Collagen amino acid supplements may lower biological age in people and improve health and lifespan in living organisms

To determine whether the prolonged collagen homeostasis of [3 Gly : 1 Pro : 1 Hyp] supplementation corresponds to lifespan benefits, we measured the lifespan of C. elegans fed this ratio of amino acids post-development, starting from young adulthood. We found that [3 Gly : 1 Pro : 1 Hyp] supplementation significantly increased lifespan by 6–27% across five independent trials, even on heat-inactivated bacteria (Fig. 1f–k, Supplementary Fig. 1i–k and Supplementary Table 2). In contrast to previous studies where 5–1000 µM glycine but not 5–10 mM glycine extended lifespan or 1–10 mM proline extended lifespan^41,42, in our experimental setup supplementing individually 5 mM glycine or 5 mM proline, or 5 mM hydroxyproline were insufficient to increase lifespan (Supplementary Fig. 1l–n and Supplementary Table 2). To compare this increase in lifespan with collagen supplementation, we fed C. elegans with 400 µg/ml of rat tail collagen supplemented in their food and assayed both conditions, whole collagen versus [3 Gly : 1 Pro : 1 Hyp] supplementation, in parallel. We found that collagen supplementation increased mildly the mean lifespan but robustly the maximum lifespan, whereas [3 Gly : 1 Pro : 1 Hyp] supplementation increased both the mean and maximum lifespan (Fig. 1i–k and Supplementary Table 2). Thus, the [3 Gly : 1 Pro : 1 Hyp] supplementation outperformed collagen supplementation in promoting longevity.

The lifespans shown above were performed on solid media, where C. elegans are typically less motivated to move and rest in the food during aging⁴³. To measure their active movement capability during aging (i.e., healthspan), we assayed their thrashing rates in liquids (i.e., swimming) during their entire lifespan. We found that most ratios exhibited prolonged movement capability during aging in at least one of the two trials (Supplementary Fig. 2a, b and Supplementary Table 3). The [3 Gly : 1 Pro : 1 Hyp] consistently improved activity during aging. In one trial, the only two combinations showing higher swimming rates during mid-life than [3 Gly : 1 Pro : 1 Hyp] were [3 Gly : 1 Pro : 0 Hyp] and [1 Gly : 1 Pro : 3 Hyp] (Supplementary Fig. 2b and Supplementary Table 3). Interestingly, at the very old age of >24 days of adulthood at the end of the lifespan, [3 Gly : 1 Pro : 1 Hyp] fed C. elegans were the most active (Supplementary Fig. 2 and Supplementary Table 3).

Taken together, this suggests that increasing the levels of each amino acid is sufficient to increase lifespan further, but combining those three amino acids in a 3 to 1 to 1 ratio has synergistic effects that are greater than the sum of their parts.

Hydroxyproline is essential for collagen triple helix stability⁴⁰. During collagen synthesis in the endoplasmic reticulum, the prolyl-hydroxylase adds hydroxy to the proline in the collagen chain⁴⁰. A key and rate-limiting co-factor for prolyl-hydroxylase activity is the Krebs cycle intermediate α-ketoglutarate, which promotes collagen stability^44,45. Supplementing human dermal fibroblasts with α-ketoglutarate increases collagen production⁴⁵. Supplementing C. elegans or mice with α-ketoglutarate increases their lifespan^46,47. We confirmed that supplementing C. elegans with α-ketoglutarate from young adulthood increased their lifespan (Supplementary Fig. 3a–c and Supplementary Table 2). Intriguingly, combining α-ketoglutarate with [3 Gly : 1 Pro : 1 Hyp] showed additive effects on their longevity (Supplementary Fig. 3a and Supplementary Table 2). We wondered whether the longevity upon α-ketoglutarate supplementation would be mediated through its co-factor activity for prolyl-hydroxylase. We used a mutant dpy-18(ok162) of the prolyl-4-hydroxylase alpha subunit that eliminates its propyl-4-hydroxylase activity⁴⁴. The longevity upon α-ketoglutarate supplementation is completely abolished in dpy-18(ok162) mutants, suggesting that prolyl-hydroxylase activity is required for α-ketoglutarate-induced longevity (Supplementary Fig. 3b and Supplementary Table 2). Intriguingly, [3 Gly : 1 Pro : 1 Hyp] supplementation was still sufficient to increase lifespan (Supplementary Fig. 3b and Supplementary Table 2). These results suggest that α-ketoglutarate and [3 Gly : 1 Pro : 1 Hyp] work through parallel pathways for their additive and synergistic effects on promoting healthy aging.

