A new preprint on bioRxiv shows how researchers have developed transcriptomic, cell type specific aging clocks from the regenerative zones of mouse brains [1].
Measuring age at the level of a single cell
Aging is multi-faceted and difficult to quantify. While chronological age is straightforward, it doesn’t necessarily capture the underlying biology, as individuals naturally age at different rates. Additionally, various lifestyle factors and interventions can affect the rate of biological aging; for example, smoking can speed it up and exercise can slow it down.
Machine learning is being utilized to estimate biological age in order to better understand aging, predict remaining healthspan and lifespan, and more efficiently test longevity interventions. These “clocks” can have a variety of different inputs. Most commonly, DNA methylation patterns are used, but newer clocks have incorporated transcriptomics, proteomics, and lifestyle factors, among others.
The source of the biological material is also a major consideration when developing these clocks. Most commonly, bulk tissue is used, averaging measurements from many cells from a specific species and tissue. Recent advancements in technology now allow for these measurements to be made at the single-cell level [2].
This high-resolution approach has opened the door to many questions that were difficult or impossible to answer previously, but it has not yet been used to investigate the aging of different cell types. Do different types age at different rates? Which types are the most useful for predicting chronological and biological age? How do various interventions impact the biological age of different cell types?
A research team based out of Stanford University has used this technology to investigate these cell type-specific contributions to aging in the subventricular zone (SVZ) in mice [1].
The SVZ is regenerative and contains many cell types
The SVZ is a brain region located on the outer wall of the lateral ventricles. It is home to proliferating neural stem cells and is one of the only known sites of neurogenesis in the adult mammalian brain. Because of this important role, its aging progression in mice has been well studied by previous research [3].
In this study, single cell RNAseq was performed on the SVZs of 28 different mice ranging from 3 to 29 months of age. 11 distinct cell types were identified based on each cell’s transcriptome. In particular, oligodendrocytes, microglia, endothelial cells, astrocytes/quiescent neural stem cells (which clustered together), activated neural stem cells/neural progenitor cells (which also clustered together), and neuroblasts were the most abundant. The researchers detected a decline in proliferating neural stem cells with age, validating previous findings regarding the aging SVZ.
Predicting chronological age with the transcriptome of SVZ cells
Next, models were built from this data to predict the chronological age of these mice using the six most abundant cell types. First, these models were built with three cohorts of mice and validated on a fourth. Each cell type had a correlation coefficient (r) between 0.71 and 0.92 with an error between 1.6 and 5.4 months. Bulk models were also created with the same dataset by pooling the single-cell data by cell type. For each cell type, the single-cell models outperformed the bulk models. Their models were also validated and performed well on external datasets from two other studies [4,5].
As a part of the training process for these clocks, the most predictive genes were identified and weighted based on their relative contributions. The researchers’ analyses were able to single out the most important genes and biochemical pathways for each cell type. Most genes and pathways were cell type specific, but others such as AC149090.1 and Ifi27 were key in all or most cell types. This finding shows some ways in which cell types age differently and highlights the value of single-cell analysis relative to bulk tissue methods.
Predicting biological age
The primary function of the SVZ is to maintain proliferating neural stem cells, a capacity that declines markedly with age. This functional capacity was quantified in the neural stem cells, neural progenitor cells, and neuroblasts by measuring the percentage of cells which were in a proliferative cell cycle phase (G2/M and S). Proliferating cells ranged from 5 to 30% of the cell population, depending on the age of the mouse and the cell type. As expected, proliferative fraction was negatively correlated with age (r = -0.8).
This proliferative fraction value was used to train new clocks using similar methods as the chronological age clocks. The biological aging clocks were not as correlated or as accurate at predicting biological age as the chronological aging clocks were at predicting chronological age. However, they still had admirable performance with correlation coefficients between 0.41 and 0.89 and errors between 2.3 to 4.6 months for each cell type. They also performed well on the two external datasets.
Cell type-specific effects of heterochronic parabiosis on chronological and biological age
The researchers used their chronological age and biological age clocks to investigate interventions known to improve age-related decline. First, heterochronic parabiosis (where blood is shared between young and old animals) was conducted on 18 mice.
