Close

IMMClock reveals immune aging and T cell function at single-cell resolution

Yael Gurevich Schmidt, Di Wu, Sanna Madan, Sanju Sinha, Sahil Sahni, Vishaka Gopalan, Binbin Wang, Saugato Rahman Dhruba, Alejandro A. Schäffer, Nan-ping Weng, Nicholas P. Restifo, Kun Wang, Eytan Ruppin

Posted on: 19 January 2025 , updated on: 20 January 2025

Preprint posted on 15 November 2024

Time flies when you’re… a cell!

Selected by Jessica Chevallier

Background

Age is not just a number anymore! Aging clocks, tools specifically designed to infer your biological age, how old your cells actually are, are gaining in popularity. So much so that the Biomarkers of Aging Consortium created a contest to benchmark and identify the top clock, or should I say clockmaker, in exchange for fame and a generous cash prize (1).

Methylation clocks, a popular type of aging clock, analyze methylation patterns at CpG sites in the DNA sequence. While accurate, obtaining whole-epigenome data at the single-cell level remains challenging. With this in mind, Schmidt and colleagues developed IMMClock (IMMune Cell clock), a machine learning model that infers the biological age of immune cells from single-cell (sc) and bulk transcriptomic data. The authors then used IMMClock to gain insight into the age-old question: “how does aging influence T cell functionality?”.

Key findings

Study design: embracing large and diverse cohorts

Schmidt and team developed individual IMMClocks for CD4+ T cells, CD8+ T cells, natural killer (NK) cells, B cells and monocytes. The development of each IMMClock can be split into a training, validation and testing phase. Unlike previous studies that trained aging clocks on small datasets, Schmidt et al. trained all IMMClocks on OneK1K- a massive single-cell dataset of ~1 million peripheral blood mononuclear cells (PBMCs) from a total of 982 healthy donors, ages 19-97 years-old. During the training phase, diversity is key. Training a model on representative unbiased data enables it to make accurate predictions later on.

Next came the validation phase, during which a different set of data was used to assess model accuracy and compare model performances during training. This step was used to finetune model parameters. The authors used scRNA-seq and bulk RNA-sequencing data from a total of seven large, diverse cohorts of healthy donors’ blood immune cells to validate the different IMMClocks. Finally, the testing phase approximated the model’s true performance on data it had not seen before.

The biological age of cells reflects the chronological age of individuals

Aging of the immune system can be attributed either to 1) differentiation, whereby a naïve cell differentiates into a specialized functional cell shifting the abundance of both cell states 2) intrinsic age-related changes that occur within the cell itself or 3) a combination of both. Consequently, Schmidt and colleagues generated these different models for each immune cell-type. The models were used to predict the biological age of cells and, subsequently, to assess whether these predictions correlated with chronological age. Predictions from the integrated IMMClock model, which considers both immune cell-type differentiation and intrinsic age-related changes, correlated highly with individual chronological age for CD4+ T cells, CD8+ T cells and NK cells. A lower correlation was observed for both B cells and monocytes, partially attributed to the diminished impact of age on these subsets.

The authors then focused their analysis on T cell aging. Four IMMClock models were trained to account for one of the following: (1) intrinsic age-related changes in naïve/central memory T cells, (2) intrinsic age-related changes in effector memory T cells, (3) shifts in cell-state abundance or (4) both intrinsic age-related changes and cell-state abundance, called the integrated model. All models were tested on six independent scRNA-seq datasets, a total of ~800 CD4+ and CD8+ T cell samples. The integrated IMMClock model was clearly the best predictor of biological age for both CD4+ and CD8+ T cells across cohorts, with the naïve/central memory T cell model not far behind. Models relying strictly on cell-state abundance generally performed poorly. Across cohorts, predictions from the integrated IMMClock model correlated highly with individual chronological age.

Aging hallmarks: CD8+ T cell vulnerability increases with age  

Aged T cells exhibit gene signatures specific to telomere shortening, senescence and exhaustion. The authors wanted to assess the correlation between the predicted biological age of cells and age-related gene signatures specifically in CD8+ T cells. As expected, a higher biological age was associated with a reduced expression of telomere maintenance genes and with a higher expression of both senescence- and exhaustion-related gene signatures. Importantly, the predicted biological age of naïve CD8+ T cells was lower than that of memory CD8+ T cells owing to the fact that biological age increases as T cells differentiate.

