Biomarkers of geroprotection and cardiovascular health: An overview of omics studies and established clinical biomarkers in the context of diet

Abstract

The slowdown, inhibition, or reversal of age-related decline (as a composite of disease, dysfunction, and, ultimately, death) by diet or natural compounds can be defined as dietary geroprotection. While there is no single reliable biomarker to judge the effects of dietary geroprotection, biomarker signatures based on omics (epigenetics, gene expression, microbiome composition) are promising candidates. Recently, omic biomarkers started to supplement established clinical ones such as lipid profiles and inflammatory cytokines. In this review, we focus on human data. We first summarize the current take on genetic biomarkers based on epidemiological studies. However, most of the remaining biomarkers that we describe, whether omics-based or clinical, are related to intervention studies. Then, because of their promising potential in the context of dietary geroprotection, we focus on the effects of berry-based interventions, which up to now have been mostly described employing clinical markers. We provide an aggregation and tabulation of all the recent systematic reviews and meta-analyses that we could find related to this topic. Finally, we present evidence for the importance of the “nutribiography,” that is, the influence that an individual’s history of diet and natural compound consumption can have on the effects of dietary geroprotection.

Supplemental data for this article is available online at https://doi.org/10.1080/10408398.2021.1975638

Graphical Abstract

Supplemental data for this article is available online at https://doi.org/10.1080/10408398.2021.1975638

Keywords:

• Geroprotector
• biological age
• healthy aging
• cognitive aging
• inflammation

Diet was shown to have a measurable impact on both omics and clinical biomarkers. Inflammation, lipid and glucose metabolism are among the main molecular processes impacted, and a positive effect of diet or natural compounds on these and other processes can result in beneficial effects on health and survival, that is, dietary geroprotection. Effects are indicated by arrows, suggesting causal influences, though it is important to consider that the entities and relationships shown here are only part of a complex meshwork of direct and indirect effects. Dashed arrows indicate evolutionary effects of diet on the genome, and effects of diet on aging-associated molecular processes mediated by the genome.

Introduction

The quest for the identification of biomarkers of aging is still in its early stages. Many different candidates have been proposed, and in many cases, the best results were obtained with composite biomarkers (Shamir 2015). Some candidates have become quite popular, such as epigenetic clocks (Ryan 2020; Bell et al. 2019; Horvath 2013), or phenotypic age (Liu et al. 2018) and Young.AI (Mamoshina et al. 2018a), which are biomarker signatures based on clinical (laboratory) marker measurements. A biomarker of aging should not just be able to predict future health and survival better than chronological age (Fuellen et al. 2019; Baker and Sprott 1988); it should also be sensitive to the effects of protective interventions, including healthy lifestyles and diet (Galkin et al. 2020a).

There is ample evidence that non-pharmaceutical interventions may delay aging. The adoption of an exercise program or of healthy eating patterns, such as a Mediterranean, MediterrAsian, or a calorie restricted diet, have all been reported to ameliorate age-related decline (Belsky et al. 2017; Gensous et al. 2020; Rebelo-Marques et al. 2018; Pallauf et al. 2013). Some diets and the consumption of some natural compounds have been linked to improved health and survival in humans and animals alike. In this review, we prioritize human studies, and only mention animal data to highlight similarities and differences with respect to humans. However, the impact of the consumption of many natural compounds, such as the polyphenols fisetin or resveratrol, as described in some but not all studies, on the health and survival of laboratory animals (Bhullar and Hubbard 2015; Yousefzadeh et al. 2018) has not yet been validated in human interventions. Many more issues are still in need of further research. For example, there is an ongoing debate if and to what extend polyphenols such as resveratrol act as calorie restriction mimetics, thus mediating life-prolonging properties (Pallauf et al. 2016). It is unclear if health-promoting properties of polyphenol-rich diets can be replaced by purified compounds as dietary supplements (Egert and Rimbach 2011) and if and to what extend data in model organisms can be translated to humans.

Natural compounds are compounds produced by living organisms. These may be consumed through a diet enriched in appropriate foods, or by supplementation of purified extracts or synthesized compounds. We define as geroprotective any diet or natural compound with a demonstrated effect in delaying, inhibiting, or even reversing, aging. Due to their simple implementation and safety record, diets and their natural constituents are of ever-increasing importance to maintain health in old age and to extend (healthy) lifespan; we call this approach “dietary geroprotection.”

Among the frequently studied dietary geroprotectors are the omega-3 fatty acids found in oily fish and nuts, vitamin D that is also found in oily fish but largely synthesized endogenously in response to sunlight exposure (Ferri et al. 2019), dietary fiber found in plant-based foods, and polyphenols, that is, plant secondary metabolites whose dietary sources are, for example, herbal teas, cocoa, coffee, vegetables and fruits including berries (Fraga et al. 2019; Shahidi and Ambigaipalan 2018; Slavin 2013). The identification of biomarkers that enable us to evaluate geroprotective effects in humans is of utmost necessity. Dietary geroprotection and biomarkers supporting its successful implementation are important not least because we are facing a “gray” wave of increasing numbers of old and very old citizens that could pose economic and other challenges proportional to their health status.

This narrative review deals with omics and clinical biomarkers that could be used to evaluate the geroprotective effects of diets and natural compounds. The first section will focus on genetic differences among individuals that can make them more or less sensitive to dietary geroprotection, while the following sections will focus on candidates for epigenetic, transcriptomic, microbiome-related and clinical biomarkers to judge successful dietary geroprotection. We have chosen to focus on omics biomarkers (specifically those related to gene expression and its regulation) because of the recent success of epigenetic clocks, and also because they offer the possibility to identify known and new molecular mechanisms involved in the aging process. In principle, implicating known aging-associated processes in the success of a dietary geroprotection can lessen or even alleviate the need for a placebo control group in a dietary trial, assuming that the up- or downregulation of specific molecular processes or pathways (such as aging-associated inflammation and cellular senescence; the more specific the better (Campisi and Robert 2014; Franceschi et al. 2018; Ou, Yang, and Liu 2020; Santoro et al. 2020a)) cannot easily be accomplished by some mind-body interaction that may give rise to a placebo effect.

Box 1: Placebo effects and biomarker sources in dietary geroprotection studies.

As described in the main text, the need for a placebo group may be lessened or even alleviated in a dietary trial, if an up- or downregulation of molecular processes or pathways can be demonstrated that is so specific to the aging-associated intervention that it is not plausible that the psychological expectation of benefits is the main driver of these effects. Handling such potential placebo effects is important in dietary geroprotection; at least, many food items cannot be faked easily, so a placebo intervention is not always feasible. This is specifically true for berry-based interventions, to which we devoted an extra section and a table compiling all recent systematic reviews and meta-analyses that we could identify. However, most of these publications make exclusive use of clinical biomarkers, and they do not attempt to pinpoint specific molecular processes or underlying signal transduction pathways of effect. Then again, at the same time, they usually refer to studies that include placebo groups based on fake berry powder, extract, or shakes. In any case, by studying the impact of an intervention using different omics modalities (e.g., epigenetics and transcriptomics), its effects can be understood even better and in a more detailed fashion.

In the context of dietary interventions, most studies were carried out using Peripheral Blood Mononuclear Cells (PBMC) or venous whole blood as biomarker sources since the blood draw needed for PBMC, and venous blood sampling in general is comparatively noninvasive and inexpensive. In fact, almost all biomarkers discussed here, except for the microbiomic ones, are based on a blood draw. However, the use of capillary blood (e.g., dried blood spots) is expected to gradually replace blood draws, given the ever-increasing sensitivity of laboratory measurements. Another promising development is swabs of various kinds (oral, nasal, skin, etc.) (Fuellen et al. 2020).