Given the positive results of [3 Gly : 1 Pro : 1 Hyp] supplementation for extending C. elegans health and lifespan, promoting collagen homeostasis also in human skin cells, and [3 Gly : 1 Pro : 1 Hyp] supplementation being safe and well tolerated in old mice, we next worked on a formulation to maximize potential health benefits in humans. To support collagen synthesis, vitamin C is essential⁴⁰, and with acerola fruit juice extract, 48 mg of vitamin C is provided per serving. In a randomized clinical trial, the antioxidant astaxanthin, in combination with collagen peptides, improved facial skin elasticity and hydration and reduced matrix metalloproteinase expression while increasing endogenous collagen expression⁵⁴, suggesting enhanced collagen homeostasis. Furthermore, supplementing astaxanthin in yeast⁵⁵, C. elegans^56,57, Drosophila⁵⁸, and male mice⁵⁹ increased their lifespan⁶⁰. We provide 4 mg of astaxanthin per serving in the form of algae powder. Lastly, given our additive lifespan effects with the combination of alpha-ketoglutarate and [3 Gly : 1 Pro : 1 Hyp] supplementation for extending C. elegans' lifespan, we provide 1 g of calcium alpha-ketoglutarate per serving. With these additional supporting ingredients, we generated a novel formulation, which we will refer to throughout the text as a "collagen activator", to test its safety and efficacy in humans.

Not surprisingly, for the short duration of the 3-month study, wrinkle severity showed a downward trend from baseline (34.07%, 95% CI: 26.80–41.34%) to the final visit (30.00%, 95% CI: 23.80–36.20%, Data Source File 1). However, this change did not reach statistical significance (p = 0.2640; Data Source File 1). By contrast, we found that supplementing Collagen Activator improved skin texture after 1 month and improved it for the rest of the study. The skin texture score increased significantly from baseline (59.93%, 95% CI: 52.98–66.88%) to the first visit (64.76%, 95% CI: 58.14–71.38%, p = 0.0435) and continued to increase until the final visit at month 3 (65.05%, 95% CI: 58.35–71.76%, p = 0.0116) (Fig. 6b and Data Source File 1).

Similarly, after 1 month of Collagen Activator intake, skin hydration improved. We measured cheek skin hydration in relative corneometer units (AU) and observed significant improvements throughout the study, increasing from a very dry baseline level (35.89 AU, 95% CI: 32.40–39.39 AU) to a sufficiently moistened state by the first visit (46.72 AU, 95% CI: 43.71–49.74 AU, p < 0.0001) and further improving by the final visit (48.32 AU, 95% CI: 45.63–51.01 AU, p < 0.0001) (Fig. 6c and Data Source File 1). Furthermore, forearm hydration showed significant improvement from baseline (41.96 AU, 95% CI: 39.18–44.74 AU) to the first visit (45.88 AU, 95% CI: 43.07–48.68 AU, p = 0.0046). However, this improvement was not maintained at the final visit (44.62 AU, 95% CI: 42.24–47.01 AU), with no significant difference observed between the baseline and final visit (p = 0.0792) (Data Source File 1). Remarkably, hydration of the upper arm improved significantly over time, increasing from baseline (36.02 AU, 95% CI: 33.46–38.58 AU) to the first visit (42.39 AU, 95% CI: 39.57–45.20 AU, p = 0.0001) and remaining elevated at the final visit (40.91 AU, 95% CI: 38.44–43.38 AU, p = 0.0059) (Data Source File 1). This suggests an improvement in skin hydration in multiple body parts.

Moreover, we also observed improvements in skin elasticity, as measured by the R2 viscoelasticity index on the cheek, which significantly improved throughout the study. The R2 index, expressed as a percentage, increased from the baseline (66.79%, 95% CI: 64.58–69.01%) to the final visit (70.16%, 95% CI: 67.75–72.56%, p = 0.0401), indicating enhanced skin elasticity (Fig. 6d and Data Source File 1).

Notably, we also observed a few sex-specific improvements upon Collagen Activator intake. For instance, significant changes were observed for brown spots within females between Baseline vs. Visit 1 (p = 0.0218) and Visit 1 vs. Final Visit (p = 0.0084; Data Source File 1). For red areas, males exhibited a significant reduction in red areas between Baseline vs. Final Visit (p = 0.0189; Data Source File 1).