Within the IMMClock models, genes are categorized based on their weights. If a gene contributes to predicting the biological age of an immune cell it’s weight may increase. Conversely,  if a gene does not contribute to predicting biological age it’s weight may decrease.  Schmidt et al. selected genes considered predicters of biological age and performed a pathway enrichment analysis on these genes to identify pathways up and down regulated in different immune cell types. Focusing on CD8+ T cells, pathways associated to the positive regulation of cell processes, cell proliferation and translation were down regulated. Pathways related to the regulation of T cell cytotoxicity, differentiation and activation were also downregulated. Upregulated pathways included pathways related to adaptive immune responses and cytokine-mediated signaling, both contributors to “inflammaging”, a low-grade inflammatory state associated with age.

Biological age: an important determinant of health

Schmidt and colleagues investigated whether the biological age of cells can be linked to various clinical outcomes using data from the Framingham Heart Study. The IMMClock predicted an elevated biological age for CD4+ T cells, CD8+ T cells and NK cells in individuals with a history of cancer, heart disease and stroke. Next, the authors analyzed a scRNA-seq dataset of matched blood PBMCs and tumor-infiltrating lymphocytes to compare the biological age of circulating T cells to those of the tumor microenvironment in squamous cell carcinoma patients. Interestingly, the biological age of tumor microenvironment- associated CD8+ T cells was higher than that of circulating T cells.

The influence of T cell activation on biological age

The authors used a CRISPR activation (CRISPRa) perturb-seq dataset of a study by Schmidt, R. et al. to investigate the relationship between the biological age of T cells and their functionality. The naïve/central memory (CM) IMMClock was used to predict the biological age of T cells in this dataset. Plotting the biological age of unstimulated cells against re-stimulated cells revealed that most targets influence biological age in the same manner. However, targets associated with the negative regulation of cytokine production (e.g. MUC1, MAP4K1 and LAT2) behaved differently across conditions. These targets resulted in an elevated biological age in re-stimulated cells compared to unstimulated cells. The authors also demonstrated that biologically-young T cells generally have a higher activation score and vice versa. As suggested by the authors, it would be exciting to pinpoint which perturbations specifically generate young and highly activated T cells for therapeutic purposes.

Why is this work important

Unlike previous aging clocks, the IMMClock was trained on a large-scale scRNA-seq dataset contributing to its robustness. The authors manage to extensively demonstrate that their model accurately predicts the biological age of CD4+ T cells, CD8+ T cells and NK cells. The use of the IMMClock to analyze a previously published CRISPRa perturb-seq dataset to gain insight into T cell functionality with age, demonstrates the versatility of this tool in identifying “rejuvenated” T cells using solely their transcriptomic profile.

As we age, our innate and adaptative immune system fall short. This phenomenon, coined for the first time in 1969 by Dr. Roy Walford as “immunosenescence”, is characterized by a complete remodeling of the immune system resulting in an increased susceptibility to infections, cancers and autoimmune diseases in older individuals (2). Understanding aging at both the cellular and molecular level may pave the way for the development of novel therapeutic strategies that promote healthy aging- that is, growing older without disease onset.

Questions for the authors

1. It appears that most researchers develop aging clocks using only one data type, such as epigenomic or transcriptomic data. Do you envision developing a multi-omic aging clock that takes as input different “omics” – the epigenome, transcriptome and proteome? Would such a clock better predict the biological age of a cell or is a simpler model more appropriate?

2. Did the IMMClock discover any cell-type specific “aging signatures” that were largely conserved across individuals?

References

(1) “Biomarkers of Aging Challenge”. The Biomarkers of Aging Consortium, 2024, https://www.agingconsortium.org/challenge

(2) Liu, Zaoqu, et al. “Immunosenescence: molecular mechanisms and diseases.” Signal transduction and targeted therapy 8.1 (2023): 200.

Tags: machine learning, scrna-seq, t cell aging

doi: https://doi.org/10.1242/prelights.39515

Read preprint (No Ratings Yet)

Author's response

The author team shared

1. It appears that most researchers develop aging clocks using only one data type, such as epigenomic or transcriptomic data. Do you envision developing a multi-omic aging clock that takes as input different “omics” – the epigenome, transcriptome and proteome? Would such a clock better predict the biological age of a cell or is a simpler model more appropriate?

We recognize the potential of incorporating multi-omics data, such as epigenomics, transcriptomics, and proteomics, to create a comprehensive aging clock. While transcriptomics alone provides robust predictions with IMMClock, combining multiple omic layers could potentially enhance accuracy and uncover novel cross-layer interactions in aging. However, such approach presents challenges, including the variability of biological signals across different omics, and computational complexity of integrating heterogeneous data types. Despite these obstacles, we view multi-omic integration as a promising direction for future research, particularly as multi-omics datasets become more widely available.