Genetics of human aging: From epidemiological studies to biomarkers

Health and survival are partially determined by genetics. And indeed, the maximum lifespans of animals are predominantly species-specific and can be predicted quite accurately from the genomes of those species (Mayne et al. 2019). Furthermore, the existence of long-lived mutants, such as the Ames dwarf mouse or the daf-2-deficient C. elegans may suggest that perhaps, similar mutations in human beings could lead to a longer-lived phenotype (Kuningas et al. 2008). Then again, a recent family tree study, including hundreds of millions of people, found evidence of a strong influence of assortative mating in heritability estimates and it suggested that the true genetic heritability of longevity is below 10% (Graham Ruby et al. 2018). Overall health can be assumed to have a stronger genetic basis, but we are not aware of studies of, e.g., healthspan, that considered assortative mating as a confounder. Nevertheless, various human mortality studies show that longevity, here defined as surviving to the top 10% of a given birth cohort, is heritable (van den Berg et al. 2019; Deelen et al. 2019). Moreover, both long-lived individuals and their first and second-degree relatives show lower morbidity, in addition to lifelong reduced mortality (van den Berg et al. 2019). In any case, genetic factors do not have a dominant influence on the healthspan and lifespan of the average human individual, suggesting that there is a lot of space for improvement by tackling the influence factors that can be modified.

Cognition is of specific interest in this context (see also Box 1), because there is not just a genetic influence on cognitive abilities and a plethora of studies demonstrate the effects of dietary geroprotection on cognition. What is more, cognitive abilities influence the choice of a healthy diet, demonstrated by the genetic influence on food choice and the effects of cognitive decline on the quality of the diet that is consumed.

Human longevity genes

A wide variety of candidate longevity genes have been identified by genome wide association studies (GWAS), but many of them could not be replicated (Pilling et al. 2017; Deelen et al. 2019). One common approach to investigate longevity is the use of extreme age groups (e.g. contrasting centenarians to young individuals), another one is using parents’ attained age as a proxy for longevity for an individual (Pilling et al. 2017; Van der Auwera et al. 2021). Replication failure could be explained by different definitions of the longevity phenotype, or different gene versus environment interactions in different cohorts (Franceschi et al. 2020). Another reason could be sexual dimorphisms influencing the aging process, masking the potential association between a gene and sex-specific aging outcomes. In fact, Zeng et al. identified genes that seem to be associated with longevity only in a single sex (Zeng et al. 2018). Notably, the genes with highest reproducibility and effect size among different studies, APOE and FOXO3, are both involved in nutrient sensing (Revelas et al. 2018).

APOE, longevity and diet

The APOE gene encodes a protein involved in lipid metabolism and transport, including fat-soluble vitamins like vitamin D and E, as well as cholesterol, through the lymphatic and vascular systems (Mahley, Weisgraber, and Huang 2009; Huebbe and Rimbach 2017). It is also involved in cholesterol transport to the brain and the resolution of inflammation (Yin et al. 2019; Liao, Yoon, and Kim 2017). The most common allele of this gene is APOE3, while the other two alleles, APOE2 and APOE4, have population-specific frequencies that are usually smaller (Yassine and Finch 2020). Overall, the APOE2 allele appears to be protective against age related diseases, while the APOE4 allele shows the opposite effect (Shinohara et al. 2020).

We here provide some more details of what is known about the molecular implications of some APOE variants on lipid metabolism and how these are thought to explain health-related outcomes. Since many dietary geroprotectors modulate lipid profiles, we can expect that the effects of at least some nutritional interventions on, e.g., longevity, are influenced by APOE genetics. Confirming this expectation, Griffin et al. (2018) have recently shown, after a 24-week long dietary intervention involving 389 subjects, that APOE4 carriers (in comparison to homozygous E3 carriers) experience a more significant decrease in total cholesterol and APOB when saturated fatty acids are replaced with low glycemic index carbohydrates (Griffin et al. 2018). Moreover, Yassine and Finch (2020) have recently summarized the available evidence on the impact of the APOE4 genotype on the response to dietary interventions and nutritional supplementation. From their analysis, APOE4 carriers were shown to be more responsive to diet and lifestyle treatments and more sensitive to omega-3 deficiencies (Yassine and Finch 2020). Nevertheless, Minihane et al. have reported that a 6-week fish-oil intervention led to adverse effects on the lipid profile of APOE4 carriers, with a significant increase in total cholesterol and a trend toward a decrease in HDL cholesterol. At the same time, subjects that were homozygous for APOE3 benefited from the supplementation of fish oils, with a significant reduction in fasting and postprandial triglycerides and a decrease in percent medium size LDL particles (LDL-3) (Minihane et al. 2000). Thus, APOE4 carriers, despite their sensitivity to omega-3 deficiencies, may not always benefit from fish oil supplementation. Furthermore, Abdullah et al. (2021), after reviewing the available literature, have recently shown APOE to be among the genes involved in the inter-individual variability of circulating cholesterol concentrations in response to diet (Abdullah et al. 2021). Some more evidence of the impact of APOE genotype status on the response to diet comes from transgenic targeted gene replacement mice. Huebbe, Jofre-Monseny, and Rimbach (2009) have shown that the transport of alpha-tocopherol to the lungs is affected by ApoE genotype, Boesch et al. (2009) have shown that quercetin has a smaller effect on inflammation and blood lipids in transgenic ApoE4 mice compared to their transgenic ApoE3 counterparts, and Graeser et al. (2011) have demonstrated that Nrf2-related gene expression is decreased in transgenic ApoE4 mice when compared their transgenic ApoE3 counterparts. Mouse evidence in favor of nutrition-related interventions for the amelioration of health in ApoE4 mice was provided by Wiesmann et al. (2016) and Yanckello et al. (2021), who have shown ApoE4 mice to be sensitive to the protective effect of a multi-nutrient supplement (Fortasyn) and to the prebiotic fiber inulin, respectively.

Investigating the effects of APOE genetics on health and survival, it was found that homozygosity for the APOE2 allele is correlated with decreased risk of dementia compared to the other genotypes (Li et al. 2020), with longevity (Shinohara et al. 2020), but also with an increased risk of atherosclerosis (Javvaji, Can, and Sharma 2021), while both homozygosity and heterozygosity for the APOE4 allele are associated with an increased risk (Marais 2019). In particular, the concentration of LDL cholesterol decreases when one APOE2 allele is present, presumably due to reduced binding of specific lipid transporters (chylomicron remnants) to APOE receptors and the compensatory upregulation of LDL receptors in liver cells, and reduced APOE-receptor-dependent conversion of LDL precursors into LDL. In the presence of two APOE2 alleles, however, the reduction of the APOE receptor activity is so large that atherogenic lipid transporters can accumulate in the presence of cofactors such as obesity, leading to atherosclerosis. By contrast, both homo- and heterozygous APOE4 carriers show higher cholesterol levels and other abnormalities that could explain their propensity to disease: they have lower serum HDL, higher inflammatory marker levels, and significantly different microbiomes compared to APOE3 and APOE2 carriers (Masana et al. 2017; Morrill and Gibas 2019; Patrick 2019; Yassine and Finch 2020), similar to transgenic lab mice that carry the human APOE4 allele (Boesch-Saadatmandi et al. 2009; Wiesmann et al. 2016). Overall, individuals that are homozygous for APOE4 could be responsive to dietary geroprotections that specifically improve lipid and inflammation markers. Thus, the APOE4 genotype could be used as a selection criterion for the identification of individuals that would benefit particularly from this type of interventions (Farmer, Johnson, and Hanson 2019; Yassine and Finch 2020), which warrants further investigations.

FOXO3, longevity and diet

The other robustly reproducible longevity gene, FOXO3, encodes a transcription factor involved in numerous pathways related to longevity, such as autophagy, inflammation, and cellular replication. It is one of the most important players in the geroprotective effects of caloric restriction in mice (Shimokawa et al. 2015). Various SNPs associate higher transcriptional levels of FOXO3 and longevity (Sanese et al. 2019; Flachsbart et al. 2017). Flachsbart et al. also confirmed that two FOXO3 SNPs (rs4946935 (A/G) and rs12206094 (C/T)) are associated with longevity in multiple cohorts. The authors showed that both variants are related to higher FOXO3 transcriptional levels in expression Quantitative Trait Loci (eQTL, Nica and Dermitzakis 2013) databases, and they investigated these variants in vitro. One interesting finding of their study is that cells transfected with the second variant (rs12206094 (C/T)) feature stronger EGCG-mediated FOXO3 overexpression (epigallocatechin gallate (EGCG) is a polyphenol contained in green tea). Studies in rodents shed further light on the in vivo role of this transcription factor, as its knockout causes increased incidence of atherosclerosis, faster exhaustion of the hematopoietic stem cell pool and higher inflammatory markers, while its overexpression is correlated with smaller cardiomyocyte size and enhanced autophagy (Morris et al. 2015). A couple of mouse and rat studies have shown that many natural plant compounds such as resveratrol, withaferin A and ECGC upregulate FOXO3 transcriptional levels in vivo (Bale, Pulivendala, and Godugu 2018; Franco et al. 2014; Shankar, Marsh, and Srivastava 2013; Wang et al. 2017; Wu et al. 2017). Further, Pallauf et al. (2017) have also shown, in vitro, that flavonoids, but not antioxidant vitamins such as ascorbic acid and alpha tocopherol can stimulate the activities of transcription factors belonging to the Foxo family, but also Nrf2 and PPARγ, two other longevity-associated transcription factors (Pallauf et al. 2017). Thus, in the context of a geroprotective intervention based on polyphenols, the FOXO3 genotype could play a significant role.

