Differential DNA methylation clock ages across buffy coat (BC), peripheral blood mononuclear cells (PBMC), and saliva in individuals in early-to-mid adulthood

In a study of 91 individuals in early-to-mid adulthood, DNA methylation clocks revealed that while saliva yields significantly higher age estimates than blood-derived tissues (buffy coat and PBMC), the latter two are comparable to each other, and all tissue types remain suitable for within-tissue aging research despite notable differences in heritability patterns.

The Gist

Imagine your body is a massive, bustling city. Every cell in your body is a building, and inside each building, there is a library of instructions called DNA. Usually, these instructions stay the same throughout your life. However, there are little sticky notes attached to these instructions called methylation. As you age, more sticky notes get added or removed, changing how the library operates. Scientists have figured out how to count these sticky notes to create a "Biological Clock" that tells you how old your cells feel, which might be different from your actual birthday (chronological age).

This paper is like a quality control inspection of three different "reporting stations" in our city to see if they tell the same time. The researchers wanted to know: Does it matter if we check the clock using blood cells or saliva?

Here is the breakdown of their findings using simple analogies:

1. The Three Reporting Stations

The researchers collected samples from 91 people (mostly in their 30s) from three different sources:

• PBMC (Peripheral Blood Mononuclear Cells): Think of this as the VIP Lounge of the blood. It contains the elite immune cells.
• BC (Buffy Coat): This is the General Assembly of the blood. It's a mix of the VIPs plus some other cells like platelets and white blood cells.
• Saliva: This is the Street Corner. It's easy to get (just spit in a cup), but it's a mix of mouth cells and some immune cells that have wandered over.

2. The Big Discovery: Blood vs. Mouth

The team tested 15 different "clocks" (formulas to calculate biological age) on all three samples.

• The Blood Twins (PBMC and BC): These two stations were best friends. They told almost the exact same time. If the VIP Lounge said you were 35, the General Assembly said you were 35.01.

  • The Takeaway: If you have old blood samples stored in a freezer (Buffy Coat), you can trust them just as much as fresh, high-quality blood samples (PBMC). They are interchangeable.

• The Mouth Stranger (Saliva): The saliva station was running a different time zone. Almost every clock said that people were biologically older when measured by saliva than by blood.

  • The Analogy: Imagine you are 30 years old. The blood clocks say, "You feel like a healthy 30-year-old." The saliva clock says, "You feel like a tired 40-year-old."
  • Why? Saliva is a "mixed bag" of different cell types, and the way it was stored (at room temperature for years) might have made the "sticky notes" look a bit more worn out than they actually are.

3. The Exception: The "Super Clock"

There was one special clock called DunedinPACE. It's a next-generation clock designed to measure the speed of aging, not just the age.

• The Result: This was the only clock that said, "Hey, the time is the same in the blood and the mouth!" It was the only one that didn't get confused by the different sample types.

4. The Twin Test: Nature vs. Nurture

The researchers also looked at twins to see how much genetics influences these clocks.

• Identical Twins (MZ): They share 100% of their DNA.
• Fraternal Twins (DZ): They share about 50% of their DNA.

The Finding: Identical twins looked much more similar to each other when using blood samples than when using saliva.

• The Metaphor: If you look at the blood clocks, the identical twins were like two peas in a pod (very similar). But when you looked at the saliva clocks, the peas looked a bit more different, as if the "pea shell" (the mouth environment) was adding some noise to the signal.
• Conclusion: Blood samples give a clearer, more reliable picture of your genetic aging than saliva does.

5. The Verdict: Which Clock Should You Use?

The study concludes that while saliva is easier and cheaper to collect, it is not a perfect substitute for blood when trying to measure biological age accurately.

• If you have blood (Buffy Coat or PBMC): You can use them interchangeably. They are reliable.
• If you only have saliva: You can still use it, but you have to be careful. The "age" it reports might be artificially high. However, some of the newer, smarter clocks (like ZhangQ, PCGrimAge, and DunedinPACE) work reasonably well across all tissues, making them the best tools for the job.