2. Did the IMMClock discover any cell-type specific “aging signatures” that were largely conserved across individuals?

Yes, the IMMClock identified cell-type-specific “aging signatures” that are largely conserved across individuals. As illustrated in Figure 4a, the model demonstrates cell-type specificity by achieving the highest performance when trained and tested on the same cell type, compared to models trained on different cell types. If “signature” refers to the list of genes and their associated weights that contribute to the model, we have detailed these signatures for each cell-type-specific clock in Supplementary Tables S2–S8. These gene weights represent each gene’s contribution to predicting chronological age within the IMMClock model for each respective cell type. Furthermore, the conservation of these signatures across individuals is supported by the model’s strong generalizability, as evidenced by its robust performance in both cross-validation and independent external validation. This indicates that the identified aging signatures are consistently relevant across different individuals, underscoring the reliability of IMMClock in capturing fundamental aging processes specific to each cell type.

Have your say

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.

Sign up to customise the site to your preferences and to receive alerts

Register here

Also in the bioinformatics category:

Adenine DNA methylation associated to transcription is widespread across eukaryotes

Pedro Romero Charria, Cristina Navarrete, Vladimir Ovchinnikov, et al.

Selected by 13 January 2025

Francisco Falcon

Evolutionary Biology

Deep learning-based predictions of gene perturbation effects do not yet outperform simple linear methods

Constantin Ahlmann-Eltze, Wolfgang Huber, Simon Anders

Selected by 11 November 2024

Benjamin Dominik Maier

Bioinformatics

Functional Diversity of Memory CD8 T Cells is Spatiotemporally Imprinted

Miguel Reina-Campos, Alexander Monell, Amir Ferry, et al.

Selected by 22 August 2024

Marina Schernthanner

Bioinformatics

Also in the genomics category:

IMMClock reveals immune aging and T cell function at single-cell resolution

Yael Gurevich Schmidt, Di Wu, Sanna Madan, et al.

Selected by 19 January 2025

Jessica Chevallier

Bioinformatics

Adenine DNA methylation associated to transcription is widespread across eukaryotes

Pedro Romero Charria, Cristina Navarrete, Vladimir Ovchinnikov, et al.

Selected by 13 January 2025

Francisco Falcon

Evolutionary Biology

A fine kinetic balance of interactions directs transcription factor hubs to genes

Apratim Mukherjee, Samantha Fallacaro, Puttachai Ratchasanmuang, et al.

Selected by 23 July 2024

Deevitha Balasubramanian

Genomics

Also in the immunology category:

Discovery of a Potent and Selective Inhibitor of Human NLRP3 with a Novel Binding Modality and Mechanism of Action

Kevin Wilhelmsen, Aditi Deshpande, Sarah Tronnes, et al.

Selected by 13 February 2025

Roberto Amadio

Biochemistry

IMMClock reveals immune aging and T cell function at single-cell resolution

Yael Gurevich Schmidt, Di Wu, Sanna Madan, et al.

Selected by 19 January 2025

Jessica Chevallier

Bioinformatics

Target cell tension regulates macrophage trogocytosis

Caitlin E. Cornell, Aymeric Chorlay, Deepak Krishnamurthy, et al.

Selected by 04 January 2025

Felipe Del Valle Batalla

Biophysics

preLists in the bioinformatics category:

‘In preprints’ from Development 2022-2023

A list of the preprints featured in Development's 'In preprints' articles between 2022-2023

 



List by Alex Eve, Katherine Brown

9th International Symposium on the Biology of Vertebrate Sex Determination

This preList contains preprints discussed during the 9th International Symposium on the Biology of Vertebrate Sex Determination. This conference was held in Kona, Hawaii from April 17th to 21st 2023.

 



List by Martin Estermann

Alumni picks – preLights 5th Birthday

This preList contains preprints that were picked and highlighted by preLights Alumni - an initiative that was set up to mark preLights 5th birthday. More entries will follow throughout February and March 2023.

 



List by Sergio Menchero et al.

Fibroblasts

The advances in fibroblast biology preList explores the recent discoveries and preprints of the fibroblast world. Get ready to immerse yourself with this list created for fibroblasts aficionados and lovers, and beyond. Here, my goal is to include preprints of fibroblast biology, heterogeneity, fate, extracellular matrix, behavior, topography, single-cell atlases, spatial transcriptomics, and their matrix!

