with cytokine production in the African cohort (Supplementary Figure 7).Among those, ximenoylacetone, KB2, and 14,16-nonacosanedione, all food-derived metabolites, were strongly correlated with ex-vivo IFNγ, IL-6, and TNFα production (Fig. 4C).However, such associations were not found in the European cohort (Fig. 4D), arguing for population-specific consumption or biological effects.Only ximenoylacetone concentrations among these metabolites negatively correlated with S. aureus-induced IFNγ production.Of note, circulating concentrations of all three metabolites were much lower in the European cohort compared with the African counterparts (Fig. 4E), which could explain the lack of association with cytokine production capacity in Europeans.14,16-nonacosanedione was also found in lower levels in the older individuals of the European validation cohort, although KB2 and ximenoylacetone were not among the detected metabolites (Supplementary Figure 2).
These metabolites also differed in their association with the lymphoid response to poly(I:C) in women and men with African ancestry (Supplementary Figure 5B).There was no correlation in men, while the metabolites were negatively correlated with IFNγ production in women.Moreover, a positive association of the metabolites with C. burnetiiinduced IFNγ production was only observed in women.
Altogether, these findings suggest that the changing metabolic landscape as people age is linked to the modifications in immune function with some sex differences, particularly for antiviral response.Hippurate and 2-phenylacetamide emerge as metabolites possibly contributing to immune dysfunction in old age in both populations.On the other hand, the food-derived metabolites ximenoylacetone, KB2, and 14,16-nonacosanedione might have protective effects against the aging of immune cells, specifically in the African cohort in which they are at higher concentrations.

Hippurate and 2-phenylacetamide are associated with faster epigenetic aging and shorter telomeres in males
Epigenetic clocks based on DNA methylation, such as Horvath's clock, and telomere length have been strongly associated with immune aging.These measures correlate poorly; however, each has been linked to age-related disorders (Jansen et al., 2021).Epigenetic age acceleration (EAA) analysis based on DNA methylation of whole blood was performed for the European cohort.Also, average telomere length was assessed from whole blood DNA samples for 98 participants, half of whom were female and half male.Because of the consistent associations of hippurate and 2-phenylacetamide with immune function, we assessed their relationship with either of the aging clocks.
In both analyses, another sex difference in the associations of these two metabolites was observed.In males but not females, both hippurate and 2-phenylacetamide concentrations showed a positive correlation with EAA (p = 0.0321 and 0.0339, respectively) (Fig. 5A).Similarly, the circulating concentrations of these two metabolites were negatively correlated with telomere length in males but not females (Fig. 5B).Of note, circulating concentrations of both metabolites were not significantly different between males and females (Fig. 5C).These associations support the possibility that hippurate and 2-phenylacetamide might contribute to aging, while the sex-specific findings should be confirmed with larger sample sizes.

Lifestyle differences lead to distinct metabolic profiles in older individuals, possibly contributing to a different immune response
In order to identify the metabolic profile unique to the older individuals of the African cohort and to determine if that can explain their lower innate immune response compared to the young individuals, we compared the metabolomes of the 24 and 32 participants over 50 years of age in the European and African cohorts, respectively.370 metabolites were significantly higher in the Africans, while 353 were higher in the Europeans (Fig. 6A).When we assessed the sources of the top metabolites with the biggest fold change in either direction, 10% and 32% were purely endogenous, while 10 and 23% of metabolites were strictly diet-derived in Africans and Europeans, respectively (Fig. 6B).A total of 10% of the metabolites higher in Europeans were related to medication use and environmental contaminants.
Pathway analysis of the 370 metabolites more abundant in older individuals of the African cohort revealed that many pathways related to amino acid metabolism, including glycine, serine, threonine, alanine, aspartate, glutamate, arginine, and histidine, were overrepresented (Fig. 6C).Primary bile acid biosynthesis, linoleic acid metabolism, and glutathione metabolism were also overrepresented.Among the metabolites more abundant in older individuals of the European cohort, only phenylalanine, tyrosine, and tryptophan biosynthesis was significantly overrepresented.
When the top 50 differentially found metabolites in Africans were correlated with ex vivo cytokine response, traumatic acid, d-erythro-lgalacto-nonulose, and epinephrine glucuronide showed a robust negative correlation with IFNγ, IL-6, and TNFα production (Supplementary Figure 8 and Fig. 6D).Collectively, these data imply distinct regulation of amino acid metabolism in the two populations, reveal the impact of diet and environmental factors on the metabolome of different groups of older individuals, and identify metabolites that might contribute to the modulation of innate immune responses in the African cohort.

