Associations of 2923 Olink proteins with demographic, lifestyle, environmental and health characteristics in middle-aged Chinese adults

Andri Iona, Baihan Wang, Jonathan Clarke, Kwok Hung Chan, Maria Kakkoura, Charlotte Clarke, Neil Wright, Pang Yao, Mohsen Mazidi, Pek Kei Im, Maryam Rahmati, Christiana Kartsonaki, Sam Morris, Hannah Fry, Iona Y. Millwood, Robin G. Walters, Yiping Chen, Huaidong Du, Ling Yang, Daniel Avery, Dan Schmidt-Valle, Feifei Li, Canqing Yu, Dianjianyi Sun, Jun Lv, Michael Hill, Liming Li, Robert Clarke, Derrick Bennett, Christopher Chen, China Kadoorie Biobank Collaborative GroupEuropean 2025 · DOI

TL;DR

This study analyzed the plasma levels of nearly 3,000 proteins in middle-aged Chinese adults to understand how non-genetic factors like age, sex, lifestyle, and health influence them. They found that age, sex, and BMI were strongly associated with many proteins, and importantly, these associations largely replicated in a European population, highlighting common biological mechanisms.

What Problem Does This Paper Address?

While genetic factors influencing protein levels (protein quantitative trait loci or pQTLs) have been extensively studied, the impact of non-genetic, environmental, and lifestyle factors (the "exposome") on the human plasma proteome is less understood, especially in diverse populations. Previous proteomics studies often focused on a limited number of proteins or exposures. This paper aims to bridge this gap by comprehensively investigating the associations between a wide array of exposome factors and a large number of plasma proteins in a Chinese cohort, and then validating these findings in a European cohort. Understanding these non-genetic influences is crucial for deciphering disease mechanisms and developing preventative strategies.

Methods

The researchers utilized the Olink Explore platform to measure plasma levels of 2,923 proteins in 2,006 middle-aged Chinese adults from the China Kadoorie Biobank (CKB). They then employed linear regression analyses to assess the associations between these protein levels and 37 different "exposures," which included demographic characteristics (age, sex), lifestyle habits (diet, smoking), environmental factors (temperature), clinical measurements (BMI, blood pressure), and health indices (frailty index). These analyses were adjusted for potential confounders and multiple testing using Bonferroni correction. To further explore independent associations, mutually adjusted models were used. The findings were then replicated in a subset of the UK Biobank cohort.

Datasets

  • China Kadoorie Biobank (CKB): A prospective cohort study of over 512,000 adults recruited between 2004 and 2008. For this study, 2,006 participants (without prior cardiovascular disease) were analyzed for proteomic profiling.
  • UK Biobank: Used for replication of findings. Approximately 35,000 participants were included in the replication analyses.

Key Insights and Ideas

  • Ubiquitous Influence of Exposome: A significant portion of the measured proteins (65%) were associated with at least one exposure, underscoring the broad impact of non-genetic factors on the proteome.
  • Dominance of Age, Sex, and BMI: These three demographic factors were the strongest drivers of protein level variation, each associating with hundreds of proteins. This highlights their fundamental role in shaping our biological makeup.
  • Frailty as a Key Health Indicator: The frailty index, a composite measure of age-related deficits, showed a substantial number of associations with proteins, suggesting its utility as a marker for overall health status and its link to systemic biological changes.
  • Sex Differences in Proteomics: Significant differences in protein associations were observed between males and females, particularly related to lifestyle and reproductive factors, pointing to distinct biological pathways influenced by sex.
  • Cross-Population Consistency: A remarkable finding was the high degree of replication (over 90%) of associations observed in the Chinese cohort within the European UK Biobank cohort. This suggests that many of the identified exposome-proteome relationships are conserved across different populations, pointing to fundamental biological mechanisms.
  • Limited Impact of Diet and Physical Activity: Interestingly, diet and physical activity showed relatively few associations with protein levels, contrasting with the stronger influence of age, sex, and BMI. This might suggest that their effects are more indirect or require longer-term accumulation to manifest in protein levels.