 



List by Osvaldo Contreras

Single Cell Biology 2020

A list of preprints mentioned at the Wellcome Genome Campus Single Cell Biology 2020 meeting.

 



List by Alex Eve

Antimicrobials: Discovery, clinical use, and development of resistance

Preprints that describe the discovery of new antimicrobials and any improvements made regarding their clinical use. Includes preprints that detail the factors affecting antimicrobial selection and the development of antimicrobial resistance.

 



List by Zhang-He Goh

Also in the genomics category:

Early 2025 preprints – the genetics & genomics edition

In this community-driven preList, a group of preLighters, with expertise in different areas of genetics and genomics have worked together to create this preprint reading list. Categories include: 1) bioinformatics 2) epigenetics 3) gene regulation 4) genomics 5) transcriptomics

 



List by Chee Kiang Ewe et al.

End-of-year preprints – the genetics & genomics edition

In this community-driven preList, a group of preLighters, with expertise in different areas of genetics and genomics have worked together to create this preprint reading list. Categories include: 1) genomics 2) bioinformatics 3) gene regulation 4) epigenetics

 



List by Chee Kiang Ewe et al.

BSCB-Biochemical Society 2024 Cell Migration meeting

This preList features preprints that were discussed and presented during the BSCB-Biochemical Society 2024 Cell Migration meeting in Birmingham, UK in April 2024. Kindly put together by Sara Morais da Silva, Reviews Editor at Journal of Cell Science.

 



List by Reinier Prosee

9th International Symposium on the Biology of Vertebrate Sex Determination

This preList contains preprints discussed during the 9th International Symposium on the Biology of Vertebrate Sex Determination. This conference was held in Kona, Hawaii from April 17th to 21st 2023.

 



List by Martin Estermann

Semmelweis Symposium 2022: 40th anniversary of international medical education at Semmelweis University

This preList contains preprints discussed during the 'Semmelweis Symposium 2022' (7-9 November), organised around the 40th anniversary of international medical education at Semmelweis University covering a wide range of topics.

 



List by Nándor Lipták

20th “Genetics Workshops in Hungary”, Szeged (25th, September)

In this annual conference, Hungarian geneticists, biochemists and biotechnologists presented their works. Link: http://group.szbk.u-szeged.hu/minikonf/archive/prg2021.pdf

 



List by Nándor Lipták

EMBL Conference: From functional genomics to systems biology

Preprints presented at the virtual EMBL conference "from functional genomics and systems biology", 16-19 November 2020

 



List by Jesus Victorino

TAGC 2020

Preprints recently presented at the virtual Allied Genetics Conference, April 22-26, 2020. #TAGC20

 



List by Maiko Kitaoka et al.

Zebrafish immunology

A compilation of cutting-edge research that uses the zebrafish as a model system to elucidate novel immunological mechanisms in health and disease.

 



List by Shikha Nayar

Also in the immunology category:

Community-driven preList – Immunology

In this community-driven preList, a group of preLighters, with expertise in different areas of immunology have worked together to create this preprint reading list.

 



List by Felipe Del Valle Batalla et al.

Journal of Cell Science meeting ‘Imaging Cell Dynamics’

This preList highlights the preprints discussed at the JCS meeting 'Imaging Cell Dynamics'. The meeting was held from 14 - 17 May 2023 in Lisbon, Portugal and was organised by Erika Holzbaur, Jennifer Lippincott-Schwartz, Rob Parton and Michael Way.

 



List by Helen Zenner

Fibroblasts

The advances in fibroblast biology preList explores the recent discoveries and preprints of the fibroblast world. Get ready to immerse yourself with this list created for fibroblasts aficionados and lovers, and beyond. Here, my goal is to include preprints of fibroblast biology, heterogeneity, fate, extracellular matrix, behavior, topography, single-cell atlases, spatial transcriptomics, and their matrix!

 



List by Osvaldo Contreras

Single Cell Biology 2020

A list of preprints mentioned at the Wellcome Genome Campus Single Cell Biology 2020 meeting.

 



List by Alex Eve

Autophagy

Preprints on autophagy and lysosomal degradation and its role in neurodegeneration and disease. Includes molecular mechanisms, upstream signalling and regulation as well as studies on pharmaceutical interventions to upregulate the process.

 



List by Sandra Malmgren Hill

Antimicrobials: Discovery, clinical use, and development of resistance

Preprints that describe the discovery of new antimicrobials and any improvements made regarding their clinical use. Includes preprints that detail the factors affecting antimicrobial selection and the development of antimicrobial resistance.

 



List by Zhang-He Goh