Discussion
Aging impacts the function of various organs and systems in the body, determines longevity, and strongly influences morbidity and quality of life.Understanding the factors that influence aging can thus have significant implications for improving the quality of life and reducing the burden of disease at the societal level.Proper immune system function is crucial for the homeostasis of the human body, and aging of the immune system influences morbidity and disease complications in humans (Alpert et al., 2019;Sayed et al., 2021).In the present study, we investigated the age-related changes in the function of the immune system in East African and Western European populations.
We identified a general decrease in T-cell function with age in both populations, similar to other studies (Goronzy and Weyand, 2017), whereas innate immune responses declined with aging only in the African population but not Europeans.One main aim of the study was to assess the circulating metabolites associated with decreased immune function in older individuals.Although the circulating metabolome differed between African and European populations, we identified several metabolites that increased with age and were associated with decreased immune function in both populations, as well as metabolites associated with potent immune responses but were deficient in individuals of older age.
Circulating metabolites have an important impact on the immune response (Koeken et al., 2022), and we aimed to identify metabolites that change with age while influencing immune responses.We identified 61 common metabolites with higher plasma concentrations in older individuals in both populations and pathway analysis revealed that pentose phosphate pathway (PPP) metabolites were enriched among them.By providing redox equivalents and nucleotide precursors, PPP plays an essential role in feeding the demands of immune cells (Nagy and Haschemi, 2015).The upregulated pathway might reflect the increased demand against high infectious burden in old age or mirror chronic low-grade inflammation (Franceschi et al., 2018).
Metabolites of the steroid hormone biosynthesis pathway were enriched among the 81 metabolites with reduced circulating concentrations in the older individuals of African and European ancestry.Agerelated decrease in steroid hormone biosynthesis is highly likely to have immunological consequences, considering the impact of steroid  hormones on immune cell function (Hoffmann et al., 2023).Steroid hormones like estradiol, testosterone, and DHEA regulate both innate and adaptive immune responses and are responsible for many immunological differences between males and females (Klein and Flanagan, 2016).DHEA and its sulfated form DHEA-S are the most abundant circulating steroid hormones in humans, primarily produced by the adrenal cortex but also synthesized in the brain (Powrie and Smith, 2018;Webb et al., 2006).Prior studies have linked low DHEA levels to aging, various age-related comorbidities, and neurodegenerative disorders (Ravaglia et al., 2002).As a result, DHEA is a popular anti-aging supplement, available over-the-counter and online, although sufficient clinical trial data to show its effectiveness is missing.Experimental data O. Bulut et al. have shown anti-inflammatory effects, such as inhibiting leukocyte migration and reducing neural inflammation (Alexaki et al., 2018;Ziogas et al., 2020).A relatively small study in elderly men receiving 20 days of DHEA supplementation reported higher monocyte numbers and lymphocyte response to mitogens than before treatment (Khorram et al., 1997).However, a 3-week treatment with DHEA in post-menopausal women reduced T helper cell numbers and mitogenic response, but enhanced NK cell cytotoxicity (Casson et al., 1993).
Our study suggests DHEA and DHEA-S have complex regulatory effects: while reducing systemic inflammation in both cohorts, they also seem to support the homeostasis of cytokine production in the Tanzanian cohort, maintaining an effective response against microbial stimuli.These data may thus support an approach in which DHEA supplementation might be beneficial as an anti-aging supplement, maintaining the tonus of the immune system while inhibiting systemic inflammation.However, large randomized controlled trials assessing the immunological effects of DHEA supplementation are needed to validate this assumption.
We also identified several metabolites with higher circulating concentrations in the older individuals in both cohorts, including tyrosol-4sulfate, propionyl carnitine, hippuric acid (hippurate), and 2-phenylacetamide, which were associated with lower cytokine production in the African cohort.These data suggest that higher concentrations of these metabolites observed in the circulation of older individuals might contribute to their declined innate immune responses.The inverse associations with cytokine production were replicated for hippurate and 2phenylacetamide in the European cohort, although sex-specific analysis revealed this only in females.Interestingly, the concentrations of these metabolites were higher in the Africans than in the Europeans, which might be one of the reasons why Africans exhibit more decline in innate immunity as they age.Lastly, both metabolites were associated with higher EAA and lower telomere length in European males, supporting the hypothesis that they are involved in aging.However, separating the sexes diminished the statistical power of the analyses, so the observed sex discrepancies should be confirmed with larger sample sizes.