Lastly, participants reported self-perceived improvements in their skin condition up to 6 months (Fig. 6e and Data Source File 1). Taken together, oral supplementation of Collagen Activator improved skin texture, hydration, and elasticity within 1–3 months.

As the natural progression of time, the chronological age increased by 0.5 years from the mean chronological age of 47.46 ± 8.33 years at baseline to 47.88 ± 8.33 years at 6 months. By contrast, the mean biological age at baseline was 47.07 ± 7.21 years, while the mean biological age at 6 months was 45.70 ± 7.57 years (mean difference: −1.37), demonstrating a statistically significant reduction (p = 0.040, paired t-test, two-tailed). When stratified by gender, the average decrease in biological age was similar between females (−1.33 years) and males (−1.48 years), with no statistically significant differences in the extent of biological age reduction between genders observed (Fig. 7b, c). The maximum decrease in biological age observed after 6 months of collagen supplementation was 8.9 years for males and 12.1 years for females (Fig. 7c).

To determine individual changes over 6 months of Collagen Activator supplementation, we plotted each participant's chronological age minus their biological age at baseline and 6 months (Fig. 7d). A negative value indicates that the individual's biological age exceeds their chronological age (i.e., 'older' biologically), whereas a positive value indicates the biological age is younger than their chronological age. For example, one participant had a baseline difference of −13.36 years, meaning their biological age was approximately 13 years older than their chronological age. After 6 months, this difference improved to −1.26, suggesting they 'gained back' about 12 years in 6 months. In contrast, another participant showed a +8.82 at baseline (i.e., their biological age was 8.82 years younger than their chronological age), which changed only slightly to +7.92 by the end of the study, indicating no benefit from supplementation.

This might suggest that individuals with a low biological age at baseline might not benefit as much as individuals with a high biological age at baseline. To test this idea, we compared the 25 and 75 percentiles, i.e., low vs. high biological age at baseline, and their change in biological age after 6 months of Collagen Activator supplementation. We found that individuals starting with a higher biological age than their chronological age showed the most substantial improvement in reducing their biological age after 6 months (Supplementary Fig.). However, because the analysis conditions on an extreme baseline and uses a single retest, part of this pattern is expected from regression‑to‑the‑mean (RTM) rather than differential treatment response. Accordingly, we interpret this subgroup contrast as descriptive and hypothesis‑generating. To assess whether participants' baseline biological age (BA) or their change in BA (∆BA) was related to improvements in skin parameters, we performed correlation analyses between (1) baseline BA and final-visit percentage changes in skin measures, and (2) ∆BA and final-visit percentage changes in skin measures. 5a

Notably, baseline BA showed a positive correlation with the percentage change in cheek dermis thickness (r = 0.335, p = 0.024) but a negative correlation with the percentage change in cheek hydration (r = –0.298, p = 0.047). By contrast, ∆BA (i.e., the difference between 6-month and baseline BA) was significantly negatively correlated with forearm dermis thickness changes (r = −0.384, p = 0.009), indicating that individuals who experienced a greater reduction in BA tended to show larger increases in forearm dermis thickness. Other skin parameters demonstrated weaker or non-significant correlations. Together, these results suggest that although most skin changes were not strongly related to BA shifts, certain dermal thickness and hydration measures did show significant associations with BA or ∆BA.

Next, we wondered what else could influence the improvement from Collagen Activator supplementation. The frequency of Collagen Activator use (daily vs. every other day), exercise frequency, physical activity levels, sleep quality, and education level all showed no significant effects (Data Source File 1). Although 66% of participants use other supplements, we did not find any significant difference in the biological age of participants who had taken other supplements vs. non-supplement users (Supplementary Fig.), suggesting that the improvements observed with the Collagen Activator are independent of or synergistic with other supplements. By contrast, individuals who consumed more than one drink per week had fewer improvements with the Collagen Activator than individuals who did not drink any alcohol (Supplementary Fig.), suggesting a less favorable interaction between the Collagen Activator and alcohol consumption. Thus, after 6 months of using the Collagen Activator in this clinical study, we observed, on average, a 17-month reduction in biological age. 5b 5c

Caenorhabditis elegans strains were maintained on NGM plates and OP50 Escherichia coli bacteria. The wild-type strain was N2 Bristol. Mutant strains used are described at www.wormbase.org↗: TJ1060 spe-9(hc88) I; rrf-3(b26) II., JK2729 dpy-18(ok162) III. Transgenic strain LSD2002 spe-9(hc88) I; xchIs001 [Pcol-144:: GFP; pha-1(+)] X³⁸.