Summary in One Sentence

Blood samples (whether fresh or frozen) tell a consistent story about your biological age, while saliva tends to exaggerate how old you are, though some modern "smart clocks" can still make sense of the saliva data.

Technical Summary

1. Problem Statement

Epigenetic aging, measured via DNA methylation (DNAm) clocks, is a critical biomarker for health and cognitive aging. However, a significant methodological gap exists regarding the comparability of DNAm clocks across different tissue types. Researchers often rely on biobanked samples (saliva, buffy coat, or peripheral blood mononuclear cells) due to ease of collection and storage, but it remains unclear whether these tissues yield equivalent aging estimates.

• Saliva is a composite tissue (epithelial cells + leukocytes) often collected non-invasively but may suffer from storage degradation.
• Buffy Coat (BC) and Peripheral Blood Mononuclear Cells (PBMC) are blood-derived tissues; while BC is easier to store long-term, PBMC is often the "gold standard" for disease and aging research.
• The Gap: Previous studies comparing these tissues have been limited by small sample sizes, wide age ranges, or lack of simultaneous collection. It is unknown if saliva-based clocks are direct substitutes for blood-based clocks, particularly in established adulthood (28–41 years), and how genetic factors (twin similarity) influence these measurements across tissues.

2. Methodology

Study Population and Design:

• Sample: 91 individuals from the Colorado Adoption/Twin Study of Lifespan (CATSLife), collected between 2015–2021.
• Demographics: Mean age 30.90 years (range 28.07–41.13); 50.5% female.
• Genetic Structure: Included 64 twins (18 Monozygotic [MZ], 14 Dizygotic [DZ] pairs) and 20 sibling pairs (biological and adoptive), plus 7 unpaired individuals.
• Tissue Collection: Every participant provided simultaneous samples of Saliva, BC, and PBMC.

Laboratory and Data Processing:

• DNA Extraction:
  • Saliva: Collected in tubes with dried reagents, stored at room temperature (3–5 years), extracted via modified phenol-chloroform/ethanol precipitation.
  • BC: Stored at -80°C; extracted using Maxwell 16 kits.
  • PBMC: Stored at -80°C; extracted via modified phenol-chloroform/ethanol precipitation.
• Methylation Profiling: Measured using the Illumina EPIC v1 array (850k CpG sites).
• Quality Control (QC):
  • Used minfi for preprocessing (detection p-value < 0.05, missingness < 1%).
  • Removed probes mapping to SNPs.
  • Applied background correction (preprocessNoob) and normalization (dasen).
  • Result: 11 saliva samples (9 individuals) were removed due to high missingness; final N = 91 for BC/PBMC, N = 82 for Saliva.
  • Cell type proportions were estimated using EpiDish (blood) and BeadSorted.Saliva.EPIC (saliva) and adjusted for in models.

Statistical Analysis:

• Clocks: Calculated 15 DNAm clocks using the dnaMethyAge R package (including Hannum, Horvath, Horvath2, PhenoAge, PCGrimAge, DunedinPACE, ZhangQ, etc.).
• Correlations: Calculated rank-based Spearman correlations ($r_s$) between tissues and with chronological age, using 100 iterations of random family member selection to avoid familial confounding.
• Mixed-Effects Models: Used lme (R nlme package) to test differences in methylation age by tissue type, treating Family ID as a random effect and Chronological Age as a fixed effect.
• Twin Analyses: Calculated Spearman correlations within MZ and DZ pairs. Applied meta-analytic procedures (DerSimonian–Laird) to aggregate correlations across clocks and tissues.