Hippurate levels were previously reported to increase in healthy aging but not in frailty (De Simone et al., 2021), and it has been associated with metabolic health (Brial et al., 2021).Its concentrations are elevated upon consumption of polyphenol-rich dietary sources such as wine, fruit juices, coffee, and tea.The intestinal microbiota is critical in hippurate synthesis from these sources (Lees et al., 2013).Our results suggest potential anti-inflammatory effects of hippurate.However, this may be associated with defective responsiveness of immune cells to microbial challenges, and it remains to be studied whether this may be disadvantageous for susceptibility to infections.There is not much prior biological research on 2-phenylacetamide, which can also be derived from plant-based food sources, but it was shown to be elevated in older mice (Osada et al., 2003).More research is needed on this metabolite associated with decreased immune function along with genetic and epigenetic aging in our study.The gut-derived source for these metabolites may support the hypothesis that manipulating diet and microbiota to regulate plasma metabolite composition and aging-related processes could be an attractive strategy.
Among the metabolites with lower plasma concentrations in both European and African older individuals, KB-2, ximenoylacetone and 14,16-nonacosanedione, all food-derived metabolites, were associated with higher adaptive and innate immune responses in the African cohort.Interestingly, this association was not observed in the European cohort, probably due to the lower concentrations of these metabolites in the Europeans, lacking thus the variability necessary to identify the stimulatory effects.The higher concentration of these metabolites in Africans is likely a diet-induced difference.For instance, KB-2 is a flavone derived from tropical breadfruit cultivated in Tanzania (Mgembe and Maerere, 2007).These diet-derived metabolites might exert an immunoprotective effect by maintaining the homeostasis of the immune responses.Temba et al. also previously reported immunomodulatory effects of food-derived metabolites in the African cohort, with rural residents showing a significant abundance of plant-based polyphenols, such as apigenin, known to have anti-inflammatory effects (Temba et al., 2021).
We observed intriguing sex differences in how age-associated metabolites relate to antiviral lymphoid cytokine responses.IFNγ production against the viral ligand poly(I:C) was positively correlated with tyrosol-4-sulfate, propionyl carnitine, and epinephrine glucuronide levels in women, contrary to the negative correlation with responses against all other pathogens.Hippurate and 2-phenylacetamide also showed a similar trend, but were not statistically significant.Men and women differ in their susceptibility to viral infections and the strength of antiviral responses induced by vaccines (Klein and Flanagan, 2016).Overall, women mount a stronger antiviral response after vaccination.The human X chromosome contains many immune-related genes including TLRs 7 and 8, which are crucial for viral recognition (Fish, 2008), and both genes can escape X chromosome inactivation (Souyris et al., 2018;Youness et al., 2023), likely contributing to the more robust antiviral response in women.The metabolites associated with immune aging might support a stronger antiviral response in women, potentially leading to both lower susceptibility to initial infection, but more severe disease in later stages of infection.Furthermore, ximenoylacetone and 14,16-nonacosanedione were negatively correlated with poly(I: C)-induced IFNγ production in women.These metabolites with seemingly restorative effects on immune responses to other pathogens might also help dim overactive antiviral processes in women and maintain immune homeostasis.
Lastly, the circulating plasma metabolomes differed significantly in individuals over 50 years of age from Tanzania and the Netherlands.The metabolic differences highlighted the importance of diet and lifestyle.Notably, more metabolites that derived from medication use and environmental contaminants were found in the European older individuals compared to Africans.Three out of 50 most abundant metabolites in older Africans strongly correlated with lower IFNγ, IL-6, and TNFα production.Traumatic acid is a regenerative plant hormone that is produced upon tissue damage (Farmer, 1994), D-erythro-L-galacto-nonulose is derived from avocado (Sephton and Richtmyer, 1963), and epinephrine glucuronide is a natural derivative of epinephrine converted in the liver (Axelrod, 1959).These metabolites might be involved in the declining innate immune response as people age in Tanzania.
The cohorts analyzed in this study were previously recruited as part of the Human Functional Genomics Project (HFGP).A limitation of the present study is that aging-related analyses were not the initial focus of HFGP, and the age distribution of the cohorts is not ideal for exploring aging processes.Another limitation, due to the design of the cohorts, was that stimulations were performed only in whole blood stimulation assay for the Tanzanian cohort, while cytokine stimulation assays in either whole blood or PBMCs were used in two Dutch cohorts.Nevertheless, the analyses presented here provide consistent and valuable information regarding the age-related differences in metabolome and their impact on the homeostasis of immune responses.An important strength of our study is the investigation and comparison of metabolicimmune interactions in cohorts of African and European ancestry, which allowed us to identify potential general mechanisms of immune modulation in human populations.
In conclusion, this study identified several common and specific agerelated metabolites in people with African and European ancestry that are likely to mediate the changes in the immune system as they age.Although correlations of metabolite concentrations with immunological parameters were statistically robust, future in vitro and in vivo functional validation studies are necessary to establish whether these metabolites causally influence immune response.Studying diverse populations with advanced omics and immunological methods is necessary to improve our understanding of the shared mechanisms of aging and the particular biological processes in each population linked to genetics, environmental factors, and diet.