For determining the most effective amino acid ratios (Figs. 1a–e and S1c–f) 0.1 M stocks of glycine (Sigma-G7126), proline (Sigma-P5607), and hydroxyproline (Sigma-56250) were prepared in filter-sterilized via Millipore 0.2 μm pore membrane (SARSTEDT-83.1826.001) mQ-H2O. Afterward, these stocks were added into NGM agar during preparation (with either carbenicillin or ampicillin and nystatin; see WormBook nematode maintenance protocols), and mixed thoroughly prior to solidifying. Thus, corresponding amino acid concentrations were achieved: either 50 μM, 500 μM, or 5000 μM. Finally, 5 mM corresponds to "one ratio point" later (3-1-1 = 15 mM glycine – 5 mM proline – 5 mM hydroxyproline). After determining 1 ratio point, all further experiments were done in the same way, except for stocks, where amino acids were solubilized in mQ-H2O to their maximal solubility points: 4 M for glycine, 3.17 M for proline, and 198 mM for hydroxyproline, correspondingly. This was done to decrease excessive dilution of NGM with water, especially in experiments with 10-1-1, 1-1-2, and 1-1-3 ratios, to avoid high salt imbalance in nematode growth media. Control and amino acid-containing plates were kept at room temperature for 2–3 days to dry after preparation. Fresh OP50 bacteria were grown a day before the experiment to achieve their stationary phase (8 h or overnight). Then, it was heat-inactivated (to exclude metabolizing amino acids by bacterial cells) at 60 °C for a minimum of 45 min and cooled down to RT. Then, it was concentrated 10-fold with 2809 rcf/15' centrifugation and seeded onto plates after resuspending the bacterial pellet. One ml, 300 μl, or 120 μl of bacterial concentrate were added to 15 cm, 6 cm, or 3.5 cm plates, respectively. After seeding, the plates were ready to use once the NGM absorbed all the excess liquid. The quantification was performed according to Statzer et al.³⁸.

Manual lifespan assays were performed at 20 °C on heat-inactivated OP50 bacteria, seeded onto NGM plates containing nystatin if otherwise stated. N2 and JK2729 worms were grown until the L4 stage on plates with heat-inactivated OP50, then transferred to 50 μM FUdR-containing plates to prevent progeny development. On the other hand, TJ1060 and LSD2002 worms were kept at 25 °C until the young adulthood stage, the restrictive temperature for spe-9(hc88) temperature-sensitive mutation, and then, after confirmation of the C. elegans' sterility, re-picked to plates with corresponding conditions, as stated in the experiments. C. elegans were observed/counted every second day until day 12 of adulthood (AD12), with daily quantification later on until the end of the experiment. All animals that dried on the plate walls deeply dug into agar, as well as C. elegans with protruding vulva and matricide phenotype, were documented as censored for further analysis. Manual lifespan data were calculated using the Kaplan–Meier estimator in GraphPad Prism™ 8.0.1 software. Log-rank (Mantel–Cox) statistical analysis was used to determine the significance. Mean lifespan changes were calculated separately with a t-test or One-way ANOVA and SEM (if not stated otherwise), utilization for statistics in the same software package.

