An atlas of the aging lung mapped by single cell transcriptomics and deep tissue proteomics

Ilias Angelidis, Lukas M Simon, Isis E Fernandez, Maximilian Strunz, Christoph H Mayr, Flavia R Greiffo, George Tsitsiridis, Elisabeth Graf, Tim M Strom, Oliver Eickelberg, Matthias Mann, Fabian J Theis, Herbert B Schiller

Preprint posted on June 20, 2018

Ageing at single cell resolution: a combined single cell transcriptomic and bulk proteomic map of the mouse lung reveals epigenetic dysregulation and cell type-specific effects of ageing.

Selected by Rob Hynds



The incidences of chronic lung diseases including chronic obstructive pulmonary disease (COPD), pulmonary fibrosis and lung cancer increase with age and lung function declines even during healthy ageing. As a complex organ consisting of over 40 cell types, the process of lung ageing at cellular and molecular levels is poorly understood.

A previous study used single cell RNA sequencing (scRNA-seq) to resolve cell types in the embryonic and adult mouse lung. In their pre-print, Angelidis et al. now extend this work to investigate the changes that occur in the adult lung during healthy ageing. They compare the cellular composition of the lung between young (3 months) and old (24 months) mice using scRNA-seq of almost 15,000 cells and integrating bulk proteomics data.

Key findings

Unsupervised clustering of scRNA-seq data revealed 30 distinct cell types – comparable to the published Mouse Cell Atlas lung data – and the cell types found were consistent between young and old mice. Interestingly, the authors found deregulated transcriptional control in older mice that was independent of cell type, supporting the emerging hypothesis that cell-cell transcriptional variability and the consequent drift of cellular identity within a cell type, rather than a co-ordinated transcriptional programme, underlies ageing.


Figure 1A-D. Workflow of scRNA-seq experiment, unsupervised clustering and transcriptional noise analysis.


The authors performed multi-omics integration of their scRNA-seq data with mass spectrometry proteomics data from an independent cohort of mice by aggregating scRNA-seq data in silico. Pathway analysis performed on both datasets revealed similar hallmarks of ageing, including those that are known, such as decreased mitochondrial function, increased expression of pro-inflammatory markers and extracellular matrix (ECM) remodelling. Analysis of ECM components by mass spectrometry is challenging but the authors use their recently described ‘quantitative detergent solubility’ protocol’, in which proteins are isolated in increasingly harsh detergents, to enrich ECM proteins in the final fraction. The ECM data also provide a reminder of the importance of integrating RNA and protein datasets: expression of collagen IV genes are downregulated during ageing but collagen IV protein is more abundant in old mice.

In addition to these bulk analyses, the resolution of scRNA-seq allowed investigation of cell type-specific ageing effects. Using alveolar type II (ATII) cells – the progenitor population of the alveolar epithelium – as an example, the authors show upregulation of major histocompatibility complex (MHC) class I genes, consistent with an interferon gamma signature identified in bulk analyses. Increased presentation of self antigens in older age might represent a protective mechanism against cancer as these cells are a candidate cell of origin for lung adenocarcinoma. Additionally, ATII cells in old mice also showed higher expression of acyl-CoA desaturase I (Scd1), an enzyme that converts saturated fatty acids to monounsaturated fatty acids. This finding is worthy of further investigation since the enzyme is considered a central link between lipid metabolism and adaptive stress signalling and has been worked up a potential therapeutic target in diabetes and obesity.

Cholesterol biosynthesis is an important process in the alveolus as cells produce and secrete surfactant, a biological film which lowers surface tension at the site of gas exchange. The authors’ pathway analysis suggested that altered lipid metabolism was a feature of aged lungs and they were able to pin this effect down to alterations in ATII cells and lipofibroblasts. In support of the idea that altered lipid biosynthesis might promote inflammation in the ageing lung, the aged mice seem to phenocopy those in which Insig1 and Insig2 – negative regulators of lipid metabolism – were deleted using an ATII-specific Cre-driver, causing lipid accumulation, inflammation and tissue remodelling.

What I like about this work

This study is the first to address adult lung ageing at single cell resolution. Reassuringly, it identifies many familiar mechanisms of ageing, both in the lung and in other organ systems. The power of this approach is demonstrated by the fact the authors unravel new effects of ageing on lung biology and can interrogate these in a cell type-specific manner. While the authors follow up on several of these in their pre-print, there is much more to be gleaned from their dataset. Helpfully, the pre-print is accompanied by a searchable online webtool in which you can check the behaviour of specific cell types, genes and/or proteins in the experiments. The tool will also be useful for the identification of lung cell type-specific markers, for example in studies aiming to characterise the heterogeneity of specific cell types in depth.


There are significant differences between the cellular composition of the murine and human lungs (for example, no basal cells, a significant progenitor population in human airways, are detected in these experiments) so the opportunity to apply the strategies described here to human tissues during normal ageing and chronic lung disease as part of the Human Lung Cell Atlas project will be fascinating.

Another exciting aspect of this work is the prospect of technological improvement. Here, the authors are able to integrate data from scRNA-seq with a bulk proteomics experiment but recent technological advances, such as CITE-seq, are beginning to allow us to derive protein information from the same single cells that are sequenced. Technologies for RNA sequencing in intact tissues – overcoming the requirement for enzymatic digestion – are also rapidly becoming available.

Further reading

Meiners, S. et al. Hallmarks of the ageing lung. European Respiratory Journal, 2015. 45(3): p. 807-27.

Questions for Authors

Few alveolar type I (ATI) cells are detected in scRNA-seq experiments. What is the cause of the unexpected ratio of ATI to ATII cells? Does this also explain why the bipotent alveolar progenitors identified in the Mouse Cell Atlas are not found here?

The changes in the club cell to ciliated cell ratio are interesting. Do the authors interpret this as a decrease in stem cell numbers during ageing?

Scd1 expression is highly upregulated in the RNAseq data but the protein is not detected in the old mice by proteomics. Do the authors think this is a technical limitation of the mass spec or could this be biological?


Tags: ageing, proteomics, respiratory, transcriptomics

Posted on: 11th July 2018

Read preprint (No Ratings Yet)

  • Author's response

    Herbert B. Schiller shared

    Thank you for this very nice summary of our work – we liked your comments and try to answer your questions below.

    We are aware of the AT2/AT1 ratio issue in this dataset. In human lungs, we got precisely the 2:1 ratio that we would expect from historical studies of human lungs. In mouse, we found out that the 500g centrifugation of single cell suspensions after isolation that we initially used was the main problem. In new experiments (not in the pre-print), the AT2/AT1 ratio improved (more AT1) using only 300g centrifugations. However, the ratio is not 2:1 as in humans, which may be either true biology or mouse AT1 cells are harder to isolate. We did not find the bipotent alveolar progenitors as a distinct cluster, but this may be possible by sub-clustering the alveolar epithelial cells in our data with optimized PCA parameters.

    The decrease in number of club cells that we observed was accompanied by a slight increase in ciliated cell frequency. This observation could be consistent with (1) an altered regulation of club to ciliated cell differentiation (which has been shown to depend on Notch signalling) and/or (2) altered club cell self-renewal capacity. We do not currently have data to address this question.

    The missing values for protein are most likely a technical limitation – we do not yet reach deep enough to quantify the whole proteome with the limited measurement time/fractionation used for this experiment.

    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