MTHFR and diet

A gene that has a definitive impact on the response to dietary intervention is the MTHFR gene. Unlike FOXO3 and APOE, it has yet not turned up in longevity GWAS studies. This gene encodes a flavoprotein that catalyzes the reduction of 5,10-methylenetetrahydrofolate to 5-methyltetrahydrofolate (Trimmer 2013). A very common polymorphism, the 677 C > T variant, results in a thermolabile enzyme with decreased functionality. Individuals that carry the 677 C > T variant tend to have higher homocysteine levels and are at risk of folate deficiency (Dean 2012). Interestingly, Sae-Lee et al. (2018) have shown that supplementation with folic acid and vitamin B12 decreased epigenetic aging measured through Horvath’s clock (see below) only in women with the MTHFR wildtype variant, who responded to the intervention with higher serum levels of folate than men of the same genotype or individuals with the MTHFR wildtype (Sae-Lee et al. 2018). Of note, folate-rich food such as green leafy vegetables can easily be consumed as part of the diet.

The fact that polymorphisms in APOE and FOXO3 can influence both the aging process and the response to dietary interventions shows how much aging and nutrition are intertwined at the genetic level. Thanks to the growing wealth of genetic data related to human longevity, identifying other genes that sit at the intersection between nutrition and aging is becoming more feasible. In addition, resources such as the LongevityMap (Budovsky et al. 2013), one of the databases belonging to the HAGR collection (Human Aging Genomic Resources, Tacutu et al. 2018), give access to this type of data. We thus expect that more and more genetic analyses will be part of studies regarding dietary geroprotection.

Box 2: Genetics, cognition and nutrition.

Cognitive improvements and the amelioration of cognitive decline by diet, by polyphenols in general and polyphenol-rich berries in particular is a long-standing topic in nutrition research. Studies of blueberries started decades ago, and similar efforts were undertaken with strawberries, many other berries, fruit and vegetables (Vauzour 2012). Since genetics influences health, cognition and diet, complex relationships can be expected, and as shown here, a specific role of lipid profiles, inflammation and immunity has already been demonstrated. Moreover, cognitive ability shows its effect on food consumption over the whole lifetime, also on the socioeconomic side, as nutrition highly depends on income and educational level (Fergus, Seals, and Holston 2021).

GWAS identified numerous risk genes for aging-related neurodegenerative diseases such as Alzheimer’s Disease (Lambert et al. 2013; Jansen et al. 2019). Besides the APOE4 locus (as described elsewhere in this review), SNPs with smaller effect size were identified indicating risk association to genes expressed in immune-related tissues and cell types. Jansen et al. (2019) performed additional analyses that indicated further biological mechanisms for dementia which involved lipid-related processes and degradation of amyloid precursor proteins. Furthermore, the emerging search for genetic loci that interact with environmental factors to increase or decrease the risk of disease is highly intriguing as these factors can be tackled. As we demonstrate in this review, biomarkers related to immunity and inflammation as well as lipid profiles are frequently improved by dietary geroprotection. In turn, mediated by genetics, these positive effects can improve cognition, health, and diet.

As an example of the influence of diet on cognition, recent evidence suggests that healthy nutrient intake can positively affect specific aspects of brain integrity. Prinelli et al. (2019) investigated dementia-free older adults from a population-based Swedish National study with MRI data and data on dietary intake using a food frequency questionnaire, estimating the intake of 21 nutrients. Principal component analysis yielded five nutrient patterns. The nutrient pattern characterized by high fiber, vitamin C, E, β-carotene, and folate was associated with the lowest white matter hyperintensity (WMH) volume, compared to the pattern characterized by saturated fatty acids, trans fats, monounsaturated fatty acids, and cholesterol. Prinelli et al. conclude by suggesting brain-health specific nutrient combinations comprising higher intake of fruits, vegetables, olive and seed oils, oily fish, lean red meat, and poultry but less intake of milk and dairy products like cream and butter, and less intake of processed meat. Since WMH is also influenced by genetics (Sargurupremraj et al. 2020), gene-diet interactions are likely involved in the determination of WMH outcomes.

As an example of the influence of genetics on diet, likely mediated by cognition, the genetic background was shown to have an impact on the individual preference and amount of food consumption. In particular, Cornelis et al. (2015) identified 6 novel genetic loci that were associated with higher coffee consumption. Interestingly, loci near GCKR, MLXIPL, BDNF and CYP1A2 that were associated with higher coffee consumption have previously been associated with smoking initiation, higher adiposity and fasting insulin and glucose but lower blood pressure and favorable lipid, inflammatory and liver enzyme profiles, thus demonstrating a complex behavioral and metabolic pattern.

Epigenetic biomarkers

A wide variety of epigenetic changes accompany the aging process, including post-translational modifications of the amino-terminal tail of histones, accumulation of aging-related histone variants, and covalent modifications of DNA bases (Unnikrishnan et al. 2019; Villeponteau 1997; Wang, Yuan, and Xie 2018; Yi and Kim 2020). The latter, especially the methylation of CpG islands, proved themselves as surprisingly reliable biomarkers of aging (Galkin et al. 2020a). Epigenetics knowledge is becoming abundant also because of the recent development of omics assays that can measure DNA methylation on a large scale. Furthermore, the methylation state of CpG islands is stable so that samples can be stored long-term and analyzed when needed (Hollegaard et al. 2013).

Epigenetic clocks as predictors of chronological and biological age

Various epigenetic biomarker signatures (also called “clocks”) for the prediction of chronological age have been created using machine learning methodologies, such as elastic net regression, based on the methylation status of CpG islands. The earliest clocks predicted chronological age from such methylation marks (Hannum et al. 2013; Horvath 2013). Epigenetic age acceleration can then be defined as the difference between the predicted and the actual chronological age of a given individual, and used as a biomarker of aging, thus estimating biological age. One leap forward was made when the learning was not performed based on chronological age but on surrogates of morbidity or mortality (Field et al. 2018), yielding clocks such as PhenoAge (Levine et al. 2018), GrimAGE (Lu et al. 2019), and “disease” clocks (Lin and Wagner 2015; Zhang et al. 2017). McCrory et al. (2021) recently compared the ability of four clocks, that is, Horvath’s and Hannum’s 2013 clocks, PhenoAge and GrimAge to predict all-cause mortality and they investigated the association of these clocks with clinical measures related to morbidity and frailty. GrimAge, the best performing clock according to their analysis, could predict mortality with a high confidence and was significantly associated with almost all clinical phenotypes taken into consideration (McCrory et al. 2021).

Nutritional effects on the epigenome

The epigenome is sensitive to various dietary influences, for example to the intake of B-vitamins, polyphenols, and dietary choline (Khan et al. 2018; Mahmoud and Ali 2019; Mandaviya et al. 2019). Thus, we suggest that epigenetic biomarkers are useful in the study of dietary geroprotection. Many studies report the effects of nutritional interventions on the methylation status of loci involved in inflammation and age-related diseases. For example, Arpón et al. (2016) reported changes in the methylation state of eight genes related to inflammation and immunocompetence, after a five-year-long Mediterranean diet intervention, while Kok et al. (2015) identified epigenetic changes in regions of the genome related to carcinogenesis and development, in elderly subjects supplemented with folate and vitamin B-12 for two years. A few studies report the effect of diet on epigenetic clocks. Quach et al. (2018) carried out a cross-sectional study involving 4,173 post-menopausal women. The authors found that a higher extrinsic epigenetic age acceleration, a measure of age acceleration sensitive to age-related changes in blood cell composition based on Hannum’s clock (Chen et al. 2016), is inversely associated with fish intake, alcohol consumption, and blood carotenoid levels. On the other hand, intrinsic epigenetic age acceleration, a measure of cell-intrinsic aging based on a modified version of Horvath’s clock independent of blood cell composition (Chen et al. 2016), is negatively associated with poultry intake (Quach et al. 2018). Fitzgerald et al. reported a reduction of Horvath’s DNA methylation age of 1.96 years measured from saliva samples in a cohort that followed an 8-week diet and lifestyle intervention that targeted sleep, exercise, stress, and nutrition (Fitzgerald et al. 2021).