First, the population of 2000–5000 C. elegans were synchronized at day 1 to day 4 of adulthood (AD1-AD4) and bleached¹⁰⁰. One μl of ampicillin (100 mg/mL) and 0.5 μl of tetracycline (100 mg/mL) per 1 ml of buffer with bleached animals were added. The animals were left overnight for hatching and L1-synchronization. Second, the synchronous L1 populations of C. elegans were filtered via an 11.0 μm pore size, hydrophilic nylon membrane (Millipore—NY1104700). The clean population of L1s was counted under a stereomicroscope in 3 ×10 μl drops to define C. elegans concentration. Meanwhile, 50 ml of stationary-phase fresh heat-inactivated OP50 bacteria were concentrated, and all LB-media leftovers were discarded. Then, the bacterial pellet was resuspended thoroughly in 10 ml of S-complete buffer in aseptic conditions <10 cm from a gas burner. Before the experimental procedure, all the pipettes and pipette boys were sprayed with 70% ethanol. A total of 3000–4000 animals from the filtered L1 populations were concentrated at 1800 rcf/1' in 15 ml Falcon®, and M9 buffer was discarded until almost a packed worm pellet was reached. Afterward, 10 ml of S-complete with food was added to C. elegans and well-mixed. Later, this tube was sprayed with 70% ethanol, transferred under a sterile laminar-flow hood, and poured into a 10 cm sterilized tray. Using a multichannel pipette, 100 μl of buffer with C. elegans was resuspended and added into the U-bottom 96-well plate per well, achieving ~ 30–40 animals per well. Plates with worms were left in small sealed plastic bags at 20 °C (or 25 °C for TJ1060) on a working orbital shaker (450–500 rpm) for internal aeration. Animals were checked every day for the presence of contamination and abnormalities in growth and development. After reaching the L4 stage, 100 μM FUdR of final concentration was added for N2 C. elegans. Corresponding amino acids were added at AD1 in the mentioned concentrations. Animals were fed with additional heat-inactivated bacteria (but 50% of the initial amount) every 5th day until adulthood day 15 (AD15), when they drastically dropped food consumption. Motility (fitness) measurement was performed using WMicrotracker ONE ® (from InVivo Biosystems™). We used 3 ×30 min "time-buckets" for measurements, and the mean value was calculated for each well. In turn, we utilized eight wells per condition for robust quantification. XY-graphs (health-span curves) were built in GraphPad Prism™ 8.0.1 software for absolute motility readings. The area under the curve (AUC) for descriptive statistics and the two-way ANOVA test were used to calculate the significance between curves, and mean values of 8 wells of the same condition were used as reference points.

Human dermal fibroblasts (HDF, neonatal, Sigma-Aldrich, SKU 106-05N) were grown in a fibroblast growth medium (Sigma-Aldrich, SKU 116-500). Before 3-1-1 supplementation, cells were harvested using accutase cell detachment solution (w: 0.5 mM EDTA, w: Phenol red; Pan-Biotech; P10-21500) and centrifuged (200 × g, 5 min, RT). Cells were resuspended in fibroblast growth medium without supplementation or with a mixture of 3 mM glycine, 1 mM proline, and 1 mM hydroxyproline (sterile filtered). Subsequently, HDFs were seeded in 10 cm plates (30% confluency). Cells were cultivated for 2 h, 8 h, and 24 h and lysed on the plate using 1 ml TRIzol-Reagent (ambion®, ref# 15596018). Cells harvested at 0 h were centrifuged again (200 × g, 5 min, RT) immediately after resuspension and lysed using 1 ml Trizol. RNA extraction was performed using a standard phenol-chloroform method. Briefly, 1 mL Trizol lysate was combined with 0.2 mL chloroform, mixed, and centrifuged (10,000 × g, 10 min, 4 C). The upper layer (~0.4 mL) was taken, combined with 0.5 mL of isopropyl alcohol, mixed, and centrifuged (10,000 × g, 10 min, 4 C). The supernatant was discarded, and the pellet was washed twice with 80% ethanol, dried, and resuspended in RNase-free water.

RNA quality was confirmed using a NanoDrop 1000 spectrophotometer and an Agilent 2100 Bioanalyzer. cDNA libraries were prepared using the Illumina TruSeq Stranded Total RNA Sample Preparation protocol. A total of 150 bp paired-end reads were generated on an Illumina NovaSeq with a read depth of 20 million paired reads per sample. Reads were aligned and annotated to the Ensembl Homo sapiens GRC38.p13 reference genome and associated GTF annotation file using the align and featureCounts functions from the Rsubread package (2.10.4) in R (4.2.0). Differential gene expression analysis was performed using the voom and eBayes functions from the EdgeR (3.38.1) and Limma (3.52.2) packages. Gene set enrichment analysis was performed using the gseGO function from the clusterProfiler package (4.4.4). RNA-seq data can be found in the NIH SRA database under the accession number PRJNA856884.

Male C57BL/6J mice were ordered from Janvier between the ages of 18 and 21 months, equivalent to approximately 55 to 60 human years of age. All ordered mice were included in the study except for the ones that reached termination criteria before the start of the study, as defined in the animal license. Mice were allowed to acclimate to the facility for approximately 1 month before baseline measures were taken and experimental diets were initiated. The Swiss Federal Food Safety and Veterinary Office approved the study, and the animal license ZH149/21 was accepted via animex-ch (https://www.animex-ch.blv.admin.ch/↗).