Exposure to young blood especially rejuvenated the activated neural stem cells/neural progenitor cells. This population had an average chronological age prediction of 4.5 months younger and a biological age prediction of 2.5 months younger. Other cell types trended towards rejuvenation but to lesser extents. Meanwhile, young mice exposed to old blood were predicted to be older across cell types, especially by the chronological aging clocks.
Exercise also has cell type-specific effects on chronological and biological age
Exercise has many well-characterized, beneficial effects on aging. The researchers used young and old mice with or without running wheels for 5 weeks for these experiments (n=15 total). Oligodendrocytes had their chronological age reduced by 1.4 months in young mice and 2.0 months in old mice and their biological age reduced by 0.6 months in young mice and 0.8 months in old mice. Activated neural stem cells/neural progenitor cells also experienced some rejuvenation (chronological: 1.9 months in young, 0.3 months in old; biological: 1.4 months in young, 1.2 month in old).
For each cell type, heterochronic parabiosis had a larger benefit than exercise. Since activated neural stem cells/neural progenitor cells responded to both interventions, they were used to compare between the two. There was minimal overlap between the transcriptome of these two interventions. This suggests that, despite both interventions targeting aging in a specific cell type, their effects were brought on by largely different mechanisms.
We generated 21,458 single-cell transcriptomes from the neurogenic regions of 28 mice, tiling ages from young to old. With these data, we trained a suite of single cell-based regression models (aging clocks) to predict both chronological age (passage of time) and biological age (fitness, in this case the proliferative capacity of the neurogenic region). Both types of clocks perform well on independent cohorts of mice. Genes underlying the single cell-based aging clocks are mostly cell-type specific, but also include a few shared genes in the interferon and lipid metabolism pathways. We used these single cell-based aging clocks to measure transcriptomic rejuvenation, by generating single cell RNA-seq datasets of SVZ neurogenic regions for two interventions – heterochronic parabiosis (young blood) and exercise. Interestingly, the use of aging clocks reveals that both heterochronic parabiosis and exercise reverse transcriptomic aging in the niche, but in different ways across cell types and genes.
Conclusion
This study is the first to use single cell transcriptome analyses to develop aging clocks. It is full of useful information to longevity researchers, particularly the datasets on aging, heterochronic parabiosis, and exercise in the SVZ. These three datasets should be useful for future research into the genes and pathways involved in both aging and rejuvenation.
This single-cell, high resolution approach highlighted many key findings that are typically washed out in the noise of bulk tissue analysis. In particular, the researchers showed throughout the study that the effects of aging, heterochronic parabiosis, and exercise were cell type-specific: aging looked different depending on which cell type in the SVZ was being evaluated. Additionally, the interventions did not uniformly rejuvenate all cell types. The heterogeneity between cell types is an added layer of complexity with which longevity researchers will need to continue grappling.
The study’s measure of biological age, proliferative fraction, was less predictive than chronological age. Additionally, it showed smaller changes than chronological age in response to the two interventions. Proliferative fraction is certainly an age-related functional outcome, especially in the SVZ, but it only captures a small piece of biological aging. Ultimately, better outcome measurements are still needed to quantify biological age and to train aging clocks.
Finally, it must be noted that, while exciting, these results are from a pre-print article. This means that they have only been scrutinized by the authors of the paper and have not yet undergone peer review. Additional, clarifying experiments or changes to the interpretation of results may be required before this data is published in a scientific journal.
Literature
[1] Buckley, et al. Cell type-specific aging clocks to quantify aging and rejuvenation the regenerative regions of the brain. bioRxiv pre-print (2022). https://doi.org/10.1101/2022.01.10.475747
[2] Singh, S.P. et al. Machine learning based classification of cells into chronological stages using single-cell transcriptomics. Scientific Reports (2018). https://doi.org/10.1038/s41598-018-35218-5
[3] Navarro Negredo, P., Yeo, R.W. & Brunet, A. Aging and Rejuvenation of Neural Stem Cells and Their Niches. Cell Stem Cell (2020). https://doi.org/10.1016/j.stem.2020.07.002
[4] Dulken, B.W. et al. Single-cell analysis reveals T cell infiltration in old neurogenic niches. Nature (2019). https://doi.org/10.1038/s41586-019-1362-5
[5] Harris, L. et al. Coordinated changes in cellular behavior ensure the lifelong maintenance of the hippocampal stem cell population. Cell Stem Cell (2021). https://doi.org/10.1016/j.stem.2021.01.003