Another epigenetic biomarker that is sensitive to both aging and diet is the methylation state of transposable LINE-1 elements. The transcriptional accessibility of LINE-1 elements is negatively correlated to the methylation state of their CpG-island rich internal promoters. There are many studies that link hypomethylation of LINE-1 elements to various human diseases, from cancers to autoimmune diseases (Zhang, Zhang, and Yu 2020). LINE-1 element methylation state is inversely correlated with chronological age (Cho et al. 2015; Zhu et al. 2012), is lower in senescent cells compared to non-senescent ones (De Cecco et al. 2019) and, importantly, their methylation has been positively associated with dietary patterns rich in fiber and plant phytochemicals (Piyathilake et al. 2012). An interventional study observed slower changes in the methylation state of LINE-1 elements after the adherence to a one year-long Mediterranean diet intervention (Martín-Núñez et al. 2014). LINE-1 hypomethylation has also been correlated to a poorer response in terms of weight loss (Garcia-Lacarte et al. 2016) and increased systolic blood pressure (Ferrari et al. 2019) to diet. Notwithstanding the poorer response, dietary geroprotection is expected to modify the methylation state of LINE-1 elements for the better even in such cases.

The epigenome is profoundly affected by both diet and the aging process. Therefore, we expect that epigenetic biomarkers will increasingly be used in the evaluation of different types of dietary geroprotection. Furthermore, employing epigenetic clocks that reflect biological age can be a significant aid in measuring the magnitude of the geroprotective effects of a given dietary intervention.

Transcriptomic biomarkers

The transcriptome, just like the epigenome, shows many biologically relevant changes during aging. Age-related gene expression changes offer sufficient signals to enable the creation of “transcriptomic clocks.” Various transcriptome-based clocks have been created, using samples coming from different kinds of tissues. Examples are the clocks by Peters et al. (2015) and Huan et al. (2018) that predict from whole blood or PBMC, the clocks by Fleischer et al. (2018) and LaRocca, Cavalier, and Wahl (2020) that predict from skin fibroblasts and the clock by Mamoshina et al. (2018b) that predicts from muscle tissue, while the ones by Ren and Kuan (2020) and Shokhirev and Johnson (2021) are pan-tissue clocks (Hartmann et al. 2021). All the current transcriptomic clocks were trained to predict chronological age, just like the first generation of epigenetic clocks, so their usefulness in the context of geroprotection is not optimal, because age acceleration, or the difference between predicted age and chronological age, can only be moderately informative of biological age (Mitnitski and Rockwood 2019). Nevertheless, a major advantage of investigating transcripts in general and transcriptomic clocks in particular is the possibility, in principle, to more easily apply enrichment analyses directly to the genes that are used as features by these clocks, and to derive insights into aging and the effects of nutrition based on the resulting enriched biological terms/pathways.

Further down the gene expression cascade, proteomic biomarkers of aging are in development. Based on specific blood parameters, Sayed et al. described iAGE (Sayed et al. 2019, BioRxiv. doi: 10.1101/840363), a predictor based on immunological parameters including cytokines, chemokines, growth factors and cytomegalovirus infection status, and both Tanaka et al. and Lehallier et al. created clocks based on plasma proteins (Johnson et al. 2020; Tanaka et al. 2018, 2020). Proteins result from gene expression, and they can thus be used for enrichment analyses, just like transcripts, and help in the elucidation of the molecular mechanisms involved in aging and in the effects of nutrition on aging.

Nutritional effects on transcription

The effects of diet on transcriptomic clocks are yet to be described in detail. However, various studies have shown that dietary changes, such as the adoption of a Mediterranean diet or a Nordic style of eating, induce wide-reaching changes in the transcriptome of PBMCs and whole blood. For example, Myhrstad et al. (2019) showed, using the transcriptome of PBMCs, that a three to four month-long Nordic diet intervention modulates inflammation in individuals with metabolic syndrome by downregulating the expression of many pro-inflammatory and upregulating the expression of a few inflammation-resolving genes; genes coding for electron transport chain components and cellular proliferation are downregulated as well (Myhrstad et al. 2019). As another example, Esser et al. (2018) showed that a four week-long daily supplementation of 100 mg epicatechin, a flavonol that can be found, e.g., in cocoa, strawberries and apple peels, down-regulates the transcriptional level of inflammatory genes in PBMCs, while Barrera-Reyes et al. (2019) obtained similar results analyzing the PBMCs of subjects two hours after the intake of high polyphenol cocoa powder (Esser et al. 2018; Barrera-Reyes et al. 2019). In whole blood, postprandial changes in glycolytic and inflammatory gene expression were found after the consumption of yogurt and acidified milk in a cohort of young men, pointing to an anti-obesity effect and a biphasic modulation of inflammation, with an upregulation at two hours and a down regulation at six hours after the intake of these foods (Burton et al. 2018).

Several studies also reported positive transcriptomic changes in PBMCs after the consumption of algal or fish oils, which are rich sources of omega-3 fatty acids; the transcriptional processes most influenced by the consumption of fish or marine oils are related to inflammation (Bouwens et al. 2009; Myhrstad et al. 2014; Myhrstad et al. 2016; Souza et al. 2020). According to Schmidt et al. (2012), immunomodulatory effects of omega-3 fatty acids are more evident in dyslipidemic subjects. Further, Polus et al. (2016) studied the effects of omega-3 fatty acid supplementation, in a cohort of obese women, a group characterized by dyslipidemia and chronic inflammation. At the end of the three-month intervention, they observed, based on PBMCs, an up-regulation of the transcriptional levels of ALOX5, a gene that codes for an enzyme involved in the production of pro-resolving DHA derivatives, and of target genes of the transcription factors NRF2 and PPAR-α, both involved in fatty acid beta-oxidation, phospholipid synthesis, mitochondrial electron transport chain, and antioxidant defense. These gene expression changes were accompanied by a decrease in circulatory cytokines, adhesion molecules, and acute-phase proteins and an increase in pro-resolving mediators, indicating an anti-inflammatory effect elicited by the omega-3 fatty acids. GWAS of fish oil supplementation effects recently revealed new gene-diet interaction loci, some of which implicated in LDL metabolism (Francis et al. 2021).

Long-term changes to the transcriptome

Transcriptional changes may be limited to a certain temporal window, as already described for acute effects by Burton et al. (2018), see above. Other changes induced by a dietary intervention may only appear long-term, sometimes long after its conclusion. For example, Milella et al. (2020) sampled the blood of 20 individuals instructed to consume approximately 350 g of grapes per day for 21 days at three different time points: at the start of the dietary intervention (T0), at its end (T1), and after a washout period (T2). While changes in the expressional levels of genes related to blood coagulation were apparent only at T1 and were absent at T2, the ones in genes related to inflammation were shared by both time points. These changes were even stronger at T2, and the ones in cellular housekeeping such as autophagy, mitochondrial biogenesis, and DNA repair became apparent only at T2.

Dietary interventions can affect transcriptomic processes related to aging and disease, such as inflammation, antioxidant defense and stress response. To understand the underlying molecular mechanisms, transcriptomic analyses should be included in geroprotective trials. Moreover, transcriptomic clocks help interpret transcriptomic data, complementing enrichment analyses by giving a concise representation of the magnitude of the geroprotective effects. Lastly, it is worth noting that the transcriptome tends to change faster in response to interventions than the other types of biomarkers discussed here, with the advantage that it can reflect acute changes and the disadvantage that it can be much noisier.