Key Results

  • 31 out of 37 exposures were associated with at least one protein.
  • Age (1154 proteins), sex (827 proteins), and BMI (869 proteins) showed the highest number of associations.
  • Frailty index (597 proteins), systolic blood pressure (479 proteins), and random plasma glucose (387 proteins) were also significantly associated with numerous proteins.
  • 65% of the 2,923 proteins examined were associated with at least one exposure.
  • 25 proteins were associated with 10 or more exposures, with five proteins (CDH2, ADGRE2, ADGRD1, ACY1, MEGF9) remaining significant after mutual adjustments.
  • Over 90% of the significant associations found in the Chinese cohort were replicated in the UK Biobank cohort.

Limitations and Future Work

The authors acknowledge several limitations:

  • Cross-sectional Design: The study design prevents the confirmation of causal relationships between exposures and protein levels.
  • Power for Rare Exposures: While adequate for common exposures, the study had limited power for rare binary exposures (e.g., liver disease, cancer).
  • Potential Residual Confounding: Despite extensive adjustments, residual confounding may still exist.
  • Recall Bias: Self-reported data for some exposures (e.g., prior diseases, diet) may be subject to recall bias, although efforts were made to mitigate this through resurveys.
  • Lack of Independent East Asian Replication: The absence of similar large-scale proteomics datasets from independent East Asian cohorts limits direct replication within this specific demographic.

Future work suggested includes:

  • Longitudinal Studies: Incorporating longitudinal data and genetic approaches (like Mendelian Randomisation) to clarify causality.
  • Prediction Models: Utilizing penalization methods (e.g., LASSO) to build prediction models and handle the high dimensionality of exposome and proteomic data.
  • Mechanistic Exploration: Investigating biological mechanisms linking exposures to specific proteins, potentially through integration with genetic data.
  • Broader Exposome Data: Future blood-based studies should collect more comprehensive exposome information to enhance analysis robustness.

Summary updated

Abstract

Previous studies have identified a large number of genetic variants affecting plasma levels of proteins, but little is known about the non-genetic factors influencing plasma levels of proteins in diverse populations. We measured plasma levels of 2923 proteins, using Olink Explore platform, in 2006 participants (mean age = 50.8 years; 62% female; mean body mass index = 23.9 kg/m2) in the China Kadoorie Biobank, without prior cardiovascular diseases. Linear regression analyses were used to assess the associations of individual proteins with 37 major exposures across multiple domains (e.g., socio-demographic, lifestyle, environmental, sample processing, reproductive factors, clinical measurements, and health-related indices), adjusted for potential confounders and multiple testing. These were further replicated and compared with findings in UK Biobank. Overall 31 exposures were associated with at least one protein, with age (n = 1154), sex (n = 827), body mass index (n = 869) showing the highest number of associations, followed by frailty index (n = 597), systolic blood pressure (n = 479), random plasma glucose (n = 387), ambient temperature (n = 292), and hepatitis B surface antigen positivity (n = 282), but with diet and physical activity showing little associations. Likewise, of the 2,923 proteins examined, 65% were associated with at least one exposure, with 25 proteins associated with ≥ 10 exposures, including five (CDH2, ADGRE2, ADGRD1, ACY1, MEGF9) after mutual adjustments. The patterns of associations were similar after further mutual adjustments for exposures examined but differed by sex, chiefly due to differences in lifestyle and reproductive factors. Most of the observed associations were replicated in the Europeans. For the purpose of Open Access, the author has applied a CC-BY public copyright license to any Author Accepted Manuscript version arising from this submission.
BibTeX
@inproceedings{iona2025associations,
  author = {Andri Iona and Baihan Wang and Jonathan Clarke and Kwok Hung Chan and Maria Kakkoura and Charlotte Clarke and Neil Wright and Pang Yao and Mohsen Mazidi and Pek Kei Im and Maryam Rahmati and Christiana Kartsonaki and Sam Morris and Hannah Fry and Iona Y. Millwood and Robin G. Walters and Yiping Chen and Huaidong Du and Ling Yang and Daniel Avery and Dan Schmidt-Valle and Feifei Li and Canqing Yu and Dianjianyi Sun and Jun Lv and Michael Hill and Liming Li and Robert Clarke and Derrick Bennett and Christopher Chen and China Kadoorie Biobank Collaborative Group},
  title = {{Associations of 2923 Olink proteins with demographic, lifestyle, environmental and health characteristics in middle-aged Chinese adults}},
  booktitle = {European},
  year = {2025},
  doi = {10.1007/s10654-025-01311-z},
}
Figure