The animal rooms were kept at 20–22 °C and 50–70% humidity. The mice were held in individually ventilated cages with a 12:12 h light-dark cycle, with the dark period occurring from 8 a.m. to 8 p.m. All measurements were performed during the dark period. Up to four mice were housed in one cage with ad libitum access to food and autoclaved drinking water. All animals were provided enrichment, including houses, nestlets, and chew blocks. Mice cages (including water bottles and enrichment) were changed weekly in a laminar flow hood, and body weight and food intake were recorded for all mice. Mice underwent daily monitoring during the study according to the parameters set out in Appendix 4. Mice reaching humane endpoint termination criteria were euthanized according to approved ethical protocols. Before starting the study, all mice were fed a standard AIN-93G-based diet for up to 1 month, during which time the baseline measurements were performed. A mixture of 3:1:1 ratio glycine:proline:hydroxyproline was mixed into the diet at 334 mg/kg of diet to achieve a final dose to mice of approximately 100 mg/kg body weight. All study diets were obtained from a commercial vendor (Research Diets, New Brunswick, NJ). After baseline measurements, mice were given either a control or 3:1:1 diet for 6 months. Cages were randomly assigned to a diet group. At the end of the study period, mice were euthanized by carbon dioxide inhalation, and tissues were collected for weighing and analysis. Organ weights were recorded for liver, kidney, heart, epididymal adipose tissue, interscapular brown adipose tissue, and spleen (Data Source File 1). There were no significant differences in organ weights other than the epididymal adipose tissue (Data Source File 1). A gross necropsy was performed to note macroscopically visible pathologies, e.g., tumors, cirrhosis, etc. No significant group differences were identified in these assessments.

The mFI is a tool to quantify general health in mice over several organ systems, which has previously been shown to correlate with mortality risk. To perform the mFI, all mice were assessed non-invasively for the presence or absence of a total of 33 health parameters, and the severity for each was graded as 0, 0.5, or 1. The original mFI developed by Whitehead et al.^101,102 includes 31 parameters, two of which were added in this study: the ledge test¹⁰³ and hind limb clasping. For the ledge test, the mice were evaluated for their ability to climb on a grid, which is a direct measure of coordination that can be impaired in ataxias and neurodegenerative diseases. Hind limb clasping is a phenotype associated with neurological dysfunction, which was observed in this study's aged mice and, therefore, incorporated into the mFI. The clasping was assessed by suspending the mice by the tail and scoring the degree by which the hind limbs were clasped towards the body. The ledge test and hind limb clasping were graded as 0, 0.33, 0.66, and 1. To calculate the mFI, the scores for each parameter were added up, and the sum was divided by 33 to get a value between 0 and 1, which allows a straightforward interpretation of the mFI. For more details on the scoring system, see Appendices 1–3.

One sachet of Collagen Activator contains active ingredients: patented 8400 mg Colgevity™ (5000 mg L-Glycine (Xi'an SLT Biotech Co., Ltd., Chemical synthesis), 1700 mg L-Proline (Xi'an SLT Biotech Co., Ltd., Corn fermentation), 1700 mg L-Hydroxyproline (Xi'an SLT Biotech Co., Ltd., Glucose Fermentation)), 1000 mg calcium alpha-ketoglutarate (Botanic Healthcare Pvt Ltd., Chemical synthesis), 200 mg algae powder (providing 4 mg astaxanthin; Botanic Healthcare Pvt Ltd., Chemical synthesis), and 150 mg acerola fruit juice extract (providing 48 mg vitamin C; Vidya Herbs Pvt. Ltd., Acerola fruit juice extraction).

Supporting ingredients were selected for their functional properties. Silicon dioxide was included as an anti-caking agent to ensure the powder's free-flowing nature, improve manufacturing processes, maintain ingredient uniformity, and enhance the stability of the final product. Citric acid was added as an acidity regulator to stabilize pH, enhance the flavor, improve solubility, prevent microbial growth, and protect the product's active ingredients, ensuring a high-quality and enjoyable final product. Steviol glycosides derived from stevia serve as a natural, calorie-free sweetener to enhance the taste, mask unpleasant flavors, and improve palatability.

The ingredients were sourced globally for quality control, focusing on high bioavailability and activity. Every batch underwent rigorous double-testing for purity by an independent Swiss lab (Data Source File 1).