Microbiome biomarkers

The microbiome comprises the ensemble of bacteria, viruses, fungi, and archaea that live symbiotically with the human body. It has a profound and wide-reaching effect on health; in fact, dysbioses of microbial communities in the skin, lungs, oral cavity, and the gut have all been implicated in the susceptibility to various inflammatory and neoplastic syndromes (Lloyd-Price, Abu-Ali, and Huttenhower 2016; Gilbert et al. 2018). The microbiome of the gut is the best-studied one, and it is also the most impacted by dietary influences (Hills et al. 2019). Many insights are emerging about the changes that the microbiome undergoes during aging. Galkin et al. recently used deep learning to construct a predictor (dubbed a “microbiome clock”) of chronological age based on microbial taxonomic profiles from more than 4000 individuals aged 18–90 years (Galkin et al. 2020b). Many properties of the aging microbiome were identified, some of which may underlie such predictors. For example, with aging, there is an increase in the Firmicutes to Bacteroides ratio (F/B ratio) that resembles the pattern seen in the obese. In particular, Vaiserman et al. have recently shown, in a cross-sectional study that involved 1550 healthy participants, that the F/B ratio increases gradually from childhood to old age (Vaiserman et al. 2020).

Nutritional effects on the microbiome and the microbiome-host interaction

Rodríguez-Morató et al. (2018) have recently shown through a randomized, placebo-controlled, double-blind, crossover study involving 11 young and healthy subjects that an animal-based diet leads to a pronounced increase in the F/B ratio mentioned above. Moreover, the animal-based diet led to a decrease in stool short-chain fatty acids and an increase in bile acids. These changes were attenuated by the daily supplementation of 30 grams of freeze-dried whole cranberry powder (Rodríguez-Morató et al. 2018). This is evidence of the possibility of rejuvenating the F/B ratio through the addition of dietary geroprotectors.

During youth, the coupling between the human body and the microbiome is highly functional, but during aging, a slow but steady uncoupling is observed (Bauer, Rees, and Finlay 2019; Esser et al. 2019; Santoro et al. 2020b). This is exemplified by the age-related loss of tryptophan and indole production by the microbiome that has recently been reported (Ruiz-Ruiz et al. 2020). Further, Zhu et al. (2020) have shown, in a small controlled crossover trial involving ten young and healthy subjects, that a short four day Mediterranean diet intervention can lead to a significant increase in fecal tryptophan metabolites accompanied by an increase in short chain fatty-acid producing bacteria belonging to the Lachnospiraceae family or the genus Butyricicoccus (Zhu et al. 2020). Also, Ulaszewska et al. have shown through a randomized crossover trial involving 40 mildly hypercholesterolemic subjects, that bacterial tryptophan synthesis and metabolism could be increased by daily intake of two apples every day for eight weeks (Ulaszewska et al. 2020). There is also evidence, from animal models, that tryptophan metabolism is essential for the maintenance of colonic barrier integrity and the prevention of inflammaging (Gao et al. 2018; Powell et al. 2020; Shimada et al. 2013; Taleb 2019). Many tryptophan metabolites interact with the aryl hydrocarbon receptor (AHR), an important player in the maintenance and differentiation of the stem cells of the gut and in the activity of resident immune cells (Sun et al. 2020).

Short chain fatty acids (SCFA) are another important metabolite produced and secreted by the gut microbiota, whose levels decrease during aging; in particular these are acetate, butyrate and propionate (Salazar et al. 2019). SCFA can improve gut barrier integrity, enter the circulation, and have various immuno-modulatory effects that are protective against age-related diseases (Morrison and Preston 2016). Since the gut microbiome produces SCFA from plant components that are still undigested (or partially digested), there are many possible avenues to accelerate their production. The simple addition of fiber-rich foods can increase the abundance of SCFA producing species. For example, Holscher et al. (2018) showed that a 3 week intervention that consisted of the daily supplementation of 41 g of walnuts to 18 healthy men and women with “normal” diets, led to a decrease in fecal content of bile acids and a concomitant increase in the relative abundances of butyrate-producing Firmicutes species. These changes were accompanied by a decrease in circulating LDL cholesterol and the non-cholesterol sterol campesterol (Holscher et al. 2018). On the other hand, the levels of the microbial-derived metabolite Trimethylamine N-oxide (TMAO), a factor possibly contributing to the pathogenesis of cardiovascular disease and dementia, tend to be higher in older adults (Brunt et al. 2020). Wu et al. (2020) created a predictor of TMAO production that was based on the presence of Emergencia Timonensis or Ihubacter Massiliensis in the feces of human subjects. This predictor could distinguish high TMAO producers from low producers with 43% sensitivity and 97% specificity. A high producer was defined as an individual that shows high plasma levels of TMAO after the intake of carnitine, an amino acid that is highly present in red meat (Wu et al. 2020).

Thus, the biomarkers useful to evaluate dietary geroprotection based on investigating microbiomes range from “black-box” predictors based on deep learning, to the more specific F/B ratio. Taxa that produce tryptophan, indole, short-chain fatty acids, bile acids, and TMAO may also give rise to candidate biomarkers.

Nutritional effects are influenced by the gut microbiome

Microbiome-related biomarkers may also be used to identify, a priori, who would benefit from dietary geroprotection. The geroprotective activity of many polyphenols is partially explained by their microbiota-mediated conversion into bioavailable metabolites (Kawabata, Yoshioka, and Terao 2019). This is exemplified by the response to the consumption of isoflavones and ellagic acid derivatives, which differs between equol and urolithin producers and non-producers, respectively. Three groups exist based on their microbiome’s ability to convert ellagic acid into urolithins: Urolithin-A metabotypes, who can convert ellagic acid into Urolithin A, Urolithin-B metabotypes, who can convert it into both Urolithin-A and Urolithin-B, and Urolithin-0 metabotypes, who are unable to convert it into either. Regarding equol, two groups exist: producers and non-producers.

García-Mantrana et al. (2019) have shown that the microbiome of Urolithin-B metabotypes is more sensitive to modulation after short-term walnut consumption. Specifically, an increase in Blautia, Bifidobacterium, and members of the Coriobacteriaceae family was seen only in Urolithin-B metabotypes (García-Mantrana et al. 2019). González-Sarrias et al. (2017) have also shown that Urolithin-B metabotypes are the only ones that responded to the consumption of pomegranate extract in their double-blind, crossover, dose–response, randomized, placebo-controlled trial of 50 participants. Specifically, a change in various clinical biomarkers of cardiovascular health was seen in Urolithin-B metabotypes, with decreases in total cholesterol, LDL-cholesterol, small LDL-cholesterol, non-HDL-cholesterol, apolipoprotein-B and oxidized LDL (González-Sarrías et al. 2017). Regarding equol, Zheng et al. (2019) have shown that both content and metabolic activity of the gut flora differed between equol producers and non-producers. Furthermore, equol producers showed lower prevalences of dyslipidemia, which suggests the important role that equol might play in lipid metabolism related to the gut microbiota (Zheng et al. 2019).

The microbiome can be informative both before dietary geroprotection, as a way to select or stratify study participants based on their microbiome’s metabolism of the nutrients under investigation, and after dietary geroprotection, as a way to analyze the beneficial effects in more detail. In addition, since host genetics also plays a role in the microbiome-related effects of diet, it will be interesting to combine genetics, microbiomics, and clinical outcomes in future studies.

Clinical biomarkers

Commonly administered clinical laboratory tests may give rise to biomarkers of aging that are based on highly standardized measurement procedures, easily available, and often cheaper than the other types of biomarker signatures discussed here (Hartmann et al. 2021). The most recent biomarker signatures (“clocks”) originating from laboratory tests are based on deep learning (Putin et al. 2016; Ashiqur Rahman et al. 2021; Galkin et al. 2020a), while some older ones, such as the “Phenotypic Age” (Liu et al. 2018; Levine 2013) were based on multiple linear regression (Voitenko and Tokar 1983), principal component analysis (Jia et al. 2016), Hochschild’s method (Hochschild 1989), or Klemera and Doubal’s method (Klemera and Doubal 2006). Of note, the latter set of methods tends to be specific for the small sets of measurements usually established in the laboratory, while deep learning and regression are also employed to learn epigenetics, transcriptomics and microbiome-based biomarker signatures.