Fig. 1 .
Fig. 1.Aging leads to distinct profiles of cytokine production capacity in Africans and Europeans.Correlation of age with ex-vivo cytokine production upon stimulation in the (A) African and (B-D) European cohorts, and (E) age distribution of each cohort.All data was controlled for sex.Red represents a positive correlation with age in the heatmaps, while blue represents a negative correlation.Multiple testing correction was performed.Solid lines depict the best fit in correlation dot plots, and dashed lines depict the 95% confidence interval.r: Spearman's correlation coefficient.*p<0.05,**p<0.01,***p<0.001.

Fig. 2 .
Fig. 2. Aging influences the distinct metabolomes of Africans and Europeans.(A-B) Principal component analysis of metabolome data and (C) volcano plots depicting the differential abundance of metabolites in different age groups in the African and European cohorts.In the PCA plots, significance stars reflect the difference between the two groups' distribution along each PC axis.In the volcano plots, significantly different metabolites are depicted in pink.PC: principal component, ns: not significant, FC: fold change.*p<0.05,**p<0.01,****p<0.0001.

Fig. 4 .
Fig. 4. Hippurate and 2-phenylacetamide are aging-related metabolites associated with low cytokine production in both cohorts.Correlation of ex-vivo cytokine production with levels of selected metabolites commonly (A-B) more abundant or (C-D) less abundant in people over 50 years of age in the African and European cohorts.(E) Levels of ximenoylacetone, KB2, and 14,16-Nonacosanedione in the two cohorts.Only the metabolites with the strongest correlations are shown in the heatmaps.Red depicts a positive correlation, while blue depicts a negative correlation.Multiple testing correction was performed.Solid lines depict the best fit in dot plots, and dotted lines depict the 95% confidence interval.r: Spearman's correlation coefficient.*p<0.05,**p<0.01,***p<0.001,****p<0.0001.

O
.Bulut et al.

Fig. 5 .
Fig. 5. Hippurate and 2-phenylacetamide are associated with faster epigenetic aging and shorter telomeres in males.Sex-dependent associations of hippurate and 2phenylacetamide on (A) epigenetic age acceleration (B) average telomere length of the cells in whole blood in the European cohort.(C) Hippurate and 2-phenylacetamide levels in males and females of the European cohort.In correlation plots, solid lines depict the best fit and dashed lines depict the 95% confidence interval.In violin plots, dashed lines depict the median and quartiles.EAA: epigenetic aging acceleration, r: Spearman's correlation coefficient.

Fig. 6 .
Fig. 6.Lifestyle differences might lead to distinct metabolic profiles in older individuals, possibly contributing to a different immune response.(A) Volcano plot depicting the differential abundance of metabolites in individuals over 50 years of age in two cohorts, (B) sources of the 50 metabolites with the biggest fold change that are higher in Africans or Europeans, (C) pathway enrichment analyses of metabolites significantly higher in Africans or Europeans, (D) correlation of ex-vivo cytokine production in the African cohort with levels of selected metabolites that are overrepresented in African individuals over 50.In pathway analysis graphs, node colors reflect the p values, and diameters reflect the pathway impact values.Only the statistically significant pathways are labeled.In the heatmaps, only the metabolites with the strongest correlations are shown.Red shows a positive correlation, while blue shows a negative correlation.Multiple testing correction was performed for correlations.*p<0.05,**p<0.01,***p<0.001.