Swissethics (Kantonale Ethikkommission Zürich) granted ethical approval for the study (BASEC ID number 2023-00953). The clinical observational trial was registered at ISRCTN registry. Clinical trial registry name: Reverse the Clock: A Clinical Trial of Collagen Activator Effects on Biological Age and Skin Quality. Registration number: ISRCTN93189645. Date of registration: 03/07/2025. CONSORT checklist provided in Supplementary Information. Written informed consent was obtained from all participants after a full explanation of the study. An observational human trial was conducted, with 66 subjects recruited and informed to consume one sachet (powder) of Collagen Activator daily for 6 months. Inclusion criteria: (1) Healthy adults aged above 35 years old, regardless of gender, ethnicity, or socioeconomic background; (2) Voluntarily purchasing the collagen activator (from Avea) for 6 months with the intent to take it daily (3) able to commit to 3 visits at the dermatology clinic Hautwerk in Zürich. The exclusion criteria included: (1) chronic skin conditions; (2) chronic medical conditions, including diabetes, cardiovascular diseases, kidney disease, liver disease, gastrointestinal disease, cancer, and autoimmune diseases; (3) pregnant and breastfeeding; (4) taking chronic medications; (5) people who had any cosmetic procedures (intense pulse light, medical peelings, laser therapy, microneedling, Botox, fillers) before 3 months of the study; (6) people who had taken other collagen supplements 3 months before the first screening date; (7) insufficient knowledge of the German or English language. Participants were recruited through a dedicated landing page on Avea Life website, newsletters, word of mouth, social media channels, and recruitment materials distributed at relevant conferences. Each participant underwent three skin measurements at the baseline, after 4 weeks, and after 12 weeks, respectively. Before each skin measurement, subjects were instructed to refrain from washing their face for 6 h, use any skin care products for 12 h, and acclimatize for 10 min before the visit. Self-assessment questionnaires of the subjects were collected monthly for 6 months. The biological age was measured via an epigenetic methylation clock test. Participants were asked to take the test via saliva collection at the baseline, and after 6 months, they began taking the Collagen Activator.

Study data were collected and managed using REDCap electronic data capture tools hosted at ETH Zurich¹⁰⁴. REDCap (Research Electronic Data Capture) is a secure, web-based software platform designed to support data capture for research studies, providing (1) an intuitive interface for validated data capture; (2) audit trails for tracking data manipulation and export procedures; (3) automated export procedures for seamless data downloads to common statistical packages; and (4) procedures for data integration and interoperability with external sources.

A Visia Cam (Canfield Scientific Inc.) was used to capture in-depth insights into skin conditions. The system is based on Canfield's RBX® technology, which utilizes cross-polarized light and UV illumination to assess reference values for spots, wrinkles, texture, pores, UV spots, red areas, and porphyrins. After image acquisition, specialized software processes the data using algorithms that detect and quantify skin parameters by identifying image patterns. The results are then compared to a comprehensive database of individuals grouped by demographic factors such as gender, age, and skin type. This database enables comparative analysis, with the software calculating a percentile for each parameter, positioning the patient's skin condition relative to the corresponding demographic group.

A Corneometer CM 825 (Courage and Khazaka, Cologne, Germany) was used to assess skin hydration of the cheek, forearm, and upper arm. Utilizing a capacitive measurement principle, the Corneometer quantifies the moisture content of the stratum corneum by detecting changes in capacitance that occur due to water's dielectric properties. Three measurements were performed for each location, and the values were averaged.

A Cutometer MPA 580 (Courage and Khazaka) was used to assess skin elasticity of the cheek, forearm, and upper arm, using a 2 mm probe in Mode 1. The device operates based on the suction method, where negative pressure is applied to the skin, and the resulting deformation is measured to evaluate the biomechanical properties of the skin. For each measurement, a target pressure of 450 mBar was applied, with an on-time of 3 s and an off-time of 3 s, allowing the skin to recover between cycles. Five repetitions were performed at each measurement site to ensure reliability and consistency in the results. The R2 value, representing gross elasticity, assessed the skin's overall elasticity. This parameter reflects the skin's ability to return to its original shape after deformation, combining immediate and delayed elastic recovery.

Measurements with both the Corneometer and the Cutometer were conducted on three standardized areas: on the left cheek, at a point 5 cm from the mouth corner along the line connecting the mouth corner to the tragus, at the midpoint of the right volar forearm, and the midpoint of the medial side of the right upper arm.