Clinical biomarker-based clocks were used to judge the effects of caloric restriction (CR). In particular, the “Klemera-Doubal Biological Age” (KD age), calculated from the CALERIE trial blood chemistry and blood pressure data following the method of Klemera and Doubal, can be impacted by two years of CR (Belsky et al. 2020). The slowdown of predicted age was especially evident in subjects with a higher baseline KD age; within two years, they showed a standstill in KD age when put on CR, while the control group aged by two years. Since many natural compounds have CR-mimetic effects, it is plausible that KD age is suitable for assessing some approaches to dietary geroprotection (Martel et al. 2019; Yessenkyzy et al. 2020).

Nutritional effects on clinical biomarkers

Many clinical biomarkers reflect established cardiovascular risk factors such as lipid profiles (LDL and HDL cholesterol, triglycerides), high blood pressure and obesity, but also emerging risk factors such as high homocysteine and low vitamin D levels, as well as CRP, other inflammation-related cytokine/chemokine levels, glucose metabolism and oxidative stress, and all these can be influenced substantially by dietary geroprotection (Carrizzo et al. 2020; Unuofin and Lebelo 2020; Zhou et al. 2021). More specifically, high LDL and low HDL cholesterol levels are common among the elderly, often reflecting metabolic syndrome (Walter 2009), and high LDL-cholesterol levels are significantly associated with incident cardiovascular disease in both middle and old age (Félix-Redondo, Grau, and Fernández-Bergés 2013; Lind et al. 2018; Zhang et al. 2020a). Moreover, overall, human aging is associated with a rising incidence of insulin resistance and type 2 diabetes (Cheng et al. 2013; Creatore et al. 2010; The DECODA Study Group 2003), and insulin as well as hemoglobin A1c and glucose measurements were all shown to be improved by diet (Castro-Barquero et al. 2020), also see Table 1.

Table 1. Effects of berries on clinical biomarkers, based on systematic reviews and meta-analyses.

Author	Type of review	Cohorts included	Type of intervention	Number of trials	Number of participants	Impacted biomarkers	Duration of interventions
Daneshzad et al. 2019; doi:10.1016/j.clnu.2018.06.979	Meta-analysis	MetS, athletes, PM women, OW/OB, HCL, NAFLD, T2D, healthy, glaucoma patients	Various berries, and anthocyanin rich foods and extracts	19	1432	↓HOMA-IR; ↓TC and ↓LDL if (dosage > 300 mg/day) and (duration > 12 weeks)	1 to 96 weeks
Fairlie-Jones et al. 2017; doi.10.3390/nu9080908	Systematic review and meta-analysis	Healthy, pre-HTN/HTN, HCL, at CVD risk, MetS, CHD	Various berries, anthocyanin rich foods and extracts, and purified anthocyanins	24	720	↑FMD following both acute and chronic supplementation; ↓PWV for acute supplementation only	Acute (single-dose), or 1 week to 6 months
Fallah et al. 2020; doi:10.1016/j.fct.2019.110922	Meta-analysis	Healthy, post MI, MetS, OB and IR, pre-HTN, T2D, HLD, pre and type 1 HTN, AA and HLD, pre T2D, MI, OB with knee OA	Various berries, and anthocyanin rich foods and extracts	32	1744	↓CRP, ↓IL-6, ↓TNF-α, ↓ICAM-1, ↓VCAM-1, ↑adiponectin; Stronger ↓ for CRP, IL-6, TNF-α, and VCAM-1 if (dosage > 300 mg)	3 weeks to 6 months
Fallah, Sarmast, and Jafari 2020; doi:10.1016/j.foodres.2020.109379	Meta-analysis	Post-MI, DLP, MetS, NAFLD, HTN, T2D, OW/OB, healthy	Various berries, and anthocyanin rich foods and extracts	40	1823	↓FBG, ↓2-h PPG, ↓resistin, ↓PAI-1; ↓HbA1c, and ↓HOMA-IR; Stronger ↓FBG, ↓2-h PPG, ↓HbA1c, and ↓HOMA-IR if (duration >8 weeks) and (dosage >300 mg/day); ↓FBG if (healthy or T2D); stronger ↓2-h PPG and ↓HbA1c if (T2D); ↓HOMA-IR if (T2D or OW/OB).	2 weeks to 6 months
Gao et al. 2020; doi:10.1017/S000711452000121X	Meta-analysis	Middle aged individuals with OB + OA, T2D, pre & stage 1 HTN, OW, OW HLD, MetS, AA, T2D, adolescents	Strawberries (both fresh, frozen, freeze-dried powder and beverage)	11	357	↓CRP; ↓TC if (baseline levels >5 mmol/L); ↓LDL if (baseline levels >3 mmol/L)	1 to 12 weeks, median 6 weeks
Hadi et al. 2019; doi:10.1039/c9fo01684h	Meta-analysis	DLP, knee OA, HTN PM women; T2D, MetS, healthy	Strawberry beverage, fresh and freeze dried	20	578	↓Circulating oxLDL, ↓MDA, ↓CRP, ↓TC, and ↓DBP, ↑FBG	3 to 12 weeks
Heneghan et al. 2018; doi:10.1002/mnfr.201700645	Systematic review	HTN, HLD, male SMK, CAD, healthy, HTN females, MetS, at CVD risk, IR, T2D males, OB + IR	Various berries (raw, juices, powders, etc.)	23	1168	In multiple studies: ↓SBP, ↓DBP, ↓PWV, ↓TAGs, ↓TC, ↓LDL, ↓oxLDL,	4 to 16 weeks
Huang et al. 2016; doi:10.1038/srep23625	Meta-analysis	Healthy, T2D, CVD risk, DLP, PM women, OB + MetS, MetS, CAD risk, HCL	Various berries (raw, juices, supplements, etc.)	22	1251	↓LDL, ↓SBP, ↓FBG, ↓HbA1c, ↓BMI; ↓TNF-α.	2 to 24 weeks
Martini et al. 2020; doi:10.1093/nutrit/nuz053	Systematic review	Healthy, OW, OB, stable CAD, pre-HTN or stage 1 HTN, ED + at CVD risk, pre-HTN and light SMK, MetS, healthy OW, healthy, males at CVD risk, athletes, male SMK	Various berries (raw, juices, supplements, etc.)	22	735	In multiple studies: ↑FMD, ↑RHI after acute intervention, ↓PWV, ↓AAI after chronic intervention	Acute (single-dose), or 1 to 8 weeks
Rahmani et al. 2019; doi:10.1002/ptr.6398	Systematic review and meta-analysis	MetS, healthy, athletes, HTN and post-MI	Aronia berry	7	286	↑HDL, ↑DBP, ↓LDL; ↓LDL more significantly and ↓TC if (duration <10 weeks); ↓TAGs if (dosage >300 mg/day)	4 to 24 weeks
Rocha et al. 2019; doi:10.1080/10408398.2018.1430019	Systematic review	T2D subjects not taking insulin	Blueberries and cranberries, extract, pure or juice	7	270	In multiple blueberry studies: ↓FBG,↓2h-PPG and ↓HbA1c if (duration ≥ 8 weeks); In single cranberry studies: ↓FBG if (duration ≥ 12 weeks); both acute blueberry and cranberry: ↓PPG, ↓Insulin-PP	Acute (single-dose), or 8 to 12 weeks
Shah and Shah 2018; doi:10.1155/2018/8450793	Systematic review and meta-analysis	Untreated T2D, HCL, NAFLD, T2D SMK, healthy, HLD, pre-HTN men, HCL + OB, CAD, MetS, HLD, DLP	Various berries, and anthocyanin rich foods and extracts	17	1573	↓TAGs, ↓LDL, ↓apoB, ↑HDL, ↑apoA1	4 to 24 weeks
Wallace, Slavin, and Frankenfeld 2016; doi:10.3390/nu8010032	Systematic review	Healthy, MetS, pre-HTN, Post MI, HLD, DLP, HCL	Various berries, and anthocyanin rich foods and extracts	10	1003	In multiple studies: ↓LDL if (not healthy or hyperlipidemic)	3 to 24 weeks
Wang et al. 2021; doi:10.1007/s00394-020-02299-w	Systematic review and meta-analysis	Healthy, DLP, HTN, CVD/CHD, MetS, OW, mild to moderate CI, COPD, T2D, ESRD	Various berries, cherries and citrus fruit (raw, freeze dried, juices and powders)	45	1130	↓SBP, ↓DBP, ↑sVCAM-1 if berry	7 to 180 days
Yang et al. 2017; doi:10.3945/an.116.014852	Systematic review and meta-analysis	T2D, pre-T2D, OW and OB, AA, MetS, DLP, HLD, HCL, NAFLD, HTN, CVD risk, CAD, healthy	Various berries, and anthocyanin rich foods, juices and extracts	32	1491	↓FBG, ↓2 h-PPG, ↓LDL ↓TC; ↓HbA1c; HOMA-IR if (overweight or obese); stronger ↓HbA1c, ↓TAGs if (dosage >400 mg/d)	2 to 24 weeks

See the Glossary (Table 2, for abbreviations and acronyms.