3. Key Contributions

1. Direct Tissue Comparison in Adulthood: This is one of the first studies to simultaneously compare Saliva, BC, and PBMC in a genetically informative sample of adults (28–41 years), controlling for genetic relatedness.
2. Validation of Biobanked BC: The study provides strong evidence that Buffy Coat (BC) is a statistically comparable substitute for PBMC in DNAm aging studies, validating the use of existing BC biobanks.
3. Saliva vs. Blood Divergence: It quantifies the systematic bias where saliva-based clocks yield significantly "older" methylation ages than blood-based clocks, cautioning against treating them as direct substitutes without correction.
4. Clock-Specific Performance: It identifies specific clocks (ZhangQ, PCGrimAge, DunedinPACE) that demonstrate higher cross-tissue consistency, guiding future tool selection.
5. Genetic Architecture: It demonstrates that twin similarity (heritability proxy) is higher for blood-based clocks than saliva-based clocks, suggesting saliva measurements may be more susceptible to environmental noise or measurement error.

4. Key Results

A. Methylation Age Estimates:

• Correlation with Chronological Age: Moderate correlations across all tissues ($r \approx 0.37–0.41$).
• Tissue Comparability:
  • BC vs. PBMC: Highly correlated ($r_s = 0.71$) and statistically indistinguishable in mixed models ($p > 0.44$).
  • Saliva vs. Blood: Saliva showed significantly older methylation ages compared to BC/PBMC across 14 of 15 clocks.
    • Examples: Saliva was ~12.5 years older (Hannum), ~7.4 years older (Horvath), and ~19.7 years older (PhenoAge) than BC.
  • Exception: DunedinPACE (a rate-of-aging clock) showed no significant difference between tissues ($p=0.486$), indicating comparable aging rates despite absolute age differences.

B. Twin Similarity (Reliability):

• MZ vs. DZ: MZ twins were more correlated than DZ twins overall ($r_{meta} = 0.49$ vs $0.38$), but MZ correlations showed significant heterogeneity by tissue.
• Tissue Effect on Heritability:
  • Blood (BC/PBMC): Higher MZ correlations ($r_{BC} = 0.64$; $r_{PBMC} = 0.49$).
  • Saliva: Lower and more inconsistent MZ correlations ($r_{Saliva} = 0.33$).
• Clock Generation: Next-generation clocks (e.g., PCGrimAge, DunedinPACE) generally showed stronger MZ correlations and better cross-tissue consistency than first-generation clocks.

C. Specific Clock Performance:

• PCGrimAge: Showed the highest cross-tissue correlations ($r_s = 0.92$) and the strongest MZ twin similarity ($r=0.85$).
• ZhangQ: A multi-tissue clock that showed the most comparable age distributions to chronological age across all three tissues.
• DunedinPACE: The only clock showing no significant mean difference in aging rates between saliva and blood.

5. Significance and Implications

• Biobank Utility: Researchers can confidently use stored Buffy Coat samples as a proxy for PBMC in epigenetic aging studies, expanding the utility of existing biobanks where PBMC was not isolated.
• Saliva Limitations: While saliva is non-invasive, it is not a direct substitute for blood in absolute methylation age estimation. It systematically overestimates biological age. However, it remains useful for relative comparisons within saliva-only studies or when using specific robust clocks (like DunedinPACE).
• Tool Selection: The study highlights that multi-tissue (ZhangQ) and next-generation (PCGrimAge, DunedinPACE) clocks are more robust to tissue heterogeneity than older, single-tissue trained clocks.
• Future Directions: The findings suggest a need for cell-type-specific clocks and further investigation into why saliva yields "older" ages (potentially due to epithelial cell composition or storage degradation at room temperature vs. -80°C for blood).

Conclusion:
The study concludes that while BC and PBMC are interchangeable for DNAm aging research, saliva requires careful interpretation. For studies prioritizing cross-tissue comparability or genetic signal detection, blood-based tissues (BC or PBMC) and next-generation clocks (PCGrimAge, DunedinPACE) are superior choices.