The measurements were taken under controlled environmental conditions of 22 °C and 45% humidity, with no daylight or windows in the room to ensure consistent conditions. Additionally, participants were instructed to refrain from contact with water and the use of cosmetics before the measurements (6 h and 12 h, respectively) to eliminate any external factors that could influence the results.

Saliva samples were taken twice: at the baseline, they were collected at Hautwerk dermatology clinic under the doctor's supervision. After 6 months, participants collected them at home using commercially available TruMe sampling kits. After collection, the samples were shipped to TruMe Labs for analysis. DNA methylation and DNAm age have been calculated, as previously reported by Demidenko et al.⁶¹. In particular, the TruMe TruAge epigenetic clock is a DNA methylation-based assay designed to estimate biological age by analyzing methylation patterns at nine specific CpG sites. Saliva-derived DNA samples were analyzed using locus-specific Sanger sequencing. The TruAge™ test, an epigenetic clock developed by TruMe Inc., estimates an individual's biological age by quantifying DNA methylation (DNAm) levels at selected CpG loci within the genome. Although the specific CpG identifiers (csIDs) and the complete algorithmic formula remain proprietary, the test operates on a well-established principle: a weighted mathematical combination of methylation measurements at preselected genomic sites known to correlate with aging.

The initial development of the TruAge clock involved screening over 900 CpG sites identified from publicly available DNA methylation (DNAm) datasets, chosen for their age-related variability and biological relevance. Candidate markers were filtered based on (1) above-median variation in beta values and (2) strong correlation between methylation status and chronological age (R² ranging from 0.38 to 0.95). From this pool, nine CpG sites were selected for locus-specific validation based on their high predictive value and genomic localization within regulatory regions of aging-associated genes. These sites exhibited methylation changes up to 40% across the human lifespan. A broader TruAge™ Index algorithm was also constructed using a panel of approximately 33 CpG sites, with the validated nine forming the foundation of the core predictive model. The TruAge™ clock estimates biological age via a regression-based formula:1\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\mathrm{Biological}\,\mathrm{Age}=C+\mathop{\sum }\limits_{i=1}^{n}{w}_{i\,}\times {M}_{i}$$\end{document}BiologicalAge=C+∑i=1nwi×MiWhere:C is a model intercept (constant),w₁ … w_n are site-specific weights (regression coefficients),M₁ … M_n are the beta values of DNA methylation at each CpG site, andn is the number of CpG sites used in the model (nine in the validated TruAge™ test).

Weights were derived using statistical learning methods, including multivariable regression and machine learning, trained to optimize age prediction accuracy.

TruAge™ utilizes saliva samples as the input biospecimen. Saliva was selected due to its non-invasive collection, 70% DNAm concordance with blood, and logistical convenience for at-home self-sampling protocols. Methylation quantification is performed using Sanger sequencing or pyrosequencing, both of which are standard high-resolution methods for detecting DNA methylation. The TruAge™ model was initially validated in a cohort of 105 healthy adults, achieving a Mean Absolute Deviation (MAD) of 4.67 years in chronological age prediction. Subsequent refinement of the model, incorporating data from over 1000 clinical trial participants (TruMe customer database; results pending publication), improved predictive performance to a MAD of 3.8 years. To assess analytical validity, TruAge™ predictions were compared against the MyDNAge® clock (ZymoGenomics), demonstrating strong agreement with a correlation coefficient (R² = 0.95, unpublished). Beyond correlation with chronological age, the TruAge test has demonstrated sensitivity to biologically meaningful changes in aging status. While established epigenetic clocks—such as Horvath's pan-tissue clock, Hannum's DNAm age, Levine's PhenoAge, GrimAge, and DunedinPACE—remain key benchmarks in the field, the TruAge Index was selected for this study due to its validated performance in saliva samples, ease of use, and robust responsiveness to biological age-modifying interventions. Its strong predictive capacity and concordance with established biomarkers of aging make it a suitable tool for our cohort and study objectives.

Online questionnaires were sent monthly to assess product intake frequency and overall health and skin conditions.

Shapiro–Wilk test was used to assess data normality. Repeated measures ANOVA or the Friedman test was applied to evaluate changes over time. For comparisons, paired t‑tests (or Wilcoxon signed‑rank tests) were performed for matched data, and unpaired t‑tests (or Mann–Whitney U tests) were performed for independent samples. Linear regression and Pearson correlation examined associations between baseline chronological and biological age. A chi-square test was employed to evaluate associations between categorical variables. All analyses were conducted in GraphPad Prism and R, and statistical significance was set at p < 0.05.