Table 2. Glossary.

Abbreviation/acronym	Meaning
↑	Increase
↓	Decrease
2hPPG	Two Hours Post Prandial Glucose
AA	Abdominal Adiposity
AAI	Arterial Augmentation Index
AD	Alzheimer’s Disease
AHR	Aryl Hydrocarbon Receptor
AI	Aortic Augmentation Index
ApoA1	Apolipoprotein A1
ApoB	Apolipoprotein B
BaPWV	Brachial-ankle Pulse Wave Velocity
BMI	Body Mass Index
BW	Body Weight
CAD	Coronary Artery Disease
CHD	Coronary heart disease
CI	Cognitive Impairment
COPD	Chronic Obstructive Pulmonary Disease
CR	Caloric Restriction
CRP	C-Reactive Protein
CVD	Cardiovascular Disease
CVT	Cerebral Venous Thrombosis
DBP	Diastolic Blood Pressure
DLP	Dyslipidemia
ED	Endothelial Dysfunction
ESRD	End-Stage Renal Disease
F/B ratio	Firmicutes to Bacteroides ratio
Fasting Blood Glucose	FBG
FMD	Flow-Mediated Dilation
GWAS	Genome Wide Association Studies
HCL	Hypercholesterolemia
High-Density Lipoprotein Cholesterol	HDL
HLD	Hyperlipidemia
Homeostatic Model Assessment for Insulin Resistance	HOMA-IR
HTN	Hypertension
ICAM-1	Intercellular Adhesion Molecule 1
Insulin-Postprandial	Insulin-PP
IL-6	Interleukin 6
IR	Insulin Resistance
KD age	Klemera-Doubal Biological Age
LDL	Low-Density Lipoprotein Cholesterol
MCI	Mild Cognitive Impairment
MDA	Malondialdehyde
MetS	Metabolic Syndrome
MI	Myocardial Infarction
NAFLD	Nonalcoholic Fatty Liver Disease
OA	Osteoarthritis
OB	Obesity/Obese
OW	Overweight
oxLDL	Oxidized Low-Density Lipoprotein
PAD	Peripheral Arterial Dysfunction
PBMC	Peripheral Blood Mononuclear Cells
PM	Post Menopausal
MI	Myocardial Infarction
PPG	Post Prandial Glucose
PWV	Pulse Wave Velocity
RHI	Reactive Hyperemia Index
SBP	Systolic Blood Pressure
SCFA	Short Chain Fatty Acids
SMK	Smokers
T2D	Type 2 Diabetes
TAGs	Triglycerides
TC	Total Cholesterol
TMAO	Trimethylamine N-oxide
TNF-α	Tumor Necrosis Factor-alpha
VCAM-1	Vascular Adhesion Molecule 1
WMH	White Matter Hyperintensity

Berries for dietary geroprotection

We chose to specifically focus on the effects of berries and their constituents, such as anthocyanins, as an example of dietary geroprotection. Berries are a widely consumed food, and they can be added easily to one’s diet. They can be eaten fresh, avoiding the loss of nutritional value caused by processing (Nayak, Liu, and Tang 2015). All of the widely consumed berries (such as strawberries, blueberries, raspberries, and cranberries) represent a hypocaloric source of fiber and polyphenols. The various classes of fiber, including cellulose, hemicellulose, arabinans, and arabino-xyloglucans that are present in berries are also implicated in their beneficial properties (Rodríguez-Daza et al. 2020). Putative beneficial properties of the polyphenol flavonoid fisetin were mostly derived from animal studies. For example, fisetin reduced cellular senescence markers in multiple tissues in mice, effects that are often comparable to the more often studied dasatinib plus quercetin combination (Yousefzadeh et al. 2018). Also, procyanidin C1, a polyphenolic compound common to various berries and specifically found in grape seed extract, is believed to act as an anti-aging dietary factor in mice (Xu et al. 2021, BioRxiv. doi: 10.1101/2021.04.14.439765). More generally, polyphenols such as anthocyanins were shown to exhibit anti-inflammatory properties (Hidalgo et al. 2012) and to be promising candidates for the amelioration of obesity and age-related inflammation (Joseph, Edirisinghe, and Burton-Freeman 2014; Tsuda 2016; Azzini, Giacometti, and Russo 2017; Land Lail et al. 2021). The positive impact of berries on health is usually explained by their constituent polyphenols as well as the metabolites thereof, which can enter into the circulation and exert anti-inflammatory effects. Further below, we will refer to the extensive literature regarding studies of berry effects in humans, often involving biomarkers of inflammation.

Referring to a notable combination of in-vivo with in-vitro work in human, Rutledge et al. (2019) recently showed that the serum of subjects supplemented with either strawberry or blueberry powder can downregulate inflammatory signaling in vitro. Their study also hinted at a cumulative effect of berry polyphenols, as the sera of the berry consumers were more anti-inflammatory after 90 days of intervention than after 45 days; the sera kept their anti-inflammatory potential when the participants were in a fasted state. Rodriguez-Mateos et al. (2019) also found evidence for the importance of circulating anthocyanin metabolites. Their study, consisting of a one-month-long wild blueberry intervention, showed that of the 63 circulating anthocyanin metabolites under scrutiny, 14 correlated with acute flow-mediated dilation (FMD) improvements, and 21 with chronic FMD improvements. Berries and berry fractions can also shape the microbiome away from the dysbiotic states typical to obesity, as shown recently by Rodríguez-Daza et al. (2020) in a mouse model of high fat diet-induced obesity. Also, the polyphenols contained in berries can synergize with each other. For example, Nichols et al. (2015) have shown that epicatechin and quercetin, two common polyphenols in berries, have superior neuroprotective properties than epicatechin or quercetin alone when used in a mouse model of hypoxic-ischemic brain injury. As for the mechanisms involved in the health-protecting effects of berry polyphenols, hormesis (Hayer 2007; Franco, Navarro, and Martínez-Pinilla 2019) is one more recent focus. Polyphenols such as anthocyanins can stimulate NRF2-mediated antioxidant gene expression and augment cellular and organismal resistance to stressors, either internal, such as metabolic stress and aging, or external, such as pollutants and UV radiation (Son, Camandola, and Mattson 2008; Leri et al. 2020). In fact, there is evidence that anthocyanins and oligomeric procyanidins protect the skin against UV-B damage (Li et al. 2019, Saliou et al. 2001), that they lower systemic oxidative stress (Ullah et al. 2019), and that they protect against various kinds of environmental pollutants (Lagoa et al. 2020).

Since there is abundant data about the favorable effects of berries on human physiology, we decided to summarize the recent (from 2015 up until today) systematic reviews and meta-analyses on human dietary berry and anthocyanin interventions and the biomarker effects observed, see Table 1. These meta-studies covered 144 different original studies, with an average coverage of only 14% of these original studies by any one meta-analysis (see Table S1 in supplementary material). One of the most important observations among the various meta-studies is that berries have more potential to modulate a given biomarker when its level is not in its optimal range at baseline. For example, Gao et al. (2020) reported that LDL and TC levels decreased after berry intake only in subjects with higher baseline levels. Yang et al. (2017) reported a positive decrease in the HOMA-IR index after berry intake only in overweight and obese subjects. Wallace, Slavin, and Frankenfeld (2016) reported a similar observation for LDL, with a decrease seen only in subjects suffering from diseases related to obesity. Another frequent observation is that berry-based interventions were effective only above a certain dosage and in studies of sufficient duration. Rahmani et al. (2019) reported a decrease in TAGs only for doses higher than 300 mg/day, Fallah, Sarmast, and Jafari (2020) reported more significant decreases in FBG, 2-h PPG, HbA1c, and HOMA-IR for studies that had a duration of eight weeks or more and dosages higher than 300 mg/day, Daneshzad et al. (2019) reported significant decreases in TC and LDL only for doses of anthocyanins higher than 300 mg/day and durations of 12 weeks or more, Fallah et al. (2020) reported that doses of anthocyanins higher than 300 mg/day are more effective at decreasing the levels of CRP, IL-6, TNF-α, and VCAM-1, and Yang et al. (2017) reported stronger reductions in HbA1c and TAGs levels with doses of anthocyanins higher than 400 mg/d.

Most of the reviews we investigated did not focus on any single berry but covered various berries, anthocyanin-rich foods, and extracts. One meta-analysis, the one by Fairlie-Jones et al. (2017), also covered the effects of purified anthocyanins. Gao et al. (2020) and Hadi et al. (2019) focused on strawberries, be they fresh, frozen, freeze-dried, or in beverage form. Rahmani et al. (2019) focused specifically on Aronia berries, while Rocha et al. (2019) focused on blueberries and cranberries. Regarding the types of cohorts, most of the reviews covered heterogeneous populations, characterized by both healthy and diseased subjects. The most common diseases among the study cohorts were related to metabolic syndrome, such as type 2 diabetes, dyslipidemia, pre-hypertension, and cardiovascular disease. Rocha et al. (2019) were the only ones to focus exclusively on subjects suffering from type 2 diabetes. The biomarkers covered by the reviews span five main biological processes: inflammation, vascular health, oxidative stress, lipid metabolism, and glucose control (see Figure 1). No difference was evident between the various types of berries in their ability to modulate the various biomarker classes. This could be because of the similarities between their constitutive polyphenols, often belonging to the anthocyanin class. Also, as reported in the first part of this section, berry polyphenols may often confer their protective effect in a similar manner, specifically by stimulating NRF2 activity. This observation could also be prevalent because of an investigator bias toward stable, validated biomarkers of a limited set of pathologies known to be influenced by diet, such as cardiovascular disease. Nonetheless, the fact that clinical biomarkers related to aging and disease can be influenced by berry consumption is instructive for future studies of dietary geroprotection. In any case, clinical measures should accompany transcriptomic, epigenetic, and microbiomic biomarkers to have a full-spectrum view on the beneficial effects of dietary geroprotection. A summary of the effects of berries and anthocyanins on human physiology is thus given in Figure 1. The detailed results are given in Table 1. See the glossary (Table 2) for abbreviations and acronyms.

Figure 1. The effects of berries on biomarkers of cardiovascular health and aging. See the glossary (Table 2, for abbreviations and acronyms).

Box 3: Nutribiography.

Following the notion of an “immunobiography” (Franceschi et al. 2017), we suggest that the term “nutribiography” describes the combination of type, dose, frequency and temporal sequence of nutrients that an individual has been exposed to in their lifetime.

As the most straightforward example, malnutrition can cause non-optimal long-term outcomes, characterized by a heightened risk of diseases related, e.g., to metabolic syndrome. Subjects exposed in utero to the “Dutch Hunger Winter” provide a rather extreme case. Two different responses were observed: subjects exposed during early gestation were born with an overall healthy phenotype. In contrast, subjects exposed during late gestation were born underweight. However, during later life, the former experienced a heightened risk of cardiovascular disease, an altered lipid profile, and a propensity to obesity, while the latter were metabolically healthier and leaner than the normal population (Schulz 2010). Epigenetic changes of, e.g., insulin signaling were reported as underlying causes (Tobi et al. 2014). More subtle differences, such as maternal deficiencies in vitamins and minerals, can also impact the child’s future risk of developing non-communicable diseases (Langley-Evans 2015). Various other aspects of diet can affect the epigenome. In particular, polyphenols are known to influence the epigenetic machinery, so that subjects with a high lifelong intake of polyphenols have an epigenetic memory of their intake (Arora et al. 2020; Leri et al. 2020). It will be interesting to see if people with a higher lifelong polyphenolic intake appear epigenetically younger compared to same age peers, even during periods of lower intake.

Another nutribiographic phenomenon is the “obesogenic memory,” a propensity to hyperphagia and weight gain in ex-obese individuals, potentially triggering Yo-Yo dieting cycles (Contreras, Schriever, and Pfluger 2019). Persistent changes of the gut microbiome induced by obesity may play a role in obesogenic memory. Accordingly, Fragiadakis et al. showed, in a subset of participants (n = 49) of the DIETFITS weight-loss trial, that the human microbiome, once set, can be resilient to diet-induced changes. They observed significant diet-induced changes at three months relative to baseline. However, the subjects’ microbiome returned to its pre-intervention state throughout the study’s later phases (six months, nine months, twelve months), in individuals consuming both a low-fat and a low-carbohydrate diet, despite adherence to the dietary guidelines and decreasing weight throughout the study (Fragiadakis et al. 2020). Thaiss et al. showed an obesogenic memory also in the microbiome of mice (Thaiss et al. 2016). The obesogenic microbiome of the mice decreased the bioavailability of the flavonoids apigenin and naringenin, but combined supplementation with both flavonoids normalized intestinal apigenin and naringenin levels and ameliorated the rate of secondary weight regain. Thus, dietary geroprotection can work despite a disadvantageous nutribiography.

As stated in the paragraph dedicated to the microbiome, equol and urolithin metabolizers can benefit considerably from consuming soy isoflavones and ellagitannins, respectively. However, apart from genetic factors, past dietary habits have also a strong influence on such traits; for example, diets rich in daidzein are positively correlated with equol producer status (Iino et al. 2019), and a month-long intervention with an ellagitannins-rich supplement can convert non-producers of urolithin into producers (González-Sarrías et al. 2017). Similarly, more short-chain fatty acids (acetate, butyrate and propionate) are produced from fiber by vegetarians and vegans than by habitual meat-eaters (Sheflin et al. 2017; Tomova et al. 2019). An individual’s nutribiography can thus influence their response to dietary geroprotection through differences in their microbiome or epigenome.

Conclusions

There is ample evidence available to state that aging-associated processes such as inflammation can be impacted by dietary interventions, throughout the omics spectrum, and, as known for a long time, in standard lipid and inflammatory cytokine/chemokine profiles. The sensitivity and specificity of many of the resulting biomarker signatures are dependent on an array of factors, such as the timing of the measurements, the strength of the intervention, and the study cohort, and sensitivity is often higher when the biomarkers are measured following an intervention with a higher dose, or in a population suffering from age-related diseases, as compared to the healthy elderly.

Some biomarkers are particularly suited for trial cohort pre-selection, while others for judging the effects of dietary geroprotection. Genetics can be employed specifically in the pre-selection phase, microbiomic and clinical biomarkers can be used for both cohort pre-selection and evaluation, while epigenetic and transcriptomic biomarker signatures work best to evaluate outcomes, affording a chance to understand the mechanisms underlying the effects observed. The development of clocks created through deep learning could also be a valuable tool for the evaluation of dietary geroprotection; new approaches to interpretable machine learning can provide further mechanistic insights.

As a suitable candidate for dietary geroprotection, we have chosen berries. Thus we summarized the recent systematic reviews and meta-analyses in terms of berry-related interventions. Furthermore we report biomarkers to be significantly changed in response to berry intake. We expect those biomarkers to be suitable candidates for future dietary geroprotection studies.

Finally, we suggest that the response of an individual to a diet in general, and dietary geroprotectors in particular, partly depends on the food they have been exposed to in the past. This "nutribiography," a term inspired by "immunobiography" (Franceschi et al. 2017), is thus defined by the combination of type, dose, frequency, and temporal sequence of nutrition that an individual was exposed to over their lifetime. Therefore, the “nutribiography” should be investigated in more depth in future studies in the context of dietary geroprotection, see Box 3.

Disclosure statement

HJG has received travel grants and speakers honoraria from Fresenius Medical Care, Neuraxpharm, Servier and Janssen Cilag as well as research funding from Fresenius Medical Care. None of the other authors report any relevant financial or non-financial competing interest.

Additional information

Funding

This work was supported by Karls Erdbeerhof (Karls Strawberry Farm), Rövershagen, Germany, and the “Interdisziplinäre Fakultät” (Department AGIS) of Rostock University. GF is further supported by the BMBF (FKZ 01ZX1903A and 1DS19049) and the European Commission (Aging with elegans, grant agreement 633589).